*•,%■« 1^ / PLANNING DOCUMENTATION VOLUME '^ Project Severe C^ivironmenfai Storms and Mesoscale Experiment January 1977 Editor, Douglas K. Lilly o Boulder, Colorado ^c y-<^^ ^ January 1977 v. \ PLANNING DOCUMENTATION VOLUME Sev Stc Ml E: January 1977 Editor, Douglas K. Lilly 6 o Q Q UNITED STATES DEPARTMENT OF COMMERCE Juanita M. Kreps.* Secretary ■■^ ^^W3S/»,, NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION Robeft M White, Administtatof Environmental Researcti Laboratories Wilmot N Hess. Director "*»ENT Of "^^ NOTICE The Environmental Research Laboratories do not approve, recommend, or endorse any proprietary product or proprietary material mentioned in this publication. No reference shall be made to the Environmental Research Laboratories or to this publication furnished by the Environmental Research Labora- tories in any advertising or sales promotion which would in- dicate or imply that the Environmental Research Laboratories approve, recommend, or endorse any proprietary product or proprietary material mentioned herein, or which has as its purpose an intent to cause directly or indirectly the adver- tised product to be used or purchased because of this Envi- ronmental Research Laboratories publication. ii FOREWORD This volume is intended as a supplement and accompaniment to the Project Development Plan for Project SESAME, dated September, 1976. It contains the significant group and individual planning reports received by the SESAME planning office since the original draft Project Development Plan (May- June, 1974) and the "Open SESAME" meeting proceedings, issued about 18 months ago. These include three reports from a planning conference on boundary layers and regional scale modelling, another on the storm scale program, two smaller planning group meetings on gravity wave interactions and satellite data and two significant reports by committees not directly associated with SESAME. Three reprints of relevant journal articles are also included. It is hoped that this material will provide both a survey of the current understanding of severe storm structure, dynamics, and observational and predictive capa- bilities, and a measured view of what can and should be done in the future. Other sources of somewhat concentrated information on the current status of severe storms research are the special cloud dynamics issue of Pure and Applied Geophysics (Vol. 113, issues 5/6 pp. 713-1084, 1975) and preprint volumes of the A.M.S. Severe Local Storms Conference of October 1975 and the A.M.S. Weather Forecasting and Analysis Conference of May, 1975, both of which are available from the American Meteorological Society. As indicated in the foreword of the current Project Development Plan, funding for the major elements of SESAME has been delayed, apparently due to competition with the First CARP Global Experiment and other major national and international commitments. Several of the recommended preliminary ob- servational, theoretical, and modelling studies are in progress, however, and the results of these will enhance the conduct of the major program ele- ments when augmented funding becomes available (as I expect it ultimately will). As the editor of this volume and one of the principal managers of the SESAME planning effort, I would like to express my gratitude to the roughly 200 scientists who participated in this effort. In most cases this was done without direct sup'port from the SESAME office except for travel costs. I hope it turns out to have been worth the trouble. Douglas K. Lilly Consultant to NOAA for Project SESAME ill CONTENTS FOREWORD Page 1. Severe Storms and Storms Systems: Scientific Background, 1 Methods, and Critical Question. Douglas K. Lilly (Reprint) 2. Aspects of Cumulonimbus Study, F. H. Ludlam (Reprint) 23 3. Reports of the Boundary Layer and Regional Scale Modeling Workshop 3.1 Theoretical Boundary Layer Group 29 3.2 Observation and Instrumentation Group 35 3.3 Regional Scale Modeling Group 49 4. Report of the Gravity Wave Workshop 61 5. Reports of the Thunderstorm Scale Workshop 5.1 Working Group I on the development and utilization 85 of thunderstorm scale observing techniques, with Appendices A-F 5.2 Working Group II on the theoretical understanding 146 and numerical modeling of thunderstorms 5.3 Working Group III on thunderstorm-environment interactions 176 5.4 Working Group IV on cloud microphysics and electrification 209 5.5 Working Group V on the problem of tornado development 214 6. Report of Satellite Data Working Group 226 7. Interactive applications of satellite observations and 234 mesoscale numerical models, Carl W. Kreitzberg (Reprint) 8. Report of the Panel on Short-Range Prediction to the 241 Committee on Atmospheric Sciences, National Research Council 9. Draft report of FGGE Workshop C, Midlatitude Regional 291 Problems Digitized by the Internet Archive in 2012 with funding from LYRASIS IVIembers and Sloan Foundation http://archive.org/details/projectsevereenvOOboul Pageoph, Vol. 113 (1975), Birkhauser Verlag, Basel 1 . Severe Storms and Storm Systems : Scientific Background, Methods, and Critical Questions By Douglas K. Lilly ^ Abstract The structure and dynamics of severe convective storms and their mesoscale environments is described on the basis of current literature. Numerical modeling of regional and cloud-scale meteor- ology is reviewed with respect to its contribution to the understanding of convective storm evolu- tion. Observation techniques are surveyed briefly. Critical questions, principally on the triggering of convective storms, are listed and a U.S. national program (Project SESAME) aimed at answering them is briefly described. 'Blow, winds, and crack your cheeks! rage! blow! You cataracts and hurricanoes, spout Till you have drench'd our steeples, drown'd the cocks ! ' - King Lear. 1. Introduction The evolution of a severe storm system depends on interactions between motion and thermodynamic fields on scales ranging from that of the midlatitude cyclone with dimensions of thousands of kilometers, to that of submicron-diameter cloud nuclei. In the following 1 have attempted to survey the current state of scientific knowledge, the observational and theoretical techniques now becoming available, and the most critical current questions and needs for developing a useful understanding of severe storms and storm systems, with the emphasis on the scales ranging from a few to a few hundred kilometers. The majority of this survey was originally assembled to aid in the planning of a new U.S. national program in severe storm and mesoscale meteor- ology, which is described briefly in Section 6. The description of the current state of knowledge is divided into Section 2 on the structure of severe storm cells themselves and Section 3 on the mesoscale interactions and evolution of arrays of storms. Section 4 describes the current status of prospects ») The National Center for Atmospheric Research, Boulder, Colorado, 80303. NCAR is sponsored by the National Science Foundation. 714 Douglas K. Lilly for improved numerical modeling of severe storms and their mesoscale environments. Section 5 very briefly surveys the observational methods suitable for investigation of these phenomena. In Section 6 we attempt to summarize the critical questions and the observational and theoretical approaches needed to provide answers. 2. Severe storm structure Current understanding of the nature of intense thermal convection is often con- sidered to date from the 'Thunderstorm Project' of 1946 (Byers and Braham [1, 2]). From the results of this multi-agency field program, the nature of the small-to- moderate amplitude 'air mass' thunderstorm was defined as a somewhat randomly organized composite of subelements or cells, which individually went through a life cycle of order 30 minutes, although the storm itself might have a substantially longer life. Such storms are nearly everyday summertime occurrences in southeastern U.S. and other areas in which warm humid low-level air and weak wind shears prevail. The severe and potentially damaging storms most frequently found in the U.S. midwest in spring typically develop in the presence of strong ambient shear. These storms evidently can organize themselves in such a way as to maintain a quasi-steady- state structure as they advance across the countryside, affecting large areas and contin- ually processing new air. The structural details vary substantially with the ambient wind and thermodynamic conditions, and also are apparently not unique within a given environment. Examples of several types recently documented by Marwitz [3] include the multi-cell storm, the so-called supercell storm, and a variety of the supercell which develops in extremely strong shears. The multi-cell thunderstorm in a sheared environment is related to the ordinary multi-cellular air-mass storm, but with an organized structural sequence such that each new cell develops on one side of the storm system, moves through it, and passes out the other side, as illustrated in Fig. 1 (Browning and Ludlam [4,5]; Newton and Fankhauser [6]). The supercell storm, named and described in detail by Browning [7, 8], consists of a single giant cell propagating to the right of and slower than the mean wind. This type of storm is often (perhaps always) rotating significantly, and is apparently the cause of most severe tornadoes (Barnes [9]). Another type of storm was described by Marwitz from two examples which developed and persisted in the presence of extremely large shears. This type consists of a single cell oscillating in a quasi-periodic manner, apparently being partially suppressed and forced to re- generate when the precipitation shaft is carried out in front of the updraft. One of the most noteworthy features of all severe storms, as observed by radar, is the 'echo-free vault' (Browning and Ludlam [4]) or 'weak echo region' (Chisholm [10]) which apparently coincides with the maximum updraft region of the cell, at least in the early and mature stages of the storm. The radar echo is missing or weak because the updraft is too fast for large drops to either form in or fall through the lower half Severe Storms and Storm Systems 7 1 5 DIRECTION OF TRAVEL OF ECHO- MASS Figure 1 Showing the influence of cell propagation on movement of storm. Each subsequent cell development occurs to the right of the previous cells, giving an effective right-hand propagation of the storm. (after Brown and Ludlam [4]). of the cloud. This updraft region is usually on the right or growing side of the cell and is the most probable location of tornadoes, if present. The presence of strong updraft rotation and probable tornado formation has been detected from radar by the obser- vation of a 'hook' echo extending part of the way around the weak echo region, sometimes enveloping it as the updraft ages (Burgess and Brown [11]; Barnes [12]; FujiTA [13]). Browning's [8] schematic illustration of the development of such an echo is shown on Fig. 2. The question of the frequency of occurrence of rotation in thunderstorms and its relationship to motion and propagation remains somewhat uncertain, but is very important to the tornado problem. Presumably all significant tornadoes generate within rotating cells, though not all rotating cells produce tornadic vortices. It has been frequently suggested that cyclonically-rotating severe storms move to the right of the mean wind because of the Magnus effect. An alternative hypothesis (Newton and Fankhauser [6]) is that such storms do not really move to the right but rather grow in that direction, from whence comes the moist inflow in the typical environmental conditions of southerly low-level winds veering to westerly in the upper troposphere. It is not quite clear, however, whether these two hypotheses are really independent of each other. A few storms have been observed to move to the left of, and faster than, the mean wind (Hammond [14]), and these apparently have a sort of mirror image structure to the right-moving storms, with the updraft and weak echo region occurring on the front left sector and perhaps anticyclonic storm rotation. Such storms have also been observed to contain tornadoes whose direction of rotation was unknown. 716 Douglas K. Lilly V ^ 1520 ECHO IN LOWER TROPOSPHERE ^ ECHO IN MIDDLE X V^TROPOSPHERE >' \ ECHO IN UPPER i \ TROPOSPHERE / / / 10 MILES V-H WINDS RELATIVE TO STORM Figure 2 Schematic diagram illustrating the development of the hook echo with successive positions of the hook being shown at times 1510, 1520, 1530 and 1540 CST for a particular storm development. The arrows on the lower right refer to wind direction in the low, middle, and high troposphere (from Browning [8]). Several examples have been observed (Newton and Fankhauser [6]; Fujita and Grandoso [15]; Charba and Sasaki [16]; Burgess and Brown [11]) of a pair of thunderstorms (or radar echoes) initially close together, which then diverged (AcHTEMEiER [17]; Lemon [18]; Jessup [19]), one moving to the left of the mean wind wind and one to the right. Fujita and Grandoso [15] suggested that such a pair may develop downwind of a pre-existing single storm as counter-rotating storms rather like von Karman vortices (see Fig. 3). The ever-present effects of the earth's rotation would seem, however, to preclude strict left-right symmetry. Recent aircraft flights and radar chafiF-tracking studies have shed considerable light on the structure of the principal updraft in the weak echo region. At cloud base and several thousand feet above it, the updraft has been observed (Marwitz and Berry [20]) to be almost nonturbulent, i.e., considerably smoother than the subcloud layer away from the storm. This is presumably due in part to the turbulence-suppressing effects of acceleration as occurs in the inlet constriction of a wind tunnel, and also perhaps because the inflow air has travelled for some distance in the cloud shadow so that boundary-layer thermals are no longer active. However, Battan [21] observed turbulent flow throughout the updraft of Arizona hailstorms. It is not known whether this result represents a difi"erence in kind or is due to the different observational methods. The updraft at cloud base is found (Marwitz [3]; Grandia [22]; Davies- Jones [23]) to have essentially the same potential temperature as that near the surface. Because of the slight thermal stability of the subcloud environment (as first shown by Malkus [24]), the updraft air at cloud base is negatively buoyant by two or three 4 Severe Storms and Storm Systems 717 degrees. Thus the updraft near cloud base is actually an energy sink for the cloud, which must be made up by greater buoyancy in the middle and upper levels. The lower portions of the updraft are essentially undiluted by entrainment of drier ambient air (Barnes [9]; Davies- Jones [23]). As the updraft enters the cloud, the cloud entrain- ment from the edges eats into it (Grandia and Marwitz [24]), reducing the equivalent potential temperature and introducing high turbulence levels. In the middle and upper levels of the cloud, turbulence observed from penetrating aircraft is severe to extreme, and it becomes more difficult to identify a single primary updraft region. Nevertheless, in active cells visible updraft turrets penetrate above the primary outflow anvil (Sinclair [26]) extending well into the stratosphere on occasion and then rapidly descending back into the main cloud mass. Observations of the total latent plus sensible heat content of thunderstorm down- drafts indicate that most of the air within them must originate from middle tropos- pheric levels, becoming negatively buoyant from the effects of evaporating rain. The actual source and history of the downdraft has not been directly observed, however. It has been suggested by Fujita [13] that the rebounding turrets from former updraft Figure 3 Schematic diagram showing steps in the process of storm splitting. The vertical cross sections of the anticyclone storm (top) and those of the cyclonic storm (bottom) are drawn so that they can be compared at each step with the plan views shown in the middle (from Fujita and Grandoso [15]). 718 Douglas K. Lilly maxima produce the initial impulse of strong downdrafts. Barnes [12] suggests that an additional source of impulsive downdrafts may be evaporative cooling of the high horizontal momentum midlevel air immediately adjacent upwind of the main precipita- tion area. Strong outflow currents occur at both the top and bottom of severe storms. The upper outflow is of interest in part because of the possibility of estimating the total mass flow in the cloud from satellite observations of the cirrus clouds in it (Sikdar and SuoMi [27]). In addition, it is a striking example of turbulence collapse in a stable environment, a phenomenon also observed in submarine wakes in the oceanic thermo- cline. The lower outflow is also sometimes visible from satellite pictures as an expand- ing arc cloud line or series of wave clouds, which is discussed further in the next section. 3. Mesoscale interactions and severe storm arrays The interaction of convective clouds and their larger scale environment remains one of the most challenging problems of meteorology. From the viewpoint of the large- scale meteorologist or climate theorist, buoyant convection is a turbulent exchange process to be treated statistically by parameterization. In the tropics and in the summer season in midlatitudes, when horizontal temperature gradients are relatively small, vertical overturning by cloud convection is the principal method by which the atmos- phere rectifies the radiative imbalances between short-wave surface heating and long- wave atmospheric cooling. Because of the ability of water substance to exist in three phases in the atmosphere and to exchange considerable amounts of energy during phase changes, the overturning takes on far more complicated forms than would exist in a dry atmosphere. For this reason the parameterization process is extraordinarily difficult to formulate, and is occupying the attention of many scientists concerned with problems of the global circulation and tropical meteorology. It must also be of serious concern to those responsible for short- and medium-range prediction over temperate and subtropical regions, since the heat, moisture and momentum exchanged by convective processes are often of comparable magnitude to those associated with larger-scale disturbances (Newton [28]). To the meteorologist desiring to understand, predict, or modify convective systems themselves, the scale interaction problem is equally complex and compelling. Almost nowhere can cloud convection be considered as either completely random or com- pletely organized. Over the open tropical oceans, where surface and synoptic-scale forc- ing exist but are very weak, major convective clouds usually occur in aggregates, ' cloud clusters,' apparently forced by those very weak larger-scale disturbances. After the initiation and early organization of the clusters, the energy released by buoyancy apparently overwhelms that of the initial forcing mechanism, however, and the large- scale meteorological characteristics of the cluster become products of the integrated fluxes of the convective elements (Reed and Recker [29]; Ruprecht and Gray [30]). Large islands or peninsulas in the tropical oceans produce the most reliable organiza- Severe Storms and Storm Systems 719 tional structures since the diurnal heating cycle generates strong sea-breeze convergence and excites thunderstorms over land at nearly the same time each day. Even there, however, it is found that the majority of the rain falls out of a small minority of the storms (RiEHL [31]), principally including those which develop during periods of larger-scale disturbance conditions. In a few regions of the world and in certain seasons - notably including the central U.S. in spring - strong buoyant convection develops within the environment of moderate-to-strong larger-scale disturbances. This occurs when the pre-convective state includes a stable layer preventing moist low-level air from rising into a condition- ally unstable upper troposphere. When the stable layer is eliminated by large-scale or mesoscale uplift, the buoyant release then occurs suddenly and violently. The large- scale conditions which favor this release are well known and recognized (Fawbush, Miller and Starrett [32]), and useful 6-12 hr forecasts or watches can be made for areas of order 10^ square kilometers. The problem is that the actual development of convective cloud arrays occurs on a considerably smaller scale and often with a degree of organization which is clearly nonrandom but also largely unresolvable from con- ventional data processed in conventional operational ways. After its organization, the convective storm array develops very high energy density and one may suspect that, as in the tropical case, the larger-scale environment loses control and the tail wags the dog. In contrast to the tropical situation, however, the preconvective atmosphere has large horizontal gradients of temperature, moisture, and other parameters critical to convective energy generation, and also some of these parameters respond strongly to the diurnal cycle. Thus the convective storm groups must eventually run out of fuel, usually within 12 hr of their generation. A number of mechanisms have been proposed for the mesoscale triggering and organization of severe storm ensembles. In some cases they simply form along large- scale fronts, utilizing the frontal uplift to provide the destabilization necessary to start convection. More often, however, they develop along lines more-or-less parallel to a cold front, but some distance ahead of it. Tepper [33] proposed that the triggering mechanism was an internal gravity wave generated somehow by the front but running out ahead of it. Although some barometric evidence for such a mechanism was pre- sented, it was later realized that the pressure fields developed by the convective ele- ments themselves were so strong as to reduce the reliability of such evidence. Recently, however, Uccellini [34] has revived the gravity wave theory with new evidence. The control of convective storm developments by surface and boundary layer nonuniformities is so strong in clear-cut situations such as land-sea breeze and heated mountain regimes, that the boundary layer influence may be suspected to be an organ- izing or triggering effect even in much more complex situations. Studies have shown positive correlations between surface temperature and subsequent tornadoes. Most recently, satellite observations described by Purdom [35] have given evidence that the existence of early morning fog or low clouds exerts a strong organizing effect on after- noon convection. The proposed mechanism is that the cloud-covered region is delayed 720 Douglas K. Lilly in its diurnal surface heating cycle, leading to low-level mass and moisture divergence with compensating convergence and early and strong convective cloud development in the previously clear regions. Further evidence is needed, especially with respect to the reliability of the effect. The most frequent impulse for triggering additional new convective cells is the outflow from old cells. Although this effect should be strongest at short distances and undoubtedly aids in maintaining the storm as a coherent entity, under some (not clearly recognized) circumstances the outflow or a trapped gravity wave generated by it can extend out hundreds of kilometers from its parent storm and trigger new storms long after the original source element has died. This sequence has been detected from satellite pictures with the outflow boundary appearing as an arc line of clouds. The arc cloud line was apparently first described by Zipser [36] from satellite observations of tropical cloud clusters, and in aircraft flights through it he found it to be essentially an air mass boundary with the inside consisting of cooler and drier air originating from cloud downdrafts. Purdom [35] has recently described circumstances in which two arc cloud lines intersect, or a line intersects a frontal surface or region of pre-existing clouds or moisture, leading to the rapid development of a new convective storm. Also a recent example has been shown (see Fig. 4 from Erickson and Whitney [37]) of an apparent arc cloud line at the head of a train of short gravity waves, suggesting that the original mass outflow had become ' linearized ' into a travelling wave disturbance, probably propagating in a pre-existing stable layer or under a strong shear. A special situation related to severe storm formation in the southwest plains states is the existence of the 'dry line,' a front-like discontinuity between moist air moving northward from the Gulf of Mexico and dry desert air from the southwest. A close association between the position of the dry line and severe storm development imme- diately to its east has been long recognized (Fawbush et al. [32]). The dry line is not a density front in the usual sense, but it tends to develop some frontal features due to the strong contrast between the radiation and convection conditions and consequent diurnal cycles on opposite sides of it. Schaefer [38] has produced the most complete study of the structure and motion of the dry line in relatively undisturbed conditions. Danielsen [39] has shown that the deep dry convective layer on the west side of the line transports westerly momentum downward rapidly when a jet stream maximum passes over it, and that this downward westerly momentum is capable of producing severe dust storms. Sasaki [40], in looking at a different aspect of the same phenome- non, showed that the downward momentum transport has strong effects in generating and directing the motion of severe storm lines in the moist air to the east. 4. Theoretical and numerical approaches a. Cloud and severe storm models A recognition of the basic nature of thunderstorm convection has existed almost from the beginnings of scientific meteorology (Humphreys [41]). Nevertheless, a 8 Severe Storms and Storm Systems 721 z o u <<-■ o a "3 M >. 3 o 722 Douglas K. Lilly quantitative understanding of the dynamics of convective storms remains somewhat undeveloped. This is due in part to the highly nonUnear nature of cloud dynamics and the strong interactions between the widely different scales of motion that occur in severe storms. Probably the principal difficulty, however, has been the lack of obser- vations of the three-dimensional motion, thermodynamic and microphysical structure suitable for developing and testing theoretical and numerical models. Because of this lack of real data, the tornado in particular is perhaps the last frontier in tropospheric meteorology - the only intense and easily identifiable phenomenon whose internal structure and dynamics remain highly speculative. A recent account of the nature and status of the tornado problem has been provided by Davies- Jones and Kessler [42]. Most current efforts at theoretical description of buoyant cloud systems involve numerical models of one kind or another. The models can be conveniently divided into two classes. The zero- or one-dimensional models (Kessler [43]) which involve rather gross simplification of the fluid dynamics but may demonstrate careful attention to the microphysical processes, and the two- or three-dimensional models in which an attempt is made to simulate the convective dynamics with some degree of realism but usually the microphysics is treated rather crudely. In a zero-dimensional cloud model, the growth of cloud and precipitation particles is studied by integrating the time-dependent equations of condensation, coalescence, and other microphysical processes with the vertical velocity of the air specified as con- stant or a known function of time. Studies such as Berry's [44] integrations of the coalescence equations have been done in this way. The method serves to isolate the processes under consideration and to allow clear identification of numerical problems, theoretical inconsistencies, etc. Attempts to draw significant conclusions regarding the structure and development of real clouds from such models usually meet with skepticism, because of their inability to simulate entrainment effects, the fallout of precipitation into regions having a different history, and numerous other inadequacies. One-dimensional cloud modeling has an honorable history, an active present, and probably a useful future, but with important limitations. The one-dimensional models (with z and / and the independent variables) trace their lineage to the early models of turbulent jets, wakes, and mixing zones, and slightly more recently to models of buoy- ant plumes and thermals (Morton, Taylor, and Turner [45]). The basic assumptions of the plume-type equations are that the motion is lineally- or axially symmetric, that horizontal pressure gradients can be ignored (strictly valid only for a long thin plume), and that entrainment through the sides can be estimated from gross properties of the plume, such as its mean vertical velocity or turbulent intensity. Having accepted those limitations, or a similarly confining set for the isolated thermal models, the systems of equations are quite flexible and allow almost any degree of complexity of micro- physics to be simulated within a plausible fluid dynamic framework. In the severe storms context, the first assumption - lineal or axial symmetry - is sufficient to cast serious doubt on the adequacy of one-dimensional (and also most two- 10 Severe Storms and Storm Systems 723 dimensional) models. It essentially forecloses the possibility of modeling simultaneous updrafts and downdrafts except, perhaps, for the coaxial updraft-downdraft system which may occur in the absence of mean shear. The second assumption - the lack of consideration of horizontal pressure gradients in wide clouds - has been thought to be removable in principle, and Holton [46] has developed a method of calculating the horizontal pressure field which may point the way to significant improvements in such models. The entrainment assumption has periodically been subject to attack on obser- vational grounds, most recently by Warner [47], but a strong defense has been made (Simpson [48]; McCarthy [49]) and the matter remains unresolved. Despite their Umitations, the one-dimensional models are extremely convenient for both operational (e.g., cloud seeding) purposes, because of the ease and rapidity of a computer simulation, and for theoretical studies, because of the ability to incorporate rather sophisticated microphysics into a model that might be valid at least for the early part of a cloud lifetime. In the latter class, recent studies by Ogura and Takahashi [50] and by Danielsen, Bleck and Morris [51] are notable and should be studied carefully by those who would wish to improve the microphysics of more complete dynamics models. The first attempts at numerical solutions of a complete set of dynamic equations relevant to buoyant convection were made about 15 years ago in an attempt to simulate the buoyant thermal experiments of Scorer, Turner, and their students. These attempts have continued but have still not succeeded to complete satisfaction, perhaps mainly because of lack of adequate spatial resolution. Nevertheless, the results were encourag- ing enough that a large number of moist cloud simulation experiments have since been conducted, including some aimed at severe storm dynamics (Wilhelmson [52]; Steiner [53]; Hane [54]; Schlesinger [55]; Miller and Pearce [56]; Clark [57]). Most of these latter have been done using three-dimensional models, although in all cases the available resolution and/or numerical methods were inadequate to treat scales much smaller than that of the entire cloud cell, and the microphysical assumptions were crude except for Clark's two-dimensional model. The rather successful three- dimensional simulations of the buoyant planetary boundary layer and trade wind moist layer accomplished by Deardorff [58, 59] and Sommeria [60] may show the way toward the successful incorporation of small-scale turbulent interactions in future three-dimensional cloud models. The principal problems to be overcome in development of a three-dimensional model of a severe storm appear to lie largely in the area of modeling approximations and numerical methods. It appears possible in principle to incorporate all the dynamic and microphysical processes which have been previously studied in isolation into a more complex model, but the problem is to fit such a model into any available com- puter and integrate it in a reasonable time at a reasonable cost. If there are remaining serious physical problems (and this may not be determinable until the numerical problems are solved) they probably lie in some of the remaining uncertainties of the microphysical processes such as ice particle growth, drop breakup and electrical 11 724 Douglas K. Lilly interactions, and in the formulation of an open boundary condition, which can allow interactions with other storm elements and the larger-scale environment. The importance of developing improved numerical modeling techniques can be readily shown. The use of ordinary second order difference equations for approximat- ing advection terms requires 5 to 10 mesh points to adequately resolve a single Fourier mode. Thus, to simulate only the largest scale elements of a severe storm cell and those an order of magnitude smaller would require at least 50 points in each hori- zontal dimension, with a spacing of roughly 500 meters. Assuming the same spacing to be appropriate in the vertical direction, which is probably somewhat optimistic, indicates a need for at least 30 vertical points. This is just for the cloud, however. Recognizing that the environment for a distance of at least one cloud diameter in each horizontal direction is a critically important part of the storm system, one should increase the volume by a factor of 9, requiring 675,000 uniformly spaced mesh points. Five or six fields are required at each point to describe the velocity and the thermo- dynamic variables, with probably at least two additional fields for the submesh-scale turbulence structure. Thus more than 5x10® variables would have to be manipulated at each time step. This is considerably greater than the fast-access memory on any existing computer. Added to this, however, are all the cloud physics variables, which for any degree of reaUsm are likely to require at least 10 descriptors at each cloud point, i.e., another 10® variables. The solution to this problem must lie in the use of spectral or other functional space discretization, which can reduce the storage requirements for a given accuracy level, plus an effective recognition that the cloud environment, while important, need not be treated with the same spatial resolution as the cloud. Thus it should be possible to reduce the number of resolution elements in the cloud by nearly a factor of two in each direction, and outside the cloud by perhaps a factor of four in the horizontal and two in the vertical. This would bring the storage requirements per time step to some- what less than 5x10^ variables, more-or-less within the capabilities of the largest presently available computers. To state the solution in the above terms is easy enough, but to actually develop it will require a high level of scientific and creative programming skill. b. Regional and mesoscale models Numerical models for the simulation and forecasting of regional and mesoscale phenomena are being developed by a number of groups. There is no single pattern to such development because of the variety of mesoscale phenomena of interest or con- cern. Many such models are designed with boundary layer pollutant transport in mind, in which case the upper troposphere is essentially ignored. The important environment of severe storms is, however, the entire troposphere, and includes a rather large variety of physical processes. Thus a model successful in describing the development and progress of a severe storm array is likely to be as complex as a global circulation model. 12 Severe Storms and Storm Systems 725 The processes which such a model should describe accurately include frontal-scale baroclinic development, strong boundary layer transports and topographic irregulari- ties, gravity waves, and parameterized convective transports, at least for the pre-squall line convection. The lateral boundaries should be open to inflow, with boundary conditions determined from the time-dependent results of some larger scale model. Radiation above the boundary layer and stratospheric processes can probably be safely ignored. SAN LAX PGU OAK I hi SAN LAX PGU Figure 5 Vertical cross section between San Diego and Medford, Oregon showing a typical frontal zone (stippled). Dashed lines: isentropes (K); solid lines: isotachs (kt). Vertical coordinate is pressure on the left and potential temperature on the right. Note the much greater ability of the ^-coordinates to resolve the wind gradients in the frontal zone. (Furnished by Bleck [66] and Shapiro [67].) One of the more promising developments in regional and mesoscale modeling is the revival of isentropic coordinates for both analysis and forecasting by Danielsen [61], Eliassen and Raustein [62, 63], Bleck [64], and Shapiro [65]. (The particular advantage of the use of isentropic coordinates in meteorological data analysis is their ability to efficiently and economically represent strong gradients in frontal zones, as demonstrated in Figs. 5a, b [66, 67]). In forecast models these advantages are retained and, in addition, certain numerical problems are eased by the elimination, in adiabatic flow, of the vertical advection term. This is especially desirable when simulating un- saturated flow over orography. In early experiments Bleck [64] indicates that some of the advantages are retained in a moist model, even though the vertical advection term reappears. The principal problem thought to be associated with isotropic coordinates - the lower boundary condition - has been efTectively treated using methods developed by Eliassen and Raustein [62, 63]. 13 726 Douglas K. Lilly Probably the two most difficult problems for the development of mesoscale models in severe storm environments, convective parameterization and the open boundary condition, are current subjects of research by C. Kreitzberg, D. Perkey, and their colleagues at Drexel University. In their current model the open boundary condition problem is being solved in an engineering sense by a process of smoothing fields out as they approach the boundary (Kreitzberg and Perkey [68]). There remains, however, a fundamental applied methematics problem of well-posedness of boundary conditions (Oliger and Sundstrom [69]), and one suspects that such simplified solutions may come undone at unexpected times. The convective parameterization method being employed by Kreitzberg and Perkey is to separately integrate a cellular one-dimen- sional convection model of the Asai-Kasahara type (called H dimensional by Ogura and Takahashi [50]) at each point where buoyant convection is predicted to exist and to calculate the feedback to larger scales from the convection model output. It remains to be seen whether this rather straightforward and brave attempt will succeed scientifi- cally and economically. 5. Observational methods a. Radar and satellite pictures The most important tool for observing the structure of severe storms over the last decade probably has been meteorological radar. Until very recently most radars were only capable of measuring the reflected power received, which by accepted theory corresponds to the mean 6th power of raindrop radius. Thus radar reflectivity echoes tend to record the integrated effect of a convective storm's past history, as measured by its ability to produce large raindrops or hail. The signals are often rapidly changing and difficult to interpret in terms of the energy generation and dissipation processes of the storm. The recent development of doppler systems promises to greatly increase the quantity of useful information receivable from radar. Figure 6 shows the kinematic structure of a thunderstorm along a particular horizontal-vertical section, as inter- preted from the data obtained by a doppler radar pair. Besides the beautiful com- parison with the earlier conceptual model, it is clear that such data is highly suitable for correlation with or possibly initialization of a numerical cloud model. It is now becoming an urgent necessity for cloud modelers to learn how to use this new data source most effectively. The time and space resolution of satellite imagery of visible and infrared light has now reached the point where such data are a valuable complement and supplement to weather radar in defining storm location, development, and structure. While the satel- lite pictures do not have the resolution of radar and can only see the cloud tops, their geographical coverage is unbroken, and they show clouds too thin to have radar echoes. As with the radar echoes, however, satellite cloud pictures tend to represent the integrated effect of dynamic processes occurring over some previous interval. In 14 Severe Storms and Storm Systems 727 E 13 O CD > O < SI X 14 12 10 8 6 4 2 y = l2.6 km lOm/s "*' '^ "T i "^ '* * ■ i ' " — '• ' ' n ^1 i» 33 36 39 42 45 48 x- Distance East of Baseline (km) #► Stornn Motion lOr km Anvil ii.'i' >/","/* it'iii Precipitation ^Gust Front Outflow Figure 6 The upper picture shows an x-z cut of the velocity vectors measured by doppler radar in a thunder- storm in eastern Colorado. The lower picture shows a conceptual model of a thunderstorm as visualized by various earlier works (from Kropfh and Miller [70]). the case of convective storms, the effective resolution of sateUite imagery is often degraded by the shadowing effects of heavy cirrus anvil tops spreading over and con- cealing the active cloud cells. Nevertheless, the pictorial representation of a major storm outbreak like that of 3 April 1974 (see Fig. 7 from NOAA-N ESS [71]) is dramatic and revealing. At present two SDS/GOES (I and II) earth-stationary satellites are providing continuous visible and IR coverage over the central U.S. 15 728 Douglas K. Lilly Figure 7 ATS-III satellite photograph, 3 April 1974, 2100 GMT, showing three intense squall lines which produced large hail and numerous destructive tornadoes. Five intense tornadoes were in progress at the time of the photograph. (Furnished by NOAA-NESS [71].) b. Surface-based direct and remote sensing Because the energy source of convective storms is rising warm moist air, the bound- ary layer probably exerts a stronger control over their development than is normally the case in large-scale dynamics. For this reason surface and boundary layer nets are pivotal to the success of any severe storms observational program. During the last decade the U.S. National Severe Storms Laboratory (NSSL) has experimented with different net spacings of stations measuring the conventional surface meteorological parameters (pressure, temperature, humidity, and wind). In a recent report Barnes [72] has looked at the data from one such net and attempted to determine the spacing which optimizes the information content, i.e., provides a definitive analysis with the minimal number of stations per unit area. He found that for the wind field which is probably the most important for definition of convective storm environments, the optimization occurred with a spacing of order 10-15 km. The latter value was appro- priate when high time resolution was used to augment the limited spatial resolution with the aid of an assumption of steady-state structure of moving storm systems. Barnes's results were dependent on the time resolution of his analysis, which was five minutes. With a longer data averaging time, a stable analysis could presumably be obtained with fewer stations, but the ability to distinguish individual storm calculations would then be diminished. Since acquiring and monitoring surface data at such fine intervals becomes prohibitively expensive for either an operational or research net covering a large area, 16 Severe Storms and Storm Systems 729 these results indicate the need for further development of remote sensing techniques. There are several methods of ground-based remote sensing which produce poten- tially useful data on boundary layer structure, but none in immediate view which are capable of measuring boundary layer mean air flow over hundreds or thousands of square kilometers, as a doppler radar pair does, but in all weather conditions. Probably the closest approach thus far is the cross-wind laser anemometer, a simple and rela- tively inexpensive device for measuring the flow velocity components perpendicular to the laser beam and averaged over the extent of the beam. Although exceptional accura- cies have been shown in comparison with conventional anemometers spaced along the base (see Fig. 8 from Lawrence et al. [73]), reliable results have thus far been limited to about 10-km path length, and the device becomes inoperative in conditions of poor visibility. Acoustic and FM-CW radars have shown the ubiquity and variety of tur- bulent structures in both heated and cooled boundary layers, previously only observed from extremely powerful conventional radars. These systems have, however, mostly been utilized only in a vertical priority mode. NORTH 11/6/70 Figure 8 24-hr comparison between the wind measured by a cross-wind laser anemometer (upper curve) with the mean of seven propeller-type anemometers spaced evenly along the 1-km test path (from Lawrence et al. [73]). 17 730 Douglas K. Lilly c. Vertical soundings Vertical soundings by balloon-borne sondes remain the observational foundation of large-scale meteorology, and are also useful - in fact, presently irreplaceable - for defining the immediate environment and energy potential of convective storms. Again the NSSL has, in its spring observational programs, used special nets of about 10 sondes, spaced at different intervals in different years. The 1966-67 net, in which sondes were located at about 80-km intervals, was probably the most effective for observing the organized subsynoptic environment of the storms. In subsequent years the stations were arranged in a finer scale net, with spacing about 35 km in 1968 and 28 km in 1969 and 1970. The data from these networks have been error-checked and digitized. Although some of it has been utilized in various case study analyses (Barnes [12]; Fankhauser [73]), it remains a rich data source for further study. Sounding data is also available from meteorological satellites, using infrared and microwave radiance measurements to determine temperature and moisture profiles and cloud motion detection to determine wind vectors. Although the vertical resolution of such determinations is limited to a few degrees of freedom in the troposphere, the horizontal resolution may be higher than that of any practical sonde net. It is yet to be determined whether this large horizontal resolution can be used in an effective way in mesoscale meteorology and severe storm environments. d. Research aircraft The airplane is a versatile platform for meteorological observation. Some of the earliest middle tropospheric soundings were made by instrumented aircraft, but even after 50 years these platforms have still to reach their optimal utilization in meteor- ology. The recent incorporation of high-grade inertial navigation systems into the motion-sensing systems (Lilly and Lenschow [75]) has greatly enhanced their ability to define mesoscale kinematics, which has been most recently utilized in the GARP Atlantic Tropical Experiment (GATE), where 5-10 research aircraft made coordinated flights through active tropical convective arrays. The analysis of this data should tell us a great deal about the interaction of cloud systems with their larger scale environ- ments in the tropics. It is unlikely, however, that the same tactics can be applied to the much more energetic and dangerous convective storms in the U.S. Only especially heavily-built and armored aircraft have been used for planned flights through such storms (Sand and Schleusener [76]). The use of more conventional research aircraft in a storm circumnavigation mode has been described by Foote and Fankhauser [77]. The mounting of scanning doppler radar systems on an aircraft is the next major hoped-for development relevant to convective storm observation. With this addition, a single airplane could seek out a storm and fly over or beside it, continuously mapping out the velocity field within! It is hard to imagine a more significant increment to the tools now available for storm investigation and monitoring. 18 Severe Storms and Storm Systems 731 6. Critical questions and methods of attack a. What are the most important synoptic and mesoscale factors leading to forma- tion and guiding the propagation of a severe storm array ? Is the initial setup of a strong conditional instability controlled entirely by the synoptic scale, or do mesoscale moisture and temperature irregularities contribute importantly? What are the causes and effects of the 'streaky' humidity fields frequently observed? Is the final triggering often produced by gravity waves and, if so, what is their origin and condition of propa- gation ? What is the role of nonuniform boundary layer forcing ? After the outbreak of the first storms in an area, do the same processes continue to work or does outflow from the existing storms become the principal triggering mechanism for new storms? Answers to these questions must arise from a synthesis of existing and new fine- scale data, from meteorological satellites, surface mesonets and sensors, and arrays designed to detect and measure the structure of gravity waves. The nature of the arc cloud line must be determined more completely, by surface and upper-air stations and probably by instrumented aircraft. The data so obtained should be used as initial and verifying conditions for fine-mesh numerical models designed to experimentally pre- dict the initial severe storm outbreaks. b. What is the entire sequence of events leading to tornado formation and other damaging winds ? Is that sequence predictable on a useful time scale, and can intensity and location predictions be made from environmental and pre-tornado storm con- ditions? Why do some storm cells within the same environment develop cyclonic rotation and rightward propagation and others the reverse? Laboratory experiments have shown the importance of the 'swirl ratio,' the ratio of volume inflow to rotary circulation, for vortex development (Davies- Jones [23]; Jischke and Parang [78]). Can such a ratio be deduced from pre-tornado storm conditions and will it lead to early prediction of tornado development? Can the internal dynamics of tornadoes always be described by ordinary fluid dynamics and thermodynamics, or must unusual energy-focussing mechanisms such as electric current concentration be invoked to explain observed intensities? A program encompassing all of the above problem areas is being developed by the U.S. National Oceanic and Atmospheric Administration and other agencies. Entitled 'Project SESAME' (Severe Environmental Storms and Mesoscale Experiment), it is intended to provide a definite set of measurements of the severe storm environment over the scale range of about 5 to 1000 km. The observational techniques will include conventional, but upgraded, sonde and surface networks as well as sophisticated research aircraft and ground- and space-based remote sensing techniques. The Draft Development Plan (NOAA [79]) also envisages the utilization of two or more pairs of doppler radars for determining the air motion fields in convective storms over all or a substantial part of their lifetimes. The current plan proposes the first major field observation period to take place in the spring of 1979. 19 732 Douglas K. Lilly References [1 [2 [3 [4 [5 [6 [7 [8 [9 [10 [11 [12 [13 [14 [15 [16 [17 [18 [19 [20 [21 [22 [23 [24 [25 [26 H. R. Byers and R. R. Braham, Jr., Thunderstorm structure and circulation, J. Meteorol. 3 (1948), 71-86. H. R. Byers and R. R. Braham, The Thunderstorm (U.S. Government Printing OflBce, 1949), 287 pp. J. D. Marwttz, The structure and motion of severe hailstorms. Parts I, II and III, J. Appl. Meteor. 11 (1972), 166-201. K. A. Browning and F. H. Ludlam, Radar analysis of a thunderstorm. Tech. Note No. 5, Imperial College, London (1960), 106 pp. K. A. Browning and F. H. Ludlam, Airflow in convective storms. Quart. J. Roy. Meteor. Soc. 88(1962), 117-135. C. W. Newton and J. C. Fankhauser, On the movements of convective storms with emphasis on size discrimination in relation to water budget requirements, J. Appl. Meteor. 3 (1964), 651-668. K. A. Browning, Airflow and precipitation trajectories within severe local storms which travel to the right of the winds, J. Atmos. Sci. 4 (1964), 634-639. K. A. Browning, The evolution of tornadic storms, 3. Atmos. Sci. 22 (1965), 664-668. S. L. Barnes, Some aspects of a severe, right-moving thunderstorm deduced from mesonetwork rawinsonde observations, J. Atmos. Sci. 27 (1970), 634-648. A. J. Chisholm, Alberta hailstorms. Part I: Radar case studies and airflow models. Meteor. Monogr. 14 (1973), 1-36. D. W. Burgess and R. A. Brown, The structure of a severe right-moving thunderstorm: New sample doppler radar evidence. Preprints, AMS 8th Conf. on Severe Local Storms, Denver (1973), 191-196. S. L. Barnes, Papers on Oklahoma thunderstorms, April 29-30, 1970, NOAA TM ERL-NSSL- 69(1974), 125-151. T. FuJiTA, Proposed mechamism of tornado formation from rotating thunderstorm. Preprints, AMS 8th Conf. on Severe Local Storms, Denver (1973), 191-196. G. R. Hammond, Study of a left moving thunderstorm of 23 April 1964, IERTM-NSSL-31, ESSA (1967), 75 pp. T. FuJiTA and H. Grandoso, Split of a thunderstorm into anticyclonic and cyclonic storms and their motion as determined from numerical model experiments, J. Atmos. Sci. 25 (1966), 416-439. J. Charba and Y. Sasaki, Structure and movement of severe thunderstorms of April 3, 1964, ERLTM-NSSL-41, ESSA (1968), 47 pp. G. L. Achtemeier, Trajectories of thunderstorms and accompanying severe weather over the midwest on August 25, 1965, Proc. AMS 5th Conf. on Severe Local Storms, Boston (1967), 229-236. L. R. Lemon, Thunderstorm wake vortex structure and aerodynamic origin, NOAA TM ERL- NSSL-71 (1974), 17-43. E. A. Jessup, Interpretations of chaff trajectories near a severe thunderstorm, Mon. Wea. Rev. 700 (1972), 653-661. J. D. Marwitz and E. X. Berry, The weak echo region and updrafts of a severe hailstorm. Preprints, AMS 14th Radar Meteorology Conf., Boston (1970), 43-47. L. J. Battan, Doppler radar observations of a hailstorm, 3. Appl. Meteor. 14 (1975), 98-108. K. L. Grandia, Aircraft observations in the weak echo region of a thunderstorm. Preprints, 8th Conf. on Severe Local Storms, Denver (1973), 6-9. R. P. Davies- Jones, Discussion of measurements inside high-speed thunderstorm updrafts, Appl. Meteor. 13(1974), 710-717. J. S. Malkus, On the structure of the trade wind moist layer. Pap. Phys. Oceanog. Met., M.LT. and Woods Hole Oceanog. Inst. (1958), 47 pp. K. L. Grandia and J. D. Marwitz, Observational investigations of entrainment within the weak echo region, Mon. Wea. Rev. 103 (1975), 227-234. P. C. Sinclair, Severe storm air velocity and temperature structure deduced from penetrating aircraft. Preprints, AMS 8th Conf. on Severe Local Storms, Denver (1973), 25-32. 20 [27 [28 [29 [30 [31 [32 [33 [34 [35 [36 [37 [38 [39 [40 [41 [42 [43 [44 [45 [46 [47 [48 [49 [50 [51 [52 [53 [54 [55 Severe Storms and Storm Systems 733 D. N. SiKDAR and V. E. Suomi, Time variation of tropical energetics as viewed from a geo- stationary altitude, J. Atmos. Sci. 28 (1971), 170-180. C. W. Newton, The role of extratropical disturbances in the global atmosphere. In Global Circulation of the Atmosphere (ed. G. A. Corby), Salisbury Press, London (1969), 137- 158. R. J. Reed and E. E. Recker, Structure and properties of synoptic-scale wave disturbances in the equatorial Western Pacific, J. Atmos. Sci. 28 (1971), 1117-1133. E. RuPRECHT and W. M. Cray, Analysis of satellite-observed tropical cloud clusters. Paper No. 219, Dept. of Atmos. Sci., Colo. State Univ., Fort Collins (1974), 92 pp. H. RiEHL, Tropical Meteorology, McGraw-Hill, New York (1954), 372 pp. E. J. Fawbush, R. C. Miller and L. G. Starrett, An empirical method of forecasting tornado development. Bull. Amer. Meteor. Soc. 32 (1951), 1-9. M. Tepper, a proposed mechanism of squall lines: the pressure jump line, J. Meteorol. 7 (1950), 21-29. L. W. UccELLiNi, A case study of apparent gravity wave initiation of severe convective storms, Report No. 73-2, Dept. of Meteor., Univ. of Wisconsin, Madison (1973), 70 pp. J. F. W. Purdom, Meso-highs and satellite imagery, Mon. Wea. Rev. 101 (1973), 180-181. E. J. ZiPSER, The role of organized unsaturated convective downdrafts in the structure and rapid decay of an equatorial disturbance, J. Appl. Meteor. 8 (1969), 799-814. C. O. Erickson and L. E. Whitney, Jr., Picture of the Month - Gravity waves following severe thunderstorms, Mon. Wea. Rev. 101 (1973), 708-711. J. T. ScHAEFER, The motion of the dryline. Preprints, AMS 8th Conf. on Severe Local Storms, Denver (1973), 104-107. E. F. Danielsen, private communication (1974). Y. Sasaki, Mechanisms of squall-line formation as suggested from variational analysis of hourly surface observations. Preprints, AMS 8th Conf. on Severe Local Storms, Denver (1973), 300-307. W. J. Humphreys, The thunderstorm and its phenomena, Mon. Wea. Rev. 42 (1914), 348-380. R. Davies-Jones and E. Kessler, Tornadoes. In Weather and Climate Modification (ed. W. N. Hess), Wiley & Sons, New York (1974), 552-595. E. Kessler, On the distribution and continuity of water substance in atmospheric circulations. Meteor. Monogr. 10 (1969) 84, pp. E. X. Berry, Cloud droplet growth by collection, J. Atmos. Sci. 24 (1967), 688-701. B. R. Morton, G. I. Taylor and J. S. Turner, Turbulent gravitational convection from maintained and instantaneous sources, Proc. Roy. Soc. (London) A234 (1956), 1-23. J. R. Holton, One dimensional cumulus model including pressure perturbations, Mon. Wea. Rev. 110 (1973), 201-205. J. Warner, Comments ' On cumulus entrainment and one-dimensional models.^ J. Atmos. Sci. 29(1972), 218-219. J. Simpson, Reply, J. Atmos. Sci. 29 (1972), 220-225. J. McCarthy, Field verification of the relationship between entrainment rate and cumulus cloud diameter, J. Atmos. Sci. 31 (1974), 1028-1039. Y. Ogura and T. Takahashi, The development of warm rain in a cumulus model, J. Atmos. Sci. 30 (1973), 261-111. E. F. Danielsen, R. Bleck and D. A. Morris, Hail growth by stochastic collection in a cumulus model, J. Atmos. Sci. 29 (1972), 135-155. R. Wilhelmson. The life cycle of a thunderstorm: Three dimensions. Preprints, AMS 8th Conf. on Severe Local Storms, Denver (1973), 279-286. J. T. Steiner, a three-dimensional model of cumulus cloud development,}. Atmos. Sci. 50(1973), 414-435. C. E. Hane, The squall line thunderstorm: numerical experimentation, J. Atmos. Sci. 30 (1973), 1672-1690. R. E. Schlesinger, A numerical model of deep moist convection. Part I: Comparative experi- ments for variable ambient moisture and wind shear, J. Atmos. Sci. 30 (1973), 835-856. 21 734 Douglas K. Lilly [56] M. J. Miller and R. P. Pearce, A three-dimensional primitive equation of cumulonimbus convection. Quart. J. Roy. Meteor. Soc. 100 (1974), 133-154. [57] T. L. Clark, Numerical modeling of the dynamics and microphysics of warm cumulus convection, J. Atmos. Sci. 30 (1973), 857-878. [58] J. W. Deardorff, Three-dimensional numerical study of the height and mean structure of a heated planetary boundary layer, Boundary-Layer Meteor. 7 (1974), 81-106. [59] J. W. Deardorff, Three-dimensional numerical study of turbulence in an entraining mixed layer, Boundary-Layer Meteor. 7(1974), 199-226. [60] G. Sommeria, Numerical simulation of turbulent processes in the tropical boundary layer. In The GARP Programme on Numerical Experimentation, Progress Report, April 1973, WMO/JOC Working Group on Numerical Experimentation, A. Robert, Chmn. [61] E. F. Danielsen, The laminar structure of the atmosphere and its relation to the concept of a tropopause. Arch. Meteor. Geophys. Bioklim. All (1959), 293-332. [62] A. Eliassen and E. Raustein, A numerical integration experiment with a model atmosphere based on isentropic coordinates. Meteor. Annaler, J (1968), 45-63. [63] A. Eliassen and E. Raustein, A numerical integration experiment with a six-level atmospheric model with isentropic information surface. Meteor. Annaler, 5 (1970), 429-449. [64] R. Bleck, Short-range prediction in isentropic coordinates with filtered and unfiltered models, Mon. Wea. Rev. 102 (1974), 813-829. [65] M. A. Shapiro, A numerical simulation of upper-level frontogenesis with a 20-level isentropic coordinate primitive equation model, Mon. Wea. Rev. (1975) (to appear). [66] R. Bleck, from current work at the National Center for Atmospheric Research. [67] M. Shapiro, op. cit. [68] D. J. Perkey and C. W. Kreitzberg, A time-dependent lateral boundary scheme for limited area primitive equation models, J. Atmos. Sci. (1975) (submitted). [69] J. E. Oliger and A. Sundstrom, Theoretical and practical problems in formulating boundary conditions for a limited area model (to be submitted), 1975. [70] R. A. Kropfli and L. J. Miller, Thunderstorm flow patterns in three dimensions, Mon. Wea. Rev. 70i(1975), 70-71. [71] NOAA-National Environmental Satellite Service photograph, first published in Bull. Amer. Meteor. Soc. 55 (1974), front cover. [72] S. L. Barnes, Meso-network array: its effect on thunderstorm flow resolution. October 1974, NOAA TM ERL-NSSL-74, 16 pp. [73] R. S. Lawrence, G. R. Ochs, and S. F. Clifford, Use of scintillations to measure average wind across a light beam, Appl. Optics, 11 (1972), 239-243. [74] J. C. Fankhauser, The derivation of consistent fields of wind and geopotential height from meso- scale rawinsonde data, J. Appl. Meteor. (1974), 637-646. [75] D. K. Lilly and D. H. Lenschow, Aircraft measurements of the atmospheric mesoscales using an inertial reference system, ISA 1st Symp. on Flow, Its Measurement Control in Science and Industry, Pittsburgh, May 1971. [76] W. R. Sand and R. A. Schleusener, Development of an armored T-28 aircraft for probing hailstorms. Bull. Amer. Meteor. Soc. 55 (1974), 1115-1122. [77] G. B. FooTE and J. C. Fankhauser, Airflow and moisture budget beneath a northeast Colorado hailstorm, J. Appl. Meteor. 12 (1973), 1330-1353. [78] M. C. JiscHKE and M. Parang, Properties of simulated tornado-like vortices, J. Atmos. Sci. 31 (1974), 506-512. [79] Anon. Draft of SESAME Project Development Plan, NOAA-ERL Boulder, Colorado (1974), 77 pp. (Received 21st April 1975) 22 Reprinted from Bull. Am. Meteorol . Soc Vol. 57, No. 7, July 1976 aspects ol cumulonimDus study F. H. Ludlam Imperial College London, England Abstract Problems set by cumuloiiiinl)iis clouds demand recogni- tion of air motion on many scales and in three dimensions. Evolution of the storms and their precipitation depend essentially on the travel of particles across the air motion. The observation and computation needed to improve theory are formidable, and their expanse is so great that support is sought in the p(>tei\tial reward of artificial modilication. (However, it is likely that for a long time technique will be tentative and |>roof of effect, elusive.) Nevertheless, progress has been impressive and will con- tinue apace if the magnitude of the problems is faced and if some ruthlessness of generalisation is applied in the acquisition and synthesis of data. Kxamples are given of apparently simple questions difficult to answer with con- fidence; they should continually be asked to expose mis- conceptions and provoke fresh ideas. 1. General Very great advances have been made during the last tv^fo or three decades in the study of cumulonimbus clouds and of the violent phenomena that sometimes accompany them. The title of the proposed observa- tional study SKSAME (.Severe Eiivironmciital Storm and Mesoscale Experiment) is an indication of tlic modern recognition that satisfying progress rccpiires these studies to embrace a great range of scales, extending from the synoptic dov^iii to the microscopic; they can not profitably, as formerly imagined, be separated simply into a microphysics of the cloud particles and a macrophysics of the motion of the cloudy air. .\s the scale of observation changes, .so also does ilic ap- propriate method of measurement; especially in the intermediate scales, where the methods are unconven- tional, heavy penalties have to be paid for llexibility of deployment, elaboration of technique, and organi- zation of collaboration. .Vbove all, perhaps, ii is the great expense involved liiat imposes :i widely felt need for justification in terms of an cvitlent social beiiclit lo recompense the public purse and ilicrcby, all too olieii, a regrettable sense of strain on the workers. During the rise of meteorology as a science, this justi- fication was the oi»vious potential benefit of forecasting weather a day or so ahead. Particularly as an aid in marine and later in aerial navigation, this benehi was won, and it has amply repaid the cost of establishing synojjtic ob.serving networks. Even so, the worst weather hazards were overcome by ;idvances in technology rather than by improved forecasting: for example, the dangers of gales on to coasts by the replacement of sails by • A slightly amended version of an invited paper picseriu-d in absentia at the .SKSAMK. Ihundeislorm Scale W Oi ksliiip held S-:> March 1970 in Norman, Okla. 774 marine engines, and of airfield fogs by landing-aids or easy diversions. The weather hazards associated with cumulonimbus clouds have even smaller scales and dura- tions, and. con.sequently. useful forecasting here has very limited feasibility and application. Improved technology is generally, at least temporarily, limited to the lightning rod (perhaps the most impressively suc- cussful of all defenses against dangerous weather) and river control. (Nevertheless, where serious hazards like tornadoes affect wealihy tomnninitics. advanced systems of vvealhcr watch and rapid warning arc winning public a])pieciali()n.) 2. Particle physics Ihiis the justification for c\|)eiisive cumulonimbus studies has increasingly been sought in the possibility of the artificial control of thunderstorm behavior, es- |)ecially the crop-damaging falls of hail. The discovery of cllicieiit tLchiii(]iies of ice-particle nucleation, at .ibout the same time as the recognition of the prevalence of sn|)eicooled water in clouds and the theory that hail si/c vv.is coiurolkd by the sharing of this water .uiioiigsi a few ice particles, led to the idea that hail toulil be tliniinished or even eliminated inexpensively \i\ sowing the artifici.il nuclei, rcntative experiments with this .lim have been in progress for at least two (kt.ides, with results often greeted as encouraging but, ill ietios|)ect, always inconclusive. Ill recent years, strong doubts have been cast upon I lie idea that the sii|)ply of supercooled water is shared f.iiily ainoiigsi ihe droplets lli;il freeze first. liut it seems obvious tli.it if .ill the cloud droplets could be in- diiceil to liee/c .is soon as they become supercooled, ihiii iiDiie loiild ever bitonie sulhcieiilly large to ar- iive at llic ground with dani.igiiig size or without ineliiiig. or peril. ips even to reach the ground at all lieloie ev.iporaling. On the other hand, rough calcula- liiiiis suggest that to achieve such results artificial ice iiiulei would be needed in cpiantities .so large as to be more e\peiisive than any damage caused by the storms aiul would bring the risk of diminishing welcome rain- f.ill. riic possibility of a beneficial result must, there- lore. (Itpeiid critically upon the demonstration that iniicli smaller (pi.intilies wouhl .somehow hinder the process of li.iil growth. Ihe natural developineiil ol prec ipil.ition in cumu- lonimbus clomls is .ilu.ivs preceded by the generation ol jirecipiiation in iiinuiliis clouds. The interesting (iimulus (louds practically always have buses well below ihe ()"(: level, and the generation of precipitation be- gins by coalescence among tloud droplets. Ihe purely I'ol. 57, \o. 7, July 1976 23 Bulletin American Meteorological Society ns micropliysical aspccis of this process appear now to be represeiitctl by theory ratlier well, and it is generally understood that the evolution of the droplet si/e dis- tribution must be regarded as a statistical |}rocess, whose pace depends rather sensitively upon the proportion of the condensed water contained in droplets with a radius of at least several micrometers. Accordingly, in maritime clouds, which grow in rather clean air, the pace is rapid and may result in rain ap|>cariiig before the cloud tops have become supercoolctl. (The vertical dimension of cumulus clouds seems still to be the prop- erty most readily related to the development of the droplet spectrum, probably because in a i)i)|nilation of cumulus clouds all those reaching the same maximum height have contained similar updraft sj)eeds and there- fore have ])rovided a similar time for the development of the spectrum.) In contrast, cumulus clouds grown in air that has been over land a tlay or two have com- paratively large initial concentrations of dro|)leis, have a small average droplet radius, ami reach well above the ()°C level before particles sufliciently large to be regarded as |)recipiiati()n develop. A proportion, es- |)txially of the largest, are then likely to be frozen, and at first the rate of growth of frozen particles is more rapid than the rate of growth by coalescence. The |)rocesses of growth by coalescetice and by riming, and the development of secondary ice particles during rim- ing, then all proceed together and may be inextricably interwoven. Howe\er, it still seems likely that in many clouds the early progress of the particle accretion is by coalescence, and the level at which some particles attain the size of precipitation still seems to depend rather upon the time spent during the rise of air from the cloud base than upon its ri.se from levels where crystals first appear in significant coiucntrations. Inevitably, however gooil the llieory of particle aggregation appears to be, it can not be applied suttcssfully for lack of a sufficiently good model of the air motion in and near clouds. It has long been obvious that the concentrations of condensed water inside cunudus clouds imply th.it the motion is f.ir from adiabalic, but ihe evolution of the |>ariicle spectrum de])ends strongly not only upon the degree but also upon the manner in which dear air is lepears to be in a rather shallow layer, in spite o( the obvious dilution of the updraft with clear air as convection on an extending range of scales develops during ascent; this process must indicate progressive and significant departure from adiabatic conditions. 27 Bulletin American Meteorological Society 779 Fic. 1. Organized three-dimensional cumulonimbus clouds over Alberta (reproduced by courtesy of the Alberta Research Council; see Warner, 1976). Second, the cloud in the flow at the foot of the up- draft (near A) is notably free from small-scale detail, and the flow is therefore almost certainly locally 'forced,' against buoyancy, over a downdraft outflow. Where this downdraft is generated by precipitation is not obvious, but there are signs of glaciation or precipi- tation already among the cloud details above posi- tion C. Third, presuming from the low elevation of the sun that the picture was taken in the late afternoon look- ing a little east of north, the updraft inflow (near A) was probably from the southeast, and the outflow away from and to the right of the observer, toward the north- east: that is, the inflow was a southeasterly current and the outflow a southwesterly one. The downdraft that develops in precipitation (hidden) beyond the leaning towers at C (which grow into the dome D) evidently near the ground has an outflow toward and also to the right of the observer, in order to produce the cloud near A. The twisting of the main updrafts and downdrafts into a three-dimensional configuration is thus apparent. 6. Conclusion Numerous stimulating contributions to the study of cumulonimbus clouds have been made in recent years. The phenomena are so rich and varied that progress must be anticipated to occupy many more years. The description of tlie detail of these phenomena is itself an exacting task, but the greater need is for generali- zations that will include as much as possible in a set of simple laws: the objective must be not only analysis but also synthesis, and its attainment will require in- spired imagination as much as diligence and new technique. References Lilly, D. K. (Ed.), 1975: Open SESAME. Proceedings, SESAME Opening Meeting. Boulder, Colo., NOAA/En- vironmental Research Laboratories, 499 pp. Ludlam, F. H., 1976: Cloud Physics. University Park, Pa., The Pennsylvania Slate University Press (in press). Warner, C, 1976: Wave patterns with an Alberta hailstorm. Bull. Amer. Meteor. Soc, 57, 780-787. 28 3.1. THEORETICAL BOUNDARY LAYER GROUP REPORT November 5-7, 1975 L. Mahrt, Chairman J. Deardorff J. Wyngaard G. V. Rao J. Young W. Hooke 29 The scientific area of interest is the influence of the planetary boundary layer and "planetary boundary layer-free flow interactions" on the initiation of moist convection. The scale of interest includes meso or mesosynoptic (roughly 10 - 300 km, .3 - 10 hrs). Unfortunately, this problem involves several component problems whose observation requires widely different measurement systems. We plan to focus on the bulk behavior of the pre-storm planetary boundary and its interaction with the overlying free flow. These aspects of the planetary boundary layer will be examined in terms of their influence of the initiation of moist convection. While the important problem of "severe moist convection-boundary layer inter- action" and small scale flux parameterization can be examined with the observations prepared below, we propose not to include special instrumenta- tion necessary for extensive examination of these problems, at least for the first observational year. Our reasoning is as follows: (1) Interactions between the highly disturbed boundary layer and severe moist convection will, in part, be considered by the SESAME cloud scale group. With such interaction, the strongly disturbed airflow near the ground loses its boundary layer character and becomes more easily defined as part of the thunderstorm circulation. The proposed measurement program will address the problem of interaction between nonsevere cumulus convection and the boundary layer. (2) Flux-gradient parameterizations and turbulence higher-moment budgets require substantially different observational systems and are best done first under synoptical ly inactive situations. We expect improved flux- gradient rel iationships to result from existing programs (Haswell, 1975, JMOF, GATE). Special consideration will be given to this subject prior to SESAME (see last section). Of course, flux profiles resulting from tethered balloon, aircraft, and certain remote sensing measurements proposed for SESAME will yield useful information on the flux parameterization problem. The intention here is to avoid obligation of SESAME resources to this complex problem. 30 Meso-boundary layer events which are expected to be most applicable to the prediction problem, most definable and most likely successfully described by an initial observational program, are now discussed. The dry line and boundary layer west of the dry line involve rapid growth of the mixed layer to depths of several kilometers and possible entrainment of mid-tropospheric jet streams. Questions of special interest are: a. What is the dynamic response or adjustment to the rapid entrain- ment of momentun? b. Is this deep convectively "mixed layer" well mixed in momentum and specific humidity in spite of the baroclinity and rapid entrainment? What are the causes and variability of such gradients if they exist? c. On the high plains, is the eastward advection of mixed layers from higher elevations important? The influence of the boundary layer on the synoptical ly driven flow of moist air from the Gulf varies diurnally. On clear nights, the boundary layer may be decoupled from the stronger southerly moist flow above the boundary layer. In this case, the air above the thin nocturnal boundary layer can moisten more rapidly. The growing daytime mixed layer engulfs the moist flow of higher levels. In this case, the downward entrainment of moist air across the mixed layer top may be an important influence on the mixed layer moisture in addition to horizontal moisture convergence and surface evaporation. The moist flow may be significantly deflected by topographical features such as the Ozarks. This inflection is expected to involve substantial diurnal variation. A low level nocturnal jet may also be driven by accelerations associated with diurnal modulation of the turbulence and horizontal pressure gradients. How does this jet influence the divergence and moisture fields and thus influence thunderstorm development or suppression? Does the shallow stable nocturnal boundary layer have any significant influence on initiation of moist convection? 31 Pre-cloud convergence which is frequently thought to involve boundary layer processes, acts to destabilize the troposphere as well as concentrate moisture. Subsynoptic structure of this vertical motion field appears to be particularly important. For example, daytime low-level stability, concentrated mainly in an inversion capping the "mixed layer", may be strong enough to largely suppress convection. When such stability is weakened only locally, a few well organized moist convective systems can develop with exclusive rights to the moisture which is trapped below the inversion. Inhomogeneities in the hydrostatic horizontal pressure gradient can generate mass convergence. The determination of such convergence from pressure fields is difficult on meso-synoptic scales due to the influence of flow history and neighboring flow associated with the importance of temporal and advective accelerations. Examples of generation of horizontal inhomogeneities in the horizontal pressure gradient field include, (1) differential heating due to heating over sloped terrain or horizontal variations in surface insolation (variations in cloud cover) or variations in surface properties such as water content, albedo, etc., and (2) meso- synoptic structure (fronts, drylines, old squall lines or edges of thunder- storm-induced-mesoscale-cold-core-highs). Meso-scale generation of vertical motions also includes direct mechanical orographical uplifting and conver- gence associated with various low level jets discussed above. Local destabilization may occur due to a local maximum in surface heating. Horizontal inhomogeneities in surface heating can be generated by horizontal variations in surface conditions such as albedo or moisture content or horizontal variations in cloud cover. The efficiency of a given heating rate in destabilizing the lower troposphere is increased at a local maximum in surface elevation. Triggering of moist convection by internal gravity waves merits special attention. Important wave sources might include topographical forcing (rising motion, heating), ageostrophic adjustment, shear instability, penetrative convection, moist convection and wave coupling with moisture- energy release. Examination of wave propagation includes the possibilities 32 of free propagation, energy trapping, partial reflection, critical levels, nonlinearities as well as boundary layer effects. Of interest is whether surface observations plus knowledge of mean wind and temperature profiles suffice to specify the wave fields. The mechanism of wave triggering of severe-storm development probably involves direct lifting of the inversion to the mixing condensation level and the influence of Kelvin-Helmholtz wave generation on inverstion dynamics. The first mechanism likely involves wave induced boundary layer pressure gradients. General goals of the measurements are to identify the bulk planetary boundary layer characteris- tics on the a and 3 scales for the pre-storm environment, including the relationship with the overlying air flow characteristics. Local quantities which need to be measured on the mesoscale include: a. Conventional surface measurements (PAM) as well as surface moisture and heat fluxes and downward fluxes of moisture and heat in the mixed layer particularly near the mixed layer top. We anticipate a grid of roughly 10 km for such observations. b. Monitoring the height and strength of the inversion capping the mixed layer and height of the lifted condensation level. We propose to devote extra attention to representative wind measurements which will allow useful calculations of divergence and its influence of the mixing depth, mixed layer moisture budget and vorticity. Local exposure problems must be minimized. We also plan to consider time and space averaging to reduce influences of low frequency turbulent flucturations on measurements of mean quantities. The following analyses will be carried out prior to the observational program: a. Test gravity wave monitoring ability of microbarograph networks (see working report of gravity wave group (Hooke)). b. Analyses of existing data on the high plains boundary layer west of the dry line (project DUST STORM (Danielson, NCAR) and NHRE)). c. Local study of structure of the growing mixed layer (Haswell, 1975, (Lenschow, NCAR)). 33 d. Climatology of low level divergence over Great Plains (Barber and Mahrt). e. Numerical modeling of mesoscale inhomogeneous boundary layers (Wyngaard). In addition, we feel that the magnitude and scope of these problems and the uncertainties they represent are quite great. Failure to improve our understanding of these will result in significant inefficiencies in SESAME execution. Continuing research on mesometeorology and micrometeorology might be done with an eye toward its possible application to SESAME. 34 3.2. OBSERVATION AND INSTRUMENTATION GROUP REPORT SESAME F. Hall S. Garstang G. Goff F. Hall M. LeMone D. Lens chow J. Marwltz J. Purdom R. Reinking J. Telford S. Williams 35 I . INTRODUCTION The scientific objectives of the boundary layer observation program are to measure the undisturbed, fair-weather atmosphere as well as the conditions during the life cycle of severe storms. Specifically, the measurements will be designed and coordinated to provide the data set necessary to answer the questions of greatest importance in improving our understanding of boundary layer triggers for severe storms and the inter- action of the boundary layer with the storms as they evolve. In the previous sections on theoretical considerations and on modeling of boundary layer-storm processes, critical meteorological parameters were identified, the accurate measurement of which will enhance the achievement of SESAME goals. These same parameters enter into descriptions of the boundary layer conditions given in Appendix A where the characteristics of storm evolution, maturity, and decay are discussed together with the capabilities of various instrumentation methods. A synopsis of these parameters is listed along the left column of Table I-l in descending order of their relative impact and priority for SESAME. Thus there seems to be general agreement among the theorists, modelers, and observationalists that wind fields are the most important in understanding storm development and that derived parameters such as divergence, both mass and moisture, are of the first priority when the inevitable trade-offs of feasibility and cost of measurements are to be considered. Listed across the table are the candidate instrument systems, also in descending priority toward the right. The immediate question arises, do we really need all of these instruments? Would not a preliminary, small-scale, "pre-SESAME experiment" wherein comparisons are conducted among the various systems be a good invest- ment in enhancing the success of the full-scale experiment? Clearly the utility and number of independent measurements each instrument system can perform varies widely. Listed in the table are estimates of the absolute accuracy and the frequency response of the sensor. For the more conventional sensors, these figures are reasonably well known and reliable. For some of the newer instruments, accuracy figures are more tentative and need to be better determined. The advantage of the remote sensing instruments is their ability to achieve measurements at a number of different height 36 increments or ranges nearly simultaneously; clearly a combination of Zn ^aMi and remote sensors will provide the optimum boundary layer instrumentation integrated system. (For some of the remote sensing systems, the maximum range is also listed.) There will be some overlaps between Table I-l for the boundary layer systems and for those measurement techniques which the cloud physics group will find desirable. 37 SESAME APPENDIX A Planetary Boundary Layer Observations and Instrumentation A-I. INTRODUCTION Before we identify the critical questions which the boundary layer observations are designed to answer and before we consider the tools required to make such observations, it is relevant to review the status of our knowledge of the boundary layer during three identifiable and different stages: • The evolution of the boundary layer on suppressed days or before radar echoes appear. • Boundary layer characteristics during the storm stages. • The boundary layer during the decay of storms. We then need to convolve the present knowledge of the boundary layer with the identified needs of the theoreticians and modelers, discussed in the previous sections, to define those fields of meteorological parameters which the boundary layer observations must provide. Then we will be in the position to discuss the critical products of these parameters and the goals of the analyzed observations. Only then can we define the tools necessary to supply these measurements. A-II. EVOLUTION OF THE BOUNDARY LAYER There are two ways the planetary boundary layer may change during the suppressed convection or pre-storm condition: Because of the diurnal change in solar radiation, and because of advection of different air masses into the SESAME region. 38 a. Diurnal Development A typical sequence of events occurring after daybreak and the inception of insolation is the development of a well-mixed layer caused by surface heating which, in turn, grows by entrainment of the stable air aloft resulting in the transfer of heat and moisture from the surface to the upper part of the boundary layer, while momentum from winds aloft is transported downward to the surface. To describe this process we need to measure continuously in time the fields of temperature, moisture, and winds. Large local horizontal gradients may occur in these fields, fronts or dry lines, which are important triggering mechanisms for severe storms. Gravity waves or inertial oscillations may produce fluctuations in these fields and may also be important storm initiators. The intrinsic variables of temperature, moisture, and winds are determined by the following inputs: • Surface fluxes of heat, moisture, and momentum. • Horizontal advection of heat and moisture. • Mesoscale and synoptic scale pressure gradients. • Entrainment processes at the top of the boundary layer. These four inputs are in turn determined by: • Insolation. • Surface characteristics such as orographic effects, roughness, albedo, soil moisture, and vegetation. • Cloud cover and atmospheric turbidity which produce, in turn, shading of the surface and the atmospheric absorption of solar and terrestrial infrared radiation. 39 • Stability of the overlying free atmosphere, characterized by the gradient Richardson number and by vertical gradients of temperature, moisture, and wind, b. Advectlve Development In parallel with diurnal development of the boundary layer, advectlve processes can significantly modify the boundary layer. Even In the absence of solar heating, advectlon effects may trigger storms. Even differential advectlon may destabilize the atmosphere. The nocturnal jet may be Important in this process and Inertlal oscillations, resulting from changes in the boundary layer input forcing may play a role. A-III. BOUNDARY LAYER CHARACTERISTICS DURING THE STORM STAGES Characteristics of the boundary layer structure during storms may be conveniently divided into two spatial classes, those characteristics observed near and under storms, where the boundary layer interactions with the storm environment are marked, and those characteristics of the boundary layer in regions removed from the Immediate vicinity of storms, perhaps 10-20 km away, but representing boundary layer structure on the disturbed days. For the first case, near and under storms, we clearly need a better understanding of the mass and moisture convergence fields and the roles of mesoscale low level jets and moisture tongues in maintaining the thunderstorm circulation. The influence of propagating gravity waves on storm enhance- ment and the variation of such wave effects as a function of amplitude and wave length of the waves needs to be better understood. The effects of inhomogeneltles in the stability and vertical structure of the boundary layer on thunderstorm persistence needs closer investigation. A knowledge of the vorticity in the boundary layer, the interaction of this vorticity 40 (conservation, advectlon, and convection) leading to rotating storms and tornadic vortices have all been modeled, but never adequately measured. For rotating clouds, we need measurements to understand the coupling of the boundary layer and the thunderstorm updraft which results in producing a condensation funnel. Better classification of tornadoes including the structural characteristics, intensity, and life-span of such vortices is needed. We need better measurements of the coupling of the mesocyclone with the boundary layer and how the outflow shear boundaries influence the intensity and life of the cyclones. Finally, better measurements are required to understand the coupling and influence of the density current or cold air outflow from storms on the updrafts into storms. Investigations of the boundary layer environment unperturbed by the immediate influence of thunderstorms, but during periods of moist convection in the region, should include supplemental measurements at distances of a few tens of kilometers from the storms. Again, the three-dimensional field of winds, temperature, and moisture and fluxes of momentum, heat, and moisture should ideally be known throughout the well-mixed layer. The diurnal variation in depth and character of this layer needs documentation to determine the influence of z^ on the storms and the feedback mechanisms on z. from the storms. During periods of nocturnal thunderstorms, the presence of the nocturnal jet and the characteristics of the boundary layer capping inversion should be monitored. A-IV. THE BOUNDARY LAYER DURING THE DECAY OF STORMS Given a well-developed storm, there are a variety of processes which may lead to its decay. SESAME can provide the first concerted effort to under- stand the relative importance of the various boundary layer-related 41 mechanisms on storm decay. Adverse factors which affect the storm include internal interactions within the storm, interactions among storms, inter- actions between the storm and its environment, and mesoscale stabilization of the environment. Internal interactions which may cause storm decay include excessive water- loading of the updrafts, horizontal momentum entrainment so that the updrafts become tilted with the precipitation falling into the updrafts, entrainment and ingestion of a new nuclei population upsetting the established steady- state precipitation process, and finally cut-off of the inflow to the storm by either the cold air outflow increasing in depth up to the cloud base or extension of the outflow nose so far ahead of the storm as to not reinforce the updrafts. Adverse interactions among storms are possible in a variety of ways. Overhanging cirrus anvils from upstream storms may seed developing storms, causing glaciation at an earlier stage than would occur naturally. Shadows from the anvils may stabilize the boundary layer, cutting off the dry convective roots of developing storms. The interception of available high energy air in the sub-cloud region by storms located upstream may rob downstream storms of their energy source. The cold air outflow from up- stream storms may also interfere with the moist convergence into developing storms, leading to their demise because the cold air outflow has low energy and is statically stable. Storm interactions with the external environment which may lead to decay include the processes of return flow external to the cloud, leading to stabilization of the surrounding atmosphere. Dry air entrainment into developing storms is another mechanism. Limited entrainment fails to produce 42 sufficient evaporative cooling, degrading the cold air outflow which can reinforce the updrafts. Excessive entrainment , on the other hand, may destroy the updrafts in the cloud by excessive evaporative cooling as well as producing excess tilting from the wind shear and conservation of momentum. A last factor which may cause storm decay is thermodynamic stabilization of the environment irrespective of the storm interactions. The storm may simply use up the available high energy air in the sub-cloud layer if it is not replenished. Diurnal cooling of the planetary boundary layer may destroy the convective roots of the storm. Advection of warm air parcels aloft over cooler air below will lead to stabilization of the environs of the storm. Finally there may be a decoupling of the upper level divergence pattern with the lower level convergence associated with the storm. In all these mechanisms it is clear that the fields of winds, moisture, and temperature and the associated fluxes in the boundary layer, in the free atmosphere, and within the storm need to be known. A-V. DEFINING THE FIELDS OF METEOROLOGICAL VARIABLES From the foregoing description of the interaction of the boundary layer with storms and from the sections on theory and modeling, it is clear that the fields of winds, moisture, and temperature are of primary importance in determining the initiation of moist convection. Other parameters are given in Table I-l. It is necessary to measure these fields accurately enough, often enough in time, and densely enough in the three spatial dimensions to provide those sets of measurements required to compare with theoretical or modeling calculations of these fields. Furthermore, the measurements must be of sufficient quality to allow the calculations of the differential properties of the fields, such as convergence and vorticity. 43 It would be desirable if the measurements were obtained at the same or smaller increments than any scale parameterized in modeling efforts so as to furnish the sufficiently resolved picture of actual boundary layer conditions. Although it does not show specifically in Table I-l, we need to define the fields of meteorological parameters from the earth's surface through the lower cloud layers. Surface characteristics such as temperature, albedo, insolation, and infrared radiation may need defining on the horizontal scales on meso-y horizontal scales and on time scales of a few tens of minutes, measurements which may be possible with ^n -i/Ctu sensors and by meteorological satellites. In the atmospheric surface layer, to heights of perhaps 30 m, the same sort of spatial and temporal scales are appropriate. Thus if the 1978 experi- ment envisions a 40 km triangle, some of the instrumentation should be placed on considerably closer centers. VI. CRITICAL PRODUCTS INTRODUCTION The problem of describing the planetary boundary layer by measurements in sufficient detail to be sure important changes are adequately represented is both one of obtaining sufficient measurements and separating our measure- ments in time and space so as to best elucidate our current concepts. While it does not seem necessary to worry about obtaining too many measurements, this concern is in fact associated with another problem. The number of measure- ments obtained in any useful measurement program will be enormous and these must be processed in such a way that they can be assimilated by theoreticians and used by experimenters to plan the best use of the observational capability during the experiment. 44 Thus, the product must be evaluated in terms of real time displays, quick look data, and finally the data product output processed to a stage where further work is essentially conceptual evolution rather than routine data processing. This approach postulates a field data processing center where graphical output can be prepared for evaluation in periods like hours. Let us postulate unlimited real time support and from this see what might be done, and then consider what is most desirable. THE TOP OF THE BOUNDARY LAYER AND ITS STATE One particularly simple measurement is the height of the top of the planetary boundary layer. This can be measured with acoustic sounders, FM-CW radar, balloon ascends and from tethered balloons, as well as from aircraft ascents as shown in Table I-l. A constantly updated real time contour map of this height with superimposed potential virtual temperature contours and lifting condensation level contours would be extremely useful. We do not know to what detail this depth is a function of position and consequently we may meet variability which makes automatic smoothing and contour drawing impractical. On the other hand it may be possible to fit these three variables to a linear function of position and time and so greatly simplify presentation. The automatic generation of this display, showing contours of depth and potential virtual temperature with an outline of regions where the lifting condensation level is within the mixed layer, and of areas with cloud cover, is of top priority. THE PROGRESS OF ADVECTION Since we are attempting to determine the initial onset of instability it is important to display the density of the overlying air and its advection 45 rate relative to the planetary boundary layer. The boundary layer is nearly homogeneous but above it the density structure is much more complex. All soundings should be processed to give mixing ratio, potential virtual temperature and wind magnitude and direction, as a function of height; possibly these should be dumped on the plotter immediately. However, we would suggest that the first 1000 m or so immediately above the convective layer in each sounding be fitted to a straight line and the results displayed as suggested for the boundary layer itself, with the addition of the slope of the potential virtual temperature added to indicate stability. Wet adiabatic instability should also be displayed. To complete this form of display this data should be combined with the boundary layer data and projected in time using the best winds so regions of increasing stability of the boundary layer relative to the overlying air can be delineated. Similarly, contours of decreasing stability could be drawn showing time until instability occurs. MANUAL EVALUATION The scheme for a useful "quick look" product in less than an hour behind the measurements may provde to be impractical and in any case will need to be evolved with human interaction. Such a data display could be simulated by human evaluation of graphical soundings and time series. If several scientists are available for data assessment, they could be assigned a set of sounding sites each for coordinating. NON-REAL TIME If this form of "quick look" product cannot be sustained, we lose the advantage of directing the aircraft at regions of incipient instability. However, such analysis seems to be essential for fast estimates. Furthermore 46 such analysis and real time work does insure that the data is processed and that the data does become readily available. THE MATURE STORM While the mature storm phase may be expected to be more complicated, the same displays as discussed above should be attempted. We now will have radar picture to use together with the other data. Care should be taken so all products are directly comparable and plotted maps are always on the same scale and can be matched to radar photographs by direct overlap. 47 r — DERIVED -^ r~ ■A i i I ■■»: "I 5! I ^1 I d ^ ^ ^ ^ "^ ^ § ^ ^ ^ ^ ^ :\. .:2 H ^ ^ o o $ I J? .^ at ?: t 5< > ^ ^ >1 i! o ill ^ ^ 5; ^ (:) ^ 5 95 o O ? Go ^ 5 Hi 3 P 0: 3 o H, (7) -"■ 1 ^ I -a: ii ~!T 5, A, CAR FA '/I RAWiNSONr L'-: C0N\/f:.\<7'CAj-U. OR MEiTRAC A! RCRAF'7 NCAR OR N^AA TETH5R5D BALLCCA! DRONE. AIRCRAFT LASER TRANS,VER6E WIND A m o o 3 ^1 I p ^ ^ s .^ » t/* f^ 5^ b>r^ 3 ^ i^rt ^ i I o o C3 Oh ^ & iJ-h o t. (5 Ik fr ^c o o AC0U6TIC SOUNDER , DORPLER_ 'puL'sed'^' DORPIER RADAR FM- CIA/ RADAR, DOPPLER MICRO (5) I>1 r I -0 i DOPPLER L/DAR 48 3.3. REPORT OF THE REGIONAL SCALE MODELING GROUP November 5-7, 1975 THE ROLE OF HYDROSTATIC MESOSCALE MODELS IN PROJECT SESAME Richard A. Anthes, Chairman Edwin Danielson Carl Kreitzberg Jan Paegle Roger Pielke Joe Schaefer Wilson Shaffer 49 1. Introduction Numerical models of atmospheric circulation patterns that range from the large, synoptic scale (wavelengths L > 2500 km) through the three (arbitrary) divisions of the mesoscale Meso - a 250 < L < 2500 km Meso - 3 25 < L < 250 km Meso -Y 2.5 me^an wind speed in PBL (y) , and— mean wind pro- file (-J"^) important atmospheric parameters to determine response of at- mosphere to flow over rough terrain. Horizontal subgrid-scale Covariances such as fluxes of heat, momentum , and moisture u'v' ,u'G u'q' Current mesoscale models do not realis- tically parameterize these eddy fluxes. 53 2. 1 The planetary boundary layer (PBL) Because knowledge of the time-variation of the structure and height of the PBL is thought to be essential to the understanding of the severe storms problem, the realistic parameterization of the PBL is of primary importance in the models. While the possibility of utilizing a simple "mixed-layer" parameterization may be considered, the consensus is that a number of levels within the PBL will be necessary to resolve possibly important vertical variations of momentum and moisture. With surface fluxes of heat, moisture and momentum computed from a surface energy budget calculation and established similarity theory, vertical transfers of these quantities will be calculated in most models by "K" theory, in spite of the non-rigorous nature of the flux- gradient relationship. Higher-order closure schemes are potentially superior to the "K"-theories, but their expense and complexity make their use at this time unrealistic. The vertical mixing coefficients will depend on such parameters as the height of the PBL, surface heat and momentum fluxes, and locally on the vertical wind shear and stability. The predictive, as opposed to diagnostic, calculation of the surface fluxes of heat and moisture requires a simple energy budget which utilizes for input only external parameters such as time of day, year, latitude, and condition of soil. This budget is necessary to calculate the portion of the insolation that is utilized to evaporate surface moisture and heat the air in the contact layer. The surface flux of heat is necessary to calculate the Monin-Obukhov length, L, which is important in establishing the momentum and temperature structures in the surface layer. The surface flux of moisture is important in augmenting the water vapor supply in the PBL, which in turn exerts a strong influence on the evolution of moist convection. This addition of water vapor through evaporation helps to maintain high surface dew points in a boundary layer thst is rapidly srowing and entraining dry air through the inversion. The verification of the model's predictions of surface temperatures and moisture will be an important part of the overall verification program. 2 .2 The parameterization of shallow and deep cumulus convection The parameterization of the effects of moist convection of the larger- scale atmosphere is one of the important problems to be addressed by project SESAME. Because the hydrostatic models cannot explicitly consider moist convective circulations, the cumulative effect of these cloud-scale motions 54 on the temperature, moisture, and momentum structure of the mesoscale atmosphere must be related to the circulations that are resolvable by the model. Both theoretical and observational results indicate that the mesoscale horizontal convergence of water vapor is the most important parameter in determining when and where organized moist convection will occur. Values of moisture convergence -4 -1 -1 as high as 10 gm kg s may occur on the mesoscale. The mesoscale moisture convergence over an area provides an integral constraint on the amount of precipitation and associated latent heating that are possible. The vertical and horizontal distribution of the heat, moisture, and momentum fluxes by the convection is determined by the spectrum of clouds present. Current cumulus parameterization schemes vary from simple ones in which a number of cloud types are permitted. To aid in the verification and refinement of the cumulus parameterization schemes, it would be useful to determine the characteristics of the cloud population, including (a) the height and radius of cloud bases, height of cloud tops; (b) mean cloud vertical velocity, temperature, liquid water, horizontal momentum, and vapor content at a number (say 5) of levels. These cloud properties should be averaged horizontally over the area of the cloud; (c) the average temperature, horizontal momentum, and water vapor content in the environment of the clouds. From the above information, the vertical eddy fluxes of heat, moisture, and momentum can be determined. 2.3 Long- and short-wave radiation Most mesoscale models do not yet contain the effects of radiation flux divergence in the free atmosphere; only the radiation budget at the ground is considered. This neglect is probably justified for most mesoscale predictions because of the short time scales involved and the dominance of other physical processes in determining temperature changes in the free atmosphere. Radiation would probably be most important in the evolution of the nocturnal PBL. Reasonably simple radiation models are available to treat this process. The most important variables that determine the long-wave flux are cloud cover (depth and height) and the average vertical profiles of temperature and moisture. 2.4 Terrain effects Because irregular terrain affects the low-level flow in a number of ways (e.g., through generation of gravity waves, production of circulation by differential heating and friction), mesoscale models will contain a smoothed 55 topography. Variations in terrain with horizontal wavelengths less than approximately four times the grid size will be eliminated; therefore, individual hills will not be represented by the meso-a and 3-scale models. 2.5 Initialization The initialization of mesoscale models remains a big problem. While there is general agreement that the horizontal wind is the most important variable to initialize mesoscale models, it is not known how much of the meso- scale variability in the wind field must be included in the initial conditions. It is a hope that the smaller-scale perturbations in the V7inds will evolve with time from the larger-scale circulations, and the physical forcing functions. Nevertheless, as much resolution in the initial wind field thst is economically practical is very desirable. Because the radiosonde appears to be the only reliable method of determining the wind at a number of levels under all weather conditions, it is desirable that an increased number of radiosondes be utilized in a way that provides the mesoscale models with wind data on the scale required by these models. For model initialization purposes, an average radiosonde spacing of ~ 200 km would be the most efficient use of additional radiosondes because the domain size is in the order of 1500 km (Fig. 1). Also, since high-resolution temperature data are less important than wind data, economical sounding systems that provide only wind data would be extremely valuable. Because there is considerable evidence that upper-level disturbances crossing the Pacific Coast and moving eastward over the Rockies play a fundamental role in generating many types of severe weather, it is important to monitor the wind and temperature structure upwind of the Great Plains. Thus, a somewhat increased density in radiosonde stations along an approximately north- south cross section through Arizona, Utah, Idaho, and Montana would be highly desirable (Fig.l). Such a cross section would be useful in the initialization of the meso-a scale models and in the real-time forecasting of severe weather. One possible arrangement of additional radiosondes would be between Great Falls, Montana, and Salt Lake City, Utah; another between Salt Lake City and Winslow, Arizona; and a third south of Tucson, in Mexico. Other approximately north- south paths would be acceptable. For models on the meso-a scale, which will normally be initialized approximately 12-24 hours ahead of the outbreak of severe weather, a fairly coarse analysis of moisture will probably be sufficient. For the smaller-scale models which may be initialized only six hours prior to the development of convection, a finer resolution of the moisture field may be necessary. Observa- 56 tions indicate that the moisture field has significant horizontal variability down to scales as small as 100 km in the vicinity of severe weather. Satellite data may provide some important information on the horizontal variability of moisture, but the satellites cannot provide very much detail in the vertical. Because the meso-B models cover less area than the meso-a models, the high- resolution data need to be collected over a region approximately 400 X 400 km. Thus, aircraft observations may be feasible to determine the horizontal variability in the region of interest. Finally, the initial depth of the PBL will be useful in the initialization of the mesoscale models on both the a and 6 scales. If the models are initialized at night, both the depth of the nocturnal inversion and the depth of the previous day's PBL would be useful. 2.6 Verification The same variables that are required for the initialization of the models will be useful for the verification. In addition, detailed temperature observations in the vicinity of baroclinic zones will be helpful. The time- history of the patterns in the mesoscale convective systems as observed by radar and satellites will be useful, as will accumulated precipitation over the domain. Data on the small scale (L < 100 km) that enable budget calculations for the calculation of cumulus and turbulent fluxes are needed to verify and improve parameterization techniques. This testing of parameterization schemes is the main way that the meso-y-scale observations will be useful to the meso-a and meso-3 models. Although relatively unimportant for initialization of these models, data that enable calculation of vertical and horizontal fluxes by sub- grid scale eddies in the PBL and determination of representative horizontally- averaged properties of a spectrum of clouds will be very useful in evaluating and improving the parameterization of the PBL and cumulus convection. 3. Use of various types of observations in mesoscale models The preceding sections on the parameterization of physical processes in the mesoscale models and on the initial and verification data requirements have indicated the types of data that will be most useful to the models. Table 3 summarizes the potential utility to the mesoscale models of the observation systems proposed for project SESAME. 57 Table 3 Utility of Observation Systems in Mesoscale Models Radiosonde; A. Temperatures useful in initializing models and in establishing absolute temperatures at points for calibrating satellite-derived temperatures. B. Winds at many levels are essential in the initialization of models, C. Moisture information useful in initialization of models and in budget studies. Chief advantage - Most reliable system for providing simultaneous wind data at many vertical levels. Chief disadvantage - Expense, difficulty of resolving horizontal gradients on small scales. Surface data; A. Surface (~ 3m) wind data useful in determining low-level convergence of water vapor. B. Surface pressure data useful in monitoring gravity waves as well as intensification or decay of mesoscale pressure systems. Surface pressure change patterns reflect total mass divergence and may be of some use in the initialization of models. C. Surface temperature and dew point useful in determining surface fluxes of heat and moisture. A short (3m) tower at each radiosonde station with temperature, moisture, and wind at two levels (e.g., Im and 3m) would be useful in testing flux-gradient relationships. An independent measurement of vertical fluxes would be important. D. Surface temperature and dew point may be used with knowledge of height of PBL during day to interpolate data to model levels within PBL. Chief advantage - Provide lower boundary conditions for models. Chief disadvantage - Surface data, no matter how dense, are insufficient to initialize and verify models. Satellite: A. Temperatures: Most important in establishing large-scale mass field and monitoring changes for verification. High resolution data are of some use in verification. B. Moisture: Important in initialization and verification of models. 58 C. Cloud development: Monitor temporal and spatial distribution of convection for verification. May also be useful in initializing position of squall lines if method can be developed to incorporate divergence on this scale in the initial wind analysis. D. Cloud motions: May be of some use in enhancing wind analysis. E. Surface temperatures: Could be important in initialization and verification. Could be used in research mode to provide lower boundary condition for models. Chief advantage : Gives horizontally averaged data over whole domain on scale demanded by models. Chief disadvantage : Provides least information about most important variable, wind, and provides marginal quantitative information in cloudy areas. Aircraft: A. Locate and follow key features of mesoscale circulations such as jet cores, dry lines, baroclinic zones, and squall lines. B. Provide detailed horizontal analyses of mean horizontal and vertical fluxes at several levels in the vicinity of convection to aid in the evaluation and development of parameterization schemes. C. Refine horizontal analysis of temperature and winds over limited, but possibly crucial, regions, such as across jet stream. Chief advantage : Provides mobile system capable of measuring detailed horizontal variations at single level. Chief disadvantage : Cannot provide nearly simultaneous observations over a large area at many levels. Radar : A. Locate and follow convective patterns to aid in model verification. B. Possibly useful for short-range forecast (0-6h) as a tool for locating mesoscale convergence patterns. C. There is some doubt whether radar will be adequate to provide quantitative precipitation estimates, since reflectivity is not a unique indicator of total water content, but instead depends on the distribution of the water (or ice) particles. Chief advantage : Remote sensing and mapping of precipitation cells will be useful in the verification of the model's predictions of convection. Chief disadvantage : Radar data are of limited use in the meso-a and meso-3 scale models, which do not depend critically on the details of the precipitation characteristics of individual thunderstorms, 59 Lidar: Continuously monitor the depth of moist layer east of the dry line and the depth of the dust-laden deep boundary layer west of the dry line. Chief advantage : Provides time history at a point of the above parameters which are crucial to the development of severe weather, Chief disadvantage : Height of inversions not sufficient for initialization of models. 4. Use of mesoscale models in the planning and operation of SESAME field programs There are many types of numerical modeling experiments that can and should be done prior to SESAME with the purpose of defining the data densities and accuracies that are required for the initialization of the models. Observational simulation experiments are one type of experiment that would be useful. Repre- sentative severe weather cases could be run to generate realistic data sets. Given a particular experimental design, the appropriate data from this design could be retained (perhaps modified by the superposition of "errors") and the rest discarded. The remaining "measurements" would then be the basis for an attempted recovery of the full dynamic fields through the analysis and initializa- tion schemes. The goal, of course, would be to reproduce the control experiment. A second type of experiment that is needed involves the testing of the models to the sensitivity of the various parameters in the model. Diabatic models with moisture are often quite sensitive to small changes in the parameters; it is, therefore, important to identify these parameters and test the models under the range of parameters likely to be encountered. Because of the importance of the initialization phase of mesoscale models, testing of various initialization schemes should begin immediately. Research is needed to determine ways of utilizing non-conventional data such as satellite or aircraft data. Finally, there is an important role for the operational use of mesoscale models during the field experiment to aid in the prediction of the location and timing of severe weather. If models are to be used during the field program as forecasting tools, considerable testing of the models prior to the experiment will be necessary before model forecasts will be credible. 60 4. REPORT OF THE GRAVITY WAVE WORKSHOP October 15-16, 1975 GRAVITY-WAVE INTERACTIONS WITH SEVERE STORMS: WORKSHOP RECOMMENDATIONS FOR PROJECT SESAME F. Einaudi R. S. Lindzen W. R. Peltier W. H. Hooke, Workshop Chairman 61 1. Introduction As presently contemplated. Project SESAME (the Severe Environmental Storms and Mesoscale Experiment) will be an intensive study of severe storm systems, intended to provide the scientific foundation upon which improved severe-storm forecasts and warnings can be developed. The project will con- sist of two concentrated field measurement periods, using combined surface-, upper-air-, airborne-, and remote-sensor data, followed by periods of analysis and research. The project arises in response to an urgent need; severe-storm forecasts and warnings are currently inadequate. Using the present surface- and rawinsonde networks supplying data to NOAA's National Weather Service, it is possible to predict only the synoptic area or areas within which storm development is likely to occur. The actual development of the convective cloud systems, however, occurs on a much smaller scale within such synoptic areas — and exhibits a considerable degree of organization that is largely unpredictable from operational data processed following current practice. Thus, foremost among the forecasting improvements we hope will follow from SESAME is an improved ability to predict and monitor the development of the severe storms on the mesoscale. At present, more is suspected than actually known about just how meso- scale meteorological processes govern the triggering and development of severe storms. The draft Project Development Plan (PDP) for SESAME, written in June of 1974, catalogs the following mechanisms: 1) frontal uplift 2) the dry line 3) low- level jets 4) nonuniform boundary- layer heating 5) orography (producing lee waves) 6) propagating gravity waves. 62 It is the intent of the SESAME planners to investigate the role of each of these mechanisms in triggering severe weather. Clearly, the study of each of these processes places different demands on the SESAME experimental design. However, in some cases those demands remain relatively poorly de- fined. It is the goal of this report to consider in detail one mechanism for severe-storm organization on the mesoscale — propagating atmospheric gravity waves — and to delineate the theoretical and experimental work that should be carried out as part of SESAME in order to study their role in storm initiation and development. Interest in the interaction between gravity waves and severe storms stems from two sources. The first is the simple fact that wave- triggering of severe storms is observed to occur (the relevant observations are summarized briefly in the next section) . The second is that wave motions are inherently predictable. If the dispersive properties of the medium are known, then once a wave packet has been observed, it is possible to predict with some accuracy how it will evolve and move with the passage of time. This is true even if the wave source mechanisms are poorly understood; however, should it prove possible to identify the situations spawning the important waves, the prediction of severe-storm development can be carried one step further. Thus, to the extent that mesoscale triggering of severe storms can be traced to the action of gravity waves, it would appear that there might be considerable potential for predicting storm motion and development. What follows is a status report on what is known about gravity-wave interactions with severe storms, together with recommendations for future research of these interactions, both theoretical and experimental, and both for SESAME itself and for preliminary work. The report was prepared following a two- day workshop held October 15-16, 1975 and attended by the authors, who also received some input from D. K, Lilly, 63 2. Current Status of Understanding of Wave Storm Interaction Here we will review the primary observational data relating to the interaction between internal inertia- gravity waves and mesoscale storms, and will attempt to isolate those characteristics of the storm systems, associated waves, and background environment in which they exist which seem most germane to the theoretical understanding of their inter-relation. There are three principal questions which require resolution. Firstly, what are the dominant mechanisms responsible for mesoscale wave generation? Secondly, what are the properties of the background atmosphere through which the waves are observed to propagate? Finally, and this is perhaps the most difficult question, how is an incipient wave affected when the convergence field associated with it leads to the development of (perhaps severe) convective activity? Having gleaned what we can from the observa- tions concerning these questions, we will proceed in Section 3 to suggest a series of theoretical studies (analytic and numerical) which should help in clarifying the remaining issues by focussing clearly upon them. Evidence from arrays of surface mi crobaro graphs indicates that some severe convective storms act as sources of internal atmospheric waves. The evidence has been reviewed by Bowman and Bedard (1971). The signals they discuss have rather short periods (6- 300s) and are thus basically acoustic waves. The source mechanism responsible for their emission is thus likely an aerodynamic one as discussed for instance by Stein (1967) . These waves have such small amplitude and weak associated convergence that they are unlikely to be important dynamically. Evidence for the existence of longer period and horizontal wavelength waves associated with outbreaks of convective activity also exists. In the following we shall use the words "gravity waves" not only to indicate the kind of periodic oscillations recorded by a network of barographs as presented by Curry and Murty (1974) in which each pressure record reveals a number of oscilla- tions, but also to describe those events for which the barographic records or the synoptic scale data reveal the presence of a jump in the pressure and/or velocity diagrams as a function of time. While such a jump is in some cases preceded and/or followed by a number of oscillations much 64 smaller than the jump itself, these records do not resemble at all those of Curry and Murty, for example, and in fact they may reflect a different generation mechanism or perhaps a propagation phenomenon. Wagner (1962) has reported, on the basis of surface pressure observations, the propaga- tion of a wavelike disturbance over New England. The propagation speed was of the order of 25-45 ms . He speculates that the source of the wave may have been associated with the initiation of convective activity in the vicinity of a cold front over eastern Texas late on the preceding day. This association is somewhat obscure. The wave amplitude was relatively large (2-4 mb) , and the propagation path was characterized by the presence of a strong low level inversion. As we shall see, the existence of such a region of strong low level stability is common to most observed wave events. It apparently defines a duct along which the disturbance can propagate without significant energy loss due to vertical leakage. Wagner interpreted his observation in terms of Tepper's (1950) 'pressure jump line' hypothesis and obtained phase velocities from this model which were roughly in accord with measurement. Significant convective activity was associated with wave passage, and this activity was that characteristic of a squall line. This raises anew the original objection to Tepper's hypothesis. Convective processes associated with such disturbances may be sufficiently intense to dominate the associated surface pressure fluctuation. In such circumstances, it is difficult to comprehend how the disturbance may be described as a propagating internal wave. This objection to Tepper's hypothesis has never been reconciled satisfactorily. Implicit in this objection is the assumption that whenever the conver- gence field associated with an internal wave leads to the outbreak of convection, the wave is destroyed in the process. Observations of the relation between large scale waves and cumulus activity in the tropical atmosphere appears to provide evidence to the contrary. Reed and Recker (1971) have described an apparent feedback and control process in their study of synoptic scale tropical systems in which the wave and its attendant convection are mutually reinforcing. Whether an equivalent process operates on the mesoscale in mid- latitudes is an open question. 65 Ferguson (1967) has described a further later winter instance of gravity wave propagation over the Great Lakes region. This case was similar in most respects to the earlier studies of Brunk (1949) and Pothecary (1954). He concluded that the observed wave had probably been triggered by an outbreak of convective activity further to the southwest. Additional convective activity was associated with the passage of the wave. Ferguson interpreted his observation in terms of a simple set of model equations which had been derived much earlier by Goldie (1925) . Bosart and Cussens (1973) point out that these model equations are erroneous since they predict rising pressure associated with strong wave induced ageostrophic flow. In Ferguson's observation, the wave period was 45 min to 2 hr 15 min, with the period and thus the wavelength increasing over the path (the observed propagation speed was constant and equal to 13 ms ), As in Wagner's observation a strong low level inversion was prominent throughout the region. Ferguson notes that in the apparent region of wave generation the cold front was accelerating and according to Tepper (1950) , this could also serve as a mechanism for excitation of the wave. Bosart and Cussens (1973) have described another case of a gravity wave accompanying cyclogenesis in the eastern U.S. The study is notable for the large amplitude (y 7 mb) of the disturbance. The observed propa- gation speed (corrected for advection) was again on the order of 13 ms The vertical sounding is characterized by a low level inversion as before. Tepper* s hypothesis was again used to calculate a disturbance propagation speed, and again the speed agreed quite closely with that deduced from the model. The direction of propagation was perpendicular to the orienta- tion of the cold front (as in Ferguson, 1967) , and the ageostrophic wind maximum was coincident with the pressure minimum. The basis for the interpretation of such observations in terms of Tepper' s hypothesis is essentially the barometric data alone. The main reservation concerning the validity of this hypothesis has been mentioned previously. In spite of this reservation, the simple wave guide model appears to predict the propagation speeds correctly. It could be, however, that the agreement is fortuitous. 66 Eom (1975) has recently described a case of gravity wave occurrence in the midwest. Again the wave appeared to emanate from a region of active frontogenesis to the south but followed a path which was parallel to and in the same direction as the upper level flow. The disturbance path lay somewhat to the east of the 300 mb jet axis in the cold sector. Along this path the soundings were again characterized by a strong low level in- version. The observed propagation speed was '^ 50 ms (not corrected for advection), the period 3-4 hrs and the maximum pressure fluctuation 14.6 mb. The wave length was thus "^j 500-600 km, much longer than those observed previously. The surface wind maximum was in phase with the pressure minimum. In spite of its long horizontal wavelength, the disturbance was confined laterally to a channel "^ 200 km wide. Eom interpreted her results in terms of the linear eigenmodes of a simple two layer waveguide and found an internal gravity mode with a phase speed which approximately fit the observed speed of propagation of the disturbance. Although cloudiness, wind and pressure variations were well correlated (strong wind- low pressure- decreased cloudiness; weak wind-high pressure-increased cloudiness), there was no close association with convective activity observed. No postulate was made as to the source of the wave. This is in marked contrast to the observation in the case study described by Uccellini (1975), in which a wave of similar wavelength and period was shown to be responsible (apparently) for the initiation of severe convective storms along its propagation path. In several of the previously mentioned studies (Wagner, Ferguson, Bosart and Cussens) , such activity also occurred, but no cause and effect relation could be discerned. The evidence for such a relation in Uccellini 's study, on the other hand, is rather convincing. As in Bom's study, the wave period observed was '^^ 3 h, the wavelength '\' 500 km, and the propagation speed 'V' 50 ms . Again, a strong low level inversion is apparent in the soundings. In this case the disturbance appeared to propagate along the surface of a developing cold front and appeared to both reinforce previously existing regions of convective activity and to trigger new outbreaks wherever the 67 background state was sufficiently near instability. Satisfactory agreement was found between the observed wave phase speed and Horn's simple two layer waveguide model. Thus the wave characteristics did not seem to be dependent upon the convection it produced. Uccellini remarks in passing upon the possible source of the wave systems which he observed. Although one of the disturbances appeared to originate in a region in which organized con- vection did exist, others apparently had no such clear cut association. The existence of the strong jet stream and developing cyclone was noted. Using weather radar data, Marks (1975) observed a mesoscale precipi- tation pattern, associated with a New England coastal front, which was characterized by three easily discernible bands moving at about 18 ms These bands were approximately 90 km apart and about 30 km wide. Similar results have been obtained for 5 other events. Again the vertical structure revealed a strong low level inversion. Above the inversion there was a layer of stable warm moist air followed by a relatively thin convectively unstable layer. Finally, above the convectively unstable layer, there was a well mixed layer. Marks invokes the mechanism proposed by Kreitzberg and Brown (1970), Harrold (1973) and Browning et al. (1973), which attributes the lifting necessary to cause the mixed layer to develop, to the broad synoptic scale vertical velocity that occurs as a result of the baroclinic process within a cyclone. Marks recognizes, however, that this type of lifting, while explaining the release of convective instability, does not account for the occurrence of precipitation areas in bands. He does not pursue this point further, and in fact his analysis does not include information concerning the wave dynamics which might be revealed by microbarograph records in the area. Einaudi and Lalas (1975) have developed a rigorous linear theory of the propagation of gravity waves and of the development of gravity wave- like instabilities in a stratified atmosphere which is close to saturation over some height interval. They include a single analysis in their paper of the vertical mode structure to be expected of a wave propagating through an atmosphere such as that corresponding to Uccellini 's observations. Be- cause the region above the low level inversion is near 100% hiomidity, it 68 appears that wave induced condensation in this region can lead to a self- ducting of the wave such that the eigenmode has large amplitude in the near surface layer but rapidly decreasing amplitude through the region of high relative humidity. Their earlier papers (Einaudi and Lalas, 1973; Lalas and Einaudi, 1973, 1974) describe in considerable detail other effects on the stability and propagation of waves in wet atmospheres with shear. It should be pointed out, however, that their methods are restricted to the consideration of circumstances in which the effect of latent heat release is essentially a second order effect. It is not clear that such methods may be generalized, but they certainly warrant further study. Although the source of such waves remains an open question, it would appear that it is at the very least highly plausible that such waves may themselves trigger the onset of convective activity. The strong association between severe convective storms and propagating internal gravity waves has also been pointed out in studies of summer and winter mesoscale systems over western Japan. Matsumoto et al. (1967a, b), Matsiomoto and Akiyama (1969) and Matsumoto and Tsuneoka (1969) contend that the pulsating tendency of such storms is due to their organization by an internal gravity wave. The periods of such pulsation have been found to lie in the range 2-3 hrs and the dominant wavelengths associated with them to be on the order of 100-200 km. In these studies no attempt has been made to speculate upon the origin of the wave. Further references can be found in the papers by Matsumoto and Ninomiya (1965, 1969, 1971). Given the existence of the low level inversion revealed in most such studies, it is easy to see that an internal wave once excited will be able to propagate freely. However, the two previously mentioned enigmas remain. How are the waves excited? How is the wave able to maintain its coherence or almost maintain its coherence when the convergence field associated with it leads to the development of severe convection? One set of observations which may shed light upon the three stage interaction process, storm -> wave ->■ storm, are those pertaining to the arc cloud line. The most frequently observed cause of the generation of new convection cells appears to be the outflow from old cells. Under 69 some as yet unclear circumstances the outflow from an old cell, or a wave generated by this outflow, can extend several hundred kilometers from the parent cell and trigger new storms. Usually such new storms appear along an arc in the warm sector ahead of the cold front and appear to propagate at right angles to the front itself. Zipser (1969) appears to have been the first to describe this phenomenon from satellite observations of tropical cloud clusters. Purdom (1973) described the development of new storms at the intersection of two outflow arcs or at the intersection of an outflow arc and a frontal surface. Tepper (1950) had originally proposed such circumstances as being likely locations for tomadogenesis — if one interprets the outflow arc as an example of one of Tepper 's 'pressure jump lines'. Erickson and Whitney (1973) have recently shown a high resolution satellite photograph in which an apparent arc cloud line appears as simply the first of a large number of cylindrical ly spreading cloud arcs with a wavelength separating them on the order of 10 km. Donnegani (1975), and Peltier and Donnegani (1976) have interpreted this observation in terms of a linear wave packet launched by the parent storm. With the dominant wavelength in the wave packet determined by the lateral scale of the original storm, the group velocity of the packet was found to be in accord with the observed propagation speed of the disturbance (^ 28 ms ) . Raymond (1975) has developed a linear theory for the prediction of the propagation speed of severe storms, also based upon the concept that this speed is to be interpreted as the group velocity of some equivalent wave packet. Although this theory appears also to predict correct propaga- tion velocities for squall lines, split pairs, multicell storms, etc., there are several objections which can be raised to it. Firstly, accepting the particular parameterization which Raymond adopts for the convection in terms of the field variables describing the wave, he solves only an incomplete version of the resulting eigenvalue problem for the group velocity C„ valid in the limit that the dominant waventunber k ->■ (i.e., the infinite wavelength limit) . The group velocity he determines thus bears no relation to that for a fastest growing mode of the system, and the physics 70 is thus rather arbitrary. Secondly, there is the question of the validity of the particular parameterization of convection which he adopts. This question is no less pressing in respect to mesoscale systems than it is for large scale tropical systems. It is an unresolved question which lies at the root of understanding the interaction between waves and severe storms . Raymond's model is actually very similar to the wave-CISK model proposed by Lindzen (1974) as an explanation of tropical cloud clusters. It differs somewhat in the particular parameterization of the convection in terms of the wave variables but is equivalent in most other respects. In regard to the proper parameterization of convection, we would be remiss in failing to mention the recent diagnostic studies of Ogura and Cho (1973) on the determination of cumulus cloud populations from observed large scale variables in the tropical atmosphere. This theory has recently been applied by Lewis (1975) in a study of the cloud population within a prefrontal squall line (similar to that described by Ferguson (1967) with some success. Further case studies of this type and extensions of them are necessary if we are ever to be able to resolve the question of wave cumulus interaction satisfactorily. To what extent can we consider a squall line to be a propagating wave and how do the properties of such a wave depend upon the convection associated with it? Are all prefrontal squall lines generated in the manner proposed by Tepper (1950) or by some closely related process, or is some other mechanism operative? In summary of the above brief review of current understanding of the problem of wave storm interaction, the following points seem important. (1) Although the wavelike disturbances observed cover a fairly broad band of frequencies and wavelengths, there appear to be at least two identifiable classes of wavelike dis- turbance. The first class consists of those disturbances which we will call prefrontal. These are invariable seen in the warm sector preceding a developing or existing front 71 and include arc cloud lines and squall lines, both of which are oriented more or less parallel to the frontal surface. Such disturbances usually have spatial scales less than 100 km. The second class consists of much longer wavelength disturbances which propagate more or less parallel to the frontal surface in the direction of the upper level flow, often parallel to the axis of a strong 300 mb jet stream. The spatial scales of these disturbances are normally greater than 100 km but certainly much less than synoptic scale. (2) The mean sounding for both of these disturbance types appears to be characterized by the presence of a strong low level inversion. (3) The origin of such disturbances is unknown. They may be produced in response to organized convection along the frontal surface, as a result of geostrophic adjustment in the frontogenesis process itself, or by some as yet unidentified instability of the frontal flow. These appear to be the most likely candidates, but others come readily to mind. (4) The nature of the feedback of convection generated by the wave onto the wave itself is also poorly understood. The elucidation of this process will require further diagnostic studies. 3. Recommended Theoretical Studies There are three distinct areas in which much further theoretical study is required. These concern the generation of waves, their propagation in low level ducts, and their interaction with convective activity. Both analytical and numerical studies will be required if meaningful models are to be constructed. 72 Concerning the question of wave generation, at least four good candidates have been previously identified; organized convective activity along the front, the geostrophic adjustment process, the frontal instability and atmospheric shears. The first two candidates are most likely to produce signals which might account for the observation of prefrontal disturbances while the third seems most closely associated with the ob- servation of disturbances travelling in the along- the-front direction. In discussion of wave generation by existing cumulus cells, in first order theory, one simply replaces the convection by an effective heat and momentum source and calculates the wave output to the external environ- ment. Such studies elucidate the connection between source time and space scales and those of the emitted wave packet and provide crude estimates of the effective source strengths required to produce observed pressure and velocity variations. The main question involves the proper parameteriza- tion of the source; is it to be considered primarily as a region of transient heating, or is it best taken as a source of momentum? A complete treatment of this problem would optimally entail a series of detailed numerical ex- periments with a full three dimensional cloud model and an investigation of the wave producing capability of model clouds. Unfortunately, cloud models themselves are in such an early state of development that such an experiment would be rather premature. Simplified representations of the process are required. The geostrophic adjustment mechanism, as reviewed for instance by Blumen (1972), is one which has long been recognized as a fundamental source of gravity wave motions. Regions of strong frontogenesis are, of course, regions which are highly non-geostrophic, and they can therefore be expected to act as efficient wave sources. Unfortunately, the calculation of the radiation pattern to be expected from a 'frontal antenna' is in no sense a trivial problem. However, the subject deserves study. Perhaps the nonlinear cross front geostrophy solutions of Hoskins and Bretherton (1972) may be usefully employed in the investigation of this problem. There is also some hope that numerical models of frontogenesis, if they are of sufficiently high resolution, may contribute to the understanding of this problem. 73 The observations of Uccellini (1975) discussed above are particularly suggestive of long wavelength gravity waves propagating along the frontal surface. It has long been recognized (see review by Kuo, 1973) that the sloping boundary between two air masses is a particularly favorable location for the development of short wavelength waves due to dynamical instability. Perhaps these previous investigations should be looked at anew from the point of view of the more recently available data to see whether these data are compatible with such an interpretation. The frontal region is so complex in its dynamical structure that it is hard to imagine perturbations which it is not potentially capable of amplifying. Shear layers in general and jet streams in particular, both tropo- spheric and low level ones, have been known to be sources of gravity waves. It appears important to attempt, wherever possible, to correlate speed, amplitude and, more importantly, direction of propagation of the disturbance to the velocity structure and direction of the jets aloft. The presence of the low level inversion in the previously mentioned observational studies seems to be one of the most common denominators amongst them. This region can clearly serve as a wave guide for the propagation of wave energy once this energy has been trapped within it. Unless there exists a critical level in the overlying near adiabatic region, such trapping could not be complete. There is some evidence in the observations of Eom (1975) and Uccellini (1975) that a critical layer may have been present over much of the path, and this could account for the maintenance of wave amplitude during propagation. Detailed analyses of the response of such stratified systems to various forcings and studies of the structure of their normal modes of oscillation would be extremely useful. Although analytical methods are probably best suited for such studies, it is also possible to simulate such propagation effects using fully nonlinear numerical models, particularly when one focuses upon the propagation of a horizontally periodic disturbance, for then periodic boundary conditions may be employed in the model, and reflection from the up and downstream ends of the domain will not introduce spurious effects. 74 Such numerical studies are particularly desirable insofar as assessment of the interaction between waves and cumulus convection is concerned, although further analytical work along the line suggested by Einaudi and Lalas (1975) may prove fruitful. Although such studies describe correctly the form the disturbance takes for times sufficiently short after the initiation of condensation by the wave, the later stages of evolution involve nonlinear processes not included in the theory. It would be useful to study the further evolution of the disturbance from a numerical point of view. There exist wet anelastic models which should prove useful in such analyses. The models could be initialized using the linear theoretical solutions for the field variables, and their tendencies are employed to study system development. The main difficulty in carrying forward such an experiment concerns the ability to properly model the cloud produced by the wave field. Since cloud models are still in an infant state of development because of the huge computer storage required to correctly account for microphysical processes, it may be some time before a realistic simulation of wave-cumulus interaction is completed. However, this should not be allowed to impede the construction of simplified models of the basic interaction, A further problem which should prove amenable to numerical modelling concerns the generation of the arc cloud. This phenomenon has apparently not been simulated to date. Since it is an example of cloud-wave interaction in some (as yet unclear) sense, in which an apparently high degree of symmetry exists, it should prove amenable to numerical description. 4. Recommended Experimental Studies We recommend that six experimental studies be carried out both as part of SESAME itself and as preliminaries to it. These include: I. The continued analysis of past NSSL data. II. The continued analysis of available data on rainbands in Massachusetts. III. Preliminary deployment of microbarographs and laser wind- convergence sensors in the SESAME area. 75 IV. The SESAME mesoscale y experiment. V. The SESAME mesoscale 3 experiment. VI. The SESAME final experiment. Before we get down to the specifics of these experimental studies, however, some preliminary remarks are in order. To begin with, we note that the research areas in which further experimental study is required match those of theoretical interest. Recall that these concern the wave generation and source mechanisms, the nature of the wave propagation, and the wave inter- action with convective activity. Of these, experimental observation and study of both (a) the wave generation process itself and (b) wave-storm interactions are the most demanding. The theoretically attractive wave sources include organized convective activity along fronts, geostrophic adjustment, frontal instability, and atmospheric shears. The sources of wave motions observed within the SESAME region may often lie outside of it themselves. For this reason alone it will be most difficult to establish the relative importance of certain of the above-mentioned source mechanisms using SESAME data. This is particularly true of the geostrophic adjustment process. Wave generation by existing cumulus should be somewhat more amenable to experimental study. Experimental verification of this process involves estimation of the heat and momentum inputs from the cumulus into the atmosphere, as well as the spatial (both vertical and horizontal) distribution and time history of these inputs, and specification of the re- sulting wave field for comparison. The latter problem is the more familiar, and will be treated below, under propagation. The former, however, is a problem of considerable complexity, and the precise observational re- quirements associated with it are not clear at present. In what follows, however, we shall take the view that the other requirements on cumulus observation associated with SESAME are so stringent as to require the best observation possible, and that as a result this data set will be suitable for treating that aspect of severe storm physics encompassing wave- storm interaction. We shall focus our attention instead on the 76 dual problems of satisfactorily specifying the gravity-wave state over the SESAME region as well as the dispersive characteristics of the medium. Wave generation resulting from instability of frontal surfaces and from atmospheric temperature and wind structure generally will be even easier to treat. Here the observational demands are essentially the same as those imposed on studies of wave propagation. Experimental requirements for the study and analysis of gravity-wave propagation are quite familiar. These include specification of the wave state and the properties of the background medium. The wave state is most conveniently specified by the use of surface networks or arrays of in situ sensors (such as mi crobaro graphs, anemometers, thermometers, humidity sensors, rain gauges, etc.) for observation of wave-associated fluctuations in surface atmospheric fields. The space- time correlation properties of these fluctuations are then examined, first to separate wave fluctuations, which do exhibit correlation properties over the array, from turbulence, which does not, and second to establish the observed wave spectra and dispersion. The surface array data can be supplemented with data from balloons and from remote- sensing systems, which indicate vertical structure in the wave-associated perturbations in meteorological fields. The same networks and instrument systems also provide data on the back- ground fields of temperature, wind, and moisture; these suffice to describe the dispersive character of the medium. With these preliminaries, we are ready to discuss the six experimental studies recommended for SESAME, I. Analysis of past NSSL data. This work has already begun (Barnes and Lilly, 1975); what is re- commended here is the continuation of this work with several specific goals in mind. The Barnes and Lilly work is an analysis of spatial covariance of atmospheric fields over the NSSL area — the site of the proposed SESAME experiment. It would seem to be useful and important to extend this to a full space-time covariance analysis. This would have several advantages. It would enable a clear separation between fluctuations of meteorological fields over the array that were wavelike in character and V those that were not. It would build up a climatology of the wave state over the SESAME area, as well as indicating the relative degree to which the wave motions were revealed in the humidity, wind, temperature, pressure, and rainfall records. This in turn would help us to determine which meteorological parameters can be observed optimally for SESAME wave studies, what instriiment sensitivities are required, and what array deployment is most desirable (separation distance, orientation, etc.). The work would have immediate application to the optimization of NCAR's Portable Automated Mesonet (PAM) for use in SESAME (the electronic processing has been determined but final choices for the meteorological instrumentation itself have not been made), and for optimization of the METRAC sounding system. The space-time correlation work is currently underway within the Wave Propagation Laboratory of NOAA, which has developed computer software from the space-time analysis of array data (Young and Hoyle, 1975) . It should be continued throughout calendar year 1976. The Geoacoustics Research Program Area feels this work is of high priority but is currently under- funded. It would be desirable to find funding in the amount of $15 K for this analysis, to cover salaries and computer costs. II. Analysis of Massachusetts rain band data. This data set comprises weather radar coverage of the Boston area together with some supporting meteorological data. It is of interest primarily as a basis for comparison with NSSL studies, since the Massachusetts data cover storms of lesser severity, and since the geo- graphical location and the meteorological processes at work are somewhat different in nature. Such data sets, emphasizing storms of lesser severity, may yield more readily to theoretical understanding, guiding our first steps toward a theory for wave-severe storm interaction. This work too is already underway, and will require no SESAME funding. III. WPL preliminary experiments. NOAA's Wave Propagation Laboratory (WPL) should deploy microbarographs and laser wind- convergence sensors within the SESAME area with several objectives in mind. The work will enable WPL to gain experimence concerning 78 optimal deployment of the mi crobaro graphs for studies of waves of interest to SESAME. The waves of interest occur on a broad range of scales, with correspondingly great demands upon the array instrumentation (sensitivities, timing accuracies, etc.). The group must also find ways to overcome thermal contamination of the pressure data due to instrument heating and cooling during the course of the day. This has proved to be a serious problem in earlier joint NSSL-WPL work using an inferior instrument; ideas for coping with the temperature extremes (different instrumentation, instrument burial, etc.) need to be checked out. Finally, the research provides an opportunity for comparison between the pressure data and the laser wind convergence instrumentation. The former presumably measures the height- integrated convergence in the spatial -frequency range of interest, while the latter measures the surface convergence. In addition, knowledge of the wave spectra together with data on wind and temperature profiles permit a calculation of the vertical variation of wave-associated convergence, so that the pressure- data and surface convergence data may be checked for consistency. It is of interest to examine the correlation between the two. Data comparison will also permit checks of the impedance relation for deducing wave properties such as wave phase speed. Plans are underway for such comparisons in the spring of 1976, with operation coinciding with the NSSL spring experiment. Two mi crobaro graph arrays will be deployed; one roughly on the scale of the laser wind- convergence triangle, the other on a larger scale. Again, SESAME funding the order of $15 K will be required for this work. IV. SESAME mesoscale y experiment. The first-year SESAME experiment will focus on the mesoscale y processes We recommend that gravity-wave work during this experimental period focus on four aspects of wave- cloud interaction. The first involves study of cloud arc lines and wave motions in clouds on the scale reported by Erickson and Whitney (1973). In addition to the appropriate ground-based observations, it would seem desirable to provide high temporal-resolution satellite scanning of the SESAME area during such periods, to provide data not only on the spatial structure of such clouds but their temporal evolution. It 79 is our understanding that this would be feasible except during periods when NOAA satellites are exercising surveillance of tropical storms. The mesoscale y work should also examine the generation of waves by shear instability in the SESAME area, investigating the probability of such events in the spring and fall, the role played by moisture, if any, and nature of the wave spectra, looking for evidence of both evanescent and internal sequences as predicted theoretically. Experimental studies should also focus on wave transport of motion and energy from the convective planetary boundary layer through the overlying inversion to the free atmo- sphere above. Finally, experimental study should focus on wave breakups of elevated inversions inhibiting storm development. Details of the wave stage over the SESAME meso y area can be supplied by surface arrays of microbarographs. However, study of wave dynamics in elevated inversions will require the use of the FM-CW radar, preferably with Doppler capability for measuring wave-associated motion fields in the inversion layer. V. SESAME mesoscale 3 experiment. By contrast, the wave studies during the SESAME mesoscale 3 experiment should focus on the very large-scale waves of the type discussed by Uccellini. Here the array deployment should be an order of magnitude larger in scale (many tens of kilometers in spacing) but the instruments need be less sensitive. Emphasis should be placed on careful measurements of atmospheric temperature and wind structure to determine the nature of the wave energy trapping, if any. Again, satellite coverage of the SESAME area with high temporal resolution will be essential to the success of the project. Extensive use will be made of METRAC soundings or their equivalent. VI. SESAME final experiment. It is difficult to be specific about what should be done during this final experimental phase to study wave-storm interactions, since so much depends upon the outcome of the preliminary experiments. However, in general, emphasis should be placed on scale interactions between wave motions of 3 and Y mesoscale, and on clarification of questions which will inevitably be raised by the earlier experimetal work. 80 References Barnes, S. L., and D. K. Lilly, 1975: Covariance analysis of severe storm environments. Ninth Conf. Severe Storms, 301-306. Blumen, W. , 1972: Geostrophic adjustment. Rev, Geophys. Space Phys.y 10^, 485-528. Bosart, L. F., and J. P. Cussen, 1973: Gravity wave phenomena accompanying east coast cyclogenesis. Mon. Wea. Rev.y 101 , 446-454. Bowman, H. S., and A. J. Bedard, 1971: Observations of infrasound and subsonic disturbances related to severe weather. Geophys. J. R. astv. SoQ.y 26, 215-242. Browning, K. A., M. E. Hardman, T. W. Harrold, and C. W. Pardoe, 1973: The structure of rainbands within a mid- latitude depression. Quart. J. Roy. Meteor. Soc.j 99, 215-231. Brunk, 1. W. , 1949: The pressure pulsation of 11 April 1944. J. Meteor., 6, 181-187. Curry, M. J., and R. C. Murty, 1974: Thunderstorm-generated gravity waves. J. Atmos. Soi., 31^, 1402-1408. Donnegani, J. E. , 1975: Acoustic gravity wave propagation from a severe convective storm. M.Sc. thesis. University of Toronto, Dept. of Phys. , Jtine, 1975. Einaudi, F., and D. P. Lalas, 1973: The propagation of acoustic gravity waves in a moist atmosphere. J. Atmos. Soi.j 30, 365-376. Einaudi, F., and D. P. Lalas, 1975: Wave-induced instabilities in an atmosphere near saturation. J. Atmos. Sd., 32^, 536-547. Eom, J. K. , 1975: Analysis of the internal gravity wave occurrence of 19 April 1970 in the Midwest. Mon. Wea. Rev., 103, 217-226. Erickson, C. 0., and L. F. Whitney, 1973: Picture of the month. Gravity waves following severe thunderstorms. Mon. Wea. Rev., 101 , 708-711. Ferguson, H. L. , 1967: Mathematical and synoptic aspects of a small-scale wave disturbance over the lower great lakes area. J. Appl. Meteor. , 6, 523-529. Goldie, A.H.R., 1925: Waves at an approximately horizontal surface of discontinuity in the atmosphere. Quart. J. Roy. Meteor. Soo.j 51^, 239-246. 81 Harrold, T. W., 1973: Mechanisms influencing the distribution of precipi- tation within baroclinic disturbances. Quart. J. Roy. Meteov. Soc^ 99, 232-257. Hoskins, B. J., and F. Bretherton, 1972: Atmospheric frontogenesis models: mathematical formulation and solution. J. 'Atmos. Soi.j 29, 11-37. Kreitzberg, C. W. , and H. A. Brown, 1970: Mesoscale weather system within an occlusion. J. Appl. Meteor. ^ £, 417-432. Kuo, H. L., 1973: Dynamics of quasi-geos trophic flows and instability theory. Adv. Appl. Meoh.^ 1_3, 247-330. Lalas, D. P., and F. Einaudi, 1973: On the stability of a moist atmosphere in the presence of a background wind. J. Atmos. Soi.^ 30^, 795-800. Lalas, D. P., and F. Einaudi, 1974: On the correct use of the wet adiabatic lapse rate in stability criteria of a saturated atmosphere. J. Appl. Meteor., 13, 318-324. Lewis, J. M. , 1975: Test of the Ogura-Cho model on a prefrontal squall line case. Mon. Wea. Rev., 103 , 764-778. Lindzen, R. S. , 1974; Wave-CISK in the tropics. J. Atmos. Soi., 51 , 156-179. Marks, F. D. , 1975: A study of the mesoscale precipitation patterns associated with the New England coastal front. M.Sc. thesis, Massachusetts Institute of Technology, Dept. of Meteorology, June, 1975. Matsumoto, S. , and K. Ninomiya, 1965: Mesoscale disturbance observed in the vicinity of a cold vortex center. Papers in Meteor, and Geophys., 16^, 9-22. Matsumoto, S. , K. Ninomiya, and T. Akiyama, 1967a: A synoptic and dynamic study on the three dimensional structure of mesoscale disturbances observed in the vicinity of a cold vortex center. J. Meteor. Soa. Japan, 45, 64-81. , 1967b: Cumulus activities in relation to the mesoscale convergence field. J. Meteor. Soo. Japan, 45, 292-304. 82 Matsumoto, S. , and T. Akiyama, 1969: Some characteristic features of the heavy rainfalls observed over the Western Japan on July 9, 1967. Part I: mesoscale structure and short period pulsation. J". Ueteov. Soc. Japans 47, 255-266. Matsiomoto, S., and Y. Tsuneoka, 1969: . Part II: displacement and life cycle of mesoscale rainfall cells. J. Meteor. Soa. Jccpan^ 47, 267-278. Matsumoto, S. , and K. Ninomiya, 1969: On the role of convective momentum exchange in the mesoscale gravity wave. J. Meteor. Soa. Japan, 47, 75-85. , 1971: On the mesoscale and medium-scale structure of a cold front and the relevant vertical circulation. J. Meteor. Soo. Japan, A9_, 648-662. Ogura, Y., and H.-R. Cho, 1973: Diagnostic determination of cumulus cloud populations from observed large scale variables. J. Atmos. Sai., 30, 1276-1286. Peltier, W. R. , and J. E. Donnegani , 1976: Case study of a squall line trigger mechanism. To be submitted for publication. Pothecary, I.J.W. , 1954: Short -period variations in surface pressure and wind. Quart. J. Roy. Meteor. Soc, %Q_, 395-401. Purdom, J.F.W., 1973: Picture of the month. Meso highs and satellite imagery. Mon. ^ea. Rev., 101 , 180-181. Raymond, D. J., 1975: A model for predicting the movement of continuously propagating convective storms, J. Atmos. Sci., 3^, 1308-1317. Reed, R. J., and E. E. Recker, 1971: Structure and properties of synoptic- scale wave disturbances in the equatorial western pacific. J. Atmos. Soi., 28, 1117-1133. Stein, R. F. , 1967: Generation of acoustic and gravity waves by turbulence in an isothermal stratified atmosphere. Solar Phys., 2^, 385-432. Tepper, M. , 1950: A proposed mechanism of squall lines: the pressure jiimp line. J. Meteor., 7_, 21-29. Uccellini, L. W. , 1975: A case study of apparent gravity wave initiation of severe convective storms. Mon. Wea. Rev., 103 , 497-513. 83 Wagner, A. J., 1962: Gravity waves over New England, April 12, 1961. Mon. Wea. Rev.^ 90, 431-436. Young, J. M. , and W. A. Hoyle, 1975: Computer programs for multidimensional spectra array processing, NOAA Tech. Rept. ERL 345-WPL 43. Zipser, E. J., 1969: The role of organized unsaturated convective downdrafts in the structure and rapid decay of an equatorial dis- turbance. J. Appl. Meteor, y 8, 799-814. 84 5.1. REPORT OF WORKING GROUP (1) ON THE DEVELOPMENT AND UTILIZATION OF THUNDERSTORM SCALE OBSERVING TECHNIQUES SESAME STORM SCALE WORKSHOP Will iam Shenk H. Michael Mogil Kenneth Crawford Scott Williams L. Jay Miller Robert Serafin John McCarthy Griff Morgan John Marwitz R. Craig Goff James Purdom Ken Wilk Ed Pearl Ed Zipser N. E. LaSeur 85 Working Group (1) devoted much of its time to discussions of various components of a thunderstorm-scale observing system. There was some uncertainty as to the appropriate dimensions of "thunderstorm-scale" because of the inherent differences associated with individual storms, super-cells and organized meso-scale systems. In the course of discussions, all scales from what qualifies as gamma through beta to alpha were touched upon. In general, the Working Group (1) concurs with the proposed step-wise approach to the ultimate implementation of a full-scale SESAME field effort. They endorsed the 1978 Boundary Layer Pilot Experiment and the 1979 Regional Scale Experiment as planned, but there was also a feeling that SESAME planning should exploit other programs such as those planned by NSSL, NCAR (NHRE), CAP, etc. Any opportunity to acquire relevant data and/or test instruments should be exploited to the fullest extent possible. Also provision should be made for the rapid and complete analysis of data collected in order that results can be of benefit to the ultimate SESAME Field Program. A general philosophical point which was expressed by several Working Group members was that SESAME should not be designed to depend upon untested, unproven instrument systems for necessary data. Unless proposed instrument systems have been shown to be capable of acquiring necessary data, or such performance can be proven in the 1978 and 1979 pre-SESAME experiments, they should not be deployed in the SESAME field programs. Some specific examples of this potential problem are addressed below, based upon Working Group (1) members' assessments of the present status and capabilities of various system components in light of SESAME data requirements. In general. Working Group (1) endorses the nested grid approach to SESAME network designs as well as the proposed location and timing. Working Group (1) deliberately avoided the question of a proposed detailed design of a SESAME observing network. There were at least two reasons for this deliberate omission: (1) network design studies to optimize the final system should be carried out, first, and (2) results of further testing of some proposed system components could well have important impact 86 on the final design. Extensive discussion by Working Group (1) of various SESAME observing system components gradually led to grouping these in three categories: 1: mandatory system components which must be available with proven capabilities in order for SESAME to be viable; 11: desirable system components which appear highly probable of proven capability to provide valuable data; and (111): ancillary or auxiliary system components which are either uncertain as to proven capabilities or will provide useful but not critically important data, or both. In the following summary of discussions on various system components, these categories are used. A general characteristic of the observing system which should be pursued is to make sufficient data available in real-time to allow appropriate decisions on such questions as aircraft deployment, activation of special system components, etc., in order that optimum data collection will result. Summary of Conclusions and Recommendations on SESAME Storm-scale Observing System Components A. Radar I. MANDATORY - an array of several (more than 3, less than 10) ground- based Doppler radars deployed in an appropriate array. Comment - At least nine Doppler radars exist and more a're likely to be available by the time of SESAME. Results already obtained from arrays of two and three radars scanning the same volume are highly encouraging; this system should be able to provide the data required for describing at least the kinematics of storms. Processing the vast quantities of data from such systems, especially real-time processing, presents potential problems that need attention. A more detailed discussion of Doppler radar, prepared by a sub-group of Working Group (1) is included in Appendix B. II. DESIRABLE - surveillance radar for longer range detection and tracking could be very useful in SESAME operations. FMCW radar has capabilities (as outlined in 1974 PDP) that could contri- bute valuable data to SESAME. 87 111. ANCILLARY - Airborne Doppler radar, if developed and available in time, should be given consideration because of its mobility. B. Satellite Imagery I. MANDATORY - imagery from geo-synchronous meterological satellite(s) both visible and IR with maximum resolution in space and time are considered necessary to SESAME. The techniques of combining successive images into movie loops should be of special value. II. DESIRABLE - supplemental imagery and atmospheric sounding plus surface remote sensing from orbiting satellites such as TIROS-14 DMSP, NOAA, etc., can provide much useful data for SESAME. Appendix C contains more detailed discussions of satellite inputs to SESAME. C. Balloon-borne sonde systems 1. MANDATORY - a combination of rawinsonde and METRAC sondes is needed for SESAME. Testing remains to be done on the METRAC system, failure of which could result in removing that system for this category. Problems of baseline check procedures humidity accurancy and wetting effects need attention. Appendix D contains additional discussion on this system component. D. Surface Network 1. MANDATORY - a network of surface observing stations of the NSSL type augmented which the NCAR RAM system is required for SESAME. The latter system is subject to further testing, but the former has a long history. E. Aircraft 1. MANDATORY - Several research aircraft such as those of NOAA (P-3, C-130), NCAR (Queenair, Electra, Sabreliner) and NASA (C-990, B-57) are required in SESAME. Supplementation of such as these with leased twin- engine light aircraft should be given consideration. Direct penetration of severe convective elements may be too hazardous. Much previous experience with research aircraft should be drawn upon in designing utilization in SESAME. Appendix E contains some supplementary information on the capabiliti of the NASA B-57 efforts to collect similar information on other aircraft should be undertaken. 88 11. DESIRABLE - armored and highly stressed aircraft suitably instru- mented for direct penetration of severe convective elements. F. Optical Indirect Sensors 11. DESIRABLE - ground-based laser anemometers deployed to measure integrated path wind components. HI. ground-based LIDAR for cloud boundary detection; airborne LIDAR for similar purposes; LIDAR for aerosol measurements. G. Acoustic Indirect Sensors 11. DESIRABLE - acoustic sounding of boundary layer structure. H. Micro-wave Indirect Sensors II. DESIRABLE - satellite microwave sounding of atmospheric structure and ground-based microwave sounders. I. Photogrammetry 1. MANDATORY - airborne and ground-based time-lapse cinematography. Comment: although much of these data are primarily of Qualitative value, a considerable amount of quantitative results can be derived in certain instances. Simply as a visual record of events this type of data is justified. 11 or 111. DESIRABLE or ANCILLARY - balloon-borne time-lapse photo- graphy. J. Miscellaneous 1. MANDATORY - (if intensive gravity wave investigation included) - an appropriate array of micro-barographs. 111. ANCILLARY - measurement of sferics and other electrical parameters; measurement of aerosols and various nuclei. 89 APPENDIX A Attendees at Working Group (1) di 1 . William Shenk 2. H. Michael Mogil 3. Kenneth Crawford 4. Scott Williams 5. L. Jay Miller 6. Robert Serafin 7. John McCarthy 8. Griff Morgan 9. John Marwitz 10. R. Craig Goff 11 . James Purdom 12. Ken Wilk 13. Ed Pearl 14. Ed Zipser 15. N.E. LaSeur scussions: NASA/GSFC NOAA/NWS University of Oklahoma NOAA/CEDDA NOAA/WPL NCAR/FOF University of Oklahoma Illinois State Water Survey University of Wyoming NOAA/NSSL NOAA/NESS NOAA/NSSL University of Chicago NCAR NOAA/NHEML 90 THUNDERSTORM SCALE WORKSHOP THE ROLE OF DOPPLER RADAR IN SESAME APPENDIX B Sub Group on Doppler Robert Serafin Radar L. J. Miller S. Barnes J. Golden C. Murino N. LaSeur P. Ray 91 1. Introduction In his opening remarks Doug Lilly asked how Doppler radar may be used to serve the research objectives of SESAME. In his executive summary * Doug has indicated that the Doppler radars are to be used for the measure- ments of dynamics in clear air using chaff as a tracer and for studies of the dynamics in storms from their developing through their decaying stages. We will attempt to outline here the capabilities of multiple Doppler systems, studies under way which may expand these capabilities in time for SESAME, and we will make some suggestions for preparation for the field programs in 1978 and 1980. Doppler radars used in pairs or sets of three or more can measure the three-dimensional structure of air motion in precipitation, with the horizontal fields being quite accurately determined and the vertical motion estimates less accurately determined. The radars may also be used to measure the Doppler spectrum and its variance; the latter being related to the turbulent motion and shear at scales of the order of the radar sampling volume and smaller. Thirdly, the reflectivity factor is measured, which is related to precipitation intensity. Variables derived from the flow fields include horizontal convergence, vorticity, turbulence and structure of measured or derived variables. There exist relatively few meteorological radars yhich are capable of participating in a multi-radar net within SESAME. Among the most prominent are the two NSSL 10 cm Doppler radars, 2 WPL 3 cm systems, the 2 NCAR 5 cm Dopplers, the University of Chicago and Illinois State Water Survey 10 cm CHILL system and the 5 and 3 cm Dopplers of the University of Miami. * A preliminary document produced for the purpose of introducing the working group of SESAME planning progress - Ed. 92 While these systems do not all use common processing techniques, they are all capable of making the fundamental measurements necessary for determination of wind field data and may be deployed in conjunction with radars of another variety or design, i.e., it is possible to make a multi-Doppler network using virtually any number and combination of the 9 radars listed above although the selection of specific sub-sets is likely to result in greater data compatibility. In all cases, however, the data processing is likely to involve both hardware and software techniques. These matters will be discussed in more detail below. By measuring the storm dynamical characteristics with reasonably high resolution, the Dopplers will provide information related to the structure and important scales of the flow fields, and the interactions between precipitation growth and dynamics. In the 1978 experiment the clear air Doppler studies, by virtue of their "snapshot" capability, should be extremely useful in comparing the various techniques deployed for measurements of convergence on the 3 scale. Because the Doppler measurements are essentially contiguous in space, the 1978 results will be most valuable as inputs to experimental design efforts aimed at the 1980 multi-scale experiment. 2. Performance Expectations 2.1 Storm Environment We shall try now to provide some guidelines as to the state- of-the art of multiple Doppler radar utilization. We differentiate between the rather turbulent non-homogeneous storm environment and the more calm boundary layer environment. The estimates given below repre- sent "intelligent guesses" but we must point out that much of the analytical work comparing the performance of various configurations of multi-Doppler systems is still under way and little has been published concerning optimum objective analysis of the data set. Following then, are our performance estimates for the Storm Environment for one triple Doppler system with approximate spacing of 50 km. 93 Single Radar Surveillance Area - 150 km radius Areal coverage for high resolution 50 km radius winds (7500 km^) Spatial resolution < 1 km Horizontal wind accuracy 0.3 to 0.5 m sec Horizontal divergence measurements > -„-4 -1 - 10 sec and vertical vorticity Vertical wind accuracy 5 m sec >Aw> 1-2 m sec Note that in the storm environment the single Doppler radar coverage may be over a circle of radius greater than 150 km provided that the radar pulse repetition frequencies are sufficiently low to avoid range ambigui- ties. The areal coverage for detailed wind field analysis is substantially less. Actually, this coverage also will be greater than 50 km except that at long ranges, the radar beams will degrade the spatial resolution to values greater than 1 km. We should also note that the accuracy of the horizontal wind measurements will depend upon the objective analysis technique employed and the characteristics of the Doppler spectrum. These estimates have been based upon the assumption that each Doppler velocity estimate on a 1 km grid is obtained with a 3-dimensional filtering technique whereby of the order of 10 independent samples have been averaged. We believe this assumption is reasonable. 2.2 Boundary Layer Investigations VThen the pre-storm planetary boundary layer can be filled 2 (over an area of =^1000 km ) with chaff, the Doppler radars can be used to determine the boundary layer kinematic structure. Included are the mean flow features at scales of - 100 to 500 m and bulk properties such as eddy dissipation rate at smaller scales. This lack of small scale resolution is a consequence of the spatial filtering of the radial velocities by the radar beam illumination function. Momentum transport processes are readily determined from the Doppler radar measurements. However, other flux quantities (particularly turbulent fluxes such as moisture and mass) are less readily determined 94 because the necessary measurements for moisture and mass are presently not possible at the same scales as the radar wind measurements. Mean moisture and mass flux convergence can be quantified but remain somewhat restricted by these same measurement limitations. The simultaneous use of aircraft and other instruments in the 1978 3-sc^le pilot experiment may remove these limitations and certainly the combined instrument set should provide the information necessary to determine how moisture con- vergence can be satisfactorily measured on the 3 scale. Because the boundary layer experiments are likely to cover a smaller area, more is known about the structure, and the chaff fall speeds are known, the measurements made can be expected to be more accurate than in the storm environment. Following are our estimates of radar system per- formance in the boundary layer-chaff mode. 2 Areal coverage 1000-2000 km (determined by chaff coverage) Spatial resolution < 500 m Horizontal wind accuracy 0.05 - 0,2 m sec Divergence > 5 x 10 sec Vertical velocity 1-2 m sec 2.3 Dual vs Triple Doppler Considerations Several groups are currently investigating the improvements realized by adding a third or more Doppler radars. ^^Thile we cannot provide fully comprehensive comparisons it is possible to make the following statements. • A triple Doppler system can measure the horizontal fields directly. • The quality of the horizontal wind estimate in a triple Doppler system is virtually independent of height above the ground. • The triple Doppler coverage tends to be circularly symmetric around the radar array. • None of the above is true of a dual Doppler system. The triple Doppler system will provide better performance over 95 the total volume scanned. However, it should be noted that, although a triple Doppler system can in principal measure the vertical component directly, the statistical uncertainty of these estimates is very large for low antenna elevation angles resulting in unreliable measurements over much of the scanned volume. Thus, we must still use computations of horizontal divergence to compute vertical velocities. With a triple Doppler system, the integrations can begin at the top of the storm working downward, a significant advantage since ground clutter effects and earth curvature effects are minimized. 2.4 Current Multiple Doppler Radar Investigations Among the active groups in this area are the University of Miami, NSSL, the University of Chicago, ISWS, NHRE, WPL and NCAR's Field Observing Facility. In August 1975 Lhermitte (2 radars) and NCAR's FOF (1 radar) collected triple Doppler radar data within the framework of FACE. Lhermitte is currently constructing a system for off-line computation and display of the three-dimensional winds. Lhermitte and his students are also making theoretical comparisons of dual and triple Doppler systems. In the spring of 1976 NSSL and the University of Chicago-ISWS team will conduct a tri-Doppler experiment in Norman, Bohne and Srivastava in Chicago are also making theoretical evaluations of multi-Doppler system performance. At NSSL Ray and Stephens at Florida State have developed objective techniques for processing dual and triple Doppler radar data. They also allow for the incorporation of data from other sources, such as additional radars, aircraft, balloon soundings, numerical models. In the summer of 1976 a 4 or 5 Doppler radar array will be deployed at NHRE using WPL and NCAR radars. The participants will be the NHRE team, WPL and NCAR's FOF. Software for analysis of this data is now being developed. It is clear that substantial efforts are under way and that by 1978 we can expect with confidence to provide adequate performance criteria for optimum deplojonent of a set of Doppler radars for SESAME. 96 3, Data Processing This is an area which cannot be ignored. The wide bandwidths of radar data demand that efficient processing techniques be employed. Virtually all of the groups engaged in Doppler radar research use or plan to install real-time mean velocity estimation techniques. Most prominent is the pulse pair or auto-covariance technique which can be implemented at modest cost and can simultaneously compute mean velocities and spectrum variances in a large number of contiguous range gates (of the order of 10 ) in an interval commensurate with the scanning beam's target dwell time which may typically be of the order of 100 millisec or less. It is also satisfactory to record digitized raw data on digital tape for later processing. This technique generally requires slower antenna scan speeds but is advantageous when ground clutter signals are of the order of the meteorological signals, i.e. when the signal to clutter ratio is less than 10. A third technique utilizes wide band video recording to record the raw analog video signal for later digitizing and processing. This technique is very attractive since it is simple and few real-time decisions need be made as to how the data are to be processed, This technique, still under development at the University of Chicago and NCAR, may suffer from dynamic range limitations. Whatever the method used to obtain data in real-time, the post processing of a number of storms represents a task of major proportion. Efficient software and/or hardware processing techniques must be made available before the field experiment, and a sufficient number of people (we estimate about 5) must be dedicated to specific tasks associated with the processing. V7e are hopeful that the current efforts mentioned earlier will point the way to an efficient processing methodology for SESAME. A. Experimental Design Lilly's Executive Summary notes the experimental design effort, now beginning, by the University of Oklahoma, related to the placement of surface stations. Deployment of a Doppler radar array also deserves serious design considerations. As a minimum, the storm climatology 97 within the 3-scale array must be determined in order to establish statistically, the number of severe storms and tornadoes which can be expected to pass through the area covered by a multi-Doppler system. Storm velocities, tracks, and lifetimes must be established in order to effectively deploy the observing system. We urge that such studies be a part of SESAME' s experimental design effort. 5. Other Considerations A number of other factors deserve consideration in the planning for SESAME. The radar wavelength is very important in determining system performance. Shorter wavelengths permit more accurate measure- ment of the Doppler velocities for a given integration time. Thus other factors being equal an X-band radar can scan a storm more rapidly than C or S band radars. However, shorter wavelengths are subject to attenuation in precipitation, thus resulting in poor reflectivity estimation and loss of signal in heavy precipitation. Short wavelengths also reduce the unambiguous velocity inteirval, a problem usually re- solved by increasing the pulse repetition frequency and consequently reducing the unambiguous range. Pulse pair processing in contiguous range gates provides excellent spatial continuity of the Doppler velocity field which tends to mitigate the effects of velocity ambiguities. In the boundary layer studies, the shorter wavelengths will be attractive if a premium is placed on accuracy in order to measure and/or derive estimates of relatively small vertical motions (of the order of 1 m sec ) It is also worthwhile to note that multiple wavelength and/or polariza- tion observations may offer information related to the nature of the precipitation and a key to where hail occurs within the storm. Dual and triple wavelength work now underway within NHRE can be expected to be of value to SESAME in 1980. Operational strategies must also be carefully planned. Radar scans and methodologies must be centrally controlled and synchronized. Radio and telephone communications should be provided at each radar site with all radars sharing a common radio frequency for communi cations. In order to minimize mutual microwave interference it is desirable that 98 radars within a given band separate their transmitted frequencies. Redundancy, where possible, should be incorporated into the system. Experience indicates that commerical power is unreliable in the severe storm environment, thus the probability of failure is greatest just when the data are most interesting, If possible, back-up diesel generators should be employed to ensure operations under the worst conditions. Data redundancy should also be considered. The University of Chicago and Illinois Institute of Technology and NCAR are currently developing low cost techniques for recording the raw coherent video on video tape and later analysis. The success of this work will pro- vide a low cost, insurance policy so that valuable data are not lost due to processing system failure. 6. Concluding Remarks and Recommendations We feel that the use of Doppler radars in the severe storm en- vironment is a proven technique as evidenced by the work of Lhermitte, WPL and NSSL and that triple Doppler experiments nov; underway will prove to further advance our observing capabilities in time for SESAME. Data processing and analysis techniques will continue to improve and, as a consequence, the Doppler radar component of SESAME will prove to be of immense value and is absolutely essential to the program. We do have several recommendations although all of them are really subsets of the first. « A SESAME radar planning group should be established to carefully consider details of the 1978 and 1980 experiments and to work closely with other design groups within SESAME. • The severe storm climatology must be accurately assessed. • The radars which are deployed should be capable of scanning as rapidly as is consistent with desired accuracy. This implies that they be fitted with pulse pair processors or equivalents. • Real-time coordination requires that quantitative real-time displays be available at several of the radars. If possible the reflectivity, Doppler velocity and spectrum variance 99 should be displayed simultaneously in order to effectively select the most interesting storm or storms within a squall line. Color display technology similar to that developed by NCAR or other displays now under evaluation by NSSL should be incorporated into the observing network. • New developments at WPL in processing adaptively spaced pulse pairs or by the addition of a high PRF capability will be valuable to SESAME if incorporated into some of the radars. These will remove the velocity ambiguity problems which are especially troublesome in the severe storm environment. • Some mobile system capability should be provided if possible. An airborne Doppler radar for high resolution studies of the tornado vortex region would be especially important for ob- serving storms which occur and/or have moved out of the region of ground based coverage. • Software development for processing and analyzing the radar data should begin now. If possible, at least one high level programmer should be dedicated to this task under direction of the radar planning group, • We do not yet know enough about thunderstorm structure and its temporal variability. These of course dictate scanning strategies and layer objective analysis techniques. We urge that, whenever multiple Doppler radars are deployed, the investigators take time to conduct sub-experiments whose specific purpose is to determine these structure functions. We also urge that research into optimum objective analysis of the data sets be continued. This is clearly not an exhaustive list but is a reasonable starting point, The authors will be pleased to work further to help with SESAME *s planning. 100 SATELLITE DATA AND SESAME APPENDIX C James F. W. Purdom Applications Group, NESS 101 I. Introduction Both the polar orbiting and geostationary spacecraft operated by NESS will play major roles in SESAME. Data from these satellites must play an integral part in the planning and execution of SESAME, as well as the research which will be done both during and after the field phases of SESAME. We will first look at the type data that will be available from the various satellites, and then look at some of the ways in which that data might enter into the various portions of SESAME. II. Polar Orbiting Satellites A. General The next generation polar orbiting satellite, TIROS-N, should be in operation during SESAME. This new system, scheduled to replace the current ITOS system in 1978, will be made up of two sun synchronous polar orbiters instead of the current one. Local sun times for the equator crossings of the two satellites are planned for 0730 descending - 1930 ascending for one spacecraft, and 0330 descending - 1530 ascending for the other spacecraft. With those orbits, one spacecraft will take data over the SESAME area (central Oklahoma) between 1000 and 1100 GMT in the morning and 2100 to 2200 GMT in the afternoon, while the other will be over the area in the early afternoon between 1400 and 1500 GMT and in the early morning between 0200 and 0300 GMT. The major advances expected over the present system are better and more temperature and moisture soundings, as well as a high resolution multispectral viewing capability. B. TIROS-N Vertical Sounder 1. Instrument Characteristics The TIROS-N Operational Vertical Sounder (TOVS) is expected to provide better data for deriving temperature and moisture profiles of the atmosphere than are currently available from the ITOS Vertical Temperature Profile Radiometers. This is due to the TOVS having added channels, improved resolution, and a limited microwave capability. Three separate and indepen- dent subsystems make up the TOVS: (a) the Basic Sounding Unit, (b) the Stratospheric Sounding Unit, and (c) the Microwave Sounding Unit. As a direct consequency of studies conducted using Nimbus-6 data, a decision 102 has been made to fly a modified HIRS (High resolution Infrared Radiometer Spectrometer) instrument on TIROS-N at the earliest feasible time, probably early 1979. Three channels are being added to the HIRS for this TIROS-N spacecraft. These additional channels should improve observations of total ozone concentration, tropospheric water vapor profiling, and surface temperature specifications. The 20-channel HIRS, called the Tropospheric Sounding Unit, will also contain a number of mechanical and optical improvements as compared with the Nimbus-6 HIRS. Characteristics of each sounding unit is given in Tables 1-4. Table 1 . Basic Sounding Unit Characteristics Parameter 14 channels, wave length \m Length of scan line Steps per scan Resolution at subpoint Resolution at scan end (1127 km) Distance between scan lines Value located at 3.70, 4.26, 9.71, 11.12, 13.33, 13.61, 13.99, 14.29, 14.49, 14.75, 14.95, 18.80, 23.15, 29.41 1127 km either side of satellite subpoint 56 total, 28 on either side of satellite sub- point Circle 21.8 km in diameter El ipse 73.2 km along scan line, and 37.3 km perpendicular to scan line 21 km at satellite subpoint decreasing to 4 km at end of scan line (1127 km) Table 2 . Stratospheric Sounding Unit Characteristics Parameter Value 3 channels centered at 14.97 ym Length of scan line Steps per scan Resolution at subpoint Resolution at scan end Peak emissivity received from 3 levels in in stratosphere: 15 mb, 5 mb and 1.5 mb 737 km either side of satellite subpoint 8 total, 4 on either side of satellite subpoint Circle 147 km in diameter El ipse 244 km along scan line, and 186 km perpendicular to scan line 103 Parameter Value Distance between scan lines 63 km at satellite subpoint decreasing to 26 km at end of scan line (737 km) Table 3. Microwave Sounding Unit Characteristics Parameter 4 channels in 5.5 mm oxygen band Length of scan line Steps per scan Resolution at subpoint Resolution at scan end Distance between scan lines Value Center frequencies of 50.3 GHz, 53.74 GHz, 54.96 GHz and 57.95 GHz 895.6 km either side of satellite subpoint 11 total, one in center and 5 on either side of satellite subpoint Circle 83.7 km in diameter Elippse 183.9 km along scan line, and 120.1 km perpendicular to scan line 85 km at satellite subpoint decreasing to 48 km at end of scan line (895.6 km) Table 4. Tropospheric Sounding Unit Characteristics Parameter 20 channels, wave length ym Other parameters Value Located at 0.69, 3.72, 4.00, 4.24, 4.40, 4.44, 4.52, 4.57, 6.71, 7.25, 8.'14, 9.71, 11.11, 13.35 13.66, 13.97, 14.25, 14.49, 14.73, 14.95 Other characteristics will be very similar to those given for the Basic Sounding Unit. 2. TOVS Data and SESAME TIROS-N TOVS data will provide valuable information for use in the planning, execution, and research portions of SESAME. In the planning phase, TOVS data should be especially useful in helping determine how to best arrange the additional radiosondes from one year's experiment to the next. Since only a limited number of special radiosonde stations will be available from year-to-year during SESAME, we will want to locate them as strategically as possible for each year's experiment. To aid in doing this, all available 104 sounding data from the 1979 regional scale experiment must be combined to determine how to use the TOVS data to effectively interpolate down to smaller scales between the existing radiosondes. Hopefully, this would then allow us to place the extra radiosondes much closer together for the first multiscale experiment. In a like manner, the same type analysis should be done between the first multiscale experiment and the second multiscale experiment. Being able to do this is important for both the SESAME program and the regional and mesoscale models of the future. TOVS data should have a variety of uses for many of the studies that will be undertaken both during and after SESAME. The data should be especially valuable in the building of regional and mesoscale models that will be under construction during SESAME, as well as those that will follow SESAME. Hope- fully, some of those models will be built to take advantage of the special characteristics of satellite sounding data; i.e., high spatial resolution, one accurately calibrated instrument taking all of the measurements, internal consistency between measurements, and very precise accuracies in the hori- zontal gradients of the radiances that are measured by the various channels. C. TIROS-N Multispectral Scanner 1. Instrument Characteristics The Advanced \lery High Resolution Radiometer (AVHRR) planned for use with TIROS-N is a multispectral scanner. The resolution of the data is 1 km at the satellite's subpoint, and gradually decreases along each scan line to 2.5 km at a distance of 1600 km from nadir. The instrument will take measurements in the following spectral intervals: (a) 0.55 - 0.9 ym, (b) 0.725 - 1.1 ym, (c) 10.5 - 11.5 ym, (d) 3.55 - 3.93 ym, and (e) a channel at 11.5 - 12.5 ym will be added to the system in early 1982. 2. TIROS-N Multispectral Data and SESAME The high resolution multispectral data taken by TIROS-N should provide useful information for both the research and real-time operational portions of SESAME. The channels were chosen primarily for hydrological reasons: to allow differentiation between clouds, snow and ice, and melting snow and ice. However, there is still a variety of information concerning convection and severe storms that can be gleaned from this multi -spectral imagery. In cloud-free areas as small as 1-2 km in diameter, we should be able to: (a) determine earth surface temperature; (b) delineate surface 105 wet areas from surface dry areas; (c) determine total moisture in an atmospheric column, still at l-2km resolution. Although observations will be taken only four times per day over the SESAME area, they will still be useful, especially when used in conjunction with GOES (Section III), for the location of dry lines, areas of deeper moisture and stronger moisture con- vergence, areas at the surface with a greater potential for rapid heating, and areas of stronger surface heating, just to name a few. The determination of total moisture will improve during the daytime with the addition of the 11.5 - 12.5 ym channel; this is because of contamination problems due to reflected solar radiation in the 3.55 - 3.95 ym channel during daytime. Much of the mesoscale information available on surface temperature and surface temperature gradients should be very useful to numerical modelers working on the mesoscale, and meteorologists interested in the boundary layer. Indeed, this measurement of surface temperature should be more representative than any number of probes making point measurements of surface temperature. The high resolution information on total moisture should be \/ery valuable for both numerical modelers and shortrange forecasters, especially when used in conjunction with the lower resolution TOVS and GOES VAS data (see Section III E). D. TIROS-N Data in Support of SESAME for Operations and Research Plans for what data products will be available routinely from TIROS-N data are still in the developmental stage at NESS. SESAME should arrange with NESS for research-oriented support with all data processed at the highest possible resolution in real-time. The mix of products from the TOVS, multispectral data and combinations of the two might include the following (1) surface temperature; (2) total moisture; (3) moisture between various levels of the atmosphere; (4) surface wet and dry area delineation, qualita- tive if possible; (5) imagery from the multispectral channels; and (6) sounding information from the NESS facility at the University of Wisconsin (see Section III F). SESAME must make sure that all TIROS-N data, sounding and multi-spectral, taken over the SESAME area of interest are archived in such a way as to be easily accessible by SESAME for follow-on research. TIROS-N data products, which are derived for real-time support of SESAME, will also be an important part of the SESAME data archive. 106 Ill Geostationary Satellites and SESAME A. General The great advantage of the geostationary satellite over the polar orbiting satellite is its ability to view the same area of the earth at frequent intervals. With the geostationary satellite's imagery, destructive weather systems from the synoptic scale down through the mesoscale may be observed virtually in real-time. In its unique way, the geostationary satellite is the only observing tool in SESAME that can simultaneously observe the various meteorological scales and their interactions, which is the very essence of the tornado problem. The heart of the present Geostationary Operational Environmental Satellite (GOES) is its Visible and Infrared Spin Scan Radiometer (VISSR) which takes high resolution imagery with a visible channel during daytime, and a thermal infrared channel both day and night. The resolution of the visible channel is 0.9 km, while the resolution of the 10.5 to 12.5 micron thermal infrared channel is 9 km. A follow-on system, GOES with VAS (VISSR Atmospheric Sounder), will allow for day and night imaging, as with the present GOES, as well as a multispectral capability and a sounding capability. The spacecraft is planned for launch in early 1980, and should provide wery valuable data for the SESAME multi scale experiments. B. Current Operational System Data are currently used in real-time to support the National Severe Storm Forecast Center and the National Hurricane Center in their unique missions. Weather Service Forecast Offices throughout the Nation also use GOES imagery in real-time to help them in their many faceted forecast problems. On normal weather days, GOES takes pictures at 30-minute intervals. On severe weather days, GOES' mode of operation is changed to provide imagery at more frequent intervals to field users. Depending on the intensity of the expected outbreak, the satellite takes images over the USA either once every 15 minutes or once every 7H minutes. The 7i^-minute interval imagery is reserved for days when the stronger type family tornado outbreaks are expected. The most rapid interval imagery that the WSFOs can accommodate is the 15-minute interval type, while the Kansas City Satellite Field Service Station supporting the National Severe Storm Forecast Center is capable of receiving imagery at y^s-minute intervals. Proper use of the 107 7^-minute interval imagery requires special interpretation equipment which is presently only available at the Satellite Field Service Station at Kansas City. C. Thunderstorm Generation Mechanisms Observed by Geostationary Satellite During the past few years, observations from geostationary satellites of severe storm areas have revealed many different mesoscale systems which exert strong influences on thunderstorm development. Even though we do not fully understand many of the mesoscale systems which we are observing, they are making every substantial contributions to our forecasting ability of severe thunderstorms. Some of our greatest insights into thunderstorm- environment coupling and tornadic thunderstorm development have come from 5-minute interval GOES imagery. Our experience with this type data is very limited, and as we observe more tornadic storms with very frequent interval imagery, we expect to find many more indications of triggering mechanisms which initiate and modulate severe convective storms. These observations should help better define many of the questions facing SESAME, and thus help sharpen the observational strategy which will be used during SESAME. Some of those phenomena are listed below. 1. Convective Line Intersections and Convective Mergers. In more dynamic regions, thunderstorms tend to form in lines, with these lines tending to lie in regions of low level moisture convergence. These lines are often detectable very early in their life with high resolution satellite imagery as lines of fair weather cumulus. Initial thunderstorm formation will often occur along those lines where they intersect with other convective areas or lines. Additionally, areas along an active squall line which are intersecting other convective lines are usually where the longest lived and most intense thunderstorms will be found. What do these various lines represent? What governs their location and orientation? Why does stronger activity develop where lines merge? 2. Cloud Arc Lines. The arc clouds observed by satellite appear to be the leading edge of the thunderstorm cold dome early in their life; yet many arcs appear to have sufficient energy to travel several hundred miles away from their origin. Many times an arc cloud passage is accompanied by severe winds, while other 108 times it is not. The leading edge of the arc represents a preferred place for new thunderstorm formation, especially where other convective lines merge into it. Where an arc intersects another major boundary, such as a front or another arc's boundary, is a favored spot for severe thunderstorm develop- ment and tornadoes. Which arcs have damaging winds and why? What determines how long an arc will last and how far it will move? How does the arc change character during its lifetime? Why is a favored place for severe and tornadic thunderstorms located where an arc intersects another major boundary? 3. Cyclonic Shear Boundaries and Thunderstorm Interaction. Using the 5-minute interval satellite imagery, we seem to have detected one major class of severe tornadic thunderstorm triggering mechanism. Using 1 km GOES imagery taken at 5-minute intervals, we are able to follow the motions of individual cumulus clouds. In tracking these clouds, we have located long and narrow lines of concentrated and well -organized cyclonic shear ahead of the severe squall line. Thunderstorms along the squall line which interact with this boundary almost always have intense tornado activity associated with them. What is the structure of these lines in depth? What is their origin? How concentrated is the cyclonic vorticity along them and across them? What, in their interaction with the severe thunderstorm, leads to tornado genesis? During the line's interaction with the severe thrunder- storm, what determines when the tornado will occur? 4. Thunderstorm Relative Flow. Using 5-minute interval 1 km imagery, we have been able to make cloud motion movies in which the thunderstorms in the squall line are held stationary and other clouds are observed to move relative to the thunderstorms. This type movie makes it easier to see what is causing the thunderstorm to develop, as well as determine how it is interacting with and affecting its environment. Using motion of this type, it appears that tornadic thunderstorms have a different relative inflow from the non-tornadic ones. In one case a very strong relative flow in the low level clouds occurred along the upwind side of the thunderstorm just prior to tornado development: was this acceleration in the low level flow due to the thunderstorm? Studies using this type of motion need to be continued. 109 6. Radiation Effects. Differential heating and cooling effects exert a strong influence on thunderstorm development and squall line evolution. How effective the sun's energy is depends on both the character of the earth's surface and the distribution of cloudiness. Satellites have observed that under certain conditions the distribution of cloudiness has a dominant role in determing where thunderstorms will form, as well as controls how thunderstorms will evolve along an existing squall line. Satellites of the future should be able to detect differences in surface characteristics, such as wet and dry areas, which should in turn affect boundary layer heating rates. 7. Anvil Characteristics. Nearly all violent thunderstorms have anvils which seem almost to explode, and turrets which rapidly protrude above the anvil and just as rapidly collapse. At times, a pause in anvil growth has been noted prior to a severe event occurring on the ground. We can detect all of these phenomena from satellite data; however, we do not know enough about their frequency or their connection with tornadoes, hail, and severe winds. 8. Importance of Cirrus. Many times cirrus appears to play an important role in determining the location of severe thunderstorms. In many cases, the severe thunderstorm will form along the edge of a cirrus band. We do not understand the mechanism by which the cirrus influences severe storm generation. Is the storm's formation due to differential heating between the cirrus overcast and clear skies, cirrus seeding, or is the cirrus only a reflection of a localized dynamic mechanism which triggers severe convection? 9. Role of the Subtropical Jetstream's Cirrus Shield. In a great number of cases, the southern most extent of a severe storm outbreak will be where the southern end of a squall line meets the northern edge of the cirrus shield associated with the subtropical Jetstream. Many times a wery strong tornado will be produced by the thunderstorm which is located at that intersection. Again, is this due to shading, seeding or dynamics? 10. Thunderstorm and Environment Interactions. Using the 5-minute interval 1 km resolution imagery, and taking advantage 110 of the large aerial coverage of the satellite pictures, thunderstorm and environment interactions are readily detectable. Among the most striking phenomena observed has been the modification of one thunderstorm's environ- ment by another thunderstorm. Cases have been observed when tornado-producing squall lines have abruptly ceased development upon coming into contact with air which has been stabilized by previous thunderstorm activity. In one case, that modified air was advected over 200 miles before coming into interaction with the tornadic squall line. 11. Small Pertubations within Large Vorticity Center. In some severe weather situations, small pertubations rotating around the major vorticity center seem to trigger tornado activity when they merge with the back edge of the squall line. D. Some Uses of GOES Data in SESAME 1 . Preplanning Phase As mentioned previously, observations from geostationary satellites have revealed many different mesoscale systems which exert strong influences on convective development. These observations are already helping to better define many of the questions facing SESAME, and should continue to do so up to and during the SESAME field program. Most of the operational weather forecasters throughout the Nation are aware of many of the benefits to be gained from high resolution GOES imagery in the mesoscale convective forecast problem: this is because they use the imagery daily in their forecast and are familiar with it. These benefits are not recognized by some of the more research-oriented meteorologists who are providing input to the planning of SESAME: this is due mainly to a lack of exposure to the capabilities offered by high resolution imagery. Because the technology with high resolution satellite imagery is expanding so rapidly, people in both operations and research who are using the data have a responsibility to both publish their results and present their findings at professional meetings. 2. Use of GOES Data during the Field Phases of SESAME Possibly the most important use of GOES data during SESAME will come from its operational uses during the field phases of the SESAME program. Here it will find two main uses: (1) nowcasting for the deployment of equipment and personnel to the proper areas for investigation--several hours lead-time 111 may be available, depending on which SESAME subprogram is to be investigated on a given day. One major use, which has already been done successfully in a joint NESS, NASA, and University of Chicago experiment, will be vectoring aircraft to areas along a squall line where a greater chance of tornadic activity exists; and (2) debriefing of field scientists after they have spent a day in the field taking observations--in doing this, the scientist making a particular observation will be able to better see how the various meteorological scales interacted to effect what he saw and measured, thus making his observa- tions much more meaningful. To insure that the above goals are met, SESAME must have the ability to receive, display, and interpret GOES imagery at the SESAME forecast location. On the more important field days, data will be needed at very frequent intervals around 3 minutes between pictures. Proper display equipment which will provide real-time animation of the rapid interval imagery for the forecaster is also necessary. Personnel trained in the interpretation of GOES imagery for severe storm forecasting are also needed to insure that maximum benefit is derived from this unique source of data. The animation sequences made on the display equipment must be saved, perhaps by a video tape system, for use in debriefing personnel after a day of field measurements and also for use in future research. In addition, SESAME should arrange with NESS for operational "research oriented" support which would include the following: hourly printouts of temperature and temperature change observed by GOES over the SESAME area, as well as winds, divergence and vorticity computed from rapid interval imagery cloud tracking. These would be needed in real-time as well as in the follow-on research portions of SESAME. 3. Use of GOES Data During the Research Portions of SESAME Careful planning during the field phases of SESAME will make it much easier to save the satellite date from GOES (and other satellites) for use in the follow-on research portions of SESAME. Much of the data which is received and used operationally during SESAME will be valuable for use in the follow-on research phase. The video tape of the animation sequences made after each day's operation would be yery valuable for anyone doing case studies, and for use in the general interpretation of many of the features previousely mentioned. Hourly winds and temperatures would also be important 112 for use by researchers. Satellite data received during the SESAME field program should also be stored on magnetic tape* so that it may be fully recovered at a future time. To recover the data from the magnetic tapes will require special types of equipment; however, this equipment would pro- bably be the same that was used in the operational portion of the SESAME field program. E. GOES with the VISSR Atmospheric Sounder (VAS) The GOES VAS instrument will be first flown on the GOES-D spacecraft in 1979. The first VAS is unique in that it is a NASA funded mission to be conducted on a GOES satellite operated by NOAA/NESS. The VAS is an extension of the current VISSR imaging capability and includes additional thermal bands for the determination of atmospheric temperature profiles by spectral selection in the CO2 absorption bands. Water vapor profiles are also determined from measurements in the H2O absorption bands. In addition, cloud heights are obtained from CO2 absorption band data, while the earth's surface temperature and low level moisture are measured using window measure- ments at 3.7 ym, 11 ym, and 12 \sn. The VAS will have three modes of operation (a) the VISSR Mode; (b) the Dwell Sounding Mode; and (c) the Multispectral Imaging Mode. It should be noted that mode (a) cannot operate simultaneously with modes (b) or (c). 1. VISSR Mode. The VISSR operating mode is the same as in the current VISSR system. As in the VISSR, the visible channel resolution is 0.9 km, however, the spatial resolution of the infrared channel has been improved to 6.9 km. 2. Dwell Sounding Mode. This is the sounding mode for the VAS experiment. The ultimate purpose of the VAS experiment is to evaluate the VAS instrument performance and demonstrate the sounding capabilities of the VAS prototype system. In the dwell sounding mode, up to 12 spectral regions covering from 14.7 vim through 3.7 ym can be sampled while the VISSR is dwelling on a single scan *Some systems presently ir. operation can store up to 2h hour's worth of data on a single magnetic tape. One such system is operated by NESS for NASA. Currently all data from one of the two GOES spacecraft are recorded and saved. Perhaps SESAME could come to some arrangement with NASA for the saving of GOES data during the SESAME field programs. 113 line. Typically, 100 spins per scan line are required to obtain the desired sounding data for that line. The above number of spins per line should be adequate to obtain soundings having a resolution of 30x30 km, and requires around 1 minute. Thus in a half-hour, a north to south swath of approximately 900 km in length can be sampled (the swath is from horizon to horizon). This time may be decreased by deleting some of the stratospheric channels. Both VISSR Mode data and Multispectral Mode data are unavailable when the satellite is operated in the Dwell Sounding Mode. 3. Multispectral Mode. This is the principal imaging mode for the VAS experiment. In this mode, the high resolution visible imagery is always available. In addition to the visible, one may receive any one of the following: (a) four of the IR channels, each with a resolution of 13.4 km; (b) two of the IR channels, each with a resolution of 6.7 km; (c) one IR channel at 6.7 km resolution and two IR channels at 13.4 km resolution. F. The GOES VAS Experiment and SESAME Although the GOES VAS experiment is still in the planning stage, there are several areas of the experiment which are of direct interest to SESAME. Initially, GOES VAS dwell sounding data and multispectral data will not be received for operational use, but will undergo an evaluation phase to determine how they might best enter into the operational system. There is a good chance that the first GOES with VAS will be placed in orbit as a back- up spacecraft for the operational GOES spacecraft. If this is the case, the GOES VAS will most probably be positioned at geostationary altitude between 90 and 95 degrees West longitude. By operating independent of the operational system, the spacecraft would be able to dwell sound or multispectrally image at wery rapid intervals; however, visible imagery resolution is reduced to 7 km since the data are not received at the NESS GOES Command Station. If the GOES VAS is operated as a part of the operational system, the research and development evaluation of the multispectral mode and the dwell sounding mode will be done at times when they do not interfere with the prime imaging mode. 114 Research into the VAS ground data processing will be performed at two locations: (1) NOAA/NESS and the University of Wisconsin at a VAS dedicated ingest and processing center proposed to be located at the University of Wisconsin, and (2) by NASA at the Goddard Space Flight Center. Some of the goals of the GOES VAS experiment to be performed by NESS relate directly to the goals of SESAME. At the NESS facility, man-machine interactive sounding techniques and systems required for mesoscale severe storm applications will be developed. During VAS experimental days all dwell sounding and multi- spectral data will be saved at the NESS facility, and TIROS-N TOVS data over the U.S.A. will be processed in a mesoscale array and archived. The NESS group will be working toward developing a real-time capability with the VAS data, thus information of importance to SESAME for daily operations may be available. Both SESAME and the GOES VAS experiment stand to benefit one another through a coordinated research effort. There are many ideas** about information which may be extractable from GOES VAS multispectral data and dwell sounding data: these need to be tested, and will depend heavily on verification data which would be available from SESAME. SESAME will benefit from the year-to-year knowledge gained from the GOES VAS experiment, and may be able to use some of the information in its daily operation. **Among the most promising theories to be checked out are: (1) the determina- tion of the divergent part of the wind (on scales as small as 50 km) from the time variation of the mass field using a second order wind equation. If this can be demonstrated, an extremely important source of data will be available for use by both operational forecasters and mesoscale numerical modelers; and (2) local changes in stability through changes in the moisture field and the vertical temperature field; again, this would be very important for both operational and research scientists. 115 POSSIBLE GEOSYNCHRONOUS SATELLITE SYSTEM CAPABILITIES WILLIAM SHENK NASA/GSFC Appendix CI 116 INTRODUCTION Rapid progress has been occurring and is expected to continue in the tech- nological capability to improve remote sensing from geosynchronous orbit. The improvement will primarily result from larger spacecraft carrying bigger telescopes, better spacecraft attitude control and determination of 3 axis systems, and the development of sizeable microwave antennas that can meet stringent design tolerances. Coupled with the expected space instrument and vehicle improvements are rapid advances in the development of interactive computer systems which will allow man to select which data need to be analyzed and to quickly extract the important meteorological parameters from it. GENERAL FUTURE CAPABILITIES THAT NEED TO BE DEVELOPED From the meteorological requirements and satellite system considerations some general needs emerge that should establish the trend of how best to exploit these-new technological developments. There are: (1) Improve spacecraft altitude control and determination systems such that unknown errors of only a few urad are left (1 urad is 36 m at the sub- satellite point) . (2) Increase-temporal resolutions so that areas of about 1000 x 1000 km can be viewed at about one minute intervals. (3) Increase spatial resolutions up to diffraction limits. (4) Extend spectral coverage into the microwave. (5) Reduce the noise in each spectral interval such that relative differences of a few tenths of a degree can be sensed at low scene tempera- tures. Correspondingly, the discriminating capability of reflectance measurements should be a few tenths of a percent. (6) Sharply increase the use of interactive computer systems that can serve as the main data analysis tool. Items 2, A, and 5 are substantially influenced by using a 3 axis stabilized system. Radiometric efficiency is considerably improved with a 3 axis system since only the area desired on the earth is scanned instead of a 360° sweep which is necessary from a spinning satellite. This improved efficiency plus larger optics will permit large spatial resolution increases up to practical diffraction limits. Stellar reference systems and improved momentum wheel packages should allow the achievement of the necessary attitude control/ determination accuracies. 117 SPECIFIC MISSION COMPARISONS Three new types of missions are approved or planned beyond the current SMS/ GOES series through GOES-C. The GOES D, E and F missions will carry the VISSR Atmospheric Sounder (VAS) which will provide the first vertical temperature and moisture profiles from geosynchronous orbit. Beyond VAS the Stormsat mission is planned which will improve many of the VAS visible and infrared sensing capabilities and microwave observations are being considered. Finally, the Synchronous Earth Observatory Satellite (SEOS) is expected to make further improvements beyond Stormsat, using larger optics, for visible and infrared measurements. The SEOS mission will be shared with earth resources objectives. The VAS improvements over VISSR will come from the addition of spectral inter- vals to obtain vertical temperature and moisture profiles. These channels will also provide better measurements of other quantities when some of them are used to implement multispectral techniques. In general, the improvements expected from the visible and infrared sensing portion of the Stormsat mission emanate from the use of a 3 axis stabilized system since the telescope currently proposed is a modified VISSR with the same 40 cm optics. The greatest improvement over the VAS is in atmospheric sounding. Another major Stormsat improvement will come if microwave sounding and imaging can be added. The SEOS mission is expected to have about a 150 cm optic on a 3 axis platform. While further gains in infrared sounding performance are achievable the greatest change from Stormsat will be higher resolution imaging since the diffraction limit is nearing completion and major changes are no longer possible. The detailed comparisons between these missions and the GOES VISSR (where appropriate) will be shown in a series of tables. The specific data shown for each mission is presented for comparison only. Significant alterations can be made to Stormsat and SEOS missions whereas the VAS sensor is nearing completion and major changes are no longer possible. While the tables can present many of the fundamental differences the accompanying text will describe details that are important and cannot easily be shown in the tables. Finally, the differences will be related to the requirements presented earlier. 118 TEMPERATURE AND MOISTURE PROFILES Table 1 shows the comparisons of the VAS, Stormsat, and SEOS systems for measuring these parameters. The microwave capability to measure the profiles in cloudy areas is separated from the clear to partly cloudy determinations that can be made from infrared techniques. All three sounding systems meet or exceed the minimum requirements limits for spatial and temporal resolution. The Stormsat and SEOS systems have maximum sounding spatial resolutions that are near the bottom of the requirements range and the temporal resolutions exceed the range needed for general mesoscale measurement. None of the systems meet the vertical level require- ments and the Stormsat and SEOS accuracy estimates are closer than VAS to what is needed although they are still +0.1-1.0°C short in absolute accuracy. Relative accuracies of <^1.0°C are expected from all of the systems. A complete on board optical system calibration for the Stormsat telescope is being studied. If it can be implemented this will give Stormsat a simpler (and probably better) calibration method than the VAS which should help to improve sounding accuracies. The Stormsat infrared capability will add the 4.3 ym CO2 sounding channels that have been proven by the Nimbus 6 HIRS to substantially benefit the temperature sounding. These channels along with the 15 ym CO2 spectral intervals, improve the temperature sounding accuracy and the vertical resolution. When the 15 ym channels alone are used high resolution imaging can be done simultaneously with the temporal resolution that is needed for cumulus cloud tracking over land (to obtain low level winds) and for the monitoring of cloud top changes associated with thunderstorms. The high temperature sounding frequency and spatial resolution will permit the close examination of the interactions of thunderstorms with the environmental air surrounding them where significant changes are expected in a few minutes. Clear column radiances need to be determined in partly cloudy situations as part of the temperature sounding retrieval process. The techniques for estimating these from polar orbiting sounders involve comparing pairs of instantaneous fields of view over a spatial array. The high temporal resolutions obtainable from Stormsat when the 15 ym channels are being employed in the high frequency sounding mode offer the possibility of 119 TABLE 1 COMPARISON OF GEOSYNCHRONOUS SATELLITE OBSERVING CAPABILITIES (750 X 750 Kin2 areas) PARAMETER Temperature Profiles (clear to partly oloudy conditions) 1. Maximum Horizontal Spatial Resolution 2. Vertical Resolution 3. Temporal Resolution 4 . Accuracy Temperature Profiles (cloudy areas) 1. Maximum Horizontal Spatial 1 . Resolution 2. Vertical Resolution 3. Temporal Resolution 4 . Accuracy Moisture Profiles (clear to partly cloudy conditions) !• Maximum Horizontal Spatial Resolution 2. Vertical Resolution 3. Temporal Resolution 4 . Accuracy Moisture Profiles (cloudy areas) 1. Maximum Horizontal Spatial Resolution 2. Vertical Resolution 3. Temporal Resolutoon 4. Accuracy VAS SATELLITE SYSTEM Stormsat 30 Km 13.5 Km 4-5 Km 2-3 Km (with 4.3 ym + 15 ym chan- nels) 4-5 Km (with 15 ym channels) 33 min. 20 min. (with 4.3 ym + 15 ym channels) 1.5 min (with 15 ym channels) +2.5«K +2-K c s 10 •H -H CO !0 -O V^ 04 It •a 0) o o u 0; -P ^ *J T3 Ui V4 C O '0 a It ^ o» c J5 O4 o c to o» c S 0) -H J3 •H O ■p J= -w ■p c j; -H o ■P X -H u o 3 0) .P o u CU Q) E W •H 4 C o -rt •H •a « I a 'O M « -H c o 4) 9 •a •o 4J V4 a U S wi X 4J -H 43 •P ;c Cfl 0) '0 u A c e 4i 8 M 4J (0 rH i-» 13 »J U) ■P 3 « 3 M B ^ "O -H rH 9 tp *» •H CO (0 n rH < 0< g u JJ jp CO CA > 10 M S z (0 B M > CO s§ a »4 -H a u •H *i M ^■^ ti « CQ (0 M M 0) :^ sa u V4 g to 9 H (0 a M < M »JfN U} u Ta- )J g 1 ll) M 10 •J < r-t CO (/) in ■p g CO r^ s n 3 D X > Z Q in « r- m X ^^ •d 3 g n 5" b 9 U 10 c Z IM OS ■0 CA w CO M > h 8 II in o ^ in e e in ^ in d d o o o o o> o in o II o o CT> O O H I • hi 3 ■P U ■P CA "O 3 O O c u fl» -H •H O O •H • i 01 « 0> E t-{ UJ IM 10 4J a •H 3 rO (U c M <0 lO U 5 (U E •H kH •H y S H X •H ■P 0. Oi X (0 <*^ ^.^ E -* a 1 3 z: u- '»-* ^-^ H < "■^ *ii^ rH • • • rH u 10 Ji u u • • 5 S 1 «M 9 ^ :^ > i § 5 5 s Vt « 9 *» "d 4J «) a A >l I H •0 • ^ 3 10 u u Q •H •H ■p 4J -H DiH rH IB •H « g A o» {^ lO Vl g- <0 rH 5 U IM r-t +) 9 ifl |S CO D> c g a X Vl U 4J CO ■p fl •H a. JS g H H •-• 126 Table 4 only contains information that is pertinent to cloud top height determinations. The expected accuracies for ice and water cloud top heights are educated estimates. If the Oxygen A band reflectance technique can be added to Stormsat (it is being studied) then ice cloud accuracies close to 0.5 km should be achieved. Precipitation The use of high frequency microwave for precipitation measurement is in the theoretical stage. Reflectance measurements and infrared threshold techniques have shown some promise, especially in tropical areas. It is likely that the combination of microwave and these other techniques will yield the best results since the microwave should be able to define the rain areas and the other methods applied where it is definitely raining. Table 5 shows the performance for the four systems. It is not possible to specify SEOS yet. It is mostly qualitive for the reason given above. The accuracy of precipitation estimates for the microwave is based on theoretical considerations . Hydrology There is the probability of satisfying a number of hydrological applications' with Stormsat. The attached memorandum indicates which are the most likely. The proposed sensor changes will help the meteorology objectives as well. Some of them have been discussed earlier. 127 s in O B O 6 rH S Q rH S o 1 O (Q 1 o n ■P 0) (d ■P Q> < u h » U M > 0) Id o « Id M rM M »^ •H Wl VI O a U4 «M U Q U^ «H u g M n 0) C -H n 0) c •H (A H M a B H Vt -H s ^ M c "* 0) in 2 43 u 0) in o # U) 0) «a o> u 'g' O o •H u "a* o o in o u X >4 C 1 o n CN U) C 1 o n cs ^; H ■P 4J -0 +J ^^ (3 '0 'a- 1 1 U) ^ Id -a 'a- 1 1 H flj •H M-l 0) Id N ■p (1) 1 N N o o ■P 0) 1 N N S CO > o vi *— ' X U Vl N X S 0,^ U Vl N X X r* }j H e Id o M > o VM U-l o o •H ^ M-l U-l o o C g S 4J o -H C (d CN (U C •<3' T CN Id '» 0) C T "a- fM fM.OI S W a ■p •H » fN Vl -H l-t (N Cm as U -H a\ H CM v|a u g o U} z X w (0 > s u P6 (>] H Ui 10 w (0 o „H^ vJ M o > K PS u 0) H n CO CO Eh u < 13 10 CO M lO U] H M »J o > > •-5 CM 1^ % >4 3 3 o 9 »4 9 * W in ^ •H It 1 8 1 o in D > Pm CO oi CO A O X z o o « m a r^ 1 •o o U « z >< 0) r-t VI 1 1 s U4 0) )4 en o a\ 6 u ;4 c 1 o 1 o o tT> •H n 0) o O 1^ V o b< c •o ■o C 1 «= J. o c Id o (d -O ■p « Id "d ■P o> z 3 JS u u o u o b: «) (0 Q> Id 0) Id w M > VI -H a 2 »4 c o -H 10 •H u c Id ■P 0) ■p ^^ Vl «M «M 0) c 55 u 1 1 4J O •a •H r-l O ■P 2 e o rH VI «M «M V c ^ -H Id r-l C Id o 4J •»< 4^ o in A e X> C +J ■P C *i 0) O 3 > Id 3 J* -H •0 N r-« o *i N rH u ■U •H O c •H ■H s q: Id o V4 (A w Q< M 09 u *» c O « 1 •H 0) o H •H S M c c U X V4 < w •H U u a • • « •H • o •H P V u • • (d • I I A. a i 4 I «M ! •8 Id • t4 Vi •tj «H Id c *M ■H \ n « 4 U 1 ■P u 4 « ^ ■-4 IM (t O -.4 u 13 0) O ^ -.^ ■p a •H IM u o o •0 >t u a 2 n a u o Id § ^ o a g « 128 Meteorological Information Extraction Capabilities The Information Extraction Division of the Applications Directorate is developing hardware and software systems to provide the capabilities required to support the Severe Storms Research Program (SSRP) through 1980. The systems developed for the SSRP will also serve as test beds for development of the technology and for sizing the system required to meet Stormsat and related meteorological information extraction needs through the 1980' s. Present meteorological information extraction capabilities include use of special software packages which execute on the Atmospheric and Oceanographic Information Processing System (AOIPS) and on a 360/91 computer. The AOIPS hardware configuration consists of two interactive image display and analysis terminals, a GE IMAGE 100 and a Hazeltine Image Analysis Terminal, interfaced to a PDP-11/70 computer. The AOIPS IMAGE 100 includes a PDP-11/45 computer and related peripherals. One unique AOIPS system feature will be the attachment of a single, NASA standard High Density Digital Tape Drive (HDDT) in September 1976. The HDDT will be used for storing SSRP related data. Figure 1 indicates the AOIPS hardware configuration and Table 1 lists HDDT characteristics. During the first half of CY 76, software is being developed on the AOIPS computer system to extract and analyze wind vector fields at several cloud height levels from SMS images. This software will allow a meteorolo- gist to select clouds and edit resulting wind vector fields interactively as the processing takes place. Analysis programs to calculate and produce plots, contours, listings and image overlays of wind fields, divergence, vorticity, stream functions, and related parameters are also being developed for interactive use. The IBM 360/91 computer will be interfaced to the AOIPS by a 4800 bit per second telephone line connection. This connection will allow the AOIPS to interactively submit work to and interrogate the 360/91 computer system. The 360/91 computer will be used to perform non-real time high bulk data processing operations and to perform large computational jobs including the operation of sophisticated meteorological models. 129 SSRP/VAS/Stormsat Plans During FY 77 and FY 78 efforts will be concentrated on developing the analytical capabilities and computer systems needed to meet the following information extraction requirements: - Computing atmospheric temperature and humidity profiles in preparation for SSRP/VAS and Stormsat/AASIR studies. - Correlating sounding profiles with wind vector fields, derived para- meters and conventional data from NOAA and other sources. - Developing severe storms forecasting statistical prediction algorithms. The AOIPS facility will be augmented by adding additional display, storage and computing power to allow simultaneous sounding extraction, multispectral analysis and wind vector processing operations in preparation for the VAS phase of the Severe Storms Research Program. Satellite data will be directly received from the GOES-D VAS and SMS VISSR preprocessing facilities for near real time data analysis and operational demonstration experiments. Recorded data will also be available from these facilities for data archive and longer term analysis functions. During the late 70 's, AOIPS will be used in limited, short duration operational demonstrations of SSRP/VAS project capabilities. This experience will provide a test bed for developing information extraction techniques needed for Stormsat. Phase B of the Stormsat program data processing study requirements will be met using the augmented AOIPS facility as a test bed for hardware and software system development. Stormsat data analysis functions will be simulated on AOIPS to fully satisfy mission objectives. A Stormsat data processing facility will be designed and specified to conclude the Phase B study effort. Applications Information Processing System (AIPS) Present Goddard computers including the IBM 360/91 are run as open shop, batch oriented systems and are unable to meet the near real time information extraction requirements of the Applications Directorate. The Applications Information Processing System (AIPS) will provide computational and data management support during the 1980 's for projects in the Weather and Climate, Earth and Ocean Physics, Earth Resources Survey, Environmental Quality and Communication/Navigation disciplines. The large 130 scale AIPS computer complex will be used to supplement existing computer systems at Goddard and to provide the near-real time, interactive, computa- tional and data base capabilities needed to support applications projects such as Stormsat, Nimbus-G, Landsat, EVAL, etc. AIPS will function as a large scale central data processor with the AOIPS, the Stormsat processor, and the Intralab processor connected to AIPS as peripheral processing terminals. Weather and climate data base models will be stored, updated, and accessed as required from AIPS to support information extraction functions being performed on the AOIPS, SSRP/Stormsat and other peripheral processors. The Stormsat processor will receive standardized AASIR data innuts from the Goddard Image Processing Facility via wideband communications interface during near-real time demonstration periods and via High Density Digital Tapes (HDDT's) during other periods of research and development with less stringent turnaround requirements. This processor will also have an input current conventional meteorological data from NOAA's planned Automated Field Operations System (AFOS). The processing of the AASIR sounder data and the statistical prediction* modeling will be performed on the AIPS processor. During real-time demonstrations, approximately 1 percent of the sounding radiances collected will be selected for inversion to temperature profiles. During research periods, however, some of the sounding data that had not been used in the real-time demonstrations will be examined in greater detail. Computer based statistical prediction models will be demonstrated using AIPS in a quasi-real-time environment. The statistical models will input: a) the most currently available analyses of temperature and wind fields, b) temperature profiles extracted from AASIR sounding data at a rate of about 25 profiles/minute, (approximately 15 sec of 360/91 time per min) , c) wind vector sets derived from the cloud tracking sequence at a rate of about 10 winds/minute, d) radiance analysis maps at specific mb temperature levels, and e) other pertinent data sets such as sea surface temperature analyses and radar reports as they become available. 131 In summary, the Stormsat dedicated processor in combination with the AIPS computer facility will be utilized to fully support severe storm forecasting objectives through the 1980' s. 132 FUTURE SATELLITE SYSTEMS VISSR Atmospheric Sounder By 1980 the VISSR Atmospheric Sounder (VAS) is scheduled for launch. The VAS will provide the first temperature and moisture profiles from geosynchronous orbit plus improve the measurements of cloud structure, winds from cloud motions, and surface temperature over what can be obtained from the GOES VISSR. Approximately three independent levels of information are possible for temperature and moisture soundings in the troposphere. The subsatellite-point horizontal spatial resolution ranges from 30 km in clear areas to near 100 km in regions with partial cloud cover up to 60%. An area that is 750 km in the north-south direction times the width of the earth can be scanned in about half an hour. The expected absolute tempera- ture profile accuracy is +2.5K with relative errors from a known point in time and space of +1 K. STORMSAT Currently, NASA is studying a new R&D mission called STORMSAT which could be launched as early as late 1981. Therefore, if the schedule is met, it would be available for the second SESAME multiscale experiment. A joint NASA/NOAA Working Group is in the process of making recommendations for STORMSAT and therefore, no specific comparisons can be made between it and the VAS. However, by going to a three axis stabilized concept the following general improvements over the VAS should be possible: (1) Higher horizontal resolution imagery - both visible and infrared, probably by about factor of 1-h to 2. (2) Simultaneous sounding and imaging. (3) More accurate temperature soundings and temperature and moisture soundings with higher spatial and temporal resolution. (4) Higher vertical resolution temperature soundings. (5) Temperature and moisture soundings in regions of non-precipitating clouds using microwave techniques. (6) Ability to distinguish between precipitating and non-precipitating clouds on a horizontal scale of about 40 km. A more detailed comparison of the VAS and STORMSAT systems will be con- tained in the final report of the NASA/NOAA STORMSAT Working Group. 133 RADIO/WIND SONDE REQUIREMENTS APPENDIX D Scott Williams, CEDDA 134 Requirements: a. Wind - rms vector error of h mps b. Temperature - rms error _ 0.5C c. Relative humidity - rms _ 5% rh d. Ability to obtain data in the first IflO (or as good as we can get) meters e. Real-time automated data reduction is essential. GMD-1 : The GMD-1 flatly fails requirement (a). It fails requirement (d) because the radar tracking unit cannot lock on to the target until the sonde is well above the horizon and away from ground clutter. I have to reverse what I said in the meeting this afternoon. In BOMEX, using modified GMD-2B radiosonde sets, we received met data (temperature and relative humidity) beginning 30 seconds before launch. Through a non- foolproof device, launch time was generally obtained to within h second. In GATE, we generally began receiving met data prior to launch and inserted a flag into the recorded met and NAVAID data at launch time. Requirement (e) can be met with the GMD-1 only if a high failure rate can be tolerated. The baroswitch gives trouble at launch and it also gives trouble in flight any time the signal is lost for more than a few seconds. METRAC : The METRAC sonde appears to be the only one on the horizon capable of meeting requirement (a). With a little care, it should also meet require- ments (d) and (e). The use of special shaped balloons, being considered for SESAME, should also allow winds to be measured over shorter time intervals than is the case when standard balloons are used. The absence of a baroswitch simplifies the computer programming required for data processing. Also, with the improved accuracy of height determination (which must be verified) of the METRAC system, pressures can be determined with great accuracy by hydrostatic computation. Requirements (b) and (c) face the same difficulties in any system. I recommend the procedures CEDDA has used in three major experiments (BOMEX, 135 IFYGL, and GATE): a. Have baseline measurements made at the factory. This removes the biggest source of error in field operations. Baseline resistances can be measured individually for thermistors. A simple and cheap additional step is to determine calibration curves for each manufacturing lot of thermistors. The combination should reduce temperature errors to a quarter degree or less. Carbon hygristors can have baseline resistances measured individually at the factory, but calibration can only be done by lot sampling. I recommend both. b. Use a mid-scale reference, in conjunction with a precision resistor, which allows sonde internal resistance to be computed for every data cycle. This adds significantly to data-reduction accuracy. Sonde internal resistance is not only seldom equal to the standard value, but it changes in flight. c. Use a clock commutator (we didn't have this for BOMEX) to ^ive a data cycle rate of once per second or faster. Unfortunately, the METRAC system is as yet untried. The 1977 field test should be devoted primarily to insuring that it will meet dependa- bility and accuracy requirements under field conditions. Some of the conditions which must be met are: a. The system must be operable and maintainable by trained but inexperienced crews such as we are apt to have for the multi-scale experiments. b. Ninety percent of sondes reaching the field should operate to balloon burst or experiment limits (whatever they might be, 100 mb?, 10 mb?). c. Frequency separation and control must be such as to preclude interference between sondes. Test at least a half-dozen sets in a pre- dominately up-, down-wind orientation. d. Ground equipment must be fully operable for 95% of scheduled releases. e. A significant number (20 or more) of soundings should be made with two or more sondes on the same train in order to test relative accuracy. The wind measuring equipment should also be tested by having 136 as many receivers as possible follow the same sonde on several soundings. Any failure of the above tests that cannot be corrected by fairly simple and straightforward modification should be sufficient grounds for postponing SESAME indefinitely. Some additional operating requirements are: a. Receiver distribution at each site to be such as to meet wind accuracy requirements up to experiment height limits, possibly with some relaxation of requirements above 500 millibars. b. Sonde-signal reception to begin 30 seconds before launch to: 1 . Insure good signal 2. Preclude unmeasured travel of sonde 3. Avoid loss of boundary- layer data 4. Allow sensor lag corrections to be made from the beginning of the sounding. c. Adequate provisions to be made for monitoring of data quality during flight. d. Backup recording of all sonde signals to: 1. Allow post-processing of data by more sophisticated methods 2. Allow data recovery in case of computer failure during flight. Miscellaneous : The following items will probably receive more extensive treatment in some other section of the final POP but deserve mention in the present context: Real-time data processing programs should be as simple as possible but still give output of acceptable quality for operational control and quality control purposes. More sophisticated programs, using methods of the highest accuracy and allowing for manual intervention when desirable, should be saved for post-processing. All processing programs expected to be needed should be designed and written before the field tests. Refinements can then be made on the basis of field-test results. This will not prevent bugs from showing up when multi-scale data are processed but might well reduce the processing time by a year or more. A plan for achieving and distributing the output data should be completed before the field phase begins. In view of the large quantity of data 137 involved, some sort of computerized mass storage should be considered. All conventional sensors (at least) should be subject to regular calibration with standards traceable to the National Bureau of Standards. All sensor systems whose output might be combined for analysis must have intercomparison tests made to assure data compatibility. A data management plan, with proper milestones, monitoring, and follow- up action must be designed and implemented to assure that all the good things mentioned above actually happen. 138 EXPECTED REMOTE SENSING CAPABILITY ON A NASA WP-57F APPENDIX E William Shenk, NASA/GTFC 139 1978 1. Cloud Top Scanning Radiometer Measurements of cloud top temperatures, surface temperature and water vapor distribution in the upper troposphere. Visible, 11 ym» and 6.7 \m channels with 110 m spatial resolution on the ground with the aircraft at 18 km. 2. Rain Mapping Radiometer Measurements of rainfall coverage and rate at 19.5 GHz. Mostly used over water. 3. Storm Structure Microwave Spectrometer Measurements of temperature profiles at 60 GHz with short range plans of adding 118 GHz channels for the same purpose. 4. 94/183 GHz Radiometer The 183 GHz measurements will be used to measure water vapor profiles. A window channel at 94 GHz will be used to test ideas concerning hail detection, rain-snow line detection and rainfall coverage and rates. 1980 The following are sensors which are expected to be ready for flight by 1980. Some may be ready in 1979. 1. Cloud Physics Radiometer Determinations of cloud top height using reflectance measurements in the oxygen A band near 0.76 ym and measurements of cloud phase using two reflectance channels: one visible and one near IR. 2. Infrared Temperature and moisture profiles from high spectral resolution measurements plus the examination of other important quantities that can be measured in the 3-15 ym interval. The following instruments are planned for flight on an aircraft other than the WB-57F. 1. Microwave Doppler - vertically pointing 140 Measurements of vertical velocity within convective cells. Most useful when velocities exceed the fall velocities of raindrops. 2. Infrared Doppler - side looking Measures horizontal winds in the clear air using backscattered radiation from aerosols. Will be particularly useful for determining wind fields surrounding thunderstorms. 141 LIDAR & LIGHTNING MEASUREMENTS WITHIN SESAME APPENDIX F R. Serafin (NCAR/FOF) A. Michael Mogil (NWS) W. L. Taylor (ERL) 142 Lidar Within SESAME Lidar is among the few techniques we have for remotely probing the clear (non-precipitating) atmosphere. Combined with acoustic sounders and microwave techniques we have methods for sensing aerosols, temperature structure, velocity structure and humidity structure. We recommend that systems of such variety be deployed within the 1978 experiment. Lidars can serve SESAME in the following ways. 1. A vertically pointing or scanning lidar can measure the relative distributions of aerosols in the boundary layer and the troposphere. These measurements will be extremely valuable in furthering work begun by Danielsen at NCAR in assessing the role that dust plays in severe storm development. 2. A scanning lidar may also be used to measure cloud base heights and to observe early cloud development within the 3-scale network. These measurements can also be made by scanning or vertically pointing 0.8 cm radars. If high resolution measurements are required throughout the 3-scale array a number of ground based systems would be required. In this event a side looking airborne lidar or short wavelength radar should be given serious consideration. 3. Doppler lidar appears to be an attractive tool for measurements of air motion in clear air. Such system should be tested as part of the 1978 experiment. Ground Based Sferics Systems The Environmental Research Laboratory (ERL) and the National Weather Service (NWS), both NOAA components, are currently evaluating several sferics systems in a semi -operational mode. These systems are designed to detect significant levels of thunderstorm generated electromagnetic energy emitted within specified distances of the sferics unit. Each system operates at a different frequency. The Taylor Electromagnetic Tornadic Storm Detector, of which there are both directional and omni-directional versions, will be tested during 1976 at NSSL and at three NWS offices. The omni-directional unit operates at a frequency of about 3 Mhz while the directional unit operates at a 143 frequency between 20 and 80 Mhz. The maximum effective range of the Taylor detector is 70 km. These systems are designed to detect tornadic storms only. The Kohl Sferics Monitor is designed to detect severe convective storms at ranges up to approximately 300 km. The Kohl Monitor operates at a frequency of 500 Khz and will be tested at two NWS offices during 1976. It is desirable that the following studies be made part of the activities of SESAME. Ground and Aircraft Photoelectric Observations of Lightning Data on the nature location and frequency of lightning strokes should be obtained by the use of simple directional photoelectric systems such as the one described by Griffiths and Vonnegut, Weather, 1975. This apparatus consisting of a photocell used in place of the microphone on a tape recorder provides a means for detecting and recording lightning even in clouds illuminated by bright sunlight. Such equipment should be used by both ground and airplane observers to characterize electrical activity in intense sferics. Satellite Observations of Lightning Activity It is desirable that instrumentation be installed on a satellite to provide information both day and night on the frequency of lightning dis- charges from severe storms. Observation from the DAP satellite showing highly unusual electrical activity from the storm system that produced the "Jumbo" tornado outbreak of April 3, 1974 indicates that observations of this kind would identify the development and location of extremely intense storm activity. A variation of the photoelectric apparatus now being used for ground and airplane lightning observations would be suitable method for detecting lightning activity. Balloon Borne Time Lapse Photography of Storms from High Altitude Balloons Information on the time scale convective processes in storm can be obtained by time lapse photography from balloons flown over the storm at altitudes of the order of 30 km or higher. Atkinson and Vonnegut have shown that this simple technique provides a unique view of the convective movements in the clouds below. Such observations made at night should provide new information on the relationship between the lightning 144 that is occurring and the convective structure of the storm, 145 5.2. REPORT OF PANEL 2 SESAME MEETING Norman, Oklahoma March 3-5, 1976 THE THEORETICAL UNDERSTANDING AND NUMERICAL MODELING OF THUNDERSTORMS John Brown Frank Murray William Cotton Harold Orville - Chairman Mike Fristch Peter Ray Tzvi Gal-Chen David Raymond Carl Hane R. Srivastava Joe Klemp Robert Wilhelmsen 146 1. Introduction A major goal in thunderstorm modeling is the understanding and prediction of severe storms. This includes the understanding and prediction of heavy rain, hail, strong straight winds, tornadoes, and lightning. Storm modeling provides a link between the initiation of convection determined from the mesoscale and the actual occurrence of those phenomena. The problems of solution are vast, ranging from submicron microphysical processes, to synoptic scale forcing of smaller scale dynamic systems, ultimately resulting in an entity called a thunderstorm. The storm is a manifestation of important internal interactions among precipitation, cloud particles, and the cloud air- flow and involves further the external interactions of the thunderstorm with the lower boundary layer, with the earth and atmosphere's electric field, with mesoscale pressure systems and other dynamic systems. Since the thunderstorm system is the "engine" by which severe weather phenomena are produced, it seems reasonable that a major portion of the SESAME should include explorations into the "unique" physical/ dynamic characteristics of the severe thunderstorm. Because direct probing of such storms is quite hazardous and, furthermore, the storm is such a large scale system, a detailed observational analysis of the storm internal structure is not likely to be obtained during SESAME. It is thus clear that mesoscale A or thunderstorm scale models can be of considerable value to the SESAME experiment. In order for either the experimentalist or theoretician to have confidence that the model simulated data do indeed have some resemblance of reality, these theoretical data should be verified against detailed cloud observations. We obviously have a problem, however, since as we mentioned above, detailed structural observations on the cumulonimbus scale are not likely to be obtained from the SESAME experiment. It is, therefore, recommended that prior to the application of thunderstorm scale models to SESAME, these models be tested against detailed cloud observations on a hierarchy of convective cloud systems ranging from cumulus-conges tus 147 to moderate intensity cumulonimbus, to moderate-severe cumulonimbus systems, Once such a simulation/verification base has been obtained, one can then apply these models to the severe tornado-producing thunderstorms with a great deal more confidence that the models do indeed simulate "real" convective systems. Furthermore, the only verification data that will then be needed from SESAME, will be general observations obtained of storm structure by Doppler and conventional radar, satellite observations and boundary layer observations. Current work along some of the lines is being done in the NSF's National Hail Research Experiment (NHRE) and the Bureau of Reclamation's High Plains Experiment (HIPLEX) . Coordination with these projects is important. Once the scientific community has developed a high confidence level in the simulation capability of thunderstorm models, we can then use such models to clarify and test various hypotheses concerning the physics and dynamics of severe convective storms. The understanding and semiquantitative information gained from such simulations can also be useful in formulating parameterizations of the convective scale on the mesoscale (a, 3), and as a tool in formulating simplified models or statistical severe weather forecast schemes, assist in the planning and conduct of observational programs and in the post analysis of field data. The techniques used to obtain solutions to such complex problems as cloud models have been primarily numerical. Important insight also comes from the skillful application of more analytical techniques, such as linear or perturbation methods. Wave Cisk theories, etc. The computer requirements are extensive for most numerical cloud models, but may be modest if certain simplifying assumptions are made, such as a steady-state storm. The computer requirements for several cloud model types are listed in Table I. The terminology "bulk micro- physics" or "detailed microphysics" refers to the complexity of the cloud microphysical simulation, whether an assumed size distribution is used or whether many categories are used to represent the cloud and 148 precipitation particle. Slab or axisymmetry refers to the basic geometry of the model. The domain for most models will extent to 20 km in height and to 50 km or more in bredth. The implied grid spacing is then 100 m to 200 m for most of the models, the 3D models requiring some- what larger spacings (500 m to 1000 m) . TABLE I: Thunderstorm Scale Models SECONDS (CDC 7600) Core Storage-Peripheral Storage ID SB IDT Bulk Micro Detailed Micro 2-20 100 1500 10- 3 X 10"* 2DT Slab Sym. Bulk Micro 1000 - 3000 5 X 10** 5 X 10^ 2DT Axisym. Detailed Micro 1000 - 3000 or more 5 X 10** 5 X 10^ 3DT Bulk Micro 50,000 10" 10^ 1, 2, 3D - 1, 2, or 3- dimensional SS - Steady-state T - Time- dependent This report summarizes the panel's discussions regarding important thunderstorm scale studies related to SESAME, listing the solution techniques applicable (1, 2, 3DT numerical models or AN for analytic techniques) and the importances attached to the task (either an A, B, or C rating) . The most appropriate numerical model is listed first and the analytic technique listed last, but this is not to imply a lower priority to the analytical methods. Equal weight is given to the unders tanding and forecasting functions. 149 2. Important Thunderstorm-Scale Studies Related to Sesame 2.1 Input to Parameterization of a or g Scale (Hydrostatic) Models (3DT. 2DT) A Because cloud processes are an important source of meso -a- and meso -3- scale motions, which in turn may trigger additional convective activity, the effects of the convection on the meso-a and meso-3 scales and the resultant feedback to the convection mudy be correctly reproduced by an adequate parameterization. The main aspects of this interaction process are (a) redistribution of mass, entropy, water substance and momentum by individual clouds, and (b) control of the area density and vigor of clouds by their environment. We recognize that current treatment of each aspect is inadequate for incorporation, into mesoscale models. Present parameterizations have been constructed mainly for applica- tion to tropical convection. Since the tropical atmosphere is normally weakly sheared, they generally do not include a parameterization of momentum transport by convection. In addition, most of the formulations ignore the effects of precipitation and assume that the convective system is in equilibrium with its larger scale environment. We know, however, that a severe storm is characterized by strong vertical transport and generation of horizontal momentum as indicated by intense downdrafts. The intensity and, perhaps, initiation of the downdrafts is controlled to a large extent by the weight of total condensate as well as the evapora- tion and melting of precipitation. Another feature of severe storm convection is that once such a convective system is initiated by a meso- scale disturbance, it becomes such an intense system that it appears to propagate and survive somewhat independent of the mesoscale initiation mechanism. From the point of view of a convective parameterization on the mesoscale 3, Oi , this implies that the parameterization cannot be an equilibrium one, but must, in fact, be stochastic. 150 Considering the difficulty of observing the statistical structure of severe storm systems, it seems likely that one of the most powerful tools available to the theoretician is the numerical model on the cloud scale. Just as cloud physicists have used detailed numerical simulations of droplet coalescence to formulate parameter izations of precipitation formation, so also can we use detailed simulations of thunderstorm dynamics to formulate parameterizations of their effects on the mesoscale a, 3. Some of the formulations may be deduced from the "insight" gai^ned from performing cloud simulations under a variety of environmental conditions. Alternately, one might perform detailed statistical analyses of the Reynold's stresses, cloud kinetic energy and covariances computed by averaging cloud-scale simulated data on scales comparable with the truncation scales of mesoscale Band a models. These data could then serve the basis for formulating a truly stochastic parameterization of the effects of severe storm convection in the mesoscale models. We suggest that development of an experimentation with non-hydrostatic models is an important technique to improve the parameterizations and we list four studies: 1. Upward and downward transports by an individual model cloud during its lifetime are available from current 3D models, and greater accuracies in these transports can be anticipated as models are improved. It is not anticipated that such complex models will be used in a manner analogous to the ID models in, e.g., Arakawa and Schubert (1974). Rather, the incorpora- tion of downward transport by negatively buoyant downdrafts should be attempted, using the 3D model results (and observations) as guidance. Suitably designed ID models would likely be incorporated into the parameterization to describe physical and dynamical processes within updrafts and downdrafts. 2. Results from 3D, Y~scale models, capable of describing motion scales (50 km, say) which overlap with those resolvable by hydrostatic meso-3 scale models, may provide an opportunity to evaluate the realism of various hypotheses regarding control of the area density and intensity of convection by the meso-3 scale. 3. The question, cited in the report of the gravity-wave workshop, of how a gravity wave is modified by the deep convection which it initiates, may also be examined in the context of 3D, yscale models. 151 A. The extent to which microphysical processes, such as the presence of various types of aerosol in varying concentrations, may control the evolu- tion of the meso- 3 scales through their influence on convection may be a feasible problem to address via sensitivity tests with models. 2.2 Studies to Understand Processes on the y-Scale 2.2.1 Origin and importance of rotation (3DT, AN) A Tornadoes generally appear to originate from a somewhat larger scale parent circulation called a mesocy clone. The formation of the mesocyclone is therefore of fundamental interest in SESAME. The laws of physics allow two plausible mechanisms for the generation of vorticity and the associated rota- tion. These are the tilting of horizontal vortex tubes associated with verti- cal shear, and the concentration of ambient absolute vorticity by stretching in updrafts. If tilting were the dominate mechanisms, one would expect cyclonic tornadoes from right moving storms and anticyclonic tornadoes from left movers. Since the latter is not observed, the circumstantial evidence is against tilting acting alone. The recent satellite observations (Purdom, 1975) indicating correspondence between tornado genesis and the encounter of a severe storm with the concentrated vorticity of a horizontal shear line are therefore greeted with great interest. It is recommended that the encounter between a severe storm and horizontal shear of varying strength be simulated. It is further recommended that particular attention be paid to this process in the SESAME observational program. 2.2.2 Origin of the downdraft air in severe thunderstorms (3DT, 2DT, AN) A A knowledge of the origin of downdraft air in severe thunderstorms is important because the downdraft is believed to be the storm component necessary for the propagation of the storm. The downdraft transports horizontal momen- tum from higher levels toward the ground producing very strong convergence along the "gust front" and consequently, the upward motion which is the "root" of the updraft. The source of the downdraft air is generally believed to be at or near the level of the equivalent potential temperature (9^) minimum (middle levels) , since the evaporation of rain into dry air produces the negative buoyancy which is the major driving force of the downdraft. The source level is also determined by the level of maximum relative inflow into the storm of environmental air, the most ideal situation being that when the 152 maximum relative Inflow and 0g minimum occur at the same height. To determine the origin of downdraft air by means of a two- or three- dimensional cloud model, one must simply calculate the trajectories of air parcels which have arrived in the downdraft region. These model calculations might be used to point toward areas where observations are needed for verification purposes. Three-dimensional air motions within the downdraft region obtained from multiple Doppler radar systems should provide much of the needed information. Aircraft wind measure- ments and chaff releases by aircraft circumnavigating the storm at the nearest safe distance at several levels should provide much of the necessary information outside the rain area. More redundancy might be added by making chemical tracer releases by aircraft in likely source regions, collecting rainfall for chemical analysis from a network of ground stations, and relating locations of different tracers to deposition patterns of those different tracers on the ground. Conceptual thunderstorm models often indicate the source of downdraft air to be located either on the upshear side or on the right flank of the storm. It has also been suggested that a portion of the downdraft air is comprised of air which has risen in the updraft. In this case, tracer releases in the updraft region and estimates of the elapsed time between release and deposition could provide the answer to this last question. 2.2.3 How do larger scale motions determine the size and distribution of convective elements (3DT) A This is perhaps the most important and most intractable problem associated with convection in the atmosphere today. Though most meteorologists believe that mass and moisture convergence control convection, the details of this process remain elusive. We are still in an exploratory stage on this subject, and feel that observation can presently contribute more to theory than vice- versa. For this reason, we strongly recommend simultaneous observations of cumulus populations and the associated low-level convergence. This perhaps could be undertaken in a preliminary fashion during the 3 scale pilot experi- ment. Penetrations of gust probe-equipped aircraft at various levels above and below cloud base could be used with satellite cloud census techniques to define the cloud populations, while the numerous techniques described 153 elsewhere would measure boundary layer convergence. Using such data as an input into and constraint on cloud-scale models, one could perform numerical experiments under a variety of environmental conditions in order to determine cause-effect relationships. 2. 2. A Boundary layer fluctuations and surface characteristics (3DT, 2DT)A An important application of mesoscale y models would be to determine the sensitivity of severe storms to anomalies in planetary boundary layer fluxes and energetics produced by surface temperature anomalies as well as anomalies in orography and surface roughness on various scales. One would anticipate (and hope) that the genesis and intensity of a severe storm is relatively insensitive to small horizontal-scale surface anomalies. The question is how small (i.e. 0.1 km, 1.0 km, 10.0 km)? A determination of the importance of small scale surface anomalies has direct bearing on the severe storm prediction problem and will largely determine the maximum grid resolution that one can use to obtain "improved" prediction of severe storm propagation and intensity. An important part of such an analysis would be an evaluation of the importance of surface temperature and water vapor anomalies produced by cloud processes, or by differences in surface characteris- tics, such as forests and lakes. These should include wetting and cooling of the surface by precipitation, and the effects of cloud shading (in particular, cirrus anvils) on thermally initiated severe storms. 2.2.5 Energetics (3DT, 2DT) A If one is to understand the processes that go on in the atmosphere, one must study their energetics. This includes the energy budget of the system as a v/hole and the transport and transformation of energy within the system. Atmospheric energy can be broadly divided into two classes — kinetic and static. The latter is further divided into potential energy and enthalpy. Finally, enthalpy may be subdivided into the thermal enthalpy (sensible heat) of each constituent of the atmosphere - dry air, water vapor, liquid drops, and ice particles - and latent enthalpy. At any given scale there is also turbulent energy, which is resolved into the other classes on the next smaller scale. In the course of development of a severe storm there is a continual transformation from one class of energy to another, the transforma- tion being different in different parts of the storm and at different stages of development. Also, advection results in transport of energy from one 154 part of the storm to another. Moreover, there is continual interchange of energy between the different scales of motion through a variety of inter- active mechanisms. The nature of these varied and complex interactions is known in a most general way, and in a quantitative sense only crudely. Yet the values are enormous. One estimate cited during this conference was that a severe storm can release energy at the rate of lO-'--^ watts (presumably this refers to the transformation of latent enthalpy to thermal enthalpy, some of which then goes into kinetic energy) . The most accessible way of getting at precise numbers is through numerical models. On the cloud scale this would mean the use of two- or three- dimensional, time-dependent models. Such models give the wind components, temperature, humidity, liquid (and perhaps ice) content, and pressure at each grid point and each time step. This is sufficient to calculate the complete partition of energy at any given time and its variation as times goes on — that is, the entire history of the energetics of a severe storm and its environment. The latter is a most important point. It is not sufficient to know the energentics of the storm itself. The energetics of the environment is equally significant. What, then, are the uses of these computations? In the first place, they assist us in understanding how a severe storm might be initiated, sustained, and destroyed. This lies at the heart of the whole problem of severe storms. In the second place, they assist us in understanding how one or more severe storms interact with their surroundings, and in particular how the different scales of motion are coupled. In recent years a major concern in numerical modeling has been how to represent important processes of smaller scale, and among the most important of these processes is thermal convection. Implicit in this attitude is that the goal is to model the atmosphere on a large scale (ranging from planetary in the case of general circulation models down to a or 3 mesoscale) . It is manifestly impossible to treat thermal convection fully and directly on these scales, so parameterization is necessary. Since the basic inter- action of scales is an interchange of energy, it is natural to develop the parameterization in terms of energetics. Some parameterizations are developed explicitly in an energetics framework, but without a detailed study of the 155 energetics of the convective scale. Others are based on simplified modeling of the convective scale in which the energetics may or may not be adequately represented. It would seem that the rational approach would be to study the energetics of the convective scale in detail, and then use that know- ledge to parameterize the convective scale explicity in energetic terms. Only then can we be sure that the y-scale is adequately represented in models of a- or 3-scale. However, it is possible and desirable to look at the problem from the other end. An individual severe storm does not grow in isolation, but is influenced by the large scale situation. In particular, the interchange of energy between cloud scale and the larger scales can have a profound effect on the development of the cloud or storm. Hence, parameterization can go in two ways. A cloud model might "parameterize" the larger scale merely through its bou ndary conditions, but it is important to know what types of energy are transported through those boundaries and in what amounts. The conditions may or may not be stated specifically in terms of energetics, but in any case they reduce to those terms. It is therefore recommended that any numerical modeling activities undertaken in connection with SESAME include explicit computations of energetics on the scale concerned and of interchanges of energy with other scales. 2.2.6 Microphysical-dynamics interactions (2DT, IDT, 3DT) B Important microphysical processes have already been mentioned — the water loading, evaporation and melting processes to form downdrafts for example. Other processes of significance involve the microphysics of rain and hailstone evolution, the production of ice crystals, the interactions of various size particles, etc. The cloud's vertical velocity field and the environmental airflow are influenced greatly by the creation and accumulation of water masses in the cloud and the appearance of precipitation outside of the cloud. Vertical shear of the horizontal wind may cause the appearance of a sloping updraft, which then allows characteristically different relation ships between precipitation loading and updraft evolution. Such situations may be one part of the key as to what transforms multicell storm situations into longer lasting supercell storms. Proper simulation of a severe storm 156 requires adequate resolution and representation of these processes in the various models. More microphysical detail can be afforded in the lower dimension models because of computer core and time requirements. Quite likely studies in ID and 2D models will be used to develop better micro- physical parameterizations in the 3D cloud models. 2.3 Initialization and Verification of Models (3DT, 2DT, IDT) . A The observational emphasis of SESAME offers a valuable opportunity to obtain rather complete and self consistent data sets for use in analysis of convective storms. Utilizing Doppler radar with appropriate use of chaff (Particularly for the boundary layer and updraft regions) the three- dimensional, kinematic fields can be determined within the cloud and the under- lying boundary layer. With a knowledge of the time-dependent structure of the wind field we can then seek to use the equations of motion to infer the thermodynamic structure of the cloud and its surroundings. Detailed observational data sets would be of great value in testing the ability of three-dimensional numerical cloud models to simulate the important features of severe cumulus storms. Such data, obtained prior to and during the development of the storm may help to remove some of the uncertainties in evaluating model results and aid in determining the most critical areas in which improvements are needed. For example, model studies verify that the convective development is very sensitive to the initial temperature and moisture distributions within the boundary layer and to low-level diver- gence fields. By incorporating a realistic initial forcing in the model one can avoid the use of artificial or arbitrary perturbations which may spurriously affect at least the timing of the storm development, if not its ultimate intensity. For verification purposes, having a fairly complete observational analysis of a storm one can assess the model performance with regard to a variety of important aspects of the cloud structure. Such detailed compari- sons provide much Improved opportunity for analysis than is possible when only global data, such as total rainfall and cloud top height are available. If global properties of a cloud do not verify adequately, little Insight is provided to determine what aspects of the model are most likely to be responsible. Furthermore, even if global verification is satisfactory, one 157 cannot be sure that features of the cloud such as the updraf t-downdraf t structure, and vertical transports of momentum and energy have been properly simulated. With the generation of kinematic and thermodynamic data sets, the possibility of new kinds of model testing would arise. For example, one could seek to initialize the numerical model at a point during the develop- ment of the cloud, integrate forward in time, and then compare the simulation with a second data set. In addition, the sensitivity of model results to uncertainties in observed and analyzed data could be tested. As mentioned above, the structure of non-steady numerical models requires "complete" initial conditions which are not available, using current observational techniques. The following problems are often cited as short- comings of the observational systems. 1. Current remote sensing devices can measure winds, but not temperature and liquid water content. 2. In non-steady situations the wind data is not given simultaneously, increasing the speed of the scanning usually results in decreasing resolution 3. The wind coverage is often not complete, and there are areas which are occasionally void of data. The fundamental problem is then: to what extent is it possible by continuous updating of the wind field, to bring the model to a stage in which it is completely initialized (i.e., all relevant variables are specified' It should be emphasized that even if the model lacks predictive capability due to inadequate physical formulations, it still can serve as a diagnostic test. The fact that some regions of the storm are occasionally void of data is the most difficult problem to tackle. It is clear that in this case some subjective judgements will have to be made. For instance, one may assume that in these regions, the air is basically environmental, and the buoyancy is zero. The kinematics of the flow in this region may be roughly inferred from nearby aircraft sound data, and/or space-time extrapolation. It is clear that before analyzing real data, careful feasibility and sensitivity studies should be conducted. In feasibility studies one may first create data sets by using a numerical model, and then try to retrieve the temperature field using the "observed" wind field. In order to identify 158 difficulties, problems 1, 2, and 3 mentioned above, may be treated separately in the following manner: a. At first one assumes that the wind field and its time derivative is given and from this using the horizontal vorticity equation, obtains the temperature field. The vertical vorticity equation may be used to estimate the lateral mixing. At this stage the sensitivity of the computed temperature, to the wind errors should be assessed. b. Next one assumes that the data is given bit and bit and then either by "brute force" updating or space-time transformation,* verify that the algorithm outlined in (a) is still convergent. It should be emphasized that 4D data assimilation and initialization techniques have been used successfully in large scale meteorology, and there is reason to believe that a similar approach should work in small- scale meteorological processes. There was not unanimous agreement in the panel in the philosophy of measuring "everything, everywhere, all of the time" to provide initializa- tion and verification data sets. The task of trying to simulate and compare with measurements a particular cloud of moderate intensity is enormous, involving the many difficulties detailed above. An alternate approach is to obtain as much data available as possible from surface and uppel: air soundings and use this for initializing the models, which would then generate a storm through realistic surface boundary conditions and larger scale forcing such as mesoscale convergence. Doppler and conventional weather radar, aircraft penetrations, satellite and ground observations are then used to determine the general convective characteristics, atmospheric electrical development, and main precipita- tion processes operative on the particular day and compared with the "global values" predicted by the cloud models, such as heights of cloud base and top, rain and hail rates and amounts, first echo heights, radar reflectivity factor values, updrafts and field strength, cloud distribution, etc. No direct comparison of model clouds with particular clouds in a severe *A coordinate transformation whereby the "Cauchy Problem" is transformed into initial value problem. 159 storm environment is attempted, but important forecasts and understanding of cloudy convection may be attained without tremendous pressures put on observ ing networks and instruments and the theoretical, numerical, and programming efforts to reconstruct the "complete" and perhaps unattainable initialization and verification data sets. 2.4 Development and Testing of Numerical Techniques and Physical Approximations for Large Cloud Models (3DT, 2DT, IDT) A Cloud modeling requires both a system of physical equations and a means of solving them. Finite difference techniques have been the primary means for solving the dynamic components of the system and will probably remain so for some years. It is important in using these techniques to understand both the stability of the finite difference system and how the finite difference system treats the types and scales of motion that are being modeled. As modelers we would like to avoid as best we can the interpretation of numerical errors as physical phenomena. Research is underway that is directed toward analyzing and testing various numerical techniques for solving simple cloud systems. New techniques are also being developed. It now appears possible to use at least second order time and space approximations for solving a three-dimensional model of reasonable proportion. There remains, however, work that should be carried on. Careful consideration must be given to the treatment of lateral boundaries. Although well-posed sets of boundary conditions have been theoretically derived for inflow and outflow boundaries, these conditions do not insure that a desired behavior will occur at the boundary. Further investigations must be conducted to provide conditions which allow gravity waves to prop- agate through the lateral boundaries without reflection and at the same time permit desired information from larger scales to be specified. The lateral boundaries might also be moved as far as possible from the storm center. Practically this requires some kind of grid stretching or nesting of grids in three dimensions. Reflections of gravity waves occur in these systems. These reflections may be serious and should be investigated. 160 It is also important to consider how to keep the storm near the center of the grid since the domain typically used would not permit a moving storm to be followed for a significant period of time. In summary, modelers need to continue to develop confidence in their numerical and physical approximations. This is a continual process as model objective as well as the models themselves change. New (or old) techniques that improve accuracy or computational speed should continue to be investigated in light of the needs of three-dimensional storm simu- lation. 2.5 Forecasting of Thunderstorms The panel considers the forecasting function to be equal in Importance to the understanding function with regard to severe storms. A project of the magnitude of SESAME should have important feedback to improving the forecasting of severe storms. 2.5.1 Analytical models for prediction purposes A- It has been shown that certain analytical or nearly analytical models of severe convection have the ability to predict storm motions and approxi- mate flow patterns from environmental sounding information alone. Since the cost of these models is relatively low, consideration should be given to the feasibility of using them on an operational forecast basis. The most immediately useful information would be the prediction of severe storm movement. Additional research may reveal correlations between various types of soundings and the resulting class of storm; i.e., supercell, multicell, splitting family, squall line, etc. The greatest forecast potential could probably be extracted by combining these models with statistical techniques described below. 2.5.2 Use of IDT and 2DT models plus NMC data to predict "global" values A- Evidence is accumulating that one- and two-dimensional models, initialized with a single atmospheric sounding may be able to depict the general characteristics of deep moist convection and the primary precipi- tation processes. A judiciously selected "representative" sounding is needed to initialize the ID models which simulate single clouds. A sound- ing near in time and space to storm development is desirable for the 2D 161 models, which have the capability of simulating important interactions with the lower boundary layer as well as the environmental winds (to a limited extent) and may result in multiple cloud and storm formations. NMC has developed over the past years statistical techniques, using conven- tional weather data, to predict the onset of severe convective weather. The time seems propitious, within the framework of SESAME, to apply these modeling techniques to the problem of predicting severe storms. The model- ing and NMC techniques could be applied either separately or combined using the improved, more timely data set provided by METRAC and new ground-based observing instruments. The primary predicted variables would be rain and hail amounts and intensities, high wind (possibly), and lightning. Presently, lightning requires the prediction of charge buildup which is in the model develop- mental stage at this time, but may be far enough advanced by 1980 to pro- vide meaningful forecasts. Alternately we may learn enough about the general properties of the thunderstorm and its environment which are conducive to lightning formation to predict lightning intensity and fre- quency as a function of predicted global properties of the storm. 2.5.3 Combination of above techniques for predictive purposes A The usefulness of cloud-scale models for the prediction of severe storms has been largely overlooked. This is not too surprising, however, since the cloud models are either far too simplified (IDSS) to simulate severe storm dynamics or they are not able with present computer capa- bilities to simulate a storm in real time (2DT) much less be a prediction aid. If one considers the collective merits of a number of simplified cloud and mesoscale models, on the other hand, it is possible that a linear combination of their model output statistics, may provide a fore- cast tool which has quite a high skill level. For example, the one- dimensional, steady-state and time-dependent models are extremely sensitive to moderate changes in thermal stability. They cannot, however, simulate the dynamic interaction of the storm with its environment. The linearized "Wave Cisk" model, on the other hand, cannot respond to changes in thermal stability, but is a good predictor of the phase speed and direction of propagation of supercell storms. The 2D deep convection models may do a 162 better job than the ID models in predicting the formation and fallout of precipitation and the type of precipitation, whether rain or hail. The simplified mesoscale models cannot respond realistically to the effects of deep moist convection, but they do predict the time and spatial evolution of low level moisture convergence which appears to be a prime initiator of and an integral constraint on thunderstorm convection. It is thus suggested that the SESAME data may provide a basis for formulating a sta- tistical regression forecast model in which the component predictors would include ID and 2D cloud model output, "Wave Cisk" model output, mesoscale convergence, as well as conventional NMC analysis and predictions, and local observations such as soundings. The SESAME data would provide the necessary input or data against which the various observed and model output data could then be screened. An alternative way to use the different models as a forecast tool would be to solve in real time a mesoscale model producing predicted atmospheric soundings which would then be entered into ID or 2D cloud models to predict convection and precipitation characteristics. These "multiple model" runs would be pursued only at those mesoscale regions and grid points showing "favorable" convective soundings, thereby requiring less computer capabili- ties. 3. Observational Requirements Many observations have been suggested in previous subsections of this report. We discuss here a few more observations necessary for the successful application of thunderstorm studies to SESAME. To provide for either two-dimensional or three-dimensional, time- dependent model initialization and verification, the observational program should be planned to include the possibility of measuring environmental and (to the extent feasible) cloud scale parameters over the life history of clouds which eventually become severe storms. For initialization purposes the perturbation should be determined as early in the time evo- lution as possible (possibly may mean cumulus congestus or clear air perturbation) . This would require the use of a mesoscale surface and upper air network for the measurement of such things as moisture convergence and surface temperature, moisture, and pressure anomalies in the area where 163 initiation is likely (e.g., in the vicinity of the dry line). Rawinsonde ascents every two hours or less may be necessary for case study situations. In order to attempt at the same time to analyze the severe storm pre- cursors, it would seem to be necessary to have available on a routine basis weather radar with long range capability and possibly high resolution satellite photography. Conventional radar data should provide information on first echo height, base and top (as defined by the 10 or 20 dBz contour) and the evolution of the radar echo patterns giving the maximum dBz's as a func- tion of space and time. The areal distribution of echoes should be recorded every 30 minutes or so. Using multi-Doppler radar coverage, the mean Doppler velocities can describe the morphology of storm motions and the complete Doppler spectra can be useful for analyzing the turbulent structure of the storm. Aircraft data is valuable for determining the partitioning of water substance in the storm's updrafts and downdrafts. The extent of ice or liquid precipitation is needed as well as information on vertical veloci- ties, humidities, and temperatures. In particular, the analysis of the spectral distribution of storm kinetic energy and fluxes is essential for the development of turbulence parameterization schemes for severe storm models and/or the use of optimum grid resolution in such models. Surface rainfall and hail are important products of the storms and must be measured for verification of models and forecasts. Both precipita- tion rates and total accumulation are necessary. The location, intensity, and propagation of tornadoes should also be documented in order to deter- mine what, if any, predicted thunderstorm scale features correlate with observed tornadoes. Visual cloud characteristics can be measured by time lapse photographs and satellite. The satellite data provide additional information on the spatial distribution of cumulus and storm populations. Observations are needed for detailed budget studies (mass, energy, water) of large thunderstorms or thunderstorm complexes. The purpose is to delineate the final disposition of mass which it processes by the storm system. Emphasis should be on observing the cold air outflow at the surface and estimating the relative magnitude of the air which is taken 164 from the planetary boundary layer and detrained aloft and air which descends in the downdraft. Ideally one would like to be able to estimate from observations the cloud induced eddy stress and eddy flux terms which appear in the hydrostatic models when averages are taken over the meso-a or meso-3 scales. To verify the model results dealing with subsidence in the environ- ment of severe thunderstorms certain observations should be undertaken. Systematic circumnavigations at several levels and several proximities to the storm should be performed to obtain the horizontal winds, tempera- ture, and humidity distribution. At the same time chaff packets might be released to determine vertical velocities in the clear air environment. Should circumnavigation be deemed impossible, certain volumes on preferred sides of the storms as selected from the results of two- and three-dimen- sional modeling, might be the object of intensive observations. 4. Concluding Remarks The study of thunderstorms within SESAME is considered a high priority item. Much interaction is needed among the mesoscale and thunderstorm scale modelers so that the prediction and understanding of favorable environments for severe storms is vastly improved. The impact of severe storms on society through loss of life and destruction of property will most certainly increase in the future. The meteorological profession must be prepared to alleviate the problems by giving proper and timely warnings to the public, made possible by better observations and prediction techniques and more complete understanding of the phenomena. 165 APPENDIX SUMMARY OF COMMENTS AND RECOMMENDATIONS (FRITSCH) I. Structure of Mid-Latitude Traveling Thunderstorms in Shear ; There is considerable observational evidence (Newton, 1966; Kropfli and Miller, 1975: Fritsch, 1975b) which shows that the type of deep convection oc- curring with mid-latitude traveling weather systems sometimes takes on substantially different characteristics than isolated or "air mass" con- vection (the type most frequently modeled by one and two-dimensional cloud models). These observations, along with theoretical modeling studies, (for example, Takeda, 1965) indicate that for at least some deep convective clouds and their particular environment, major differences show up in the structure of the internal cloud circulation. The differences appear to be closely tied to the three dimensionality of storm's circulation. Specifically, relatively strong vertical shear and intruding middle level dry air seem to generate an organized circula- tion structure (like the Browning or Newton model clouds) which persists in a pseudo-steady state form. The updraft and downdraft develop side by side and are tilted in the vertical so that condensate generated in the updraft may readily enter the downdraft area where it nourishes the evaporational cooling of the intruding middle level dry air. Upon enter- ing the boundary layer and interacting with the surface, the moist down- draft forces additional boundary layer air upward which re-enforces and perpetuates the updraft. The traveling nature of the storms along with the larg^ scale horizontal flow to the clouds, and the directional shear in the environment, allow the storm to persist in a pseudo-steady state manner for periods of time considerably longer than the "typical" life- 166 time of the relatively stationary or slow moving airmass type convection. Although this asymmetrical internal circulation structure is ob- served quite frequently for thunderstorms in shear, it is not well under- stood what controls whether the storms develop in multicellular clusters or are dominated by large single or "super" cell systems. Furthermore, the control on the number of these storms in a given area is also not understood (although it appears to be related to the amount of buoyant energy available and the ease with which it can be released). In some instances, squall lines or cumulonimbus systems appear to induce an intermediate scale organization in the form of meso-3 highs and lows. This interaction of cloud scale and the larger environmental scale may strongly affect the organization and evolution of the cumulus clouds, particularly with regard to the formation of severe weather. The mechanism which produces the meso high appears, for the most part, to be the result of evaporational cooling in downdrafts, however, the explana- tion of how meso-3 lows are produced remains rather elusive. There is some evidence that, under certain conditions, the warming needed to pro- duce the meso lows may be caused by compensating subsidence ahead of the thunderstorms. Indeed, if subsidence develops in an organized manner in the lower stratosphere and upper troposphere, substantial warming (suffi- cient to cause surface pressure falls on the order of mbs per hour) can occur in the region beneath the subsidence. The downwind subsidence also could have the effect of capping other convective development ahead of the existing convection and then explosively releasing the boundary layer air as the pre-existing convective system arrives. In addition, develop- ment of meso-3 lows (whether they are induced by compensating subsidence or 167 not) serves as a mechanism for increasing low level convergence and further amplifying the convective activity as well as providing a source of rotation for generating the meso-cyclone and possible tornadic activity. II. Parameterization of Deep Convection; Incorporation of the effects of deep moist convection on the synoptic (and intermediate) scale requires either the simultaneous resolution of both the convective and synoptic scales (directly or by nesting grids) or knowledge of relationships between the synoptic forcing and the convective response (i.e., parameterization). Several parameterization techniques have been developed which relate the amount of convection to mass, moisture, or energy budgets of the larger scales. For example, Arakawa and Schubert (1974) relate the amount of con- vection to the rate of destabilization by the large scale. Fraedrich (1974) uses total energy and potential vorticity as controls, while Ceselski (1974) regulates the convection by using large scale lov/-level convergence as the mass flux into cloud base. These types of techniques clearly allow for interaction of the convection with larger scales. However, the or- ganized structure of mid-latitude thunderstorms in shear is such that the updraft and downdraft can "cooperate" to consume the boundary layer air at a rate significantly larger than the rate at which the large scale is supplyi it (see, for example, Newton and Fankhauser, 1964; Fankhauser, ly71, or Fritsch, 1975a). Conversely, the same structure which permits the very efficient consumption of moist boundary layer air also acts to reduce the precipitation efficiency of the thunderstorm since a substantial portion of the condensate is evaporated in the production of the moist downdraft. Precipitation efficiency here is defined to be the ratio of precipitation 168 to vertical moisture flux into the updraft. Thus, for these storms the precipitation efficiency may be low but the net precipitation may still be relatively high since much more mass and moisture are processed through the storms. From the standpoint of parameterization, there is i: large potential error which may be introduced into the larger scale models if this characteristic structure and precipitation inefficiency are not accounted for in one or two dimensional cloud models used in parameterization. Specifically, if the amount of mass entering cloud base is regulated by the large scale low level or boundary layer convergence, and convective cloud dimensions or microphysics are then adjusted so that the cloud model precipitation agrees with the observed, clearly the resulting cloud model characteristic will be in error (this same argument applies for cloud ensembles). More importantly, if then the "tuned" cloud model is used in parameterizing the effects of the convection on the large scale, the magnitude and vertical distribution of the thermodynamic and momentum adjustments may be substantially different from the actual effects of the convection. The fact that the cloud scale vertical mass transfer rate can be several times larger than the large scale supply rate in- dicates that significantly more vertical stabilization and downward momentum transfer is occurring than that indicated by the large scale budgets. III. Recommended Objectives for SESAME In view of the importance of the three-dimensional characteristics of the traveling mid-latitude thunderstorm in shear, it would be rather 169 difficult for one and two dimensional cloud models to provide useful analytical and predictive information for parameterization (other than timing and location of cloud initiation) unless the strong effects of the three dimensionality are parametrically included. Otherwise, the one and two dimensional models may: 1. substantially underestimate the amount of mass and moisture processed through the clouds, 2. fail to predict the magnitude and location of compensating subsidence (both moist and dry downdrafts), 3. fail to predict the location and development of meso-3 systems (with possible rotation), and 4. significantly underpredict the thermodynamic stabilization and vertical momentum transfer. Each of these four items might produce a significant adjustment in the large scale structure. Consequently, two of the primary objectives of SESAME, from the cloud modelling standpoint, should be 1. to provide sufficient initial condition and succeeding informa- tion so that three dimensional cloud models, and their inter- action with the near environment, can be adequately tested, and 2. to identify and quantify those three dimensional cloud character- istics and their relationship to the large scale environment, so that one and tv/o dimensional cloud models may be used in con- ' vective parameterization. 170 There is considerable overlap in the information required for developing convective parameterization techniques and cloud model testing. Primary objectives of SESAME from the convective parameteri- zation standpoint, should be 1. identification of the large scale to cloud scale relationships which determine the type of convection which will occur, 2. quantification of the relationships between synoptic or meso-a mass and moisture convergence and the amount and rate of cloud scale mass and moisture consumption, (budget studies) and 3. quantification of the relationships between available buoyant energy (positive area on thermodynamic diagram just as con- vection begins) and the amount of mass and moisture processed by the clouds. This may require identification of meso-6 enhancement. IV. Recommendations for Observing Network: 1 . Time Framev/ork SESAME essentially addresses the scale interaction problem and the scales of interest range from synoptic through meso - a, 3, and y . Clearly, to understand the contribution of each scale, it is necessary to simultaneously measure all of these scales over a period of time which will encompass the "character- istic time" of each scale. Thus, the necessary (longest) period of time is defined by the largest (synoptic) scale and this places a minimum of intensive effort of at least 12 hours for each event. 171 2. Spatial Considerations In general, the two larger scales (synoptic and meso-a) can be satisfactorily measured by a radiosonde network. Sev- eral observing networks have been suggested. Among these are: a) SESAME PDP network (1975) b) Numerical Modeling Group network, SESAME Boundary Layer Working Conference (Nov., 1975) c) TDL Network, SESAME Workshop (Mar., 1976) Examination of these networks indicates adequate synoptic coverage; however, an important region of meso-a coverage appears to be left out. Specifically, the southv/estern portion of the proposed large scale ob- serving networks, would have great difficulty resolving wavelengths on the order of 200 to 1000 km, and these waves over the Southern rockies frequently trigger deep convection several hours later as they approach the moist boundary layer over the plains. As now proposed, the sounding networks would not resolve meso-a waves until they had moved eastward to near the center of the network, and already begun their interaction with the meso-3 and meso-y scales. This would make the characteristics and structure of meso-a feature rather difficult to analyze and under- stand, and would overlook a fundamental aspect of the prediction problem. A new network, adjusted to compensate for this problem, is suggested in an attachment. Measurements for studying the smaller scales, meso-3 and <»', should include both radiosondes and other less conventional observational systems such as doppler radar, aircraft, and satellites. It is likely 172 that the high density sounding network in the center of the proposed SESAME experimental area will adequately measure the upper portion of the meso-B scale, however vertical circulations forced in the near environment of the convection (e.g. compensating subsidence) may or may not be sensed by a sounding network fixed in time and space. Con- sequently, aircraft observations of temperature, horizontal wind, humidity and hopefully vertical motion, are necessary in the region extending about 100 km radius from the active convection. A dense (L <_30 km) surface network, which measures wind, temperature, pressure and humidity, is necessary for observing meso-3 scale features. Of particular importance are surface wind and pressure since these parameters can detect meso-3 scale development. As far as the meso-y scale is concerned, aircraft and doppler radar observations are a must if we are to close the physical loop from synoptic down to the cloud scale. Direct observation of cloud characteristics (such as dimensions, vertical motion, cloud base, etc.) are needed for cloud modeling and measurements of actual cloud vertical mti^^^ and moisture fluxes are necessary for budget and parameterization studies. These types of measurements can most easily be made using the aircraft sounding platform and dual doppler radar. 173 PROPOSED UPPER AIR NETOORK FOR SESAME' ' ■• The largest square delineated by a dotted line encloses the NWS stations that would make special upper-air measurements for the SESAME data base. The middle-sized square, which is about 1?.00 kin on a side, defines the domain for initializing mesosca.'- prediction models. The average station spacing is about 200 km when only the regular WS rawinsondc stations and the special rawinsonde stations at V.'cather Service Office.* (WSO's; solid circles) are considered. The smallest square, about 500 km on a side, encloses a net of upper-air stations spaced 100 km apart and a net of surface stationf (not shown) spaced 50 km apart. (Those upper-air stations denoted by small hollow circles would require special siting.) Data from within the 500 km square would be used mainly for developing statistical prediction schemes, studying storra-environ- raent interactions, verifying mcsoscale model forecasts, developing cumulus param- eterization schemes, etc. (Data from the three rawinsondcastations in Central Oklahoma, spaced about 60 kn apart, would be used to test the accuracy and spatial resolution of satellite soundings*) 174 Arakawa, A. and W. H. Schubert, 1974: Interaction of a cumulus cloud ensemble with the large-scale environment. Part I. J. Atmos. Sci., 31, 674-701. Ceselski, B. F., 1974: Cumulus convection in weak and strong tropical disturbance. J. Atmos. Sci., 31, 1241-1255. Fankhauser, J. C, 1971: Thunderstorm-environment interaction determined from aircraft and radar observations, Mon. Wea. Rev., 99, 171-192. Fraedrich, K. , 1974: Dynamic and thermodynamic aspects of the para- meterization of cumulus convection: Part II. J. Atmos, Sci., 31, 1838-1849. Fritsch, J. M., 1975a: Synoptic-meso scale budget relationships for a tornado producing squall line. Proc. Ninth Conf. on Severe Local Storms, Oct. 21-23, 1975, Norman, Oklahoma. Fritsch, J. M., 1975b: Cumulus dynamics: Local compensating subsidence and its implications for cumulus parameterization. Pageoph, 113, 851-867. Kropfli, R. A. and L. J. Miller, 1975: Thunderstorm flow patterns in three dimensions, Mon. Wea. Rev., 103, 70-71. , Newton, C. W., 1966: Circulations in large sheared cumulonimbus. Tellus, 18, 699-712. , C.W. and J. C. Fankhauser, 1964: On the movements of convective storms, with emphasis on size discrimination in relation to water-budget requirements. J. Applied Meteor., 3, 651-658. Takeda, T., 1965: The downdraft in convective shower clouds under the vertical wind shear and its significance for the maintenance of convective systems. J. Meteor. Soc, Japan, 43, 302-309. 175 5.3. THUNDERSTORM-ENVIRONMENT INTERACTIONS Report of Working Group #3 SESAME Workshop on Thunderstorm Scale Norman , Oklahoma 3-5 March 1976 Group Members: Chester Newton (Chairman) Charles Anderson Alan Betts William Bonner Charles Chappell Jerome Charba Michael Fritsch Isidoro Orlanski Frederick Sanders Yoshikaza Sasaki Joseph Schaefer Peter Sinclair Contributors to discussion: Stanley Barnes Francis Bretherton Douglas Lilly 176 1.0 Introduction In considering thunderstorm-environment Interactions, we are concerned with the following broad aspects: Synoptic structure of the broad-scale environment, and physical processes that modify It, leading to Initiation of severe storms Evolution of severe storm systems. Including patters of propagation and" movement Mesoscale synoptic features, both Independent of and resulting from convectlve systems Airflow and thermodynamic structures within and In the near environments of storms Statistical properties of ensembles of storms (dimensions, Intensity of vertical motion and precipitation, etc.) relevant to parameterization for numerical models Processes and quantitative evaluation of the Interchange of momentum, heat and water between storms and the larger-scale environment Adequate description of these processes Is an essential step toward physical understanding, which In turn Is a necessity for prediction, either by purely numerical methods or man-machine Interactive schemes, whether by subjective or numerical methods. Storm-environment Interaction Is a two-way process. The character of the environment Influences the form taken by a convectlve storm. In turn, convectlve storms Influence their synoptic-scale environment by releasing heat, and by transferring both heat energy and horizontal momentum between low and high levels. The processes are different, not only quantitatively but qualitatively, In disturbances that are dominated by convectlve and by stable upgllde cloud systems. The character assumed by convectlve storms also differs. In respect to their types and configurations; their Intensities, longevities and movements; and their patters of propagation, depending on the wind shear, stability and water vapor distribution In their host environment . 177 The proposed SESAME program offers the opportunity to study the interactions among scales ranging from turbulent- through storm- to synoptic- scale, in detail as well as over the life histories of storm groups and squall lines, from initiation to dissipation. Through the knowledge gained by intensive analysis during a period of abundant observations, we can hope to develop methods that will be useful when specialized observing tools have been removed or diminished and "conventional" observations remain. These methods may be empirical or numerical, or, as is likely for a long time to come, a combination of numerical prediction calculations of the environmental field combined with empirical knowledge of typical storm behaviour and with objective statistical procedures relating diagnosis and predictions of the synoptic-scale structures to weather events. The proposed observation program will contribute alike to the discovery of empirical knowledge and to development of niimerical prediction procedures. 178 2.0 General Aspects of Storm-Environment Interactions Problems in the following three categories may be considered: 1. Stage setting ; necessary and sufficient conditions for thunderstorm development a) role of synoptic and mesoscale convergence • magnitude and horizontal scale of vertical currents • release of potential instability by layer lifting • triggering by mesoscale features in planetary boundary layer b) role of vertical wind shear • in differential advection • as an inhibitor of cumulus development in early stages c) role of subcloud-layer diabatic and advective processes 2. Environmental control over behavior of storms a) influence of stability and water vapor stratification b) influences of wind shear on storm characteristics, including the configuration of updrafts and downdrafts, and intensity and persistence c) influences on movement and patterns of propagation d) determinants of large hail, damaging wind, intense rain, severe lightning, tornado genesis 3. Feedback ; significant alteration of the environment by storms a) development of meso-systems b) gravity waves, arc-lines, duststorms c) barrier effects, detrainment, scale of environmental compensation d) effects of cirrus shields e) vertical eddy exchange of properties by clouds, and influence upon mesoscale and synoptic-scale processes. 179 In the following sections, some aspects of these general areas will be discussed, without any attempt at a comprehensive review. Along the way, a few specific recommendations will be given. An attempt has been made to avoid duplicating discussions in the PDP or by other groups (such as in the Boundary-Layer Workshop) , but some repetition or elaboration is desirable for emphasis where specific comment is called for . 2.1 The Pre-Storm Environment The role of differential advection in a baroclinic air mass in which the wind turns with height, in contributing to changes of potential stability of the layer between the planetary boundary layer (PBL) and the middle and upper troposphere, is well known. This is principally a synoptic-scale process, although it is not known to what extent smaller-scale features are a part of it. Owing to physiographic controls, the process is enhanced over the Great Plains region by the development of intense low-level jets (LLJ) , making this region especially susceptible to rapid changes of stability. The maximum-speed core is about 500 m above ground level and thus in the layer where advection of moist air with high static energy (or high wet-bulb temperature, 6 ) is important for generation of potential instability. The limited special observations that have been made (by pilot balloons) suggest that when the LLJ is well developed its width is around 600 km, while at times when the winds are comparatively weak the LLJ tends to be broken up into fingers that are as little as 200 km wide. Systematically, there are large diurnal variations in both the along-streara and cross-stream velocity components, which commonly exceed 10 m/s; thus on occasion the LLJ velocity may vary from 20 m/s 180 in midaftemoon to AO m/s around 0300 local time. This variation is due to a strong downward eddy transfer during daytime that abstracts momentum from the LLJ, setting up an inertial oscillation (upon release of the frictional coupling) that results in supergeostrophic winds in the early-morning hours. There may also be subsidiary influences owing to diurnal variations of the thermal field. The diurnal variations of wind velocity are best developed near latitudes where the period of the inertial oscillation is in phase with the diurnal cycle of insolation (30** latitude) , or somewhat farther north where the general velocity field is on the whole stronger. Along with the diurnal variation of the LLJ, the low-level convergence field changes, and this has been held accountable for the generation of thunderstorms at night. It may be noted also that, in situations when other conditions (e.g. the presence of a capping inversion) inhibit the formation of thunderstorms at night, the rapid northward advection of high 9 during the night may enhance the intensity of convection on the following day. Although the properties of the LLJ itself have been accorded a fair amount of study, it appears that insufficient attention has been given to the quantitative influence of LLJ's upon changes of potential stability. Therefore, We recommend that , prior to full-scale SESAME operations, studies be supported to ascertain the advective influence of the low- level jet stream and its diurnal oscillations, in combination with radiative influences, upon changes of potential stability in the layer between the moist planetary boundary layer and the middle troposphere. 181 In severe storm forecasting practice, it has been observed that rapidly-moving and rather small-scale troughs and ridges in upper levels (small compared with ordinary cyclone-scale upper waves) frequently set off convection. Through the linkage between their upper-level divergence patterns and low-level convergence, it can be suinnised that thickening of the moist layer, with destabilization, probably takes place as such upper-level features approach the maritime tropical (mT) moist tongue. These features should be discernible through the meso- scale rawinsonde observations and ancillary observations. For special soundings to be planned, advance detection of the small upper level perturbations, by observations west of the SESAME network, is required. As a second related consideration, it has been suggested that convergence along the dry-line may be enhanced when daytime heating leads to an adiabatic lapse rate in a deep layer of the lower troposphere west of the dry-line, resulting in strong downward transfer of momentum there. Furthermore, convection is often set off when cold fronts from the Pacific encounter the western boundary of the mT airmass . The structure of such fronts as they pass over the Rockies onto the plains , and the connected circulations, are not well known; there are also timing problems that might be resolved if the characteristic behaviour of the cold fronts were elucidated. In light of the above three features (need for detection of small upper-level perturbations, momentum transfer west of the dry-line, better observations of cold fronts) , We recommend that the network of aerological stations making serial ascents be enhanced not just over the region of 182 expected convective activity, but also to the west of the normally- observed western boundary of maritime tropical air. Convective storms are commonly set off by a variety of mesoscale features as noted in the PDF, including cold-dome boundaries residual from earlier convective disturbances that have died out. There are also indications that storms may form preferably where "hot spots" exist as shown by surface temperature and dew-point analyses, although the association is not firmly established. The proposed enhancement of the surface network, as described in the PDP, should be adequate to delineate most of the mesoscale features (including also small-scale cyclone waves that sometimes form on fronts). Also, satellite observations will make it possible to test the generality of the tendency that has been noted, for convection to form earliest in cloud-free regions with enhanced insolation. A probable exception to this adequacy of observation is the dry-line. While a "synoptic-scale" feature in the sense of representing the extensive western boundary of the ml air mass, the dry-line is a sharp discontinuity (less than 1 km wide). It is, as is well known, a favorite locale for the initiation of convection, that, in the form of a squall line, may advance eastward some hundreds of kilometers over a broad area that experiences severe weather. The dry-line is not necessarily a straight-line feature, and it has been suggested that a wavy form, in which certain parts may favor convection, may be present. (These may be associated with roll convection oriented along the wind in the dry air to the west when it is heated, bringing down more westerly momentum in the descending branches of the rolls and 183 thereby enhancing convergence along the regions where these intersect the dry- line) . With the ordinary surface network, and possibly with the enhanced mesoscale network, the dry-line and its perturbations, and the migration of such features, are not easy to resolve with confidence. While satellite observations help, they are most useful after convection is initiated. Limited observations have also indicated that, in the region east of the dry-line, pronounced perturbations in the depth of the mT air are present, probably standing gravity waves initiated at the dry-line convergence zone . Considering the importance of the dry-line as an initiator of convective storms, its fine scale and difficulty of location, and the desirability of examining its nature along with that of the associated standing gravity waves which may also trigger convection. We recommend that on suitable occasions systematic aircraft traverses of the dry-line be undertaken, to establish its location and delineate its physical structure and circulation at the dry-line and for some tens of kilometers to either side. While the dry-line has been singled out as a particular feature worthy of intensive investigation because of its frequent association with developing convection, there are a variety of other mesoscale features that also should be given special attention. Perturbations on a scale of 100 to 200 km in analyzed fields of surface temperature, pressure and/or moisture convergence are often associated with thunderstorm development. These features can appear several hours before strong 184 convection becomes apparent from satellite photography. When they develop in conjunction with synoptic-scale discontinuities (i.e. , fronts, dry-lines) it has been shown that they serve as a focal point for severe thunderstorm occurrences. However they also can form at locations considerably removed from synoptic scale surface features. If the atmosphere is conditionally unstable, such a feature may also serve as a triggering mechanism for severe convection. Considering the present limited understanding of the perturbations mentioned above, we recommend that on some occasions when these precursor mesoscale features are evident from surface observations (or other indications) , aircraft be directed to them to make traverses to determine the three- dimensional structures as necessary to investigate their physical nature. Indirect sensors will be useful in monitoring changes in the depth of the moist layer, on a continuous basis to detect slow or rapid changes that take place between aerological stations and between the times they make observations. Monitoring by satellites, of water vapor fields and of the temperatures of the tops of stratocumulus decks often present before deep convection sets in, will be useful to provide continuity between aerological stations. (Conversely, other SESAME observations can aid in assessing the fidelity of satellite- observed features in the water-vapour and temperature fields and the significance of their mesoscale structures in relation to storm development, looking toward the use of satellite observations in forecasting after SESAME.) It would also be desirable to use small aircraft to make extensive traverses alternately ascending and 185 descending through the stable layer atop the moist air, to assess changes in its height and intensity both prior to convection and in its general vicinity after its onset. Since this monitoring problem is discussed elsewhere and its importance is generally recognized, we consider it unnecessary to make any specific recommendation. 2. 2 Environmental Influences on Convective Storms 2.2.1 Boundary layer-storm layer interactions . Recent aircraft observations have shown that, just below the cloud base of hailstorms, updraft air is negatively buoyant but rising. This indicates the presence of nonhydrostatic pressure fields associated with the cloud itself, i.e., present only after the cloud forms. The cloud may have formed in a region where the usual processes were adequate to form it, for example from surface heating; or where there were special conditions conducive to formation, such as the convergence at a dry-line or line of confluence; but as shown by the observations it may have moved into a region unfavorable for convection to be initiated (or renewed) except through the action of the preexisting storm itself. A squall line or a large severe storm may move a large distance through such an "unfavorable" environment; the conditions under which it may continue to persist, or will die out, is thus a major forecasting problem. Considering this behaviour, considerable emphasis should be put into further examination of the processes in the subcloud layer and just above it, in the vicinity of active storms. Observations indicate that the air feeding the updraft may rise from the lower few hundred meters with greatest 9 (rather than the subcloud layer as a whole); at distances up to 20 km or so from the locus of the updraft at cloud base; and that subcloud updrafts generally over 5 m/s and up to 15-20 m/s exist despite the observation that the subcloud updraft air is 1-3° colder than the environment at the same level. Determination of the motion field, the 186 distribution of thermodynamic properties, and the pressure field, are essential to understand the processes. The boundary-layer group indicated the importance of establishing the pressure field as well as the mass- and water-vapor-divergence distributions at the surface and in the PBL. The opportunity should not be lost, during the intensive study of the PBL by other types of observations, to use these as a base for making detailed explorations on the storm scale, using aircraft. Of the desired quantities, good instrumentation is at hand for measuring the horizontal and vertical wind components, and the temperature and water vapor distribution, directly. Probably the best method of estimating the pressure field is to calculate it from the air motions (both vertical and horizontal accelerations of air approaching the updraft intake indicate nonhydrostatic pressures of the order of 1-2 mb) . This would require extensive mapping by aircraft flying at two or more levels in the subcloud layer (in addition to the surface observations) , with traverses both in the fore-and-aft direction and laterally across the subcloud updraft region. Repeated observations of this sort might be useful in establishing the degree of validity of steady-state assumptions about the subcloud airflow relative to a storm. They would also be useful in assessing the influence of storms, perhaps passing nearly over and perhaps at a considerable distance, upon the measurements taken by a surface or upper- air sounding station. This assessment is important in respect to the representativeness of observations used for calculating divergence, advection or other quantities on the synoptic and mesoscales. Aircraft 187 observations in the subcloud air feeding a storm are, of course, also essential for understanding the processes within the cloud. Observations of the airflow and thermodynamic properties in down- drafts (whose characteristics are less known than those of updrafts) below as well as above cloud base should also be made. Measurements of liquid water content and of drop-size distributions should be included. These would aid in the evaluation of the evaporation processes that (along with the weight of water and ice particles) drive the downdraft. The relative contributions to evaporation and conductive cooling of the air from large water particles (which fall rapidly) and from small droplets (which evaporate most effectively but fall slowly, unless they are dragged along by the larger particles) need to be assessed. Traverses should also be made in the horizontal outflow regions (as part of storm circumnavigations which will assess the total budget of properties in the subcloud region) , to get quantitative appraisals of outflow influences on modifying the general planetary boundary layer. We recommend that two or more aircraft be committed, either by SESAME or its cooperators, to make systematic measurements of the airflow and thermodynamic properties in the updraft intake and downdraft exhaust regions, both directly beneath the drafts and, as necessary, up to 20-30 km distant from the draft bases. For measurements (including drop-size distributions) in the active downdraft region, highly stressed and armored aircraft will be necessary. While planned or chance rawinsonde ascents will provide some information inside convective systems, and ground-based Doppler radars 188 will volumetrically observe storm circulations, aircraft must be employed to get extensive direct measurements of the internal structures of individual storms. Horizontal profiles of vertical and horizontal velocity, temperature and humidity are wanted both directly beneath the updraft intake and at several levels in the storm, right up to the top. Comparison of the properties, such as 9 at the various levels along with the mass fluxes, can be used to deduce the influence of entrainment and the extent to which entrained air affects the outer sheath and the core of the updraft (as well as the downdraf t) . While it is generally accepted that the major properties of a convective cloud (buoyancy and intensity of vertical motions, cloud top height, response of draft shape to horizontal drag forces associated with vertical shear, etc.) are inverse-diameter dependent, the details of the relevant processes are not well established. In addition to measurements in a sampling of individual clouds, a "cloud census", especially of the spectrum of cloud diameters and top heights (by use of aircraft and RHI radar) should be obtained within the context of the other observations by SESAME. In connection with budget studies tying in convective fluxes with assessment of horizontal fluxes on the larger scale, the assumption of the form of the updraft profile strongly affects the quantitative flux calculated, with a given amount of total mass flux in the updraft, or downdraf t. (See Section 2.3.) Fast updrafts are visible as cloud turrets that may extend up to a few kilometers into the stable air above an anvil top. From the height and horizontal dimensions of a turret, the updraft velocity in 189 the upper part of a storm may be estimated, and, with certain assumptions, the dimensions and mass flux of the updraft. The method requires validation, by direct measurement of vertical velocity profiles by aircraft penetrating turrets near the anvil-top level. If found valid, photography and infrared radiation measurements of cloud tops by overflying aircraft, along with satellite observations, would be an economical way to estimate the updraft properties of populations of clouds such as in squall lines. The provision of airborne downward- pointing Doppler radar would yield measurements of vertical velocities both in updraft and downdraft, along with the shapes of tilted drafts and the circulations in regions where particles are expelled into the anvil or into the precipitating part of the cloud system. Direct in-cloud measurements, or indirect sensing of this kind, will also provide a check on the vertical velocity structure obtained by integration of cop lane-scanning Doppler radars. The variation of updraft velocity with height above cloud base is a matter of some controversy. Chaff, released into updrafts below cloud base and tracked by radar, has indicated maximum upward velocities in middle levels of the cumulonimbus. If the vertical velocity continued to decrease above such a level, this would be at odds with the heights of cloud tops that are observed, which require large vertical draft speeds in the upper parts of the cloud. It has been suggested that there may be two levels of maximum vertical velocity in updrafts: one in middle levels where upward acceleration is suppressed by the accumulated weight of condensed water, and another one higher up where this effect is diminished relative to that of buoyancy. The actual 190 nature of the updraft profile seems to be poorly established, and aircraft measurements at several levels could help to resolve the uncertainties. This question is, of course, also tied physically to the degree of entrainment. The assessment of this, as pointed out earlier, would also be aided through determination of the horizontal profiles of properties by aircraft traverses at several levels. Considering the need for information on the interior structure of severe storms in connection with several fundamental physical questions as outlined above, We recommend that SESAME obtain, perhaps partly through the cooperation of agencies with an interest in assessing turbulence and other hazardous features of severe storms , the services of two or more highly-stressed and well- instrumented aircraft for measurements" of vertical and horizontal wind and gust components, temperature and other elements, at various levels in convective storms. 2.2.2 Influences of Vertical Shear . The old notion that thunder- storms could not survive the presumed disruptive effects of vertical shear has been replaced by the observation, first put forward by severe weather forecasters, that severe thunderstorms actually have a predilection to form in the vicinity of low- level jets and upper- tropospheric jet streams, where the vertical shear is large. This is partly because synoptic-scale differential advection and vertical motions, that generate and release potential instability, are most pronounced in these regions. It has also been demonstrated that a large- diameter storm with a fast updraft can resist deformation by the wind 191 shear, because of its large mass and the tendency for horizontal momentum to be conserved in the vertically-moving air. In recent years, the concept has emerged that storms influenced by strong vertical shear differ in fundamental respects from the typical "air-mass" thunderstorm in a relatively stagnant environment. Basically different forms are taken by the internal circulations, with the result that thunderstorms in a strongly-sheared environment tend to be more persistent than air-mass storms. They are also more vigorous in all respects, partly because they form in regions where the synoptic influences mentioned earlier favor great instability, but also because of the influences of shear upon the cloud itself. Vertical shear has two principal influences upon a storm. One is that the drafts of sheared storms do not stand vertically, an effect attributed to storm movement and conservation of horizontal momentum in the drafts. Owing to their sloping configuration, the updraft and downdraft do not compete for the same space and displace one another, as is the case with an air-mass thunderstorm cell in which the updraft evolves into a downdraft. By contrast, in a sheared thunderstorm the sloping drafts can persist side-by-side, with the condensing updraft shedding water into the downdraft. Partial evaporation of the precipi- tation, along with the weight of the water, drives the downdraft circulation. The diverging downdraft, in turn, lifts moist air to renew the updraft. Thus in this kind of storm the updraft and downdraft act in a cooperative manner, each contributing to maintain the other. A second aspect of shear in the environment is that the winds at low and middle levels transport moist and dry air to the storm. In 192 contrast to an air-mass storm that consumes the boundary- layer air beneath it and nearby, a storm in a sheared environment has air supplied continuously to it by the general horizontal flow, sustaining a large flux of energetic air in the drafts. This is partly satisfied also by the movement of the storm, which may enhance the intake of air. It has not been fully explained why persistent single-celled storms seem to dominate on some occasions while multicellular evolving clusters are preferred on others. It has been suggested that the preferred behaviour is a function of the strength of the vertical shear, and also the character of the hodograph; that is, that a linear hodograph may produce a different storm type than a strongly curved hodograph. A Richardson-number concept has been proposed, the general principle being that "steady-state" storms can only exist if the vertical shear is properly matched to the degree of instability. If the air is very unstable, the updraft flux will be very large, and if the vertical shear is weak, not enough air is supplied to the storm to sustain the updraft. The notion is attractive, and needs to be tested against observations. Recent numerical studies have also indicated that the cross-stream circulation that precedes a modeled squall line is strongly correlated with low Richardson number in the frontal region, and that the intensity of such a circulation also depends on the component of vertical shear across the front. The closed circulation is developed as a result of ageostrophic vorticity generation associated with forced symmetric instability. Observational evidence is as yet inconclusive, and the 193 hypothesis could be tested against the detailed soundings afforded by SESAME. In addition to considerations of the effect of shear on individual cells, there are influences on the patterns of propagation, on at least two scales. Considering the scale of a squall line that may consist of an array of multicellular clusters, there is a tendency for new clusters to form on the south or southwest end of the squall line. Considering an individual cluster, there is also an apparent preference for new cells to form (in most cases) on the right rear side, with marked influences on the migration of the cluster as a whole. These forms of behaviour have been described, but are inadequately documented quantitatively and incompletely understood. They can best be examined in the context of an adequate network of surface and upper- air stations, along with radar observations and other supporting observations such as by aircraft, that will be afforded by SESAME. To examine the above influences, We recommend that during SESAME various quantities be computed and mapped routinely over the whole aerological network, to describe the flow field and energy-related quantities for interpretation of storm behaviour. These include mean vector wind in the troposphere, vector wind shears, integrated vector products to show the horizontal moisture flux both in the whole troposphere and in the planetary boundary layer separately, divergency of moisture flux, kinematic divergency and vertical motion, measures of potential stability, and Richardson. numbers defined in various ways . 194 2.3 Influences of Convectlve Storms upon the Environment Various disturbances that are initiated by severe storms, some of which may travel an appreciable distance from the origin and may influence later formation of severe convection or other weather phenomena, are noted in Section 2.0. These are also mentioned in the PDP and discussed, with a comprehensive review especially of gravity waves, in the report of the Boundary Layer Workshop. Hence we shall not elaborate on this aspect of the influence of storms on the environment. It should be noted, nonetheless, that mesosystems such as squall lines overturn the environment over large areas, at times comparable in extent to a major part of the proposed SESAME aerological network, so that the water vapor and temperature (and correspondingly the 6^) as well as the wind field may be completely modified. Over the region from the Great Plains eastward to the Great Lakes, thunderstorms occur most frequently during the late evening and early morning hours. (Thus aircraft investiga- tions of active storms will probably for the most part be constrained to late afternoon and mainly to the western part of the region.) For a comprehensive study of planetary boundary-layer modification by thunder- storms, following through during the period of restoration by diabatic processes, and advection and vertical motions, serial ascents should be made for protracted periods including the night hours. This is, of course, also necessary for studying such features as the low-level jet stream, and the large ageostropic perturbations imposed upon the wind field in the upper troposphere by convective disturbances. We stress also that, in respect to budget studies as outlined below, which are essential to link the influences of convective clouds to the 195 mesoscale and synoptic scale, it is essential to study the full range of circumstances from marginal to obviously explosive synoptic situations. Also, on days when significant convection is expected, soundings should be initiated well before the onset so as to obtain budgets during quiescent periods, through periods of rapid convective development to fully-developed convection. From the convective parameterization standpoint, desirable objectives are (1) to determine the type of convection likely to be dominant under given large-scale conditions of stability and shear; (2) quantification (through budget studies) of relationships between synoptic or a-scale mass and moisture convergency and cloud processes; and (3) relating the available buoyant energy as shown by soundings to the amount of mass and energy pro- cessed through the convective clouds. To understand the contributions of and interrelations among the different scales, it is necessary to simul- taneously make measurements on time and space scales appropriate to each. 2.3.1 Convective Mesoscale Budget Studies; Meso-g Scale Rawinsondes 1. Purpose A mesoscale budget analysis permits the computation of the convective net transports (and sources) of mass, heat, water, momentum and vorticity from sequential sets of rawinsonde data adequate to define the mesoscale field. It provides a quantitative value for the feedback of the convection (whether cumulus or severe storms) on the next larger scale (meso-a) . This is needed: a) As a quantitative adjunct to descriptive and conceptual models based on mesosynoptic analyses, and quantitative cloud-scale analyses (e.g., based on multiple Doppler radar). 196 b) To provide diagnostic data for comparison with fine-scale hydrostatic model convective transports using parameterization theory, to improve such parameterizations where necessary to Improve forecasts. c) As an approach to some integral properties of a severe storm system, which may be difficult or impossible to derive by integration from the microscale. For example, it is very difficult to compute net condensation and evaporation from the details of the microphysical-dynamical interaction, but much easier from a meso-3 budget. 2. Requirements a) In general, a rawinsonde network which will provide good budget data will also provide excellent meso-3 fields for analysis of inflow and circulation of severe weather systems. However, the budget requirements are in general more stringent. b) Horizontal fields of data are essential at all pressure levels from the surface to say 100 mb (above the level of the convection, where w or o) = 0). We need fields of p, T, q, u, v and liquid water (e.g., quantitative 4-D radar coverage of the entire budget volume, not sector scans). Liquid water fields are needed because cloud storage terms are large during periods of convective development and indeed, comparable with the convective transports themselves. c) Horizontal resolution: Essentially the assumption of linearity must be applied between soundings, even with a field analysis. For severe storm 197 environments a horizontal resolution of 50-75 km (or closer at a few stations) seems desirable to resolve the variance in the humidity and wind fields. Some compromise between spacing and area covered by the SESAME meso-3 net is clearly essential and must receive careful consideration. In our opinion, 100 km is too large a spacing for the meso-$ research net. d) Time resolution: Rapid evolution of convection can be expected, and the capability of a 60-90 min sonde frequency at all sites is essential. Budget analyses require two time intervals and probably either further time averaging or suitable averaging of similar events or situations. This is however, related to the sonde accuracy (particularly wind for divergence) and how well the field , analysed from the separate soundings , represents the divergence. Sufficient numbers of days must be sampled at the highest time resolution feasible for continuous operation. Sondes can be precision sensors and be precalibrated (e.g., as in GATE). e) The whole spectrimi of convective states should be adequately sampled, from suppressed conditions to understand the evolution of the mixed layer, to the initiation and development of severe storms. 198 f ) Sensor accuracy : Temperature ± 0.2°C ^ ^ . , , ^, 5 mb vertical resolution Humidity < ± 5% (at warm temperatures) * Pressure ± 1 mb Wind ±0.5 m s~ ^ (or better), 25 mb vertical resolution (or better) * If the METRAC system gives absolute height to better than 10 m then correspondingly higher accuracy in the pressure sensor might be desirable to enable some investigation of non-hydrostatic pressures near severe weather (e.g., on special sondes). g) Sensor lags, systematic errors, etc.: The thermal lags of thermistor and hygristor and humidity lag of the hygristor should be tested in the laboratory as well as field tested. The sensors should be checked for radiation errors also. METRAC must be very carefully tested in the field for interference between sondes in adverse conditions of sonde drift, electrical interference, and ducting, etc. h) Analysis plans: Field (possibly combined with radar fields) analysis of meso- scale features is a clear requirement. Some time and space interpretation is needed. The thermodynamic budget analysis is reasonably well understood. The dynamic analysis (momentum and vorticity) has not been adequately studied, and study of this is clearly a prerequisite to the understanding of tornado cyclone initiation. Budget studies must be done in conjunction with careful mesosynoptic and radar studies of the evolution of the convection. Close cooperation between diagnostic budget studies and cloud modelling 199 is needed both to assist the modelers by specifying constraints, and to provide conceptual model guidance for the further interpretation of the budget transports and sources in terms of cloud scale processes, i) Adjunct observations: • 4-D quantitative volumetric measurements of total liquid water (e.g., by radar) to estimate cloud storage changes. This is needed for the heat and water budgets. It is also necessary to estimate the percent of the budget volume saturated at a given time . • Doppler radars could provide air motions on the cloud scale. They might also provide a vertical flux of liquid water/ice as a partial check on the budget computations of the vertical flux of water integrated over the volume. • Cloud and system motion fields (derived from radar or satellit are essential for comparison with air motions. • An adjunct aircraft observation program is needed to resolve finer than rawinsonde grid scale and provide turbulent flux estimates and independent estimates of mass divergence, etc. at crucial levels, as discussed in the next section. • Surface flux measurements or estimates: (i) Of precipitation to check the water budget [with (iii)]. (ii) Of radiation and ground storage to provide surface energy balance. (iii) Of sensible and latent heat and hence moist static energy to check the thermal energy balances. 200 (The thermodynamic budgets give the coupled fluxes S+Lq, L(q+£) , S-Lil, where S is sensible heat, q and Z are vapor and liquid water content, and L is heat of condensation. Of these, two are independent.) We recommend that studies for the purpose of assessing convective cloud fluxes in relation to the mesoscale and synoptic scale energy and momentum budgets be supported by serial rawinsonde ascents at frequent intervals, during situations wherein weak as well as strong convective activity is anticipated, and throughout the periods from the quiescent pre-convection state through the time of fully-developed convection. The commencement of serial soundings prior to onset of convection is also essential to examine the question whether organized convection consumes boundary-layer air rich in thermodynamic energy, at a rate significantly greater than the rate at which the synoptic-scale convergence is supplying it. It is conceivable that overturning due to the cooperative updraf t-downdraft coupling in sheared thunderstorms (Section 2.2.2) is so effective as to lead to this result. The question is significant in respect to modeling, affecting assumptions about the relation between vertical transports in clouds and boundary-layer convergence, and also the contribution of cloud-induced subsidence to stabilization and to the vertical transfer of properties. 2.3.2 Direct Evaluations of Vertical Fluxes in and Near Storms The vertical transfer of sensible heat, as expressed by the correlation between temperature and vertical velocity, depends not only upon the mean value of these properties in the cloud but also upon their distribution within the cloud. It is essential to make direct 201 measurements of the appropriate quantities in a population of storms, to serve as a guide for energy-flux parameterization schemes as well as a basis for deducing the quantitative influence of entrainment upon properties at varying distances from the updraft core, and in the downdraft. The importance of such measurements can be seen by comparison of two extreme assimiptions about the shapes of the horizontal profiles of vertical velocity and of cloud temperature deviation from the environment, within an updraft. One is the "top-hat" assumption used in one-dimensional cloud modeling, i.e. that the vertical velocity and energy properties are uniform across the diameter of the updraft (an assumption that leads to inconsistencies of updraft intensity, diameter, cloud top height, etc.) Aircraft penetrations of hailstorms have, on the other hand, indicated that the vertical velocity profile, and also the temperature profile, are well approximated (on the average) by a bell-shaped curve. In this case, there is a correlation between w' and T' within the updraft. A simple calculation shows that, for the same mass flux, an updraft with a bell-shaped profile transports about twice as much heat as one with a top-hat profile, owing to this in-cloud correlation. The same consideration applies to the vertical eddy flux of water vapor and horizontal momentum. The maximum vertical velocity and other properties of updrafts depend on the characteristics (stability, water-vapor content, and wind shear) of the cloud environment; and also, because of the effects of entrainment, upon the horizontal dimensions of the updraft. Presumably, the rate of production of precipitation (which can be 202 estimated from radar observations combined with surface rainfall measurements) can be related to the updraft mass flux. If, then, some generally valid profile characteristics can be established by aircraft measurements at various levels, in a reasonably large sample of clouds with varying dimensions, this would serve as an aid in assessing the energy and momentum fluxes accomplished by a population of clouds. Observations by aircraft should not be confined to the cloud itself, but traverses should encompass the region up to at least two storm diameters distant from the cloud boundary. This is essential not only to evaluate the vertical fluxes that may be quantitatively significant in the clear environment disturbed by a cloud, but also to answer certain qualitative questions. It is not known to what extent subsidence compensating the updraft mass flow takes place in the near vicinity or distant from a storm, and, if near the storm, how this subsidence is distributed spatially. The question is relevant to individual storm behaviour; i.e. , subsidence nearby a storm might suppress other convection in its vicinity, as suggested by some satellite and other observations. This would favor a natural selection process in which active large storms grow at the expense of smaller-scale convection. There is suggestive evidence that in cases of strong vertical shear, subsidence may be induced as much as 50 km or so downwind from convective storms, contributing to mesoscale surface pressure falls there; and also to "capping" of the moist layer, focusing the explosive release of the thermodynamic eddy energetic air in the convective system. In examining these questions, dropsondes in the neighborhood of storms would also be useful. 203 With the above considerations, and in line with a related recommendation in different terms in Section 2.2.2, We recomnend that an extensive aircraft sampling program be undertaken in conjunction with the SESAME observations, to measure the vertical velocity, horizontal momentum, temperature and water properties in the updrafts and downdrafts of storms of varying sizes and at several levels, to obtain information essential to generalizing their properties for the purpose of parameterizing vertical fluxes by a population of clouds. 3.0 Comments on Aerological Network Two proposed arrays have been presented: (1) Fy79 SESAME meso-a network, and (2) BL workshop meso-a network. The former is oriented with its long dimension W-E, perhaps the most suitable arrangement for numerical modeling (?) and for including eastward- moving disturbances. The latter augmented network extends farther north and south. If a choice had to be made between having the long dimension S-N or W-E, the greater range of latitude would be preferable for the following reasons: (1) This would permit optimum analysis of "Colorado cyclones", i.e., of disturbances that amplify in the lee of the Rockies and are responsible for development of (2) Low-level jets, which are oriented primarily S-N because of terrain effects, and should be detailed as fully as possible by the observations, from their source over the Gulf of Mexico or the eastern plains of Mexico to the northern plains. The low-level jet is responsible for rapid destabilization through (3) Differential advective processes, with strong influences of 204 (A) Inertial oscillations, which vary with latitude and should be detailed over as large a range of latitude as feasible. The low-level jet, and the associated inertial oscillations, contribute in an important way to the (5) Divergence field (of mass and moisture) . Since a leading object of the mesoscale experiment is to study and model this field prior to onset of convective disturbances, concentration of observing facilities over the plains and neighboring states would be justified by the fact that most organized convective storms start in the western part of the region. Also, as noted earlier, the preferred daytime operation of aircraft for storm exploration is most favored in the western part of the severe storm region, as is also, for example, cloud photo- gramme try . (6) Such a concentration would also furnish the most complete history of air-mass modification (due to insolation and eradiation, advective influences, and vertical motions associated with synoptic disturbances and sloping terrain) during its primarily meridional flow. With a southward extension of special soundings to the Gulf Coast, this would include a variety of processes ranging from influences of stratus decks typical inland from the coast, to the broken low cloud evolving to deep convection farther north, as the inversion is weakened. A suggested network layout is shown in Fig. 1. This includes, in addition to the inner mesoscale array and added stations interspersed 205 among the existing NWS rawinsonde stations over the Great Plains and eastward, some enhancement of the density of upper-air stations farther west. The rationale for this was outlined in Section 2.1. Desired measurement precisions for mesoscale budget calculations were specified in Section 2.3.1. In the outer, less dense array of stations, wind accuracies of the operational rawinsonde system would be tolerable. To secure compatibility of soundings within the inner core network of mesoscale stations, it would be essential to arrange, for the duration of the experiment, for the regular rawinsondes (specifically at Oklahoma City and Dodge City) to be replaced by the METRAC system. It would be desirable, for an extended period of SESAME operations, for the routine sounding frequency at the regular NWS stations to be increased to four per day, with three-hourly soundings on days when severe weather is expected. By relating synoptic-scale structures at more frequent intervals (than now afforded by 12-hourly sounding) to severe weather occurrences, researchers could more adequately test forecasting parameters and rules in operational use. Such studies could also give an idea of the upper bound to be expected from incorporating fine-mesh forecasts in severe weather predictions. Soundings at an enhanced frequency would be valuable for assessing the accuracy of model results at intervals during a 12-24 hour forecast period, and for tests in which time varying boundary conditions (for the nested grids) can be based upon observed data. Supplementary (Lagrangian) observations by constant-level balloons (set to drift within the PBL) would be valuable in delineating air trajectories including the effects of inertial oscillations, and 206 (if provided with pressure sensors) gravity waves or updrafts and downdrafts on chance encounter with these phenomena. 207 hL nAl50^l.P|4^ : ■VKJOfri'g 4 ti^^^l^'^ PROPOSED UPPER AIR NETWORK FOR SESAME ' The largest square delineated by a dotted line encloses the KWS stations that would siake special upper-air measurements for the SESAME data base. The middle-sized square, which is about 1000 km on a side, defines the domain for initializing mesosc« prediction models. The average station spacing is about 200 km when only the regula: NWS rawinsonde stations and the special rawinsonde stations at Weather Service Offic (WSO's; solid circles) are considered. The smallest square, about 500 km on a side, encloses a net of upper-air stations spaced 100 km apart and a net of surface static (not shown) spaced 50 len apart. (Those upper-air stations denoted by small hollow circles would require special siting.) Data from within the 500 Ian square would le used mainly for developing statistical prediction schemes, studying storm-environ nent interactions, verifying mososcale model forecasts, developing cumulus param- eterization schemes, etc. Data from the three rawinsonde^stations in Central Oklahoma, spaced about 40 km apart, would be used to test the accuracy and spatial resolution of satellite soundings. FIGURE 1 208 5.4. REPORT OF SESAME WORKING GROUP IV ON CLOUD MICROPHYSICS AND ELECTRIFICATION March 1976 GROUP MEMBERS: R. L. Lavoie, Chairman R. Arnold R. Braham M. Brook J. Dye B. Foote E. Kessler W. Taylor B. Vonnegut 209 It is the consensus of our working group that microphysical variations do not play a central role in triggering the development of a severe storm; however, we have insufficient knowledge about electrical processes to make judgments about the possible role of electricity in these storms. It is probable that microphysical factors play some role in modulating the intensity of severe storms. For example, the natural variations in aerosol populations may influence hail size and occurrence, the intensity of downdraft-induced surface winds, or the frequency of lightning strokes. However, it is our conclusion that microphysical aspects probably control only a small fraction of the total variance of storm severity. Larger scale static and dynamic factors certainly dominate. While it is important to determine how precipitation and hail develop in severe storms, other projects, such as NHRE, HIPLEX, and DUSTORM, are examining microphysical development and the role of microphysics in modu- lating the intensity of cumulonimbus. In view of the "critical mass" of manpower and equipment needed to undertake such detailed microphysical obser- vations, it does not seem prudent to duplicate this effort within the limited resources of SESAME. However, once the SESAME project area and plans are well defined, it would seem worthwhile to encourage the involvement of these other observational systems (particularly aircraft) for a two to three week period. This would provide at least a limited amount of microphysical data to initialize and verify numerical cloud models and to test algorithms used in connection with remote probing devices planned for SESAME. Probably the only routine microphysical observations that can be made without great cost and complexity, but still could provide useful information, would be the measurement of cloud condensation nuclei (CCN) and ice nuclei. Techniques for CCN measurements are well established, and representative d ata enable one to estimate cloud droplet spectra at cloud base. There is less agreement among specialists about techniques for measuring ice nuclei and about the diagnostic value of ice nucleus data. Nevertheless, ice nucleus measurements using any of the better-known methods would contribute to basic knowledge, and might prove to be valuable to other goals of SESAME at relatively little expenditure of funds and manpower. 210 Turning to electrification, we find that there are virtually no lightning data from the midwest storms which can be used to characterize storm severity or activity. Sferics measurements are available, but because of a lack of simultaneous lightning data, their origin and meaning are not clear. We propose a program of measurements to explore relationships between electrical activity and thunderstorm dynamics. As a first step, the measure- ments are aimed at devising a thunderstorm activity index based upon electrical parameters which correlate with variables such as damaging wind, large hail and tornado activity. To establish the activity index, we need to observe the following characteristics: A. Cloud to Ground Flashes 1. number of return strokes per flash; 2. total flash duration; 3. flashing rate; 4. polarity of electric field changes; 5. burst rate of electromagnetic impluses; 6. space time mapping of lightning streamers from high frequency electromagnetic radiation. B. Cloud flashes 1. field change structure; 2. total flash duration; 3. burst rate of electromagnetic impulses; 4. space time streamer mapping from electromagnetic radiation. The above enumerated characteristics are readily measured with equipment now available. The NOAA WPL has developed a burst rate detector which was operated for several years at 20 locations distributed throughout southern and central U.S. The space-time lightning mapper is a NOAA WPL instrument which will be employed at two locations within the Cape Canaveral area during the Thunderstorm Project-2 (TP-2) experiments in 1976 and following years. The electric field change instrument is a simple device which is readily available from lightning investigators at such institutions as Univ. of Arizona (E.P. Krider), New Mexico Tech (M. Brook) and Univ. of 211 Florida (M. Uman). Lightning discharge (VLF C-G strokes) locations can be determined in real time by means of automatic stations covering a range out to approximate! 250 km. These instruments have also been developed by WPL and may be desirable for thunderstorm tracking in the SESAME program, especially when used in combination with radar echo overlay. A second program more directly concerned with relating electrical activil to thunderstorm dynamics and precipitation growth would involve, in addition to the above measurements, a network of a minimum of twelve field change stations to determine the position and magnitude of charges involved in individual lightning flashes. The charge location, along with electro- magnetic space-time maps, would be superimposed on high resolution radar reflectivity profiles and Doppler wind fields. Possible interactions betweer electric fields and precipitation growth would be evaluated by noting the rate of change of reflectivity in the region of the lightning charge volume. 212 ADDITIONAL STATEMENT BY B. V0NNE6UT It is desirable that the following studies be made part of the activities of SESAME. Ground and Aircraft Photoelectric Observations of Lightning Data on the nature location and frequency of lightning strokes should be obtained by the use of simple directional photoelectric systems such as the one described by Griffiths and Vonnegut, Weather , 1975. This apparatus consisting of a photocell used in place of the microphone on a tape recorder provides a means for detecting and recording lightning even in clouds illumin- ated by bright sunlight. Such equipment should be used by both ground and airplane observers to characterize electrical activity in intense sferics. Satellite Observations of Lightning Activity It is desirable that instrumentation be installed on a satellite to provide information both day and night on the frequency of lightning dis- charges from severe storms. Observation from the DAP satellite showing highly unusual electrical activity from the storm system that procuded the "Jumbo" tornado outbreak of April 3, 1974 indicates that observations of this kind would identify the development and location of extremely intense storm activity. A variation of the photoelectic apparatus now being used for ground and airplane lightning observations would be suitable method for detecting lightning activity. Balloon Borne Time Lapse Photography of Storms from High Altitude Balloons Information on the time scale of convective processes in storm can be obtained by time lapse photography from balloons flown over the storm at altitudes of the order of 30 km or higher. Atkinson and Vonnegut have shown that this simple technique provides a unique view of the convective movements in the clouds below. Such observations made at night should provide new information on the relationship between the lightning that is occurring and the convective structure of the storm. 213 5.5. REPORT OF WORKING GROUP # 5 (TORNADO CYCLONE) THE PROBLEM OF TORNADO DEVELOPMENT 3-5 March 1976 Group Members: Joe Golden, Chairman Rodger Brown John McCarthy Ed Pearl Bill Shenk Bob Davies-Jones John Marwitz Charles Anderson 214 TORNADO CYCLONE GROUP I. INTRODUCTION The essence of the problem of tornado development is the structure and evolution of the tornado cyclone. It is agreed that we should limit measure- ments to the tornado cyclone scale, as opposed to tornado scale. We need to document the tornado with respect to the tornado cyclone. We urge that the first multiscale experiment be conducted over April, May and June. SESAME will require chase teams to document tornado cyclone features in real time and communicate these back to base and aircraft. These features should include wall cloud, and flanking line of congestus clouds. The scales of motion to be considered depend primarily upon our conception of the mesocyclone. Doppler data indicates a solid rotation core (single Doppler) of '\^5 km diameter, range of 3-10 km in the mesocyclones. Dual- Doppler radars detect circulation over a diameter of 20 km in some cases. Definition of mesocyclone: a closed cyclonic circulation relative to storm, with continuity in vertical depth (several km) and time (tens of minutes), average diameter of 10 km and range 1-20 km. Typical peak tangential windspeeds as determined by Doppler radar are 15-25 m/sec. It may be detected in several ways: Doppler radars (1, 2, 3), surface V, p, possibly satellite cloud patterns, or visual documentation of rotating lowered cloudbase. Tornado cyclones (containing one ore more tornadoes) are here considered to be a subset of mesocyclones. We must be cautious about using hook- shaped PPI radar echoes as necessary and sufficient identifiers of mesocyclones II. FUNDAMENTAL QUESTIONS Tornado Cyclone Group It should be understood that a few of the questions outlined by this group are currently ongoing pursuits, whose solution may be forthcoming prior to the start of the fir$t multiscale SESAME experiment. In the following (Part II), * = work currently ongoing, but definitive answers not expected prior to first multiscale experiment, = answers expected prior to first multiscale experiment. No symbol = full SESAME research period required. 215 Mesocyclone Key Questions (1)* What is the source of circulation? (Storm and outside air and lower boundary. a. In what regions of the initial cloud or thunderstorm is rotation first evident? (2)* What is the life-cycle and 3-D structure of mesocyclone? a. By what mechanisms does mesocyclone intensity or extend to the surface? (3)* How does the pre-thunderstorm environment change with space and time leading to tornadic storm development? (4) What is the relationship between the dryline bulge (and other sub- synoptic disturbances, including subsynoptic lows), and mesocyclones? (5)* Why do some thunderstorms produce mesocyclones while others do not? (6) What determines the ultimate size and intensity of a mesocyclone? a.* Can tornadoes form without a detectable mesocyclone? (7) What is the relationship between mesocyclones and severe weather phenomena, especially hail and wind? (8)* What is the relationship in space and time between the tornado, meso- cyclone, and gust front (include documentation of tornado morphology, times, locations). a.* What is the structure of the flanking line(s)? What is its role in tornadogenesis? b.* How is the lowered, rotating cloudbase ("wall cloud") related to the mesocyclone? c. What is the relation between the hook echo and the mesocyclone? (9) What role do new cell developments in the flanking line play in the maintenance and structure of the storm system? (10)* What is the relation between cloud growth (overshooting and collapse, horizontal changes) and tornado development? (Include relation to tornado life-cycle). (11)* What is the role of merging and splitting thunderstorms in tornado- genesis? 216 The observational systems needed to answer each of the above questions on tornado cyclones are given in Table 1. Note that we categorize the expected impact of each observing tool on a particular question according to whether it is mandatory, needed or useful (but not required). It should be apparent that questions # 1-3 are most fundamental in nature and place additional requirements on the horizontal, vertical and time resolution of the observing systems, especially the surface and sonde networks. The requirements for each presently conceived observing system are listed in Table 2. We recognize that some of these system requirements may have to be reconciled with conflicting needs from the other SESAME Working Groups and budget constraints. An outline of the secondary research plan for Tornado Scale Measurements in SESAME is attached. Also attached is a summary of recommended atmospheric electrical observations on tornadoes by Brook and Vonnegut. 217 9> 00 iO in iU 00 5ci X x:i xl x| x| X xl x! X X| X X xl xl X x| x| xj X xl X X X| X X X X X x! xl xl X x| X x| XXX f. a> o 0) B co evJ xl X x| xl X X X X xl xl xl xl X| X X x| X X x| X xl x| x| X X x| x| X x| X xl • X X X X X X X x] • X X| X X X x| xl • X X X X • X X X xl x| X x| x| X ^— XJ en o c L. 0.-0 c (/> 3 •r* o JZ 1/1 ■*-> Q) o ■tJ *r— X> fM> 3 It* «4- Ol -o C/l c 3 (0 >, M c 1— o fO o *P* r^ •«-> ^M c ^» O «« *J X X\ X 01 c--^ lA 10 >i •a i. 1— J= lO Ql O. J.-0 1- «o 9) la f- i- ^ »- QC cC O) %. Q. o o o. 0> T3 +J ^ o u V O •M o c 4-> x: OJ lO C Q. z Vt- g . r— ^- Oi Q.T- 3 t— u •r- CJ «f- L. U «1 •M > «»- 4J C <♦- .— U (O VI •<- i. 3 3 x: C^-' 3 s:co o HH (O OJ ■o e o ■ to > r* 01 Xk Jk£ T> •«-* I_ §g •4- § o 1. 4-1 (O O) u CJ c 0) 1. Z o — 4-> •T" •r- "O «^ < t. >» ■M C fmm « 0) 9) (/) 01 ^— 0) ■M v» > 3 a. (U c •M to u O (U ♦J o O f 3 U TJ . TD r— +- J- C O) - (O > E XI Mt •0 o>II^ J3 ♦'^ 01 0> C J3 t- nj 1 9) • CL "O CL ■0 •tJ^ in 3 10 H- c J= 0) f— t> o> t 0) M •r- "r- 01 V» M C 01 1- t/> ^ ^ ■=s "5 1. C lO 6 "O u Or- •»- M 3 T- 1 +* i • 0) •M 0) &. r- Crt c E S> U 41 •^ flj-r- > 1.1 • t. 3 1— ■»-* ""^ lO 01 to 01 "ox: M Ql o-o k +J x: ^ «o^^ ex tn m •r— «rt ^■^ -0 J- Ul 3 «4- ■M C 3: 0; £ 2: 0) ««- «/> E t. 4J Q r^ •^ 3 3 0) C ^ ^ cr> t- r"> • •— OJ 1- ^ • 0) CL U ^■■^ >- «4- >» u "O E • 3 0.— c c x: o; c ^« <« •r- CL (/> +J ^*^ t- ro 01 t. CJ> « LU c I. 01 +J trt c cn s 0) • 3 ■•-> 1— c 0. •0 10 4J 0: C3^ c — > IB ^.^ 3 •r- •r- X> 10 ^ J- «♦- ■M O. A x: U. 3 J3 •I- •»-> 0; 0; „_^ 1— 3 +J 10 J- (O 0. t/) 3 ^ c 3 _j E u o> h- ** ■M •^ •♦-> !a Q. -0 sz •r" c z to (/) oO «/) to •^ 3 «*- OJ •*>> +j •T" Ul •^ s- •r- s- T3 s: V> T3 3 .- Q w 1 OJ 3 0> 3 Ul i- <0 •f- £ 1— +J > ^— 0£ 3 o> ■- OH- Q. E OJ «t- u ^ CM ►—1 c W) A i- E «f- 0) c 01 =5 x: t- «/> i- ro +J QJ £ •r" •M Ul cy •M (rt T3 a. 01 ^>w^ C t— 4J C ^ Ul vo +-> ■0 c 01 < •M >♦- -0 c ^. >_«l* •<- C «* O) M (/) 2 I— +J o; s- 3 -0 • ■0 l/> tt> -r- ^— 0) ^M 3 in Q. la E a> x: •p- (o 4J > C o. i. to ^.^M* E i. •0 ■0 »■% •r- (/I o> •r- 3 «♦- 1- J- E (U OJ C 01 t ■4-> +j --^ N to CM oc t. z U c^ -0 0^ «t OJ 3 (O E to D1 ■v^ •^ 00 ■M ■0 «-' to CO ■M 1 v C 4J i. «4- C M i. E C M (U 01 c * (/) 4J O) J= 1 "D re «I 4-> Jii ••- , • lO "O ■•-> E 1 (A 01 •t- W» s. to >— ^ ''^^ "»Si^ > x: •r- E E i. /\^ 4-> >, •r- OJ J- a. sl V) Q. 4J s- + Q. s. O) . V) (O -Q t/t Q. UJ tr> <4- s- r-M "3 +1 +J cvj t^^ >, 10 E •M E 3 > u > q; ro c\j -o 0) f- • « c s- (O >>**- +-> 10 M 'r- J- Q. U ^— v> • c i~ , 1/1 -r- 0) E I0 *r- +j U to ^^ ^>^ ^ O) •1— q: c r^ U CL3 sz T3 -a ir> -r- > 4J OJ JC (O to E s- ■u •r- (Q •r- t- •r- a. OJ 3 •«-> u n ^ +J •f— s- — fm^ •r- (O > > ^ rg **- •!- • • F~* •«-» 0) t. ^M r^ Q. ir> ■M cr 10 E -M r— CL «/> 10 E , 4J •« •r- E 0. /vl h- q: a: r— Q OC XI a. (/) =) 2: Q C_) eC a. a: B •• 0. If Q « 0. s • • c «5 M «i 0) •• as ♦*<-» 0» ♦» •r- Id • • V «i ^M ■M M 0. •• « u ^« (/I S «^ k 10 0) 4J «»- g t •M 0) e »— a too 3 JS •f- 9 (/) ^-^ £ u < t/l 219 <0 > LO ^ C\J -^ i- O Q) O > f— O ^"■■^ 1 O) o «/i r— VO — (T3 .'-^ E O C_) ? c «/> ca: •r- IT) ta I a: X E • 4-> ca f- < O -^ LU O •a 2: -o OO 4- E •r- 1— O T3 ^ >> s- 1 0) I— cno >, N in x> cn a •t— r— , J3 (U •*-> s- •.- V O +J (0 3 Oi J- «/) u •1- X Q. (U 4-> o Q < * — '^ <een used subjec- tively in many ways and objectively in many forecasting aids. In NWP, climaiological statistics are used in ob- taining the weigliting functions in optimal inter|)oiation objective analysis and as the first guess in some iterative objective analysis tecliniques (Mcl'hcrson, 1975). In the tropics, climatology is an important part of statistical forecast techniques and is even used as the southern boundary condition in dynamical prediction models (Kesel and Winninghoff, 1972). In very short-range terminal forecasting, climaiological records are the basis of the widely used station conditional probability tables. Because of its high information content, climatology can be used as a reference level against which skill can be measured in forecast verification schemes. In mesoscale weatiier forecasting, mesoclimatology could be used in several rather different ways: 1) as an aid in specifying the first-guess field to in- corporate local geographic effects between observ- ing sites in the objective analysis of moisture and boundary layer structures; 2) ill the refinement of larger-scale forecasts to re- flect local geographic effects; 3) in development of knowledge that would permit extension of mesoscale forecasting to regions of sparse data from studies in regions having similar geography and adequate observations in order to verify models and their parameterization schemes; 4) in regional air quality simulation modeling for studying pollution abatement strategies. Several types of mesoclimatic information are needed, including: 1) local wind flows for model initialization and tra- jectories for air quality studies; 2) local variations in boundary layer thermal and moisture structure and surface fluxes for model initiali/iition and for parameterization of subgrid scale proces.ses; 3) local anomalies in surface temperature and winds as well as cloudiness and precipitation to aid in forecasting these variations. Satellite and radar data are ideal sources of meso- climatic data because of the completeness of their cover- iige. Satellite data are i)cing archived on the mesoscale as described by Booth and Taylor (1969) and discussed by Barrett (1974, p. 50). Radar data are l)eing coded and archived using the manually digitized radar scheme of Moore el id. (1974). .Although these encoding techniques require comi)romises and loss of resolution, they are es- sential for accumulating mesoclimatic information in the face of an otherwise overwhelming data load. As automatic radar digitizers are installed at more sites, these data will become available for mesoclimatic studies (McGrew, 1972; Greene, 1975). It is less clear how soon mesoscale numerical models will become effective tools for generating mesoclimatic information through simulation studies. On the large scale, general circulation models have a role in studies of large-scale climatic processes, but it has been difficult to develop models whose climatologies compare favor- ably with observed climatologies in very great detail (U.S. Committee for GARP. 1975, Appendix B). r<) the extent that mesoscale NWP models make accurate forecasts, they will obviate the need for meso- 237 Bulletin American Meteorological Society 683 climatology for forecasting purposes. This is clearly the goal of the mesoscale modeler. Yet, experience with large-scale NVVP has demonstrated that predictions can always be improved by statistical modification of model output wherein tiie dynamic prediction is treated as a source of predictors and the weather prediction is the predictand (Leith, 1974). Otiier predictors are combined witii the dynamic model predictors through model out- put statistics (MOS) (Klein and Glahn, 1974) in calcu- lating the predictand. Of course, the statistical relation is a form of climatological information. The iiigh cost of explicit mesoscale NWP increases the potential value of mesoscale climatic simulation studies. If a limited number of idealized cases could be com- puted ahead of time and the solutions referred to when the conditions arise, then the computer costs could be amortized over many forecasts. ,A map-type classifica- tion scheme, such as tliat used by Paegle (1974), could i)e used to identify the large-scale conditions under which a particular mesoscale simulation run should be used. Such tcchnicpies have never lieen competitive with ex- plicit NWP for 1- to 3-day large-scale forecasts. It is more likely that this technique will be practical in meso- scale air quality prediction (Hosier, 1975) than in meso- scale weather prediction iiecause fewer meteorological variables and, iience, fewer very different situations need be considered for air quality prediction than for weather prediction. There will ije a substantial proliabilistic component to predicting mesoscale phenomena, particularly in relation to convective activity. It will be useful to determine the frequency of mesoscale events within different large-scale storm systems. Such statistics, compiled from a set of 24 occluded frontal systems, are shown by Kreitzberg (1964, p. Gfi). I-'or important large-scale systems or map types, the statistical distribution of mesoscale phenomena should be devclojied along with the statistical distribu- tion of convective .scale phenomena, such as maximum snowfall or rainfall rates, wind gusts, or severe weather probability (for example, sec Novlan and Gray, 1974). It is also important to determine if any particular sizes of mesoscale rainbands predominate as a function of the large-scale structure (for exaniple. see .\kiyama, 1974). Empirical climatology by its very nature involves con- siderable data collection, archiving, and processing. The revolution in minicomputer technology is bringing computer power to the field forecaster with the Auto- mation of Field Operations and Services (.AFOS) system (Johnson and Giraytys, 1975). This power could be used for mcsoclimatology if appropriate hardware, software, and training were available at the field sites. The stor- age, accessing, and manipulation of an extensive meso- climatic data base at a Weather Service Forecast Office (WSFO) are not simple tasks. Neither is it easy for the National Climatic Center to collect, evaluate, and utilize all the mesoscale data available from the innumerable different special observing sites and networks around the country. Obviously, the more the WSFOs deal with mesoclimatology, the more rapidly they will utilize that information in mesoscale forecasting. 5. Special-purpose simple models and empirical relations For special mesoscale phenomena, simple models or forecast schemes will provide necessary additions to dy- namic mesoscale predictions. Experience has clearly shown that large-scale NWP, statistical forecasting, and subjective forecasting are .complimentary methods, not alternate schemes. The point is simply that. NWP can be extended down in scale from a 200 km grid size over the hemisphere to a 20 km grid size over a region 2000 km on a side by 1980. The subsequent refinements to convert the NWP product into improved and more specific forecasts will still be necessary and possible. The NWP forecasts on a 20 km mesh will be made at a central computer facility, and the output fields could be available to the field forecast AFOS computers for each hour during the forecast period. The field forecaster will always have more recent and more detailed information than is incorporated in the dynamic forecast. He can u.se this additional informa- tion and mesoclimatology with MOS, conditional prob- ability, or simple linear models to improve the 1-6 h forecasts and generate the specific forecast products needed at that location and time. The expansion of the VHF continuous weather broadcasts along with .\FOS will alleviate the weather forecast dissemination bottleneck; these improvements, in turn, will amplify the demand for rapidly produced special-purpose fore- casts. .\ pilot |)rogram to test several components of a future very short-range weather forecast system was conducted during the summer of 1975. This Chesapeake Region AFOS/Saiellite Weather Broadcast (CRAB) test (Giraytys ct al., 1976) included a preliminary shakedown of sev- eral components of the system described in this paper but tiealt with time ])eriods shorter than those for which a mesoscale NWP model woidd be appropriate. Iherc are numerous examples of simple models and cm|)irical relations that have been or could be developed including those lor gravity and lee waves, cumulus con- vection, orographic channeling, flash flooding, wind gusts, and frost. .V good deal of careful training is re- (juired to use the models appropriately if the benefits that could be gained from forecaster experience and in- genuity are not to be forfeited by elimination of his judgment (Kreitzberg, 1969). 6. Conclusions 1 echuological advances in satellite sensors, large and small computers, as well as data and forecast communica- tion systems combined with the rapid advances in large- scale observing and NWP systems will continue, or ac- celerate, the rate of change of the whole weather forecast system. Dynamic NWP will handle progressively smaller scales for shorter periods over smaller domains. The large 238 684 Vol. 57, No. 6, June 1976 mesoscale features in the moisture and motion fields detected by the satellites will be incorporated in the initial conditions in the NWP models. The large meso- scale fields output by the models will be used as the foundation upon which small mesoscale satellite and radar data will be interpreted. Mesoclimatology will advance in support of mesoscale forecasting and as a ' result of understanding gained from mesoscale numerical simulation studies. Improvements in communications of both data and forecasts and the availability of mini- computers to the field forecasters will greatly increase the needs and capabilities for diverse special-purpose simple models and empirical relations. This revolution will be hastened or retarded by the financial resources committed to it and the skill with which it is managed, but the developments will occur. All the pieces of the forecast system do exist in some stage of development and will most probably evolve into an operational system within 15 years or so if no special effort is made to assemble the system earlier. If the effort were begun immediately, the components of the fore- cast system could be assembled within two years and well developed within five years. A rash of special meetings have been convened re- cently that dealt with these aspects of mesometeorology, including: National Center for Atmospheric Research Colloquium (Shapiro, 1974); SESAME Opening Meeting (Lilly, 1975); Third Symposium on Meteorological Ob- servations and Instrumentation (Giraytys, 1975); First Conference on Regional and Mesoscale Modeling, Analy- sis, and Prediction (Randerson, 1975); and Workshop on Satellite Studies of Severe Storms, June 24-26, 1975, El Paso, Tex. These meetings were in addition to the normal conferences on air pollution, severe weather, cloud physics, and weather modification. The revised Project Development Plan for tlie Severe Storms and Mesoscale Experiments (SESAME) is being reviewed. Plans are being formulated for a special severe storm satellite STORMS.AI" for mesoscale coverage in tlie early 1980s. A large component of the scientific and observational support for understanding and model- ing mesoscale phenomena and assessing their predictabil- ity will be in association with SES.AME. How well this new mesoscale forecasting system will work is not known, since mesoscale pre(lictai)ility re- mains to be evaluated. Undoulnedly, the advances will be substantial, and sul)stantial research will occur in support of and in response to these advances. Further impetus to mesometeorology is l)eing provided by recog- nition of new regional problems associated with energy production and distribution, including environmental implications, and recognition that weather modification, particularly precipitation augmentation, must be under- stood in terms of mesoscale interactions. The above developments provide ample evidence that after 18 years of big mesometeorology plans and little systematic implementation, we are on the verge of dramatic advances. It should be clear, now as never before, that the goals of the Department of Defense, the Energy Research and Development Administration, NASA, NOAA, the National Science Foundation, the Department of the Interior, the Department of Trans- portation, and the sponsors of these agencies — the tax- payers — all would be well served by a mesoscale forecast system based largely on nested numerical models and satellite observations. Acknowledgments. This report was prepared with finan- cial assistance from the Scientific Services Program, Battelle Columbus Laboratories, Durham, for the U.S. .Army Electronics Command, Atmospheric Sciences Lab- oratory, White Sands Missile Range, N. Mex. These ideas are based, in large part, on research performed under the Atmospheric Sensing and Prediction Project at Drexel University with support from DOD, NSF, ERD.A, and NASA; the respective contract numbers are F19628-69-C-0092, GA-34()93, E (1 ]-l)-2360, and NAS8- 31235. References Akiyama, T., 1974: Mesoscale organization of cumulus con- vection in large-scale rainbands in the Baiu season. }. Meteor. Soc. of Japan, 52, 4-18-451. Allison, L. J., A. Arking, \V. R. Bandeen, \V. E. Shenk, and R. Wcxler, 1975: Meteorological satellite accomplishments. Rev. Geophys. Space Pliys., 13, 737-745. Barrett, E. C, 197-?: Daily and monthly rainfall estimates from weather satellite data. Moti. Wea. Ren., 101, 215-222. , 197-1: Climalology from Satellite.'!. London, Methuen and Co. Ltd., 418 pp. Booth, A. L., and V. R. Taylor, 1969: Mesoscale archive and computer products of digitized video data from ESSA satellites, flull. Amer. Meteor. Soc, 50, 431-438. Fujita, T. T., E. W. Pearl, and W. E. .Shenk, 1975: -Satellite- tracked cumulus velocities. ;. Appl. Meteor., 14, 407-413. Giraytys, J., 1975: Ihird symposium on meteorological ob- servations and instrumentation. Hull. Amer. Meteor. Soc, 56, 981-991. , R. Derouin, L. Mcrritt. and V. Bare, 1976: The report on CRAB-75. Syst. Develop. Office, NOAA/NWS. Silver Spring, Md., 55 pp. Goff, R. C., and H. R. Hudson, 1072: The thermal structure of the lowest half kilometer in central Oklahoma. Decem- ber 9, 1966-May 31. 1967. NOAA Tech. Memo., ERL NSSL- ,">8, 53 pp. (Available fion\ National Technical Information Service (NTIS), Springfield. Va.; refer to COM-72-11281.) (;rcene, 1). R.. 1975: Hydrological application of digital radar (lata. Preprints of Papers, 16th Radar Meteorology Cottferc/ice (Houston). Boston. .Amer. Meteor. Soc. 353-360. Hosier. C. R.. 1975: The meteorology program of the En- vironmental Protection Agency. Hull. Amer. Meteor. Soc, 56, 1261-1270. Johnson. R. E.. and J. Giraytys, 1975: The AFOS experi- mental system. UMO Hull., 24, 86-92. Kasaliara, .\.. 1971: Various vertical coordinate systems used for inimcrital weather prediction. Mon. Ilea. Rev., 102, .'•)09-522. Kescl. 1'. G., and K. J. Winninghoff, 1972: The Fleet Numeri- cal WeatlRi Central operational primitive equation mo- R and thus that R = < yx*""<^xx* > Here x and y being anomalies their ensemble means vanish, <^x ^ = /y> = 0; 3on asterisk indicates a transpose so that x* is a row vector; and <'yx*.> and <''xx*>are covariance matrices of which only •<^xx*> need be square. 259 12 In practice, the infinite ensemble averages indicated by the brackets ^ ^ are not known but niust be estimated from a finite sample of many cases of which N, say, are effectively independent. Theoretical analysis of the effect of sampling errors on the utility of a regression equation is still incomplete, but it indicates that sampling errors increase the fractional variance of estimation of a predictand by an amount (k+l)/N where k is the number of predictors used. Thus unless a particular predictor leads to a real incremental reduction in fractional variance by an amount 1/N it is not worth using. For meteorological time series an effectively independent sample occurs about once every three to eight days depending on geographical location (Madden, 1976). The regression equations for MOS methods are developed from a data base of as many past forecasts as are available for a particular numerical model. Sampling errors associated with such a finite data base seem to limit the number of significant predictors for each predictand to be of order ten, and stepwise screening procedures have been developed to select a signifi^:^nt "group of predictors. In thir screening process, model output predictors must compete with other potential predictors taken from any observations available before the time that the forecast is made. It is important to note that the regression equations provide automatically for at least a partial cor- rection of those systematic model errors that have been revealed empirically in the available past forecasts. An important aspect of regression methods is that chey. can provide forecast information in probabilistic terms. Thus forecasts can be given of a best estimate of a meteorological variable together with an 260 13 associated measure of uncertainty such as the standard or root mean square error of the estimate, or they can be given in terms of the probability of occurrence of various weather events. It is this kind of probabilistic information that is of value as input to the decision- making process of users who wish to maximize their expected gain or to minimize their expected loss. A number of combined dynamical and statistical methods are currently being used by the National Weather Service (NWS) in the operational forecasts of severe weather up to 24 hours in advance. These utilize a number of different numerical models and choices of predictors. Forecasters have traditionally associated the occurrence of severe weather with moist hydrostatic instability of the atmosphere, and numerical indices quantifying potential instability are natural choices as predictors. The various indices uhat have been used operationally have been summarized by /JLaka et al. (1973) . Among these are the Showalter (1953) index which, was supplcmted in 1969 by the Lifted index (Galway, 1956; Stackpole, 1967), the K-index (George et al . , 1960) , the Alaka-Reap i.ndex (reap and Alaka, 1969) , the tornado like- lihood index (Willis, 1969), and the Total Totals and Sweat indices (Miller, 1972) . A new predictor is suggested by the work of Darkow and Livingston (1975) who have shown that the small-scale variation in the surface static energy field may also bear a relationship to the location and timing of severe storm activity. Operational forecasting procedures developed in the Techniques Development Laboratory (TDL) of the National Weather Service (NWS) combine several of these indices with predictors taken from three 261 14 operational numerical prediction models, namely, the Primitive Equation (PE) Model (Shuman and Hovermale, 1968), the Limited Area Fine-Mesh (LFM) Model (Howcraft, 1971)^ and the Trajectory (TJ) model (Reap, 1972). The Itodel Output Statistics (MOS) procedure uses a linear regression prediction equation based on a screening, selection, and weighting of the best of these available predictors (Reap, 1974; Reap and Foster, 1975). At present such 24-hour forecasts, which give probabilities of thunderstorms and conditional probabilities of severe thunderstorms, are produced at the National Meteorological Center (NMC) in Washington and transmitted once daily to the National Severe Storms Forecast Center (NSSFC) in Kansas City. A typical forecast is shown in Fig. 1. Another procedure, developed by David (1971, 1973) at NSSFC, uses multiple screening regression to derive equations for forecasting severe weather up to 50 hours in advance. The predictors in this case are synoptic rawinsonde and surface observations at the initial time and forecast values taken from the PE model. On a smaller scale, Charba (1974, 1975) has used multiple screening regression to develop an objective method that yields forecasts of severe weather from two to six hours in advance in square areas 150 km on a side. In the most recent version of the method, predictors are derived from observed surface atmospheric variables, manually digitized radar data, local climatic frequencies of severe weather, and basic variables as predicted by the LFM model. Predictor quantities computed from these data are appropriately positioned relative to the predictand areas and are "linearized" to enhance their correlations with the predictand. 262 15 During the spring and summer of 1974, experimental forecasts for 17-21 CST, produced by this scheme, were transmitted from NMC to NSSFC. In 197 5 forecasts were transmitted three times daily; these v^ere valid at 11-15 CST, 14-18 CST, and 17-21 CST. The sample forecast shown in Figure 2 is for the extensive tornado outbreak of 3 April 1974. In 1976 forecasts are being prepared by an improved scheme based on predictors obtained from manually digitized radar reports, LFM forecasts, and objectively analysed surface data. Recommendation; The Panel strongly recommends the continued development of combined statistical-dynamical methods of objective forecasting such as the method of Model Output Statistics. 263 16 Section 4. Regional Forecasting Models The special concern of this report is the more accurate prediction of small scale weather events for a forecast period out to 12 hours. To thi3 end it is natural to consider the benefits of numerical models with increased horizontal resolution. The general benefit of resolution in providing increased accuracy on all scales has been described already, but there is also the special benefit in being able to resolve smaller scales of motion that are expected to be more highly correlated with events of interest. The benefit can be measured by the greater value of the predictors provided by the model to the regression pro- cess leading specifically to more sharply defined probabilities of occurrence of small scale events. For this improved skill to be realized, it is neccessary, of course, that the finer resolution regional models describe the detailed behavior of the atmosphere more accurately. As with coarser models, this capability depends on the accuracy of the observation and analysis of the initial state and on the accuracy of the model in simulating the behavior of the atmosphere. The relative importance of an accurate initial analysis depends on the characteristic lifetimes of smaller scale structures in the atmosphere. According to theoretical scaling arguments based on a -3 power energy spectrxom and on direct observations, the characteristic lifetime of disturbances in a Lagrangian sense is of the order of a day and is not strongly scale dependent for scales down to about 50 km. At a fixed observing station or model gridpoint, however, the Eulerian lifetime of a disturbance is determined by advection processes and is 264 17 shorter in proportion to its scale. For short-range prediction models the principal requirement is for the advection Bad propagation of disturbances to be computed with an accuracy that matches that with which they can be observed. There is some evidence from the studies of the sensitivity of forecasting skill to model resolution (Puri and Bourke, 1974) that the nonlinear coupling between different scales of motion can lead in the course of a model forecast to more skill in the description of small- scale details than was present in the initial analysis. This may be especially true if known small-scale surface influences such as orography, albedo, and surface roughness are properly modeled. Most of the physical processes included in the finer resolution regional models are scaled-down versions of those included in the coarser planetary scale models. These include orographic and boundary layer effects, condensation and precipitation processes, and radiative heating and cooling. Experiments with regional models show that details of cumulus convective precipitation are particularly important. Many other model forecast sensitivity ejcperiments will be required to determine the relative importance of these various processes to the skill of short-range forecasts. Among the special problems of regional models is that of proper initialization. Erroneous gravity wave modes with associated vertical motion can trigger erroneous convective scale disturbances. This is an especially delicate problem since there is some evidence that real gravity wave modes may do just that, and one must try to remove a false effect without removing a real one. 265 18 Another special problem of regional models is that of embedding them in a larger scale flow. This has been done by choosing boundary values for a regional model from a concurrent forecast carried out with a coarser model. Some mathematical and numerical problems remain in the proper specification of such boundary conditions. It is theoretically desirable to permit the regional model results to influence in turn the coarser model, and such two-way interaction schemes are being investigated. Although spectral models permit in many ways a more natural separation of scales, little work has been done on the problem of embedding a regional spectral model into one treating larger scales. Recommendation: The Panel strongly recommends the continued development of mesoscale regional models by a number of research groups and urges their use for predictability and observing system simulation experi - ments of the sort that have been carried out with planetary scale models . Parameterization and Empirica] Correction In addition to the explicitly computed variables in an atmospheric model, the real atmosphere is influenced by many subgrid scale variables that describe scales of motion or physical processes that are too small to be resolved. Even the explicit variables may be influencing the atmosphere through processes which are not properly described in the model. A central problem in model develop- ment has been the devising of parameterization procedures to estimate the average influence of hidden variables and processes on the evolu- tion of the explicitly computed variables. Most efforts to solve this problem have examined separately different physical processes such as convection, boundary layer turbulence, ?66 19 or radiative heat transfer. An alternate approach is to seek an empirically detenr^ined modification of the dynamical equations of the model in order to account for all the hidden effects which are causing model forecast errors. The first-order modification involves the addition of forcing terms to the model dynamical equations in order to counteract any tendency for the mean field of an ensemble of forecasts to drift away from the observed climate mean field. Such forcing terms preserve the first moment properties of the atmosphere in the forecast model and can be interpreted as a parameterization of the many fixed or slowly changing surface forcing influences which determine the mean climate. In this way, for example, the surface temperature averaged over many forecasts would reproduce the observed mean diurnal temperature cycle. The second-order modification involves the addition of terms linear in the model variables which are the regression estimates of the dis- crepanci 3S between the observed and computed rates of change. Such linear terms can be shown theoretically to preserve in the forecast model the zero time lag second-moment properties of the atmosphere such as the distribution of variance over different scales. They can be interpreted as the best empirical linear fit to the parameterization of hidden effects in terms of model variables. The practical determination of a second-order modification is subject to the sampling error limi- tations of any regression estimate that were described in Section 3. It is necessary to find by a judicious screening process the few most significant predictors for each predictand, and it is not yet clear how 267 20 effectively this can be done for existing forecasting models. It seems likely, however, that since correlations are local the choice of predictors will be more effective in a gridpoint rather than in a spectral representation of model variables. Higher-order modifica- tion is probably not feasible with present models , but it is not as important for short-range as for extended-range iorecasts. Boundary- Layer Models Processes in the planetary boundary layer are known to have an important effect on the occurrence and severity of convective storms. And, of course, the values of meteorological variables in the boundary layer are those that affect the most people. The planetary boundary layer is characterized by its turbulent behavior which is produced by thermal convection and mechanical shearing processes. It undergoes the most pronounced diurnal variation of an^ region of the lower atmosphere The speci=>.l problems of devising numerical models of the plane- tary boundary layer include the detailed balancing of solar and ter- restrial radiative heating and cooling effects, the estimation of the divergence of vertical turbulent fluxes of heat, moisture, and momentum, and the proper specification of large-scale forcing influences. Much progress has been made in recent years in the development of such models that have succeeded in simulating the diurnal change in boundary- layer properties in agreement with data from a number of experiments carried out in selected areas. The most complex of the boundary- layer models (e.g. , Deardorff , 1972) compute rather detailed three-dimensional flow, temperature, and 268 21 moisture fields and would be far too coinputationally expensive to cover the domain cf interest for regional forecasts. But the complex models are being used to test simpler schemes for making boundary- layer forecasts which can be included in regional forecasting models. Such schemes produce predictions of the mean value of wind, temperature, and humidity in the boundary layer within the context of a regional- scale moJel. A boundary-layer model is being developed by TDL for operational implementation at NMC which will contain 10 levels. Pre- dictions from the model should be valuable in their own right and as input to MOS forecasts of thunderstorm probability. 269 22 Section 5. Data Requirements In stating data requirements for short-range prediction.- it is necessary to take account of the size and nature of the phenomena to be predicted, the technological and economic feasibility of the required observing systems, and the purpose for which the data are gatliered, whether for operational use or research. Ideally, one would like to observe all pertinent atmospheric variables with sufficient density, frequency, and accuracy to resolve completely the phenomena under con- sideration. For many of the mesoscale systems of importance in short- range prediction, observations with a horizontal spacing of 10 km or less are required to achieve this objective. Even if the technological means could be found to obtain meaningful descriptions of atmospheric fields on such a small scale, the costs of maintaining a suitable net- work of regular observations covering a substantial geographic area v;ould be prohibitive. Clearly, any practical statement of data require- ments for short-range prediction must recognize the severe technological and ecoiiomic constraints involved and must differentiate between oper- ational and research goals. This report focuses on the meso- scale patterns connected with convective activity and severe storms, which may generate thunderstorms, tornadoes, hail, strong winds, and heavy rains. The basic purpose of this report is to suggest means for improved prediction of the mesoscale systems, in both the short term, through application of presently available technology, and in the long term through research that will increase knowledge and understanding of the mesoscale systems and their interactions with larger- and smaller- 270 23 scale systems. It thus seems appropriate to consider observational requirements from the point of view of both short- and long-term objectives. (a) Data requirements for immediate improvement in short- range prediction of mesoscale convective systems Short-term advances in prediction can be pursued by follow- ing three promising routes: (1) better identification and tracking of the mesoscale features, (2) development of better statistical relation- ships between severe storm occurrence and synoptic-scale fields, and (3) development of improved limited area, fine -mesh numerical predic- tion models for more accurate prediction of the larger-scale environ- mental changes that determine and control the convective activity. The associated data requirements are as follows: (i) Identification and tracking of the mesoscale features Weather surveillance radars (10 cm) and geostationary satellites offer the best tools for monitoring convective activity. Recommeiidation: The Panel recommends that automatically digitized radar data be provided with complete coverage of the areas of the United States susceptible to severe storm development . The radar display should indicate echo depth and intensity, and a search continued for even better radar indicators of severe storm activity. Surveillance should be continuous during periods of severe weather. Recommendation: The Panel recommends that geostationary satellite digitized images, both visible and infrared, be provided at half -hourly intervals, with the highest possible resolution . The resolution should be about one mile for the visible and five miles for the infrared. Infrared data should give equivalent black body or cloud-top temperatures, 271 25 convective outbreaks. However/ it would seem premature to recommend additions to the upper- air observing network until more is known about the performance of the LFM's in convective situations and also until more is known of how the performance is affected by the grid spacing. Simply put, no sound guidelines exist at this time for recommending a finer observational grid for routine use. Until suitable experiments are run to yield the required information on the effect of grid spacing on forecast accuracy, it seems advisable to maintain the net- works in their present form. Plans are currently under way at TDL to run an operational boundary- layer model in conjianction with NMC's LFM, the latter being used to provide lateral and upper boundary conditions for the former. Data requirements are similar to those of the larger model except that •more horizontal resolution is needed. Horizontal resolution may be aided by using output from the previous forecast. Initialization requires fairly detailed profiles of temperature, win^ and h\imidity from the surface to 2 km. Approximately 10 levels of data with conventional accuracy will suffice. Additional desirable initial data include: (1) rainfall from the previous day, (2) soil temperature, (3) cloud analysis, including not only clouds within the lowest several kilometers but also cloud cover (high, middle, and low fraction) above the boundary layer, (4) geostrophic winds, aind (5) tendencies of the wind vector, temperature and moisture for use in a variational initialization technique. Remote sensing of clouds and water vapor from satellites should provide much of this information. 272 26 Since much of the severe weather in the central United States occurs between 21Z and 03Z3 it is probable that upper air data at 18Z would greatly benefit short-range forecasting. Recommendation; The Panel recommends that consideration be given to the addition of routine rawinsonde observations at 18Z in the central United States during the severe storm season. Dynamical and statistical techniques would have to be developed to utilize these data, and at least two seasons' data will be needed to develop and test the techniques. This data collection would cost about $100K per month (22 sites, 30 soundings, $150 per sounding). These data could be used for: 1. A last-minute update for statistical prediction schemes valid in the 21Z - 05Z time interval. 2. An up-to-date three-dimensional synoptic -scale data base on which the midday and afternoon satellite and radar data can be superimposed for more complete interpolation and as ground truth for calibration of operational satellite data. 3. A test of the benefits of reinitializing the LFM and boundary -layer models with midday observations. 4. An improved synoptic -scale data base for use in analysis of severe storm field experiments of NOAA and NASA. (b) Data requirements for research in short-range prediction of mesoscale patterns The following data requirements are stated with specific research objectives in mind. These objectives are: (1) to describe more clearly the mesoscale thermodynamic and kinematic structures 273 27 associated with organized intense convective systems that produce severe weather including tornadoes. (2) To define the key atmospheric phenomena or processes responsible for organizing these mesoscale structures and determine their relationship to the larger scales. (3) To develop an improved capability for forecasting where and when these mesoscale convective patterns will form and the movement and structural changes of already existing patterns. To carry out these objectives requires the establishment of special data-gathering networks and the use of the most sophisticated instru- mentation available. To keep costs within boxonds, the data gathering must be limited both in time and space. A region of about 1 million 2 km in which severe storms have a high probability of occurrence, and in which special observational facilities already exist, and a time period of two or three months during the season of greatest activity would seem desirable for an initial experiment. It is expected that the data obtained would be used to test and improve fine-mesh numerical prediction models and to develop models of smaller convective features whose structure cannot be resolved with the fine-mesh grids but whose behavior can be related to resolvable larger-scale features. The data requirements are as follows: (i) Surface observations Wind, temperature, humidity, pressure and precipitation rate, cloud cover, and cloud base height are required with the usual accuracy and frequency at regular synoptic stations. Continuous recordings of the first five elements, with microbarographs used for 274 28 pressure, are needed in a special network of more closely spaced stations. Sonie 100 stations spaced about 25 km apart are desirable. (ii) Rawinsonde observations Supplementary rawinsonde measurements are needed 2 within the roughly 1 million km experimental area. The number added should be sufficient to reduce the spacing between the regular stations by a factor of two or more. A mesh size of about 200 km, requiring about 25 additional stations, is desirable. The rawinsondes will measure wind, temperature, pressure, and humidity with the usual accuracy. Releases should be made every three hours during disturbed weather or upon its approach. (iii) Indirect sensors High-resolution satellite photos, visual and infrared, and surface teniper^ure are required at half- hourly inteirvals. Also required are vertical temperature and moisture soundings and cloud wind measurements at hourly intervals. Effective black body or cloud-top temperatures are required in digital torm. Geostationary satellites will be capable of providing these data in the near future. Four types of radar data are required: (1) composite digitized echo maps that give a three-dimensional representation of the echoes, (2) dual Doppler radar measurements of horizontal wind fields, (3) dual wave length radar measurements of liquid water, and (4) FM-CW radar measurements of the fine structure of precipitation and the structure of gravity waves and other small-scale features of the clear air. 275 29 (iv) Boundary- layer observations The proposed special rawinsonde network would provide valuable boundary-layer data. Within a limited region, these data should be supplemented by acoustic soundings, which will give a record of boundary- layer depth, by measurements of fluxes of heat, moisture and momentum by gust-probe aircraft and by measurements of ground temperature made radiometrically from aircraft. Flux measurements from a central mast would also be useful. Solar and terrestrial radiation measurements are needed with sufficient accuracy to determine radiative flux divergences. (v) Terrain description Data are required on terrain height, large-scale surface roughness length, surface albedo, average soil conductivity, and heat capacity and soil moisture, (c) SESAME Program There is presently being planned within the Environmental Research Laboratories of NOAA the SESAME Scientific Program with the principal aim of "observing, experimentally predicting, and understanding the formation, evolution, and meteorological impact of severe convective storms". The major observational components of the SESAME program are two multiscale experiments planned for the springs of 1980 and 1982. T\ preliminary experiments are also planned; the first, a boundary -layer systems test in 1978; the second, a regional-scale experiment in the spring of 1979. In the regional-scale experiment, about 20 rawinsonde stations would be added over the central United States to reduce 276 30 spacing to about 200 km, as shown in Figure 3. The resulting data set is to be used primarily for the initialization and testing of regional models during periods of anticipated strong convection. The special merit of conducting the regional-scale experiment during 1979 lies in the existence during that time of the enhanced data network over midlatitude oceans provided by the First GARP Global Experiment (FGGE). The SESAI4E program provides the best opportunity to satisfy the data requirements for research in short-range prediction of mesoscale patterns described in Section 5(b). Therefore, the Panel strongly endorses the SESAME program and, in particular, the conduct of the regional rawinsonde experiment in the spring of 1979. 277 31 Section 6. Summary and Recommendations The most important steps that could be taken immediately to improve prediction of severe storms are (1) better detection and tracking of convective activity, (2) development of better s^-atistica; relationships between convective storms and the large mesoscale disturbances that produce them and (3) an improvement in the dynamica prediction of these large mesoscale disturbances. The Panel on Severe Storms has dealt fully with Item (1) in its report comprising Part II of this document. This Panel strongly endorses the recom- mendations of the Panel on Severe Storms relating to the importance of developing and applying the various new sensing schemes to the identification and tracking of convective activity and to the importance of designing an effective yearning system. An analysis of recent experience with numerical weather prediction models indicates that horizontal resolution is of great importance to the improvement of forecast skill. Recommendation 1: The Panel recommends, in light of the marked impiovement in skill of the limited-area fine-mesh model (LFM), that a study by NOAA be made of the potential benefits in providing greater computing power to MC for a further increase in model resolution (Section 2, page 10). The present use of statistical relationships in predicting regional probabilities of severe storm outbreaks has led to a considerable improvement in skill beyond that achievable by purely dynamical methods. 278 32 Recommendation 2; The Panel strongly recommends the continued development of combined statistical-dynamical methods of objective forecasting such as the method of Model Output Statistics (Section 3, page 15). The use of statistical approaches to the parameterization and empiric!al correction of dynamical modes should be explored further . Higher resolution regional models are able to resolve smaller scales of motion. These scales are expected to be more highly correlated with events of interest and thus to provide more valuable predictors for statistical regression and to lead specifically to more sharply defined probabilities of occurrence of small-scale events . Recommendation 3: The Panel strongly recommends the continued development of mesoscale regional models by a number of research groups and urges their use for predictability and observing system simulation experi- ments of the sort that have been carried out with planetary scale models (Section 4, page 18). Two important data sources that will contribute to the better identification and tracking of mesoscale and convective activity' are weather surveillance radars (10 cm) and geostationary satellites. Recommendation 4; The Panel recommends that automatically digitized radar data be provided with complete coverage of the areas of the United States susceptible to severe storm development (Section 5,a,i; page 23). Boundary layer research will contribute in an important way to the development of mesoscale models. 279 33 Recommendation 5: The Panel recommends that geostationary satellite digitized images, both visible and infrared, be provided at half -hourly intervals, with the highes t possible resolution (Section 5,a,i; page. 23). Much of the severe weather in the central United States occurs between 21Z and 03Z, and it is probable that upper air data at 18Z would greatly benefit short-range forecasting if utilized with proper dynamical and statistical techniques. Recommendation 6; The Panel recommends that consideration be given to the addition of routine rawinsonde observations at 18Z in the central United States during the severe storm season (Section 5, a, iii; page 26). An early experiment in the SESAME Scientific Program is a region scale experiment with a doubling of the density of rawinsonde stations over the central United States. This experiment contributes directly to the understanding of the statistical relationships between convectjve storms and the large mesoscale disturbances which produce them. The Panel strongly endorses the SESJU^ Program and, in part i cular, the conduct of the regional rawinsonde experiment in the spring of 1979. 280 34 FIGURE CAPTIONS Fig. 1 Computer -drawn map of thunderstorm probability (solid) and conditional probability of tornadoes, large hail, or damaging winds (dashed). The probabilities are valid for each manually digitized radar (MDR) block during the 21-27 hr interval following 0000 GMT initial time, or + 3 hrs from 0000 GMT the next day- Observed tornadoes = V. , . hail = • , and damaging winds = ■ . Fig. 2 Two-to-six hour probability of severe weather. The severe weather reports plotted are valid for the forecast period. The symbols are: T tornado; • hail :^ 3/4 in. dia.; and m damaging surface wind gusts. Fig. 3 SESAME rawinsonde network for 1979 regional scale experi- ment (from Fig. 3.3, Project Severe Environmental Storms and Mesocale Experiment, Environmental Research Laboratories, National Oceanic and Atmospheric Administration, 1976) 281 35 •H 232 36 •H 283 37 S5 §gg uj n. *- c/} 00 O •H 284 REFERENCES Alaka, M. A., W. D. Bonner, J. P. Charba , R. L. Crisci, R. C. Elvander, and R. M. Reap (1973). Objective techniques for forecasting thunderstorms and severe weather. Report No. FAA- RD-73-117. Techniques Development Laboratory, Silver Spring, Maryland. 97 pp. Atlas, D. , and C. W. Ulbrich (1974). The physical basis for attenuation rainfall relationships and the measurement of rainfall parameters by combined attenuation and radar methods. J. Rech. Atmos. 8:275-298. Barge, B. L. (1974). Polarization measurements of precipitation backscatter in Alberta. J. Rech. Atmos. 8:163-173. Boone, L. M. (1974). Estimated crop losses due to hail. Agrucultural Economics Report 267. Economic Research Services, U. S. Department of Agriculture, Washington, D. C. Brouillette, J. (1955). A tornado warning system: its functioning on Palm Sunday in Indiana. Research Report No. 15. Disaster Research Center, Ohio State University, Columbus, Ohio. Carbone, R. E. , D. Atlas, P. J. Eccles , R. Fetter; and E. A. Mueller (1973). Dual wavelength radar hail detection. Bui].. Am. Meteorol. Soc. 54:921-924. Charba, J. P. (1974). Objective forecasts of severe thunderstorms from observed surface predictors. TDL Office Note 74-1. Tech- niques Development Laboratory, Silver Spring, Maryland. 12 pp. Charba, J. P. (1975). Operational scheme for short-range forecasts of severe local weater. Preprints of 9th Conf. on Severe Local Storms. Am. Meteorol. Soc, Boston, Massachusetts, pp. 51-57. 285 93 Darkow, G. L. , and R. L. Livingston (1975). The evolution of the surface static energy fields on April 3, 1974. Preprints of 9th Conf . on Severe Local Storms. Am. Meteorol. Soc. , Boston, Massachusetts, pp. 254-269. David, C. L. (1971). A statistically generated aid for forecasting severe thunderstorms and tornadoes. Preprints of 7th Conf. on Severe Local Storms. Am. Meteorol. Soc, Boston, Massachusetts, pp. 42-44. David, C. L. (1971). An objective method for estimating the probability of severe thunderstorms using predictors from the NMC (PE) numerical prediction model and from observed surface data. Proc. of 7th Conf. on Severe Local StonT\s. Am. Meteorol. Soc, Boston, Massachusetts, pp. 223-225. Deardorff, J. W. (1972). Numerical investigation of neutral and unstable planetary boundary layers. J. Appl. Meteorol. 11:91-115. Donaldson, R. J., Jr., R. M'. Dyer, and M. J. Kraus (1975). Opera- tional benefits of meteorological Doppler radar. AFCRL-TR-7 5-103. Air Force Cambridge Research Laboratories, Hanscomb Air Force Base, Massachusetts. Eccles , P. J. (1976). Remote measurement of mass and kinetic energy of hial using dual wavelength radar. Preprint of 18th Radar Meteorol. Conf., Oct. 26-29, 1976. Am. Meteorol. Soc, Boston, Massachusetts, pp. 192-199. Environmental Research Laboratory (1976). SESAME Project Develop- ment Plan. National Oceanic and Atmospheric Administration, Department of Commerce, Boulder, Colorado. 42 pp. 286 95 I Fujita, T. T. (1975a). Super outbreak tornadoes of April 3-4, 1974, University of Chicago, Chicago, Illinois. Fujita, T. T. (1975b). Detection of severe storms by satellites. Proc. of 3rd Symposium on Meteorol. Observations and Instrumenta- tion. Am. Meteorol. Soc. , Washington, D. C. Ga3.way, J. G. (1956). The lifted index as a predictor of latent instability. Bull. Am. Meteorol. Soc. 37:528-529. George, J. J. , et al (1960). Weather Forecasting for Aeronautics . Academic Press, New York. 673 pp. Georges, T. M. , and G. E. Green (1975). Infrasound from convective storms. Part IV. Is it useful for storm warning? J. Appl. Meteorol. 14:1303-1316. Grotjahn, R. , and J. J. O^Brien (1976). Some inaccuracies in finite -differencing hyperbolic equations. Mon. Weather Rev. 104:180-194. Haltiner, G. J., and R. T. Williams (1975). Some recent advances in nuinerical weather prediction, Mon. Weather Rev. 103:571-590. Harold, T. W. , E. J. English, and C. A. Nicholas (1974). Accuracy of radar -derived rainfall measurements in hilly terrain. Quart. J. R. Meteorol. Soc. 100:331-350. Howcroft, J. G. (1971). Local forecast model: present status and preliminary verification. Nat. Meteorol. Center Office Note 50. National Oceanic and Atmospheric Administration, Suitland, Maryland. 6 pp. Jameson, A. R. (1976). Use of the ratio of radar reflectivity factors at 10 and 3 cm in hailstorm observations. Preprints of .17th Radar Meteorol. Conf . , Oct. 26-29, 1976. Am. Meteorol. Soc, 287 96 Boston, Massachusetts, pp. 173-180. (Also to appear in Ph.D. dissertation by same author, Dept. of Geophy. Scis. , University of Chicago, Chicago, Illinois, 1977). Klein, W. H. , and H. R. Glahn (1974). Forecasting local weather by means of model output statistics. Bull. Am. Meteorol. Soc. 59:1217-1227. Leith, C. E. (1975). Numerical weather prediction. Rev. Geophys. Space Phys. 13:681-684. Madden, R. A. (1976). Estimates of the natural variability of time -averaged sea -level pressure. Mon. Weather Rev. 104:942- 952. Merilles , P. E. (1975). On horizontal truncation errors in spectral models. Mon. Weather Rev. (to be published). Mileti, D. S. (1974). A normative causal model analysis of disaster warning response. Unpublished Ph. D. dissertation, Dept. of Sociology, University of Colorado. Mileti, D. S. , T. E. Drabek, and J. E. Haas (1975). Human systems in extreme environments: a sociological perspective. Mono- graph No. 021, Institute of Behavioral Science, University of Colorado. Mileti, D. S. (1975). Natural hazard v;arning systems in the United States: a research assessment. Monograph No. NSF-RA-E-75- 013, Institute of Behavioral Science, University of Colorado. Miller, R. G. (1972). Notes on analysis and severe storm forecasting procedures of the Air Force Global Weather Central. Technical Report 200 (Rev.). Air Weather Service, U. S. Air Force. 170 pp. 288 97 National Oceanic and Atmospheric Administration (1974). The widespread tornado outbreak of April 3-4, 1974. Natural Disaster Survey Report 74-1. Depart;nent of Commerce, Washington, D. C. National Weather Service (1975). Program development plan: automation of field operations and services. National Oceanic and Atmospheric Administration, Department of Commerce, Washington, D. C. Puri, I., and W. Bourke (1974'i. Implications of horizontal resolution in spectral model integrations. Mon. Weather Rev. 102:333-347. Reap, R. M. , and M. A. Alaka (19G9). An objective quasi -lagrangian index for predicting convective weather outbreaks. Proc. of 6th Conf. on Severe Local Storms. Am. Meteorol. Soc. , Boston, Massachusetts, pp. 119-124. Reap, R. M. (1972). An operational three-dimensional trajectory model. J. Appl. Meteorol. 11:1193-1202. Reap, R. M. (1974. Thunderstorm and severe probabilities based on model output statistics. Preprints of 5th Conf. en Weather Forecasting and Analysis. Am. Meteorol. Soc, Boston, Massa- chusetts, pp. 266-269. Reap, R. M. , and D. S. Foster (1975). New operational thunderstorms and severe storm probability forecasts based on model output statistics (MOS) . Preprints of 9th Conf. on Severe Local Storms. Am. Meteorol Soc, Boston, Massachusetts, pp. 58-63. Showalter, A. K. (1953). A stability index for thunderstorm forecasting. Bull. Am. Meteorol. Soc. 34:250-252. Shuman F. G. , and J. B. Hovermale (1968). An operational six-layer primative equation model. J. Appl. Meteorol. 7:525-547. 289 98 Stackpole, J. D. (1957). Numerical analysis of atmospheric soundings. J. Appl. Meteorol. 6:454-467. Taylor, B. C. , and K. A. Browning (1974). Towards an automated weather radar network. Weather 29:202-215. Wells, P. I., and P. V. Tryon (1976). The attenuation of UHF radio signals by houses. OT Report 75-98. Institute for Telecommunication Sciences, Office of Telecommunications, U.S. Department of Commerce. Available from NTIS , PB 258447 /AS. White, G. F. , and J. E. Haas (1975). Assessment of research on natural hazards. Massachusetts Institute of Technology Press, Cambridge, Massachusetts. Willis, T. G. (1969). Characteristics of the tornado environment as deduced from proximity sounds. Proc. of 6th Conf. on Severe Local Storms. Am. Meteorol. Soc. , Boston, Massachusetts, pp. 222-229. Additional Reading Committee on Science and Astronautics (1974). Hearings on H. R. 9811, November 5, 7, and 9, 1973, No. 23, 93rd Congress, 1st Session, before the Subcommittee on Space Science and Applica- tions. U.S. Government Printing Office, Washington, D. C. National Advisory Committee on Oceans and Atmosphere (1973). A report to the President and the Congress, 2nd Annual Report. U. S. Government Printing Office, Washington, D. C. 290 8. DRAFT 15 March 1976 MIDLATITUDE REGIONAL PROBLEMS; Report of FGGE Workshop C National Center for Atmospheric Research Boulder, Colorado 80303 26 January 1976 Organizer: C. Kreitzberg, Drexel University 215-895-2726 Rapporteur : D. Bavunhefner, NCAR 303-494-5151, Ext. 669 FTS: 322-5669 Principal Participants: Observers : R. Barry, University of Colorado R. Bleck, University of Miami E. Danielsen, NCAR J. Fein, University of Oklahoma R. Inman, University of Oklahoma D. Johnson, University of Wisconsin A. Kasahara, NCAR D. Lilly, NCAR/ERL-NOAA Y. Lin-, St. Louis University D. Martin, St. Louis University J. Paegle, University of Utah D. Perkey, NCAR J. Perry, NRC Y. Sasaki, University of Oklahoma T. Schlatter, NCAR M. Shapiro, NCAR S. Tracton, NMC-NOAA H. van Loon, NCAR J. Winston, NESS-NOAA B. Bliesner, NCAR F. Carr, SUNYA P. Downey, NCAR B. Fogle, NSF G. Hermcui, SSEC, University of Wisconsin J. Kimpel, University of Oklahoma E. Zipser, NCAR Sponsored by the National Science Foundation. 291 TABLE OF CONTENTS Introduction 1. Charges 2 . Premises 3. Research Topics 4. Influence of FGGE Data 5. Recommendations Appendices A. Regional Studies in Southern Hemisphere Midlatitudes B. Regional Studies in Southern Hemisphere Midlatitudes C. Regional Studies During FGGE in Anticipation of GARP-II, Climate Studies D. Coastal Effects on Cyclone Development E. Cyclone Development Related to Latent Heating F. Scale Interactions in Extratropical Cyclones G. Problems Associated with Orographic Effects in a Niimerical Prediction Model H. Objective Analysis and initialization I. FGGE Satellite Data Useful in Studies of Midlatitude Regional Problems J. Satellite Soundings vs. Rawinsondes, Dropsondes amd Radar Tracking K. Proposed SESAME Fy79 (Spring) Field Program L. Data Analyses and Initialization of FGGE Data 292 Y Introduction The atmosphere contains a full spectrum of interacting scales which are often studied separately because of limitations in theory, understeuiding , observations emd computer power. Scale separation is a powerful analytic tool, but a treacherous one because scale inter- actions cire essential in the many developments importemt in, weather forecasting. Most generally, the cyclonic scale determines what, when and where smaller-scale developments occur, and the small scale, in turn, inqpacts evolution of the large scale. We therefore need to incorporate smaller-scale phenomena into modeling research to accom- plish the first GABP objective — to extend large-scale weather fore- ccusting to a longer time period. Examples of several in^xsrtant mechanisms whereby the small scale impacts evolution of the larger scale include orography, coastal effects, jet streams, fronts (upper and lower level), heat, moisture cuid momentum fluxes from organized convective disturbances. In the classical fluid mechanical sense of primary and secondary mechanisms, the smaller scales normally have a secondary effect on the large- scale circulations. However, the scale interaction may be extremely important in a nuniber of particular situations. Having incorporated the primary leirge-scale processes in current numerical weather pre- diction models, it may well be that satisfying the GT^BP first objective will depend upon better understamding of small-scale processes and their relationship to the synoptic scale. Midlatitude regional studies for FGGE are therefore basically concerned with understanding 293 the synoptic-cyclonic scale of motions and their interaction with sub- synoptic phenomena. The knowledge gained from such work can be applied to the large-scale global models to effectively extend large- scale predictability. This report is divided into five sections: (1) describes the charges to the Hidlatitude Regional Workshop; (2) defines the premises agreed upon in deliberations by the principal participemts ; (3) briefly summarizes topics identified in the Workshop that may influence large- scale predictability; (4) outlines the effect that FGGE data will have on the solution of the problems identified in (3) ; and (5) makes a series of recommendations that the Workshop participants feel will increase the chances of success with regard to the first GARP objective. Tliroughout this report, the reader is referred to appendices for more detailed information on a particular subject. The appendices repre- sent various individuals' opinions on the siibject in question and are not a consensus of the participants in the Workshop. 1.0 Charges At the direction of FGGE Workshop A, which met at GFDL in November, 1975, we interpreted our charge for Workshop C, which dealt with Midlatitude Regional Problems, to be threefold: (1) identification and consolidation of FGGE-related research topics, (2) evaluation of the FGGE cA>serving, processing and archiving plans in relation to the research identified in (1), euid (3) communication of these ideas to all interested in participating in FGGE. 294 This report does not represent a comprehensive answer to the stated charge, but is only a first attempt to stimulate contributions from the many other scientists expected to participate in the research on these questions. The ideas herein are primarily personal opinions and do not necessarily imply a consensus of all participants or their respective organizations . 2 . Premises The following five statements summarize the premises upon which the panel based its deliberations. A majority of the principal par- ticipants was in agreement on these statements. 2.1 Research in the United States on midlatitude regional-scale problems in the next five years will deal primarily with the Northern Hemisphere because of its higher conventional data density. We will therefore consider only the Northern Hemisphere midlatitudes ; how- ever, it is essential that the knowledge gained be transferred as rapidly as possible to Southern Hemisphere problems (Appendices A & B) . While our attention will focus on the continental United States, it is recognized that other regions may exist with comparable data density and have many of the same characteristics; thus, our con- clusions could apply to these regions as well. 2.2 The GARP II objective — vxnderstanding and predicting climate time-scale events — may have many regional-scale implications. FGGE Cein only begin to answer GARP II objectives because of its limited observing period. Regional climate questions will not be considered further and are deferred to Appendices A and C. 295 2.3 FGGE will contribute svibstantially to the present Southern Hemisphere and tropical data coverage. In the northern midlatitudes , principal new information during FGGE will be increased satellite coverage over both land and ocean areas. The FGGE Special Observing Periods, as presently constituted, will not intact midlatitude regional studies. The primajry benefit of FGGE data for midlatitude regional research will be to provide a single-soxirce archive in the form of level Il-b and Ill-b data sets. 2.4 Although it remains to be shown what scales of motion the satellite system will ultimately be able to resolve, the improvement in analyses for zonal wavenumbers greater than 10 in mid-latitudes should be an important FGGE goal. During FGGE, increased efforts to collect and process conventional data such as AIKEPS, SECONS etc. in level Il-b and Ill-b data sets should also svibstantially enhcuice cyclonic-scale analyses over the northern midlatitudes . 2.5 Midlatitude regional studies in the post-FGGE era will undoubtedly deal with cases during FGGE because of the improved definition of large-scale processes and the in^roved avedlability of consistent and convenient data sets. 3.0 Research Topics The Workshop identified four areas of regional-scale research that might have a significcuit impact on attainment of the GABP I objective through FGGE — (1) cyclone development (2) scale inter- action, (3) regional-scale models and (4) observational capabilities. 296 3.1 Cyclone development The development and decay of cyclonic-scale weather systems are still inadequately understood phenomena. They occur in favored geographical areas, such as in the lee of mountain ranges or near coastlines or over relatively warm water (e.g.. Gulf of Mexico and the Gulf Stream) (Appendix D) . Certain physical processes sometimes play an in^ortant role in the development of these systems (Appendix E) , but there is considerable disagreement as to the exact mechcmisms and energetics of cyclonic-scale amplification ouid decay. The present numerical models are able to predict development of baroclinic waves to a first approximation only. Unfortunately, the secondary effects cu:e needed to extend the current range of predictability. 3.2 Scale interaction The problem of interactive scales of motion, as discussed in the introduction, may well be a primary source of error in numerical weather prediction. The net effect of small scales on the synoptic scale, such as mountain waves, convection euid gravity waves, in terms of heat, moisture and momentum fluxes is essentially xinknown. The parameterization of these small-scale phenomena in regional numerical models has begun to show its relative iii^>ortance in accurately foreccusting the cyclonic scale. The effect of rapid cyclonic-scale amplification on all scales, the planetary-scale waves as well as the mesoscale features, may be an overlooked area of research with respect to extension of predict- ability (Appendix F) . Certain global features such as blocking may 297 be triggered by regional-scale developments. 3.3 Model development Most modelers of regional- scale phenomena agree that model- dependent errors rather than external errors are still the dominant source of discrepancies found when forecasts are compared to the real atmosphere. Improvements in numerical methods of integration, increased horizontal and vertical resolution, methods of handling mountainous terrain (Appendix G) , physical parameterization schemes and boundary-layer treatments are of primary in^portance. In addition, in^rovement of the data analysis scheme emd cissociated initialization over a limited geographical area are also areeis of research that may contribute to a more accurate prediction (Appendix H) . 3.4 Observational capabilities The usefulness of the proposed FGGE observing system will be highly dependent upon the capabilities of prototype remote observing technology. For regional-scale analysis and prediction, it is essen- tial to know the acc\iracy, scales of motion measured and confidence levels (error variance) by which the various atmospheric variables are being measxired. Our current knowledge, as expressed in Appendix I, gives a preliminary feeling about the overall capabilities, but more quantitative studies are needed. We should determine, for example, how well jet streams, temperature gradients etc. can be resolved with the proposed FGGE observing system. 298 A direct comparison of the new observing system with redundemt information provided by conventional techniques is highly desirable (Appendix J) . To provide a more accurate analysis of the data, information gathered from these studies, such as the relative scales observed, their accuracy and usefulness, could then be applied to regions where no red\indant information exists. 4.0 Influence of FGGE Data It was generally felt by the participants of this Workshop that the FGGE observing system and subsequent data archives as presently designed will not significantly aid the research described in Section 3. This conclusion prevails because of the need for data on smaller scales than are currently being plemned for FGGE and because of a skepticism toward the new observing system capabilities. Therefore, the most important contribution that FGGE-related research will make to midlatitude regional problems will be em understanding of the capabilities of the observing system itself. There are also some potentially beneficial spinoffs from FGGE data. For regional modeling, glc^cQ-scale motions can be used as lateral boiuidary conditions for finear-scale reseeurch and prediction. A solid verification base of laxge-stale information with which to compare modeling and diagnostic results can be derived from the pro- posed data archive. 299 5.0 Recommendations 5 . 1 Regional studies We recommend the following midlatitude regional studies that will contribute strongly to FGGE and the first GARP objective; 5.1.1 Recommendation ; Diagnostic studies of cyclogenesis in the lee of the Rockies, in the Gulf of Mexico, over the Mississippi Valley and over the Gulf Stream should be conducted. 5.1.2 Recommendation ; Subgrid- scale processes that might iinpact global-scale forecasts should be diagnosed. Verification and improvement of GCM parameterizations should be en^hasized. Since MflEX is an example of such a parameterization study, the experience and knowledge gained in this stu<^, as well as GATE, should be carefully considered in plarming and performing 5.1.2. 5.2 FGGE contributions to regional studies In order to conduct regional studies to meet the goals of improving and extending weather forecast skill and to meet the needs of . small-scale studies, we recommend; 5.2.1 Recommendation ; The routine FGGE data collection should be expanded during selected periods of par- ticularly important weather developments within or near the continental United States and also during periods of particularly comprehensive regional observations in conjunction with other projects, especially SESAME (Appendix K) . The expanded data set could include high- resolution satellite data, extra reconnaisance missions etc. It cannot be stressed too strongly that researchers will be severely hampered emd limited in their attack of problems that 300 contribute toward GARP I objectives or benefit in emy direct way from the FGGE data if this recommendation is not accommodated. 5.2.2 Recommendation : The expanded FGGE data sets should be assembled for three 3-day periods in each season in addition to the SESAME periods. These special periods should, if possible, be selected on the basis of the occurrence of certain intermittent weather phenomena. Coordi- nation with other mesoscale field projects should be encouraged to take maximum advantage of other data- gathering activities and observing systems. It would be particularly useful if dropsondes could be deployed along the coasts in those cases dealing with coastal phenomena. 5.2.3 Recommendation ; The expanded FGGE data archives should include high- resolution satellite sounder data, satellite winds, NMC operational model output, sta- tistical products, AFGWC model output including their mesoscale cloud diagnoses , airway observations , radar observations, manually digitized radar data, hourly precipitation data and storm data. 5.2.4 Recommendation : The expanded FGGE data sets, cor- responding to level Ill-b output, should be provided, but in the region of augmented data should be gridded independently of dynamic models (Appendix L) . 5.2.5 Recommendation ; Research on four-dimensional data assimilation schemes appropriate for the augmented data density should be encouraged so that the gridded data can be analyzed utilizing dynamical techniques like those developed for the level Ill-b standard FGGE grid (Appendix H) . 5.3 Observing systems In order to properly assess the capabilities of the FGGE observing system emd to make it more useful to the research community, we have outlined a third series of recommendations. 301 5.3»1 Recommendation ; A systematic assessment should be made of remotely sensed satellite data, such as temperature and moisture profiles, wind vectors, sea surface ten^)- erature etc., by direct coinparison with rawinsondes, dropsondes and other conventional data over different latitudes and regions of the United States and over neeurby ocecuis . The Gulf of Mexico and the Gulf Stream are of peurticular concern, but the tests should include the Gulf of Alaska and the region off the west cocist. This work should be intensified before FGGE and continue dviring FGGE. 5.3.2 Recommendation ; Top priority should be given to the collection of special rawinsonde and dropsonde data to meet the objective of systematic assessment of satellite data. The cost of special sovindings not taken on synoptic times is trivial cc»npcired to the value of unambiguous direction con^arison with the satellite systems to be used in FGGE (Appendix J) . 5.3.3 Recommendation : Special effort should be given to coordination with Project SESAME to ensure that its regional experiment (Appendix K) coincides with FGGE. This recommendation arises from recognition that minimal additional expense would be involved to collect considerably more data for both interccanpeurison and case studies. The overlap of research toward GARP I objectives with that of SESAME objectives would xondoubtedly contribute tremendously toward understanding of scale interactions. 5.3.4 Recoimnendation ; Every effort should be made to place the VISSR Atmospheric Sounder (VAS) (Appendix I) on GOES-C in time for FGGE rather than waiting to lavmch it with GOES-D. 302 Satellite enthusiasts and radiosonde enthusiasts often disagree about the merits of the two measuring systems, but a few well conducted es^eriments could settle the question. The accuracy of the radiosonde can be determined by simultaneous radar tracking. After the torroidal oscillations of the balloon and short vertical wavelength, internal gravity wave oscillations are removed by filtering; the height and time derivative of the height can be used to calibrate the mecin temperatures . The acciiracy of the satellite temperatures can then be assessed from the radiosonde data. Radar tracking is probably feasible at only a few sites but it is important for establishing a standard for both temperature and wind. E. Danielsen, NCAR 303 APPENDIX K Proposed SESAME FY79 (Spring) Field Program The current plan for the SESAME project envisages four field periods; (1) a preliminary meso-scale experiment in 1978 to test the ability of veurioxis sensors and systems to measure mass and moisture convergence in a location siabject to thunderstorm activity; (2) a regional-scale es^eriment in the spring of 1979; (3) the first coordinated multi-scale experiment in 1980, and (4) the second multi-scale escperiment in 1982. The regional-scale escperiment is designed to acquire upper air data necessaary to test the capability of regional-scale (250-2500 km) numerical weather simulation models for predicting the general atmo- spheric conditions associated with convective storm development. The National Weather Service sounding network will be augmented by 20 to 30 additional stations over a 2 x 10^ km^ area of the central United States, thereby reducing the spacing between stations to approxi- mately 200 km. During four 48-hour periods, soundings will be acquired at 3-hour intervals from all stations in the e35>eriment cirea. The periods will be selected to observe the pre- convective atmosphere associated with cyclogenetic storms. Although the experi- ment is independent of other activities, both the observational periods and the data reduction will be coordinated as seems appropriate with both storm-scale programs, e.g., the National Severe Storms Laboratory, the Chicago Area Project, and the National Hail Research E^^eriment, and with the FGGE observing system. 304 Two suggested versions of the network layout are attached. Neither should be considered final. Automatic data processing techniques will be available to process the soundings, and model test results should be available in time to incorporate modifications in the major field experiment the following spring. D. Lilly, NOAA, NCAR 305 \ -^-^ g^-U ^ ^jf-l--) -'-■i-.: J5. en <^ ^ £ I 2 I J ^^>' gw r/Sr -> o \ ^ APPENDIX L Data Analyses and Initialization of FGGE Data Objective analysis of the regionally expanded data set into a Ill-b format would greatly enhance its utility for individual users. However, if all of the Ill-b data sets are model-dependent as a consequence of idealized balance constraints or particular forecast model assumptions, this utility would diminish for many types of investigations. This would be particularly true of diagnostic studies whose conclusions should ideally be independent of model assumptions. If their conclusions are model-dependent, the nature of the model dependence should be easily understood. This, unfortunately, is not the case for four-dimensional data analyses obtained by continuous data assimilation into complicated numerical models. Therefore, it seems reasonable to suggest that some FGGE regional data be objectively archived with minimal physical constraints beyond hydrostatic balance. It appears as though the data set would be sufficiently dense over the United States to justify this approach for at least part of the augmented level Ill-b regional archive. J. Paegle, University of Utah * U.S. Government Printing Office: 1977-777-045/1230 Regions 308 PENN STATE UNIVERSITY LIBRARIES iiiillllliiilll AD0D07mqs3t2 ^