Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Satellite derived sea surface temperature : a physical approach Shin, Hae-Yong 1986

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-UBC_1986_A1 S44.pdf [ 9.67MB ]
JSON: 831-1.0053227.json
JSON-LD: 831-1.0053227-ld.json
RDF/XML (Pretty): 831-1.0053227-rdf.xml
RDF/JSON: 831-1.0053227-rdf.json
Turtle: 831-1.0053227-turtle.txt
N-Triples: 831-1.0053227-rdf-ntriples.txt
Original Record: 831-1.0053227-source.json
Full Text

Full Text

SATELLITE DERIVED SEASURFACE TEMPERATURE: A PHYSICAL  APPROACH  By H A E - Y O N G SHIN B.A.Sc, University of Waterloo, 1980 A THESIS SUBMITTED IN PARTIAL FULFILLMENT O F T H E REQUIREMENTS FOR T H E D E G R E E O F DOCTOR O F PHILOSOPHY in T H E FACULTY OF G R A D U A T E STUDIES DEPARTMENT OF OCEANOGRAPHY  We accept this thesis as conforming to the required standard  T H E UNIVERSITY O F BRITISH COLUMBIA August 1986 ©Hae-Yong Shin, 1986  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make i t freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission.  Department of Oceanography The University of British Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 Date  Sept. 22. 1986  S u p e r v i s o r : Dr. W i l l i a m J . Emery y  A B S T R A C T  In this study, a method of estimating atmospheric attenuation in the infrared remote sensing of sea surface temperature (SST) from National Oceanic and Atmospheric A d ministration ( N O A A ) meteorological satellites, by explicitly considering the physical processes involved, is developed. Unlike past methods which tend to rely on the inputs from the in situ data, this method which is entirely satellite derived, will be useful in the open oceans where in situ observations are sparse.  Using Advanced Very High Resolu-  tion Radiometer ( A V H R R ) , with atmospheric temperature and water vapor profiles from the TIROS Operational Vertical Sounder (TOVS), SST's are retrieved. T O V S sensors are carried by the same satellite as the A V H R R and provide simultaneous corrections for the A V H R R based SST estimates. SST, computed from sensor systems on NOAA-7, are compared with surface skin temperatures (from a P R T - 5 infrared radiometer mounted on a ship) and subsurface temperature measurements.  In addition to the physical method  ( A V H R R + T O V S ) , two additional split-window techniques (SWT) using channels 4 and 5 of the NOAA-7 A V H R R sensor, were investigated. These methods were: 1. Using A V H R R alone,and 2. Using A V H R R and data from the High-resolution Infrared Sounder (HIRS). The importance of scan angle correction and emittance effect to define the correct atmospheric path is discussed. The improvement in SST retrievals using sensor combinations is demonstrated with satellite versus ship skin temperature differences ranging from 0.63 to 0.89 ° C for A V H R R alone, from 0.49 to 0.80 °C for A V H R R + T O V S , and from 0.39 to 0.56 ° C for A V H R R + H I R S . The improved temperature accuracy with A V H R R + H I R S is due to the atmospheric water vapor correction possible with some of the HIRS channels. The effects of errors in T O V S derived atmospheric profiles upon the SST are significant, but using the S W T ,  these effects are reduced, improving the over all SST retrievals. Even  though T O V S profiles cannot duplicate radiosonde observations in accuracy or in vertical resolution, their overall effect on the SST is found to be within the physical limitations of the satellite sensors.  It is suggested that the physical method can effectively be used  to generate coefficients, applicable to a given location, for a S W T method such as the  AVHRR+HIRS, in order to achieve the required accuracy at a reduced computation time. The coefficients can then be updated for seasonal variations. Significant differences between ship skin and subsurface temperatures were observed with the mean deviation of -0.86 °C in the Pacific and —0.2°C in the Atlantic for a range of root-mean-square (RMS) temperature differences between 0.16 and 0.87 °C.  iii  Table of Contents 1.  INTRODUCTION 1. Sea Surface Temperature from Meteorological Satellites 2. Atmospheric Profiles from Meteorological Satellites 3. Objectives 2. B A C K G R O U N D T H E O R Y A N D S S T R E T R I E V A L T E C H N I Q U E S 1. Blackbody Radiation 2. Atmospheric Interaction In the Infrared 1. Atmosphere . 2. Absorption and Emission 3. Molecular Absorption In The IR Window Region 3. Radiative Transfer Equation (RTE) 4. SST Retrieval Techniques 1. Single Channel Technique 2. Split Window Technique 3. SWT With Scan Angle Correction 4. Extension of the AVHRR SWT Using HIRS 5. AVHRR SWT with T O V S Derived Atmospheric Profiles 1. Verticai Pro&les Retrieval 1. Retrieval Theory 2. Improved Iterative Solution 2. Calculation of TOVS Transmittances 1. HIRS/2 2. AVHRR 3. D E S C R I P T I O N A N D P R E P R O C E S S I N G O P D A T A 1. Satellite Sensors 1. Advanced Very High Resolution Radiometer (AVHRR) 2. TIROS Operational Vertical Sounder (TOVS) 1. High-resolution Infrared Radiation Sounder/2 (HIRS/2) 2. Microwave Sounder Unit (MSU) 2. In Situ Measurements 1. North Atlantic Ocean 2. North Pacific Measurements 3. Barnes PRT-5 Radiometer 3. Sensor Calibration 1. Satellite Sensor 2. Ship-borne Radiometer 4. Registration of Satellite and Ship Data 5. Error Sources 1. .Radiometer 2. Natural Sources 4. R E S U L T S O P T H E C A S E S T U D I E S 1. Comparison of Vertical Profiles 1. Radiance Synthesis 2. Retrieval Procedure 3. Pro&les Comparison 4. Sensitivity Test 2. Sea Surface Temperature Retrieval iv  1 3 6 8 11 11 12 12 13 16 18 19 19 20 22 25 28 28 29 31 33 34 36 37 37 38 39 40 42 43 43 44 46 47 48 50 50 52 55 56 59 59 60 66 70 87 91  1. Scan Angle Effect 2. Emittznce Effect 3. Comparison with Surface Skin Temperature 4. Skin vs Bulk Temperature Comparison 5. Conclusions and Recommendations BIBLIOGRAPHY Appendix A : Definition of Statistical Parameters Appendix B : Definition of Meteorological Parameters  v  91 92 93 107 114 121 133 134  List of Tables 1.1. Sea surface temperature measurements 1.2. Estimated future SST capabilities 2.1. Regression coefficients for equation (2.20) 2.2. Regression coefficients for equation (2.22) 3.1. Orbital parameters for N O A A - 7 3.2. N O A A - 7 AVHRR characteristics 3.3. Characteristics of HIRS/2 channels 3.4. Characteristics of MSU channels 3.5. HIRS/2 system parameters 3.6. MSU instrument parameters 3.7. Barnes PRT-5 performance specification 3.8. Thresholds used to detect cloud contamination 4.1. Information on RAOB co-located T O V S spots 4.2. Information on radiosonde data 4.3. Synthesized and observed HIRS brightness temperature differences for Pacific 4.4. Synthesized and observed MSU brightness temperatures differences for Pacific 4.5. Synthesized and observed HIRS brightness temperature differences for Atlantic 4.6. Synthesized and observed MSU brightness temperature differences for Atlantic 4.7. Differences in temperature profiles for Pacific 4.8. Differences in dew point profiles for Pacific 4.9. Differences in temperature profiles for Atlantic 4.10J)ifferences in dew point profiles for Atlantic 4.11.Vertically averaged RMS temperature differences for 15779 4.12.Vertically averaged RMS dew point differences for 15779 4.13.Vertically averaged RMS temperature differences for 17681 4.14.Vertically averaged RMS dew point differences for 17681 4.15. Emittance effect on SST 4.16. Results of satellite derived versus PRT-5 measured SST 4.17.TOVS retrieved AVHRR channel 4 parameters 4.18.TOVS retrieved AVHRR channel 5 parameters 4.19.Comparison of physical SWT based on T O V S and RAOB 4.20.Results of satellite derived SST versus bulk temperature 4.21.Comparison between skin by PRT-5 and bulk sea temperature  vi  3 3 23 26 37 39 40 41 42 42 47 56 61 61 . 63 . 64 . 65 . 65 76 77 78 78 80 80 81 81 93 . 94 95 96 106 109 114  List of Figures 1.1. Outgoing IR radiance spectrum near 15 pm 2.1. Vertical temperature profile for the U.S. Standard Atmosphere 2.2. The annual mean global energy balance 2.3. Infrared absorptions by atmospheric gases 2.4. Three normal modes of vibration 2.5. Differences between simulated and observed SST vs observed SST using eqn. 2.6. Differences between simulated and observed SST vs observed SST using eqn. 3.1. HIRS and MSU ground coverage 3.2. Map of the Atlantic study area 3.3. Map of the Pacific study area 3.4. Functional block diagram of the Barnes PRT-5 radiometer 3.5. Three sources of the radiance received by the PRT-5 3.6. Calibration curve for the PRT-5 used in the Pacific 3.7. AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 15770 3.8. AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 15835 3.9. AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 17667 3.10.AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 17681 3.11 .Schematic of physical processes influencing the skin temperature 4.1. Scatter plots of brightness temperatures vs. differences (a) 4.2. Scatter plots of brightness temperatures vs. differences (b) 4.3. The flow diagram of the TOVS profile retrieval procedure 4.4. Temperature and dew point profiles for N O A A - 7 15779 4.5. HIRS weighting functions for 15779 4.6. Temperature and dew point profiles for N O A A - 7 15835 4.7. HIRS weighting functions for 15835 4.8. Temperature and dew point profiles for N O A A - 7 17667 4.9. HIRS weighting functions for 17667 4.10.Temperature and dew point profiles for N O A A - 7 17680 4.11.HIRS weighting functions for 17680 4.12.Temperature and dew point profiles for N O A A - 7 17681 4.13.HIRS weighting functions for 17681 4.14.Temperature differences from the two TOVS retrieval methods (Pacific) . 4.15.Dew point differences from the two TOVS retrieval methods (Pacific) 4.16.Temperature differences from the two TOVS retrieval methods (Atlantic) 4.17.Dew point differences from the two TOVS retrieval methods (Atlantic) . 4.18.Comparison of precipitable water vapor amounts in the Pacific 4.19.Comparison of precipitable water vapor amounts in the Atlantic 4.20.Plot of precipitable water vapor vs. channel 4 transmittance 4.21.Plot of precipitable water vapor vs. channel 5 transmittance 4.22.The effect of water vapor amounts in the atmosphere, on SST 4.23.Sensitivity test on channel 4 transmittance 4.24-Sensitivity test on channel 5 transmittance 4.25.The effect of profile errors on 7 4.26.The effect of profile errors on SST 4.27.The effect of scan angles on SST 4.28.SST from the three atmospheric correction methods for orbit 15779 4.29.Temperature scatter plots from the three methods for 15779 vii  8 14 14 16 18 (2.20) 24 (2.22) 27 41 45 45 46 48 51 . . . 53 . . . 53 . . . 54 . . . 54 58 67 68 69 71 71 72 72 73 73 .74 74 75 75 . . . 83 83 . . . 84 . . . 84 85 85 86 86 87 89 89 90 90 92 97 97  4.30.Binned histogram from the three methods for 15779 4.31 .SST from the three atmospheric correction methods for orbit 15835 4.32.Temperature scatter plots from the the three methods for 15835 . . . . . . 4.33.Binned histogram from the three methods for 15835 4.34.SST from the three atmospheric correction methods for orbit 17680 4.35.Temperature scatter plots from the three correction methods 4.36.Binned histogram from the three methods for orbit 17680 4.37.SST from the three atmospheric correction methods for orbit 17681 4.38.Temperature scatter plots from the three correction methods 4.39.Binned histogram from the three methods for orbit 17681 4.40.Temperature and dew point profiles from two orbits 4.41 .SST differences measured by two consecutive satellite passes 4.42.Bulk and PRT-5 surface temperature plots for Jul. 14 4.43. Binned histogram for Jul. 14 4.44.Bulk and PRT-5 surface temperature plots for Jul. 18 4.45. Binned histogram for Jul. 18 4.46.Bulk and PRT-5 surface temperature plots for Nov. 25 4.47J3inned histogram for Nov. 25 4.48.Bulk and PRT-5 surface temperature plots for Nov. 26 4.49.Binned histogram for Nov. 26  viii  98 100 100 101 102 102 103 104 104 105 108 108 110 110 Ill Ill 112 112 113 113  ACKNOWLEDGEMENTS I would like to acknowledge the assistance of the crews of the MV Pandora II and Andy Thomas, in the collection of the in situ data, the Institute of Ocean Science for providing the PRT-5 radiometer, and the Department of Oceanography at U B C for providing the necessary hardware. This study was supported by the Canadian Natural Science and Engineering Research Council and the British Columbia Science Council. This support is gratefully acknowledged. I am also grateful to H. Woolf of NOAA/NESDIS, Cooperative Institute for Meteological Satellite Studies at the University of Wisconsin for providing the TOVS Export package, and Michael Hill of NOAA for supplying the AVHRR transmittance software. The cooperation of the Canadian Atmospheric Environment Service is also appreciated for supplying the RAOB data for the Northeast Pacific region. I am indebted to Peter Schluessel for many helpful discussions and for supplying me with data collected in the Atlantic. I would also like to thank the members of my supervisory committee for their support and special thanks to Dr. Bill Emery, my supervisor, who first introduced me to the problem of satellite derived SST and continuously encouraged me during the course of this study. Finally, I would like to acknowledge the assistance of Paris Vachon and the MacDonald, Dettwiler and Associates Ltd. and thank my family and Susan R. who were there to help at difficult times.  ix  CHAPTER 1 INTRODUCTION  Of the numerous applications of polar orbiting meteorological satellites, their capability to remotely measure sea surface temperature (SST) has perhaps received the widest attention and appreciation. SST is of importance to synoptic meteorologists and climatologists for its role in the climate system, and is of importance to oceanographers for its relationship to ocean dynamics. It is believed, among climatologists, that the ocean plays a significant role in driving global weather systems. A prominent example of this is the generation of the annual Monsoon in the Southeast Asia due primarily to seasonal differences between land and sea surface temperatures. In North America, evidence has linked extreme changes in the weather to El Nino events in the tropical Pacific, which is characterized by the anomalous increases in the SST (Emery and Hamilton, 1985). At this point it is not clear whether the anomalous SST increases cause the El Nino wind changes, or the wind changes produce the SST warming. On the practical side, highest marine biological productivities are usually found near frontal zones which are often characterized as strong temperature gradients. These areas can be efficiently identified from the SST maps derived from satellite-borne sensors. The advantage of satellite derived SST is apparent if one realizes that 71% of the surface of the earth is covered with ocean. Traditionally, global SST has been mapped by gathering data from all available sources, such as commercial vessels, buoys, research ships, using such methods as bucket sampling, engine intake monitoring, Expendable Bathythermographs (XBT), etc. Due to its vastness, it is virtually impossible to keep current SST maps over global water bodies when only these conventional methods are employed. Also, the problem of combining data from several different sources with highly variable accuracies, is not a trivial matter. In addition, complete global coverage has never been successful as the locations of data acquisition tend to heavily follow well known commercial routes or oceanographically interesting regions. On the other hand, satellites, not being restricted by 1  the geographical problems, can provide complete global coverage with spatial and temporal resolution that no other method can match. The use of operational satellites to retrieve SST, began with NASA's NIMBUS experimental satellites which carried both infrared and microwave sensors. From the technology obtained through the NIMBUS series satellites, TIROS operational satellites were introduced, first with visible sensors only, then subsequently with both infrared, microwave sensors, and an atmospheric sounder called the Vertical Temperature Profile Radiometer (VTPR). In 1978, a new generation of polar orbiting satellite (TIROS-N) became operational with more advanced sensors. The first TIROS-N satellite, later renamed NOAA-6, carried the Advanced Very High Resolution Radiometer (AVHRR) and the TIROS Operational Vertical Sounder (TOVS). The TOVS system consists of three types of sensors; High-resolution Infrared Radiation Sounder (HIRS), Microwave Sounding Unit (MSU), and Stratospheric Sounding Unit (SSU). Starting with NOAA-6, both the data acquisition and the transmission are done digitally thus retaining the wider dynamic range of the sensors and increasing the reliability of the data transmission. With the success of NOAA-6, a spur of research testing the capabilities of meteorological satellite began world wide. Since then, a number of application areas have been identified and a continuous effort to discover new areas is still very active. NOAA-7 from which data for this research was obtained from, was launched on July 1981. Even though the sensors have degraded significantly, it still is being used as a backup satellite to NOAA-9. Considering the fact that the expected lifetime of TIROS-N series satellites are about 2 years, this is remarkable, but not unusual. Even NOAA-6 continues to be used for image data at least in some of its AVHRR channels.  2  1.1 Sea Surface Temperature from Meteorological Satellites Since the first launch of a meteorological satellite with infrared sensors about 15 years ago, a number of researchers have studied the problem of deriving accurate SST from the data generated by these sensors. A list of SST measurements achieved in the past with various satellites is found in Table 1.1 (Darnell and Harriss, 1983). Given the user requirement of 0.5 ° C (Robinson et al., 1984), the satellite SST retrievals have not been widely accepted as the operational products. The estimated theoretical SST accuracies achievable in the future are listed in Table 1.2 (Darnell and Harriss, 1983).  Table 1.1  Sea surface temperature measurements Satellite Sensor Meandiff ( ° C ) GOES 2,3,4,5 SMS 2 Nimbus 6 NOAA5 DMSP Block 5C Seasat  VISSR VISSR HIRS VHRR VTPR SR SMMR  Surf, resol.(fcm)  •3 to -5 -2 to -4 1.6 •1.8 to -3.6 -0.3 to 0.31 1 -3 to -5  7.5 7.5 23.8-36.3 0.9 55.6 0.5-2.8 16-144  Table 1.2  Estimated future SST capabilities Sensor Potential accuracies ( ° C ) Satellite TIROS-N GOES ERS-1  MSU HIRS AVHRR VISSR ASTR  2 2 0.5-1.0 1.5 0.5  One of the early pioneers in satellite derived SST, Smith et a/.(1970), developed a statistical histogram SST retrieval method and found that using 3.8/xm data from Nimbus 2 and Nimbus 3, the difference between the estimated SST and the ship measurement was less than 1°C. Rao et a/.(1972) produced the first satellite global SST map in 2.5° lat/long grids. They found that the satellite SST root mean square (rms) differences, from ship SST data, were 2 - 3°C. 3  Given accurate atmospheric profiles of temperature and gases, SST can be derived accurately using sufficiently accurate transmission models already available. Shaw et a/.(1969) achieved an accuracy of less than 0.2° C to correct for the atmospheric contribution to a balloon-borne radiometer. Shaw used concurrent averaged radiosonde and rocketsonde data for the temperature and the moisture profiles. Maul and Sidran (1973) performed theoretical radiation calculations with a standard atmosphere and concluded that the atmospheric contribution can be sufficiently corrected up to 2°C. Maul and Sidran investigated the effects of various atmospheric conditions, sensor viewing angle, cloud amount, cloud height, and instrument noise. At nadir, the temperature difference with  in situ  ship  data was 2.2° C in a dry atmosphere and was as much as 10.5° C in a moist subtropical atmosphere. At the worst viewing angle of 60° these differences were up 7°C due to increased path length. The effects of cloud were unpredictable and Maul and Sidran suggested that only the clear field-of-view (FOV) should be considered. Anding and Kauth (1972), Prabhakara et o/.(1974), Deschamps and Phulpin (1979), and McMillin et a/.(1974, 1980) employed differential absorption properties at two infrared window channels to improve the SST retrieval algorithm. Prabhakara et a/.used a physical approach considering e type, p type, and line absorption in the water vapor spectra with a model atmosphere to derive mean transmission over a temperature range. Using a transmittance model developed by Weinreb and Hill (1980), Callison and Cracknell (1984) estimated SST over water surrounding Britain using radiosonde data and reported a high correlation between the satellite derived and  in situ  SST measurements.  As a fully operational product, NOAA initiated the Global Operational Sea-Surface Temperature Computational Program (GOSSTCOMP) in 1972 to produce time and space averaged global SST maps for users. GOSSTCOMP used V T P R to obtain atmospheric information for the correction of SST measured by the scanning radiometer (SR) (Brower  et a/., 1975). Compared to ship data, GOSSTCOMP SST maps achieved an accuracy of 1.5°C. Beginning November of 1981, GOSSTCOMP was replaced by the new MultiChannel Sea Surface Temperature (MCSST) algorithm (McClain 1980, McClain et al.  t  4  1982). This widely recognized method of deriving SST maps is currently being used operationally by NO A A (McClain et ai., 1985; Bernstein and Chelton, 1985). As further described in McClain et al. (1983) and compared with other approaches in McClain et al. (1985) the MCSST method is capable of producing surface temperatures with a bias of 0.3 ° C and RMS deviation of about 0.5 °C. This correction procedure employs the radiance differences in the two infrared channels of the AVHRR/2 system to remove the atmospheric water vapor contamination. The coefficients applied to the measured radiances in the two infrared channels are derived by the regression of computed sea surface temperatures on in situ measurements from ships and ocean buoys. Verification of resulting accuracies are then made by comparison with an independent set of in situ surface temperature measurements. However, the lack of physical justification for the MCSST technique has been a concern among users.  In the three SST workshops held between 1982 and 1984, sponsored by the Jet Propulsion Lab, it was found that the global, spatially averaged RMS differences of the satellite derived SST lied between 0.5 -1.0 ° C when compared to data from ships, XBTs, and buoys. The researchers agreed that the SST retrieval method should be based on an understanding of the physical processes in the atmosphere and the surface of the ocean, and the effort should be put in to reduce the measurement biases and improve the rms accuracies at any given location to the « 0.5 ° C level (Njoku, 1984).  5  1.2 Atmospheric Profiles from Meteorological Satellites Traditionally, meteorological observations of the atmospheric vertical structure have been measured using balloon-born (radiosonde) or rocket-borne (rocketsonde) instruments, or by using aircraft. Radiosonde observations (RAOB) are widely used because they are more cost effective. Instruments capable of measuring temperature, humidity, pressure, and other meteorological parameters, are carried upward by a buoyant balloon. As the radiosonde ascends, the on-board transmitter sends the measured information down to a ground station. Often the balloon is tracked to measure the horizontal component of upper wind. In most upper-air stations, radiosonde launches are made at 00:00 and 12:00 GMT,  the principal synoptic hours. The ascension rate is about 300 m/min  and therefore  it takes about one hour for the balloon and the radiosonde to ascend from sea-level to above 100 mb (« 16 km).  Due to extreme cold environment, measurements above 100m6,  typically are not made using radiosondes.  With radiosondes, measurements are made directly by contact with the atmosphere, and therefore the measurements are as accurate as the sensor systems allow. Because of their continuous measurements, radiosondes provide a very fine vertical resolution. The problem of using radiosondes (or rocketsonde) is the requirement of having a dense network of observing stations in order to achieve the desired horizontal resolution of the atmosphere. Over land areas, the World Meteorological Organization (WMO) recommends that synoptic observation network distance be less than 150 km between stations. It is obvious to see why such a recommendation is not feasible over oceanic areas. Keeping in mind that major weather systems originate over the ocean, developing a method of observing such areas with adequate spatial, vertical and temporal resolution seems vital.  The idea of obtaining vertical atmospheric profiles from remote platforms was first conceived by King (1958) and further developed by Kaplan (1959). In his paper Kaplan suggested that the vertical temperature profile can be inferred from the spectral distribution of atmospheric emission as atmospheric constituents selectively absorb and emit 6  outgoing radiation. In order to make temperature measurements, the source of such radiation should be from the atmospheric gases which should be abundant to be globally valid and uniformly mixed to reduce the complexity involving the amount of gas at different atmospheric pressure levels. In addition, the gas must have emission bands which are feasible to measure. In the earth's atmosphere, there are two such gases, carbon dioxide ( C 0 ) and oxygen. C O 3 is relatively abundant and has infrared vibrational-rotational 3  radiation bands near 4.3 pm (shortwave) and near 15 pm (longwave). Oxygen is even more abundant and has a spin rotational radiation band at 0.5 cm (microwave).  The water  vapor has absorption bands at 6.3 pm and 18 pm in the infrared and at 0.8 cm and 1.35 cm in the microwave region. Figure 1.1 is an outgoing IR radiance spectrum at 15 pm as measured by the Infrared Interferometer and Spectrometer (IRIS) on the NIMBUS IV satellite. If one examines the spectral band, the radiation received around the band center must have originated from the upper layers of the atmosphere, since the absorption is large near the band center.  Furthermore, radiation around the wings of the band must have originated from  near the surface as higher radiance suggests weaker absorption through the column of the atmosphere. The cold temperature at the band center (690cm ) represents the mini-1  mum temperature at the tropopause and the increase in temperature with the increase in wavenumber reveals warmer temperature towards the surface. Thus, by observing the atmosphere at selected wavenumbers, derivation of atmospheric profiles in the troposphere and in the lower stratosphere is possible.  7  500  600  7 0 0  800  WAVE NUMBER (cm") 1  Figure 1.1 Outgoing IR radiance spectrum near 15 pm. Observed by the IRIS on NIMBUS IV. Arrows represent V T P R channels, (from Liou, 1980) 1.3 Objectives The prime objective of this study is to accurately determine the sea surface temperature from a N O A A polar orbiting satellite using a physical model of the atmosphere without any inputs from concurrent in situ meteorological measurements. In other words, a totally independent satellite SST retrieval method based on the mathematical description of the interaction between the atmospheric constituents and the electromagnetic (EM) radiation, will be developed. This is especially important when the field of view lies over the open ocean where it is difficult to conduct in situ measurements on a regular basis. There are three reasons why a totally independent satellite SST retrieval algorithm should be pursued: 1) A statistically optimized solution tends to distort the information if incorrect indepen8  dent statistics are incorporated into a solution.(Weinreb et al., 1975) 2) The atmospheric constituents vary significantly in space and time, thus casting uncertainties on the use of past and global statistics in the solution for present, localized measurements. 3) It is difficult to estimate how much new information has been distorted to fit the empirical functions derived from the existing data either oceanographic or atmospheric (Thompson, 1981). The entirely physical method of deriving SST, developed in this study, relies on the Split Window Technique (SWT) as the basis of obtaining SST from the AVHRR with actual transmittances calculated from atmospheric profiles obtained from the T O V S . However, the performance of the T O V S in producing atmospheric profiles has not been very encouraging when compared with radiosonde measurements (Hayden et al., 1979). It seems that there are inherent fundamental limitations in the TOVS system. This brings us to the second objective. It is to compare the physical method against other SST retrieval methods in order to identify problem areas and to improve the current SST method. As mentioned earlier, with accurate descriptions of the atmospheric radiation path, one can derive SST with the highest accuracy possible by the available sensors. Thus, the accuracy of the physical approach depends on the performance of the T O V S atmospheric profiler, provided that the transmission model is accurate enough. The third study objective is to compare the satellite derived SST against in situ skin temperature rather than the usual bulk temperature which most researchers have used. The satellite radiometer only sees the top micrometer of the ocean surface and in order to properly validate the results one should compare two sets of data from similar classes. The relationship between the surface skin temperature and the bulk temperature is an extremely complicated one with many variables. Thus the validation of a SST retrieval algorithm based on the bulk temperature is subject to definite limitations. In chapter 2, basic infrared remote sensing theory is presented. Then the developments d  of several split window techniques and a T O V S atmospheric profiles retrieval method, are presented. In chapter 3, the instruments used and the preprocessing of data, both satellite and in situ, are presented. Possible error sources are defined and their estimated contributions are also presented in this chapter. In the final chapter, results from the experiments for both the T O V S profile retrievals and the SST retrievals, are presented and findings discussed.  10  CHAPTER 2 B A C K G R O U N D T H E O R Y A N D SST RETRIEVAL  TECHNIQUES  In this chapter, we examine how sensors located in the outer space can see through the atmosphere and determine the state of the atmosphere as well as the surface parameters. The interaction between the atmosphere and the E M radiation in the infrared region is examined and techniques used to retrieve atmospheric profiles and SST, including the physical method developed in this study, are presented.  2.1 Blackbody Radiation All terrestrial objects absorb and emit electromagnetic radiation. The emitted radiation at frequency v{Hz) by a blackbody (a perfect emitter) at temperature T{ °K) is described by Planck's equation  BAT)  =  2hv* 1 c exv[hu/kT]  (2.1)  -1  3  where h is Planck's constant (6.6262 x 10~ erg sec), c is the speed of light (3 x 10* m/s), 37  and k is Boltzmann's constant (1.3806 x 1O~~ ergdeg~ ). 10  l  If one integrates Planck's func-  tion over the entire electromagnetic spectrum to obtain flux density emitted by a black body, we get Stefan-Boltzmann Law.  (2.2) = aT  4  where a Stefan-Boltzmann's constant, is equal to 5.67 x 10 %  8  ergcm~*sec~ deg~ . 1  4  If one measures emitted radiation, B„(T), the temperature of the material can be determined by inverting Planck's equation. However, most objects found on Earth are not blackbodies, but are known as greybodies. The ratio of the spectral brightness emitted by a greybody to that emitted by a blackbody is defined as the spectral emissivity in the direction {0, <f>)  e {QA) v  =  B {9 <f>)/B(u) u  11  t  Maekbod9  .  (2.3)  Thus, the radiation measured by a radiometer can be quantified as £„(0, <j>) B^T) and knowing the emissivity, the temperature of the radiation source can be determined.  2.2 Atmospheric Interaction In the Infrared 2.2.1  Atmosphere  The earth's atmosphere is composed of several types of gases. More abundant atmospheric gases are Nitrogen (75.51% by mass), Oxygen (23.14% by mass), Argon (1.28% by mass), water vapor (0-0.04%), and Carbon dioxide (325 parts per million). Due to the earth's gravitational field, atmospheric gases experience downward forces exerted toward the surface of the earth. This results in more then 80% of the atmospheric mass concentrated below the tropopause which exists around 10 km in altitude. The force per unit area due to the weight of the atmosphere is expressed as a pressure. The mean atmospheric pressure averaged over the earth's surface is approximated by MAg /4irRj. 0  the total mass of the atmosphere (5.14 x 10 (9.8 m/sec), and RE  18  t  where MA is  kg), g is the mean gravitational acceleration 0  is the mean radius of the earth (6.37 x 10 m). From these values, s  one can estimate the surface pressure of the earth to be approximately 1000 mb.  The vertical temperature profile of the "U.S. Standard Atmosphere" is shown in Figure 2.1. This profile is typical of conditions found in middle latitudes. As shown in the figure, the atmosphere can be divided into four distinct layers according to its temperature characteristics:  troposphere,  stratosphere,  of these layers are known as the  tropopause,  mesosphere, stratopause,  and  thermosphere.  mesopause,  and  The boundaries thermopause,  re-  spectively. As mentioned earlier, most of the atmospheric gases reside in the troposphere and thus most of the atmospheric interactions with the E M radiation take place in this layer. The troposphere is characterized by strong vertical mixing and therefore the concentration of atmospheric gases is highly variable. The stratosphere is characterized by very small vertical mixing and most of the atmospheric ozone is concentrated in this layer. The troposphere and the stratosphere account for about 99.9% of the atmospheric mass. 12  The annual mean global energy balance for the earth atmosphere is shown in Figure 2.2. This figure shows that 51% of the incident solar radiation are absorbed by the earth's surface. The earth's surface in return disposes of this energy by a combination of infrared radiation and sensible and latent heat flux. Of the upward infrared emission from the surface, 71% (15% of the averaged solar irradiance) is absorbed in passing through the atmosphere and the remaining 29% (6% of the averaged solar irradiance) reach space. An infrared sensor positioned in space receives this 6% plus 38% net emission by H3O, C 0  3  and 26% emission by clouds. It seems highly unlikely that the surface temperature can be determined through the earth's atmosphere, but the atmosphere is remarkably transparent to infrared radiation at certain frequencies. These are known as the "atmospheric windows" and they will further be discussed later.  2.2.2 Absorption  and Emission  All atmospheric gases on earth interact with E M radiation. Since none of the atmospheric gases behave as blackbodies, the interaction process involves absorption, emission, and transmission which complicates the understanding of these processes. Atmospheric gases are also highly variable in space and in time; one cannot simply use statistically obtained values. The goal of remote sensing is to measure upwelling atmospheric plus surface radiation and infer atmospheric and surface parameters at any pressure level, from surface to any significant height, from a distance. The processes of absorption and emission result from electrons changing quantum state, say from n, to another, m. For absorption, m > n while for emission, m < n. The radiation frequency is the oscillation frequency of a charge distribution. This was postulated by Bohr in 1913 and proved later by mechanical wave theory. The frequency of radiation is, u  where E  m  and E  n  =  E  m  -  E  h  "  (2.4)  are the total energies at levels m and n respectively. The photons  absorbed or emitted have frequencies such that the product "hi/* equals the difference in energy between two levels. For a single atom, a discrete set of radiation frequencies is 13  TEMPERATURE (°K)  Figure 2.1 Vertical temperature profile for the U.S. Standard Atmosphere, (from Wallace and Hobbs, 1977)  Figure 2.2 The annual mean global energy balance. Numbers represent percentage of the globally averaged solar irradiance incident upon the top of the atmosphere (from Wallace and Hobbs, 1977) 14  observed. This is due to the fact that an electron in an atom can only be at one or another set of discrete energy levels. In addition, not all electron transitions are allowed to occur. For instance, when the average electron position is equal to zero, there is no oscillating charge and hence no absorption nor emission takes place.  When energy levels change during each transition, finite widths are observed in the spectral lines. The natural cause for the broading of spectral lines is due to the finite duration of the molecules in an excited state and due to the loss of energy in emission. Furthermore, collision among absorbing molecules and the Doppler effect resulting from different thermal velocities of atoms and molecules, also contribute in broadening the spectral width. The effect of natural broadening is practically negligible in the presence of collision and Doppler type broadening. The Doppler type broadening is not significant in the lower part of the atmosphere where due to high pressure, collision broadening dominates. The monochromatic radiation we have thus far discussed is almost never observed in nature. The total energy of the molecule includes contributions from vibration and rotation. Suppose a stable equilibrium in a molecule occurs when the nuclei are separated by a distance RQ.  If the nuclei are separated by a slightly greater or smaller distance,  R, the energy is increased and there exists a restoring force. The electrons follow the nuclei as R varies. This causes small amplitude vibration about RQ and absorption or emission takes place. The kinetic energy of the nuclei rises when molecules rotate about their common center of mass and this is referred as the rotational energy of the molecule. The rotational energies are relatively small. The lines of the pure rotational spectrum occur in the infrared and microwave regions.  The computation of the absorption spectrum is a complicated process as spectral lines due to all types of atmospheric gases, broadened by collision and Doppler processes, extend out to far wings. The spectral lines then overlap each other producing mutual effects making the theoretical description of absorption spectra virtually impossible. Thus, empirical 15  absorption model seems to be more successful.  2.2.3 Molecular Absorption  In The IR Window Region  In the 11 — 13 pm infrared region of the E M radiation spectrum, C O and H 0 are the a  2  major absorbers and these two absorption processes are examined here in detail. Ozone is another dominant absorber in the infrared region; it however plays relatively little importance in the window region considered. The major infrared absorption by atmospheric gases is shown in Figure 2.3. This figure well illustrates relative transparency in the atmospheric window regions.  V"  r 100 0  CO  1  1  i  100 0  s c  a  0 ioo  1  •  1  I I .  1  1  '  •  ' CH,  V . tl  V 1  1 1 1 1  1  V  100 v  —1  Y  100 0  is  1  0  100  1 . y:X. 1001W-AA  * HDO  100 0  7  1  2  3  4  5  6  AT"  r " '* ~V*Vv.  a  10  9  U  N  r  11  12  Aggregate  13  14  IS  W a v e l e n g t h (tun)  Figure 2.3 Infrared absorptions by atmospheric gases, (from Handbook of Geophysics and Space Environment, 1965)  C O j is a linear type molecule where three atoms form a symmetrical line with the carbon atom in the center and one oxygen atom at each side. Figure 2.4(a) shows three normal modes of vibration for such linear molecules. This type of molecule possess no permanent dipole moment and hence symmetrical motion is inactive in V\ vibration. Also, 16  there exits no pure rotational spectrum in linear symmetrical molecules. The fundamental  v<i vibration mode occurs at 15/tm and 1/3 at 4.3/xm in the E M R spectrum. The water molecule is an asymmetric top type where two hydrogen atoms and an oxygen atom form an isosceles triangle. Figure 2.4(b) shows three normal modes of vibration from the molecules of such type. The i>a fundamental is found at 6.3pux and v\, v$ are found near 2.7 pm. The pure rotational spectrum of the asymmetric top molecule has a complex spectrum and it is difficult to mathematically formulate its absorption spectrum. Bignell (1970) suggested that the absorption coefficient due to the water vapor continuum in the IR window region should take the following form:  (2.5) where p is the total pressure and e is the water vapor pressure. The first term is due mainly to collision broadening in the wings of the water vapor absorption lines. The second term is partially due to self-broadening in the wings of the water vapor line. But, a more dominant influence is from the spectrum due to water dimers. The physical reaction of the water dimer is described by 2H O ^ ( H 0 ) + E , a  3  3  (2.6)  where E is the binding energy of the dimer molecule. The existence of the second term is also confirmed by Burch (1970). Ozone contributes about 12% of the absorption in the 8 — 13 pm window region. The 9.6 pm ozone band has a narrow bandwidth. Even though it consists of a large number of moderately intensive lines, the effect in the window region is relatively small and thus ignored in this study.  17  (a)  (b)  Figure 2.4 Three normal modes of vibration, (a) linear molecules and (b) asymmetric top type molecules 2.S Radiative Transfer Equation ( R T E ) The E M radiation radiated from a surface, must travel through the atmosphere before it reaches the remote sensor on-board a satellite. In the atmosphere, the E M R from the surface comes in contact with atmospheric constituents which normally attenuate the surface radiation. The amount of interaction is a function of the atmosphere, its composition, and the wavelength of the E M R and is expressed by the radiative transfer equation. For a nonscattering atmosphere in local thermodynamic equilibrium, the steady state radiation transfer equation appears as a set of linear, first-order ordinary differential equations. An expression for a measured radiance, /(f), within an atmospheric absorption band can be expressed as  I{u) = B{u, T.)e{u)T{v p ) t  Q  + f  B[u, T(p)) dr{u,p) (2.7)  where B is Planck's function at wave-number v, r(i/,p) is transmission of radiation between pressure level p and the effective top of the atmosphere, po is pressure at surface, and p is the reflectivity of the target. The first term in equation (2.7) is due to the contribution from the surface, which is blackbody radiation at temperature T, with emittance c and IS  the transmittance r(i/,po) or r<> between the surface and the top of the atmosphere. The second term is due to the contribution by the atmosphere in the field of view (FOV). The third term is due to the downwelling radiances from the atmospheric constituents and then reflected back from the surface. Thus the measured radiance I(y) is a weighted mean of Plank's function profile. To retrieve the temperature profile T{p), equation (2.7) must be inverted to solve for the Plank's function, B(u,T ), t  given the atmospheric transmittance  r(u,p), and the measured radiances /(»>). 2.4 S S T Retrieval Techniques The SST retrieval techniques involve physically or empirically solving the radiative transfer equation from a set of remote measurements to obtain the surface blackbody radiation and retrieve the surface temperature by inverting Planck's function. In the past, attempts were made to retrieve surface temperatures from a single window channel with a limited success. Additional window channels (2 or more) have been employed with improved results. The SST retrieval method using two scan angles has also been experimented on with some successful results.  2.4.1 Single Channel  Technique  From the R T E , we consider a monochromatic radiative transfer. Solving equation (2.7) for B(T,), we have 1 -r(",Po) €(i/)r(t/,p )  (l-r(i/,p )], 0  (2.8)  0  where e is the emittance of the target,  and  C ){%r(p)l/r (^P)}^(.,p) 2  T  =  0  U the surface emittance is assumed to be unity, we have (2.9) 19  Thus, given accurate atmospheric temperature and moisture profiles, it is theoretically possible to solve each term and finally the surface temperature, T . The limitations of this 9  method are: a) One requires an extremely narrow sensor response function over the infrared window region. Over the sensor response spectrum, the atmosphere must be transparent enough such that the atmospheric contribution is less than the sensor noise. b) one requires very accurate profiles of all major atmospheric absorbing constituents over the sensor response spectrum. The profiles should have characteristics similar to those of the sensor, thus radiosonde data alone is not sufficiently representative. In addition, an accurate transmission model is required for all constituents considered. Due to such problems, the single channel method is almost never used unless additional information can be provided. The method lacks information to compensate for the water vapor effect.  2.4.2 Split Window  Technique  The Split Window Technique (SWT) uses two neighboring atmospheric windows corresponding to the two wavelengths required for correction. In the AVHRR/2 sensor, these wavelengths are channel 4 (11 pm) and 5 (12 pm). The basis of the SWT is that a relationship exists between the surface temperature and the radiances measured by the two window channels. This assumption has been verified by several investigators (Anding and Kauth, 1970; Deschamps and Phulpin, 1979; Prabhakara et al., 1974; and Ulivieri, 1984). If the surface emissivity is assumed to be unity, the emerging radiation, at the top of the thermodynamically equilibrated and non-scattering atmosphere, takes the following form. (2.10) where T' is the sea surface temperature, r is the atmospheric transmittance, p is the atmot  spheric pressure, B is Planck's function, and I{u) is the observed radiance. Rearranging 20  this equation in terms of the upwelling mean atmospheric radiation, B^(u,T ), a  Iiv)  = B{v,T$T{» po)  + Bi{u T )[l-T(»,po)] t  t  we have (2.11)  m  If the radiance for a F O V is measured at two wavelengths, V\ and v*, = B{y T',)T{vuPo)  I M  + Bi(u Ta)[l  u  i M  - T(uliPo)]  u  (2.12)  = B(^,r;)r(z/ ,po) + J J ( ^ , r « ) [ i - r(i/ ,p )]. t  3  3  0  (2.13)  In order to combine (2.12) and (2.13) to extract T'„ both equations must be transformed to a reference wavelength. If v\ is selected as the reference wavelength, only equation (2.13) needs to be transformed. When two channels are relatively close together, the dependence of Planck's function on the wavelength, dB[u,T{p)\/dv, = B{u ,T[)r{y^p )  I'M  x  is negligible. Thus, we have  + B ( i / , f ) [ l - r(i/ ,po)].  0  T  a  a  3  (2.14)  where /'(fa) is the equivalent radiance measured at i/ , and expressed in terms of the 3  reference wavelength. Through simulation, a relationship between the surface temperature and the radiances in two channels can be derived (Dalu et al., 1981; Uliveri, 1984) as  B{u X) x  - B^Ta)  = c[B{u T:) u  - B^u f )\. u a  (2.15)  The value of e is found to be 1.165 ±0.05. Combining equations (4.3), (4.5), and (4.6) and solving for B{ux,T' ), t  we get  B^X)  =h+  c  * - - l - r ^ - ' J I  (  l  < > 216  where / , is equal to /(f ) and r,- is equal to r(i/,,po)> Now, if Planck's function is inverted, t  the surface temperature T[ can be obtained. The coefficient of the correction term is a function of water vapor content. Dalu et al. (1981) found that if an independent means of estimating water vapor content is available, the SST measurement can be improved. McMillin et al. (1984) provided an alternate method directly involving brightness temperature values. If a Taylor expansion to the Planck's function about the temperature is used, the following SWT algorithm can be derived. Tt^Ti  + niTt-Ti)  21  (2.17)  or 2? = ao + (1 + 7)^1 - 7 * 2  (2.18)  where  ca-raJ-U-n)  7  and OQ is a constant to correct for a bias. The value for gamma can be derived from a sample of atmospheres, or it can be found by performing a regression against ship and buoy SST measurements. When the coefficients are derived using statistical means, the emittance effect is usually incorporated in the regression process. However, in the physical method to be described later, the emittance effect has to be explicitly corrected for. The surface blackbody radiation, B(vi,Tfl,  derived in equation (2.16) is under the assumption that the emittance of  the target is unity. By adding the correction term, we get  B(v T.) lt  = B{u ,T' ) x  =  t  B{u T' ) u  t  + p(ui)B{ui,T,)  - pMB^T^l  -p{u )B {u T )\\ x  {  u  -  a  Tjvupo)]  -  T{U1)PO)]  •  (2.19)  I-PM  The emittance corrected temperature T can then be derived by inverting Planck's function. t  2.4.3 SWT With Scan Angle  Correction  In order to study the impact of atmospheric water vapor and temperature structures, together with the scan angle variation of the radiometer, on the radiance measured by the AVHRR/2 split window channels, Schluessel and Grassl (1985) performed radiative transfer calculations for a set of 182 tropical and middle latitude atmospheres accounting for seasonal variations in the latter. Computed brightness temperatures, from AVHRR channels 4 and 5, are then used to fit the ship observation of surface skin temperature using a linear regression model to find a set of coefficients for the split window formula  T =a +a t  0  + oaT^,  x  (2.20)  where T/* represents the brightness temperatures in the split window channels (for more details see Strong and McClain, 1984) and T represents the sea surface temperature. A t  22  noise level of 0.12 °C in the brightness temperatures, was considered within the simulations to account for radiometric noise. A wide variety of atmospheric situations was included in the same regression to retrieve a quasi-global set of coefficients a,- as given in Table 2.1  Table 2.1  Regression coefficients for equation (2.20) Scan angle(degree) OQ ( ° C ) a\  a  3  RMS error ( ° C )  radiometric + atmospheric noise -0.99 0 -1.01 10 20 -1.05 30 -1.14 40 -1.21 50 -1.53  3.659 3.688 3.774 3.918 3.926 4.207  -2.641 -2.670 -2.756 -2.899 •2.904 -3.172  0.61 0.62 0.64 0.70 0.70 0.79  atmospheric noise only 0 10 20 30 40 50  4.000 4.027 4.089 4.206 4.230 4.500  -3.004 -3.031 -3.093 -3.229 -3.229 -3.485  0.34 0.35 0.38 0.43 0.43 0.53  -1.23 -1.25 -1.29 -1.37 -1.46 -1.83  from Schluessel (private communication)  The numbers in Table 2.1 demonstrate the strong scan angle dependency, with increasing standard errors at larger scan angles, indicating the impact of atmospheric structure on the radiances seen at the satellite altitude. The importance of scan angle correction, for the calculation of sea surface temperature from multi-channel AVHRR data, has been recently emphasized by Kelly and Davis (1986). Other studies (Maul, 1983; Dalu, 1985) have also investigated the magnitude of SST errors due to differences in scan angle. However, due to the discrete use of the scan angle the RMS errors are slightly smaller than those found by Strong and McClain (1984) and Strong (1984) for satellite and drifting buoy match-ups. The retrieval coefficients seem more robust against atmospheric peculiarities whose impacts increase at larger scan angles. A further study of the numerical simulations was done by comparing the observed surface skin temperatures with those computed from the synthetic brightness temperatures using (2.20). Figure 2.5 shows the differences be23  tween these two SST's plotted against the measured surface skin temperature giving a first impression about systematic errors introduced when using a simple two channel approach. Since these errors increase with increasing scan angle they must be caused by variations in atmospheric structure.  +  +  +  ! . :  +  •  +  +  +  + '  + +  +  * ^  *  T  •  + +  + +  v  ^r^F  \  +  +  +  +  +  +  --  +  ^  +  ± * +  -  +  V +  +  +  i  • * ++  +  +  +  ++ +  1  1  !  1  1 S  1  1  I  12  i  16  1  1 20  1  ;  2L  1  1 28  Figure 2.5 Differences between simulated and observed SST vs observed SST using eqn. (2.20). Systematic errors are caused by variations in atmospheric structure.  An examination of those cases with the largest differences, between measured and satellite retrieved surface skin temperatures, points to two different sources of the large errors in the split window results. First, for T ir > T„ which is a standard condition in a  coastal waters in spring, the brightness temperatures predict far higher SST values than really occur. Second, extremes of water vapor content in the atmosphere, (e.g. tropical atmospheres with especially high water vapor contents or "dry lines" just behind cold front passages showing considerably drier conditions than "normal" atmospheres), require atmospheric corrections that lie beyond the range of the split window coefficients. Consequently, an improvement of the split window SST retrievals should be possible with 24  additional information about the low level tropospheric temperature structure and the water vapor anomalies at heights that are not covered by the AVHRR window channel atmospheric weighting functions.  2.4.4 Extension of the AVHRR  SWT Using HIRS  One possible way to overcome the difficulties which arise with the simple split window approach (which suffers from a poor knowledge of the thermodynamic state of the atmosphere) is to use atmospheric sounder data from other radiometers on board the same satellite. The TOVS package with its High Resolution Infrared Radiation Sounder (HIRS) and Microwave Sounding Unit (MSU) should be able to deliver valuable information to correct the systematic errors in the split window SST retrievals caused by the atmosphere. However, the assumption of horizontal homogeneity of atmospheric temperature and moisture fields, within one sounder resolution element, has yet to be justified. The information needed for an adequate SST retrieval, even in pathological cases, can be taken from HIRS channels which mainly monitor the atmosphere rather than the surface. The HIRS water vapor channel 11 has the maximum of its weighting function at approximately 700 mb and therefore indicates properties of the vertical water vapor structure not detected by the AVHRR window channels. The use of this additional channel promises a supplemental correction in cases with extreme water vapor contents (anomalously dry and wet atmospheres). HIRS channels 6, 7, 8 with weighting functions peaking at 700, 900, 1000 mb respectively, provide information about the temperature structure in the lower troposphere. The ratio  r  where  =  (r*"-r*j'  ( 2 , 2 1 )  is brightness temperature in HIRS channel », increases with atmospheric stability  in lower tropospheric levels. One can expect cases of extreme (7 > — T ) differences, that at  9  the HIRS channels will provide a way to correct the AVHRR SST retrievals for these anomalous conditions. Hence, an appropriate model is T. = a + a Tf Q  1  + a^Tf + a T ^ 3  25  + a r 4  (2.22)  where T* is brightness temperature for AVHRR channel i. Computer simulations of the atmospheric radiative transfer for the desired N O A A - 7 channels have been performed for the above mentioned set of 182 atmospheres. The synthetic brightness temperatures are then fitted to (2.22) in order to regress the coefficients a,- given in Table 2.2. The radiometric noise again enters the radiative transfer calculations leading, together with the atmospheric errors, to the given RMS values.  Table 2.2  Regression coefficients for equation (2.22) Scan angle(degree) a ( C) a%  a  8  0  3  03  a  4  RMS error ( ° C )  radiometric + atmospheric noise 13.27 0 13.52 10 20 13.84 14.55 30 40 13.38 50 11.74  3.239 3.271 3.388 3.591 3.653 4.118  -2.261 -2.295 •2.419 -2.633 •2.688 •3.154  0.0734 0.0762 0.0849 0.1010 0.0967 0.1130  -7.021 -7.131 -7.240 -7.500 -6.882 -5.921  0.54 0.55 0.56 0.61 0.62 0.72  atmospheric noise only 0 10 20 30 40 50  4.442 4.471 4.554 4.725 4.781 5.191  -3.537 •3.568 -3.657 •3.839 -3.883 •4.288  0.119 0.121 0.128 0.144 0.132 0.142  -1.752 -1.819 -1.838 -1.838 -1.194 -0.0416  0.20 0.20 0.22 0.26 0.27 0.35  3.59 3.76 3.91 4.15 2.81 0.65  from Schluessel (private communication)  A comparison with Table 2.1 shows decreased temperature errors for all scan angles. Figure 2.6 includes, as does Figure 2.5, differences between measured skin and satellite regression temperatures. Figure 2.6 clearly demonstrates the elimination of extreme outliers and hence the potential of the extended split window method to give adequate results even for "bad* atmospheres. The improvement to the stand-alone AVHRR SWT by including water vapor information has been theoretically established by Dalu (1985) and Ulivieri (1984). To these researchers, the question remained to where the additional information should be obtained from. By including water vapor information from the HIRS water vapor channels, significant improvements in the atmospheric correction can be achieved. 26  For example, the standard errors decreased from 0.34 °C to 0.20 °C at nadir view and from 0.53 °C to 0.35 °C at a scan angle of 50° in going from a stand-alone AVHRR retrieval to an AVHRR+HIRS method. The lower parts of Tables 2.1 and 2.2 give coefficients and errors for a regression neglecting the radiometer noise which can be achieved in practice by spatial averaging of the brightness temperatures over many pixels. The remaining errors are due mainly to "atmospheric noise" which is systematic and cannot be reduced by spatial averaging. The main disadvantage of this method is the difference in the FOV of AVHRR and HIRS systems. The coarse resolution elements of the HIRS sounder must be interpolated to the AVHRR scan patterns making the SST retrievals partly subject to the interpolation method used.  i  l  -t  i  +  i  +  +  i  + ++  +  *  -u t  +  %;  +  .  + +  4 + + + V  -tt+  + +  + +  T  +  +  + + %  +  +++ + ++  I *  +  +•  +  +  + + -H-  -  — 0  i  —  i  — £  i  —  1  — i  8  I  i •r  12  IS  1  T  1  20  —i  1 24  1  1 28  T  Figure 2.6 Differences between simulated and observed SST vs observed SST using eqn. (2.22). Systematic errors are caused by variations in atmospheric structure  27  2.5 A V H R R S W T with T O V S Derived Atmospheric Profiles A physical SST correction method, developed in this study, uses T O V S profiles in order to obtain the exact transmission functions for the atmosphere and hence offers a refinement to the split window technique. This method does not require any concurrent external surface observations and is thus very attractive. Keeping in mind the emittance effect (increase in apparent SST due to downwelling atmospheric radiance), the success of this SST correction method depends on the capability of the TOVS sensors, the accuracy of the profile retrieval technique, and the correctness of the transmission model for the interacting atmospheric gases. The capability of TOVS to supplement radiosonde (RAOB) data, where observation stations are sparse, has been shown by Kelly et al. (1983). The comparison of TOVS retrieved temperature and moisture profiles with radiosonde data has been performed by Philips et al. (1979), Bruce et al. (1977), Hillger and Vonder Harr (1979), Moyer et al. (1978), and Hayden et al. (1981). Their findings were rather discouraging as large RMS deviations were found when TOVS profiles were compared with the nearest coincident radiosonde measurements. Even though the two measurements were over the same atmosphere, the measured characteristics differed significantly. Radiosondes make point measurements as they ascend continuously whereas TOVS makes horizontally and vertically integrated measurements. The RMS statistics between these two different measurements do not provide proper accuracy estimates of the TOVS retrievals. Such comparisons are, however, the only available way to gain some confidence in the T O V S retrievals. The moisture profiles produced even poorer results with the dew point RMS deviations as large as 7.3 °C. However, Moyer et al. (1978) reported that when moisture profiles were analyzed in terms of total integrated precipitable water, the deviation was only 20 % of the total precipitable water.  2.5.1 Vertical Profiles Retrieval  28 Retrieval Theory The retrieval of atmospheric profiles from a set of remote measurements is similar to many geophysical inversion problems. Ignoring the surface term in the R T E from (2.7), the radiative transfer is described as: B[u,T{p)]dT{u,p).  J(i/)=f  In the 15 ftm C 0  3  (2.23)  band, the transmission of the surface term is only significant in the  spectral wing region and progressively less important towards the center of the absorption band. The temperature dependence of Planck's function is much stronger than the frequency dependence over this absorption region. Thus, it is sufficient to represent Planck's function in terms of a linear approximation about a reference frequency, i/ , r  B[u, T{p)\ = a(u )B[u , r  r  T{p)] + b(u ),  (2.24)  r  where a{u ) and b{u ) are empirically derived coefficients. The value of v is selected near r  r  T  the center of the absorption band. Substituting equation (2.24) to (2.23), we have 9{")=  f° f{p)K[v,p)dp, 'P.  (2.25)  Jo. where 9 { U )  ~  f(p) =  a(u)  '  B[u,T{p)),*nd.  For observations made with M channels, we have  9i = [ fiP)Ki{p) 0  dp ,  » = l,2,...,Af.  (2.26)  The solution to this equation is underconstrained as the unknown profile f{p) is a continuous function of pressure and there are only a finite number of observations, g+s available. 29  Now, express f(p) as a linear combination of N variables.  /(p) = E / y % ( p ) .  (2.27)  3=1  where / / are unknown coefficients and Wj(p) are the representation function. Then, * = Z)/>/ 3=1  &) i(P) P>  W  K  » = 1,2,...,M.  d  (2.28)  *»  J  Defining a square matrix,  fw {p)Ki{p)dp 3  %  we now obtain fl = 2  1 » = l,2,...,Af  (2.29)  3=1  or in vector form, g = A f . The matrix A can be solved from equation (2.29) and it is substituted into equation (2.30). The profile f can be found by  f = A-»g = ( A A ) " A g , T  where A  T  1  T  (2.30)  represents transposed matrix of A . By substuting f into equation (2.28) we have N  /(P) = £  Wjtp)*uH  » = 1,2,..., M .  (2.31)  The above method provides an exact solution to the radiative transfer equation. This seems straight forward. However there are a number of problems associated with the exact solution. First, because the equation (2.31) is underconstrained, the solution is unstable. Furthermore, inevitably there are measurement errors introduced with the value of gt such that g, — gt + €,-. This means that there are a number of possible solutions of f3- which fall within the measurement error, e \ Secondly, the measurement error can be greatly g  amplified during the process especially when the inversion is unstable or non existent. Then the solution becomes virtually useless. Thirdly, due to overlapping of the functions representing W/(p) for the HIRS instrument, certain ambiguities are introduced into the solution which makes the solution very sensitive to the noise.  30 Improved Iterative Solution T O V S sounding of the atmosphere involves retrieving vertical profiles of temperature and moisture from a number of discrete radiance measurements. For the HIRS/2 infrared sounder on-board the NOAA spacecraft, there are 8 channels at 15 /xm and 5 channels at 4.3 fim in the C 0 absorption bands of the infrared spectrum. The radiances to estimate 3  atmospheric moisture are measured with 3 channels at 6.7/xm in the water vapor absorption spectrum. Because only a finite number of measurements are available, the problem of inverting the nonlinear R T E to solve for the surface radiance is ill-posed, that is there are a number of possible solutions. In order to converge on a single solution, additional external constraints are necessary. The additional constraints can be obtained from independent atmospheric profiles, such as those from radiosondes. This approach is usually fast and convergence is always achievable. However, the involvement of statistics limits the validity to local regions where the ROAB samples are taken. The requirement for up-to-date R A O B data, for example over an open ocean, imposes a practical limitation to the method. Also it is possible that a solution which represents the true state of the atmosphere (perhaps a strong anomaly), may be rejected because the general statistical constraints, for typical atmospheric structures, does not allow it. An alternate to the indirect, statistical approach is to directly solve the R T E using an iterative procedure. This method has the disadvantage of requiring a representative initial profile and convergence is not always guaranteed. Also, there is no way of knowing to which solution the process has converged. The iterative process used to retrieve the temperature profile is from Smith et al. (1983) and is as follows.  T»  + L  (p) = T*(p) +  s;  &  (2.32)  where T (p) is the nth iteration temperature at pressure p, m is the number of channels, n  and TB * is the brightness temperature calculated from the temperature profile SI  T (p). n  The weighting function, W {i/ p) n  where R [fi) n  is equal to  it  is the nth iteration Planck's radiance for channel  emerging at the top of  the atmosphere. The iteration is terminated when the brightness temperature difference between two successive iterations is less then 0.05° C. The iteration procedure for the moisture profile is adapted from Hillger (1984). This method is a variation of Chahine's relaxation method to handle the temperature inversion near the surface which often exists over the ocean in the summer time. The moisture profile is expressed in terms of the mixing ratio, Q[p). The iteration procedure is expressed as  E^AJ^Ar^p) Q  N + 1  (P) = Q  (2.33)  »  EAr, (p) n  8=1  where Ar (p) = r , ( p ) - T ( p t  m  i  ), A/,- = n  m + 1  h,ob»-U,caie  * \ and S is the factor for converting  from radiance change to mixing ratio change. S is defined as  ^ r" Wi -  s  =  m  sn< np)h  (234)  where U(p) is the precipitable water integrated to pressure level p. The importance of the initial atmospheric profile for the TOVS solution has been stressed by many researchers (Crosby and Weinreb 1976, Spanuch 1978, Thompson et al. 1980). Usually initial profiles are obtained from climatological profiles, forecast profiles, or standard profiles. Even an isothermal profile at a selected temperature may be used. Often it is overlooked, that the characteristics of these profiles are quite different from the satellite derived profiles. Also, there usually are large differences in time and space between these profiles and the satellite retrieval. Keeping these points in mind, the initial guess temperature profile was generated using interpolation from the climatological mean profiles stored by month and latitude zone. For a moisture guess profile, the climatological mean profiles do not serve as good initial profiles due to highly variable nature of the 32  atmospheric water vapor. The best way to estimate the initial water vapor profile for the FOV is from the HIRS water vapor channels themselves. In the atmosphere the change of lapse rate occurs in ah irregular manner such that it can not simply be described in terms of pressure (Shen et al. 1986). However, the atmosphere can be described as layers t  with constant lapse rate for each layer in hydrostatic equilibrium. For the initial water vapor profile, the measured radiances from three HIRS water vapor channels were used as the estimates for the pressure levels where those channels peak. The radiances for pressure levels in between were interpolated using the power-law decrease in mixing ratio with pressure p, Q(P) = Qmaxip/Pmaxf ,  (2.35)  where T is the lapse rate which was derived from the brightness temperature values of the HIRS water vapor and the window channels. The surface temperature during the iterative process, is estimated from 2 HIRS/2 longwave window channels (11 and 8/im) according to the following iterative relationship; r, = r ; + ^ = - p ,  (2.36)  where T& refers to the brightness temperature computed for an estimated surface temperature, Tf, and atmospheric temperature and moisture profiles, and r(p ) is the transmit0  tance above the surface pressure level, p . 0  2.5.2 Calculation  of TOVS  Transmittance*  The spectral transmission function of the radiation between an atmospheric level p and the effective top of the atmosphere is described by 1  t  v  n ( j / P ^ ) = e x p ( - - / k{v,p')q{jp')sec9dp'\ 9 Jo >  (2.37)  where g is the gravitational acceleration, k is the absorption coefficient, and q is the mass fraction of the absorbing gas. The quantity inside the square bracket is known as the monochromatic optical depth of the absorbing gas. If there are JV absorbing constituents, 33  the total transmittance is the product of the transmittance for each constituent. Thus, N  T{",P,9)  = Hri(v 9).  (2.38)  tPt  i=l  For a radiometer which responds with its finite spectral band width, the integrated transmission function takes the following form. (2.39)  The atmospheric transmittances for the TOVS are calculated using the method described by Weinreb et at. (1981). The line-by-line method of calculating transmittances requires large computational time. This method parameterizes the line-by-line method and these parameters are used in the actual calculation of transmittances. The method is currently being used at the National Environmental Satellite Services (NESS) and since the description can be found in Weinreb et at. (1981), only a brief outline is given in the following sections.  HIRS/2  In the infrared region, major interactions are caused by C 0 , N 0 , C O , C H 3  3  0 , and N as mentioned earlier. Among these, water vapor, C O 3 , and 0 3  3  3  4 >  H 0, 2  are the domi-  nant absorbers in the earth's atmosphere. For simplification, these active constituents are grouped into six categories: 1) Spectral lines of the uniformly mixed gases ( C 0 , N3O, C H , CO, 0 ) , 2  4  2) Spectral lines of water vapor, 3) Spectral lines of ozone, 4) Water vapor self-broadened continuum (dimer interaction), 5) Water vapor Ns-broadened continuum (H3O-N3 interaction), 6) Collision induced band of molecular nitrogen. 34  3  The spectral line absorptions are primarily caused by the transition between energy levels due to molecular vibration and rotation. The spectral lines have a finite width that is proportional to the finite lifetime the molecules stay at the higher energy level. The absorption bands are also broadened by Doppler shifts during transitions. The absorption lines are broadened further due to collisions between gas molecules, which is referred to as the continuum absorption. Since C 0 is uniformly mixed, the continuum absorption due to 3  C 0 can easily be estimated. However, for water vapor, these line broadenings are directly 3  proportional to the pressure and the temperature of the atmosphere, and are more difficult to account for due to their complex interactions. Molecular scattering is not important in the infrared region and is thus ignored. Also the aerosol effect is ignored because errors due to not including the aerosol effect, are relatively small (Stowe and Fleming, 1974). (1) Uniformly Mixed Gases The method of calculating transmittances in an inhomogeneous atmosphere for the uniformly mixed gases, is taken from the papers by McMillin and Fleming (1976), Fleming and McMillin (1977), and McMillin et al. (1979). It is an iterative method in which the transmittances for an arbitrary temperature profile and an arbitrary mixing ratio profile, are evaluated for 40 discrete levels from coefficients found empirically. This method is fast because the model uses only arithmetic operations. (2) Spectral Lines of Water Vapor Water vapor dominates infrared absorption in the lower atmosphere. For calculating transmittances of spectral lines, the method of Weinreb and Neuendorffer (1973) is used. This method approximates the inhomogeneous atmosphere as a succession of homogeneous layers, where pressure, temperature, and mixing ratio are assumed to be constant. The total transmittance r is computed as a function of the layer's pressure, temperature and the amount of water vapor. Then r is expressed as a polynomial, according to Smith (1969), with coefficients calculated off-line. (3) Ozone 35  Ozone has a vibration-rotation spectrum with the strongest absorption band near 9.6 /un which lies at the center of the longwave atmospheric window. The transmittances for ozone are calculated by interpolating between two transmittances that are compiled by Kunde and Maguire (1974), using the line-by-line method. The two transmittances are for the two concentration profiles at 257 and 480 Dobson units (one Dobson unit is equal to 10 cm of ozone). It is assumed that most ozone concentration profiles in the atmosphere lie between these two concentration profiles. The total ozone concentration is retrieved using channel 9 of HIRS/2. (4) Water Vapor Continua The absorption due to water vapor continua is very significant because the far wings of strong absorption lines are located near the atmospheric windows. For the self-broadened and foreign-broadened water vapor continua, the absorption coefficients are proportional to the partial pressure of water vapor and the partial pressure of the dry atmosphere respectively. The method of calculating transmittances for these absorptions are described in Weinreb and Hill (1980) with coefficients obtained from laboratory samples by Roberts  et al. (1976), Burch et al. (1973), and Burch et al. (1971). (5) Molecular Nitrogen Molecular nitrogen has a collision-induced absorption band centered at 2330 c m ! The transmittance is calculated with the method by Burch et al. (1971) and the coefficients are derived from data given by Shapiro and Gush (1966).  AVHRR  Transmittances for three infrared channels of AVHRR are calculated using the same method as for HIRS/2, except in two respects. First, transmittance for the ozone is not included. Second, transmittance through the uniformly mixed gases, is calculated using L O W T R A N (McClatchey et al. 1972 and Selby et al., 1978). Further detail on how %  L O W T R A N is used, can be found in Weinreb and Hill (1980). 36  CHAPTER 3 D E S C R I P T I O N A N D P R E P R O C E S S I N G OF D A T A  3.1 Satellite Sensors The satellite data for this study were collected from the NOAA-7, the NOAA's current operational polar orbiting satellite.  NOAA-6 is still active. However, its sensors have  significantly degraded as the satellite has passed its expected lifetime.  Thus, only the  data obtained from the N O A A - 7 were used. The N O A A - 7 data used in this study were received from 3 separate H R P T ground stations; Satellite Oceanography Laboratory at the University of British Columbia, Scripps Institute of Oceanography in La Jolla, California, and CMS Lannion in Lannion, France. The N O A A - 7 meteorological satellite follows a sun-synchronous orbit over the earth, at a nominal altitude of 833&m. The orbital parameters for N O A A - 7 are given in Table 3.1. There are two major instrument packages on-board these satellites of interest to this study. They are the AVHRR and the T O V S . The TOVS system consists of three separate sensors; the Stratospheric Sounder Unit (SSU), the High-resolution Infrared Radiation Sounder/2 (HIRS/2), and the Microwave Sounder Unit (MSU). The following sections briefly describe these individual sensors.  Table 3.1 Orbital parameters for NOAA-7 Parameter Inclination of orbit Satellite Height Nodal Regression Nodal Precession Orbits per day Orbital period Cycle duration  NOAA - 7 99.092° 833 km 25.40 degrees/orbit W 0.986 degrees/day E 14.2 101.58 min 1 day  37  3.1.1 Advanced Very High Resolution Radiometer  (AVHRR)  The AVHRR consists of scanning radiometers at five spectral bands; one in the visible (0.5 pm),  one in near infrared (NIR) (0.9pm), and two in the thermal window of the  electromagnetic spectrum (3.7 pm and 11 pm) and one (12 pm) to measure the water vapor effect on the window channels. The visible and NIR channels are useful in discerning clouds, land-water boundaries, and snow/ice covered areas. The instant field of view (IFOV) for all channels is specified to be 1.3 milli-radians which corresponds to 1.1 km at the subsatellite point.  With 2048 pixels (picture elements) across the swath per channel, the  AVHRR covers a track about 2700 km wide (see Table 3.2). The visible and NIR channels of the AVHRR use silicon detectors to measure incident radiation. The IR channels use photo detectors which convert photon energy directly into the excitation of electrons into the conduction band. The incoming radiation produces electron-hole pairs separated by a space charge field. This effect in turn produces photo current which can be detected and retrieved. The widely used photo detector is of the type Mercury Cadium Telluride (HgCdTe) which is operated at 105 ° K in the spacecraft. The spectral response covers 3 pm to 12 pm, peaking at 10 pm. The sensors of AVHRR channel 4 and 5 are of this type. For channels 3, Indium Antimonide (InSb) is used. InSb is a photo detector which also operates at 105 ° K and responds from lpm to 5.6 pm with a peak response near 5.6 pm. equivalent temperature difference [NEAT]  The noise  is better than 0.12 ° K at 300 ° K target for all  IR channels. The scanning mechanism of the AVHRR is designed to operate such that the sensor is scanned from viewing the space towards the sun which is always to the left of the spacecraft velocity vector. By pointing at the cold space before each scan, zeroing of the sensors are effectively achieved removing unwanted biases. Looking in the spacecraft velocity vector, the AVHRR scans from right to left, which translates east to west in geographical coordinates. 38  Table 3.2 NOAA-7 AVHRR Characteristic  characteristics NOAA  Latitudal Coverage Ground Coverage Maximum scan angle IFOV Ground resolution (nadir) Ground resolution (Maximum off-nadir) Data precision Channels (/tm) 1: 2 3: 4: 5:  -  77AVHRR  90° N - 90° 5 2700 km db55.4° 1.39 - 1.51 millirad 1.1 Am 2.4 km along track 6.9 km across track 10 W* 0.58 0.725 3.55 10.5 11.5 -  0.68 1.1 3.93 11.5 12.5  (from Schwalb, 1978)  3.1.2 TIROS Operational  Vertical Sounder  (TOVS)  The T O V S system senses very narrow-band portions of the thermal radiation from uniformly mixed atmospheric gases such as Oj , or CO] . Table 3.3 gives the electromagnetic characteristics of the HIRS/2 and Table 3.4 shows the MSU sounding channels. In Figure 3.1, HIRS and MSU ground coverage spots are shown for one earth scan period. No descriptions are given for the SSU instrument, because it is not used in this study. In the following sections, the two components of vertical sounding instruments are briefly described. More detailed descriptions of these instruments can be found in Schwalb (1978).  39  Table 3.3 Characteristics of HIRS/2 channels HIRS Channel Principal Channel Channel wave wavelength absorbing (pm) number number constituents Temperature soundine 1 668 2 679 3 691 4 704 5 716 6 732 7 748  15.00 14.70 14.50 14.20 14.00 13.70 13.40  CO, CO C0 CO, COj C0 /H 0 C0 /H O  Surface temperature 8 898  11.10  Window  9.70  0  Water vapor sounding 10 1217 11 1364 12 1484  8.30 7.30 6.70  Temperature soundine 13 2190 14 2213 15 2240 16 2276 17 2361  Total ozone 9  Level o f peak energy contribution  NEAN mu//(m — srcm~ )  30 mb 60 mb 100 mb 400 mb 600 mb 800 mb 900 mb  0.80 0.27 0.22 0.22 0.22 0.22 0.22  a  3  2  3  2  a  Surface  3  1  0.11  25 mb  0.16  H 0 H 0 H 0  900 mb 700 mb 500 mb  0.16 0.22 0.11  4.57 4.52 4.46 4.40 4.24  N O N O C0 /N 0 C0 /N 0 C0  1000 mb 950 mb 700 mb 400 mb 5 mb  0.002 0.002 0.002 0.002 0.002  Surface temperature 18 2512 19 2671  4.00 3.70  Window Window  Surface Surface  0.002 0.001  Cloud detection 20 14367  0.70  Window  Cloud  0.1% Albedo  1028 High-resolution  3  3  3  3  a a  3  3  3  3  3  Infrared Radiation Sounder/2  (HIRS/2)  HIRS/2 has twenty (20) channels that are primarily located in the infrared region. Very strong spectral bands of C 0 occur in the shortwave (4.3 pm) and longwave (15 pm) 3  regions. Two sets of HIRS/2 channels, one near 4.3 um and the other near 15 pm, are used to sound the atmospheric temperature profile. Channel 9, located at 9.7 pm, is used for measuring ozone content and channel 20, located in the visible spectrum (0.69 pm), is used for measuring the earth's albedo. HIRS/2 also has three atmospheric window  channels at 11.1 pm, 4.0/im, and 3.7 pm and three water vapor channels at 6.7 pm, 7.3 pm, and 8.3 pm. HIRS/2 covers 56 FOVs across the swath of which the IFOV at nadir is 17.4 km in diameter. The HIRS/2 system parameters are shown in Table 3.5.  SUBORBITAL  T R A C K  Figure 3.1 HIRS and MSU ground coverage. Larger ellipses represent MSU foot prints  Table 3.4 Characteristics of M S U channels  MSU Channel number  Frequency {GHz)  Principal absorbing constituents  Level of peak energy contribution  Specified NEAT  1  50.31  Window  Surface  0.3  Surface emissivity  2  53.73 54.96 57.95  o o, o  700 mb 300 mb 300 mb  0.3 0.3 0.3  Temperature sounding  3 4  3  a  41  (°K)  Purpose  Table 3.S HIRS/2 system parameters Paremeter  NOAA  Calibration  Stable blaekbodiea (2) and apace background ±49.5° (±1120 km) 6.4 seconds 56 1.25° 1.8° 100 milliseconds 17.4 km diameter 58.5 km cross — track 29.9 km along — track 42 km along — track 2880 bits jsecond  Cross-track scan Scan time Number of steps Optical FOV Step angle Step time Ground IFOV (nadir) Ground IFOV (end of scan) Distance between IFOV's Data rate  -  7/HIRS/2 Microwave Sounder Unit (MSU) Radiation in the infrared region is heavily attenuated by water vapor in the atmosphere. In the presence of clouds it is almost impossible to obtain vertical profiles. In the microwave region, at some frequencies, clouds have little or no effect on the measurement. The MSU has three channels in the oxygen band near 60 GHz. MSU channels take 11 FOVs per scan to cover the same scan area as the HIRS/2 (see Table 3.6 for further description).  Table S.6 MSU instrument  parameters  Paremeter  NOAA  Calibration  Hot reference body and space background each scan cycle ±47.35° 25.6 sec 11 1.84 sec 7.5° (ZdB) 320 bps  Cross-track scan angle Scan time Number of steps Step time Angular resolution Data rate  -  7/MSU  42  3.2 In Situ Measurements For a verification of the satellite SST retrieval methods, comparisons with in situ sea surface temperatures measured from ships or buoys, are necessary. In most satellite SST verification studies (e.g. McClain et ai., 1985; Bernstein and Chelton, 1985) surface bucket, ship injection (engine cooling water), drifting buoy and X B T surface temperatures have been used as measures of in situ SST. As has been mentioned earlier there can be a significant difference between these "bulk" measures of SST and the skin temperature measured by the satellite radiometer. In order to better verify satellite skin temperature, and be able to estimate the differences between this surface skin value and that at a meter or two down in the water column, two different field experiments were carried out. In both experiments, which collected observations from the North Atlantic and the eastern North Pacific, a ship-mounted radiometer continuously monitored surface skin temperature while thermistors, in the upper few meters, measured the near surface bulk temperature.  3.2.1 North At Jan tic Ocean On the cruise of RV "Meteor" which took place during the period of October 20 to November 28, 1984, temperature and radiation measurements were collected in the Northeast Atlantic Ocean between latitudes 21 ° N and 54 °N. Radiometric surface skin temperatures and water temperatures, at depths of 2, 4, and 7 meters, were monitored continuously. The skin temperature was measured with a precision radiometer (Barnes PRT-5) every 2 minutes as 1 minute means. Each other minute was used for a calibration of the radiometer. The absolute accuracy of the skin temperature was calibrated to 1/20 °C, which included the correction for reflected atmospheric radiation. The in sttu water temperatures were sensed with platinum resistance thermometers (PT200) mounted at the specified depths measuring 1 minute temperature means with an accuracy of 1/80°C. In order to match in time the ship-borne temperatures with those measured by the satellite, the ship temperatures were averaged. Since the ship's speed was about 11 knots, 5 minute averaging was used to match the F O V of a single AVHRR pixel. RAOB's were 43  collected four times daily during the "Meteor" cruise in the Atlantic. Figure 3.2 shows the map of the cruise area with the cruise track.  3.2.2 North PaciBc  Measurements  Similar to the North Atlantic measurements, during the period of July 10 to July 19, 1984, the research vessel Pandora II was employed to collect both skin and bulk surface temperature measurements in the eastern North Pacific Ocean. The cruise covered an area to the west of Vancouver Island near 49° N and 125° W. The cruise tracks covered from Juan de Fuca Strait to Barclay Sound and extended out to the continental shelf break about 200 km offshore (see Figure 3.3). Sea surface radiation was measured using a Barnes PRT-5 radiometer sensing radiation for wavelengths from 9.5 pm to 11pm in the infrared spectrum. The radiation averaged measurements were recorded at 2 minutes intervals. Unlike the Atlantic data, calibration for the Pacific data was performed only once a day with spot checks done every 30 minutes. Calibration was done by mounting the radiometer facing a well stirred water bath and measuring the surface radiation along with the temperature while the temperature of the water bath was varied from 0°C to about 25°C. The temperature of the water bath was recorded using a precision thermometer. No RAOB's were collected from the ship during the Northeast Pacific cruise, instead the nearest land station (Quillayute, Washington State), coincident with this location were used for the verification of the T O V S atmospheric profiles retrieval method. RAOB data were also collected from Port Hardy, British Columbia, which is located in the north end of the Vancouver Island. The subsurface temperature was taken at a depth of no more than 2 meters below the surface by pumping the sea water through a hose which was dragged along the side of the ship. The temperature was measured with a precision thermistor as the water flowed by and averages over one minute period, were recorded by the on-board computer.  44  i  SO"  #  ]  N  \ \ i  r  J  ^u  >/ E 10°  10"  fcCRCATOR-PROJEKTION BEZUOSBREITE. 40.00 GRAD MASS-STAB. I • 20000000  Figure 3.2 Map of the Atlantic study area. Cruise tracks were covered during Nov. 15 and 26.  130U 52N  128U  126U  124U  122U  50N  SON  48N  120W 52N  48N  -  46N 130U  128U  126U  124U  122U  46N 120U  Figure 3.3 Map of the Pacific study area. Two cruise tracks represent transects for two days (Jul. 14 and 18). 45  3.2.3 Barnes PRT-5  Radiometer  The Barnes PRT-5 used, is a chopper radiometer with wavelength response at 9.5 — 11.5/xm. It consists of lens, detector, filter, chopper, and electronic circuitry. A lens focuses incoming radiation on the detector and the filter rejects wavelengths outside the filter band. The chopper is used to block the field of view periodically so the detector is exposed to the target and then to the internal cavity alternatingly. The cavity is temperature controlled at 4 5 ° C allowing 0.05 ° C variation even though the ambient temperature varies by 60 ° C . The radiation detector is a hyperimmersed thermistor bolometer which is basically a resistor that is thermally sensitive (ie. its resistance varies according to its temperature). The functional block diagram is shown in Figure 3.4 When the target radiation reaches the radiometer, the detected radiation is amplified and digitized. Table 3.7 outlines the performance specification of the Barnes PRT-5 radiometer.  OPTICAL  UNIT  T  ELECTRONICS  ,  |  '  •  I  I  I  I  UNIT I  , |  I  I xl/llClM  to-.ooc.f  J  Figure 3.4 Functional block diagram of the Barnes PRT-5 radiometer.  Figure 3.5 shows three sources of the radiance received by the PRT-5 radiometer 46  Table 3.7  Barnes PRT-5 performance Characteristic Temperature range ( ° C ) Accuracy( °C) Sensitivity ( ° C in 0.3 cps bandwidth) at 25 ° C Response (time constant) 0.3 cps 3.0 cps 30 cps Reference temperature Operating temperature Detector Lens Filter band Filter view  specification Value +10 to +45 0.5  Better than 0.1 500 milliseconds SO milliseconds 5 milliseconds 45°C±1/2°C -20°C*o +40°C Hyperimmersed thermistor bolometer 10 mm Itran - 2, f/2.8 9.5 — 11.5 pm 2° nominal  pointed downward over the sea. The PRT-5 is subject to the similar atmospheric absorption as the AVHRR infrared channels. However, usually for path lengths less than 300 m, atmospheric absorption can be ignored in the 8 - 12 pm atmospheric window (Katsaros, 1980). Correction the error due to emittance effect of the water can be corrected by measuring the sky radiance and solving for the emittance with the RTE. However, this technique is not used for the Pacific measurement. Instead an alternate method of keeping the measurement angle at the Brewster angle is used. At this angle the reflection is at a minimum for both horizontally and vertically polarized radiation (Grassl and Hihzpeter, 1975; Grassl, 1976). The Brewster angle for water is about 57° from zenith (Katsaros, 1980). The effect of sky reflectance is effectively eliminated by performing the calibration process under the condition where there was minimum reflection of the downwelling sky radiations. This is accomplished by controlling the angle of incident radiation of the PRT-5 radiometer.  3.3 Sensor Calibration  47  Figure 3.5 Three sources of the radiance received by the PRT-5. (from Katsaros, 1980) 3.3.1 Satellite Sensor  The AVHRR radiance values obtained from the HRPT stream represent voltage counts measured by five detectors. These voltage counts are relative values due to varying environmental conditions and the sensor system. In order to relate these voltage counts to absolute radiances and then to absolute temperature values, it is necessary to calibrate the sensor measurements. Before the launch, the spacecraft and its sensors are subject to extensive calibration procedures in a simulated environment. However, calibration information obtained from the prelaunch is not adequate for the duration of the spacecraft lifetime. To overcome this difficulty, NOAA spacecraft have the capability to perform onboard calibration for the infrared sensors for both AVHRR and HIRS/2. The on-board calibration for the visible channels is not carried out and one has to rely on the prelaunch calibration values. During every scan line, the AVHRR instrument views cold space and its housing. The cold space represents near-zero radiance reference while the instrument housing is designed to be a blackbody target at a relatively constant temperature (» 290 °K). Four Platinum Resistance Thermometers (PRTs) monitor the target temperature. By measuring sensor 43  responses at these two views and knowing the accurate temperature of the internal target, two reference points can be used to calculate a linear fit of the calibration curve. The linear calibration curve is represented by the gain and the offset for each AVHRR infrared channel. The calibration curve for each pass was calculated from collecting 50 samples for each PRT and 500 samples of both the target scan data and the space scan data, which required processing of over 200 H R P T minor frames. The collected data were then averaged to calculate the gains and the offset for each N O A A - 7 pass. The samples were taken over the region where in situ measurements were conducted. The HIRS instrument automatically enters a calibration mode every 256 seconds. First, the scan mirror points to view cold space for the duration of one scan line (equivalent to 48 scan spots). All HIRS channels sample during this time. Then the mirror is moved to view the cold internal calibration target.  Data are taken for the duration of time  needed to take samples for 56 scan spots. The mirror is then moved to view the internal warm calibration target for the same duration of time. After these three steps, the mirror returns to the start of scan position at which time normal earth scan can begin. The total calibration sequence takes the time equivalent to complete three Earth scan lines. Due to solar effects, the data taken from the internal cold target cannot be reliably used, and only the space scan and the warm target data are used for calculating calibration coefficients. Similar to the AVHRR system, four thermistors are used to monitor the temperature of the warm target during the calibration mode. Again, scan data of space are used as a near-zero temperature reference and the warm target data are used as the second point in calculating a linear fit to obtain the gain and the offset for each HIRS channel, except for the visible channel (channel 20).  Each MSU scan line contains parameters necessary to calibrate the MSU sensors. Since each scan line contains one sample, usually samples from several scan lines are averaged for the calculation of calibration coefficients. For MSU sensors, the relationship between 49  Input radiance and output counts cannot be linearly approximated, and thus a nonlinear algorithm is required. There are two internal targets and for each target, two PRTs measure the temperature of these targets. The first target is for use by MSU channels 1 and 2 while the second target is for the MSU channels 3 and 4. The PRT count values are first converted to resistance (R) and then converted to temperature ( ° K ) . The target temperature used for the calculation of calibration coefficients is averaged over 25 scan lines. The target radiance and the space radiance is then used to generate the gain and the intercept for each MSU channel. 3.3.2 Ship-borne  Radiometer  In the Atlantic, the calibration procedure was integrated in the measurement process. In the Pacific, two processes, calibration and measurements, were done separately. The calibration values obtained were averaged for each 3 day period and least square fit was performed on these data to obtain a linear representation of the calibration curve. The calibration curve for the PRT-5 radiometer used in the measurements of sea surface radiation in the Pacific, is shown in Figure 3.6. The accuracy of calibration is estimated to be ±0.25 * C . 3.4 Registration of Satellite and Ship D a t a The registration or navigation of satellite imagery involves coordinate conversion from satellite coordinates (line/pixel) to geocentric (earth-centered) coordinates (latitude/longitude) and then to geographic coordinates. The final displayable products are mapped into conical projection. The satellite positional coordinate system has its x-axis along the satellite velocity vector and its z-axis along the local vertical whereas the geocentric coordinate system has its x-axis aligned along the Greenwich meridian and its z-axis pointed at the north pole. The satellite orbit is estimated using the Brouwer Osculating orbital model with input parameters from the TBUS messages which describe the satellite orbit at epoch. The Brouwer osculating orbital model differs from the simple Keplerian (mean) orbital model by including higher order terms in the gravitational potential equation reft)  o d  0.0  !  i  j  0.1  0.2  0.3  V O L T A G E  0.4  0.5  C O U N T  Figure 3.6 Calibration curve for the PRT-5 used in the Pacific. This curve is derived from calibration data from two days  suiting from assuming the earth as an oblate spheroid. For each ship measurement (PRT-5 and bulk), time of observation is recorded as the reference. The ship's position was recorded every half an hour with the accuracy of a few meters. The intermediate positions are determined by calculating the great circle distance. For the Pacific data, the surface radiation was sampled every 2 minutes (every 5 minutes for the Atlantic data). For each sample, latitude/longitude was converted into a satellite scan line/pixel by using an iterative process. Iteration is necessary as the coordinate conversion process involves non-linear inversion of the coordinate transformation. The first guess of the satellite time is taken from the first archived AVHRR scan line. Given the number of lines stored and the AVHRR scan rate of 6 lines/sec, closer estimation of the satellite time for the desired line/pixel can be achieved. The satellite time is then corrected according to 51  the difference between desired and calculated latitude/longitudes and the iteration process continues until the along track angle is « 0. The satellite timing correction was accomplished by overlaying the high resolution coastal map onto the imagery and by performing linear shifts in x and y directions until the best fit was achieved. The second and higher order errors arising from the variations in the satellite attitude (yaw, roll, and pitch) were not corrected. The error estimated from using the Brouwer osculating model is less than using the simple Keplerian orbital model and is expected to be between 1 to Zkm (MDA, 1983). However, for a small local area, which is true for images used in this project, the resulting registration error should be much smaller (within 1 km). For each scan and pixel position, weighted averaging was performed to retrieve a digital brightness count which was then calibrated and converted into a brightness temperature value. The registration of HIRS foot prints and AVHRR imagery was accomplished by registering five HIRS scan line (56 spots per line) onto the AVHRR imagery which was used as the template. Five HIRS scan lines take up exactly 192 AVHRR scan lines. The template was then moved along the AVHRR imagery and each individual AVHRR pixel was tagged with a HIRS scan and pixel number. This tag was later used to identify the nearest or coincident HIRS spot along the ship's cruise track. Figure 3.7 to Figure 3.10 show AVHRR channels 2 and 4 images of orbit 15779, 15835, 17667 and 17681, respectively, with HIRS foot prints overlayed on channel 2 images. 3.S E r r o r Sources There are a number of ways in which errors are injected into the SST retrieval process. In this study, the main emphasis is on correcting the error introduced by the atmosphere through which the SST measurements are made. The additional error sources are identified in this section. Some sources are natural and are difficult to deal with while others are inherent and can be pre-determined through experiments.  52  Figure 3.7 AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 15779. Only boundaries of each HIRS spot is shown  Figure 3.8 AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 15835. A slight mismatch in navigation can be seen along the coastal boundaries 53  Figure 3.0 AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 17667. Almost all HIRS FOVS are contaminated by clouds  Figure 3.10 AVHRR channels 2 and 4 imagery with HIRS foot prints for orbit 17681. Southern Britain and northern French coastal lines are visible. 54  3.5.1  Radiometer The radiometer mainly used in this study, AVHRR, itself contributes to the increase  in error in two ways. One is from inherent sensor noise and the second one is quantization error introduced when the analog signal is converted into the 10 bit digital representation. Combined these errors can be represented in total by N E A T which is found to be 0.12 °K determined from the pre-launch laboratory experiments (Robinson et ol., 1984). The operational procedure of calibrating the AVHRR and HIRS systems can also introduce error from the several assumptions made.  A common assumption is that the  emissivity of the blackbody cavity, on-board the satellite, is unity. Another assumption is space is a known cold source. The error from these assumptions is estimated as 0.31 °K obtained by I T T Aerospace. The calibration procedure is based on a linear fit between two measurements of radiance taken above and below the temperature range of interest. For the AVHRR, the two temperature sources are the internal target ( » 288 °K) and space(» 3 ° K ) , respectively. However due to such factors as sensor nonlinearities, measurement of internal target temperature, and target emissivity, a linear fit is not adequate for temperatures below and above the typical oceanic temperature range, 270 — 305 °K. There have been several proposed ideas to correct for this deviation. The most extensive study on this deviation was performed by Brown et c/.(1985) and a method was proposed to improve the radiometer calibration accuracy to ± 2 ° C .  In this study, however, the  non-linearity error is not corrected as the published correction values are based on the prelaunch tests on the AVHRR sensors. NOAA-7 was launched on July, 1981 and sensors have drifted since then. Brown et al., (1985) pointed out that their non-linearity correction data is not valid over the instrument's service life. Considering the fact that NOAA-7 went through a temperature increase in October, 1984 in order to reduce the noise on the channel 3 of AVHRR, and that the correction data are temperature sensitive, no attempt was made to incorporate the non-linearity correction. Besides, the sea surface temperature range during two cruises was between 10 to 15 °C, and at this range the error caused by non-linearity is near zero according to the data tabulated by Brown et al., (1985). The 55  error figures for the HIRS channels are expected to be worse than for the AVHRR due to using the band correction method instead of integrating and convolving with the sensor response function.  3.5.2 Natural Sources If the FOV is obscured by an opaque cloud, IR detection of the surface temperature is impossible. However, if only a portion of the FOV is covered with sub-pixel sized clouds, it is still possible to estimate the surface parameters at the cost of decreased accuracy. Several algorithms to deal with this situation are available such as the histogram technique and N* by Smith (1971). The estimated error from the sub-pixel clouds is 0.2 — 0.5°K and it varies with the amount of cloud contamination. In this study, HIRS FOVs were checked for cloud contamination by comparing temperature measurements obtained by the HIRS window channels with those from the sounding channels. If the HIRS estimated surface temperature is as cold as the temperature around the tropopause, the F O V is considered to be cloud contaminated and the FOV is simply rejected. A more crucial problem is the cloud detection for the AVHRR/2 on a pixel by pixel basis along the cruise tracks. Again, no "declouding" (i.e. separation of clear pixels from partially or totally cloud contaminated ones and estimation of the amount of cloud contamination), is attempted. If a pixel/spot is detected as being cloudy it is rejected and no longer used. The cloud detection scheme used visible and infrared thresholds which were determined from histograms of all radiances within the area of interest (Olesen and Grassl, 1985). Thresholds chosen for each of the NOAA-7 passes are shown in Table 3.8. The variations in the %Albedo in the AVHRR channels 1 and 2 are primarily functions of the sensor scan angle and the solar zenith angle. Table 3.8  Thresholds used to detect cloud contamination Orbit %Albedoi %Albedo r 3 - T4(°C) r 5 (°c) 2  15779 15835 17667 17680 17681  > 3.19 > 4.70 > 0.70 > 4.00 > 0.70  > > > > >  1.65 3.60 0.80 3.00 0.80  < < < < <  8.0 8.0 8.0 8.0 8.0 56  < < < < <  -3.0 -3.0 -3.0 -3.0 -3.0  Solar contamination contributes to the error during daytime satellite passes. AVHRR channel 3 (3.7 pm) is especially susceptible to solar contamination (as large as 20 °K). This is the main reason why the dual channel method, which uses two IR channels, is not considered in this study even though channel 3 is most transparent among the 3 AVHRR infrared channels. The common practice of calibrating satellite SST measurements is by comparison with  in situ measurements taken by surface bucket sampling, engine intake sampling, or by Expendable Bathythermograph (XBT) which represent the temperature at tens of centimeters, to a few meters below the surface. A radiometer typically measures radiation emitted from no more than the top 0.1mm of the water and this skin temperature is typically a few tenths of a degree cooler than the temperature a few centimeters below due to the vertical heat flux through the air-sea interface. The difference can range from 0.1 to 0.5 °C depending on the environmental conditions surrounding the target. At some instances, difference of up to 3.4 °C has been observed (Robinson et al., 1984). The physical process involved in the surface skin temperature is a complex one as illustrated in Figure 3.11. Under most circumstances, the vertical net heat flux, Qs = Qi + QIR + QG + QH + QE, is positive and the temperature deviation AT is negative (i.e. the surface is cooler than the bulk temperature). If the atmospheric correction algorithm is based on the bulk temperature, or comparison is to be made with the bulk temperature, the thermal gradient existing below the surface can become an important factor. To eliminate the possibility of introducing AT error in assessing the performance of the atmospheric correction algorithm, a radiometer is used to obtain in situ skin temperature measurements.  57  Figure 3.11 Schematic of physical processes influencing the skin temperature. This diagram illustrates complexity involved with A T . (from Katsaros, 1980)  58  CHAPTER 4 R E S U L T S OF T H E C A S E S T U D I E S  In this chapter, the results of the iterative TOVS profile retrieval method, and the SST retrieval methods, using the TOVS atmospheric correction, are presented. In addition, results from two other SWT's are presented. All satellite derived results are compared with the spatially and temporally closest in situ measurements for validation. First, TOVS profiles from both the Pacific and the Atlantic Oceans are compared against RAOB measurements and second, the corrected satellite SST computed using several methods, are compared against the PRT-5 measured surface skin temperatures and bulk temperatures. A discussion of all the SST retrieval methods considered in this study, is presented.  4.1 Comparison of Vertical Profiles The physical SST retrieval method uses TOVS derived atmospheric temperature and moisture profiles to estimate the atmospheric transmission in AVHRR channels 4 and 5. However, before the transmission functions were calculated, TOVS profiles were compared against co-located radiosonde profiles in order to assess the accuracy of the TOVS retrievals. T O V S and R A O B profiles are not expected to be identical due to different characteristics in the sensors and their retrieval methods. These differences are as follows: (1) Radiosondes make direct point measurements of the atmospheric parameters whereas the T O V S makes inferred volumetric measurements from a distance. (2) Radiosondes take about an hour to reach the height where the atmospheric pressure is about 100 mb, whereas T O V S makes an instantaneous measurements. (3) Generally within the troposphere, radiosondes ascend vertically with respect to the surface of the earth whereas the TOVS make measurements from space at an angle except at nadir. 59  (4) Both the T O V S and the radiosondes exhibit different measurement error characteristics that are instrument dependent. These points should be kept in mind when considering the results presented. Balloon borne radiosonde data were used in this study to compare the results of T O V S soundings. In the Pacific, RAOB data were obtained from two locations (Quillayute, State of Washington and Port Hardy, B.C.) at two synoptic hours (0:00 and 12:00 GMT) from July 10 to July 18, 1984. During this period, the NOAA-7 daytime passes were between 23:00 G M T and 24:00 G M T . In order to minimize time discrepancies, only RAOB data taken at 0:00 G M T were used in the comparison. In the Atlantic, radiosondes were launched directly from the ship {RV Meteor) three times daily for November 25 and 26, 1984. The co-located HIRS spots were found by calculating the location of each HIRS spot (latitudes and longitudes), and finding a spot with a minimum distance from the location where the radiosondes were launched. For all passes, this location was within or near the closest single HIRS field of view making the distance between the T O V S and the RAOB measurements less than the diameter of single HIRS field of view. Time and location information of the T O V S spots are given in Table 4.1 and the similar information for the radiosonde observations are given in Table 4.2.  4.1.1 Radiance Synthesis The physical iterative retrieval method used in this study, involves two stages; iterative solution and weighting function calculation. The errors in the iteration stage depend on the quality of the feedback mechanism and whether convergence is achievable or not. The iteration stage assumes accurate weighting functions in order to achieve convergence. The iteration stage does not have direct control over the calculation of the weighting functions which is based on the physical properties of the atmosphere. Thus, the accuracy of the transmittance function must first be evaluated. The assumption made in the development of the empirical transmittance model, is that tabulated transmittance values are accurate and physically possible. Even though accuracies have been tested by several researchers 60  Table 4.1  Information on RAOB co-located TOVS spots T  Q  v  s  Orbit  Date  Time  Line  Spot  Latitude  Longitude  15722 15750 15779 15807 15821 15835 17667  10/07/84 12/07/84 14/07/84 16/07/84 17/07/84 18/07/84 25/11/84  23:00Z 22:35Z 23:51Z 23:26Z 23:14Z 23:02Z 16:20Z  17680  26/11/84  14:27Z  17681  26/11/84  16:07Z  35 30 9 40 27 24 17 16 14 7 7 8 12 11 11  20 8 50 38 30 21 49 50 52 3 4 6 51 53 54  47.76° N 47.76° N 47.76° N 47.76°N 47.76° N 47.76°N 46.64° N 47.25°N 47.93° N 49.31°N 49.81°N 50.01°N 49.31°N 49.81°N 50.01°N  124.55°W 124.55°W 124.55° W 124.55°W 124.55°W 124.55° W 7.61° W 6.76° W 5.66° W 3.40° W 2.00° W 0.50° W 3.40° W 2.00° W 0.50°W  Table 4.2  Information on radiosonde data RAOB Date  Time  Lat  Lon  11/07/84 13/07/84 15/07/84 17/07/84 18/07/84 19/07/84 25/11/84  0:00Z 0:00Z 0:00Z 0:00Z 0:00Z 0:00Z 9:55Z 15:53Z 21:53Z 9:48Z 15.-53Z 21:51Z 9:48Z 15:53Z 21:51Z  47.79°N 47.89°N 47.61°N 47.93° N 47.94°N 47.82°N 46.69°N 47.12° N 47.98°N 49.61° N 49.87°N 49.96°N 49.27° N 49.76°N 49.81°N  124.75°W 124.61°W 124.62° W 124.50° W 124.42° W 124.62° W 7.37°W 6.83° W 5.56°W 3.08° W 2.23° W 0.46° W 3.45° W 1.91°W 0.97°W  26/11/84 26/11/84  over spectral regions where experiments were conducted, a question still remains on how the model will behave for the TOVS sensors. A way of testing the transmittance model is to synthesize radiances for the TOVS sensor based on the radiosonde inputs.  This  process isolates the iterative stage of the T O V S retrieval scheme and only the transmittance 61  calculation stage is put to the test. The amount of ozone present over the fields of view was assumed to be fixed at a seasonal mean value, since no ozone data were available. Over the two carbon dioxide absorption bands (long and short wave), the effect of wings of the ozone absorption band is relatively weak and thus the impact of the above assumption should be negligible.  To synthesize radiances, the atmospheric transmittances based on the R A O B atmospheric profiles (temperature and moisture), are directly applied to the radiative transfer equation to solve for the surface term and the upwelling path radiance in order to obtain the expected radiance emerging at the top of the atmosphere. The calculated radiance value is then filtered by the spectral response of each channel. The final radiance value for each channel reflects the synthesized version of what the satellite sensors received and measured. The downwelling radiance is ignored as this term is relatively small for the clear FOVs selected.  In Table 4.3, observed and calculated HIRS brightness temperature values for the Pacific during the experiment period, are listed. There are no synthesized radiances for the HIRS visible and ozone channels available. In the last two columns, mean difference and standard deviation values are presented for the six measurements. In Table 4.4, MSU brightness temperature differences for the Pacific are shown. As other researchers (Hillger, 1984; Hayden et al., 1981) have experienced, all water vapor channels (channels 10, 11, and 12) showed large differences. There are two possible explanations. First, the vertical resolution of the HIRS water vapor channels is too poor resulting in large volumetric measurements made with HIRS being compared with RAOB point measurements. Second, the characteristics of the HIRS water vapor channels are such that even the most transparent channel does not sense the surface where the majority of moisture exists.  Another noticeable large difference occurred in the HIRS window channels. This problem is caused by the fact that the surface temperature measured by radiosondes represent surface air temperature which is above the surface skin water temperature which the 62  HIRS window channels measure. This explanation is confirmed by examining selected HIRS FOVs overlayed on AVHRR data. Another reason, for the large differences between RAOB and HIRS values, is that for all satellite passes considered, co-located HIRS FOVs happen to cover areas along the coastal boundaries. For most FOVs, the portion covered by the water area exceeded those over the land area. Thus the HIRS surface temperature represented a mixture of water and land with a bias towards the water temperature which is considerably cooler than the land temperature in summer. This reasoning is confirmed by examining MSU channel 1 which senses microwave radiation from the surface. The brightness temperature of MSU channel 1 is generally much larger than that synthesized from radiosonde data due to highly variable and larger emissivities originated from the land. For the remaining channels, the standard deviations show reasonable differences within the acceptable limits. For most channels, synthesized radiances showed warmer (negative value since differences are calculated from AT = Tobserved — T nthe»ized) than ay  the measured radiances.  Table 4.3 Synthesized and observed HIRS brightness temperature differences for Pacific HIRS Channel 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20  Wave no 10  12  July 14  16  1.36 667.92 2.27 679.21 0.93 691.56 0.31 704.63 717.05 0.10 -0.80 733.20 749.21 -2.03 898.94 -5.64 1027.38 1224.89 -5.40 -6.01 1363.85 1489.06 -12.89 •5.06 2183.05 -1.77 2208.28 -1.08 2239.84 1.65 2271.33 -3.02 2357.55 -2.33 2512.83 2663.79 -0.24 14453.10  1.11 0.75 -0.5 -0.65 -0.66 •1.48 •2.46 •8.04  -0.55 1.99 0.27 -0.34 •0.26 •1.37 •2.53 -7.54  -0.75 1.97 0.5 -0.17 0.18 -0.56 -0.32 •0.32  (cm- ) 1  •  •6.73 •2.66 2.28 -5.66 -2.53 •2.03 1.52 -0.95 6.07 14.74 *  •6.63 -2.32 •5.08 -8.75 0.77 -16.08 -5.70 -1.77 -2.18 -0.38 -1.90 -1.67 1.72 -0.32 -2.79 -4.16 2.65 •5.29 •5.03 3.13  -  -  (all figures are in ° K ) 63  17  18  Bias  a  0.88 1.31 0.56 1.81 2.64 1.91 0.34 1.08 0.44 0.36 0.48 -0.00 0.99 0.55 0.15 0.88 -0.32 -0.61 2.15 -0.77 -0.99 2.77 -3.00 -3.63  0.95 0.64 0.56 1.45 0.58 0.86 1.79 4.27  -  -  -  1.74 -2.71 •9.89 0.66 1.28 -0.39 1.39 -3.52 4.82 5.48  -3.02 -3.54 -4.94 -3.44 -1.07 -1.22 1.46 -2.03 -0.81 1.49  -3.73 -4.79 -6.79 -3.49 -1.11 -1.38 1.24 -2.75 0.96 3.26  -  -  -  -  3.24 2.35 7.43 2.54 1.40 0.61 0.77 1.13 4.36 6.65  -  Table 4.4  Synthesized and observed MSU brightness temperatures differences for Pacific MSU Frequency July Channel (GHz) 10 12 14 16 17 18 Bias o 1 2 3 4  50.33 53.78 55.00 57.99  15.52 -1.19 0.30 1.60  35.31 0.51 0.64 -0.82  20.49 49.25 -0.29 -0.89 0.41 1.19 0.40 2.74  14.65 -0.45 0.31 0.11  13.02 -1.08 0.22 1.05  24.71 -0.56 0.51 0.85  14.52 0.63 0.36 1.24  (all figures are in ° K )  In Table 4.5, synthesized HIRS radiances and in Table 4.6, synthesized MSU radiances based on the radiosonde data from the Atlantic are shown. For November 25th (17667), all HIRS spots were partially cloud contaminated and therefore have larger differences for channels whose weighting function peak below the tropopause where clouds are known to be concentrated at.  For November 26th (17680, 17681), the HIRS spots examined  were all cloud free and much smaller differences were observed. As in the Pacific, large differences persist for the water vapor channels. Even though the atmosphere is relatively free from concentrated moisture, the effect of broad weighting functions still permeates the result. Notice the smaller differences in the synthesized and observed radiance values for the HIRS window channels. The problem of comparing two different temperature sources (land and water) present in the Pacific data, no longer exists in the Atlantic data, as the radiosondes were launched from the ship. An explanation for the large differences in the surface radiances is due to the fact the radiosonde surface temperature represents air temperature whereas the HIRS window channel measures water surface skin temperature. The table also shows large differences for most channels with peak contributions above the tropopause. An explanation for these differences is that the highest level of moisture profile available from the radiosondes was around 400 mb. and the moisture content above this level was extrapolated. Obviously the extrapolation scheme underestimated the moisture content and resulted in a smaller amount of radiation than the actual value calculated for the upper atmosphere.  In Figures 4.1 and 4.2, scatter plots of brightness temperature versus temperature 64  Table 4.5  Synthesized and observed HIRS brightness temperature differences for Atlantic HIRS Channel 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20  17667  (16:20)  15:53  21:53  •8.91 -10.60 -7.98 -7.67 -7.47 -7.00 -5.78 -5.19 -4.36 -5.52 -3.76 -5.36 -1.59 -3.80 -1.10 -2.88  -10.38 -7.94 -7.06 -3.55 •1.93 -1.09 0.83 0.48  -0.87 -0.15 -2.12 •4.95 -4.20 -5.02 -3.79 -10.64 -0.31 1.33  0.54 -0.98 -2.17 -2.00 •0.45 •1.98 •2.96 •11.38 1.85 3.12  9:55  -3.21 -5.70 •10.07 -6.75 •5.99 -6.05 •2.97 -7.97 -0.99 1.41  •  17680  (14:27)  time 15:53  21:51  RAOB 9:48  -9.87 -11.70 -7.71 •6.88 -7.60 -7.72 -5.81 -7.59 -5.08 -3.97 •2.69 •3.35 •0.39 -0.32 0.40 0.16 0.69 1.41 -2.93 •3.05 •3.05 -5.73 -6.40 -12.07 2.24 4.70  0.32 1.13 -0.24 •2.41 -2.12 -4.60 -5.29 -13.31 2.27 3.49  17681 (16:07) 9:48  15:53  21:51  -11.21 -11.60 -7.46 -7.51 -5.55 -8.29 -2.46 -6.56 -1.72 -3.15 -0.88 -1.50 1.49 0.53 0.89 -0.12  -13.14 -7.98 -8.58 -5.91 -3.62 -2.26 0.10 0.48  -11.72 -8.27 -6.40 -4.54 -4.10 -3.56 -0.76 0.01  1.59 0.68 4.36 -0.09 7.15 -5.66 -0.94 -1.37 -0.26 -0.46 -2.51 -3.45 -2.07 -10.02 -12.42 -16.09 2.84 1.38 4.16 1.71  0.64 0.54 -2.48 -1.96 -1.98 -4.59 -8.29 -16.66 1.82 1.85  0.48 2.67 4.92 -3.23 -3.86 -6.10 -6.81 -13.89 1.30 1.61  (all figures are in ° K ) Table 4.6  Synthesized and observed MSU brightness temperature differences for Atlantic MSU Channel 9:55 1 2 3 4  26.05 -7.30 -6.15 -6.22  17667  (16:20)  15:53  RAOB 21:53  17.77 -3.35 •2.85 -5.18  24.54 •1.64 •2.16 -3.83  17680  (14:27)  time 9:48  15:53  21:51  9:48  15:53  21:51  22.22 -7.77 •9.32 •4.36  13.90 •3.22 •3.33 •3.38  14.87 -1.16 1.87 -1.66  14.10 -4.26 -7.31 -4.69  16.06 -7.28 -6.41 -4.07  11.82 -4.75 -1.02 -2.86  17681 (16:07)  (all figures are in ° K ) difference are shown. In general, points on these graphs are scattered and don't show any significant definitive temperature dependent relationship between two variables. Possible exception is with channel 5 which peaks around the tropopause. The lack of relationship in these plots indicate that the biases in the channel brightness temperatures, between HIRS and the synthesized values, are not systematic but are random in nature.  This  result suggests that the differences are primarily due to sensor noise rather than error 65  in the transmittance model. The scatter in the y-direction in some of the channels are caused by the ill-posed nature of remotely sensed radiation measurements (i.e. there can be an infinite ways of forming the transmittance function with the same radiance value). Notice the large variations in AT observed from the HIRS water vapor channels suggesting inconsistencies in the water vapor measurements by HIRS.  4.1.2 Retrieval  Procedure  The atmospheric temperature and moisture profiles were retrieved from HIRS and MSU radiances using the physical iterative method described in chapter 2. A flow diagram of the procedure is shown in Figure 4.3. First, HIRS and MSU radiances coincident with a radiosonde profile, were retrieved from a file. The geographical position of the selected spot (latitude/longi-tude), and the date and the time information were used to select a initial guess temperature profile from the stored climatological mean profiles. The retrieved temperature and moisture profiles are limb corrected in order to minimize the spatial registration error between two profiles. If the coincident MSU spot was suspected of being contaminated by the radiances originating from land, then the microwave retrieval was bypassed. Often MSU data is useful when the F O V is partially or fully obscured by clouds. In this study, only clear FOVs were considered and the purpose of temperature profile retrieval based on the MSU radiance data, is to improve the initial guess profile to the next stage where the temperature profiles are retrieved based on the HIRS radiance data. If MSU data contaminated by land is included, the warmer land surface temperature detected by the MSU will affect the retrieval process resulting in warmer surface temperature and colder temperature above the surface. The water vapor profile was first retrieved based on the initial temperature and the water vapor guess profiles. The resulting water vapor profile was then used to retrieve the temperature profile until the iteration converged. Then the final water vapor profile was retrieved using an iterative process based on the temperature profile just obtained. 66  o  CH  o d  7  o -i  CJ  LO "  °.  E—•  O  az  CD ~  _ J  O  O 7  E—'  E—  1  LJ  LD I  O  IT) ~  1  o  d  O  200.0  250.0  200.0  300.0  CH  4  o  d °  o o  a  r~i i n i  o  o 200.0  BRIGHT T(C  300.0  o  O  oo o o oo  E—<  _j o LJ  o  a  1  o  d 200.0  o  Lf)  CJ  O  300.0  250.0  BRIGHT T(C) o d  CH  6  -i  LD  O  CH d ~  ~  t— 1 ]  O  LJ  L D ~~  a  |  O  d 250.0  LD  a: d  £-<  EL' TR  or ° '  E—•  E->  5  •>—' O  ^ °  10  •—'  300.0  250.0  CH  d  o  o  BRIGHT T(C)  BRIGHT T(C) d  °  CD  O  O  CH  o  o  I  o o•  8  o  ^ CU o ' O  CH  O  in ~  1  o d 200.0  250.0  BRIGHT T(C)  300.0  200.0  250.0  BRIGHT T(C)  Figure 4.1 Scatter plots of brightness temperatures vs. differences (a), (a) Channel 4, (b) Channel 5, (c) Channel 6, (d) Channel 7, (e) Channel 8, and (f) Channel 10  300.0  o  CH  CD "I  o  14  CH  o  15  CH  o'  16  o n o  o o  CJ  CJ  o o  t- o  o1  d °  r~i u i ' i  cn  o CE  o"  d °  a  o  o  f- o  o  CD  200.0  250.0  BRIGHT  s  o o •  300.0  o  ^  cn r~i  o o  ^ '  •>—' o  o in I  °o°  E— ]o  c9  a  o o i  CD ~  o o  200.0  250.0  BRIGHT  T(C)  1  LLJ  I 1  I I 300.0  200.0  250.0  BRIGHT o CD  CH  -i  300.0  T(C)  13  CJ LT) "~ E-" O  Q  250.0  T(C)  —1 • 300.0  I  o  O  9DQ  o  o  o p  BRIGH^  O  CE CD " fr-< ]O LJ Q m ~ o  o o  CD O  200.0  o  ip  LD ~  300.0  12  o o o  t-l CE  o  -e—  T(C.  o o  o m ~  CJ  °  250.0  CH  o -  o CJ  200.0  BRIGHT  11  o CP  CD  T(C)  CH  O  o  CD LD ' o  LO" I  I  ^ '  200.0  250.0  BRIGHT  T(C  Figure 4.2 Scatter plots of brightness temperatures vs. differences (b). (a) Channel 11, (b) Channel 12, (c) Channel 13, (d) Channel 14, (e) Channel 15, and (f) Channel 16  300.0  MSU  BR Temp  HIRS UU BR  DSD FOU c o n t a m i n a t e d Temp  ui th  Water U s p c r Profi le E s t i mat i on  I*mp*rstufs uith  HSU BP Temp Initial  1 and ?  B y p a s s PIU Temp R e t r i e v a l  profile  n s u  T Prof i 1 e Residual  < delta New  HIRS WO  BR  and S h o r t BR Temp  HIRS Window_ C nanne 1 s BR Tern*  7  Initial  T Pro-file  Water Uapor Prof i1e E s t i mat i o n  Temp  New  Long HIRS  yes  Uaue  WW  L_A  Prof i 1 e  L  Iterstive Temperature Profile Retrieval  "fc** S u r f a c e Tern p e r a t u r e  Surface Temper at ur-2 E s U m a t i on Residual  < delta  ? "^"Temperature  HIRS UU Channe I s BR Temp  Pro-file  1t e r a t i v e Wat e r U a p o r Pr o f il« R e t r i ewal  Residual  < delta  ?  Water  Figure 4.3 The flow diagram of the TOVS profile retrieval procedure. 69  Vapor  Profile  4.1.3 ProGles  Comparison  For temperature profiles, temperature values are calculated at 40 atmospheric levels ranging from 0.1 mb to lOOOmft plus the surface temperature. For moisture profiles, retrieval atmospheric pressure levels range from 300m6 to 1000m6, plus the surface moisture value. The output for the moisture is expressed as "dew point depression" which is defined as  T — To where To represents the dew point temperature. In Figures 4.4, 4.6, 4.8, 4.10, and 4.12 TOVS retrieved temperature profiles are presented along with radiosonde profiles for 2 days in July from the Pacific and 2 days in November (3 passes) from the Atlantic. Also shown on the same figures are dew point profiles from TOVS and the radiosondes. A set of weighting functions for each pass is shown in Figures 4.5, 4.7, 4.9, 4.11, and 4.13 below each profile to give a rough idea about the vertical averaging done by HIRS channels during the retrieval. From these figures preliminary appraisal of the performance of the TOVS retrieval scheme can be made. For obvious reasons, TOVS retrievals on clear days match closer to the corresponding radiosonde retrievals. The retrievals over the Atlantic winter atmosphere show better results compared with the Pacific summer atmospheric retrievals. For both the Pacific and Atlantic data, temperature profiles seem to fit well with the corresponding RAOB profiles. For the moisture profiles of July 14, a moist layer near the surface can be seen with a dry layer on top of it. On July 18, similar conditions existed with the exception that the moist layer was sandwiched between dry layers indicating an instability. Looking at the T O V S retrieved moisture profiles, on those two days, it can be clearly seen that small scale moisture fluctuations in the atmosphere have been completely missed by TOVS, and instead, moisture profiles which seem to represent an averaged value over a thickness are observed. In the Atlantic, the profiles for November 25 are cloud contaminated which is indicated by the 100% humidity at 750 mb level in the radiosonde profiles. The TOVS profiles better represented the state of the atmosphere for November 26. The atmosphere was relatively dry and stable without exhibiting any small scale fluctuations. Under this condition TOVS seems to have performed well. Temperature differences between the TOVS retrieved and the radiosonde profiles are 70  ATMOSPHERIC PROFILES TOVS DRTE : 840714  Figure 4 . 4 Temperature and dew point profiles for NOAA-7 15779. Radiosonde data from July 13, 0:00Z N7-15779 HIRS LONGURVE 0.0  HIRS SHORTUfWE  1.0  HIRS UflTER VfiPOR OHO SURFRCE  1.0  0.0  Figure 4 . 6 HIRS weighting functions for 15779. 71  0.0  1.0  ATMOSPHERIC PROFILES  Figure 4.0 Temperature and dew point profiles for NOAA-7 15835. Radiosonde data is from July 19, 0:00Z N7-15835 HIRS LONGURVE 0.0  HIRS SHORTURVE 1.0  NORMALIZED WEIGHT  0.0  HIRS UflTER VAPOR P.ND SURFACE 1.0  NORMALIZED WEIGHT  Figure 4.7 HIRS weighting functions for 15835. 72  0.0  1.0  NORMALIZED WEIGHT  Figure 4 . 8 Temperature and dew point profiles for NOAA-7 17667. Radiosonde data from November 25, at 15:53Z. N7-17667 HIRS  LONGUOVE  HIRS  SHORTUPVE  Figure 4 . 0 HIRS weighting functions for 17667. 73  HIRS  UB1ER  VfiPOR  -Z5 d  DEW P01-N TICI 20.0  ATMOSPHERIC PROFILES TOVS DATE 841126 START TIME 142713 END TIME 143025 LATITUDE 49.87 LONGITUDE -2.23 RAOB DATE 841126 RAOB TIME 1553  N7-I7680  SOLID - TOVS DASH - RAOB  TEMPERATURE  TEMPERRTUREICI  Figure 4.10 Temperature and dew point profiles for N O A A - 7 17680. Radiosonde data from November 26, at 15:53Z. N7-17680 HIRS  LONGUP.VE  HIRS  SHORTURVE  HIRS RND  "NORMALIZED WEIGHT  0  "NORMALIZED WEIGHT'"  Figure 4.11 HIRS weighting functions for 17680. 74  URTER  VAPOR  SURFBCE  "NORMALIZED WEIGHT'"  Figure 4.12 Temperature and dew point profiles for NOAA-7 17681. Radiosonde data from November 26, at 15:53Z. N7-17681 HIRS  LONGURVE  HIRS  SHORTURVE  Figure 4 . 1 3 HIRS weighting functions for 17681. 75  HIRS  UfiTER  VRPOR  shown in Table 4.7 for all six days of observations in the Pacific; the last two columns show the means and the standard deviations. Table 4.8 is same as Table 4.7, but for the dew point differences. Mean temperature differences indicate on overall temperature bias in the T O V S retrieved profiles with the negative values indicating a cold bias in the TOVS retrieved profile. Again, the large differences in the surface temperatures are due to the fact that the HIRS measured surface skin temperature whereas the radiosonde measured surface air temperatures over land. Table 4.7 shows warm biases below the tropopause, larger warmer biases surrounding the tropopause, and warm biases above it. The tropopause is characterized by an abrupt temperature inversion. Because this sudden change cannot be resolved by the broad HIRS weighting functions the retrieved temperature, which represents temperature around the tropopause, appears warmer than the actual one.  Table 4.7 Differences in temperature pro&les for Paci&c July Pressure {mb) 10 12 14  17  18  Bias  a  2.46 4.23 2.72 2.27 0.33 0.32 -0.50 -1.14 -0.21 0.44 0.77 1.46 0.39 3.25 3.60 3.88 2.99 2.43 1.53 1.15 9.37 1.75  1.54 3.65 3.09 1.77 -0.30 1.74 1.99 1.31 0.86 0.30 1.59 2.31 1.79 2.35 2.38 2.35 1.71 2.23 2.74 9.82 9.56 2.17  1.45 2.88 2.43 0.45 -3.55 4.41 2.52 2.20 2.20 1.49 0.96 0.25 -1.13 -1.40 -0.95 -1.05 1.32 0.85 3.68 7.92 6.23 -4.54  1.11 2.11 1.22 0.47 1.41 3.08 1.74 0.81 0.41 -0.25 -0.23 -0.46 -1.19 -0.44 -0.21 -0.21 0.30 2.04 2.69 4.80 6.38 -3.16  1.40 1.91 1.55 1.17 3.37 1.80 1.14 1.11 1.26 1.51 1.95 2.68 2.83 3.57 3.88 4.34 3.85 3.96 3.27 3.92 2.52 3.88  100.00 115.00 135.00 150.00 200.00 250.00 300.00 350.00 400.00 430.00 475.00 500.00 570.00 620.00 670.00 700.00 780.00 850.00 920.00 950.00 1000.00 1011.00  -1.85 -1.37 -0.71 -0.33 7.07 2.31 1.22 0.92 0.93 -0.13 0.45 0.71 1.36 2.82 3.87 4.80 4.82 9.94 8.38 7.83 6.46 -6.37  2.08 2.62 0.07 -0.45 4.06 5.71 2.13 -0.03 •1.91 -3.38 -4.25 -5.77 -6.27 -6.43 -6.60 -7.27 -6.96 -2.92 -2.72 -0.78 3.90 -3.81  1.00 0.65 -0.27 •0.86 0.85 3.99 3.06 1.62 0.61 -0.24 •0.92 -1.70 -3.27 -3.24 -3.57 •3.99 -2.08 -0.32 2.53 2.89 2.78 -8.18  (all figures are in ° C )  76  Table 4.8  Differences in dew point profiles for Pacific July Pressure (mb) 10 12 14 300.00 350.00 400.00 430.00 475.00 500.00 570.00 620.00 670.00 700.00 780.00 850.00 920.00 950.00 1000.00 1011.00  2.26 2.32 6.64 13.07 7.61 0.39 1.67 -4.05 19.31 -4.78 11.50 -1.03 -2.87 -2.41 -4.18 -0.00  5.01 3.40 1.66 1.59 -0.12 14.13 13.34 12.96 12.53 11.73 11.47 0.89 -5.01 -9.26 -9.87 0.14  4.88 3.21 4.24 2.74 2.53 2.09 4.50 7.97 10.88 12.36 13.71 14.95 12.55 8.43 3.33 -7.95  16  17  18  Bias  a  0.44 -0.02 2.53 1.40 1.34 1.51 -0.68 10.21 10.27 11.01 10.70 10.53 8.45 8.55 -3.49 -0.00  2.95 2.57 3.51 13.55 13.74 13.62 7.77 4.84 1.85 0.64 11.52 12.54 12.01 1.43 -2.25 0.00  4.43 2.74 2.63 8.01 14.79 18.30 16.14 15.36 15.49 15.46 -3.29 8.74 15.59 -0.79 -0.25 0.74  3.33 2.37 3.54 6.73 6.65 8.34 7.12 7.88 11.72 7.74 9.27 7.77 6.79 0.99 -2.79 -1.18  1.64 1.13 1.61 5.15 5.89 7.18 6.03 6.31 5.37 7.24 5.69 5.88 7.89 6.22 4.02 3.04  (all figures are in °C)  In Table 4.9, temperature differences between the TOVS profiles and those from the radiosondes in the Atlantic are shown; dew point differences are in Table 4.10. The values in these tables show that the T O V S profiler produced better results for the Atlantic data than for the Pacific. The temperature differences at all levels are less than 1.92 ° C and the dew point differences are less than 2.38 °C. This is an improvement over what earlier studies had found (Moyer tt al., 1979; Philips et al., 1979; and Hayden et al., 1981). From the radiosonde profiles, one can see that the atmospheric conditions for the Pacific, appear more complex than those observed in the Atlantic. This is primarily due to seasonal variations in the atmosphere. The summer Pacific atmosphere is characterized by an instability near the surface, colder tropopause temperature, and a larger temperature lapse rate indicating much more dynamic condition of the atmosphere in the summer time. On the other hand, the winter Atlantic atmosphere is characterized by a smaller lapse rate with no sudden instabilities. Both the temperature and the moisture profiles are smooth enough for the TOVS to sense the actual atmospheric conditions.  Simple mean and standard deviation statistics between the TOVS retrieved and the 77  Table 4.0  Differences in temperature profiles for Atlantic Pressure (mb)  17667  Orbit 17680  17681  100.00 115.00 135.00 150.00 200.00 250.00 300.00 350.00 400.00 430.00 475.00 500.00 570.00 620.00 670.00 700.00 780.00 850.00 920.00 950.00 1000.00 1016.00  -0.45 -1.47 •2.30 -3.23 -6.92 •2.89 -0.39 0.34 0.51 -0.24 -1.53 •2.37 -2.63 -1.35 •2.53 -2.46 -3.30 -4.17 •5.93 -6.60 -7.41 -5.56  -6.91 -7.80 -0.00 -7.77 -8.76 -7.94 -4.15 -4.07 -2.40 -0.47 -0.80 -1.65 -3.35 •1.82 0.34 1.42 1.48 0.35 -0.53 -0.63 -2.42 -3.52  -7.49 •8.64 0.09 -8.94 •9.99 -8.91 -4.97 -4.71 -2.88 -0.52 -0.87 -1.22 -3.10 •1.54 0.45 1.23 0.75 -1.16 -2.73 -3.43 -5.38 1.68  Bias  a  -4.95 3.19 -5.97 3.20 -0.74 1.11 -6.65 2.46 -8.56 1.26 -6.58 2.64 -3.17 1.99 -2.81 2.24 -1.59 1.50 -0.41 0.12 -1.07 0.33 •1.75 0.47 -3.03 0.30 •1.57 0.19 -0.58 1.38 0.06 1.79 -0.36 2.10 -1.66 1.88 -3.06 2.22 -3.55 2.44 -5.07 2.05 -2.47 3.05  (all figures are in °C)  Table 4.10  Differences in dew point profiles for Atlantic Pressure (mb)  17667  Orbit 17680  17681  Bias  a  300.00 350.00 400.00 430.00 475.00 500.00 570.00 620.00 670.00 700.00 780.00 850.00 920.00 950.00 1000.00 1016.00  2.03 1.62 2.19 2.90 2.52 1.79 4.01 4.36 1.46 -1.35 -11.57 •14.30 -16.05 -15.79 -15.29 -13.27  -2.65 -3.87 -3.33 -1.39 2.99 3.07 0.81 2.05 2.80 4.41 1.02 -0.44 -3.85 -5.77 -9.11 -6.28  -3.01 -4.21 •3.48 •1.09 3.33 3.94 1.70 3.12 3.85 5.32 1.61 -0.45 •4.40 -6.80 -6.95 0.98  •1.21 -2.15 •1.54 0.14 2.95 2.93 2.17 3.18 2.70 2.79 -2.98 -5.06 -8.10 -9.45 -10.45 -6.19  2.30 2.67 2.64 1.96 0.33 0.88 1.35 0.94 0.98 2.95 6.08 6.53 5.63 4.50 3.53 5.82  (all figures are in °C)  78  radiosonde profiles, show point differences which include errors associated with the radiosonde measurements. Since we are more interested in the overall accuracy of the T O V S retrievals, comparisons in terms of vertically averaged values at various thicknesses offer additional information about the accuracy of the retrievals. In Table 4.11, averaged RMS temperature differences at various thicknesses for pressure levels between 100m6 and the surface, are shown. The bottom half of the table is obtained by applying the TOVS retrieval process with the radiance inputs synthesized from the coincident radiosonde data. The comparison between the top and the bottom figures in the table shows the limitation of the iterative T O V S retrieval scheme. In most cases considered, RMS differences for the TOVS retrievals were better and showed less variations than those from the synthesized values. Generally, the RMS differences tend to improve when vertically averaged values are looked at rather than values at each discrete pressure levels. In Table 4.12, RMS dew point differences are presented in the same format. The data set used for Tables 4.11 and 4.12 is from the NO A A orbit 15779 on July 14. In Tables 4.13 and 4.14, RMS temperature and RMS dew point differences, respectively, for the Atlantic data (17681 on Nov 26) are shown. From the numbers in the above four tables, the uncertainty of the iterative scheme is seen for the Pacific moisture profile retrieval. In the Atlantic moisture profile, significant reduction in the RMS differences are seen when thicker atmospheric levels are considered. From the figures in the tables for both the Atlantic and the Pacific, the ability of the TOVS system to retrieve temperature profiles is very encouraging. One way of reducing overall error in the iterative process is to determine the systematic deviations. However, the number of available radiosonde and T O V S data, obtained during this study, is too small to assess any systematic errors in the iterative process.  The differences between the TOVS retrieved and the radiosonde profiles for the Pacific are plotted in Figure 4.14, along with results obtained from the NOAA/NESS's Export routine. The Export routine is a atmospheric profile retrieval algorithm developed by Smith et oi.(1983) based on the earlier paper by Smith (1970). It uses an iterative method to retrieve a temperature profile, but the moisture profile is derived directly from the 79  Table 4.11 Vertical// averaged RMS temperature differences for 15779 P(mb)  950  850  700  620  500  400  300  200  100  1.8 0.4 0.5 1.2 1.2 0.6 0.4  2.6 2.0 0.9 0.1 0.6 0.7 0.2 0.1  0.3 1.2 1.2 0.6 0.1 0.4 0.5 0.2 0.0  4.0 3.7 3.2 2.6 2.5 2.7 2.8  3.5 3.6 3.5 3.1 2.7 2.6 2.8 2.8  1.0 0.6 1.3 1.8 1.8 1.6 1.7 1.9 2.0  TOVS 200 300 400 500 620 700 850 950 1000  2.8  1.7 2.0  2.1 0.2 0.3  3.6 2.6 1.1 0.6  2.7 3.2 2.6 1.4 1.0  0.6 1.5 2.0 1.9 1.1 0.8  Synthesized 200 300 400 500 620 700 850 950 1000  3.7  3.5 3.6  1.3 2.4 2.7  0.8 1.3 2.1 2.3  2.2 1.5 1.6 2.2 2.3  3.4 2.8 2.2 2.2 2.5 2.6  (all figures are in °C)  Table 4.12  Vertically averaged RMS dew point differences for 15779 P(mb)  950  850  700  620  500  400  14.0 14.6 12.7 11.1  7.6 10.7 12.1 11.1 10.1  3.5 5.5 8.0 9.5 9.2 8.5  5.0 7.5 7.4 7.0  1.3 1.8 4.5 5.1 5.0  6.9 4.2 1.6 1.0 2.0 2.2  TOVS 500 620 700 850 950 1000  3.0  10.3 7.8  15.8 12.6 10.6  Synthesized 500 620 700 850 950 1000  4.7  8.9 7.7  10.1 9.0 8.2  (all figures are in °C) HIRS water vapor channels. The dew point differences for the two methods are shown in Figure 4.15. The temperature and dew point differences obtained by both methods for 80  Table 4.13  Vertically averaged RMS temperature differences for 17681  P(mb)  950  850  700  620  500  400  300  200  100  4.2 2.5 2.5 1.8 1.5 1.8 2.0  8.0 6.3 4.3 3.9 3.1 2.7 2.7 2.9  9.0 8.4 7.4 5.8 5.3 4.5 4.0 3.9 4.0  2.9 4.2 3.8 4.0 3.9 3.2 2.8  0.5 1.5 3.0 2.9 3.2 3.2 2.7 2.4  0.2 0.1 0.8 1.9 2.0 2.4 2.5 2.2 1.9  TOVS 200 300 400 500 620 700 850 950 1000  4.4  2.4 3.2  0.3 1.1 1.8  0.0 0.1 0.9 1.5  2.0 0.8 0.7 1.2 1.6  1.4 1.7 1.1 0.9 1.3 1.6  Svnthestzed 200 300 400 500 620 700 850 950 1000  2.3  0.3 0.6  3.8 2.0 1.2  4.2 3.8 2.6 1.8  3.4 3.9 3.7 2.8 2.2  5.0 4.2 4.4 4.1 3.3 2.8  (all figures are in ° C )  Table 4.14  Vertically averaged RMS dew point differences for 17681 P(mb)  950  850  700  620  500  400  300  4.9 4.4 3.4 2.1 1.7  2.0 2.9 3.1 2.7 1.8 1.5  0.6 1.8 2.5 2.8 2.4 1.7 1.5  4.9 4.7 3.9 3.3 2.9  6.6 5.9 5.5 4.8 4.2 3.8  5.2 5.9 5.5 5.3 4.8 4.2 3.9  TOVS 400 500 620 700 850 950 1000  2.5  1.4 1.5  1.9 0.1 0.2  4.1 2.7 1.2 0.9  Synthesized 400 500 620 700 850 950 1000  0.5  0.9 0.6  2.5 1.9 1.5  4.6 3.6 2.8 2.4  (all figures are in ° C ) the Atlantic, are shown in Figures 4.16 and 4.17. When compared to the results obtained 81  from the modified iterative processes used in this study, the Export routine produced comparable temperature profiles with generally colder biases. Significantly Large biases were observed with the Export routine in the dew point retrievals especially below the tropopause where the most of the atmospheric moistures resides. The RMS differences between the two methods were comparable for the temperature profile, however, for the dew point profile, less RMS differences have been obtained by the method developed in this study.  In Figures 4.18 and 4.19, plots of total precipitable water vapor from TOVS are compared with coincident radiosonde retrievals for the Pacific and the Atlantic, respectively. As seen in the Atlantic data the differences are small except for the orbit 17667 which was cloud contaminated. However, for the Pacific data, differences are significant as the T O V S measured precipitable water vapor amounts were as much as the twice the amounts measured by the radiosondes. During the Pacific summer cruise, radiosonde profiles revealed moist air near the surface with dry air on top of it. The change was too abrupt for the TOVS sensors to resolve. In order to compensate for this change, the TOVS profiler overestimated the moisture amounts at all levels of the atmosphere. In order to examine the effect of of water vapor in the atmospheric transmittance, the relationship between the amount of total water and the transmittance were closely examined (Figure 4.20 for channel 4 and Figure 4.21 for channel 5). An examination of these figures revealed certain linear relationships between two values. This suggests that even though TOVS may not be able to duplicate the atmospheric profiles obtainable by radiosondes, especially the moisture profiles, T O V S can still be used to estimate the total moisture amount reasonably well to which the transmittances in the IR window region are most sensitive. This implies the possibility of making atmospheric correction to surface temperatures, using TOVS derived atmospheric profiles.  The effect of total water on the surface temperature measurable by AVHRR channels 4 and 5 are plotted in Figure 4.22. It can be shown that the absorption due to atmospheric 82  LU  in  ee  Tr-  OELTfi T ( C )  -10.0  ZD  ce  ce  UJ  LU  CL  CL  LU ce  ce  ce  CL  DELTP  z:  ce LT) cn^ •  CL  -10.0  CO  UJ  T(C)  -10.0  OELTfi T ( C )  ce  -10.0  OELTfi T ( C !  -10.0  DELTA T I C )  -10.0  DELTA T I C !  Figure 4.14 Temperature differences from the two TOVS retrieval methods (Pacific). Dashed curves represent results from the Export routine and the solid curves represent results from the modified routine.  Figure 4.15 Dew point differences from the two TOVS retrieval methods (Pacific). Dashed curves represent results from the Export routine and the solid curves represent results from the modified routine. 83  .7667  17680  CO  1768)  CO  21  LU  rr  cc  ZD  ZD CO  CO  co^r •  UJ  LU  cr  CL U3 •  -10.0  DELTR  -10.0  T(C)  -10.0  DELTR  T(C:  DELTR  T(C:  Figure 4.16 Temperature differences from the two T O V S retrieval methods (Atlantic), (solid curve • modified version; dashed curve - Export routine).  17667  17680  -10.0  -10.0  -10.0  DELTR  DELTR  T(C  17681  T(C:  DELTR  T(C  Figure 4.17 Dew point differences from the two TOVS retrieval methods (Atlantic), (solid curve - modified version; dashed curve - Export routine). 84  Figure 4.18 Comparison of precipitable water vapor amounts in the Pacific. Differences are large starting from day 3.  Figure 4.10 Comparison of precipitable water vapor amounts in the Atlantic. The large difference in the first satellite pass is due to cloud contamination in the F O V . 85  4  CH  o. 0-5  1.0  1.5  TOTAL NRTERICM; Figure 4.20 Plot  of precipitable water vapor vs. channel 4 transmittance.  CH  0.5  5  1.0  T0TRL Figure 4.21 Plot  2.0  1.5  HRTER(Cn  2.0  of precipitable; water vapor vs. channel 5 transmittance. 86  water vapor is more severe in channel 5 compared to channel 4. The third and fourth curves show the susceptibility of the SWT to the water vapor absorption by taking advantage of the differential absorption characteristics of two AVHRR channels.  • o A + =  LEGEND CHANNEL 4 CHANNEL 5 SWT-TEMP SWT-DP  CO  UJ  en  o.  0.0  0.5  1.0  TOTAL  HATER(CM)  1.5  2.0  Figure 4.22 The effect of water vapor amounts in the atmosphere, on SST. A significant reduction in error can be made by applying the SWT.  4.1.4 Sensitivity Test In order to determine the effects of errors in the temperature and moisture profiles, a sensitivity test was carried out. After the retrieval of the atmospheric profiles for orbit 15779, first the temperature and then the moisture contents at each level were systematically varied by increments of ±2 °C. The effects of these variations to the transmittance on AVHRR channels 4 and 5 are shown in Figures 4.23 and 4.24 along with a curve representing the total amount of water in the atmosphere. As is apparent the effect of moisture is more pronounced than the effect of temperature and is more severe in AVHRR channel 5 than in channel 4. From the profile comparisons, the inadequacy of HIRS in retrieving atmospheric moisture profiles, is clear, especially when the retrieval process overestimated 87  the atmospheric moisture. This is evident in the figures, as under-estimating by 10 ° C causes » 4% error whereas overestimating by 10 ° C causes as much as 10% error. Typically, the RMS deviation of temperature at all levels is less than 2.0 - 3.0 ° C depending on the retrieval method, this amount of error is tolerated by the SWT. The RMS deviation for the dew point, however, is much greater (7.0 - 8.0 °C). This large deviation has a greater impact on the accuracy of the SWT. It appears that this deviation cannot be further improved with the present HIRS water vapor channels.  The effects of errors in the temperature and moisture profiles can be serious to the AVHRR transmittances. It was shown earlier (Figure 4.22) that the SWT can better estimate the amount of absorption than any single channel. In Figure 4.25, the effect in profile retrieval error on 7 found by equation (2.16) is shown. It is interesting to note that the accuracy of temperature profiles is more significant in a dry atmosphere. In contrast, as moisture builds up in the atmosphere, the accuracy in moisture profiles has stronger effect on the transmittances than the accuracy of temperature profiles. The result of varying errors in the atmospheric profiles on the retrieved SST, is shown in Figure 4.26. It is surprising to note that even biases of ± 1 0 ° C to temperature and moisture profiles result in only 0.2 ° C errors in the final SST when the SWT is used. This result is very hopeful for the physical technique of retrieving SST.  88  Figure 4.23 Sensitivity test on channel 4 transmittance. Temperature and dew point at each pressure level were varied at an increment of ± 2 ° C .  1  -10.0 -8.0  1  1  -6.0  -4.0  1  1  1  -2.0 0.0 2.0 DELTA T ( C )  1  1  4.0  6.0  i  8.0  o o  10.0  Figure 4.24 Sensitivity test on channel 5 transmittance. The test was conducted under same conditions as for channel 4. 89  Figure 4.25 The effect of profile errors on 7.  LEGEND • - TEMPERATURE o = DEW POINT  - 2 . 0  0 . 0  DELTA  2 . 0  4 . 0  6 . 0  8 . 0  10.0  TIC)  Figure 4.26 The effect of profile errors on SST. The effect is small in the final SST compared to the magnitude of errors from other sources. 90  4.2 Sea Surface Temperature Retrieval Four split window satellite SST retrieval methods are used; NOAA's SWT (McClain, 1985), AVHRR-alone, AVHRR+HIRS, and AVHRR +TOVS. In addition, results from two single channel correction methods are also presented for comparison. AVHRR-alone SST values are calculated using equation (2.20), AVHRR+HIRS SST values are calculated from equation (2.22), and AVHRR+TOVS SST retrievals involve calculation of the atmospheric transmittances for each AVHRR F O V and application of the basic SWT outlined in chapter 2. For each satellite pass, the closest HIRS spots along the cruise track were first identified. Temperature and the moisture profile retrievals were then used to calculate the transmittances, upwelling and downwelling radiances, and surface reflectivity. These values were used in equation (2.19) to generate satellite derived SST with TOVS (AVHRR+TOVS). The coefficients in tables 2.1 and 2.2 were used to generate SST for the AVHRR-alone and the AVHRR+HIRS retrieval methods. The coefficients were selected according to the scan angle of each observed sample. NOAA's coefficients, obtained by McClain, were used to estimate the sub-surface temperature later to be compared with the measured in situ data. Finally, single channel correction methods were applied using atmospheric profiles retrieved from TOVS for both the AVHRR channels 4 and 5.  4.2.1 Scan Ansle Effect The importance of scan angle effects on the satellite measured SST has been expressed in chapter 2. In Figure 4.27, the result of the scan angle effect on the SST correction technique is shown. From this figure, the significance of scan angle correction above 40 degrees is shown. At a scan angle of 5 0 ° , an error of as much as 0.5 °C can be expected. For the physical TOVS method, the scan angle effect is not a problem since both the TOVS and the AVHRR sensors view a target at similar angles and TOVS profiles should represent the atmosphere seen by the AVHRR sensors. Thus for the physical method, the retrieved profiles were not limb corrected in order to preserve the atmosphere experienced by the radiances detected by the sensors.  91  Figure 4.27 The effect of scan angles on SST. Significant increases in error are shown for scan angles greater than 3 0 ° .  4.2.2 Emittance Effect Another significant influence on the SST is the emittance effect which has a close relationship to scan angles. The result of not including emittance term in the R T E when deriving a SST algorithm has been stressed by Dalu (1984) and Ulivieri (1984). However in the past, the emittance effect has been limited to a theoretical analysis due to a lack of accurate surface temperature measurements. With the PRT-5 measurements, the emittance effect can be compared with theoretical results. The emittance effect can be derived from subtracting equation (2.9) from (2.8).  B(u,T ) 9  where B(i/,T,)  -  B(uX)  > B(v,T[).  = p{v){B{u T.) t  - B {v T.)\\ x  t  -  T(»,PO)}}  ,  (4.1)  This means that if the emittance effect is not corrected for,  the retrieved temperature will be lower than that with emittance effect corrected. The 92  emittance effect are listed in Table 4.15. The emittance effects on the SST for satellite passes considered in this study is large for all passes as most satellite observations were made at extreme angles. This result agrees well with the theoretical values calculated by Dalu (1985) for scan angles greater than 4 0 ° . From this result, the importance (0.2 to 0.4 ° C difference) of including all terms in the R T E when developing a SST retrieval algorithm, is again confirmed.  Table 4.15 Emittance effect on SST Orbit  A T ( °C)  15779 15835 17680 17681  0.39 0.41 0.45 0.44  4.3 Comparison with Surface Skin Temperature In this section comparisons between the satellite derived SST and the ship-borne radiometer measurements, are presented. The comparisons presented here are for observations from four days. The selected NOAA passes are NOAA-7 15779 on July 14, NOAA-7 15835 on July 18, NOAA-7 17667 on November 25, NOAA-7 17680 on November 26, and NOAA-7 17681 on the same day. The first two passes were over the eastern North Pacific Ocean while the latter three passes were over the North Atlantic Ocean. Other passes are not presented due to either the lack of surface data or cloud contamination. Resulting SST from all six methods were compared against the PRT-5 measured SST and some statistics were calculated. The result from comparisons from all 4 days are summarized in Table 4.16. Table 4.17 lists surface, upwelling, and downwelling radiances with surface transmittance and reflectance for channel 4 of the AVHRR. The boldfaced numbers underneath are figures obtained using the coincident or nearest radiosonde observations. Table 4.18 is similar to Table 4.17, but for AVHRR channel 5. The SST retrieved from the three methods mentioned are plotted in Figure 4.28 along with the SST measured by the PRT-5 93  Table 4.16  Results of satellite derived versus PRT-5 measured SST AVH* AVH AVH AVH NOAA Orbit +H +T +T  15779  0.73 0.89 0.51 0.95  0.33 0.56 0.45 0.95  0.71 0.79 0.36 0.96  0.28 0.44 0.34 0.96  -0.31 0.45 0.32 0.96  204  0.96 1.06 0.45 0.93  0.32 0.56 0.46 0.92  0.15 0.44 0.41 0.93  0.87 0.94 0.35 0.94  0.86 0.92 0.33 0.94  0.59 0.67 0.32 0.94  290  -0.34 0.95 0.89 0.07  -0.89 1.20 0.81 0.14  -1.07 1.25 0.65 0.34  -1.58 1.68 0.56 0.56  -2.41 2.48 0.59 0.65  110  CC )RR  0.06 0.83 0.83 0.13  BIAS( •C) RMSDl °C <r| CC )RR  0.92 0.94 0.21 0.47  0.52 0.57 0.23 0.40  0.18 0.28 0.22 0.32  -0.17 0.23 0.15 0.71  -0.57 0.60 0.20 0.80  -1.29 1.32 0.28 0.80  122  BIASf°C) RMSD  0.91 0.96 0.28 0.65  0.56 0.64 0.31 0.66  0.23 0.39 0.32 0.69  -0.39 0.46 0.24 0.54  -1.04 1.08 0.31 0.25  -1.99 2.03 0.41 0.02  158  BIAS( °C) RMSDl °C ; J c  cc >RR 17667  17680  BIAS( °C) RMSDl • c j  :4  17681  Sample Size  5  1.30 1.39 0.48 0.95  BIAS( °C) RMSDl CC )RR  15835  AVH +T  radiometer for orbit 15779. The cruise for this day started from near the shore where the water temperature was cold and headed out in the southwest direction. The temperature increased slowly moving away from the shore and warmer temperatures were observed over a warm tongue which extended from the southwest (see Figure 3.6). The temperature dropped off at the edge of the tongue at the continental shelf break. In Figure 4.29, temperature scatter plots are shown for the same pass. For all methods, there were biases with the largest from the AVHRR-alone method and the smallest from the AVHRR+HIRS method. This can be seen in the histograms plotted in Figure 4.30. The AVHRR-alone method produced a bias of 0.73 °C. Addition of HIRS water vapor data significantly improved both the bias and the RMS differences (0.33 ° C and 0.56 °C, respectively). For the AVHRR+TOVS methods, large bias and RMS differences resulted. This is mainly due to the poor moisture profile retrieved by T O V S . As mentioned in the previous section, the moisture content was overestimated. Hayden et al., (1981) reported that even the most 94  transparent HIRS water vapor channel (channel 10) does not have its weighting function peaking at 900m6 as previously predicted, but rather peaks at a higher elevation. Thus the moisture content near the surface, which accounts for most of the moisture on this day, could not be sensed. The severity of this problem worsens as the surface moisture increases. In general, an increase in atmospheric moisture is associated with an increase in the surface temperatures and the warmer surface temperatures with increases in the surface radiances. Larger surface radiance interacting with a less moist atmosphere produces less attenuation. To compensate for the large radiance calculated, the iteration process increased the moisture content above the atmospheric level where the most of the moisture should lie. This fault in the compensation results in overestimation of the moisture content at most levels.  Table 4.17  TOVS retrieved AVHRR  channel 4 parameters  Orbit  Line  Spot  B(T.)  15779  7  49  15835  22 22  19 20  17667  18 17 17 16  47 48 49 50  15 14  51 52 2 3 4  8 8  12 11  6 7 51 53  11 11  54 55  17680  17681  8 7 7  89.83  B\ 83.45  84.93  To  Po  0.90  0.00835  104.48  84.57  85.08  0.03  0.00835  05.53  80.63  82.21  0.00  0.00835  80.76  76.53  70.60  0.81  0.00835  85.67  60.08  71.10  0.03  0.00835  85.67  60.08  71.10  0.03  0.00835  90.65 89.26 85.24 85.48 84.65 85.19  85.11 84.24 80.78 82.52 84.46 85.58 86.16 88.64 89.40 89.52 89.54  (radiance values are in m — arcm a  77.56 77.49  78.55 78.42  72.47 70.22 70.54 69.45  74.02 71.37 72.04 70.75  66.16 64.01 66.15 65.49 64.13  67.13 64.83 67.05 66.32 64.90  61.58 61.67 68.57 66.88  62.32 62.52 69.60 67.71  65.11 64.03  65.90 64.82  0.94 0.94 0.88 0.91 0.89 0.90  0.92 0.93 0.93 0.94 0.94 0.94 0.94 0.92 0.94 0.94 0.94  0.00835 0.00835  0.00835 0.00835 0.00835 0.00835 0.00835 0.00835 0.00835 0.00835 0.00835  0.00835 0.00835 0.00835 0.00835 0.00835 0.00835  )  l  For orbit 15835, the satellite retrieved and ship-borne radiometer measured SST's are 95  Table 4.18 TOVS retrieved AVHRR channel 5 parameters Orbit  Line  Spot  15779  7  49  15835  22 22  19 20  17667  18  47 48 49 50  17 17  16  B{T.)  To  99.45  104.26  0.84  0.01125  110.64  102.36  104.83  0.88  0.01125  110.27  07.38  100.01  0.84  0.01125  104.18  01.03  06.40  0.71  0.01125  00.85  86.26  88.23  0.88  0.01125  00.85  86.26  88.23  0.88  0.01125  95.31 95.58  105.12 103.65 99.39 99.65 98.77 99.35  87.56 85.97 85.42 84.73  17680  15 14 8  51 52 2 3 4  17681  8 8 12  11  6 7 51 53  99.76 100.37 103.00 103.81  11 11  54 55  103.93 103.95  7 7  102.02  Po  99.26 98.33 94.66 96.51 98.57  (radiance values are in m — 3  81.66 79.73 82.24 81.76 80.57 77.65 77.34 84.51 83.17 81.44 80.29  97.02 97.17 90.21 87.91 87.99 86.95  83.28 81.08 83.75 83.17 81.84 78.87 78.79 86.21 84.53 82.73 81.60  0.90 0.90 0.81 0.86 0.82 0.85 0.87 0.89 0.89 0.90 0.91 0.91 0.90 0.88 0.90 0.90 0.91  0.01125 0.1125  0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125 0.01125  srcm~ ) l  shown in Figure 4.31, with the scatter plots, are shown in Figure 4.32. For three days preceding July 18, a very strong wind blew steadily in the southeastward direction which caused an ocean mass transport perpendicular to the direction of the wind (to the right). This resulted in increased upwelling near the coast bringing cold water to the surface. The cruise track for this day, followed to the southwest direction over the same warm tongue mentioned with the earlier pass (see Figure 3.7). Soon after passing this warm region the ship changed its course and headed southeast towards Juan de Fuca Strait. A decrease in temperature was quite noticeable as the ship approached the shore. With this pass, the largest values of bias and RMS difference were observed among the five satellite passes considered. There are several explanations for these large discrepancies. First by examining the radiosonde dew point profile and the TOVS retrieved dew point profile one can see that the T O V S dew point profile missed all the dynamic structures below the troposphere. The T O V S dew point profile seems to represent averaged dew point values. The TOVS  Figure 4.28 SST from the three atmospheric correction methods for orbit 15779.  Figure 4.29 Temperature scatter plots from the three methods for 15779. 97  N7-15779 |{RVHRR)  -1.4-1.2  H/07/84  - (PRT)  I I -0.8 -0.6 -0.4 -0.2  -1.0  I  \  0.0  0.2  DELTA (AVH+HIRS)  CO  1.4-1.2  -1.0  -  -0.8 -0.6 -0.4 -0.2  0.0  DELTA (AVH+T0VS)  -  0.4  |  |  1  1  0.6  0.8  1.0  1.2  r1 4  T(C)  (PRT)  o  CO  "\  0.2  0.4  0.6  0.8  1.0  1.2  1.4  T(C)  (PRT)  i—i O [_ (M  cc  £° CD  o  0 - |  o  I 1.4-1.2  I -1.0  1  I  |  -0.8 -0.6 -0.4 -0.2  1  1  1  0.0  0.2  0.4  DELTA  |  • " [ r  0.6  0.8  1  1  Hr a  1.0  1.2  1.4  T(C)  Figure 4.30 Binned histogram from the three methods for 15779. AVHRR+HIRS method produced the smallest bias ( « 0.4 ° C ) .  98  temperature profile is more comparable to the radiosonde profile; still small scale structures were not resolved by TOVS. During the Iteration process, the water vapor amount was again overestimated due to the poor vertical resolution of the HIRS water vapor channels. To aggravate the situation, a problem with ship navigation was also encountered with this pass. The large bias and RMS error are partly due to a mismatch between the actual PRT5 track and the calculated track over the AVHRR imagery. This is evident by examining the correlation coefficient. For the AVHRR-alone and the AVHRR+HIRS methods, the results are much better than the one from the AVHRR+TOVS method. Obviously, the atmospheric water vapor content concentrated in the mid-troposphere was well sensed by the HIRS channel 11. The binned histogram for this pass is shown in Figure 4.33 In Figure 4.34, the satellite retrieved and the PRT-5 measured SST's are presented for orbit 17680. The temperature scatter plots are presented in Figure 4.35 for the three SST retrieval methods. Similar plots are shown for orbit 17681 in Figures 4.36 and 4.37. The typical atmospheric condition, for this day, was dry and clear with the radiosonde observed water vapor amount of 0.9g/cm . Here, the AVHRR-only SST computation overestimates 9  the atmospheric water vapor impact leading to a strong positive bias in satellite-ship SST match-ups. This bias is significantly reduced by applying the AVHRR+HIRS formulae to it. The comparison shows that the TOVS profile retrievals worked very well over the dry winter atmosphere. The comparison between the TOVS corrected and the measured SST shows a negative bias of 0.39 ° C and an RMS difference of 0.46 ° C . This result is better than the result from the AVHRR-alone method and differs only by 0.05 °C from the AVHRR+HIRS method. The binned histograms are shown in Figures 4.38 and 4.39 for orbits 17680 and 17681, respectively. From the above results, it is clear that the AVHRR+HIRS method of correcting the atmospheric influences, performed the best under all atmospheric conditions encountered. AVHRR+TOVS performed well only under certain atmospheric conditions, when the atmosphere exhibited slow varying temperature and water vapor profiles exhibited no abrupt changes. Yet it was noticed that this method usually performed better than the simple 99  N7-I5835  18/07/84  O A  (AVHRR) (HVH+HIRS) (flVH+TOVS) (PRT)  20.33  21.[  Figure 4.31 SST from the three atmospheric correction methods for orbit 15835.  N7-15835  18/07/84  (RVHRR)  (flVH+HIRS)  (RVH+TOVS)  14.0  Figure 4.32 Temperature scatter plots from the the three methods for 15835. 100  N7-15835  o  (AVHRR)  -  18/07/84  (PRT)  CO  1.4-1.2 -1.0 -0.8 -0.6 -u'.4 -0.2  0.0  0.2  o'.4 o'.6 o'.8 l'.O  1.2  l'.4  O-i  DELTA  T(C)  CO  (RVH+HIRS)  -  (PRT)  1.4-1.2 -1.0 -0.8 -0.6 -0.4 -0.2  DELTA  O-i  (AVH+T0VS)  CO  0.0  -  0.2  0.4  0.6  0.8  1.0  1.2  1.4  T(C)  (PRT)  O H  CC  1.4-1.2 -1.0 -0.8 -0.6 -0.4 -0.2  0.0  DELTA  0.2  0.4  0.6  0.8  1.0  1.2 1.4  T(C)  Figure 4.33 Binned histogram from the three methods for 15835. AVHRR+HIRS method produced the smallest bias. 101  N7-I7680  26/11/84  O A + -  (AVHRR) (AVH+HIRS) (AVH+TOVS) (PRT)  SHIP TIHEIHI  Figure 4.34 SST from the three atmospheric correction methods for orbit 17680.  N7-17680  26/11/84  Figure 4.35 Temperature scatter plots from the three correction methods. 102  N7-17680  o• ro  (AVHRR)  i  -  1  (PRT)  r  -0.6  1.4-1.2 -1.0 -0  i  i  -0.4 -0.2  0.0  DELTA  o•  (AVH+HIRS)  26/11/84  -  i  0.2  i  0.4  i  0.6  r  1  1  0.8  1.0  1.2 1.4  1  -i 1.0  1.2 1.4  T(C)  (PRT)  o• CM  CC  ^ o CD o • O ~ o  CD.  i i r 1.4-1.2 -1.0 -0.  !  -0.6  (AVH+T0V5)  CO  o f—  0.0  DELTA  o• ro  -  |  |  (  -0.4 -0.2  0.2 0.  i r 0.6 0.  1  1  —  T(C)  (PRT)  o -  cr > etc  UJ O CO CD o O  o d.  i i i i i i 1.4-1.2 -1.0 -0.8 -0.6 -0.4 -0.2  1  1  1  1  1  1  0.0  0.2  0.4  0.6  0.8  1.0  DELTA  1 r 1.2 1.4  T(C)  Figure 4.36 Binned histogram from the three methods for orbit 17680. A negative bias was observed from the AVHRR+TOVS method. 103  N7-I768!  26/11/84  HIP TIHEIHI"  O  (RVHRR)  A  (flVH+HIRS)  +  (RVH+TOVS)  -  1PRT1  "°  158  "•'  »•>  Figure 4.37 SST from the three atmospheric correction methods for orbit 17681.  N7-17681  *~—i  26/11/84  1  1  1  10.011.0 12.0 13.0 14.0  PRT T(C)  ~-f  1  1  1  1  10.011.0 12.0 13.0 14.0  PRT T(C)  1  1  1  PRT T(C)  Figure 4.38 Temperature scatter plots from the three correction methods. 104  ,  10.011.0 12.0 13.0 14.0  N7-17681  o ro  (AVHRR)  -  26/11/84  (PRT)  CO  .1-1.2 -1.0 -0.8 -0.6 -0.4 -0.2  0.0  DELTA  o ro  (AVH+HIRS)  g°  -  0.2  0.4  0.6  0.8  1.0  1.2  1.4  T(C)  (PRT)  CO  i—i ( _  o • r\]  CC >  V,  CC  !H° C D  CD-  C D  ~  o  CD_  — i I I 1.4-1.2 -1.0 -0.  i  -0.6 -0.4 -0.2  I  'A I 4  0.0  DELTA  0.2  T(C)  I  I  i  0.4  0.6  0.8  I  1.0  1  1.2  1— 1.4  (PRT)  I I 1 i i I •1.4-1.2 -1.0 -0.8 -0.6 -0.4 -0.2  I  0.0  DELTA  I  0.2  1  0.4  1  0.6  1  0.8  1  1.0  1  1.2  r 1.4  TIC)  Figure 4.30 Binned histogram from the three methods for orbit 17681. AVHRR+HIRS method produced the smallest bias. 105  AVHRR-alone method. The performance of this physical method (AVHRR+TOVS) depended heavily upon the results of the TOVS vertical profile retrievals. It is difficult to evaluate the accuracy this method can achieve. In order to better assess the SST results using T O V S , the Atlantic AVHRR data were corrected using the same physical method but with transmittances calculated from the coincident radiosonde data. If the radiosonde data represent the true state of the atmosphere, the result obtained from this exercise should establish the limitation of the physical method of estimating SST. The fundamental difference in characteristics between the radiosonde data and the atmosphere seen by the T O V S sensor, as discussed before, must be recalled here. The mean and the RMS differences in the estimated SST using the T O V S and the RAOB, and the PRT-5 SST are shown in Table 4.19. From values in this table, it is evident that the performance of TOVS has been underrated by other researchers in the past. Two possible reasons are first, most past comparisons were done on a strict comparison between TOVS retrieved and the nearby radiosonde profiles, and second, researchers were more concerned with the climatological impact of the TOVS retrievals which requires high resolution profiles the T O V S cannot provide. In most passes, the physical SST method based on the TOVS profiles showed smaller RMS deviation than the one based on the radiosonde. Table 4.10  Comparison of physical SWT based on TOVS and RAOB Orbit 15779  BIAS(°C) RMSD(°C)  RAOB 0.74 0.81  TOVS 0.71 0.79  15835  BIAS(°C) RMSD(°C)  0.89 0.96  0.87 0.94  17680  BIAS(°C) RMSD(°C)  -0.26 0.28  -0.17 0.23  17681  BIAS(°C) RMSD(°C)  -0.34 0.40  -0.39 0.46  For orbit 17680, the study area was located to the west of the satellite pass nadir and for the next orbit (17681), the same area was to the east of the satellite nadir. This offered an opportunity to try the SST correction method which utilizes absorption differences 106  between two different scan angles rather than two channels as in the SWT. However, the result was very poor with a bias of 0.82 ° C and a RMS difference of 3.52 °C. Figure 4.40 shows two TOVS temperature profiles over the same area observed by the two consecutive passes. The time and the space differences ( « 100 minutes and within the single HIRS F O V , respectively) are sufficiently small enough to be ignored. Thus the differences reflect primarily the atmospheric path differences.  The earlier pass (17680) produced a smaller bias ( - 0 . 1 7 ° C compared to -0.39 °C) and RMS differences (0.23 * C versus 0.46°C). There can be several explanations for these differences. First, the atmosphere within the F O V for the orbit 17680 was predominantly over land whereas orbit 17681 was over the Atlantic Ocean where greater moisture content is expected. This is verified by examining the two T O V S retrieved dew point profiles and the total water amounts (0.838 cm va. 0.854 cm) for the same earth location. The increase in moisture in the atmosphere affected the transmittance of the AVHRR channel 5 resulting in larger RMS differences. This result is shown in Figure 4.41. Another explanation is that the atmosphere went through rapid changes in the moisture content, within two consecutive satellite passes ( « lOOmtn). This explanation is, however, unlikely. 4.4 Skin vs BnDc Temperature Comparison  The PRT-5 measured SST and the bulk temperatures measured at 2 to 4 meters below the surface are shown in Figures 4.42, 4.44, 4.46, and 4.48 for the Pacific and for the Atlantic. In these figures, SWT retrievals using the NOAA coefficients are also shown. The NOAA coefficients (or McClain's coefficients) are derived by regressing with ship, buoy, and X B T measurements and thus represent the bulk water temperatures rather than the surface skin temperatures. The resulting SST using the NOAA coefficients are comparable to results found by others (McClain, 1985; Njoku, 1985) with a bias of 0.58 °C and a RMS difference of 0.68 ° C in the worst case. The statistical measures between the satellite derived and the bulk temperatures are summarized in Table 4.20. The largest differences were observed on July 18. As described before, the cruise on this day suffered 107  TEflPERflTURE PROFILES  DEU POINT PROFILES  TOVS DRTE : 841126 STRRT TIHE: 142713 END TIHE : 143025  TOVS OflTE : 841126 START TIHE: 142713 ENO TIME : 143025  Q  • umnoE i n . s i  LOCHUOEi O • LfltinCC > 4.7S uxatu*.  C  - UtltTUOC : «9.1t LOWITUMi -3.08  a  - LATITUDE ,  «9.«  LOMITlJOEi -1.91  Figure 4.40 Temperature and dew point profiles from two orbits. Differences between 17680 and 17681 SST's are primarily due to atmospheric differences in two F O V s .  +• 26/11/84  *  + N7-17681  2.0  12.0  13.6  14.4  i 15-*  i 18-0  i IS.S  i US  i 10.4  SHIP TiMEIHI  i— t9.2  +*•+  1  1  20.0  20.0  +++  +  +  -  (PRT)  — i 21.6  1 22.4  1 23.2  1 24.0  + +  26/11/84  N7-17680  2-0  12-9  13.6  14.4  '3-2  100  IS.0  IT B  la 4  SHIP TIHEIHI  19 ?  lo.o  xj.a  -  (PRT)  21.0  72.4  2). 2  24.0  Figure 4.41 SST differences measured by two consecutive satellite passes. Data from 17680 has smaller RMS error than those from 17681. 108  from a strong southeastward wind resulting in large deviations in the PRT-5 temperature due to varying incident angles. Figures 4.43, 4.45, 4.47, and 4.49 show histograms of the bulk-skin temperature differences which occurred during the Pacific and the Atlantic cruises. In the Atlantic, the 0.22 °C mean deviation of the cool sea surface skin, from the deeper in situ temperatures, still leaves a range of values from — 0.5 °C to 0.9 °C. Taking this fact into account one may expect systematic uncertainties when comparing remotely sensed SST's to ship borne bulk temperatures, which should lie in the range of the SST (0.3 °C) accuracy required for climate applications ( W G R P / T O G A , 1984).  Table 4.20  Results of satellite derived SST versus bulk temperature Orbit 15779  BIAS( °C) RMSDl  °cj  >RR 15835  BIAS( °C) RMSDl °C)  •cj  a )RR 17667  BIAS( °C) RMSDl °C)  °C)  )RR 17680  BIAS( °C) RMSDl °C)  °C)  )RR 17681  AVH  AVH +H  AVH +T  AVH» +T  AVH +T  Sample Size  0.48 0.63 0.41 0.96  -0.09 0.45 0.44 0.96  •0.50 0.64 0.40 0.96  -0.12 0.35 0.33 0.97  -0.54 0.63 0.32 0.97  -1.14 1.18 0.32 0.97  204  -0.26 0.50 0.43 0.93  •0.90 1.00 0.43 0.92  -1.07 1.15 0.42 0.93  -0.35 0.53 0.40 0.93  -0.36 0.54 0.40 0.94  -0.63 0.75 0.41 0.93  290  -0.06 0.79 0.78 0.19  -0.45 0.95 0.84 0.14  -1.00 1.27 0.77 0.20  -1.19 1.34 0.61 0.39  -1.70 1.78 0.54 0.58  -2.53 2.59 0.59 0.64  110  0.58 0.61 0.20 0.54  0.18 0.28 0.22 0.47  -0.17 0.27 0.21 0.40  -0.51 0.53 0.14 0.76  -0.91 0.93 0.19 0.83  -1.63 1.65 0.27 0.81  122  0.58 0.64 0.27 0.69  0.22 0.37 0.30 0.70  -0.11 0.33 0.31 0.73  -0.72 0.76 0.24 0.57  -1.37 1.41 0.31 0.26  -2.32 2.36 0.41 0.02  158  NOAA  BIAS( °C) RMSDl °C  °C)  )RR  5  The ocean surface as seen by a radiometer usually has a cool skin being approximately 0.2 C to 1.5 °C colder than the water below (Saunders, 1967; McAlister and McLeish, 1969; 6  Hasse, 1971; Grass 1 and Hinzpeter, 1975; Grassl, 1976, 1977; Robinson et ol., 1984). This has to be kept in mind when comparing remotely sensed surface temperatures with ship 109  *•  N7-15779  14.0  |4.«B  14.93  N7-15779  V  V  -  V * M * +4**M*%  ^  14/07/84  13.4  13.06  10.33  - (T AT 2 M) + (MCCURIN) 1 1 1  —i  16.0  IB. 2  10.66  19.13  1  1S.S  20.06  1  20.31  1  21.0  14/07/84  - (T AT 2 M) + (PRT)  Figure 4.42 Bulk and PRT-5 surface temperature plots for Jul. 14. The top graph shows the satellite retrievals with NOAA coefficients.  N7-15779  1  1  1  ,  1  ,  -1.4-1.2 -1.0 -0.8  -0.6  -0.4  -0.2  0.0  0.2  "I  1  1  14/07/84  DELTA  ,V  0.4  ^  0.6  A/ A/X  0.8  l'.O  /\  l'2  T(C)  Figure 4.43 Binned histogram for Jul. 14. Bulk temperature is 0.8 °C warmer than the surface skin temperature 110  N7-15835  18/07/84  (T  fiT  2  M)  (flCCLHlN)  11-0  11.68  12.39  N7-15835  13.0  13.G  18/07/84  - (i ni 2 m +  (PRT)  Figure 4.44 Bulk and PRT-5 surface temperature plots for Jul. 18. The top graph shows the satellite retrievals with N O A A coefficients.  N7-15835  o to  fl)  T RT 2  CD  -  18/07/84  (PRT)  o = —. O -  en ^  en  o  .  co o  o  -  -  n  1  •1.4-1.2 -1.0  1  -0.  1  -0.6  1  -0.4  r  -0.2  0.0  DELTA  0.2  0.4  0.6  0.8  1.0  1.2  1.4  T(C)  Figure 4.45 Binned histogram for Jul. 18. The strong negative biases are partially due to varying angles of incident of the PRT-5 caused by strong winds. Ill  Figure 4 . 4 6 Bulk and PRT-5 surface temperature plots for Nov. 25. The top graph shows satellite retrieval with NOAA coefficients; Satellite data was heavily cloud contaminated.  o d  - i  N7- 17667 2 5 / 1 1 / 8 4 T RT 2 M:  O  -  (PRT)  V  -  ^  CC ^ o. CO C G oCD ~ o d. -1.4-1.2 -1.0 - 0 . 8 -0.6 - 0 . 4 -0.2  0.0 0.2  DELTR  "T  0.6  0 . 8 1.0  T(C)  Figure 4 . 4 7 Binned histogram for Nov. 25. The bulk temperature is » 0.2 ° C warmer than the surface skin temperature. 112  N7-I768I 26/11/84  - (T AT + (PRT) 6.0  T.2  9.6  10.6  |2.0  13.7  13.6  16.S  16.6  19.2  2  n> 20.4  21.6  22.6  24.6  Figure 4.48 Bulk and PRT-5 surface temperature plots for Nov. 26. Top graph shows satellite retrieval with NOAA coefficients  26/11/84  O -1 LD  cn " .  -z. o  :T  A T 2 M) - ( P R T ;  0  —'  o  £CC ro" d  >  CC  o  cn °-  m ™  o CD.  -0.7  -0.5  -0.3  -0.1  0.1  DELTA T ( C :  0.3  0.5  0.7  Figure 4.40 Binned histogram for Nov. 26. The bulk temperature is « 0.3 ° C warmer than the surface skin temperature. 113  borne measurements. The skin temperature is balanced by the total heat flux at the oceanair interface. Therefore the fastest response of the skin temperature is to variations in the net longwave radiativefluxwhich establishes a balance between the surface emitted thermal radiation and the downwelling atmospheric radiation (which affects only the uppermost 10 micrometers of the ocean (Grassl, 1977)). This surface layer will be destroyed by the white capping of wind induced waves at wind velocities above about 10 m/sec (Clauss et al., 1970). This mixes the upper layer resulting in smaller differences between skin and bulk water temperatures. This result is seen Table 4.21 (data from Nov. 25 and Nov. 26) which shows the mean and the RMS temperature differences between PRT-5 measured skin temperature and the bulk temperature. For a strongly stratified upper ocean, the diurnal heating of the upper few centimeters, by incoming solar radiation, will raise the near surface temperatures increasing the skin temperature. The result will be a surface skin that is warmer than the water at a depth of several meters. Hence, the range of the surface skin temperature will be between some tenths of a degree higher than the bulk temperature and several tenths below it. Table 4.21  Comparison between skin by PRT-5 and bulk sea temperature Date  Corr  RMS( °C)  BIAS(°C)  *(°C)  July 14 July 18 Nov 25 Nov 26  0.99 0.91 0.99 0.94  0.87 1.31 0.16 0.30  -0.86 -1.22 -0.11 -0.28  0.17 0.49 0.11 0.13  4.5 Conclusions and Recommendations Through this study two mainfindingsprovide a significant contribution to the retrieval of SST. First, an entirely independent and physical method of retrieving SST from the satellite sensors has been developed and tested. As a result, the physical verification on the theory of SWT has been demonstrated. Also, the study has indicated the possibility of having a "true" satellite derived SST method which frees the solution from the limits set by the past statistics, and which is sensitive to anomalous conditions in the atmosphere. 114  Second, the satellite derived SST's were compared with the ship-mounted radiometer SST's by which the non-linear differences due to the surface skin effect are effectively eliminated. This result provided a detailed understanding of the nature of errors contributed by the atmospheric gases in the satellite derived SST which can now be dealt with separately from the surface skin effect. In the comparisons of satellite derived SST's versus in situ SST, the error which resulted from temporal differences has not been considered in this study. The satellite measurements typically took 2 to 4 minutes to pass over the study area whereas ship measurements were made over a period of 10 to 24 hours. Thus the PRT-5 SST included variations due to surrounding environmental conditions such as diurnal variations which cannot be detected by satellite measurements. The surface skin temperature usually appears colder when the incoming solar shortwave radiation ceases to exist. The evidence for this variation can be seen when the differences between the skin and the bulk temperature are examined with respect to the ship time. A larger temperature difference ( « 0.15 °C) was observed for data collected on July 14 after the sunset. This difference also includes temperature dependent bias. An analysis of the diurnal difference is not possible in this study because the data collection strategy was not specifically designed to study the diurnal variations. It is estimated, however, that the error due to diurnal variations is around 0.1 °C based on the temperature differences between the skin and the bulk measurements. The physical method (AVHRR+TOVS) performed with the RMS errors less than 0.94 °C for three out of 4 cruises tested, however, large biases were observed. For summer cruises, the biases were positive (satellite SST > PRT-5 SST) whereas for winter cruises, the biases were negative. The large biases in the summer cruises were primarily caused by overestimating the moisture content in the atmosphere by the TOVS retrieval method. These overestimations in the atmospheric moistures resulted in both the larger downwelling radiation and the larger attenuation than what actually occurred in the "true" atmosphere. The final effect is a warm bias in the satellite derived SST. The cold biases in the winter cruises seems to be the result of systematic biases as functions of the surface temperatures 115  and do not reflect the errors in the TOVS profiles, since the total water amounts measured by the T O V S is comparable to the ones measured by the radiosondes. Although the negative biases were unexpected, the RMS differences for the winter cruise are well within the physical limitation of the AVHRR sensor (0.23 and 0.46 °C for orbits 17680 and 17681, respectively). The overestimation in the water vapor profiles for the summer cruises are due to poorly defined HIRS water vapor channels and their broad vertical weighting functions. From comparison to what other researchers have found, the limitation of the HIRS instrument has already been recognized. The only way to improve the physical method is to employ more accurate satellite-borne vertical sounders. The use of radiosondes to estimate transmittances have proved to be no better than using the TOVS system. The reason is that even though radiosondes can provide accurate atmospheric measurements, their ascension paths do not necessarily represent the HIRS FOV's.  Supported by other studies and again in this study, a way of improving the vertical temperature and moisture profiles is to reduce the width of their weighting functions thus increasing the vertical resolution. The reduction of the width of the weighting function can be accomplished by narrowing the spectral bandwidth of the radiometric instrument. However, the finite bandwidth will always cause vertical smoothing. With the HIRS system, both 15 jan and 4.3 fim regions of the CO2 absorption spectrum are used to obtain measurements throughout the entire atmosphere.  A hybrid method, as Kaplan (1976)  has suggested, uses radiometers with narrow spectral responses located in the wings of 4.3/im to sense the troposphere, and radiometers in 15 fxm region are used to sense the stratosphere. This method increases vertical resolution to » 2 km in the troposphere and the retrieval procedure does not have to rely on an iterative scheme, thus eliminating the requirement of initial profiles. Based on the empirical studies done by Thompson (1982), it was found that this new set of radiometers, named Advanced Moisture and Temperature Sounder (AMTS), indeed exhibits an advantage over the present HIRS sounder at all atmospheric levels with maximum advantage occurring near 20 mb. With AMTS, the vertical resolution of 2 to 3 km is possible within the troposphere. 116  For moisture profiles, the number of channels sensitive to atmospheric moisture, which also should include more transparent channels, must be increased in order to enhance the surface measurements. As in the temperature sensing radiometers, the weighting function for the moisture sensitive radiometers should be made narrower to increase the vertical resolution. The inclusion of data obtained from the microwave sensors should be extremely useful despite their wide FOVs. The proposed plan for the 19 channel microwave sounder, Advanced Microwave Sounder Unit (AMSU) offers high hopes in providing more accurate atmospheric moisture information even in the presence of clouds in the FOV. Once the problems relating instrumentation have been solved, the effort should be directed toward removing systematic biases in the vertical profile retrieval algorithm. In this study, the number of available data was too little to be statistically justified. As mentioned earlier, the existence of the systematic biases is apparent from the profile comparisons shown in chapter 4. An instability was encountered in the iterative method under certain atmospheric conditions. The moisture content, especially near the surface, was grossly overestimated by the retrieval method during the iteration. This problem should be corrected by studying the nature of the feedback mechanism in the iterative retrieval method. In order to prevent the vertical profile converging on an arbitrary solution, additional information can be put in to bring the solution closer to the actual atmosphere. Additional information can be obtained from more accurate surface temperatures estimated from the HIRS window channels or even AVHRR data can be incorporated to estimate the surface temperature. Another important item of information is the tropopause height, if that can be estimated from the HIRS radiance measurements. For the AVHRR+HIRS method, the RMS errors found lie well within the errors given by the retrieval formula and the radiometric noise in ship and satellite measurements. Some biases are due to systematic atmospheric effects requiring more sophisticated analysis. The errors induced by atmospheric water vapor and temperature structures need to be corrected for in the SST algorithms which should also account for scan angle dependence and non-blackbody ocean surface radiation. Computation of the AVHRR only estimates 117  from NOAA-7 data have RMS accuracies of 0.56 — 0.9 °C, on a single pixel basis if scan angle dependency is explicitly considered. Comparison with in situ skin measurements indicate strongly biased (up to 0.71 °C) AVHRR SST estimates due to atmospheric contamination. The addition of HIRS reliably decreases these biases ( « 0.3 °C) caused by extreme atmospheric water vapor and temperature structures, as demonstrated by the comparison with in situ measured SST skin temperatures. The estimated SST by both the AVHRR+HIRS and the AVHRR+TOVS generally fell within the standard deviations reported during the JPL workshop. Their RMS errors were, however, greater than what the users demand (less than 0.5 °C). But it is suspected that this demand will perhaps never be met with the current sensors available. It appears the technical limitations of the current sensors have been reached and the best they can provide is consistent RMS deviations of less than » 0.9 °C in the mid-latitude regions. The poor performance of the TOVS in a moist environment is indicative of further degradation in the tropical regions. However, this is yet to be tested. The performance of the AVHRR+HIRS and the AVHRR+TOVS confirm the necessity of current water vapor information in the basic split window algorithm. A concern for the AVHRR+HIRS method is that this method still relies on the past statistics, namely past radiosonde data from several types of atmosphere. This means that this method may not entirely be suitable at any location at any one time. One reason for the acceptable results from the AVHRR+TOVS, despite large errors in the TOVS retrieved atmospheric moisture profiles, should be credited to the SWT. The effects of atmospheric water vapor on the AVHRR channels 4 and 5 transmittances are indeed significant. This implies that retrieving accurate SST's using single channel is probably not possible. It is shown that if the split window technique is used, error up to 10 °C at all atmospheric levels, result in deviations of less than 0.2 °C in the final SST. This result is worse if the errors occur in the moisture profiles.  Another possible error source is the uncertainty in the pixel-ship co-locations. This is evident for the case of July 18. It appears that there were errors in the ship position which were manually recorded every half an hour. Obviously, single fields of view are intersected 118  by the ships course on different secants, so that the 5 minute ship tracks may fall partly out of the pixel selected. However, the error induced by this aspect should not be very large since the surface temperature seldom shows gradients of 0.5 °C per pixel. The temporal variability of the ocean surface is not easy to quantify. It depends strongly on skin-bulk temperature differences, their destruction by wind induced white caps, the diurnal heating of the surface layer and on dynamical processes like semidiurnal tides (Robinson, 1984). Nevertheless, the selected one day periods seem adequate to match up the ship and satellite measurements. One other thing to consider when looking at the SST comparisons, other than the temporal differences, is the size differences in the F O V s between the AVHRR and the PRT-5. These differences are quite large ( » 1 km versus less than 5 m ) and 2  2  their effects need to be studied further. Especially on windy days where there are large swells causing vertical mixing in the water, the satellite measurements show up warmer than what the PRT-5 measured due to larger integrated area. The main drawbacks of all SST retrieval methods considered here, is their lack of proper detection schemes for cold thin clouds, and inability to estimate marine aerosol amounts shifting satellite borne temperatures to lower values (Haenel, 1976). Both subpixel clouds and aerosol particles are able to modify the SST retrievals in a systematic way up to several tenths of a degree as computed by radiative transfer simulations of the radiometer channels. For the TOVS sounder, with its coarse spatial resolution, clouds are the limiting factor since no "declouding" procedures were used to produce clear fields of view where there are partly cloud contaminated spots. Generally, HIRS F O V s are too large to investigate mesoscale features even in fields of broken clouds. Contrary to this, significant improvement is possible for clear areas with the AVHRR/HIRS combination. As mentioned in the beginning of this thesis, the physical SST retrieval method is attractive in areas where in situ data is sparse. It has been proven that the this method is comparable to other SST retrieval methods. One disadvantage of the physical method is that it is computation intensive. It is certainly not advisable to use this method to generate SST fields over a large area. However, this method can be used to generate 119  seasonally or latitude varied coefficients to perhaps a SST retrieval method such as the AVHRR+HIRS method. This process will be effective in overcoming the short term reliability of the regression based SWT's where radiosonde or ship SST measurements are not readily available while keeping a reasonable computation time.  120  BIBLIOGRAPHY Abel, P.G.,Smith, W.L., and Arking, A.,1978: An empirical model for atmospheric transmittance functions and its application to the NIMBUS-6 HIRS experiment. NOAA Technical Memorandum NESS 99. Anderson, J . M . and Wilson, S.B.,1984: The physical basis of current infrared remote sensing techniques and the interpretation of data from aerial surveys. Int. J. Remote  Sensing, 5, pp. 1-18. Anding, D., and Kauth, R., 1970: Estimation of sea-surface temperature from space. Remote Sensing of Environ.,l(4,), pp. 217-220. Anding, David C.,1979: Overview of atmospheric effects on remote sensing applications, Atmospheric Effects on Radiative Transfer, Proc. Soc. Photo-Optical Inst. Engineers, 105, C.B. Ludwig (ed), pp. 126-134. Avra, Elton P. and Alexander, George, 1975: Empirical investigation of three iterative methods for inverting the radiative transfer equation. ECOM-5576, US Army Electronics Command, Fort Monmouth, New Jersey 07703. Backus, George and Gilbert, Freeman, 1967: Numerical application of a formalism for geophysical inverse problem. Geophys. J. Royal Astron. Soc., 13, pp. 247-276. Backus, George and Gilbert, Freeman, 1968: The resolving power of gross earth data. Geophys, J. Royal Astron. Soc., 16, p. 169-205. Barcilon, Victor, 1975: On Chahine's relaxation method for radiative transfer equation.  J. Atmospheric Sci., 32, p. 1626. Barnett, Thomas L., 1969: Application of a non-linear least-square method to atmospheric temperature sounding. J. Atmospheric Sci., 26, p.457. Barton, I. J . , 1985: Transmission model and ground-truth investigation of satellite-derived sea surface temperature. J. Climate and App. Met., 24, 508-516. Ben-Shalom, A., DeVir, A.D., and Lipson, S.G., 1983: Infrared spectral radiance of the sky. Optical Engineering, 22, pp. 460-463. Bernstein, R.L., 1982: Sea surface temperature estimation using the NOAA 6 satellite Advanced Very High Resolution Radiometer. Journal of Geophysical Research, 8 7 , pp. 9455-9465. Bernstein, R.L. and Chelton, D.B., 1985: Large-scale sea surface temperature variability from satellite and shipboard measurements. Journal of Geophysical Research, 90, 11619-11630. Brower, R.L., Gohrband, H.S., Pichel, W.G., Signore, T . L . and Walton, C C , 1976: Satellite derived sea surface temperatures from NOAA spacecraft. NOAA Tech. Memo NESS 78, 74pp Brown, O. B., Brown, J . W., and Evans, R. H., 1985: Calibration of Advanced Very High Resolution Radiometer infrared observations. Journal of Geophysical Research, 00, 11667-11677. 121  Bruce, R.E., Duncan, L.D. and Pierluissi, J.H., 1977: Experimental study of the relationship between radiosonde temperatures and satellite-derived temperatures, Monthly Weather Review, 105, 493-496. Burch, Darrell E., 1981: Continuum absorption by atmospheric H3O. Atmospheric Transmission, Proc. SPIE, 277, pp. 28-39. Chahine, Moustafa T., 1970: Inverse problems in radiative transfer Determination of atmospheric parameters. Journal of Atmospheric Science, 27, pp. 960-967. Chahine, Moustafa T., 1972: A general relaxation method for inverse soution of the full radiative transfer equation. Journal of Atmospheric Science, 20. Chahine, Moustafa T., 1974: Remote sounding of cloudy atmosphere I. The single cloud layer. Journal of Atmospheric Science, 31, pp. 233-243. Chahine, Moustafa T., 1975: An analytical transformation for remote sensing of clearcolumn atmospheric temperature profile. Journal of Atmospheric Science, 32, pp. 1946-1952. Chahine, Moustafa T., 1977: Remote sounding of cloudy atmosphere II. Multiple cloud formations. Journal of Atmospheric Science, 34, pp. 744-757. Chahine, Moustafa T., Aumann, H.H., and Taylor, F.W., 1977: Remote sounding of cloudy atmosphere HI. Experimental verifications. Journal of Atmospheric Science, 34, pp. 758-765. Chahine, Moustafa T., 1977: Generalization of the relaxation method for the inverse solution of nonlinear and linear transfer equations. Inversion Methods in Atmospheric Remote Sounding, A Deepak (ed), Academic Press, New York, 622 pp. Chahine, Moustafa T., 1978: Derivation of clear-column temperature profiles with high vertical resolution and accuracy. Remote Sounding of the Atmosphere from Space, COSPAR, 4, H-J. Bolle (ed), Pergamon Press. Chahine, Moustafa T., 1980: Infrared remote sensing of sea surface temperature. Remote Sensing of Atmospheres and Oceans, A . Deepak (ed), Academic Press, New York, p. 641. Chahine, Moustafa T., 1982: Remote sensing of cloud parameters. Journal of Atmospheric Science, 30, pp. 159-170. Chen, Y . M . , Woolf, H.M., and Smith, W.L., 1974: Vertical resolution of temperature profiles for High Resolution Infrared Radiation Sounder (HIRS). NO A A Technical Report NESS 67. Chou, Ming-Dah, 1974: An iterative scheme for determining sea surface temperatures, temperature profiles, and humidity profiles from satellite measured infrared data. J. Geophys. Res., 70, pp. 430-434. Clauss, E., Hinzpeter, H., Mueller-Glewe, J . , 1970: Messungen zur temperaturstruktur im wasser an der grenzflaeche ozean-atmosphere. *Meteor"-Forschungsergebnisse, Reihe B, No 5, 90-94. 122  Conrath, Barney J . , 1969: On the estimation of radiative humidity profiles from mediumresolution infrared spectra obtained form a satellite. Journal of Geophysical Research, 74, No. 13, pp. 3347-3361. Conrath, Barney J . , 1972: Vertical resolution of temperature profiles obtained from remote radiation measurements. Journal of Atmospheric Science, 20, p. 1262. Conrath, Barney J . , 1977: Backus-Gilbert theory and its application to retrieval of ozone and temperature profiles. Inversion Methods in Atmospheric Remote Sounding, A. Deepak(ed), Academic Press, New York, 622 pp. Crosby, D.S. and Weinreb, M.P., 1974: Effect of incorrect atmospheric statistics on the accuracy of temperature profiles derived from satellite measurements. J. Stat. Comput. Simul., 3, pp. 41-51. Dalu, G . , Prabhakara, C , and Lo, R.C., 1981: Improved accuracy of the remote sensing of sea surface temperature. Oceanography from Space, pp. 109-114, Plenum, New York. Dalu, G . , 1985: Emittance effect of the remotely sensed sea surface temperature. Int. J. Remote  Sensing,  8, 5, 733-740.  Darnell, E. L. and Harriss, R.C., 1983: Satellite sensing capabilities for surface temperature and meteorological parameters over the ocean. Int. J. Remote Sensing, 4, pp. 65-92. Deschamps, P.Y. and Phulpin, T., 1980: Atmospheric correction of infrared measurements of sea surface temperature using channels at 3.7, 11, and 12 pm. Boundary Layer Meteorology,  18, pp. 131-143.  Deschamps, P.Y., Herman, M., and Tanre, D., 1983: Definition of atmospheric radiances and transmittances in remote sensing. Short communication, Remote Sensing of Environment, 13, pp. 89-92. El-Raey, M., 1983: An integral equation method for inversion of electromagnetic soundings. Remote  Sensing  of Environment,  13, pp. 403-408.  Emery, W.J. and Hamilton, K., 1985: Atmospheric forcing of interannual variability in the Northeast Pacific Ocean: Connection with El Nino. J. Geophysical Research, 00, 857-868. Fleming, H.E. and Crosby, D.S., 1971: A new treatment of the boundary term in the inversion of the radiative transfer equation. Remote Sensing of the Atmosphere: Inversion Methods and Applications, A . L . Fymat (ed). Fleming, Henry E. and McMillin, Larry M., 1977: Atmospheric transmittance of an absorbing gas 2: A computationally fast and accurate transmittance model for slant paths at different zenith angles. Applied Optics, 16. Fleming, H.E. and Smith, W.L., 1972: Inversion techniques for remote sensing of atmospheric temperature profiles. Geophysical and Astrophysical Temperature Measurements,  203, pp. 2239-2250.  Fritz, S, Wark, D.Q., Fleming, H.E., Smith, W.L., Jacobowitz, H., Hillery, D.T., and Alishouse, J.C., 1972: Temperature sounding from satellites. NOAA Technical Report NESS 59. 123  Fritz, Sigmund, 1981: Satellite temperature soundings (microwave satellite observations. Proc. IAMAP Symposium, Hamburg, E S A SP-165, pp. 25-28. Gautier, D. and Revah, I., 1975: Sounding of planetary atmosphere: a Fourier analysis of the radiative transfer equation, Journal of Atmospheric Science, 32, p. 881. Goody, R.M., 1952: A statistical model for water-vapor absorption. Quart. J. Royal Met. Soc., 78, 165-169. Grassl, H., Hinzpeter, H., 1975: The cool skin of the ocean. GATE-Report 14, Vol. I, WMO/ISCU, Geneva, 229-236. Grassl, H., 1976: The Dependence of the measured cool skin of the ocean on wind stress and total heat flux. Boundary Layer Meteorology,10, 465-474. Grassl, H., 1977: Gemessene strahlungs- und waermefluesse ueber dem ozean. Forschungsergebnisse, Reihe B, No 5, 42-50.  "Meteor 9  Gruber, Arnold and Watkins, Carmella Davis, 1982: Statistica Assessment of the quality of TIROS-N and NOAA-6 satellite soundings. Monthly Weather Review, 110, pp. 867-876. Haenel, G . , 1976: The properties of atmospheric aerosol particles as function of the relative humidity at the thermodynamic equilibrium with the surrounding air. Advances in Geophysics, 73-188, Academic Press. Hasse, L., 1971: The sea surface temperature deviation and the heat flow at the sea-air interface. Boundary Layer Meteorology, 1, 368-379. Hayden, C.M., 1979: Remote soundings of temperature and moisture. Quantitative orological data from satellites, WMO Tech note No. 166, 1-31.  mete-  Hayden, C.M., Smith, W.L., and Woolf, H.M., 1981: Determination of moisture from NOAA polar orbiting satellite sounding radiances, J. Appl. Meteor., 20, 450-466. Heintzenberg, Jost, Muller, Heinz, Quenzel, Heinrich, and Thomalla, Edith, 1981: Information content of optical data with respect to aerosol properties: numerical studies with a randomized minimization-search- technique inversion algorithm. Applied Optics, 20, No. 8. Hillger, D.W. and Vonder Harr, T . H . , 1979: An analysis of satellite infrared soundings at the mesoscale using statistical structure and correlation functions. J. Atm. Sci., 36, 287-305. Hillger, D.W., 1984: Spatial and temporal variations in mesoscale water vapor retrieved from T O V S infrared radiances in a nocturnal inversion situation, J. Climate and Applied Meteorology, 23, 704-723. Holyer, R. J . , 1983: Summary of a two-satellite method for measurement of SST including comparisons with ground truth and MCSST values, Satellite-Derived Sea Surface Temperature: Workship-II, J P L 84-5, p. 37 Imbault, D., Scott, N.A., and Chedin, A., 1981: Multichannel radiometric determination of sea surface temperature: parameterization of the atmospheric correction. Journal of Applied Meteorology, 20, pp. 556-564. 124  Jedlovec, G . J . , 1985: An evaluation and comparison of vertical profile data form the VISSR Atmospheric Sounder (VAS). J. Attn, and Oceanic Tech., 2, 559-581. Kaplan, Lewis D., 1959: Inference of atmospheric structure from remote radiation measurements. Journal of the Optical Society of America, 40, pp. 1004-1007. Katsaros, K., 1980: The aqueous thermal boundary layer. Boundary-Layer  Met.,  18, 107.  Kelly, G.A., Forgan, B.W., Powers, P.E., LeMarshall, J.F., Hassett, M., and O'Connor, B., 1983: A satellite-based operational system for upper air analysis in the Australian region. Remote Sensing of Environment, 13, 369-390. Kelly, K.A. and Davis, R.E., 1986: An analysis of errors in sea surface temperature in a series of infrared images from NOAA6. Journal of Geophysical Research, 01, 26332644. King, Jean I.F., 19 : Nonlinear Inversion: Theory and Praxis. Remote Sensing Atmosphere: Inversion Methods and Applications, A.L. Fymat (ed).  of  the  King, Jean I.F., 19 : Inversion methods and the obsolescence of differential equation for specifying physical observables. Remote Sensing of the Atmosphere: Inversion Methods and Applications, A . L . Fymat (ed). Kneizys, F.X., Shettle, E.P., Gallery, W.O.,Chetwynd, J.H. Jr., Abreu, L.W., Selby, J.E., Fenn, R.W., and McClatchey, R.A., 1980: Atmospheric transmittances/radiances: computer code L O W T R A N 5. AFGL-TR-80-0067, Air Force Geophysics Laboratory, Hanscom A F B , M A . Koenig, E.W., 1980: Performance of the HIRS/2 instrument on TIROS-N. Remote Sensing of Atmospheres and Oceans, A. Deepak (ed), Academic Press, New York, 641 pp. Lauritson, L., Nelson, G . J . , and Porte, F. W., 1979: Data extraction and calibration of TIROS-N/NOAA radiometers, NOAA Tech. Memo., T M N E S S 107, 81 pp. Lee, A . C , 1973: A study of the continuum absorption within the 8 — 13 pm atmospheric window. Quart. J. R. Met. Soc., 00, 490-505. Liou, Kuo-Nan, 1980: An introduction to atmospheric radiation, Academic Press, New York. Manual of Remote Sensing, 1983: 2nd Edition, American Society of Photogrammetry, Chapter 5. Mateer, C.L., 1965: On the information content of Umkehr Observations. Journal Atmospheric  Science,  22, p.  of  370.  Maul, George A. and Miriam Sidran, 1973: Atmospheric effects on ocean surface temperature sensing from the NOAA satellite scanning radiometer. Journal of Geophysical Research, 78, pp. 1909-1916. Maul, George A., 1981: Application of GOES visible-infrared data to quantifying mesoscale ocean surface temperature. Journal of Geophysical Research, 86, pp. 8007-8021. 125  Maul, G. A., 1983: Zenith angle effects in multichannel infrared sea surface remote sensing. Remote  Sensing  of Environment,  13, 439-451.  McAlister, E. D., McLeish, W., 1969: Heat transfer in the top millimeter of the ocean. Journal of Geophysical Research, 74, 3408. McCartney, Earl J . , 1983: Absorption and emission by atmospheric gases: The physical processes, John Wiley & Sons, Toronto. McClain, E. Paul, 1980: Results of global tests of a two-window method for satellitederived sea surface temperature, Workshop of Application on Existing Satellite Data to the Study of Ocean Surface Energetics, Space Science and Engineering Center. McClain, E. Paul, 1981: Multiple atmospheric-window techniques for satellite-derived sea surface temperature. Oceanography from Space, J.F.R. Gower (ed), Plenum Publishing Co., pp. 73-85. McClain, E, Paul, 1981: Split-window and Triple-window sea surface temperature determination from satellite measurements. Mini-symposium on Application of Aerospace Remote Sensing in Marine Research, ICES Statutory Meeting. McClain, E. Paul, Pichel, W.G., Walton, C.C., Ahmad, Z., and Sutton, J . , 1982: Multichannel improvements to satellite-derived global sea surface temperatures. Preprint X X I V COSPAR, Ottawa, Canada, Advances in Space Research, 2, pp. 43-47. McClain, E.P., Pichel, W.G., Walton, C . C . , Ahmad, Z. and Sutton, J . , 1983: Multichannel improvements to satellite-derived global sea surface temperatures, Advances in Space Research, 2 (6), 43-47, Pergamon, New York McClain, E.P., Pichel, W.G., and Walton, C . C . , 1985: Comparative performance of AVHRR-based multichannel sea surface temperatures. Journal of Geophysical Research, 0 0 , 11587-11601. McClatchey, Robert A., 1979: Atmospheric transmission modes and measurements. Atmospheric Effects on Radiative Transfer, Proc. Society of Photo-Optical Instrumentation Engineers, C.B. Ludwig (ed), 105, pp. 2-6. McClatchey, Robert A., Benedict, W.S., Clough, S.A., Burch, D.E., Cafee, R.F., Fox, K., Rothman, L.S., and Garing, J.S., 1973: A F C R L atmospheric absorption line parameters compilation, AFCRL-TR-73-0096. McMillin, L.M., Wark, D.Q., Siomkajlo, J.M., Abel, P.G., Werbowertski, A., Lauritson, L.A., Pritchard, J.A., Crosby, D.S.,Wolf, H.M., Luebbe, R.C., Weinreb, M.P., Fleming, H.E., Bittner, F.E., and Hayden, C.M.(1973), Satellite infrared sounding from NOAA spacecraft. N O A A Technical Report NESS 65, 112 pp. McMillin, Larry M., 1975: Estimation of sea surface temperatures from two infrared window measurements with different absorption. Journal of Geophysical Research, 8 0 , pp. 5113-5117. McMillin, Larry M., 1978: An improved technique for obtaining clear radiance from cloudcontaminated radiances. Monthly Weather Review, 108, pp. 1590-1597. 126  McMillin, Larry M., 1978: Temperature and moisture profiles from TIROS-N satellites: a description of the retrieval algorithms. Remote Sounding of the Atmosphere from Space, COSPAR Advances in Space Exploration, 4, H-J. Bolle (ed), Pergamon Press. McMillin, Larry M., Fleming, Henry E., and Hill, Michael L., 1979: Atmospheric transmittance of an absorbing gas 3: a computationally fast and accurate transmittance model for absorbing gases with variable mixing ratios. Applied Optics, 1 8 . McMillin, Larry M., 1980: The split window retrieval algorithm for sea surface temperature derived from satellite measurements. Remote Sensing of Atmospheres and Oceans, A. Deepak (ed), Academic Press, 641 pp. McMillin, Larry M. and Dean, Charles A., 1982: Variance ratios, loss of energy and regression in satellite temperature retrieval. Monthly Weather Review, 110, pp. 296299.  McMillin, L.M. and Crosby, D.S., 1984: Theory and validation of the multiple window sea surface temperature technique, Journal of Geophysical Research,^, C3, 3655-3661. MDA,  1983: METDAS System Design Manual., Volume I, MacDonald, Dettwiler and Associates.  Moyer, V., Scoggins, J.R., Chou, N.M., and Wilson, G.S., 1978: Atmospheric structure deduced form routine Nimbus-6 satellite data, Mon. Wea. Rev., 1 0 6 , 1340-1352. Newman, Wiliam I., 1979: The application of generalized inverse theory to the recovery of temperature profiles. Journal of the Atmospheric Sciences, 3 6 , pp. 559-565. Njoku, Eni G., 1985: Satellite-derived Sea surface temperature: Workshop comparisons, Bullet. Amer. Met. Soc., 6 6 , 274-281.  Ohring, G., 1979: Impact of satellite temperature sounding data on weather forecasts, Bull. Amer. Met. Soc, 6 0 , 1142-1147.  Olesen, F. S., Grassl, H., 1985: Cloud detection and classification over oceans at night with NOAA-7. International Journal of Remote Sensing, 6 . Parker, R.L., 1970: The inverse problem of electrical conductivity in the mantle. Geophys. J. Roy. Astron. Soc., 22, pp. 121-138.  Paulson, C A . and Simpson, J.J., 1981: The temperature difference across the cool skin of the ocean. Journal of Geophysical Researchfib,11,044-054.  Peckham, G., 1974: The information content of remote measurements of atmospheric temperature by satellite infrared radiometry and optimum radiometer configuration. Quart. J. Roy. Met. Soc., 1 0 0 , pp. 406-419.  Philips, N, McMillin, L.M., Gruber, A. and Wark, D., 1979: An evaluation of early operational temperature soundings from TIROS-N, Bull. Amer. Meteor. Soc, 6 0 , 10, 1188-1197.  Phillips, N., 1980: Two examples of satellite temperature retrievals in the North Pacific. Bull. Amer. Met. Soc, 6 1 , 712-717.  127  6  Phulpin, T . and Deschamps, P.Y., 1979: Estimation of sea surface temperature from AVHRR infrared channel measurements. Coastal & Marine Applications of Remote Sensing, A.P. Cracknell (ed), Proc. of 6th Ann. Conf. of the Remote Sensing Soc, U of Dundee. Pierlussi, Joseph H., Yomiyama, Ken, and Gomez, Richard B., 1979: Analysis of the L O W T R A N transmission functions. Applied Optics, 18. Roberts, Robert E., Selby, John E.A., and Biberman, Lucien M., 1976: Infrared continuum absorption by atmospheric water vapor in the 8-12 .urn window. Applied Optics, 15. Robinson, I.S., Wells, N.C. and Charnock, H., 1984: The sea surface thermal boundary layer and its relevance to the measurement of sea surface temperature by airborne and spaceborne radiometers, Int. J. Remote Sensing, 5, 1, 19-45. Prabhakara, C , Dalu, G . and Kunde, V . G . , 1974: Estimation of sea surface temperature from remote sensing in the 11- to 13- micron window region, Journal of Geophysical Research, 70, 33, 5039-5044. Rodgers, C D . , 1968: Some extensions and applications of the new random model for molecular band transmission. Quart. J. Royal Met. Soc., 04, 99-102. Rodgers, C D . , 1970: Remote sounding of the atmospheric temperature profile in the presence of cloud. Quart. J. Royal Met. Soc., 06, pp. 654-666. Rodgers, C D . , 1971: Some theoretical aspects of remote sounding in the earths atmosphere. J. Quant. Spectrosc. Radiat. Transfer, 11, pp. 767-777. Rodgers, C D . , 1976: Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation. Review of Geophysics and Space Physics, 14, pp. 609-624. Rodgers, C D . , 1976: The vertical resolution of remotely sounded temperature profiles with a priori statistics. Journal of Atmospheric Science, 33, p. 707. Saunders, P.M., 1967: Shadowing on the ocean and the existence of the horizon. Journal of Geophysical Research, 72, 18, 4643-4619. Schluessel, P. and Grassl, H, 1985: Improvement of AVHRR-multi-channel surface temperature with HIRS. COSPAR workshop on Satellite-Derived Surface Temperature for Global Climate Applications, May 28-31, 1985, Washington D.C. Schwalb, A., 1978: TIROS-N/NOAA A - G satellite series, NOAA Technical Memorandum NESS 97, 75 pp. Selby, J.E.A. and McClatchey, R.A., 1975: Atmospheric transmittance from 0.25 to 28.5 «um:| computer code L O W T R A N 3, AFCRL-TR-75-0255, Air Force Cambridge Research Laboratories. Selby, J.E.A., Kneizys, F.X., Chetwynd, J.H., and McClatchey, R.A., 1978), Atmospheric transmittance/radiance: computer code L O W T R A N 4, AFGL-TR-78-0053, Hanscom A F B , Mass. 01731. 128  Shaw, J.H., Chahine, M.T., Farmer, C.B., Kaplan, L.D., McClatchey, R.A., and Schaper, P.W., 1970: Atmospheric and surface properties from spectral radiance observations In the 4.3 micron region. Journal of Atmospheric Science, 27, p. 773. Shaw, J.H., 1970: Determination of the earth's surface temperature from remote spectral radiance observation near 2600cm~ . Journal of Atmospheric Science, 27, pp. 950959. l  Sidran, Miriam, 1980: Infrared sensing of sea surface temperature from space. Remote Sensing of Environment, 1 0 , pp. 101-114. Singh, S.M., 1984: Removal of atmospheric effects on a pixel by pixel basis from thermal Infrared data from instruments on satellites : The Advanced Very High Resolution Radiometer (AVHRR). Int. Journal of Remote Sensing, 5, pp. 161-183. Smith, Lewis L., Krassner, Jerry, Egan, Walter G., Hilgeman, Theodore W., and Selby, John. E.A., 1979: Recommanded modification of L O W T R A N 4 to include first order solar scattering, Atmospheric Effects on Radiative Transfer. Soc. Photo-Optical Inst. Engineers, C.B. Ludwig (ed), 1 0 5 , pp. 60-69. Smith W.L., 1970: Iterative solution of the radiative transfer equation for the temperature and absorbing gas profile of an atmosphere. Applied Optics, 0 . Smith W.L., Koffler, P.K., and Curtis, W.R., 1970: The determination of sea-surface temperature from satellite high resolution infrared window radiation measurements. Monthly Weather Review, 0 8 , pp. 604-611. Smith W.L., 1971: Calculation of clear-column radiances using airborne infrared temperature profile radiometer measurements over partly cloudy areas. NOAA Tech. Mem. NESS 28, 12 pp. Smith W.L. and Howell, H.B., 1971: Vertical distributions of atmospheric water vapor from satellite infrared spectrometer measurements. Journal of Applied Meteorology, 1 0 , p. 1026. Smith W.L., Woolf, H.M., and Fleming, H.E., 1972: Retrieval of atmospheric temperature profiles from satellite measurements for dynamical forecasting. Journal of Applied Meteorology, 1 1 , p. 113. Smith, W., Hilleary, D.T., Fischer, J . C , Howell, H.B., and Woolf, H.M., 1974: NIMBUS-5 ITPR experiment. Applied Optics, 1 3 pp. 449-506. Smith, W.L. and Woolf, H.M., 1976: The use of eigenvectors of statistical covariance matrices for interpreting satellite sounding radiometer observations. Journal of the Atmospheric Sciences, 33, pp. 1127-1140. Smith, W.L., Woolf, H.M., Hayden, C M . , Wark, D. Q., and McMillin, L. M., 1979: The TIROS-N Operational Vertical Sounder. Bulletin Amer. Met. Soc., 6 0 , 1177-1187. Smith, W.L., Woolf, H.M., Hayden, C M . , Schreiner, A . J . , and LeMarchall, J.M., 1983: The physical retrieval T O V S Export package. First International TOVS Study Conference, Igls, Austria. 129  Smith, William L., 1983: The retrieval of atmospheric profiles from VAS geostationary radiance observations. Journal of the Atmospheric Sciences, 40, pp. 2025-2035. Spankuch, D., Timofeyev, Y . M . , and Guldner, J., 1977: Comparison of different inversion methods for the determination of vertical temperature profiles from simulated satellite measurements in narrow spectral intervals. Zeitschrifts fur Meteorologie, 27, pp. 234-242. Staelin, David H., 1980: Progress in passive microwave remote sensing: nonlinear retrieval techniques. Remote Sensing of Atmospheres and Oceans, A. Deepak (ed), Academic Press, New York, p. 259. Stowe, L.L. and Fleming, H.E., 1980: The error in satellite retrieved temperature profiles due to the effects of atmospheric aerosol particles. Remote Sensing of Environment, 9, pp. 57-64. Strand, O. N., 1974: Theory and methods related to the singular-function expansion and Landwebers's iteration for integral equation of the first kind. SIAMJ. Numer. Anal., 11, 562-689. Strand, Otto Neall and Westwater, Ed R., 1975: A numerical comparison of two inversion techniques for remote sensing of atmospheric temperature profiles. Remote Sensing of the Atmosphere, ThA.5.1-ThA.5.4. Strong, A . E . and McClain, E.P., 1984: Improved ocean surface temperatures from space - Comparison with drifting buoys, Bull. Amer. Meteor. Soc., 65, 138-142. Strong, A.E., 1984: Drifting buoys used to maintain accuracy, Trop. Ocean Atmosphere Newsletter, 25. Susskind, J . and Resenberg, A., 1980: Temperature Retrievals from TIROS-N. Remote Sensing of Atmospheres and Oceans, A . Deepak (ed), Academic Press, New York. Susskind, J . , Rosenfield, J . , Reuter, D., and Chahine, M.T., 1982: The GLAS physical inversion method for analysis of HIRS2/MSU sounding data. NASA Technical Memorandum 84936. Thompson, Owen E., Eom, Jae K., and Wagenhofer, Joseph R., 1976: On the resolution of temperature profile fine structure by the N O A A satellite vertical temperature profile radiometer. Monthly Weather Review, 104. Thompson, Owen E., 1981: Transfer and distortion of atmospheric information in the satellite temperature retrieval problem. Proc. IAMAP Symposium, Hamburg, ESASP-165. Thompson, Owen E., 1982: HIRS - A M T S satellite sounding system test - theoretical and empirical vertical resolving power. Journal of Applied Meteorology, 21, p. 1550. Thompson, Starley L. and Warren, Stephen G., 1982: Parameterization of outgoing infrared radiation derived from detailed radiative calculations. Journal of the Atmospheric Sciences, 30, pp. 2667-2680. Twomey, S., 1970: Information content and indirect sensing measurements. Journal of Atmospheric Science, 27, p. 515. 130  Twomey, S., 1977: Introduction to the mathematics of inversion in remote sensing and indirect measurements. Elsevier Scientific Publishing Co., Amsterdam. Ulivieri, C , 1984: Minimization of atmospheric water vapor and surface emittance effects on remotely sensed sea surface temperature, IEEE Transactions on Gcosicencc and Remote Sensing, G E - 2 2 , 6, 622-626. Wallace, John M. and Hobbs, Peter V., 1977: Atmospheric science: An introductory survey, Academic Press, New York. Walton, C , 1980: Deriving sea surface temperatures from TIROS-N data. Remote Sensing of Atmospheres and Oceans, A . Deepak (ed), Academic Press, New York, pp. 547579. Walton, C , 1982: Recent improvements in deriving sea-surface temperatures from the NOAA-6 satellite system. IEEE trans. Geo. Remote Sensing, G E - 2 0 , 404-408. Wang, J.Y., 1974: On the estimation of low-altitude water vapor profiles from groundbased infrared measurements. Journal of the Atmospheric Sciences, 31, p. 513. Wang, J.Y. and Goulard, R., 1975: Numerical solutions in remote sensing. Applied 14.  Optics,  Wark, D.Q., 1961: On indirect temperature soundings of the stratosphere from satellites. J. Geophys. Res., 6 6 , p. 77. Wark, D.Q., Lienesch, J.H., and Weinreb, M.P., 1974: Satellite observations of atmospheric water vapor. Applied Optics, 13, pp. 507-511. Weinreb, Michael P. and Neuendorffer, Arthur C , 1973: Method to apply homogeneouspath transmittance models to inhomogeneous atmospheres. Journal of the Atmospheric Science, 30, pp. 662-666. Weinreb, Michael P., 1979: Atmospheric transmission for remote temperature sounding. Atmospheric Effects on Radiative Transfer, Proc. Society of Photo-Optical Instrumentation Engineers, 105, C.B. Ludwig (ed), pp. 22-30. Weinreb, Michael P. and Hill, Michael L., 1980: Calculation of atmospheric radiances and brightness temperatures in infrared window channels of satellite radiometers. N O A A Technical Report NESS 80. Weinreb, Michael P., Fleming, H.E., McMillin, L.M., and Neuendorffer, A.C., 1981: Transmittances for the TIROS Operational Vertical Sounder. NOAA Technical Report NESS 85, 60 pp. Werbowetzki, Adolf (ed), 1981: Atmospheric sounding user's guide, NOAA Technical Report NESS 83. Westwater, E.R. and Cohen, A., 1973: Application of Backus-Gilbert inversion technique to determination of aerosol size distributions from optical scattering measurements. Applied Optics, 12. Westwater, E.R. and Strand, Otto Neall, 1974: A generalized inversion program. N O A A Technical Report E R L 309-WPL 34. 131  Westwater, E.R., Sweezy, W.B., McMillin, L.M., and Dean, C , 1984: Determination of atmospheric temperature profiles from a statistical combination of ground-based and operational NOAA 6/7 satellite retrievals. J. Climat. and App. Met., 23, 689-703. White, Kenneth O., 1979: Water vapor continuum absorption in the infrared. Atmospheric Effects on Radiative Transfer, Proc. Society of Photo-Optical Instrumentation Engineers, 195, C.B. Ludwig (ed), pp. 14-21. Wilson, Wayne H., 1979: Measurements of atmospheric transmittance in maritime environment. Atmospheric Effects on Radiative Transfer, Proc. Society of Photo-Optical Instrumentation Engineers, 1 0 5 , C.B. Ludwig (ed), pp. 153-159. Zuev, V.E., 1970: Atmospheric transparency in the visible and the infrared. Translated from Russian, U.S. Department of Commerce.  132  Appendix A : Definition of Statistical Parameters Four main statistical parameters are used in the comparison of TOVS profiles and satellite derived SST. Bias is equivalent to a sample mean.  1=1  Root-mean-square difference (RMSD) is defined as  == \  RMSD  (A.2)  N  Standard deviation is defined as a square root of a sample variance.  STD  =  ZftXi-YJ-BIAS]* 1=1  N  (A.3)  N-l  Correlation is a measure of the "degree of association* between two random variables, X and Y.  CORR  £(*-X)(Y.-F) =  ,  , = l  V «=i where Xi = retrieved  temperatures  i=i  and Y% = in situ  133  temperatures.  (AA)  Appendix B : Definition of Meteorological Parameters The hydrostatic equation relates height above sea level into atmospheric pressure. It is defined as rw gpdz, (f?.i) where g is the acceleration due to gravity, p is the pressure, and p is the density of the air at height z. The ideal gas equation is written in the form  p = RpT,  (B.2)  where R is the gan constant for 1 kg of gas, T is the temperature, and p is the density of the gas. The saturation mixing ratios Q(p) with respect to water is defined as  Q(p) = 0 . 6 2 2 - ^ - , p — «,  (5.3)  where p is the total pressure and e, is the saturation vapor pressure. The dew point Td is the temperature to which air must be cooled at constant pressure in order for it to become saturated with respect to a plane surface of water. The relative humidity (RH) with respect to water is the ratio of the actual mixing ratio to the saturation mixing ratio with respect to water at the same temperature and pressure. It is defined as i Q(p) nn  iva  (at temperature T and pressure p) d  = IIHJ——————  -  —  -  —.  !  1  Q(p) (at temperature T and pressure p) The precipitable water vapor U is defined as U = g~ f "Q(p)dp. l  Pm  Jo  134  (S.5)  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items