,1 -v^ NOAA Technical Memorandum NESS 78 Q <*■*: op 5»* ^ATZS 0< " *^ ■ SATELLITE DERIVED SEA-SURFACE TEMPERATURES FROM NOAA SPACECRAFT \i Washington, D. C. June 1976 noaa NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION / National Environmental Satellite Service NOAA TECHNICAL MEMORANDUMS National Environmental Satellite Service Series The National Environmental Satellite Service (NESS) is responsible for the establishment and oper ation of the environmental satellite systems of NOAA. NOAA Technical Memorandums facilitate rapid distribution of material that may be preliminary in natun and so may be published formally elsewhere at a later date. Publications 1 through 20 and 22 through 2 are in the earlier ESSA National Environmental Satellite Center Technical Memorandum (NESCTM) series The current NOAA Technical Memorandum NESS series includes 21, 26, and subsequent issuances. Publications listed below are available from the National Technical Information Service, U.S. Depart ment of Commerce, Sills Bldg. , S285 Port Royal Road, Springfield, Va. 22151. Prices on request. Orde by accession number (given in parentheses). Information on memorandums not listed below can be obtaine from Environmental Data Service (D831), 3300 Whitehaven St., NW.., Washington, D.C. 20235. NESS 38 Publications and Final Reports on Contracts and Grants, 1971. NESS, June 1972, 7 pp. (COM-72 11115) NESS 39 Operational Procedures for Estimating Wind Vectors From Geostationary Satellite Data. Mi chael T. Young, Russell C. Doolittle, and Lee M. Mace, July 1972, 19 pp. (COM-72-10910) NESS 40 Convective Clouds as Tracers of Air Motion. Lester F. Hubert and Andrew Timchalk, Augus 1972, 12 pp. (COM-72- 11421) NESS 41 Effect of Orbital Inclination and Spin Axis Attitude on Wind Estimates From Photographs t Geosynchronous Satellites. Linwood F. Whitney, Jr., September 1972, 32 pp. (COM- 72- 11499) NESS 42 Evaluation of a Technique for the Analysis and Forecasting of Tropical Cyclone Intensities Frc Satellite Pictures. Carl 0. Erickson, September 1972, 28 pp. (COM-72-11472) NESS 43 Cloud Motions in Baroclinic Zones. Linwood F. Whitney, Jr., October 1972, 6 pp. (COM- 72 10029) NESS 44 Estimation of Average Daily Rainfall From Satellite Cloud Photographs. Walton A. Follansbee January 1973, 39 pp. (COM-73-10539) NESS 45 A Technique for the Analysis and Forecasting of Tropical Cyclone Intensities From Satellil Pictures (Revision of NESS 36). Vernon F. Dvorak, February 1973, 19 pp. (COM-73-10675) NESS 46 Publications and Final Reports on Contracts and Grants, 1972. NESS, April 1973, 10 pj (COM-73-11035) NESS 47 Stratospheric Photochemistry of Ozone and SST Pollution: An Introduction and Survey of Si lected Developments Since 1965. Martin S. Longmire, March 1973, 29 pp. (COM-73-10786) NESS 48 Review of Satellite Measurements of Albedo and Outgoing Long-Wave Radiation. Arnold Grube: July 1973, 12 pp. (COM-73-11443) NESS 49 Operational Processing of Solar Proton Monitor Data. Louis Rubin, Henry L. Phillips, a Stanley R. Brown, August 1973, 17 pp. (C0M-73-11647/AS) NESS 50 An Examination of Tropical Cloud Clusters Using Simultaneously Observed Brightness and Hi Resolution Infrared Data From Satellites. Arnold Gruber, September 1973, 22 pp. (COM-7 11941/4AS) NESS 51 SKYLAB Earth Resources Experiment Package Experiments in Oceanography and Marine Science. L. Grabham and John W. Sherman, III, September 1973, 72 pp. (COM 74-11740/AS) NESS 52 Operational Products From ITOS Scanning Radiometer Data. Edward F. Conlan, October 1973, pp. (COM-74-10040) NESS 53 Catalog of Operational Satellite Products. Eugene R. Hoppe and Abraham L. Ruiz (Editors March 1974, 91 pp. (COM-74-11339/AS) NESS 54 A Method of Converting the SMS/GOES WEFAX Frequency (1691 MHz) to the Existing APT/WEFAX Fi quency (137 MHz). John J. Nagle, April 1974, 18 pp. (C0M-74-11294/AS) NESS 55 Publications and Final Reports on Contracts and Grants, 1973. NESS, April 1974, 8 p (C0M-74-11108/AS) (Continued on inside back cover) c g I NOAA Technical Memorandum NESS 78 SATELLITE DERIVED SEA-SURFACE TEMPERATURES FROM NOAA SPACECRAFT Robert L. Brower Hilda S. Gohrband William G. Pichel T. L. Signore Charles C. Walton Washington, D. C. June 1976 mO WMOsp^ UNITED STATES DEPARTMENT OF COMMERCE Elliot L. Richardson, Secretary NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION Robert M. White, Administrator National Environmental Satellite Service David S. Johnson. Director r MfNl Of PREFACE This technical memorandum should give the reader an understanding of the derivation of sea surface temper- atures from NOAA spacecraft data. A brief description of the past, present, and future of the sea surface temperature (SST) operation is followed by in-depth discussions of the scientific background for measure- ment of sea surface temperatures, the data processing system, its performance, and SST products and displays. The intent of the authors is to provide a basic document with a moderate amount of detail on satellite SST derivation and processing. It is hoped that both the casual and the serious user will find much of value in this memorandum. Each section is relatively self-contained so the entire memorandum need not be read. The first-time user may wish to read only the Overview (sec. 1) and section 5, "Sea Surface Temper- ature Products and Displays", to obtain an idea of what is available. On the other hand, the veteran user may find much new information in sections 2, 3 and 4, which detail the SST system and its per- formance. Mention of a commercial company or product does not constitute an endorsement by the NOAA National Environmental Satellite Service. Use for publicity or advertising purposes of information from this publication concerning proprietary products or the tests of such products is not authorized. li CONTENTS Preface iii List of Illustrations vi List of Tables vii Abstract 1 1 . Overview - Past , Present , Future 2 1 . 1 The problem 2 1.2 The past 2 1 . 3 The present 3 1 . 4 The future 4 2 . SST Model Theory 6 2.1 Radiometric measurement of sea surface temperature. 6 2.1.1 Sensor response and calibration 6 2.1.2 Sea surface temperature measurements 8 2.1.3 Atmospheric attenuation 10 2 . 2 SST retrieval theory 13 2.2.1 System characteristics affecting retrievals ....... 13 2.2.1.1 Resolution and sampling rate 13 2.2.1.2 System noise 13 2.2.1.3 Earth location accuracy 14 2.2.1.4 Raw versus mapped data 14 2.2.1.5 Cloud cover percentage 14 2.2.2 Rationale for retrieval technique developed 15 2.2.3 Retrieval algorithm 17 2.2.4 Histogram types and tests 20 2.2.4.1 Histogram tests 20 2.2.4.2 Histogram types 21 2.2.5 Behavior of retrieval technique 22 2 . 3 Rej ection of erroneous data 27 2.3.1 Static control 27 2.3.2 Dynamic control 28 2.3.2.1 Conceptual view of objective analysis 28 2.3.2.2 Search procedures 29 2.3.2.3 Adjacent data rejection test 31 2.3.2.4 Final merge 33 3 . SST Processing System 35 3 . 1 Spacecraft and ground equipment 35 3.1.1 NOAA spacecraft 35 3.1.2 SR subsystem 35 3.1.3 VTPR subsystem 38 3.1.4 Spacecraft tape recorder 38 3.1.5 Acquisition stations and Z-axis correction 38 iii 3.1.6 NESS central processing facility 38 3.2 Overview of SST data processing system 39 3. 3 Orbital processing 41 3.3.1 Data input 41 3.3.2 Creation of raw data base 41 3.3.3 Calibration and quality control 41 3.3.4 Earth location calculation 43 3.3.5 SST raw retrieval calculation 43 3.4 Daily SST processing 43 3.4.1 Atmospheric attenuation calculation 43 3.4.2 Objective analysis technique 45 3.4.3 Quality control 45 3.4.4 Daily archival 46 3.4.5 Processing summary and initialization 46 3. 5 Verification 46 3.6 Scheduled non-daily processing 47 4 . System Performance 48 4.1 SST verification 48 4.1.1 Comparison with ship data 48 4.1.2 Relative merits of ship and satellite measurements. 50 4. 2 System reliability 52 4 . 3 System monitoring and evaluation 53 4.4 Satellite SST coverage 54 5 . SST Products and Displays 57 5.1 Products 57 5.1.1 Observations 57 5.1.2 Analyzed Fields 57 5. 2 Displays 57 5.2.1 Latitude/longitude printout 58 5.2.2 Photographic display 58 5.2.3 Teletype transmission 58 5.2.4 Microfilm 58 5.2.5 Character printout 59 References 67 Appendix A: EDS Archive Tape Format Description 69 Appendix B: SST Analyzed Field 10-Day Archive Tape Format Description 72 IV ILLUSTRATIONS Fig. 1 Window energy vs. temperature curve for NOAA-4 sensor 1 ." 9 Fig. 2 Arrangement and dimensions of data blocks for one frame of raw SRIR data 16 Fig. 3 Clear area data histogram and its associated mean estimate histogram 18 Fig. 4 Partly cloudy data histogram and its associated mean estimate histogram 19 Fig. 5 Data histogram of unresolved temperatures and its associated mean estimate histogram 25 Fig. 6 Data histogram containing semi-transparent clouds and its associated mean estimate histogram 24 Fig. 7 Low level broken stratus data histogram 25 Fig. 8 Data histogram of cold or mixed clouds 25 Fig. 9 Gradients on geographical grid 29 Fig. 10 Search pattern on strong and weak gradient fields. 30 Fig. 11 Scanning radiometer data flow 56 Fig. 12 The NOAA polar satellite 57 Fig. 15 SST automated processing system 40 Fig. 14 SST orbital processing 42 Fig. 15 SST daily processing system 44 Fig. 16 Daily mean difference between ship and satellite sea surface temperature measurements 49 Fig. 17 Evaluation procedures 55 Fig. 18 Number of satellite observations in Northern Hemisphere during July 1975 55 Fig. 19 Number of satellite observations in Southern Hemisphere during July 1975 56 Fig. 20 Latitude longitude grid of SST 60 Fig. 21 Digital Muirhead display of SST 61 Fig. 22 Data from SST teletype tape 62 Fig. 23 Global polar stereographic microfilm map 63 Fig. 24 Polar stereographic microfilm section off Eastern U.S 64 Fig. 25 Mercator projection microfilm map 65 Fig. 26 SST analyzed field character print sheet 66 Fig. B-l SST first guess field grid 73 Fig. B-2 SST grid box orientation 73 TABLES Table 2-1 10.5 to 12.5 pm atmospheric attenuation corrections 11 Table 2-2 Empirical mean atmospheric correction 12 Table 2-3 Chart of histogram types and tests 22 Table 2-4 Retrieval statistics for 3/2/75 26 Table 4-1 Regional comparison of ship measurements vs. satellite 50 Table 4-2 Classifying satellite measurements using ship reports 52 Table 4-3 Rate of success in obtaining SST 53 Table B-l Global scale analyzed field packed data format ... 74 VI SATELLITE DERIVED SEA-SURFACE TEMPERATURES FROM NOAA SPACECRAFT Robert L. Brower, Hilda S. Gohrband, William G. Pichel, T.L. Signore, and Charles C. Walton National Environmental Satellite Service, NOAA Washington, D.C. ABSTRACT. The National Environmental Satellite Service (NESS) has developed a sea surface temperature observation system to support the operational, research, and developmental needs of oceanographers and environmental scientists. The system pro- vides daily global surveillance of the ocean's surface temper- ature structure. Sea surface temperature values are derived from Scanning Radiometer (SR) infrared data from the NOAA series of polar orbiting satellites. The technique used to obtain these temperatures is the fully automated computer procedure, GOSSTCOMP (Global Operational Sea Surface Temper- ature Computation) . Surface temperature retrievals are derived by statistical analysis and quality control techniques applied to instrument measurements covering roughly 100-km square areas. Retrieval temperatures are corrected for the effects of atmospheric attenuation by using time-coincident measurements derived from a Vertical Temperature Profile Radiometer (VTPR) . The basic product obtained is a daily set of 5,000 to 7,000 observations of sea surface temperature over the oceans of both hemispheres. A number of derived products also are gen- erated. These products include: (1) a global analyzed temper- ature field accessible on computer disk storage, (2) hard copy displays of portions of this analyzed field, and (3) a permanent archive of the analyzed field and observations. The GOSSTCOMP procedure has produced a high level of dependa- bility for product delivery. During the second half of 1974, an operational success rate of 97% was maintained. During 1974, 1,640,654 observations were produced with a global daily mean difference from ship reports ranging from -0.9° to +0.39°C, with an RMS deviation varying between 1.67° and 2.23°C. Procedures used to obtain sea surface temperatures are constantly being improved as advances are made in retrieval, atmospheric attenuation, and quality control techniques. A major improve- ment in the accuracy of GOSSTCOMP-produced sea surface temper- atures is expected in 1978 with the next generation of polar orbiting satellites, the TIROS-N series. 1. OVERVIEW - PAST, PRESENT, FUTURE The National Environmental Satellite Service (NESS) Sea Surface Tempera- ture (SST) program was developed as an observation system to support research and development needs of NOAA's oceanographers and environmental scientists. The program provides daily global surveillance of the surface temperature structure of the world's oceans by remote sensing from satellites. The goal of this NESS effort is to develop and expand these capabilities so as to be competitive with traditional measuring devices. Sea surface temperature measurements are needed in applications such as weather prediction, commercial fishing, and environmental research. In many areas of the globe where observations are not available in the conventional form of in situ measurements from ships or buoys, remote sensing from satel- lites has proven to be feasible for providing quantitative measurements. Beginning with ITOS-1 (Improved TIROS operational satellite) , launched January 23, 1970, satellites of this series (ITOS-1, NOAA-1, NOAA-2, et seq. ) have carried instruments that make routine measurements from which sea sur- face temperatures can be derived. Early in 1970, NESS established as a goal the development of a system to obtain operational sea surface temperatures from the spacecraft. 1.1 The Problem The infrared portion of the Scanning Radiometer (SR) aboard the ITOS satellite offered a unique opportunity to examine synoptically the distri- bution of surface temperatures over the oceans. This opportunity, however, was accompanied by many problems. Considerable effort has been invested in producing an operational model and in overcoming the many factors that made obtaining accurate temperatures difficult. First, the sensor was designed primarily to furnish meteorolo- gical imagery. Use of the data from this sensor for quantitative measure- ments produced many initial problems. Output signals from the instrument essentially are measurements of the combined thermal radiation emitted by the ocean surface and the atmosphere. To relate the output signal to sur- face temperature, the effects of absorption by atmospheric gases in the infrared spectral window had to be taken into account and subtracted from the total signal. A thorough knowledge of instrument characteristics, electrical processing systems, and communications links had to be gained so that the sensor output could be related to the contents on the received signal. Finally, there was the necessity to design an automated, modular, operationally feasible, and highly dependable computer processing system that would produce accurate results. 1.2 The Past Development of the operational processing of infrared (IR) data from NOAA spacecraft for determining sea surface temperatures has been continuous in NESS since late 1970. By December 1970, an early prototype of an auto- mated model had been used to process two daytime passes per day. The proto- type model was used routinely until the end of January 1971. The results from this test were used to design more efficient procedures, and computer processing time was reduced from 40 to 0.8 minutes per orbit. This break- through in processing resulted in a global processing model that came into daily operational use on March 21, 1971 and operated continuously until the NOAA-1 Scanning Radiometer failed on July 8, 1971. The experience of these three months of daily operation and post analysis of the results pointed to a major weakness in the on-line data quality diagnostics and the calibration procedures used. The period between the failure of NOAA-1 and the launch of NOAA-2 on October 16, 1972 was used to improve the processing model, establish proce- dures for archiving temperature data, and to develop an interactive pro- cessing system with atmospheric attenuation data from the NOAA-2 Vertical Temperature Profile Radiometer (VTPR) . (Temperature data from the NOAA-1 SR sensor were corrected for atmospheiic attenuation by the use of empirical tables . ) Routine daily processing of NOAA-2 data to derive sea surface tempera- tures was started on December 1, 1972. Atmospheric attenuation corrections determined from the VTPR data were used in the operational model after June 1, 1973. Quality of the sea surface temperature measurements obtained from this processing model varied with time and geographical area. Verifi- cations performed in-house, as well as those done by others, have shown that quality was related to the temperature gradient field; there were good measurements in regions of weak temperature gradient and marginal measure- ments in regions of strong temperature gradient. This variation in quality was traced to the analysis procedure. An objective analysis program (described in paragraph 2.3) was developed, tested and placed into operation on May 1, 1974 and made possible the depiction of true gradients in the temperature field. 1.3 The Present The processing model used to obtain sea surface temperatures today is a fully automated operational set of computer procedures. These procedures are highly modular in design to permit component updating as improvements are made in technology and new retrieval methodology or to meet additional model requirements. This is the GOSSTCOMP (Global Operational Sea Surface Temperature Computation) model. To achieve an efficient combination of shared functions with other products, the data manipulation for the current version of the GOSSTCOMP model is deeply integrated with the total SR data processing system. The total system data flow path from the spacecraft to output product is lengthy and involves several complex component functions. Briefly, data from the SR sensor are recorded on the spacecraft analog tape recorders. The recorders are read out at a rate of 40:1 to either of the Command and Data Acquisition (CDA) stations located at Gilmore Creek, Alaska and Wallops Island, Virginia Orbital data on readout are recorded in analog form at the CDA stations and then transmitted to NESS at Suitland via microwave ground lines at a 2:1 rate; at Suitland the signal is converted to digital format and stored on magnetic tape (a more detailed discussion of the sensors aboard the NOAA series of satellites and the ground equipment used is contained in section 3) . The orbital data next are assembled onto NOAA central computer storage for automated processing. Surface temperature values are derived by a statistical histogram analysis of 1024 instrument measurements forming a square area approximately 100 km on a side surrounding the retrieval point. The detection of cloud free areas from which measurements are to be made and the determination of the quality of those measurements are ascertained from the histogram character- istics. Accepted retrievals are assigned a geographical location and cali- brated. Corrections for atmospheric attenuation in the SR sensor's infrared spectral window (10.5 to 12.5 urn) are computed from the VTPR sensor data and applied to the retrieval temperatures. The validity of each retrieval temperature is determined in a final quality-control comparison of the Northern and Southern Hemispheres. The current version of GOSSTCOMP was designed to handle the many problems of a fully automated processing system, the numerous technologies involved in remote sensing, and produce a high level of dependability for product delivery. During the last six months of 1974, a 97% success rate and an observation accuracy to within 1.5°C of ship observations were achieved. 1.4 The Future Quantitative measurements of sea surface temperatures are still hampered by factors of scale, residual cloud contamination, atmospheric moisture, and noise in the data. There are additional uncertainties in measurement accuracy, the physical characteristics of the problem (such as the effects of changes in surface emissivity) , transmissivity functions of the atmos- pheric gases, surface texture or sea state, and surface reflectances at varying incidence angles. Some of these problem areas are under current analysis. Some of the answers will come only with the development of new sensor systems. There are great hopes for the TIROS-N spacecraft scheduled for launch in 1978. The digital recording systems and multi-spectral sensor system planned should greatly improve sea surface temperature retrievals; spatial reso- lution should improve, noise should be reduced, and there should be more accurate compensation for atmospheric attenuation. If microwave sensors are aboard, it may be possible to obtain surface temperatures through cloud cover and data for ice boundary analysis. Techniques for dealing with these problems have already been proven feasible by the research community. Revisions are being made in several parts of the processing system such as separation of retrieval techniques from dependence on a filtering field, development of real-time orbit-by-orbit applications for more critical 4 analyses, and the provision of higher resolution analysis for specific geo- graphical areas. 2. SST MODEL THEORY 2.1 Radiometric Measurement of Sea Surface Temperature The techniques used to obtain sea surface temperatures from the IR sensor data are discussed in this section. Basically, sea surface temperatures, T , are calculated from the measured radiance using the relation T s = T bb + AT ■ (1) where AT is the correction for atmospheric attenuation and T^ is the ob- served (measured) blackbody temperature. Functional procedures necessary to obtain sea surface temperatures from radiometric measurements include defining sensor response characteristics, performing data calibation, deriving clear column measurements, and correcting for atmospheric attenu- ation. 2.1.1 Sensor Response and Calibration The infrared channel of the NOAA scanning radiometer is sensitive in the spectral "window" of 10.5 to 12.5 urn. The input energy collected by the instrument (bolometer) is a function of the integrated radiation flux from the emitting surfaces of the viewed scene, the atmospheric emitting and absorbing gases, and the spectral response function of the sensor filter. The bolometer total response energy, N, in the spectra, (\i ,\i) , is the integral product of the input energy, I, and that of the sensor filter response function, (f>(A) . Mathematically, this can be expressed as •A 2 N = \ I(X)(A) the relative response function of the sensor, C the radiometer output when looking at a scene of equivalent blackbody temperature T, and m and b are calibration slope and intercept parameters calculated during processing. The value of m is given by _ N(T bb ) - N(T gp ) , c bb " c sp where T sp is the radiative temperature of deep space, N(T b b) is obtained from eq (3) evaluated for the known reference blackbody temperature, and N(T sp ) is the energy value given by eq (3) evaluated for 3.5K, the radiative tempera- ture of deep space. In practice N(T sp ) is taken as zero energy. Cbb is the radiometer output when it views the on-board blackbody at the temperature T bb , and C S p is the radiometer output when it views deep space. Having evaluated m, the value for b is obtained by substitution in eq (4) . When viewing space, N(T) is zero; therefore, = m C sp + b , (6) and it follows that b = -m C sp . (7) Substituting the values obtained for m and b from eq (5) and (7) in eq (4) , radiometer outputs are converted to energies, N(T). These are converted to equivalent blackbody temperatures by inverting eq (3) . Figure 1 contains the computed temperature-to-energy relationship of the NOAA-4 sensor 1. The thermal calibration values of m and b and the polynomial coefficients of the electronic calibration are computed for each data frame. Con- sequently, the equivalent blackbody temperature of any scene is derived by first converting the raw digital value to energy by means of the calibration algorithm N(T) = m (a + a 2 C) + b , (8) where ao and a.\ are the coefficients of the electronic calibration voltage- to-digital value relation, and C is the raw digital value returned by the sensor. 2.1.2 Sea Surface Temperature Measurements The input radiant energy of any scene viewed by a sensor is a function of (1) radiation emitted from any surface such as land, water, clouds, ice and snow; (2) radiation absorbed and radiated from atmospheric components; and (3) radiation reflected from the surface. The optical properties of the ocean surface in the 10.5- to 12.5 urn window within viewing angles of 0° to 60° approach those of a blackbody radiator. Current estimates of the emissivity of the ocean surface range from 0.967 to 0.997. Assuming a value of unity for the ocean-surface emissivity permits dropping the reflected atmospheric radiation term since there can be no reflection from a blackbody. The magnitude of reflected radiation, if other than unity, is small. The ability to eliminate cloud contaminated scenes is the most signifi- cant factor in determining the eventual accuracy of surface temperature values. Clouds are themselves radiating surfaces. Clouds absorb the radiation emitted from the ocean surface and emit radiation as specified by the Planck function. In a single-channel model, no effective means are available to infer a surface radiance value in cloud contaminated scenes. For this reason the viewed scene must be limited to cloud free areas. Considerable effort has gone into the design of the sea surface temperature model, both in the retrieval process and in post analysis, to eliminate any cloud contaminated brightness temperatures (scene energy converted to temperature using Planck's relationship). Details of these procedures are given in section 3 of this document. Mathematically, the component sources of radiation in a clear atmosphere can be written as: ifi c o m © >- e> tr uj u o 22 21 20 9 3 7 6 5 4 3 2 I 9 8 7 S 5 4 3 2 \ WINDOW ENERGY TO TEMPERATURE r N (T) = J £(\,T) ^(XtfX NOAA 4 SENSOR I J I I I I l__L J I I I I I I I L — — — — NfONfONNfV>ro M N W M w SSSSgogggggggggog TEMPERATURE (K) Figure 1. --Window energy vs . temperature curve for NOAA 4 sensor 1. 'A 2 N s = I e (X)3(X,T s )dX , (9) f 2 , Jxi s where N s is the total energy emitted by a radiating surface in a bandwidth (Xi,X 2 ) at a temperature T s , (3 is the Planck radiance, and e s is the surface emissivity, and N a = 1 g(X,T(P)) dT ( X » p ) dPdX , (10) JXiJPo dP where N a is the positive energy contribution of the atmosphere in the band- pass (Xi,X2), T(P) is the temperature of the atmosphere at pressure level P, and t(X,P) is the atmospheric transmittance from pressure level P to the top of the atmosphere. P is the atmospheric pressure at the surface. The total energy, N(T), over the bandpass measured by the sensor viewing an ocean surface is then given by N(T) = N s t s + N a = N s - (N s a s - N & ) , (11) where a s and x s are the absorptivity and transmittance, respectively, of the total atmosphere. Expressed in terms of equivalent temperatures one obtains T bb = T s - AT , (12) where T bb is the sensor observed brightness temperature, Ts is the surface temperature, and AT is the atmospheric correction. Since T s is the value to be obtained, the equation is usually expressed as in eq (1). The value of AT is obtained from the VTPR soundings. 2.1.3 Atmospheric Attenuation Although the 10.5- to 12.5-um region is a window for longwave radiation, it is known that radiation emitted by the earth and clouds in this spectral band is slightly attenuated by the atmosphere. The principal absorbers in this region are water vapor, carbon dioxide, ozone, and aerosols. The extent of attenuation is a function of absorber concentration in the radio- meter field of view. The accuracy of sea surface temperature measurements is highly dependent on the correction which must be made for the effects of these absorbers. Estimates of the magnitude of these corrections are given in table 2-1. 10 Table 2-1 10.5- to 12.5-um atmospheric attenuation corrections Absorber Range H 2 0° to 9.0°C CO 0.1° to 0.2°C 3 0.1°C Aerosols 0.1° to 0.95°C Attenuation by water vapor far exceeds that by any of the other absorbers. Other residual absorption, such as that produced by salt and dust particles, has little effect unless there is high particle concentration. Ice crystals near the tropopause are assumed to be opaque. Brightness temperatures associated with sea surface temperature measure- ments are corrected for atmospheric attenuation with coefficients derived from VTPR processing of a coincident temperature and moisture profile sounding. The coefficients are derived by integrating the effect of ab- sorption and emission of radiance by each layer of the atmosphere. This integration is based on both the VTPR sounding and theoretical trans- mittance functions. The computed coefficients are used to correct both for the straight-down single atmosphere absorption and the absorption through a slanted atmospheric path. Further details on this calculation procedure are contained in NOAA Technical Report NESS 65 (McMillin et al. 1973). Mathematically, the relationship of atmospheric attenuation is given by AT = AsecO + Bsec 2 6 , (13) where A and B are the calculated VTPR coefficients and the viewing angle measured from the satellite nadir point. Until VTPR data were available, an empirical correction for the atmos- pheric absorption was used. Statistical correction values were determined from a comparison with a conventional sea surface temperature analysis. The statistical correction was expressed as a function of observed bright- ness temperature and the satellite viewing angle. In effect, the inclusion of the observed brightness temperature in the functional relation gives a coarse approximation of the temperature-pressure state of the absorbers. This procedure is still used as a back-up to VTPR data processing. Table 2-2 shows the empirical mean atmospheric correction values in terms of observed brightness temperature and zenith angle. ] I Table 2-2--Empirical mean atmospheric correction. Brightness Correction in K for loca] L zenith angles temperature (K) 7 14 21 28 35 43 51 270 3.05 3.09 3.18 3.32 3.50 3.74 4.03 4.40 271 3.12 3.15 3.24 3.38 3.56 3.80 4.09 4.46 272 3.18 3.21 3.30 3.44 3.62 3.86 4.15 4.52 273 3.25 3.28 3.37 3.51 3.69 3.93 4.22 4.59 274 3.31 3.35 3.44 3.58 3.76 4.00 4.29 4.66 275 3.39 3.43 3.52 3.66 3.84 4.08 4.37 4.74 276 3.47 3.51 3.60 3.74 3.92 4.16 4.45 4.82 277 3.55 3.59 3.68 3.82 4.00 4.24 4.53 4.90 278 3.63 3.67 3.76 3.90 4.08 4.32 4.61 4.98 279 3.71 3.76 3.85 3.98 4.17 4.40 4.69 5.06 280 3.81 3.86 3.95 4.08 4.27 4.50 4.79 5.16 281 3.90 3.95 4.04 4.17 4.36 4.59 4.88 5.25 282 4.00 4.05 4.14 4.27 4.46 4.69 4.98 5.35 283 4.10 4.15 4.24 4.37 4.56 4.79 5.08 5.45 284 4.19 4.25 4.33 4.46 4.65 4.88 5.17 5.54 285 4.30 4.36 4.44 4.57 4.76 4.99 5.28 5.65 286 4.41 4.47 4.55 4.68 4.87 5.10 5.39 5.76 287 4.52 4.58 4.66 4.79 4.98 5.21 5.50 5.87 288 4.63 4.69 4.77 4.90 5.09 5.32 5.61 5.98 289 4.74 4.79 4.88 5.02 5.20 5.43 5.72 6.09 290 4.87 4.92 5.01 5.15 5.33 5.56 5.85 6.22 291 4.99 5.04 5.13 5.27 5.45 5.68 5.97 6.34 292 5.12 5.17 5.26 5.40 5.58 5.81 6.10 6.47 293 5.24 5.29 5.38 5.52 5.70 5.93 6.22 6.59 294 5.37 5.42 5.51 5.64 5.82 6.06 6.35 6.72 295 5.51 5.56 5.65 5.78 5.96 6.20 6.49 6.86 296 5.65 5.70 5.79 5.92 6.10 6.34 6.63 7.00 297 5.79 5.84 5.93 6.06 6.24 6.48 6.77 7.14 298 5.93 5.98 6.07 6.20 6.38 6.62 6.91 7.28 299 and above 6.07 6.12 6.21 6.34 6.52 6.76 7.05 7.42 12 2.2 SST Retrieval Theory Original research on sea surface temperature retrieval techniques (pro- cessing sensor data to locate sea surface and measure its temperature) using scanning radiometer infrared (SRIR) data was done by W. L. Smith, P. K. Rao, R. Koffler, and W. R. Curtis (1970) . A modification of this original tech- nique, first suggested by W. L. Smith and R. Koffler (1970), was later developed into the present technique by J. Leese, R. Brower, B. Goddard and W. Pichel (1971a). The current method of retrieving sea surface tempera- tures from satellite SRIR data employs a statistical technique to extract sea surface temperatures from blocks of raw data organized into histograms. The technique described herein has been used without major modification since December 1, 1972. 2.2.1 System Characteristics Affecting Retrievals Numerous characteristics of the total data handling system, including satellite hardware, data transmission, ground equipment, and the computa- tional methods used to process the data have a direct impact on the quality and quantity of SST retrievals. Those characteristics which have influenced the development of the present retrieval method are outlined below. 2.2.1.1 Resolution and Sampling Rate. The incoming IR analog signal is sampled during digitization at a rate 3.5 times greater than the rate which would produce contiguous non-overlapping samples. Thus, there is con- siderable intrascan overlap. The movement of the satellite along its orbital track and the instrument scanning period of 1.25 seconds provide contiguous coverage with little or no interscan overlap. At the subsatel- lite point, the 5.3- milliradian (mr) instantaneous field of view (IFOV) and the satellite altitude of 1,464 km produce a resolution for each data sample of approximately 7.4 km. The area represented by a data sample increases from the subsatellite point to the horizon, so, at a local zenith angle of 60°, the IFOV is about 12.9 by 19.3 km (Fortuna and Hambrich 1974). In addition to the earth scene, each scan line contains space view, instrument housing view, and thermistor telemetry data used for electronic normaliza- tion, calibration, and system noise determination. 2.2.1.2 System Noise. The measurement accuracy of the SR instrument is expressed as a noise equivalent temperature difference (NEAT) . The NEAT is the change in energy input from a scene (expressed in temperature units) which results in a change in instrument output equal to the root mean square (RMS) noise level, i.e. the NEAT is a function of instrument temperature and scene temperature. Instrument noise and its NEAT increases with increasing instrument temperature. However, as the scene temperature increases, the NEAT decreases (i.e. it is significantly better for warm scenes than for cold ones), because at higher scene temperatures a smaller temperature change is required to produce a fixed change in energy. This is a result of three factors: (1) the instrument response is linear with input energy, (2) input energy in the 10.5- to 12.5-um window is approximately proportional to the fourth power of temperature, and (3) the instrument noise is not a function 13 of scene temperature. For a nominal case, with an instrument temperature of 25°C and a scene temperature of 27°C, the NEAT is approximately 0.3°C. The NEAT increases to 1.4°C for the same instrument temperature when the scene temperature is -88°C (Schwalb 1972) . These NEAT values are representative of the instrument only. The total system NEAT is significantly larger because of additional noise introduced by on-board tape recorder fluctuation, data transmission, and analog-to-digital conversion. For the NOAA 2, 3 and 4 satellites, the space view standard deviation (which is considered to be equal to the system RMS noise level) has generally fluctuated within a nomi- nal range of 1.75 to 3.5 digital counts with most orbits showing a mean system RMS noise of between 2.0 and 3.0 counts. A system noise of 2.5 counts results in a system NEAT of 1.2°C and 3.5°C for scene temperatures of 27°C and -68°C, respectively, when the SR instrument temperature is 25°C. (See data in Adams (1972).) The system noise is considered to be random, even though this sometimes is not the case. The noise characteristics are continually monitored and adjustments to the ground equipment to compensate for tape recorder fluctu- ation (Z-axis correction) periodically are required to filter out non-ran- dom noise. 2.2.1.3 Earth Location Accuracy. The ability to locate the SR data accurate- ly has a bearing on retrieval procedures and the overall accuracy of the system. It has been estimated (Conlan 1973) that errors in earth location are kept within 20 km along the orbital track and 40 km for local zenith angles above 50°. 2.2.1.4 Raw versus Mapped Data. When one considers using satellite data, a choice must be made whether to use raw data or calibrated and mapped data. The latter data are much easier to work with since earth location, electronic and thermal calibration, and compensation for limb darkening have all been applied beforehand. However, in the mapping and calibration process, the volume of data requires the use of integer "look-up" tables with their attendent round-off and approximation problems. Attainment of the highest quantitative accuracy requires the use of raw data and high precision in the calibration, earth location, and atmospheric attenuation of these data. 2.2.1.5 Cloud Cover Percentage. On any given day, approximately 40 to 50% of the earth is obscured by clouds (Shenk and Salomonson 1972) . The per- centage of single data elements contaminated by clouds is larger than 50% because a data element containing even a few clouds must be classified as cloud contaminated. As the size of the data spot increases with zenith angle, the probability of it containing some clouds increases. There are even greater probabilities that aggregates of data spots will be con- taminated. Thus, to achieve global sea surface temperature coverage, any retrieval method working with aggregates of data elements must be capable of retrieving an accurate temperature, even though some of the data elements are contaminated by clouds, i.e., retrievals must be possible even in partly cloudy regions. 14 2.2.2 Rationale for Retrieval Technique Developed Although the primary mission of the ITOS satellite is meteorological in nature, it was recognized even before the first ITOS was launched that it would be possible to retrieve sea surface temperatures on a global basis daily. A tentative objective was established to achieve global coverage with retrievals from areas 100 km on a side and an RMS deviation from ground truth of 1.0°C. This objective has not been fully achieved with the present ITOS satellites; however, it is expected that the TIROS-N series will meet this goal. Working under the constraints imposed by the characteristics of the SRIR system, and keeping in mind the oceanographic objective proposed, it became evident that the retrieval process to be used should follow the guidelines outlined below: (1) Retrieve from raw data rather than from mapped data. The data must be calibrated and corrected for the effects of atmospheric attenuation as precisely as possible to retrieve sea surface temperatures to an accuracy of 1°C. Thus, the technique developed retrieves from raw data without using look-up tables for calibration. (2) Obtain retrievals from blocks of data rather than from individual samples. The large system NEAT necessitated the use of some averaging technique if there was to be any possibility of achieving an accuracy of 1.0°C. In examining the tradeoff between resolution and accuracy, it was found that a block of data containing 1,024 samples was sufficiently large to achieve the required accuracy. In the retrieval process, histograms are constructed from raw data blocks, each containing 64 samples from 16 adja- cent scans. Thus, each block of data contains 1,024 samples and at the sub- satellite point (because of overlapping samples along a scan line) , is approximately 100 km on a side (119 by 140 km). (3) Use data samples with local zenith angles less than 60°. Because of expansion of the IFOV at higher zenith angles, a statistically signifi- cant sample covers an unacceptably large area at local zenith angles greater than 60°. Therefore, such data samples are not used. Figure 2 shows the arrangement of data blocks from a 16 scan line frame of raw data. There are 15 blocks ranging in dimension from 119 by 140 km for block 8 at the sub- satellite point to 119 by 313 km for blocks 1 and 15 centered at local zenith angle of 51°. (4) Use a method independent of the system noise. Since the system NEAT varies extensively between orbits and within each orbit, it is necessary to seek a retrieval method which can make observations under a range of system noise. The statistical technique presently being used is independent of random system noise within its normal range for the ITOS SR system. (5) Take advantage of the randomness of the noise. Since the system noise is sufficiently random, the statistical technique used is based on a Gaussian distribution. (6) Be able to retrieve in partly cloudy area. The low probability of 15 rC 4-> T3 -J, E 4-1 C ^1 O o +-> 03 Q £I£ Zt73 103 SAT 8ST 6t>I £H OVl 1VI 6t7l 8SI SZI I0Z £T£ lo co n a, vO m -r CO CM LO to CD CD 13 E r-Ol u 4-> c o o o GO 03 a o a> X CD c u X o o 4-> 5h 3 !/) o CT E m o3 0) s +-J !m c oo 03 4-1 '4-' M "3 u < X o -r o rH r i o rO c ^ 1 — 1 rt i — i tfl +-> CD 03 rt t.0 c 4-> -3 c •H 03' •H i— 1 T3 m cti o 4-> 03 03 i/i o u C c o 00 •H o X •H u. >-o +J CO u I — 1 •H c o S 00 a; r— 1 m E E X) o (/) rC •H 05 4-> "3 X X - — i ■H 'J U 9- h T3 03 03 E o C UJ 0} 03 to 03 E 1/1 1—1 03 ■M • O O C 03 •m X i — i 0) 4J Uh 4-> 03 E 03 > o -d tfl r— I +J rj CH 3 o O 00 x 4-> d> CD •H U <1) r— 1 h 4-J 3 E P-, +-> 3 00 03 E 00 Ph • H h 03 •H c PL, IH 00 T3 •H CT. 16 having a completely clear area of 100 km on a side would make it impossible to achieve global coverage without compositing for long periods of time if retrievals could be obtained only from areas with no clouds. The technique now used can retrieve sea surface temperature in data blocks that are up to about 30% cloud contaminated by exploiting the general rule that clouds are colder than the surface. 2.2.3 Retrieval Algorithm For a uniform-scene temperature field, with only random noise affecting the sensor output, a frequency distribution of the data elements from this field will be Gaussian (fig. 3a). Simply calculating the arithmetic mean or locating the modal class of this histogram will give one the actual scene temperature. Since most scenes have some clouds present, one must be able to find the true surface scene temperature when cloud contaminated samples are present in the histogram. Figure 4a, a data histogram contaminated by clouds, is typical of many partly cloudy data blocks. The warm side of the histogram (smaller IR counts) is unaffected by the clouds. If one could find an uncontaminated histogram whose cold side is a mirror image of the warm side, the mean would be the surface scene temperature. The following formula can be derived from the normal density equation. It will yield the mean of a normal histogram if the frequencies for any three classes are known (T 3 2 -T 1 2 )ln(F 1 /F 2 )+(T 1 2 -T 2 2 )ln(F 1 /F^ T s = , (14) 2(Ti-T 2 )ln(F 1 /F3)-2(T 1 -T 3 )ln(F 1 /F2) where T s is the mean of the histogram, and F\, Fn, F 3 are the respective frequencies for any three classes, Tj, T 2 , T 3 , of the histogram. This makes it possible to find the uncontaminated mean of a histogram from a partly cloudy area by using only the classes on the warm side. Also, so long as system noise is random, the accuracy of the calculated mean will not be compromised as system noise increases. Because there is a certain degree of non-random noise in the data, two operations are performed to insure the accuracy of the calculated mean: (1) The histogram is smoothed before attempting to retrieve a mean; and (2) since a single calculation of the mean from a slightly non-Gaussian histogram could be in error, repeated calculations of the mean are made from all possible combinations, taken three at a time, of the classes (each class being one IR response count wide) constituting the warm side of the histogram. These estimates of the mean are distributed into a second histo- gram called the mean estimate histogram (MEH) . The uncontaminated mean of the data histogram is then calculated from the MEH by finding the mean value of all mean estimates which fall within two classes of the mode of the MEH. Figure 4b shows the MEH for fig. 4a. 17 — 190.0 — 180.0 < cr — 170.0 CD O — 160.0 I— CO — 150.0 •— X ID -140.0 2 LU i— — 130.0 _ < 51 oo r^. u-i O -120.0 1— CO UJ uj cr 3C 1— — UJ r .., ,, r , . O O r ■■< - i r ooooooooo © m oi/"*ov»©u"*oi/~>o ■j 1 r«"> fi ini inj «— •— A:N3H03yd u GO •H O - [XI to —190.0 — 180.0 — 170.0 sz — 160.0 < cr CD — 150.0 H- ID -140.0 g CD I— CO — 130.0 _ I oo k in o -120.0 —110.0 -100.0 — 90.0 < h- < LATITUDE: 12 LONGITUDE: -66 ZENITH ANGLE: TIME: 56 16: 4:38 RETRIEVAL: 111 1 ■' o o o 1 , 1 , 1 1 1 o o o o 1 ' 1 ' 1 ' o o o ■: o o o O OO -O o o o o -J iNJ O oO o o o < •£> -*T INI O A3N3n03dd i/i o d e 3 o CD u +-> > X o cti •H +-> IH F-t t-J -H CD rt 1 e T) e CD XI to rt e 3 ■ H h i^ +-> OO o3 ■H CD oj CD (fl T3 +-> 03 3 cti to OJ 4-> T3 ■M ■H c 03 3 u 03 O o 03 l-H Fh ej o3 C/) X cc CD 4-1 fn hH 17 3 rt 3 O U3 r— i O ■ H u ^H DO • 1 i— 1 C H 1 — i 1 03 O 03 ^ • E^ i — I O 03 C/j" w to 0) CD Tl > T3 C CD • H =5 03 ^ T3 +J +-> 3 3 03 •H u M O DO 4-> 03 •H i — i 0j ri 0) U- o X - — l X -200.0 -190.0 5L" < 180.0 L3 CD — 170.0 co 160.0 1— ZD 150.0 g ~JZ L±J 1— ■-H0.0 _ < -4- f\J O O 130.0 CO r^ «— c"> i>o IXJ u~i -*T m r\j - 120.0 110.0 LU < LU 5L" LATITUDE: LONGITUDE: ZENITH ANGLE: TIME: 56 16: RETRIEVAL: 1 1 1 1 o o | i | i , i | i | i | i ooooooooo O 1/1 U-l -J- -j- f-> ft fsj rsi «— •— A3 N3n03bd CD - •H o '-H X X.' o U •H 200.0 190.0 — 180 — 170 < CD — 160 -150 h- ZD O (_) o I— CO 1— HO CE JZ 130.0 120.0 100.0 < vT INJ O < Q LD •o LU o < < O .- > Z> i — X LU h- — o — LU , O i— I -J :- ex -t 3 bo L9 2.2.4 Histogram Types and Tests The numerous sizes and types of clouds and the complexity of the sea surface temperature pattern in strong gradient areas produce an infinite variety of histograms. However, most data distributions can be classified into one of six general types. A series of tests have been devised to in- sure that retrievals result only from histograms from clear or partly cloudy areas. Yet, no matter how many tests are used, there will still be numerous retrievals that are erroneous because of the following considerations: (1) Uniform cloud layers can produce histograms of the same shape as those obtained from a clear atmosphere. (2) If the test criteria are made very restrictive, the accuracy of the retrievals may be high, but the number will be low. (3) There are no reliable parametric limitations except the minimum temperature of sea water. The use of a first-guess field to obtain tempera- ture boundaries for acceptable retrievals is possible, but not always desir- able. These factors necessitate the use of an objective analysis technique to further screen the retrievals to produce reliable SST observations. The tests described below are designed to eliminate a large number of the kinds of histograms from which poor retrievals are consistently obtained. However, a considerable percentage of the successful raw retrievals are still contaminated because the tests are not completely restrictive. Because a large number of observations are needed for a global analysis, it was found to be more beneficial to eliminate in the objective analysis technique (described in paragraph 2.3) only those retrievals with residual contamina- tion rather than lose accurate retrievals by using overly restrictive tests. 2.2.4.1 Histogram Tests. The following series of tests was developed to discriminate clear or partly cloudy cases from histograms contaminated by clouds. Pre-retrieval and post-retrieval refer respectively to before and after the creation of the mean estimate histogram. (1) Pre-retrieval gross contamination test -- There must be at least 800 samples in the distribution with temperatures above approximately 265 K (IR counts below 170) . (2) Pre-retrieval data histogram mode test -- The modal class of the data histogram must contain at least 50 samples. (3) Pre-retrieval warm side range test -- C w is that class in the distri- bution where the cumulative sum from the warm wing surpasses 25 samples. The number of classes between C w and the modal class must be less than eleven but greater than four. (4) Pre-retrieval stratus test -- C^ is that class in the distribution which contains the 25th sample in a cumulative sum from the cold wing. The number of classes between C]< and the modal class must be greater than the 20 number between C w and the modal class. (5) Post-retrieval MEH percent test -- At least 70% of the mean estimates must be within two class intervals of the modal class of the mean estimate histogram. (6) Post-retrieval "retrieval mode" test -- The surface temperature class retrieved from the MEH must not be more than one class interval colder than the data histogram modal class. 2.2.4.2 Histogram Types. The six most common types of histograms are listed below together with descriptions of each MEH type and the applicable histo- gram tests. Table 2.3 relates each histogram type with tests designed to reject those that will not yield an acceptable surface retrieval. (1) Clear area or uniform warm cloud histogram (fig. 3a) -- This type of histogram is characterized by a normal distribution with a MEH (fig. 3b), showing a spike at the mean of the data histogram. This type passes all pre- and post-retrieval tests. Only a first guess temperature will permit identification of this type of histogram as being from a clear area or a uniform cloud. (2) Partly cloudy histogram (fig. 4a) -- The MEH (fig. 4b) from this type has a good clustering of mean estimates calculated from the Gaussian warm side of the histogram. The cold side of the distribution reflects the type of cloud contamination and is generally very complex and non-Gaussian. This type of histogram passes all tests. (3) Histogram of unresolved temperatures (fig. 5a) -- Data blocks con- taining multilayer clouds, clouds very close to the surface temperature, or strong surface gradients produce histograms which are superimpositions of normal curves. Generally, the MEH (fig. 5b) will not have a distinct mode and will fail the MEH percent test. Often the gradient is so high that the warm side range test is failed. If the histogram is from a high-gradient clear area, the histogram will not pass the stratus test and will sometimes fail the retrieval mode test. (4) Histogram containing semi-transparent clouds or clouds smaller than the IFOV (fig. 6a) -- Because the sensor integrates the incoming radiation within its IFOV, when a cloud smaller than the resolving power of the sensor or a semi-transparent cloud which permits partial transmission of surface radiation lies within the data block, the temperature sensed will lie between the temperature of the surface and that of the cloud. This type of contami- nation results in a general cooling of the entire histogram, usually accom- panied by an increase in the standard deviation. Either the MEH (fig. 6b) will fail the MEH percent test or the data histogram will not pass the warm side range test. (5) Low level broken stratus histogram (fig. 7) -- A histogram from an area containing a uniform stratus deck with breaks allowing surface views is characterized by a normal cold side with a skewed warm side. The MEH 21 is similar to that resulting from a histogram of unresolved temperatures. The stratus test and sometimes the retrieval mode test will eliminate these histograms. (6) Cold or mixed cloud histogram (fig. 8) -- This type of data area is characterized by a very unorganized, non-Gaussian histogram with a large range. The MEH is also completely disorganized. The histogram fails the gross contamination test, the data histogram mode test, the MEH percent test, and the warm side range test. Table 2-3 Chart of histogram types and tests Histogram Types App 1 icable rejection tests (1) Clear area or uniform warm cloud None (retrieval accepted) (2) Partly cloudy None (retrieval accepted) (3) Unresolved temperature Warm side range test Stratus test Retrieval mode test MEH percent test (4) Semi-transparent clouds or Warm side range test clouds smaller than the IFOV MEH percent test (5) Low level broken stratus Stratus test Retrieval mode test (6) Cold or mixed cloud Gross contamination test MEH percent test Warm side range test Data histogram mode test 2.2.5 Behavior of Retrieval Technique The percentage of incoming histograms from which successful raw retrievals are made is very consistent from day to day. Table 2-4 lists retrieval statistics for March 2, 1975, and is quite typical. Fourteen orbits con- taining a total of 67,425 data histograms were processed. Of these histo- grams, a little over 1% were eliminated because of calibration inaccuracies and tape parity errors and 40% were not processed because they were obtained over land. Successful raw retrievals were obtained from 43% of the histo- grams over the ocean. 22 -230.0 51 i— -220.0 < or CO LU t— -210.0 CD o h- LU EC LU -200.0 a:' a. LU 190.0 t - X E 2j O =3 LU 1 d . . _ - o LU >— in ■ -1*0.0 £ (— < ■ a: - 170.0 _ < 2- u-> O OO O M 9 V- r**» r*i o r*"> ro >o i u~» i/t ^r ■H '4-' j - 160.0 *SJ rsl .. o LU «- '-M -^ -150.0 z '■ UJ ■ J -1 uj o < < O = > X 1 -g — HO.O LL => 1- I UJ |_ _ H- ■■ — s: — o — ai at 1 t- M K C 1- ■■ 130.0 < O LU — LU Xj r"T"'i""i"" ,,...,..... | . ,. . . ! .... { . ! 1 1 ■ 1 1 ii 1 I'll 1 1 1 LO oooooooo o o o OW^Ot/^OfcHOu^ © irt © CD m^r^t^ror'isi^ •" ?-. 3 ADNanoaud - .- bo -H LJU !/) CD r— 230.0 5h 4-> —220.0 CO 1- a3 U CD On 6 —210.0 2r LU a CD -200.0 ex. LU O. X -a CD > — 190.0 I— o LU o ^- o to ID r— LU CD 1-180.0 2 CO < U_ 3 ■-170.0 _-< X tfi O OO O o ^H Ik vO 1 U~> W* •& S Hl i — .. <— »- 160.0 <\1 a3 < f\J U ^Bii 1 Q LU «— bO O »-150.0 DE: UDE: ANGL 56 VAL: Hist H-no.o 3 \— X UJ l_ — . H- -• — ^^ _ O — UJ CC I I i— sc z x: i— < O LU — LU 1 7 n n 03 LO i , i , i" ' i i i ■ ' i '■ ■ i ' i ■' . .,..|..i |. , - - - 1 150.0 © o © o o © o o © © © © © © © o o o o © © © CD o «o tO -* f\J o «o *c A3N3n03Ud n*- rsj a •H 23 —200.0 s: h- — 190.0 < CO LU ■ - oc 1— h- — 180.0 CD o ^ Z LU tX — 170.0 CO Q- — 160.0 1— DZ X LU ar Q 1 -150.0 g q: LJJ I— < ■< LL. — uo.o _ ^ - (XI *- oO i — i o> lo *— OO OO I— "- l/"l u-i ro *— -130.0 CO ■* LJJ -o LU B- 12 0.0 •• o -o " < o LU Z Q < XI > -110.0 LU =1 >— I LU 1— — t— 1 Art A s: < O — z z O LU lu or x: 1— — LU t- ce 1 I 11 " 11 " o o 1""' "T ' T "' -■ o o o o o o I ' I _,uu ' u 3 o o © u~» O u~i O W> © i/"» Z> u~> O LO -^ ADNanogyd CTJ bO •H O X 43 — 200.0 — 190.0 CO LU 1— — 180.0 z -170.0 — 160.0 •— ID 5:: < cr CD O h- or LU Q_ X 3= O LU -150.0 2 CO < U_ 1 — 140.0 _ X _ f\J *— oO u~i «— OO OO ^ =—130 .0 < u~i in -4- -^ u^-120.0 1—1 10.0 Q LU O 1— >— < LU • ■ CJ3 LU Z Q < X> h- X — 1— O — O LU > LU lu cr C 1— — UJ r , "<""i o o 1 ""I"" 1 ....,....,....l....,....,....,....,....,----, ■-,■■■■ oooooooot 3 o o O oo oooooooot O^rrsJOoOO^" 1 ^ A3 N3n03dd 3 >" CD PL, 1 — 1 DO c c • H CD C X •H +-> CD 4J rH C CD O I— 1 u r— 1 CD 03 £ CL> CO U CD CO -rt s 3 O o fH i — i t+H o f= fH cD o H bO CO O 13 fJ 3 CO O • H rH X O CD 4-> +J c CD CD Q fH i CD i Cu • trt CD c vC CD fH CD ■M in 1 3 •H b-0 £ •H CD C1h CO 24 p230.0 -220.0 i— t— -210.0 21 2 O 1— —200.0 < en < Z E < J — 190.0 1— o ^ O A ZD (-- LJ A -180.0 g en to CO o 0£ cr =c CD m — 170.0 _ < I — 41 .5 -39.5 51 .0 : 13 AILED M — 160.0 < m — 150.0 DE- UDE: ANGL VAL: ^ — UO.O TITU NGIT NITH RE: TRIE 130.0 < O LU — UJ 1 ' 1 1 : 1 II _i _i rvj t— cc o o ooooooooo o o ooooooooo O oO *0*TiNIOoO*0>riNI a: N3no3y d o "3! O ■H -J 4-1 O o Q i 00 o U •H 230.0 -220.0 -210.0 s: 200.0 < ■190.0 l— C_D o UJ. f- ^ p =1 1— p < -180.0 2 CO or CO Q cr in _J - 170.0 _ in o oO < ti- < ■o ro o ro h- — 160.0 < INI 150.0 UJ o " UJ o < ^ < Q ID > -UO.O 1— LU LU CC 1= t— 130.0 < O rsj h- CC 1 1 1 j II 1 ' I : 1 ' 1 I ~* O o oooooooc 3 O o o oooooooc 3 o «o rt i i 3 25 Table 2-4 -- Retrieval statistics 3/2/75. Total Percent of Percent of histograms total from which retrieval histograms w as attempted Total histograms 67,425 100 Failed calibration quality checks 465 1 Failed due to tape parity errors 105 Over land 26,848 40 Histograms from which retrievals 40,007 59 was attempted Failed gross contamination test 7,098 11 Failed data histogram mode test 3,908 6 Failed warm side range test 6,064 9 Failed stratus test 2,751 4 Failed MEH percent test 1,580 2 Failed retrieval mode test 1,368 2 Successful raw retrievals 17,238 26 100 18 10 1.5 7 4 3 13 20 2.3 Rejection of Erroneous Data The basic tool for quality control of satellite derived, sea surface temperature data is an objective analysis technique (OAT). It screens the data by means of two distinct processes, one static, the other dynamic. The SST field produced by the analysis is used as a standard against which satel- lite retrievals are compared (static process) . During the actual analysis run, retrievals are rejected as non-representative (dynamic process). Only after satellite reports have been subjected to both procedures are the data considered of sufficient quality to be disseminated. 2.3.1 Static Control The SST field produced by the objective analysis program on the previous run is the main comparison standard for two tests, a gross error evaluation and a high moisture coefficient test. The tests are designed to remove those retrievals which are cloud contaminated or have poor corrections for atmos- pheric attenuation. A cartesian grid overlay on some cartographic projection (presently polar stereographic) forms the SST field. For each intersection of grid lines, (i,j), there is a sea surface temperature, Tjj , a measure of surface gradient, Gij , and a measure of age, Ajj , since Tjj was last modified by satellite in- formation. These three parameters are only defined for integral (i,j). Ajj and Gij are extended to any point of the grid using the value of the point at (jx+.5J, |y+.5|) in place of (x,y), where jz| is the largest integer smaller than or equal to Z. The temperature Tjj definition is extended by interpolation to the desired point. A satellite retrieval, Rxy, located with respect to this grid at point (x,y) is subjected to the following control tests: (1) For R xy 8] , or (b) T xy -Rxy5G xy • (17) The retrieval is rejected otherwise. 27 (2) For R X y>T X y, the retrieval is rejected (warm side gross error check) if R xy " T xy > Max(2.5°C , G xy ). (18) Note that the deviation of the retrieval from the SST field may be greater in areas of stronger gradient and may still be accepted. The cold side bound of 1.2°C is more restrictive than the warm side bound of 2.5°C, because the actual retrieving process has a cold end bias caused by cloud contamination. This stringent restriction is relaxed if no new information is incorporated into the field at the point of interest (refer to eq(16)). In practice the warm side bound is made more restrictive to form the high moisture coefficient test. (3) For R xv > T X y (high moisture coefficient test) the retrieval is rejected if {"■ Max[0°C,(Rxy-Uxy-E xy )]) nq , R xy -T xy >Max ^0.5°C,Max(2.5°C ; G X y) where U xy is the attenuated temperature of R xy , and E X y is an empirically derived estimate of the atmospheric attenuation for the report R X y. ( cf . Table 2-2 and paragraph 2.1.3) 2.3.2 Dynamic Control The primary purpose of the objective analysis technique (OAT) is to pro- duce an SST field for use in static quality control. However, in generating this field from retrievals passing the gross error and high moisture coeffi- cient tests, further anomalous data are detected and rejected. This detec- tion is termed dynamic control. 2.3.2.1 Conceptual View of Objective Analysis. The better the sea surface temperature field used as a first guess field, the better is the screening of retrievals. An objective analysis should include (1) The use of a reliability factor for each retrieval (and a resulting reliability for field temperatures); (2) the incorporation of climatology into the field to maintain data sparse areas over prolonged periods of time; (3) the removal of anomalies from retrievals before actual field tempera- tures are computed, i.e., a contiguity check on incoming data; and (4) the incorporation of the data into an SST field dependent on gradients of the field so that the effect of a retrieval on surrounding points of the field decreases with stronger gradients. It was found that all the above considerations could be accounted for satisfactorily by using weighted averages exclusively throughout the process. By using an averaging technique, retrievals within a specified area from each grid point can be combined to yield a sea surface temperature (Tij) for 28 each grid intersection over water. In this way, randomly located obser- vations which do not map to every grid point in a 1-to-l or multiple-to-1 relationship can be interpolated to update the field. 2.3.2.2 Search Procedures. Associated with each grid point in the SST field to be merged with the latest data is an area which determines the retrievals to be used in calculating the new Tjj at that point. Only those retrievals which map into this area are considered in the weighted mean. The extent of this area must depend on the thermal gradient. (See example in figures 9a and 9b. Figure 9a--Weak gradient on geo- graphical grid. Figure 9b- -Stronger gradient on geographical grid. The temperatures at points Pi and P2 are the same for both figures. If the gradient at point Pi is defined only in terms of the information of Pi and ?2, it does not represent the thermal pattern of the surrounding areas. Con- sequently, a gradient is defined by using the information from all the points, Pi through P7, with each unit box having an associated weight. In this way, the area of search for Pi in fig. 9a will be smaller than that of fig. 9b in the horizontal direction (X + ) from Pi to P7. More precisely the gradient at point (i,j) in the positive x direction is G f] k=l[w(k)M(i,i)|T i+k j-T it1 |] 6 Ew(k) k=l (20) 29 where M(i,j) is that function which normalizes the distance represented by one grid unit to some standard if the grid is not an equal area projection (reducing the dependence of G x+ on the cartographic projection employed), and w(k) is the weight assigned to each grid box used in the gradient calculation. More specifically w(k) is 6,6,3,1,1,1 for k = 1,2,3,4,5,6. And Tij is the temperature at point (i,j). Gradients in the remaining three directions x~, y + , y" are defined in an analogous manner. The area to be searched can now be based on these gradients since the re- sulting figure will reflect the thermal pattern. The most successfully used, empirically derived function that assigns a distance in each direction (based on the gradient) is x+ ij K |g x+ /s|+i (21) where K, the maximum distance in kilometers to be assigned under any circum- stance, is set to 600 km, and s is a scaling factor (presently 0.8). The distance in the other three directions can be defined similiarly. The region used in deriving the sea surface temperature at point (i,j) is the rectangle with lower left corner (i-D*^ , j-Dyo and upper right hand corner (i+D*j, j+D^j) with the further constraint that the distance between any point (x,y) in this rectangle and point (i,j) be less than the maximum distance searched in the four directions, i.e. [(x-i) 2 + (y-j) 2 ] 7 < Max(D.-,D X .,D y :,D^t) . ■ LJ 1] !J !J (22) The associated area becomes a circle if all distances (gradients) are equal. Fig. 10 shows two typical search patterns. Figure 10a--Search pattern, weak gradient. Figure 10b- ■Search pattern, strong gradient 30 The extent of the pattern is limited so that it never crosses land barriers. 2.3.2.3 Adjacent Data Rejection Test. Retrievals are compared against themselves before being merged with the SST field to form the updated SST field. Information found to be anomalous is rejected and not used in the analysis. For a retrieval, Rxy, the procedure consists of calculating a preliminary SST at point (x,y), using all ambient retrievals in the asso- ciated area to (x,y) and excluding the temperature measurement in question. A comparison of this SST with R xy determines the quality of the measurement and its subsequent removal. For R xy , the retrieval of interest, generate a search area about point (x,y). Let R uv be a retrieval that lies within this boundary. The tempera- ture at point (u,v) affects the resulting temperature at point (x,y) . Con- sequently, some measure of the reliability of R uv is needed. This confi- dence value assigned to R U v depends on (1) Any previously known reliability C uv of R uv ; (2) the average gradient at point (u,v) G uv [«)^«0 2 ] 1/2 . C23, Certainly, a higher gradient would reduce the accuracy of estimating the change in temperature at point (x,y) . Further, by reducing the reliability of a point found in a high gradient area, the search area can be viewed as being weighted to follow the thermal pattern of the SST field more closely; and (3) the distance, P, from the points (x,y) to point (u,v). Taking the above into consideration leads to a formulation of the weight, W uv assigned to R uv with respect to some reference point(x,y), u, _ ^uv r W uv " , (24) G* P r u uv where F is a scaling factor for the grid resolution used (F=l is used at present), r is a constant set to 2, P is restricted to F if P is less than F, and G* is as follows: ' uv G * v = 1, if G UV <1; G uv = G xy if G uv otherwise G uv = G uv These restrictions guarantee that a retrieval found in searching from point (x,y) always has a weight less than a retrieval at point (x,y). Note that 31 even though the notation W uv does not explicitly mention the base reference point, it has meaning only with respect to this point. For each retrieval, R uv > found in the search area, a corresponding relia- bility W uv can now be defined. With this information a weighted average of all retrievals found, excluding the one of interest, R , is performed; this yields a preliminary temperature for point (x,y). More precisely: T xy = ^(W uv R U v ) ^ (2^ where the summation E represents all retrievals within the associated area of x,y, excluding the retrieval at x,y. In practice a modification to eq(25) is made: s _ ^[Wuv( Ruv-Tuv)] - (26) S X y - _ z w u vv uv By using the difference function rather than an absolute temperature measurement, a more conservative approach is realized. Temperature changes represented by retrievals are examined only in relation to the SST field. Consequently, changes to the field are less intense for eq(26) than for eq.(25), A retrieval is rejected when compared to its adjacent data if l S xy-( R xy-T X y)|>dl (27) The report R is accepted regardless of eq(27) if it deviates only slightly from the preliminary SST field, l T xy- R xyl (28) or if the report reliability is suffiently high, C xy >d 3 . (29) If no retrievals are found in the search area except the one being tested, a judgment is made solely on (28) and (29) . Best results are achieved when applying this "neighbor check" if multiple passes through the retrieval data are performed, restricting the values of dj, d 2 , d 3 each time accordingly. The following chart depicts current parameter settings. 32 d l d 2 d 3 First pass 1.0°C 0°C unused Second pass 0.5° 0.5° unused Third pass 0.4° 0.7° unused All retrievals that satisfy the neighbor check are considered valid obser- vations and represent the primary product of the project. 2.3.2.4 Final Merge. For each grid point, all observations located in the area determined by the gradients of the SST field (and which have passed all data rejection tests) are merged with the grid temperature. This merge forms the SST field for the analysis run and in turn the input SST field for the next run. The change to grid point (i,j) required by new observations is ATJ-j = E [ w uv( R uv" T uv)] , (30) ij m uv where the summation is taken over all observations found in the associated area to point (i,j). However, some grid points will have no new information for the analysis run. To overcome this and to maintain the validity of such areas over extended periods of time, 3% of the difference between the succeed- ing month's climatology, T?- and the grid point temperature is included in the merge, AT-' • = •03Wc(T^-T ij ) + Z[W uv (R uv -T uv )] % (31) 1J W C +ZW UV where W c is the associated reliability of climatology, (W c = 0.4). In this way, a grid point with no satellite information for 30 days will be changed to a purely climatological value. If, at any time during the month, new satellite information is inserted at the grid point, much of the effect of climatology will be removed because the weight of climatology, W c ,is chosen to be small compared to a typical value of W uy . Eq(30) can now be extended with the summation including climatology so that all grid points are updated. The final merge temperature, T-jj , is the weighted average of today's difference function, AT*_j, of weight, £W UV , with the change to the field as dictated from yesterday (i.e. zero) of weight Wj_t. T * _ £[W UV (R UV -T UV )] (32) ij « ij ' where the summation includes climatology. W|-j varies depending on its rela- tion to EW UV ; this favors the elimination of old data. Here 33 W*.=0, if W-^Dtf • and (33) W ij =W ij' if W ij> ZW uv ■ ( 34 ) The associated reliability for the new SST, T- • is, W ij = W iJ +DV uv > for the condition of eq(33) (35) 2 and t w ij = Zw uv for tne condition of eq(34). (36) W— and T^-: on next analysis become W- ■ and T^i respectively. 34 3. SST PROCESSING SYSTEM 3.1 Spacecraft and Ground Equipment The overall flow for Scanning Radiometer (SR) data from which sea surface temperatures are retrieved is shown in figure 11. Data sensed by the radio- meter are recorded on the spacecraft tape recorders and played back upon command to one of two data acquisition stations. Data are then relayed to a central site, processed in a large computer, and made available to users. This section presents a very brief description of the spacecraft and ground equipment for users unfamiliar with the present system. Only those portions of the hardware system applicable to SR data are discussed. For a more detailed explanation of the hardware and for information on the other sensor subsystems, see Schwalb (1972), Fortuna and Hambrich (1974), and McMillin (1973). 3.1.1 NOAA Spacecraft The Improved Tiros Operational Satellite (ITOS) is a second generation operational meteorological satellite. The ITOS is launched by the National Aeronautics and Space Administration (NASA) and, when turned over to the National Environmental Satellite Service (NESS) , is given the operational name "NOAA". Orbiting at a nominal altitude of 1,464 km, the NOAA satellite is in a sunsynchronous polar orbit with a nominal inclination of 101.7°. Circling the earth every 115 minutes, it always crosses the equator south- bound at a local mean time of 9:00 a.m. The NOAA satellite (fig. 12) carries the following sensors (redundant systems are used for reliability) : (1) Scanning Radiometer (SR) : described below. (2) Vertical Temperature Profile Radiometer (VTPR) : described below. (3) Very High Resolution Radiometer (VHRR) : a two-channel scanning radio- meter similar to the SR, with a resolution of 0.8 km at the satellite subpoint . Limited on-board recording capabilities do not permit global coverage from this sensor. (4) Solar Proton Monitor (SPM) : measures proton and electron flux from space in several energy ranges. 3.1.2 SR Subsystem Data used for calculating sea surface temperatures originate from the SR sensor, a two-channel instrument sensitive to radiation in the 0.5-0.7 urn visible region and in the 10.5- to 12.5 urn infrared (IR) "atmospheric window" region. An elliptical mirror rotating at 48 rpm provides an optical scan perpendicular to the satellite track. Radiation reflected by the mirror passes through a beam splitter and spectral filters to select the two chan- nels. The IR radiation is detected by a thermistor bolometer which provides a 5.3 mr instantaneous field of view (IFOV) . The spacecraft orbital and sensor scanning characteristics provide contiguous global coverage at a resolution of 7.5 km. Accurate calibration is provided by viewing space and a portion of the instrument housing containing embedded thermistors. The resolution and sensitivity of the IR portion of the SR subsystem is discussed 35 z o H < H Z <-> = 32 Q U 2£ o« o oe<2 I— 1 Q. O <4-l Kt 4-J CC T3 h O z (1) < CO e H CO o < U •H oo -a o oj tttE Ph too. W> C •H c c 03 o -» z CO o o 1 »- — CO c9_io: Q Isi I/) M PS •H « 'J o Ph H CO CO I I ■ LO •H PL, 11 (storing SST retrievals in the SST raw Retrieval Storage disk file. Simul- taneously, the VTPR orbital processing system calculates vertical temperature (soundings using the SST analyzed field to obtain a lower boundary temperature ;for its regression analysis (McMillin et al. 1973). One output of the VTPR [processing is an orbit-by-orbit disk file of atmospheric attenuation moisture ■coefficients which reflects the total moisture content of the atmospheric column. In the first step of the SST daily processing system, each SST raw •retrieval is corrected for atmospheric attenuation. The VTPR moisture ('coefficients are used for this correction if they are spatially and tempora- lly coincident with the raw retrieval being corrected. If the VTPR sensor is not able to provide moisture coefficients for a certain area of the globe, then a global map of the moisture coefficients for the last 72 hours is used to obtain an atmospheric correction. If moisture coefficients are unavail- able for more than 72 hours, then an empirically derived mean correction is used. (See paragraph 2.1 for a more detailed explanation of atmospheric attenuation procedures.) On a typical day (April 28, 1975), 79% of the raw retrievals were cor- rected using time coincident VTPR coefficients, 21% using the 72-hour field, and less than 1% using the empirical tables. 3.4.2 Objective Analysis Technique An objective analysis is performed on the moisture-corrected retrievals ,'to screen out those with residual cloud contamination. The output of the 'analysis is a disk file of SST observations that have passed all the objec- tive analysis quality checks and an analyzed SST field consisting of a Northern and Southern Hemisphere analysis on a 256 x 256 polar stereographic [grid. Each raw retrieval is compared to the SST value resident in the [previous day's analyzed field at the same location. Retrievals are rejected if they deviate from the analyzed SST by more than certain limits which are ^functions of the local gradient and the length of time since an SST obser- vation has been available at that location. Retrievals with atmospheric corrections that deviate from the empirical mean atmospheric attenuation jcorrections by more than preset limits are compared more stringently with |the associated field SST value. Retrievals are then compared and those jfound anomalous are rejected. All retrievals passing the above three tests iare placed in the SST Observation Storage. The observations are merged into ithe SST analyzed field to be used for the following day's retrieval screen- ing and for the generation of products. Finally, local gradients are re- calculated and the analyzed field is complete. 3.4.3 Quality Control To insure that the quality of the SST observations does not degrade, numerous parameters are recorded and analyzed periodically. The parameters most indicative of the state of the system are graphed. These parameters include (1) The mean difference between ship observations and the SST analyzed field, 45 (2) the average value of all raw retrievals, (3) the mean value of the moisture correction in four latitude bands (18°S to 36°S, 0°S to 18°S, 0°N to 18°N, and 18°N to 36°N) , and (4) the mean value of the SST field for both hemispheres. A quick-look photographic display of the field is also generated each day This is a grey-scale presentation of the global temperature pattern, paired with a display of the retrieval coverage. When problems that may compromise quality occur, other diagnostic tools, such as detailed printout displays of some of the SST analyzed field values and microfilm contour maps of the SST pattern in areas of concern, are activated. 3.4.4 Daily Archival The SST analyzed field and the SST are accumulated daily on separate disk for use by the scheduled non-daily processing routine. In addition, a disk back-up tape is created each day. This tape contains a copy of all the SST data files on disk, thus, it provides a recovery capability in the event of disk or program failure or a tape failure during archival. 3.4.5 Processing Summary and Initialization The last operation performed is that of summarizing the day's successful processing and initializing the directories in preparation for a new day's processing. Three summaries are produced: (1) Raw retrieval processing summary -- This summary of retrieval statis tics for the day contains orbital information, the number of histogram processed, the number failing the different pre- and post-retrieval tests, and the number of retrievals accepted. (2) Daily quality control summary -- This summarizes all the orbital quality control parameters for the day and plots the ranges of the various quality parameters for each orbit of the day. (3) Master directory summary -- This summary displays the IR and visible master directory contents and shows the orbits for the day and the processing success of all RDB data users. As a last step, the SST raw retrieval storage, the IR and visible master directories, the IR and visible quality control summary data sets, and the atmospheric attenuation moisture coefficient storage are initialized to prepare them for new incoming orbital data. 3.5 Verification Twice a day, once at 0000 GMT and again at 1200 GMT, the Navy Fleet Numerical Weather Center in Monterey, California, transmits approximately 1,000 ship reports to its Suit land office. These are recorded on magnetic tape, made available to NESS personnel, and loaded into the IBM 360/195. A computer program compares these ship data with the SST analyzed field and generates global and local area statistics. Many parameters are recorded and analyzed; one of these, the global mean deviation, is plotted and 46 summarized in an SST Quarterly Report. This mean deviation from ship obser- vations serves as an indication of the accuracy and stability of the total SST processing system. 3.6 Scheduled Non-Daily Processing Three tasks that support the daily SST operation are performed on a non- daily basis. These functions are automatically executed on a preset schedule at the frequency required by the program they support: (1) Every ten days, the contents of the analyzed field archive are trans- ferred to tape. (2) On the first day of each month, the contents of the observation archive, covering all observations for the previous month, are trans- ferred to tape and sent to the Environmental Data Service (EDS) . This EDS archive tape includes the latitude, longitude, time, and tempera- ture for each observation. (3) On the 15th day of every month, climatology for the next month pre- pared by the National Center for Atmospheric Research (NCAR) is loaded to disk for use by the objective analysis procedure. 47 4. SYSTEM PERFORMANCE A judgement of the SST system must be made from two aspects, the veracit) of its products and the reliability of the total system processing. An absolutely accurate product available at random times is of little use. It must be remembered that the project goal is an automated operational SST program. All theoretical, programming, and evaluational aspects were examined for feasibility; those not within the constraints of automation were rejected. Subjective appraisal of the status of the system is to be avoided. The system should inform the project monitor of errors, not vice versa. Only ii this way could manpower requirements be kept to a minimum, since the volume of data for coverage of the globe precludes assessment by human means. 4.1 SST Verification Verification, the comparing of data against a standard, raises an imme- diate question. What is the standard for SST measurement? When the project was in the earliest stages of development, this was not considered a problen and ship data were used. Although ship data are still used, they are con- sidered a problem. Are the ship data better or worse than satellite obser- vations? How accurate are they? The present programs used to verify SST observations began with the use of the simplest measuring criteria to ascertain the difference between satellite and ship reports, namely, means and standard deviations. The most recent criteria in use are not much better than these early methods; a completely satisfactory method of verify ing SST satellite retrievals is still being sought. 4.1.1 Comparison with Ship Data Currently, the most important verification program compares ship obser- vations against the satellite derived field. The accuracy of the field produced by the objective analysis program directly affects the validity of the individual satellite observation (paragraph 3.4.2). Consequently, a good comparison, while indirect, can be interpreted as a verification of the satellite observations. The ship reports used are those contained in the Navy "B96 tape", transmitted twice daily at 0000 GMT and 1200 GMT from Monterey, California, to Fleet Weather Control, FOB #4, Suitland, Maryland. Values rejected by the Navy (Holl et al . , 1971, pages 24-25, section 2.2.6) as unrepresentative are not used for verification. The remaining reports are subtracted from the nearest grid temperature of the satellite derived field (a negative difference implies the satellite field is colder than the standard). No interpolation is done. The mean and standard deviation of all such comparisons are then calculated. Figure 16 contains a graphic illustration of these results from July 1, 1974 through March 31, 1975. This graph is used as a control chart. If the mean difference is greater than ±0.5°C, an investigation is started to determine the cause in the SST system. Depending on the diagnosis, a correction may or may not be made. A mean difference greater than ±1.0°C requires a correction in the system when possible. These same data, extending over the globe, are divided by 48 1.5 0.0 -0.5 - II I I I I I I I I I I I I I II I I I I I I I II I I I I I I II I I I II I II I I I I I I I I I I I I I I I I I II I I I I I I I I I I I I I I I I I I I I I I I I I II I I I I JUL AUG SEP 0.0 - -1.0 - i ii i i i i i i i i i r i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i i ii OCT NOV DEC 1.5 0.5 - 0.0 - ' ' ' ' i ' I I I I I I I I M I I I I I I I I I M I I I II I I II I I I I II I I I I I II I I I II I I I I I II II I I I I I II I I I I II I I I I II I I II II I JA" FEB HAR Figure 16. --Daily mean difference between ship sea surface temperature measurements and satellite SST observations (July, 1974 to March, 1975) 19 geographic area to better define weaknesses and strenj Iths of the SST field. Table 4-1 gives this sectional breakdown for a representative day in March 1975 for the difference between <. ship reports and satellite derived f Leld. Table 4-1 -- Regional comparison of ship measurements vs. satellite measure- j ment (3/11/75). I Area Mean difference Standard deviation t Count 1. 20E-20W, 20N-90N -1.27°C 1.41 160 I 2. 20W-60W, 20N-90N -.03 1.99 94 s 3. 60W-100W, 20N-90N .00 2.70 120 i 4. 100W-140W,20N-90N -.37 1.39 60 j 5. 140W-180W,20N-90N -.53 1.61 107 t 6. 180W-140E,20N-90N -.16 2.18 146 s 7. 140E-100E,20N-90N -.55 2.07 48 , 8. 100E-60E, 20N-90N .0 .0 a 9. 60E-20E, 20N-90N 5.6 .0 1 a 10. 10E-80W, 0-20N -.15 1.60 56 | 11. 80W-120W, 0-20N .76 1.70 28 i 12. 170W-100E,0-20N -1.28 1.71 26 | 13. 100E-10E, 0-20N -1.14 .42 7 i 14. 0-360, 0-20S -1.62 1.58 61 t 15. 0-360, 20S-90S .75 1.52 56 s 16. Global Summary -.43 2.00 970 , The southern hemisphere has only two divisions due to the lack of conven- tional reports from this area. Area 3, which shows the highest scatter (standard deviation of 2.7°C), contains most of the Gulf Stream. This high degree of scatter was expected since the criteria of a global model necessi- tated limiting the resolution of measurement. The rise in the standard de- viation also is manifest in the area off the coast of Japan (area 6) which also contains a warm ocean current, the Kuroshio Current. The researcher interested in applications must always keep this limitation in mind; areas where the SST gradient is greater than 5.2 K/100 km will not be stable in character. Local and meso-scale analysis over such regions as a means to supplement the global scale field are under discussion. 4.1.2 Relative Merits of Ship and Satellite Measurements The value of ship SST reports and estimations of their errors are dis- cussed by Leese et al. (1971), and James and Fox (1972); and, while no state- ment concerning the absolute accuracy of the satellite measurements can be made, their relative merit as compared to ships' measurements can be con- sidered. The most important difference between the two is obvious. Tempera- ture reports for the entire globe are made from the ITOS spacecraft using a single instrument, whereas a multitude of instruments are used in large scale 50 I coverage by ships. A researcher using data from different ships to investi- gate a large oceanic region for sea surface temperature structure is relying ion the lack of systematic and/or large calibration errors in many individual 'instruments. In general, ships' instruments are not calibrated against one 'another. This problem does not occur in using satellite derived information. The relationship between ambient temperature information is constant, ialthough there may be systematic biases; e.g., those due to incorrect atmos- jpheric attenuation for local meteorological conditions. Let us examine the merits of ship and satellite measurement. Given a satellite observation and a ship report within 100 km of each other and i reported within the same 24 hour period, calculate the probability of agree- jment between the two measurements. The agreement will be affected by (1) Jship mislocation, (2) error in ship measurement, (3) difference in time i between compared reports, (4) sea surface temperature gradient, (5) error in (Satellite measurement, and (6) mislocation of satellite measurement. Ideally, the reduction in the probability of agreement caused by the first four items should be measured. By combining this information with the probability of agreement between ship and satellite observations, a true picture of the accuracy of the satellite measurement could then be obtained. Preliminary | attempts to do this have been unsatisfactory. One attempt consisted of comparing sea surface measurements, ship vs. ship, whenever two ship measure- ments were made within 100 km of each other, and calculating the probabi- lity of agreement on a global basis. Unfortunately, however, this knowledge is not directly applicable to the ship against satellite comparisons since [the distribution of the locations would be different between the two sample sets (affecting items 3 and 4) and the ships used in the two sample sets need jnot be the same (affecting items 1 and 2). In any event the probability of agreement between ship observations is a measure of what to expect from in- struments of known reliability when subjected to the same treatment undergone by an instrument of unknown reliability. Table 4.2 is a summary of the results for March 1975. The ship reports used are the same as those on the Navy B96 file. The probability of agree- ment between the two observations, ship and satellite, was calculated over the entire globe, whenever the satellite measurement and a ship report was made within 100 km of each other. By definition, temperatures differing no more than 1.5°C were considered to be in agreement. The probability of agreement between ship and ship was also calculated when ships were within 100 km of each other. SI Table 4-2 -- Classifying satellite measurements using ship reports. Date Ship vs. satellite Ship vs. ship Numb er of Probab ility of Number of Probability of comp arisons being within 1.5°C comparisons being within 1.5°C 3/3/75 870 0.68 1779 0.80 3/4 867 0.60 2097 0.78 3/5 794 0.64 2360 0.83 3/6 877 0.62 2536 0.79 3/7 736 0.65 2332 0.80 3/10 939 0.63 2304 0.76 3/11 829 0.63 2180 0.70 3/12 786 0.65 1789 0.80 3/14 2147 0.66 2215 0.83 3/17 808 0.66 1471 0.80 3/18 956 0.70 1600 0.75 3/19 881 0.61 2408 0.77 3/20 1240 0.70 2685 0.74 3/21 658 0.71 2119 0.77 3/24 774 0.59 2692 0.76 3/27 765 0.61 2207 0.82 Average 0.65 0.78 March 1975 is the only month for which complete statistics are available; consequently it is not claimed to be representative of other times. The agreement between ship and satellite observations is quite impressive. On the average, given any satellite observation, one can assume that it will agree with that of a nearby ship 65 out of 100 times. (Ship reports will agree with each other 78 out of 100 times.) 4.2 System Reliability The success rate of the SST system was evaluated by finding what percen- tage of orbits scheduled produced EDS-archived SST observations. Table 4-3 shows this percentage month-by-month for the period since the inception of the present SST system. It also shows the number of days on which no EDS archive file was produced. The processing system is reliable; there were only five failure days out of more than 300 days of operation. A large measure of this success is due to the redundancy of all instruments and com- puters in the system - two satellites, two radiometers on each satellite, tw orbital ingest EMR computers, two IBM 360 computers. Improvements are now directed towards reducing human intervention and support, where much of the reason for failure rests. 52 Table 4-3 -- Rate of success in obtaining SST Month Jun 74 Jul Aug Sep Oct Nov Dec Jan 75 Feb Mar Percentage 98 Number of failure days 94 95 98 97 96 99 2 10 90 100 99 2 4.3 System Monitoring and Evaluation Evaluation of SST has three purposes: (1) to detect errors in the SST system, (2) to ascertain authenticity of SST retrievals, and (3) to provide a base for decision analysis and improvements to the SST system. All evaluation programs which detect errors in the SST system can be divided into two classes; (1) those used to locate the error in one of the three basic SST subsystems - SR sensor (all processing up to and including unattenuated SST), VTPR sensor (specifically the derivation of A and B attenuation coefficients), and mapping (OAT program, and attenuation program) and (2) those which show the precise cause of the error. Once a problem has been located in one of the three branches of the SST system, further evalua- tion is made to find the specific cause. Figure 17 is a summary of the main evaluation programs and the decision paths used to locate and correct errors in the SST system. Programs which locate error good Programs which identify cause of error AUTHENTICATE SST ->|N0 ACTION bad ^L bad CHECK SR SENSOR ^ ->• INVESTIGATE CAUSE AND CORRECT i — . , K good bad CHECK VTPR SENSOR \£ ->! INVESTIGATE CAUSE AND CORRECT good CHECK MAPPING > INVESTIGATE CAUSE AND CORRECT Figure 17 -- Evaluation procedures Note that only one program need be checked each day. If a check shows that the system has deteriorated, three other parameters that show which subsys- tem of the SST is at fault are checked. At present, it takes 0.5 man-hours per day to monitor this system, i.e., to check the SST quality and to record 53 all diagnostic printouts for future references. 4.4 Satellite SST Coverage Implied within the system performance discussion is the coverage of the globe for each analysis run. The SST programs run daily on all data gathered in the previous 24 hours. Since the ITOS satellite covers every area of the earth at least twice each day (up to 25 times daily at the poles) there are two chances to retrieve an SST measurement over any given cloud- free region. This translates into 40,000 retrieval histograms before quality control tests; 6,300 remain after all data rejection tests are complete. These figures are for a representative day of operation. Figure 21, para- graph 5.1, shows the distribution of the satellite observations for a typical day. Note that one observation will affect more than its immediate area, thus extending the region changed for that particular analysis run. No sea surface temperature retrieval is attempted above the 70° parallel because of ice coverage in the region. Over the period of one month, practically every point in the NESS 256 x 256 grid overlay used in the analysis is updated. Figures 18 and 19 are FR-80 contour maps of a sample month's coverage. No contours are drawn where there was no new information for the month. This only occurs where persistent cloud cover is expected for that time of year. The large amount of data near the poles is due to orbital overlap and the map projection. 54 Figure 18. --Number of satellite observations in Northern Hemisphere durinj July 1975. 55 Figure 19. --Number of satellite observations in Southern Hemisphere during July 1975. 56 5. SST PRODUCTS AND DISPLAYS Satellite SST data are available to users either as observations or as analyzed fields. Both products are available from NOAA's Environmental Data Service (EDS) in the form of magnetic tape. In addition, users with NOAA IBM 360/195 terminals can obtain SST data directly from disk. Paragraph 5.2 describes some of the ways in which the SST data can be displayed. Although these displays are not available as products, they are included to illustrate the nature of the SST satellite data and to suggest possible applications. 5.1 Products 5.1.1 Observations Satellite sea surface temperature observations are available in three forms: (1) an EDS archive tape, (2) an observation transmission tape, and (3) an observation disk data set. All three forms consist of a collection of individual sea surface temperature observations. Each observation con- tains the latitude, longitude, time, source (satellite and sensor number), and magnitude of the sea surface temperature. (1) EDS archive tape - All SST observations for one month are collected, written to tape, and archived by EDS. Appendix A describes the format of this tape. Copies of the EDS archive tape can be pur- chased from EDS. The EDS archive tape series starts with December 1972 and continues uninterrupted to the present. (2) Observation transmission tape - When required, all the observations for the previous day can be recorded on tape for transmission purposes. The content of this tape is similar to the EDS archive tape, although the format is different. (3) Observation disk data set - This is the master file of observations from which the two tapes just described are created. This file is accessible to terminal users at any time, but is updated only once a day. 5.1.2 Analyzed Fields Once per day, all the satellite SST observations are merged into a polar stereographic field. In addition to the SST, this "analyzed field" contains land-sea tags, climatology temperatures, SST gradients, data age information, and verification temperatures. The latest field is accessible to terminal users at any time on disk. Ten analyzed fields from ten successive days are written to tape; this tape can also be purchased from EDS. With a few exceptions, this tape is available for every day since May 10, 1973. The format of this ten-day analysis field archive is given in appendix B. 5.2 Displays Numerous methods of displaying and transmitting SST observations and fields have been developed. A few of the more useful methods are described 57 in this section to familiarize the user with the character of the SST data. It must be emphasized that these displays are not products and are not available for distribution. They all are illustrative of the types of dis- plays a user can create from the satellite SST data. 5.2.1 Latitude/Longitude Printout (figure 20) This type of display shows the SST analyzed field values at half-degree intervals of latitude and longitude. The SST is in °CxlO. The one-digit number above the SST is the number of days since an observation occurred at that earth location. If an observation occurs on the day of the analysis, the number is replaced by a letter code designating the number of observa- tions used to arrive at the SST shown (A=l-4, B=5-8, . . .H=29 or more). Land is- denoted by asterisks. 5.2.2 Photographic Display (figure 21) A Digital Muirhead Display (DMD) shown in figure 21 can be produced from the analyzed field. This display permits a "quick-look" at the integrity of the field, and gives the analyst an idea of the global SST pattern and an indication of the spatial distribution of observations. The left-hand dis- play shows the SST patterns in 3.0°C grey-scale contours. The grey wedge at the top gives the temperature of each contour starting with black for 270, 271, and 272 K. Each succeeding grey level represents three degrees with white representing 300, 301, and 302 K and above. Land is black. The right display depicts the number of observations used in the analysis for each gridpoint. A black square pinpoints the position of each observation on the analysis date. The darkest grey represents the gridpoints at which more than ten observations were used to produce an analyzed temperature. The middle grey value represents areas at which from one to ten observations were used, and white areas contain no new data. The date shown on the print is the analysis date. Data used are predominately from the previous Green- wich day. 5.2.3 Teletype Transmission (figure 22) SST observations coded for transmission Over teletype lines can be entered on magnetic tape. Figure 22 is a printout of observations coded according to WMO Draft Recommendation A(73-CBS) for teletype transmission of processed data. 5.2.4 Microfilm (figures 23, 24, 25) Contour displays of the analyzed field can be made on microfilm. A global polar stereographic contoured field is shown in figure 23. Contours are labeled in °C with a two-degree contour interval. Sections of the globe can be enlarged in polar stereographic projection (fig. 24) or in Mercator projection (fig. 25). 58 5.2.5 Character Printout (figure 26) The analyzed field can be displayed as a computer printout with a letter code representing sea surface temperature intervals. Normally for maximum information, each succeeding letter of the alphabet represents one degree above 269 K (i.e. A=270.0 to 270.9, B=271.0 to 271.9, etc.). Temperatures are not entered for land areas. This display is a 256x256 hemisphere grid mesh, or exactly the resolution of the fully analyzed field. Figure 26 uses a three degree interval cycle designed to compensate for the photographic reproduction process. 59 25 24 23 22 21 20 19 IS 17 1« 15 i< 13 12 i i( 30-«2 203 •2 203 •2 1 71 •2 1 7 1 -2 1 77 ♦2 183 •2 190 '2 190 •A 197 • A 197 •A 187 •8 192 • B 192 -8 188 ♦B 191 •8 185 •8 185 ♦B 1*5 •B 180 •8 182 •8 164 -8 182 -8 1 7* •B 1 7* •B 182 ♦B 1*7 •B 177 • A .A «A .A 167 167 16* 168 •B 192 •A 192 «A 192 • A 183 •A 183 •2 1(3 • A 199 '2 18? • A 19? • A 195 •A 200 •A 200 •8 190 •8 190 •B 191 >B 187 •8 181 'B 180 «B 180 • 8 186 •B 186 •B 182 •8 188 >B 188 •B 187 ♦B 187 •B 1 77 ♦ A .A »A «A 171 17* 168 168 29-.B 192 •A •A 198 • A 198 •A 193 •A 199 ♦A 199 • A 193 •A 193 •A 195 193 ♦B 193 •B 193 •B 184 •B 184 -B 187 •A '91 •A 191 •8 180 •B 180 •8 1 78 •B 188 •8 188 •8 188 ♦B 185 .8 187 -8 187 ♦A »i 171 . •8 193 •A 183 •8 202 •8 19* •A 195 •A 201 -8 198 •B 198 •A 183 •A 185 •8 193 •B 194 •8 18* •8 184 •8 180 •A 181 • A 181 ♦A 191 •B 186 •8 186 -B 178 •8 193 •8 192 •B 192 *8 185 •B 1 77 •8 177 J 2»-«B 193 »B 202 •8 202 '8 203 •8 203 • A 201 • A 210 •i 19* •8 196 •A 183 •B 19* •8 194 • A 195 ♦A 195 •B 180 »B 198 •8 195 •B 186 • B 186 •B 191 •B 191 •8 193 • A 181 -A 181 • A 1 76 •A 1 76 '. ; •B 202 202 •A 207 •A 207 •B 212 • B 212 •A 210 • A 2 02 •2 192 •A 183 • A 190 »A 19* •A 1 96 «8 19* 196 •8 198 • % 195 •B 204 ♦B 200 •8 200 •B 198 •A 193 •A 193 •A 181 • '. • ', 27-. C 21* •B 19? • B 199 206 -8 2 0a • B 2 l 5 • 8 215 «A 202 •A 215 •A 2 15 •A 190 -2 2 1 1 •A 193 •8 196 •A 199 •A 193 •A 193 -8 204 •8 194 •B 200 •8 198 •A 186 •A 183 '. ■ 1 '. J 21* •8 209 ♦ 8 212 • 8 212 •A 215 •8 215 •8 212 • A 204 • A 204 ♦A 217 • A 21? -2 21 1 •A 194 •A 19* •A 199 •A 21 1 •8 197 •8 19* •8 19* •A 201 •A 201 •A 186 '. 1 ', '. '. ; 26-»D 21* •8 209 •C 207 •3 220 •A 215 >8 213 •8 212 •A 21? •A 223 •A 223 •A 21? •A 201 »A 201 •A 205 •A 205 -A 21 1 »B 197 •A 20* •A 208 •A 208 ♦A 188 '. t '. • • '. '. •0 21 7 •0 217 •C 20? ♦8 218 *8 218 213 •B 214 •A 21? «A 207 ♦A 216 •A 21? ♦A 210 .A 213 •A 213 • A 214 •8 210 ♦B 210 •A 20* •A 201 • A 199 ♦A 188 * • '. '. ' ' ', 25-«C 212 -0 217 • c 219 •C 219 •A 21? •A 217 .8 214 -8 210 • A 20? •A 218 •A 218 •A 210 .A 217 • A 218 •A 2 14 •A 21 1 •A 205 *A 20) •A 201 •A 189 •A 189 J '. '. '. '. *, ; 219 •0 21? • B 219 • A 200 •A 200 •A 208 •A 208 ♦ 8 210 >A 211 •A 2 1 1 •A 214 »A 214 • « 217 • A 225 • A 225 • A 21 1 • A 197 ♦A 197 ♦A 193 «A 193 I J • I ", '. • 2*-«C 219 203 •8 2 1 9 •8 2 10 •B 197 »B 197 •8 215 »B 2 1 1 •8 21 1 ♦A 207 •A 207 -A 2 10 •A 2 14 • A 21* • A 210 • A 210 •A 210 >A 188 • A 188 '. ', '. • ; '. '. '. •C 217 • B 203 •A 200 •8 210 ♦A 220 •A 214 •8 215 •8 2 1* >B 21 1 •8 199 *A 202 «A 210 •A 202 »A 212 •A 210 •A 201 • A 189 •A 189 '. 1 • '. '. '. ♦ '. ', 23-. B 210 •8 210 •A 200 -8 201 • A 220 "B 207 .8 207 .8 216 •A 199 •8 199 •A 2 06 •A 206 •A 202 «A 203 •A 203 •A 201 •A 188 •A 188 '. '. ' '. '. '. '. • '. •C 232 »B 217 ♦B 217 '8 2 09 ♦8 205 •B 205 •A 205 • A 203 •A 199 •A 208 ♦A 208 •A 216 • A 203 »A 203 • A 201 • A 201 •A 190 »■ '. '. '. '. ; ; 1 '. • 22-. C 229 2 2-9 .8 217 209 -A 216 •8 213 »8 213 •A 215 • A 20 3 •* 18? ♦4 205 .A 2 1* • A 20« •A 191 •A 198 •4 193 '.I •• 1 '. • I '. '. ', '. '. •8 217 21-. B 217 22 5 •8 218 •C 225 »8 117 • B 22 1 >B 229 •A 2 16 ■>B 229 «A 210 •A 210 .A 214 •8 219 .A 215 •8 219 •A 215 •A 217 •* 187 •A 217 • A 209 •A 207 >A 209 •A 21? .A 209 •A 21? •4 202 •4 202 •4 199 •4 197 •4 193 •4 197 :: :: I l ; | • 1 I • 1 •B 221 •8 218 •8 2 32 •B 232 •8 227 ♦8 227 ♦A 220 ♦A 223 ♦A 223 •A 215 •4 216 •4 216 •* 209 • 4 198 •4 198 •4 193 v. '.". I '. ', '. '. '. '. '. '. 20-. 8 223 •8 225 •8 225 •8 222 •8 222 218 ♦ A 216 •A 21 1 •A 213 •A 215 ♦A 215 •4 215 •4 209 • 4 206 •4 2 02 •4 187 •4 18? •7 195 '. '. ". '. '. • '. '. • •A 234 •8 226 •8 219 •8 222 •8 225 -•8 218 •A 214 • A 21 1 • A 213 ♦A 213 ♦A 215 •4 224 -4 224 •4 210 •4 199 .4 199 •4 192 '' '. '. 1 '. '. '• • 1 '. 19-. A 229 *B 226 •8 223 •8 222 •A 227 •A 22 7 »A 221 •A 221 • A 213 •>4 224 •4 224 «4 228 •6 219 •4 210 . * 209 •6 20? •4 192 .* 200 '. \ '. ', '. '. '. '. '. ♦8 223 •B lib •8 223 •A 230 • A 225 • A 22b •A 22b -4 23* • A 223 •A 223 •6 230 •4 227 •4 225 ♦4 216 •4 209 •6 200 •6 200 •4 200 '. • '. '. '. '. • '. '. 1S--B 223 .A 232 •A 232 •A 234 • A 228 «A 228 •A 231 •4 234 >* 233 •6 232 •4 230 •4 230 .4 225 »4 21 1 ♦4 21 1 • 4 202 • 4 200 '.'. '. J '. '. '. '. • '. • •A 224 •A 224 • A 23? •A 237 •A 230 •A 2 35 •4 23* •4 2 3* ♦6 235 •2 232 •2 232 •2 2 12 • 4 219 •4 219 -4 205 -4 200 •6 199 •6 202 • '. '. '. '. '. '. ' I 17-. A 237 •5 245 •A 245 • A 245 •A 2 30 •A 232 •A 234 •2 241 ♦2 241 •4 233 •2 232 •2 212 •2 218 •4 21 1 ♦* 204 •6 204 •6 199 1°. ♦ '. '. '. • '. '. ". • •A 23 7 •A 246 • A 245 •A 244 • A 244 •2 24? •2 244 «2 244 •2 243 •4 233 •6 232 •2 229 »2 2 1 3 •6 205 »6 205 •6 202 •6 203 •b 203 * • • ♦ '. ♦ ; • ; 16-. A 252 •A 254 •A 254 ♦A 2 52 •A 256 'A 2 56 •2 246 •2 24 1 •6 240 •6 240 •6 2 32 •2 235 •2 235 •6 222 •6 206 ♦6 202 •6 193 »« '. '. '. * '■ • • * '• •A 250 *A 250 • A 255 • A 255 •5 255 •2 2 59 •2 2 46 -2 251 •2 24 4 •6 244 •2 245 •2 253 •2 2 57 •6 222 6 223 •6 2 10 '■'. '.', '. '. '. * '• ♦ * *. • 15-. A 2<.7 -A 257 ■ A 257 • A 259 •5 255 -A 2 58 • A 25 8 •2 25 7 •6 244 •2 248 ♦ 2 260 '2 253 •2 235 •2 235 •2 226 •6 2 1 1 *' '.'. • • '. '. '. '. • '. * 25 24 23 22 21 20 19 18 17 1 > 1 1 i i 1 t i 1 10 Figure 20. --Latitude longitude grid SST computer print for the area off the northwest coast of Africa for January 27, 1976 (see section 5.2.1 for expla- nation). 60 Figure 21 . --Photograph Display--Digi tal Muirhead Display (DMD), 61 •109590 6210 07909 10101) 111 16100 56000 222 14415 27250 333 13412 23000) 0133 000000 290 269 271 277 266 232 222 224 214) 0205 000005 267 0325 035005 277 255 245 248 248 0402 000010 276 0504 015010 274 0622 050010 275 255 256 265 256 0704 015015 280 0824 040015 279 260 260 256 261 0901 000020 273) 1027 025020 274 273 270 266 263 1103 000025 266 1205 025025 271 1321 055025 270 270 268 267 263 1404 000030 252 1502 030030 265 1601 060030 255 1719 070030 365 268 264 266 264 1807 00 0035 244 241 240 1918 075085 266 260 257 255 256 2007 000040 233 2117 090040 270 259 254 254 253) 2204 185040 243 2307 000045 234 2424 090045 269 263 255 254 259 2507 000050 228 2624 090050 259 243 240 237 241 2706 000055 217 2821 105055 271 225 231 224 237 2906 000060 213 3019 120060 257 217 210 218 219 3107 000065 215 3219 125065 249 203 202 200 229 3307 000070 213 3420 120070 238 199 191 191 192 555 07999 10101 777 (( 1 1 GAXT14 1 CWBC 301500); ' J 67409 73011 01001)) 30950 43100 \ > 50150 90025 00025)) ) 298 287 283 283 284 276 280 280 279 281 277 277)) ' 262 )' 284 264 258 252 238 242 234 245 242 240 231 231))' 281 280 277)) ' 275 280 282 274 279 267 267 267 262 255 255 248))' 243 244 244 249 243 236 236 220))' 278))' 278 278 278))' 279 280 279 263 271 269 272 262 255 254 251 250))' 256 254 255 261 237))' 279 277 276))' 276 280 278 277 277 277 281 285 273 257 254 258))' 270 )' 275 267 256 250 258 267 251))' 272 277 274 279 280 291 285 289 278 286 283))' 265 269 265 272 266 259 248 263 279 261)) 265 262))' 273 272 272 267)) 275 281 282 276 282 288 284 280 278 277 273 274))' 265 269 274 265))' 250 251 251)) 261))' ))' 273 277 286 285 283 280 275 273 271 269 270 265))' 259 257))' 249 249 249 255))' 271 274 277 280 273 271 271 272 268 271 267 268))' 252))' 234 228 224 222 235 240)) 280 )' 236 273 274 277 278 274 279 276 269 252 259 262)) ' 242 244))' 236 221 223 228 236 226)) 276 285 284 284 279 278 274 279 268 267 266 267))' 254 249 247 238 254 248 241))' 229 217 222 215 215 224))' 263 265 275 273 273 264 263 259 254 257 254 249))' 241 239 248 242 244 245 241)) 217 204 199 183 175))' 275 277 264 257 257 258 253 243 236 233 224 221)) ' 229 238 238 241))' 215 209 209 186 177))' 255 252 247 234 230 223 225 222 220 222 218 217))' 225 217))' 215 211 203 291 279 281))' 242 237 232 226 217 218 213 213 216 215 210 206))' 209 203))' 214 209 200 190 182 179)) 239 236 227 232 223 220 210 212 212 208 206 200)) ' 241 200 295))' Figure 22. --Data from SST teletype tape (the parentheses and apostrophes represent carriage control commands). 62 Figure 23. --Global polar stereographic microfilm map for January 7, 1976, 63 21 ZZ u u " zs 25 Figure 24. --Polar stereographic microfilm section for January 7, 1976 depict- ing SST isotherms in Celsius off the Eastern U.S. coast. 64 HOW H5W now 50N 45N AONi 35N 30N 25N 20N SON 45N 40N 35N 30N ^25N 20N HOW 135W 130W 125W 120W 115W now Figure 25 . --Mercator projection microfilm map section for Jan. 7, 1976, 65 Figure 26.--SST analyzed field character print sheet, using three degree contours (January 9, 1976). (See note in section 5.2.5.) 66 REFERENCES Adams, H., 1972: F17 model ITOS scanning radiometer, Data Book , Radio Corporation of America Astro-Electronics Division, Contract No. G-OF-8256- 0201-99-F09, Santa Barbara Research Center, Goleta, California, 141 pp. Bristor, C.L., (Editor), 1975: Central processing and analysis of geo- stationary satellite data, NOAA Technical Memorandum NESS 64, National Environmental Satellite Service, National Oceanic and Atmospheric Admin- istration, U.S. Department of Commerce, Washington, D.C. , 155 pp. Conlan, Edward F. , 1973: Operational products from ITOS scanning radiometer data, NOAA Technical Memorandum NESS 52, National Environmental Satellite Service, National Oceanic and Atmospheric Administration, U.S. Department of Commerce, Washington, D.C, 57 pp. Fortuna, Joseph J., and Hambrick, Larry N., 1974: The operation of the NOAA polar satellite system, NOAA Technical Memorandum NESS 60, National Environmental Satellite Service, National Oceanic and Atmospheric Admin- istration, U.S. Department of Commerce, Washington, D.C, 127 pp. Holl, Manfred M. , Mendenhall, Bruce R. , and Tilden, Charles E., 1971: Technical development for operational sea surface temperature analysis with capability for satellite data input, Naval Weapons Engineering Support Activity Detachment (FAMOS) , Contract No. N62306-70-C-0334, Meteorology International Inc. , Monterey, California, 73 pp. James, Richard W. , and Fox, Paul T. , 1972: Comparative sea surface tempera- ture measurements, Report No. 5 , World Meteorological Organization Reports on Marine Science Affairs, Secretariat of the World Meteorological Organ- ization, Geneva, Switzerland, 27 pp. Leese, John, Pichel, W. , Goddard, B., and Brower, R. , 1971(a): An experi- mental model for automated detection, measurement and quality control of sea surface temperatures from ITOS-SR data, Proceedings of the Seventh International Symposium on Remote Sensing of Environment , March 1971, Willow Run Laboratories, University of Michigan, Ann Arbor, Michigan, 625-646. Leese, John, Pichel, W. , Goddard, B., and Brower, R. , 1971(b): Factors affecting the accuracy of sea surface temperature measurements from ITOS- SR data, Conference Preprint No. 90 on Propag ation Limitations in Remote Sensing, June 1971 , North Atlantic Treaty Organization, Advisory Group for Aerospace Research and Development, Paris, France, 13 pp. Ludwig, George H. , 1974: The future polar orbiting environmental satellite system, Proceedings of EASCON, IEEE 1974 Electronics and Aerospace Systems Conference, October 7, 8, 9, 1974 , Aerospace and Electronics Systems Society, IEEE Washington Section, Washington, D.C, 498-502. (,7 McMillin, Larry M. , Wark, D.Q., Siomkajlo, J.M., Abel, P.G., Werbowetzki, A., Lauritson, L.A. , Pritchard, J. A., Crosby, D.S., Woolf, H.M. , Luebbe, R.C., Weinreb, M.P., Fleming, H.E. , Bittner, F.E., and Hayden, CM., 1973: Satellite infrared soundings from NOAA spacecraft, NOAA Technical Report NESS 65, National Environmental Satellite Service, National Oceanic and Atmospheric Administration, U.S. Department of Commerce, Washington, D.C., 112 pp. Prabhakara, C. , Conrath, B.J., and Kunde, V.G., 1972: Estimation of sea surface temperature from remote measurements in the 11-13 urn window region, Preprint X-651-72-358 , U.S. National Aeronautics and Space Admin- istration, Goddard Space Flight Center, Greenbelt, Maryland, 15 pp. RCA (Radio Corporation of America), 1975: TIROS-N system definition study report, Submission 2 , prepared for National Aeronautics and Space Administration, Greenbelt, Md. , under Contract No. NAS5-20644, RCA Astro-Electronics Division, Princeton, N.J. , 308 pp. Ruedger, Howard W. , 1973: Simulation studies for NOAA stabilized compensa- tion program, NOAA/NESS Contract No. 2-35369 , Research Triangle Institute, Research Triangle Park, N.C., 119 pp. Schwalb, A., 1972: Modified version of the improved TIROS operational satellite (ITOS D-G) , NOAA Technical Memorandum NESS 35, U.S. Department of Commerce, Washington, D.C., 48 pp. Shenk, William E., and Salomonson, Vincent V., 1972: A simulation study exploring the effects of sensor spatial resolution on estimates of cloud cover from satellites, Journal of Applied Meteorology , 11, 214-220. Smith, William L. , and Koffler, R. , National Environmental Satellite Service, National Oceanic and Atmospheric Administration, U.S. Department of Commerce, Washington, D.C., 1970 (private communication). Smith, William L. , Rao, P.K., Koffler, R. , and Curtis, W.R., 1970: The determination of sea surface temperature from satellite high resolution infrared window radiation measurements, Monthly Weather Review , 98, 604-611. Smith, William L. , Woolf, H.M., Abel, P.G., Hayden, CM., Chalfant, M. , and Grody, N. , 1974: Nimbus-5 sounder data processing system, part I: measurement Characteristics and Data Reduction Procedures, NOAA Technical Memorandum NESS 57, National Environmental Satellite Service, National Oceanic and Atmospheric Administration, U.S. Department of Commerce, Washington, D.C, 99 pp. 68 APPENDIX A EDS ARCHIVE TAPE FORMAT DESCRIPTION (MARCH 1974 TO PRESENT) Each archive tape contains one month of SST observations obtained from NESS operational satellites. There is one file for each day in the month. If data for any day of the month are unavailable, the file for the month contains a record of zeros for that day. Each file contains one record of documentation followed by a variable number of data records. A physical end-of-file indicator signifies the end of data for each day. All records are physical records of binary data written at 800 BPI on the IBM 360/195 on nine-track tapes with odd parity. The documentation record is 3,240 32-bit integer words in length. It contains the following information: 32-bit word no. Quantity 1 Number of orbital readouts for day. 2 Number of raw SST retrievals for day. 3 Number of records in file. 4 Number of SST observations for day. 10(20, 30, etc.) Sensor number. 11(21, 31, etc.) Orbital readout number. 12(22, 32, etc.) Upper 16 bits contain Greenwich hour of year at beginning of orbital readout. Lower 16 bits contain Greenwich quarter-second of hour at the beginning of orbital readout. 13(23, 33, etc.) Upper 16 bits contain Greenwich hour of year at end of orbital readout. Lower 16 bits contain Greenwich quarter-second of hour at the end of orbital readout. 14(24, 34, etc.) Number of raw retrievals in orbital readout. 15(25, 35, etc.) Reference NASA time for orbital readout. The above documentation quantities are applicable to NOAA satellite SR sensors. Word 1 contains the number of readouts of SR data processed on that day. Word 2 is the number of raw retrievals obtained for the day. Raw retrievals that pass the objective analysis quality tests are called SST observations. Only these observations are recorded on the EDS archive tape. Word 3 is the number of records of data (including documentation record) in the file for the day. Word 4 is the number of SST observations for the day. Each orbital readout has a group of six words containing descriptive (documentation) data for the readout. Documentation data for the first orbital readout is in words 10-15; for the second readout, words 20-25, etc. Word 10 (20, 30, etc.) is the sensor number which designates the source of the orbital readout. The following sensor numbers have been assigned. As other sources of data are used, the list will expand. 69 i ! 1 - NOAA 1 SR 1 2 - NOAA 1 SR 2 3 - NOAA 2 SR 1 4 - NOAA 2 SR 2 5 - NOAA 3 SR I 6 - NOAA 3 SR 2 7 - NOAA 4 SR 1 8 - NOAA 4 SR 2 9 - NOAA 5 SR 1 - NOAA 5 SR 2 Words 11(21, 31, etc.) are the NOAA satellite orbital readout number. Words 14, 24, etc. are the number of raw retrievals in the readout. Words 15, 25, etc. contain the time at which the NASA reference satellite counter was reset to zero. Words 12, 13 and 22, 23, etc. are four 16-bit bytes. The upper 16 bits of words 12, 22, etc. contain the hour of the year and the lower 16 bits the quarter-second of that hour at the beginning of the read- out. Likewise, the upper 16 bits of words 13, 23, etc. contain the hour of the year and the lower 16 bits the quarter-second of that hour at the end of the readout. All times are in GMT. For instance, if the orbital readout contains data from 0000 :20Z on January 2 to 0240 :10Z on the same day, then word 12 would have 25 for the hour and 80 for the quarter-second at the beginning of the readout. Word 13 would have 27 for the hour and 9,640 for the quarter- second at the end of the readout. The second record of each file and subsequent records, until an end-of- file indicator occurs, are the observation data records. The number of records is variable from day to day depending on the number of SST observa- tions calculated. Each record is 3,240 32-bit words in length and contains a maximum of 810 observations. There are eight 16-bit integer bytes in each observation. Bytes 1-8 (words 1-4) contain data for the first observation, bytes 9-16 (words 5-8) for the second observation, etc. Each observation contains the following data: SST ((K - 269.9) x 10) Longitude (Degrees west longitude x 10) Latitude (Degrees x 10) 0=N. Pole, 900=Equator, 1800=S. Pole) Sensor number Day of year Hours Minutes Seconds GMT of observation The sea surface temperature is coded by subtracting 269.9 K from the SST and multiplying by 10. The longitude is in degrees West of Greenwich multiplied by 10. The latitude is in degrees South of the North Pole multiplied by 10. The sensor number is an integer designating the source of the SST observa- tion. These were defined above in the description of the documentation record. Bytes 5-8 contain the time of the observation in day of the year, hours, minutes, and seconds of the day. For example, an SST observation of 70 295.4 K at longitude 120°W, latitude 5°N obtained from NOAA-3, sensor 1 on February 3 at 1640 : 15Z hours, minutes, and seconds of the day would be recorded as follows: 16-bit coded bytes quantity interpretation remarks 1 SST 255 295.4 K (295.4K-269.9K)xl0 2 Long. 1,200 120°W 120°WxlO 3 Lat. 850 5°N 85° South of 90°NxlO 4 Sensor 5 NOAA 3, sensor 1 5 Date 34 3 February Day of the year for 3 Feb 6 Hours L6 16:00:00 16 hours 7 Minutes 40 00:40:00 40 minutes 8 Seconds 15 00:00:15 15 seconds If the SST value (Byte 1) is zero, the observation is disregarded. Zeros will be placed in the last record of each day as a fill when the number of observations is not an integral multiple of 810. However, the end-of-file indicator is the intended terminator of one day's observations. 71 APPENDIX B SST ANALYZED FIELD 10-DAY ARCHIVE TAPE FORMAT DESCRIPTION (JUNE 1, 1974 TO PRESENT) The following description of the 10-day analyzed field format applies to those fields archived since June 1, 1974. Before July 30, 1974, the year of the century and the day of the year were absent from the field. The GOSSTCOMP Global Scale Analyzed Field is a 512-row by 256-column sea surface temperature field organized in a polar stereographic grid, true at 60° lati- tude. Rows one to 256 are used for the Northern Hemisphere and rows 257-512 for the Southern Hemisphere. Origins of the Northern and Southern Hemi- spheres are row one, column one and row 257, column one, respectively. The origin is located in the lower left-hand corner as shown in figure B-l. Row 128, column 128 is the North Pole and row 384, column 128 is the South Pole. The grid for each hemisphere can be thought of as a 255 x 255 grid, centered at the pole, with the distance from pole to equator 124.817436208 grid units. Row 256 of each hemisphere is superfluous and column 256 of each row contains the row number in the last 32 bits. Each grid point represents the area surrounding it (fig. B-2) . An archive tape is produced every ten days on the IBM 360 containing the SST analyzed fields for the previous ten days. Each analyzed field is a file containing 512 records of 1,024 32-bit words each. Each record is one row of the field grid. The tape records are binary physical records with physical end-of-file indicators written at 800 BPI on a 9-track tape with odd parity. Each row in the analyzed field consists of 256 columns and each of these columns has four 32-bit words, containing information pertaining to that grid point. Column one of each row consists of the first four words of the record of that row, data for column two is found in words five to eight, etc. Thus, there is a total of 1,024 32-bit words in each row. The format for the four words in each column is given in Table B-l. Word 1,024 of each row contains the row number. The higher order 16 bits of word 1,022 of row one contain the year of the century and the lower order 16 bits of the same word contain the day of the year. The climatology temperature is the clima- tology SST at the gridpoint. The average gradient is an average two-dimen- sional gradient around the gridpoint expressed in K/100 km. The day tag is the number of days since an observation was close enough to be used in the analysis for that particular gridpoint. The land-sea tag code is designated as follows: 0, land; 1, sea; and 2, shipping lanes. If the grid point is over land, all of word one will be zero. The reliability is a function of the density and quality of observations entering into the analysis. The highest reliability is given as 51.1. Words three and four are used inter- nally in the analysis and are of little concern to users. All temperatures are coded by subtracting 269.9 and multiplying by ten. The average gradient and reliability are also scaled by ten. 72 ROW 5l2n ROW 384 * to o or ROW 257 ROW 256 ROW 128 ROW SOUTHERN HEMISPHERE NORTHERN HEMISPHERE COLM. I COLM. 256 COLUMNS 130 o o o o o 129 o 0> oo oi o COLUMNS Figure B-l (top) depicts SST first guess field grid as overlayed on a polar stereographic projection Figure B-2 (bottom) shows grid box orientation. 7.3 Table B-l -- Global scale analyzed field packed data format Word Sits* Quantity Quantity range Coded range 1 23-31 1 14-22 1 8-13 ] 2-7 1 0-1 2 23-31 2 14-22 2 5-13 2 0-4 3 0-31 4 0-31 SST Climatology temperature Average gradient Day tag Land-sea tag Climatology change Verification temperature Reliability of SST Number of observations used in analysis at this grid point Analysis parameters Analysis parameters 270. 0-320. OK 1-501 270. 0-320. OK 1-501 0.0-5. 2K/100 km 0-52 0-63 days 0-63 0-2 0-2 Vacant Vacant 270. 0-320. OK 1-501 0-51.1 0-511 0-31 0-31 * Bit is the high order bit U. S. GOVERNMENT PRINTING OFFICE : 1976 — 210-801/288 74 (Continued from inside front cover) NESS 56 What Are You Looking at When You Say This Area Is a Suspect Area for Severe Weather? Arthur H. Smith, Jr., February 1974, 15 pp. (COM-74-11333/AS) NESS 57 Nimbus-5 Sounder Data Processing System, Part I: Measurement Characteristics and Data Reduc- tion Procedures. W.L. Smith, H. M. Woolf, P. G. Abel, C. M. Hayden, M. Chalfant, and N. Grody, June 1974, 99 pp. (COM-74-11436/AS) NESS 58 The Role of Satellites in Snow and Ice Measurements. Donald R. Wiesnet, August 1974, 12 pp. (COM-74-11747/AS) NESS 59 Use of Geostationary-Satellite Cloud Vectors to Estimate Tropical Cyclone Intensity. Carl. 0. Erickson, September 1974, 37 pp. (COM-74-11762/AS) NESS 60 The Operation of the NOAA Polar Satellite System. Joseph J. Fortuna and Larry N. Hambrick, November 1974, 127 pp. (COM-75-10390/AS) NESS 61 Potential Value of Earth Satellite Measurements to Oceanographic Research in the Southern Ocean. E. Paul McClain, January 1975, 18 pp. (COM-75-10479/AS) NESS 62 A Comparison of Infrared Imagery and Video Pictures in the Estimation of Daily Rainfall From Satellite Data. Walton A. Follansbee and Vincent J. Oliver, January 1975, 14 pp. (C0M-75- 10435/AS) NESS 63 Snow Depth and Snow Extent Using VHRR Data From the N0AA-2 Satellite. David F. McGinnis, Jr., John A. Pritchard, and Donald R. Wiesnet, February 1975, 10 pp. (COM-75-10482/AS) NESS 64 Central Processing and Analysis of Geostationary Satellite Data. Charles F. Bristor (Editor), March 1975, 155 pp. (COM-75-10853/AS) NESS 65 Geographical Relations Between a Satellite and a Point Viewed Perpendicular to the Satellite Velocity Vector (Side Scan). Irwin Ruff and Arnold Gruber, March 1975, 14 pp. (COM-75-10678/AS) 10678/AS) NESS 66 A Summary of the Radiometric Technology Model of the Ocean Surface in the Microwave Region. John C. Alishouse, March 1975, 24 pp. (C0M-75- 10849/ AS) NESS 67 Data Collection System Geostationary Operational Environmental Satellite: Preliminary Report. Merle L. Nelson, March 1975, 48 pp. (COM-75-10679/AS) NESS 68 Atlantic Tropical Cyclone Classifications for 1974. Donald C. Gaby, Donald R. Cochran, James B. Lushine, Samuel C. Pearce, Arthur C. Pike, and Kenneth 0. Poteat, April 1975, 6 pp. (COM-75- 1-676/AS) NESS 69 Publications and Final Reports on Contracts and Grants, NESS-1974. April 1975, 7 pp. (COM- 75-10850/AS) NESS 70 Dependence of VTPR Transmittance Profiles and Observed Radiances on Spectral Line Shape Parame- ters. Charles Braun, July 1975, 17 pp. (COM-75-11234/AS) NESS 71 Nimbus-5 Sounder Data Processing System, Part II: Results. W. L. Smith, H. M. Woolf, C. M. Hayden, and W. C. Shen. July 1975, 102 pp. (C0M-75-11334/AS) NESS 72 Radiation Budget Data From the Meteorological Satellites, ITOS 1 and NOAA 1. Donald H. Flanders and William L. Smith, August 1975, 22 pp. NESS 73 Operational Processing of Solar Proton Monitor Data. Stanley R. Brown, September 1975. (Re- vision of NOAA TM MESS 49), 15 pp. NESS 74 Monthly Winter Snowline Variation in the Northern Hemisphere from Satellite Records, 1966-75. Donald R. Wiesnet and Michael Matson, November 1975, 21 pp. NESS 75 Atlantic Tropical and Subtropical Cyclone Classifications for 1975. D. C. Gaby, J. B. Lushine, B. M. Mayfield, S. C. Pearce, and K. 0. Poteat, March, 1976, 14 pp. NESS 76 The Use of the Radiosonde in Deriving Temperature Soundings From the Nimbus and NOAA Satellite Data. Christopher M. Hayden, April 1976, 21 pp. NESS 77 Algorithm for Correcting the VHRR Imagery for Geometric Distortions Due to the Earth's Curva- ture and Rotation. Richard Legeckis and John Pritchard, April 1976, 30 pp. PENN STATE UNIVERSITY LIBRARIES A0D0D72Q mi23 * « '^ e -i9i«> NOAA--S/T 76-1770