t55. jAXNESS (*3 NOAA TR NESS 63 A UNITED STATES DEPARTMENT OF COMMERCE PUBLICATION / V \ NOAA Technical Report NESS 63 U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration National Environmental Satellite Service Verification of Operational SIRS B Temperature Retrievals HAROLD J. BRODRICK AND CHRISTOPHER M. HAYDEN WASHINGTON, D.C. December 1972 , *ii • w NOAA TECHNICAL REPORTS National Environmental Satellite Service Series The National Environmental Satellite Service (NESS) is responsible for the establishment and operation of the National Operational Meteorological Satellite System and of the environmental satellite systems of NOAA. The three principal offices of NESS are Operations, Systems Engineering, and Research. The NOAA Technical Report NESS series is used by these offices to facilitate early distribution of research results, data handling procedures, systems analyses, and other information of interest to NOAA organiza- tions . Publication of a report in NOAA Technical Report NESS series will not preclude later publication in an expanded or modified form in scientific journals. NESS series of NOAA Technical Reports is a continua- tion of, and retains the consecutive numbering sequence of, the former series, ESSA Technical Report National Environmental Satellite Center (NESC) , and of the earlier series, Weather Bureau Meteorological Satellite Laboratory (MSL) Report. Reports 1 through 37 are listed in publication NESC 56 of this ser- ies . Reports 1 through 50 in the series are available from the National Technical Information Service (NTIS) , U.S. Department of Commerce, Sills Bldg., 5285 Port Royal Road, Springfield, Va. 22151. Price $3.00 paper copy; $0.95 microfiche. Order by accession number, when given, in parentheses. Beginning with 51, printed copies of the reports are available through the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402. Price as indicated. Microfiche available from NTIS (use accession number when available). Price $0.95. ESSA Technical Reports NESC 38 Angular Distribution of Solar Radiation Reflected From Clouds as Determined From TIROS IV Radi- ometer Measurements. I. Ruff, R. Koffler, S. Fritz, J. S. Winston, and P. K. Rao, March 1967. (PB-174-729) NESC 39 Motions in the Upper Troposphere as Revealed by Satellite Observed Cirrus Formations. H. McClure Johnson, October 1966. (PB-173-996) NESC 40 Cloud Measurements Using Aircraft Time-Lapse Photography. Linwood F. Whitney, Jr., and E. Paul McClain, April 1967. (PB-174-728) NESC 41 The SINAP Problem: Present Status and Future Prospects; Proceedings of a Conference Held at the National Environmental Satellite Center, Suitland, Maryland, January 18-20, 1967. E. Paul McClain, October 1967. (PB-176-570) NESC 42 Operational Processing of Low Resolution Infrared (LRIR) Data From ESSA Satellites. Louis Rubin, February 1968. (PB-178-123) NESC 43 Atlas of World Maps of Long-Wave Radiation and Albedo — for Seasons and Months Based on Measure- ments From TIROS IV and TIROS VII. J. S. Winston and V. Ray Taylor, September 1967. (PB-176- 569) NESC 44 Processing and Display Experiments Using Digitized ATS-1 Spin Scan Camera Data. M. B. Whitney, R. C. Doolittle, and B. Goddard, April 1968. (PB-178-424) NESC 45 The Nature of Intermediate-Scale Cloud Spirals. Linwood F. Whitney, Jr., and Leroy D. Herman, May 1968. (AD-673-681) NESC 46 Monthly and Seasonal Mean Global Charts of Brightness From ESSA 3 and ESSA 5 Digitized Pic- tures, February 1967-February 1968. V. Ray Taylor and Jay S. Winston, November 1968. (PB-180- 717) NESC 47 A Polynomial Representation of Carbon Dioxide and Water Vapor Transmission. William L. Smith, February 1969. (PB-1 83-296) NESC 48 Statistical Estimation of the Atmosphere's Geopotential Height Distribution From Satellite Radiation Measurements. William L. Smith, February 1969. (PB-183-297) NESC 49 Synoptic/Dynamic Diagnosis of a Developing Low-Level Cyclone and Its Satellite-Viewed Cloud Patterns. Harold J. Brodrick and E. Paul McClain, May 1969. (PB-184-612) NESC 50 Estimating Maximum Wind Speed of Tropical Storms From High Resolution Infrared Data. L. F. Hubert, A. Timchalk, and S. Fritz, May 1969. (PB-184-611) (Continued on inside back cover) dD gMQSfrfe U.S. DEPARTMENT OF COMMERCE Peter G. Peterson, Secretary NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION Robert M. White, Administrator NATIONAL ENVIRONMENTAL SATELLITE SERVICE David S. Johnson, Director O NOAA Technical Report NESS 63 Verification of Operational SIRS B Temperature Retrievals Harold J. Brodrick and Christopher M. Hayden WASHINGTON, D.C. December 1972 UDC 551.524.77:551.501.74:551.507.362.2 551.5 Meteorology .501 Methods of observation .74 Upper air pressure-height measurement techniques .507 Instrument carriers .362.2 Meteorological satellites .524 Atmospheric temperatures . 7 Upper air temperatures .77 Vertical temperature gradients Mention of a commercial company or product does not constitute an endorsement by the NOAA National Environmental Satellite Serv- ice. Use for publicity or advertising pur- poses of information from this publication concerning proprietary products or the tests of such products is not authorized. For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402— Price 55 cents CONTENTS Abstract 1 1. Introduction 1 2. Verification method 2 3. Results 5 k- Summary" and conclusions 12 References 12 Appendix . 13 A. Errors in verification standard 13 B. Relationship between error ratio and correlation coefficient 14 C. Technique of scaling error ratio versus correlation coefficient 14 in Digitized by the Internet Archive in 2013 http://archive.org/details/verificationofopOObrod VERIFICATION OF OPERATIONAL SIRS B TEMPERATURE RETRIEVALS Harold J. Brodrick and Christopher M. Hayden National Environmental Satellite Service, NOAA, Washington , D.C. ABSTRACT. A verification procedure is applied to deter- mine the accuracy of the Satellite Infrared Spectrometer (SIRS) B temperature retrievals by measuring their use- fulness in the context of the National Meteorological Center ( Suitland, Md.) operations. Three pressure levels are considered. It is shown that temperatures derived from satellite data for application at 200 mb are definitely useful while at 5>00 and 300 mb, under certain cloud conditions, the SIRS temperatures are moderately useful. 1. INTRODUCTION Since the advent of the first experimental vertical sounding instruments, much attention has been focused on the Satellite Infrared Spectrometers (SIRS) A and B and the operational use of the temperature profiles retrieved from their radiation data. Our purpose here is to verify the SIRS B retrieved vertical temperature profiles provided operationally to the National Meteoro- logical Center (NMC) from July 1970 to March 1972. The technique by which temperature profiles were obtained from measured radiances is described by Smith et al. (1972). For the present, it is sufficient to know that the temperature profiles were obtained by an iterative procedure wherein a guess temperature profile was perturbed to fit the measured radiances to within experimental error. Before proceeding to a description of the verification technique, the reader should understand the fundamental tenet of this report. The verifica- tion must be carried out in dense data areas to have some standard against which to measure the accuracy of the SIRS data. Nevertheless, the verifica- tion technique is designed to determine the reliability of the temperature information provided by SIRS in areas both rich and sparse in conventional data . This point cannot be too strongly emphasized. Despite the fact that the verification is conducted in good data areas, the results are applicable to poor data areas as well. To understand the verification technique, one must have some familiarity with analysis procedures at the NMC. For each synoptic analysis time (0000 and 1200 GMT), there are two principal objective analyses, one based on data collected up to 3*5 hr after synoptic time (operational analysis) and the other on data collected up to 10 hr after synoptic time (final analysis). Conventional data used in the operational analysis are reused, together with data received later, in the final analysis. SIRS data are used in both, but are processed somewhat independently in that different first-guess tempera- ture profiles are used in the retrieval process. SIRS B temperature retrievals made for the operational analysis use a guess temperature profile derived from 12-hr forecast temperature fields predicted from the previous synoptic analysis. The temperature retrievals made for the final analysis use a guess temperature profile derived from the temperature fields of the operational analysis . The verification is a comparison of two temperature differences -- the difference between SIRS-retrieved temperatures and the NMC ( 12-hr) forecast temperatures versus the difference (forecast error) between the "true" tem- perature (to be explained in the next section) and the NMC (12-hr) forecast. Moreover, because two nearly independent SIRS retrievals are made, two categories of comparisons can be made: 1. comparisons for retrievals based on forecast temperature profiles and 2. comparisons for retrievals based on concurrent analyzed profiles. Comparison 1 indicates the effectiveness of the truly operational SIRS product while comparison 2 serves in effect as a limit to which comparison 1 can be expected to approach and also as a verification of the final SIRS retrieval data, which is the archived product. Because we are evaluating the SIRS retrievals with respect to forecast error, we feel that the results of comparison 1 are applicable to sparse data areas even though they are computed in good data areas. The fact that short- term forecasts may be on the average more accurate in good data areas is of no consequence since we have been able to include cases where the forecast errors are relatively large. The only reason to suspect that SIRS modifica- tions to the guess profiles might be different over sparse data areas is that terrestrial effects, which occur in the dense data areas, introduce some addi- tional uncertainty in the retrieval procedure. For this reason, soundings over oceans might be slightly more consistent than those over land; but it is not presently possible to substantiate this hypothesis. 2. VERIFICATION METHOD In the following evaluation, we shall be primarily concerned with two verifi- cation parameters, (l) the correlation coefficient (corr. coeff.) between the forecast error and the SIRS difference from the forecast and (2) the ratio of the standard deviation (std. dev.) of the forecast error to the root mean square (rms) error of the SIRS. Determination of errors implies a knowledge of the truth. For the "truth," we used values obtained from the NMC objective analyses, but with a slight adjustment to compensate for smoothing accomplished by the objective analysis procedure. It is not our purpose here to discuss the relative merits of this smoothing but only to comment that, for the purpose of evaluating a point data source by reference to values interpolated in objectively analyzed fields, the smoothing implicit in the latter should be removed. Evidence of the smoothing can be seen in figure 1. This figure depicts for three levels the scatter of the difference from the 12-hr forecast indicated by radiosonde reports as compared to the differences indicated by colocated values interpolated in the operational objective analyses. The departure from a slope of 1 shown by the to < LU o O Z o CO q < Cl. 1 1 11 1 2 1 ♦ 2 11 111 12. 1 1 1 222 2- 1 2 12 211 131 2111 1 1 1231222112 211 1*11 1 1 115718731 2 1 11111 1141 2 2 2 39761*1 1 1 22351 31 1 1 133*175*311 1 1 11112 2 1 1 11 238 821 2 211152212211 1 3 22513 2 HI 1 1331* 1 21 1 5221222 2322 1 1 111 ♦ i 1 11 •2 1 3 11 1 1 1 1 1 1 1 1 1 1 1 1 -5.50 -3.70 -1.91 -0.11 1.68 ANALYSIS - FORECAST 500 mb 500 mb 300 mb 200 mb Corr. Coeff. 0.87 0.89 0.92 rms 0.84° 1.02° 1.23 Std. Dev. Analysis 1.32° 1.69° 264 Std. Dev. Radiosonde 1.69° 2.15° 3.13 c Regression Slope 1.11 1.13 1.09 Regression Scatter 0.83° 0.98° 1.23 i— 3.7. 1 1 co 2 11 < . 11 2* U LU 2.2. 1 111 1 12*1 1 1 1 111* 2 1 1 1 11121* 1 u_ ; 1 1 1 121 1 1 1 1 1*1 0.6 I 11123222122 LU , 1121*222 1 a , 131 13*211 z o , 5526231 11 , 1 13651213 -1.0» 1 1253531*32 1 1 CO , 1155762* 1 1 o . 31523612 1 . 11212** 1 Q . 623*2211 < -2.6. 12*2*2 2 11 , 132*12*2 • 2212 211 . 1M3 1 11 1 1 , •112 1 -*.2» 2 211 *U2 2 . • 1 122 2 • 11 1 1 . 1 1 ANALYSIS - FORECAST 300 mb 9.1< 1 ♦ 1 1 ♦ 6.9 . 1 ♦ 1 « 1 1 23. 1 111 *.7 1 1 1 1*1 111 21 123 111 1 1 112312 11 1 1221**22 1— CO < 2.5 1 1 1356 1 12 133672 1 1 1 3*53211 1 3**»321 11 U 1 11183 111 LU 0.3 2*61 31 Q£. 1 1 5633111 O 2 323*121 ' 1 21 * 26* 221 21123*12 1 -1.9 • 11*12 1 1 II 1 12363211 11 a z 22S 21 22233*1 1 1 212 2 11 o CO O -*.l • •2112211 1 •113 1 111 1 11 !!• 1 2 • 11 1 Q -6.3 • 1. 1 2 < • 1 Q£ . • 1 1 1 1 • 1 -8.5 !• -10.7 1 -12.9 -11 • 1 .6 -7.5 -3.* 0.6 *•' 8.8 Values, °C ANALYSIS - FORECAST 200 mb Figure 1 . --Radiosonde versus the objective analysis change to the 12-hr fore- cast. The line shown with the + symbols depicts 1:1 correspondence. regression slopes in the figure indicate an average smoothing of the objec- tive analysis procedure. The scatter of the points indicates a random error of either the radiosondes or the analysis. Clearly, the smoothing bias should be accounted for in the verification of the SIRS retrievals. For an ensemble, this can be done by including the re- gression line slopes of figure 1 in the statistical comparisons as follows: The expression for the rms difference between SIRS and the objective analysis values can be expressed after some manipulation as rms = a 2 + a 2 - 2ro a + (m - M \ 2 1/2 (l) s n s n \ s n ) where a is the standard deviation of the SIRS sample (SIRS-forecast), a n is the standard deviation of the analysis sample (analysis - forecast), r is the cor- relation coefficient between the two samples, M s is the mean deviation of the SIRS sample, and M n ±s the mean deviation of the NMC sample. All of the above parameters are computed directly from the samples. The final rms difference between the SIRS and the adjusted analysis values is computed from eq (l) modified such that the standard deviation of the analysis sample is multiplied by the appropriate regression line slope of figure 1. The adjusted rms value is our estimate of the SIRS error. Note that, in boldly defining the term "SIRS error," we have ignored uncer- tainties engendered by radiosonde observational error, space, and time interpolations in the analyses. These uncertainties cannot be readily esti- mated, but we feel that they make a negligible contribution to the correlation coefficient or error ratio. For the interested reader, we have included a discussion of this assumption in the appendix where we show how one may properly include his estimates of uncertainty and so personally adjust the verification parameter reported here. The error ratio (a n /rms, adjusted by the slope factor) is a straightforward estimate of the utility of the SIRS. A ratio greater than unity signifies that the forecast error is greater than error contained in the SIRS estimates. In such a case, the SIRS can be said to be useful in correcting the forecast. It is difficult, however, to define a degree of utility from the magnitude of error ratio. For this reason, we are also presenting the correlation coeffi- cient that we hope will aid the reader's interpretation of the results. A discussion of the relationship between the error ratio and the correlation coefficient is included in the appendix. In the following section, the results of this report are summarized in graphs presenting together the correlation coefficient and the error ratio. This allows the viewer to approximate the correlation coefficient correspond- ing to the point where the SIRS sample begins to improve on the forecast (i.e., an error ratio equal to unity) and thus to better judge how much the SIRS values improve on the forecast. The method of scaling the two parameters is discussed in the appendix. 3- RESULTS The results presented here are from six samples taken during the period July 20, 1970, through March 21, 1971. The data used were collected up to 6 hr prior to analysis time in areas with reasonably dense conventional data coverage. Retrievals affected by high clouds (cloud pressure < ^0 mb and . cloud amount >10 %) , as defined by the retrieval procedure, were not accepted. SIRS and forecast errors as described above were correlated for both heights and temperatures at the 5>00-, 300- , and 200-mb levels; and the thickness and mean temperatures for the £00- to 300-mb and 5>00- -50 200-mb layers. However, only the results of temperature verifications are presented in this report. Constant pressure height results are discussed elsewhere by Hayden (1971). Three cloud condition categories are used to break down each of the samples and are defined as: (l) clear, (2) lower clouds below 700 mb in height, and (3) higher clouds above 700 mb, but excluding the above-defined high cloud category. These are as inferred by the SIRS retrieval model for the final SIRS retrieval. Table 1 presents the sample size and time period for each of the six samples along with the number in each of the cloud categories. The summarized results are presented with limited discussion as it is be- yond the scope of this report to explore all causes behind the trends or relative comparisons of the verification of the various parameters. We are treating a research instrument that was used operationally; manifold changes were made in the retrieval processing during the time period we are here dis- cussing. For those interested in some detail, the basic scatter diagrams of temperature differences (uncorrected by slope factor) for the SIRS B data processed for the operational analyses and a complete set of the statistics for both operational and final retrievals are shown in the appendix. The scatter diagrams in figures 2 and 3 are illustrative of the results obtained from the study. These distributions are from the March sample for all cloud conditions combined and for only the cloud- free conditions. For the 500- and 300-mb diagrams, the regression slopes of <1 indicate that the SIRS retrievals overspecified the analyzed changes to the forecast tempera- Table 1. --Period and size of each sample with the number in each cloud category Sample Period Total size Clear Lower Higher 1 July 20-Sept. 2k, 1970 180 3k kS 101 2 Sept. 25-Nov. 8 226 26 $6 Ikk 3 Dec. 10- Jan. h, 1971 218 k3 91 8k k Jan. 9-Jan. 29 2U3 31 6y 11*3 5 Feb. 2-Feb. 22 21k 29 8U 161 6 Mar. 1-Mar. 20 258 19 70 169 < U LU Cfc o 00 oo < z < 1 1 1 1 1 1 1 1 1 11 1 1*1 1 1 1 *1 ] 11 111 1 11 1*1 1 1111 1*2 1 1 1 2 1 11 • 1 12123 111112 1 111 11 UH 11 1 112 1 11 13 11*21 1 1 2 311 *1 1 2 12 1* 1 1 1 12311 31 2 1 212 2 11121 11 11 1 12 1 111113 1 11 1 211111 2 3 1 12' I 1 1 12> 1 2 1 1 1 11 11 1 ♦ 2 1 1 1 ♦ 1 1 1* 1 1 1111 1 1 1*> 1 1 1 I 1 11 1 500 mb 300 mb 200 mb Corr. Coeff. rms Std. Dev. Analysis Std. Dev. SIRS Regression Slope Regression Scatter 0.63 0.50 0.82 1.91° 2.05° 2.43° 1.45° 2.18° 4.13° 2.23° 1.79° 2.91° 0.41 0.61 1.16 1.12° 1.88° 2.38° SIRS - FORECAST 500 mb • 11 11 •! 3.0* 1 1 1 1 1 1 1 1 ♦ ♦ 1 1 11 ♦ 11 . 1 1 111 1* 1 1.9. 1 11 211 1*1 11 2 H- • 1 1 1*1 22 1 OO . 1 1 121111 < 0.7 • 1 1 1111 121* ♦ 1 2 1 1 2 1 1 U 1 1 22 3111 LU . 21 I 1 Q£ 1 1 HI Hill o 1 •11 112111 1 -0.* ■ 1 1 11 1 1 1 11 1 1 1 ♦ 1 2 1 11 1 1 1 21 • 1 1211 1 1 OO 1'211 1 1 1 1 . 1 1 1 1 1 00 -1.6" 1 1 2 2 1 j > 1 1 1 1 —J 1 • 1 1212 < 1 11 11 1 1 1 1 1 Z -2.7. 1 11 • 1 1 1 21 1 1 1 < . 1 1 1. 11 •21 1 1 3 1 1 1 11 1 1 11 1 1 1 1 1 -6.2. 1 -».90 SIRS - FORECAST 300 mb CO < u LU o oO 0O >- < Z < i i u iiii i mi a ii i u i ♦ 1222 1 1 • 12 12 • 2 12 11. 2 I 22 ZZ 122*1 22 3*1121 121 1312*2121111 11 1 2*2 1 1 132 '-11 1 1 1 1 1* 1 11 212 1 1 1 12 1 1 2124 II 1 1*1 11 11 112 1 113 11 1 *11U11122 1 1 111 1 • 1 1 21 1 2 1 1 1 1 111111 1 • 1 1* 1 1 1 1 I 2 1 1 1 1 11 11 11 SIRS - FORECAST 200 mb Figure 2. --SIRS versus the analysis change to the 12-hr forecast for all conditions, March 1971 sample cloud 1 .drt • 1 I.«7 ♦ 1 • 1.07 : 1 • 1 1 ♦ 1 0.66 * t- # <2 0.25 ♦ < 1 u . Ul • 1 1 £* . £-0.16. • IX. 1 ^-0.5?: 1 , to . >- ♦ _J * < ♦ •7 -0.98 « i< 1 1 < •* 1 -1.38* 1 • -1.79. ♦ -2.20* 1 ♦ ■l.*l -0.13 1»16 2.** SIRS - FORECAST 500 mb 500 mb 300 mb 200 mb Corr. Coeff. 0.87 0.66 0.85 rms 1.00° 2.03° 1.94° Std. Dev. Analysis 1.15° 2.32° 3.30° Std. Dev. SIRS 1.77° 1.57° 2.65° Regression Slope 0.56 0.97 1.06 Regression Scatter 0.57° 1.75° 1.73° to < U m D£. o to >• < Z < -2.30 -1.20 -0.10 »•<•» Z' 11 3-21 SIRS - FORECAST 300 mb 5.0 1 I 1 1 3.9 1 I 1 • ♦ 1 ♦ 2.7 1 ♦ 1 l.S 1 ♦ ♦ 0.3 ♦ -0.9 1 • ♦ -2.1 • 1 ♦ -3.3 1 ♦ -4.5 I -5.7 1 -5 30 -3.*6 -1.63 0.21 2. OS 3-80 SIRS - FORECAST 200 mb Figure 3. --Same as figure 2 except this is for clear cases A S O (D) HIGHER Figure 4. --Correlation coefficient for operational 500-mb tempera- ture retrievals (lower solid line) and for final retrievals (upper solid line) ; error ratio for operational retrievals (lower dashed line) and for final retrievals (upper dashed line) tures. At 200 mb, the slope of >1 indicates underspecification of analyzed temperature change. Consideration of the radiosonde-analysis slope factor (i.e., multiplying the regression slopes by the slope factor) makes the over- specification look more reasonable at 500 and 300 mb, although it increases the underspecification at 200 mb. The graphs of correlation coefficient and error ratio for the three levels and all cloud categories are shown in figures ks $s and 6. Two sets of statistics are presented in each graph, corresponding to the operational and final retrievals. The latter are always on top since they are always "better" from the advantage of a more accurate first guess. In many cases,the sta- tistics for the final retrievals are off the graph scale, which was designed to emphasize the results of the operational retrievals. Let us consider first the operational retrievals. Note that, for the total samples of the operational retrievals, the results are superior at 200 mb (compare figs. I4A, 5>A, and 6A). Substantial improvement over forecast temperatures is demonstrated by the SIRS retrievals. It is also encouraging to note that, at £00 and 300 mb (figs, kk and 5A), the operational retrievals are comparable or tend to add some information to the forecast temperature values, particularly in the last three or four samples at 500 mb. Clearly, however, cloud contamination is reducing the effectiveness of the SIRS at these levels. The slightly better results at 500 mb compared to those at 300 mb is somewhat puzzling but may be due to the lack of sufficient vertical resolution in the area of the tropopause because of the absence of the 692 cm -1 channel radiance observations in the retrievals (Smith et al. 1972). For the three levels, the remaining portions of figures k, $> and 6 display the samples broken down according to cloud conditions. Several interesting quirks are evident, emphasizing the difficulties caused by cloud contaminations. 2.29 2.17 2.54 1.0 Q.9 0.8 0.7 6 0.5 0.4 O.3UADIOSONDE VB. ii.VW AHA! :.!', : 0.89 0.2|- J- /rms L 0.1 A S O N D (A) ALL CASES 2,14 J F M 2.10 3.04 2.73 A S O N (B) CLEAR 2 09 2.03 2.87 1.0 <\ 1 1 / " 0.9 - . N ^ / r- — "* . 0.8 0.7 0.6 \ \ _ \ y 1 \ 1 M _ 0.5 "^S. %s* ' - 0.4 3 - 2 . 1 1 1 1 1 1 1 1 &/' A S O N D (D) HIGHER Figure 5. --Same as figure 4 except this is for operational 300-mb temperature retrievals 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 - 0.1 3.31 3.35 RADIOSONDE vs. NMC ANALYSIS 1' r = 0.92 o-jrrns; 2.35 o n 6 I I I I I I S O N D A) ALL CASES 310 3.44 2.46 J F 3.54 . A S O N D J (B) CLEAR 3.86 3.81 3.50 2.23 : J I I I L 1.8 1.6 1.4 Cn/ rmS 1.2 1.0 ■08 0.6 A S O N 0) HIGHER D J F M Figure 6. --Same as figure 4 except this is for operational 200-mb temperature retrievals 10 Two points especially require some clarification, (l) While at £00 and 300 mb the operational retrievals obtained from clear-sky measurements are generally better than the retrievals obtained from cloud- contaminated mea- surements, there are periods at 200 mb when the reverse is true. (2) Con- sistently at 200 mb and occasionally at the other levels , the operational retrievals obtained with higher cloud conditions are superior to those ob- tained under lower cloud conditions. Both these results are unrealistic. On the average, the no-cloud solutions should always be best; and the solutions should deteriorate as higher clouds occur and more opaque channels are influenced. The first of the above apparent paradoxes can be explained if we hold to the hypothesis that, when the error ratio is near or00 ^3 300 mb, and to a slight extent at 200 mb, may be attributed to stratospheric warming events that could not be retrieved accurately at high levels (due to erroneous guess temperature profile), with the resultant errors propagating into the tropospheric levels. Error ratios and correlation coefficients computed for the radiosonde samples of figure 1 are shown at the bottoms of figures kA, 5>A, and 6A. It is interesting to note that, for all samples at 200 mb and in some of the cloud breakdowns in the later samples at 500 and 300 mb, the final SIRS data are more consistent with the "truth" (as defined) than are the radiosonde data. 11 Table 2. --Average rms errors for mean temperatures (in deg.) of the specified layers and for individual levels ' Mean temperature rms Temperature rms 850-500 mb 850-300 mb 850-200 mb 500 mb 300 mb 200 mb All cases 1.8 1.5 1.1 1.6 2.0 2.6 Clear 1.6 • 1.3 1.1 1.2 1.8 2.6 Lower 1.5 1.2 0.9 1.3 2.0 2.U Higher 2.1 1.7 1.2 1.9 2.0 2.7 The improvement of the statistics of the final retrievals over those of the operational retrievals demonstrates a bias of the SIRS retrievals to their initial guess temperature profiles. This is entirely to be expected since both the "minimum information" method of inversion and the method of cloud correction result in a bias toward the guess. Previous discussions of SIRS accuracy (Wark 1971) indicated expected and observed accuracies for rms temperature errors at both individual levels and in layers. For comparative purposes , rms errors (averaged for all the samples) are presented in table 2 for each of the three levels individually and for a layer from 850 mb to each level. The mean temperature rms errors were com- puted from height rms errors for the last five samples in which the 850 mb NMC height was used as the reference height to hydrostatically build up the SIRS heights. As indicated by Wark, the mean errors decrease with increasing layer thickness for all cloud conditions. Puzzling is the tendency for the lower cloud category to yield results better than the clear category; but this may again be due to the inclusion of an additional channel, improperly cor- rected, in the latter retrievals. The individual level rms errors in table 2 increase with height since the upper two levels are in the vicinity of the tropopause where errors have been consistently greater according to previous papers (Wark 1971 and Smith et al. 1972). Consequently, for layers thicker than the 850- to 500-mb layer, the mean rms errors become less than the individual level errors at the top of the layer. The magnitude of the mean errors in table 2 are somewhat larger than those suggested by Wark, but we should again blame cloud contamination. Finally, we would like to reassert our contention that, for operational retrievals, these results apply equally in data-poor as well as data-rich areas. As partial substantiation to this claim, we should mention that we have applied the verification techniques to samples with larger forecast errors simply by excluding those cases with small observed. error. In this way, we have to some extent simulated poorer forecasts that might be expected in data-sparse regions. Results for the restricted samples are almost indistinguishable from those for the total samples. 12 U. SUMMARY AMD CONCLUSIONS The NMC forecast is used as a basis for obtaining the SIRS B operational retrievals and as a first guess for the succeeding constant-pressure analyses. The theme of the verification scheme presented here is simply to measure the ability of the SIRS retrieved data to add information to the NMC analysis first guess. At 200 mb, the results are quite encouraging, showing a generally good ability to add significant information to the guess temperature profiles, except for an indication of some sensitivity in SIRS-deduced clear conditions where the lack of proper corrections to the two most transparent carbon- dioxide channels (wave nos. 73U and 750) evidently was detrimental to the retrievals. The retrieved temperatures at 500 mb show consistent ability to properly correct guess profile errors in the clear and lower cloud cases. The higher cloud cases display a slight amount of additional information only in the later samples. At 300 mb, somewhat less information is added to the guess temperatures than at the other two levels, though the clear cases give generally good and occasionally very good results. The cloud- contaminated cases tend to yield a modest correction of guess temperature errors, but the results are mixed. That is, in some samples, lower cloud conditions give better results than the higher (as in the 500-mb situation); in others, higher cloud retrievals are better (as at 200 mb). The results rather clearly indicate that a primary difficulty in obtaining accurate and useful retrievals lies in the cloud problem, both because of improperly correcting observed radiances to account for cloud contamination and because the loss of information in the retrievals when strongly cloud affected SIRS channels are not used. This difficulty should be greatly alleviated with future, higher resolution sounders that will probe between the clouds. REFERENCES Hayden, Christopher M. , "On Reference Levels for Determining Height Profiles From Satellite-Measured Temperature Profiles," NOAA Technical Memorandum NESS 32, National Environmental Satellite Service, National Oceanic and Atmospheric Administration, U.S. Department of Commerce, Washington, D.C., Dec. 1971, 15 pp. Smith, William L., Woolf, Harold M. , and Fleming, Henry E., "Retrieval of Atmospheric Temperature Profiles From Satellite Measurements for Dynamical Forecasting," Journal of Applied Meteorology , Vol. 11, No. 1, Feb. 1972, pp. 113-122. Wark, David Q., "Annex G, the Accuracy of Satellite Infrared Sounders" in "The Feasibility of the First GARP Global Experiment (FGGE) and the Criticality of Initiating Systems Planning," COS PAR Working Group 6, a Report to JOC , Global Atmospheric Research Program, Geneva, Switzerland, Jan. 1971, h pp. 13 APPENDIX In this section, we shall address three of the more technical details not covered in the main body of the text; and we shall present statistics for operational and final retrievals and scatter diagrams for cloud categories of operational retrievals. A. Errors in Verification Standard A clearly defined bias in the objective analyses of temperature that we used as the verification standard has been accounted for. Less well defined errors introduced by radiosonde measurement error or space and time inter- polation have been ignored. We shall indicate here how the latter could be included in the correlation coefficient and error ratio. A correlation coefficient that includes additional uncertainty in the objective analysis can be computed from the sample correlations as A r - v (2) where e 2 is the variance of the additional uncertainty; C se is the covariance of the uncertainty and the SIRS sample; and C sn is the covariance of the SIRS and analysis samples. Other variables retain their previous definitions. It is apparent from eq (2) that random errors would absolutely cause improvement only in the absence of a positive correlation between the SIRS indicated changes and the uncertainty. It is not likely that correlation exists from a bias in radiosonde measurements, but there could well be a correlation with the errors introduced by space and time interpolation of the objective analyses. Similar terms are added to the error ratio. Including the uncertainty, the ratio is given by L/2 (3) The error ratio will be improved (increased) in only those samples where SIRS is already shown to be beneficial (i.e., those cases where the ratio is greater than unity) and then only if the SIRS indicated changes and the uncertainty are poorly correlated. In presenting our results, we have chosen to ignore the additional uncer- tainty terms. We believe them to have a very small effect on the magnitude of the correlation coefficient and on the question of whether the error ratio is greater or less than unity. 14 B. Relationship Between Error Ratio and Correlation Coefficient If we assume that the mean forecast error is zero (as can be shown empirically),, the correlation coefficient can be expressed by £2 2 2 o^ + o - rms + M r = s n ± 2o c s n (h) If we make the approximation that the mean SIRS error is zero, the correlation coefficient can be written in terms of ratios r = — + fl - -J I (5) where R s is the ratio of the standard deviations given by the SIRS and analy- sis indicated changes to the initial guess and R is our defined error ratio. In general, the error ratio increases as the correlation coefficient increases, but there is no direct correspondence. For example, if the error ratio is held at unity, the correlation coefficient can vary as the standard deviation of the SIRS changes varies. The error ratio and correlation coefficient give equivalent information only when the standard deviations of SIRS and analysis changes are of the same magnitude. C. Technique of Scaling Error Ratio Versus Correlation Coefficient To plot the correlation coefficient and the error ratio on the same graph , we used the following scaling for figures hi $, and 6. For all graphs, the same linear distance represents a given increment in the error ratio or in the correlation coefficient. The increment in both cases is the standard deviation of all correlation coefficients or error ratios obtained from the operational retrievals of individual months, levels, or cloud categories. This ensures that, on the graphs, the correlation coefficient and the error ratio are scaled approximately equivalently. On individual graphs repre- senting a single level and cloud condition, the scale of the .error is moved such that the average operational correlation coefficient (from the 6 mo of data) lies opposite the average operational error ratio. 15 Sample l.--July 20-September 24, 1970 Cloud Corr. coeff. rms Std. dev. analysis Std. dev. SIRS Mean error t Regression conditions Slope Scatter 200 mb All cases 0.79* .96t 2.1*6° 0.99 3.00° 3.00 1.83° 3.25 -1.53° -O.36 1.29 0.89 1.85° 0.8U Clear .69 .96 3.17 1.18 2.82 2.82 1.28 2.80 -2.3U -0.88 1.52 0.97 2. 0i| 0.78 Lower .82 • 95 2.07 0.87 2.56 2.56 1.83 2.85 -1.1*3 -0.11 l.lil 0.86 1.1*8 O.76 Higher .82 .97 2.3k 0.97 3.22 3-22 1.96 3.53 -1.30 -0.30 1.35 0.88 I.82 O.82 300 mb All cases Clear Lower Higher 500 mb All cases Clear Lower Higher 0.51* .88 1.32° O.83 1.1*6° 1.1*6 1.2k 9 1.68 -0.17° - .25 .61* .77 1.23° 0.69 .61* .89 1.22 0.76 1.50 1.50 1.20 1.65 - .31 - .12 .79 .81 1.15 0.68 ^9 .89 1.19 0.72 1.39 1.39 1.11 1.55 .25 - .07 .71* .80 1.13 0.614 .51 .88 1.U1 0.89 1.1*7 1.1*7 1.31 1.73 - .32 - .37 • 58 .75 1.26 0.69 o.l*l* .80 1.23° 0.93 1.10° 1.10 1.20° 1.51 - .13° .17 .1*0 .58 .99° .62 .77 .89 1.02 0.86 .86 0.86 1.09 .1*8 .71* .62 .60 .68 .51 .89 1.16 0.67 1.22 1.22 1.08 1.1*1* - .20 .17 .57 .75 1.05 0.55 .1*0 .80 1.35 0.99 1.11 1.11 1.28 1.61 - .30 - .02 .35 1.02 0.67 ""Upper number refers to operational retrievals, t Lower number refers to final retrievals. *Mean error =M - M . (See text.) s n 16 Sample 2. --September 25-November 8, 1970 Cloud Corr. coeff. rms Std. dev. Std. dev. analysis SIRS Mean error t Regression conditions Slope Scatter 20 ° mb 0.79* 2.1+9° 3.79° 2.17° -0.1+1+° 1-38 2.31° All cases Clear Lower Clear Lower Clear Lower .96 1 1.18 3.79 i|.03 - .1+3 0.90 1.03 •57 2.63 2.51 1-39 -I.63 1.03 2.06 .90 1.39 2.51 2.55 -0.80 O.89 1.10 .80 2.08 3.27 1.91 - .18 1.38 1.9U .96 1.07 3-27 3.61 - .21 0.87 0.93 .81 2.61 1+.09 2.37 - .33 1-39 2.1*2 Higher >9? lml9 htQ9 ^ 35 _ ^ 0#91 1>03 30 ° mb .1+3 1.80° 1.68° 1.51° - .59° .U8 1.51° All cases .85 1.11+ 1.68 2.06 - .36 .69 0.87 .kk 1.71 1.57 1.06 - .89 .65 Lip. .81 1.09 1.57 1.72 - .36 .73 0.92 .2+3 1-53 1.51 1.29 - .27 .50 1.36 .87 0.90 1.51 1.81 .05 .72 0.75 h^w -U2 1-91 1-72 1.61 - .65 ,h$ 1.56 Higner <86 ± ^ 2 ± ^ 2 2 ^ _ ^ ^ Q ^ Q SOOmb .38 1.1+2° 1.18° 1.35° -10° .33 1.09 c All cases ^ -^ 1>l8 128 1#22 1>86 _ #21 #ij9 0#83 * Upper number refers to operational retrievals, t Lower number refers to final retrievals. tTCean error = M - M • (See text.) 17 Sample 3. --December 10-January 4 1971 Cloud conditions Corr. coeff. rms Std. dev. Std. dev. analysis SIRS Mean error $ Regression Slope Scatter 200 mb 0#69 * 2# £ 7 „ 3^2° 2-56° -0.3l+° 0.95 2.51+ All cases #Q0+ 1#QQ 3 ^ 2 j^ #6 q #?8 lm $ 3 Clear Lower Lower All cases Clear Lower .78 2.11+ 3.38 2.75 - .37 .96 2.11 .92 1.77 3.38 1+.08 .67 .76 1.32 .62 2.71 3.1*3 2.1+6 - .05 .87 2.69 .88 2.01+ 3-143 3.82 .99 .79 1.60 „. . .72 2.61 3.66 2.1;5 - -61; 1.08 2.53 Higher #91 1#80 3#66 ^ #l8 >39 0>7 q 1# £ 3 300 mb >36 2#2il o 1#?3 o 1#93 o #8 ^o #33 1#6l c All cases ^ 1>?3 1>?3 2# ^ #56 #53 1#l6 n1 .1+6 2.00 1.62 1.70 1.01 .iUi 1.1+3 uiear ^ ± ± ^ -^ 2>0Q Q ^ ^ 0>Q £ ,19 2.1+1+ 1.55 1.93 -97 .15 1.52 .67 1.91 1.55 2.26 .88 .1+6 1.16 „. . .1+2 2.13 1.88 1.90 .65 .1+2 1.71 Higher >?6 1;?1 1#88 2#62 #21 % t£ lm22 50 ° mb .59 1.68° 1.30° 2.06° - .19° .37 1.05° .72 1.66 1.30 2.27 - .1+2 .1+1 0.90 .71 1.1+3 1.16 1.87 .52 .1+1+ .82 .77 1.33 1.16 2.01 .01+ .1+5 .73 .67 1.1+2 1.31 1.91 - .11+ .1+6 .97 .73 1.55 1-31 2.15 - 40 .1+1+ .90 „. , .1+7 2.01 1.30 2.15 - .60 .28 1.15 higner ^ Q ^^ ± ^ ^ _ ^ ^ Q Q ^ 3 * Upper number refers to operational retrievals. tLower number refers to final retrievals. tMean error = M - M . (See text.) s rz 18 Sample 4. --January 9-January 29, 1971 Cloud tions es Corr. coeff. 0.76* .96t rms 2. hi 1.20 Std. dev. analysis Std. dev. SIRS Mean error* Regression condi Slope Scatter 200 mb All cas 3.58° 3.58 2.55° 3.93 -0.52 .28 1.06 0.87 2.35° 1.05 Clear .82 .98 2.32 0.89 3.08 3.08 2.82 3.57 -1.1*7 0.28 .90 .81* 1.76 0.61* Lower .61* .93 2.19 1.25 2.76 2.76 2.10 3.09 - .Ii7 .1*3 .81* .83 2.11 1.01* Higher • 79 .96 2.53 1.23 3-98 3.98 2.65 U.30 - -33 .21 1.18 0.89 2.1*6 1.11 300 mb All cases .91 2.11° 0.96 1.89° 1.89 1.69° 2.23 .97° .06 .51 .77 1.63° 0.80 Clear .79 .97 2.11 0.70 1.92 1.92 1.60 2.35 1.75 -0.07 .79 1.18 0.1*9 Lower .90 2.25 0.99 1.89 1.89 1.50 2.18 1.25 0.32 .51 .78 1.73 0.81 Higher • i;l .89 2.01+ 0.99 1.83 1.83 1.71 2. lit .67 - .01* • 76 1.67 0.81* 500 mb All cases .70 .86 1.68° 1.30 1.51° 1.51 2.22° 2.29 - .57° - .31* .1*8 .57 1.07° 0.77 Clear .87 .91 1.23 1.03 1.61 1.61 2.22 2.19 .1*7 .23 .63 .66 .79 .68 Lower .72 .91 1.15 0.85 1.36 I.36 1.61 1.82 - .11 - .21 .61 .68 •57 Higher .69 .81* 1.96 1.51 1.56 1.56 2.32 2.1+6 -1.02 -0.52 .1*7 • 53 1.12 0.81* * Upper "Lower number number refers refers to operational retrievals, to final retrievals. jMean error = ■ M - M s . (See n text. ) \ 19 Sample 5. --February 2-22, 1971 Cloud Corr. Std. dev. Std. dev. Mean Regre ssion conditions coeff. rms analysis SIRS error* Slope Scatter 200 mb All cases 0.72* .96 + 3-32° 1.29 l+.08° I+.08 2.75° 1+.32 -1.77° -O.I4.2 1.07 0.91 2.81° 1.15 Clear .62 .95 3.68 1.27 3.20 3.20 2.1+8 3.58 -2.63 -0.63 .79 .85 2.52 O.96 Lower .72 .96 2.89 1.06 3.66 3.66 2. 31+ 3.82 -1.35 -0.16 1.13 0.92 2.5U 1.00 Higher .75 .96 3-14-6 1.1+0 1+.39 1+.39 2.91 1+.65 -I.83 -0.51 1.13 0.91 2.92 1.23 300 mb All cases • 51 .93 2.16° 1.31 2.1+1° 2.1+1 1.78° 3.02 .28° - -$9 .69 .71+ 2.07° 0.88 Clear .73 • 9h 1.60 1.95 2. 111. 2. Hi I.82 3.39 .62 -i.m .86 ^9 1.1+5 0.76 Lower .1+8 2.15 0.88 2.1+0 2.1+0 1.51i 2.69 0.22 - .27 .75 .85 2.11 0.7^ Higher .1+7 .92 2.25 1.36 2.U0 2-U0 I.83 2.98 .25 - .66 .61 .71+ 2.12 0.91 500 < ;s - All cases .61 .80 1.81° 1.36 1.50° 1.50 2.19° 2.22 - .1+7° .02 .1+2 1.20° 0.90 Clear .82 .82 1.21 1.22 1.37 1.37 I.83 2.01+ • 58 .08 .61 .78 .79 Lower .70 .85 1.3U 1.12 1.1+1 1.88 2.03 .08 .02 .$3 .$9 1.00 0.75 Higher • 5ii .77 2.09 1.1+8 1.51+ 1.5U 2.18 2.31 - .9h .00 .38 .51 1.29 0.97 * Upper number refers to operational retrievals, +Lower number refers to final retrievals. tMean error - M - M . s n (See text. ) 20 Sample 6. --March 1-20, 1971 Cloud conditions Corr. coeff. Std. dev. Std. dev. Mean rms analysis SIRS error* Regression Slope Scatter 200 mb All cases Clear Lower Higher 0.82 .97 + .85 .97 .79 .96 .83 .97 2.1*3 C 1.25 1.91* 1.06 2.60 1.27 2.1*1 1.26 1*.13° i*.13 3-30 3-30 U.06 1*.06 1*.20 1*.20 2.91° U.55 2.65 3.89 2.68 i*.i*2 3.02 l*.6l -0.20° .29 - .86 .21 - -56 .1*6 .03 .23 1.16 0.88 1.06 O.83 1.20 0.89 1.15 0.88 2.38 c 1.08 1.73 0.79 2.1*8 1.07 2.37 1.10 300 mb All cases Clear Lower Higher • 50 .93 ,66 .95 .1*6 >9k • 53 ■ 92 2.05 0.99 2.03 0.89 2.2l* 0.92 1.97 1.03 2.18° 2.18 2.32 2.32 2.19 2.19 2.15 2.15 1.79° 2.61 1.57 2.71 1.65 2.63 1.85 2.59 .1*1° .11* 1.03 0.21 .88 .12 .11* • lit 0.61 .78 .97 .82 .60 .79 .61 .77 1.88° 0.79 1.75 0.71 1.95 0.72 1.83 0.82 500 mb All cases Clear Lower Higher .63 .78 .87 .91* • 52 .77 .61* .78 1.91° i.ia 1.00 1.05 1.69 1.38 2.06 1.1*6 i.l*5 e 1.1*5 1.15 1.15 1.31 1.31 1.52 1.52 2.23° 2.11* 1.77 2.03 1.67 1.81* 2.39 2.23 *Upper number refers to operational retrievals. tLower number refers to final retrievals. .82° ,1*0 .28 .17 .79 .71 .96 .33 .1*1 • 53 .56 .53 .1*1 .1*1 .53 1.12° 0.90 .57 .1*0 1.12 0.81* 1.16 0.96 tMean error = M 8 M ( See text . ) Sample 1 21 E o o ►— CO < u LU o CO co >- < z < E o o CO co < CO CO — I < z < E o < u LU o CO >- —J < z < 1 1 12 lift 11? .3* — .b *"-2." -.3* " '1.7 3." 5 *«• I* » I 1 1 I* 2 I • I .1 12*11 ! i * ii 30" 1 -J.*0 - 1 .97* * * * -.S- .89 ?.3I J SIRS - FORECAST CLEAR SIRS • FORECAST LOWER CLOUD SIRS - FORECAST HIGHER CLOUD 22 Sample 2 E o o to < u UJ OS o < z < E o o m t/J < u (JO < Z < E o o to < U LU a: o >- < Z SIRS FORECAST CLEAR SIRS FORECAST LOWER CLOUD SIRS - FORECAST HIGHER CLOUD Sample 3 23 11 11 E o o < u LU ex. o > I < Z < SIRS - FORECAST CLEAR SIRS - FORECAST LOWER CLOUD SIRS - FORECAST HIGHER CLOUD 24 Sample 4 21 I I i 50 -2.72 -.«S SIRS - FORECAST CLEAR SIRS • FORECAST LOWER CLOUD SIRS • FORECAST HIGHER CLOUD Sample 5 25 1 2 • 12 i 2111 mi ) II 211 1 11 2* 1 It SIRS - FORECAST CLEAR SIRS - FORECAST LOWER CLOUD SIRS - FORECAST HIGHER CLOUD 26 Sample 6 -D E O o CN < u LJJ o >- < Z < 3.". 5.7 i- 21 1 1 1 1 111 1 1 • 1221 1 • 1 I 12 |1*| 11 I* 12 11 112*1 12111 II I 231 I 131 Ml 11 11 111 11121 21 1 12.2 • 2 1 1 -7 2 —.0 8 2.S b.7 3. E o o ,.,,: CO I/) < U ex. o < z < -O l.< E o o IS) < U O CO >- — 1 < Z < SIRS • FORECAST CLEAR SIRS - FORECAST LOWER CLOUD SIRS - FORECAST HIGHER CLOUD (Continued from inside front cover) NESC 51 Application of Meteorological Satellite Data in Analysis and Forecasting. Ralph K. Anderson, Jerome P. Ashman, Fred Bittner, Golden R. Farr, Edward W. Ferguson, Vincent J. Oliver, and Arthur H. Smith, September 1969. Price $1.75 (AD-697-033) 'Supplement price $0.65 (AD-740- 017) NESC 52 Data Reduction Processes for Spinning Flat-Plate Satellite-Borne Radiometers. Torrence H. MacDonald, July 1970. Price $0.50 (COM-71-00132) NESC 53 Archiving and Climatological Applications of Meteorological Satellite Data. John A. Leese, Arthur L. Booth, and Frederick A. Godshall, July 1970. Price $1.25 (COM-71-00076) NESC 54 Estimating Cloud Amount and Height From Satellite Infrared Radiation Data. P. Krishna Rao, July 1970. Price $0.25 (PB-194-685) NESC 56 Time-Longitude Sections of Tropical Cloudiness (December 1966-November 1967). J. M. Wallace, July 1970. Price $0.50 (COM-71-00131) NOAA Technical Reports NESS 55 The Use of Satellite-Observed Cloud Patterns in Northern Hemisphere 500-mb Numerical Analysis. Roland E. Nagle and Christopher M. Hayden, April 1971. Price $0.55 NESS 57 Table of Scattering Function of Infrared Radiation for Water Clouds. Giichi Yamamoto, Masayuki Tanaka, and Shoji Asano, April 1971. Price $1.00 (COM-71-50312) NESS 58 The Airborne ITPR Brassboard Experiment. W. L. Smith, D. T. Hilleary, E. C. Baldwin, W. Jacob, H. Jacobowitz, G. Nelson, S. Soules, and D. Q. Wark, March 1972. Price $1.25 (COM-72-10557) NESS 59 Temperature Sounding From Satellites. S. Fritz, D. Q. Wark, H. E. Fleming, W. L. Smith, H. Jacobowitz, D. T. Hilleary, and J. C. Alishouse, July 1972. Price $0.55 (COM-72-50963) NESS 60 Satellite Measurements of Aerosol Backscattered Radiation From the Nimbus F Earth Radiation Ex- periment. H. Jacobowitz, W. L. Smith, and A. J. Drummond, August 1972. Price $0.25 (C0M-72- 51031) NESS 61 The Measurement of Atmospheric Transmittance From Sun and Sky With an Infrared Vertical Sounder. W. L. Smith and H. B. Howell, September 1972. NESS 62 Proposed Calibration Target for the Visible Channel of a Satellite Radiometer. K. L. Coulson and H.~ Jacobowitz, October 1972. Price $0.35 PENN STATE UNIVERSITY LIBRARIES AQDDD7aDlfl33D