Improved Model For Microwave Brightness Temperature Seen From Space OVer Calm Ocean

The Pennsylvania State University The Graduate School College of Engineering
AN IMPROVED MODEL FOR THE MICROWAVE BRIGHTNESS TEMPERATURE SEEN FROM SPACE OVER CALM OCEAN
A Thesis in Electrical Engineering by Sandra L. Cruz Pol 1998 Sandra L.Cruz Pol
Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 1998
S. Cruz-Pol
iii
ABSTRACT
An improved model for the microwave brightness temperature seen from space over calm ocean is presented. This model can be divided into two sub-models, the atmospheric absorption model and the ocean surface emissivity model. An improved model for the absorption of the atmosphere near the 22 GHz water vapor line is presented in the first part of this work. The Van-Vleck-Weisskopf line shape is used with a simple parameterized version of the model from Liebe for the water vapor absorption spectra and a scaling of the model from Rosenkranz for the 20-32 GHz oxygen absorption. Radiometric brightness
temperature measurements from two sites of contrasting climatological properties San Diego, CA and West Palm Beach, FL are used as ground truth for comparison with in situ radiosonde derived brightness temperatures. Estimation of the new models four parameters, related to water vapor line strength, line width and continuum absorption, and far-wing oxygen absorption, are performed using the Newton-Raphson inversion method. Improvements to the water vapor line strength and line width parameters are found to be statistically significant. The accuracy of brightness temperatures computed using the improved model is 1.3-2% near 22 GHz. In addition, the Hill line shape asymmetry ratio was evaluated on several currently used models to show the agreement of the data with the new model, and rule out atmospheric vapor absorption models near 22 GHz given by Waters and Ulaby, Moore and Fung which are based on the Gross line shape. In the second part of this work, a modified ocean emissivity model is presented. The
brightness temperature measured above the sea surface depends, among other things, on the oceans specular emissivity. We investigate the contribution to the brightness temperature from the specular ocean emission. For this purpose, satellite-based radiometric measurements from the
TOPEX/Poseidon project are employed together with near-coincident radiosonde profiles from fifteen (15) stations around the worlds oceans and TOPEX altimeter measurements for filtering of low wind conditions. The radiosonde profiles are used to compute the upwelling and downwelling emission and the opacity of the atmosphere. The radiative transfer equation is applied to the radiosonde profiles,
S. Cruz-Pol
iv
using the atmospheric model developed in the first part of this work, in order to account for atmospheric effects in the modeled brightness temperature. The dielectric properties of sea water are found from the modified Debye equation using salinity and sea surface temperature data from NODC ocean depth-profiles. The ocean complex permittivity model developed by Klein and Swift and, more recently, by Ellison is tested and revised. The average error in the modified emissivity model, over the range 18-40 GHz, is found to be 0.0037, which in terms of brightness temperatures, translates to a model error of approximately 1K.
S. Cruz-Pol
TABLE OF CONTENTS
LIST OF FIGURES .......................................................................................................................... LIST OF TABLES ............................................................................................................................. ACKNOWLEDGEMENTS ................................................................................................................. INTRODUCTION ................................................................................................................................... 1 1. THEORETICAL BACKGROUND ................................................................................................... 1 1-1 RADIATIVE TRANSFER EQUATIONS .................................................................................................. 1 1-1.1 Ground-Based Radiometer ...................................................................................................... 3 1-1.2 Satellite-Based Radiometer ...................................................................................................... 5 2. MICROWAVE ATMOSPHERIC ABSORPTION MODEL ....................................................... 12 2-1 ATMOSPHERIC ABSORPTION .......................................................................................................... 13 2-1.1 Line Shapes ............................................................................................................................ 15 2-2 CURRENT MODELS AND THEIR LIMITATIONS ................................................................................. 15 2-2.1 Atmospheric Absorption Line Shapes .................................................................................... 17 2-3 EXPERIMENT DESCRIPTION AND CALIBRATION ............................................................................. 22 2-3.1 Radiometer Data .................................................................................................................... 22 2-3.2 Radiosonde Data .................................................................................................................... 25 2-4 ANALYSIS AND RESULTS ............................................................................................................... 28 2-4.1 Hills Ratio Test ..................................................................................................................... 28 2-4.2 Parameter Estimation: Newton-Raphson Iterative Method................................................... 30 2-4.3 New Model Retrieved Parameters ......................................................................................... 31 2-4.4 Error Analysis ........................................................................................................................ 34 2-5 CONCLUSIONS ................................................................................................................................ 39 3. SEA SURFACE EMISSIVITY ........................................................................................................ 41 3-1 CURRENT MODELS AND THEIR LIMITATIONS .................................................................................. 41 3-1.1 Specular sea surface emissivity model ................................................................................... 41 3-1.2 Wind-roughened emissivity Model ......................................................................................... 44 3-1.3 Air-Sea Stability ..................................................................................................................... 47 3-2 DATA SETS ..................................................................................................................................... 49 3-2.1 TOPEX/Poseidon :Altimeter and Radiometer data ............................................................... 49 3-2.2 Radiosonde Data .................................................................................................................... 52 3-2.3 NODC Ocean Temperature and Salinity Profiles.................................................................. 55 3-3 ANALYSIS AND RESULTS ............................................................................................................... 59 3-3.1 Model for TB using raob, NODC, and altimeter data ........................................................... 59 3-3.2 Selection of the Maximum Time and Space Separation ......................................................... 59 3-3.3 Evaluation of the Model Performance ................................................................................... 62 3-3.4 Modified Dielectric Model Parameter Estimation................................................................. 64 3-3.5 Error Analysis ........................................................................................................................ 68 3-4 CONCLUSION .................................................................................................................................. 72 4. CONCLUSIONS ................................................................................................................................ 74 4.1 CASE STUDY: RELEVANCE OF THIS WORK TO THE TOPEX/POSEIDON ALTIMETRY MISSION ......... 74 4.2 CONCLUSIONS AND FUTURE WORK ................................................................................................. 79 REFERENCES ...................................................................................................................................... 82 APPENDIX A OXYGEN MICROWAVE SPECTRUM PARAMETERS ............................... 88
S. Cruz-Pol
vi
APPENDIX B GOFF-GRATCH FORMULATION FOR WATER VAPOR DENSITY AS A FUNCTION OF TEMPERATURE AND PRESSURE. .................................................................... 89 APPENDIX C TABLE OF MEAN, STANDARD DEVIATION AND COUNTS OF SEA SURFACE TEMPERATURE AND SALINITY PER MONTH PER RAOB STATION FOR THE PERIOD OF 1900 TO 1990. ........................................................................................................ 91 APPENDIX D FORTRAN PROGRAM FOR THE NEW ATMOSPHERIC MODEL .......... 93 APPENDIX E FORTRAN PROGRAM FOR THE MODIFIED OCEAN SURFACE EMISSIVITY MODEL ......................................................................................................................... 95
S. Cruz-Pol
vii
LIST OF FIGURES
FIGURE 1.1 THE POWER RECEIVED BY AN ANTENNA IS EQUIVALENT TO THE NOISE POWER DELIVERED BY A MATCHED RESISTOR. IF THE BLACKBODY ENCLOSURE IS AT TEMPERATURE T, THE BRIGHTNESS TEMPERATURE, TB, MEASURED BY THE ANTENNA IS DEFINED AS BEING EQUAL TO T. [CHANDRASEKHAR, 1960] ................................................................................................................... 2 FIGURE 1.2. PASSIVE REMOTE SENSING WITH (A) UPWARD-LOOKING RADIOMETER AND (B) DOWNWARDLOOKING RADIOMETER. ...................................................................................................................... 3 FIGURE 1.3. DOWNWELLING BRIGHTNESS TEMPERATURE AS OBSERVED BY AN UPWARD LOOKING RADIOMETER FOR U.S. STANDARD AND DRY ATMOSPHERIC CONDITIONS AND FOR THE HYPOTHETICAL CASE ABSOLUTELY NO WATER IN THE ATMOSPHERE. THE CONTRIBUTION AROUND 22.235 GHZ ARE MAINLY DUE TO THE RESONANT WATER-VAPOR EMISSION. .................................... 5 FIGURE 1.4. PASSIVE REMOTE SENSING WITH (A) UPWARD-LOOKING RADIOMETER AND (B) DOWNWARDLOOKING RADIOMETER. ...................................................................................................................... 6 FIGURE 1.5 EFFECTS ON THE SEA SURFACE EMISSIVITY DUE TO (A) FREQUENCY, (B) SEA SURFACE TEMPERATURE, (C) SALINITY AND (D) WIND SPEED. (UNLESS OTHERWISE NOTED, THE PLOTS ARE GIVEN FOR F= 37 GHZ, S= 35 AND TS=280 K.) .............................................................................. 9 FIGURE 2.1 WATER VAPOR ABSORPTION VERSUS FREQUENCY FOR AT A PRESSURE OF 1013 MBARS, AIR TEMPERATURE OF 290 K FOR NO WATER VAPOR IN THE ATMOSPHERE AND FOR WHEN THERE IS WATER VAPOR (2G/CM3). [ULABY ET AL., 1980] .............................................................................. 14 FIGURE 2.2 WATER VAPOR ABSORPTION VERSUS FREQUENCY FOR SEVERAL MODELS AT A PRESSURE OF 1013 MBARS, AIR TEMPERATURE OF 290 K AND RELATIVE HUMIDITY OF 50%. (L87 = LIEBE 87, L93 = LIEBE 93, W76 = WATERS 76 AND UMF81= ULABY, MOORE AND FUNG 81). ................. 16 FIGURE 2.3 (A) EFFECT OF LINE PARAMETER VARIATION BY 10% ON TOTAL ATMOSPHERIC ABSORPTION. (B) DIFFERENTIAL EFFECT OF LINE PARAMETER VARIATION WITH RESPECT TO NOMINAL L87R93 MODEL. THE ARROWS AT THE BOTTOM OF THE FIGURE INDICATE THE FREQUENCIES MEASURED IN THIS EXPERIMENT. (SAME ATMOSPHERIC CONDITIONS AS FIG. 2.2). ............................................... 21 FIGURE 2.4 ZENITH BRIGHTNESS TEMPERATURE INTERCOMPARISON BETWEEN RADIOMETER UNITS J1 AND J2 FOR ABSOLUTE CALIBRATION PURPOSES. MEASUREMENTS WERE CONDUCTED DURING DECEMBER 1991 AT SAN DIEGO, CA AT (A) 20.7 GHZ, (B) 22.2 GHZ AND (C) 31.4 GHZ. THE DATA HAVE BEEN SMOOTHED BY A 30 MINUTES RUNNING AVERAGE......................................................... 24 FIGURE 2.5 TOTAL ERROR IN CW LINE PARAMETER DUE TO A 0.5 K UNCERTAINTY IN MEASURED TB AND THE CORRECTION OF THE RAOB RELATIVE HUMIDITY READING, VERSUS NUMBER OF RAOB PROFILES USED IN THE ESTIMATION (SEE SECTION 3.2 FOR A COMPLETE DISCUSSION). A TRADE OFF BETWEEN
THE AMOUNT OF DATA USED AND THE MINIMUM ERROR IN PARAMETER ESTIMATION YIELDS AN OPTIMUM VALUE OF 21 RAOB PROFILES, PROVIDING A TOTAL OF 108 DATA POINTS. ....................... 27 FIGURE 2.6 HILL RATIO COMPARISON BETWEEN VARIOUS ATMOSPHERIC MODELS SHOWING AGREEMENT OF THE CHOSEN WATER VAPOR ABSORPTION LINE SHAPE WITH THE RADIOMETER DATA. (SEE TEXT FOR EXPLANATION OF MODELS' ACRONYMS).................................................................................... 29 FIGURE 2.7 BRIGHTNESS TEMPERATURE SPECTRA COMPARISON BETWEEN RADIOMETER DATA (WVR) AND RADIOSONDE-DERIVED DATA WITH NEW AND NOMINAL PARAMETERS FOR A VAPOR BURDEN OF (A) 2.9 G/CM2 , (B) 2.3 G/CM2, AND (C) 1.3 G/CM2. (NOTE THAT ONLY FIVE CHANNELS WERE IN OPERATION DURING THE RAOB LAUNCHES FOR CONDITIONS (B) AND (C) )....................................... 32 FIGURE 2.8 PLOT OF THE DIFFERENCE TB -TB L87R93 FOR (A) HUMID (WEST PALM BEACH), (B) MODERATE (SAN DIEGO) AND (C) DRY (SAN DIEGO) CONDITIONS. NOTE THAT TBLIEBE87 IS EQUIVALENT TO OUR NOMINAL MODEL (CL = CW = CX =1.0 AND CC = 1.2). (ONLY FIVE CHANNELS WERE IN OPERATION DURING THE RAOB LAUNCHES FOR CONDITIONS (B) AND (C) )......................... 33 FIGURE 2.9 PERCENTAGE ERROR IN THE IMPROVED MODEL FOR ATMOSPHERIC ABSORPTION USING THE 1962 U.S. STANDARD ATMOSPHERE AT SEA LEVEL WITH RH = 50%. THE ERROR IN THE IMPROVED MODEL MODEL ERRORS ARE DUE TO BIAS AND RANDOM MEASUREMENT UNCERTAINTIES IN RADIOMETER TB, THE CORRECTION FOR RAOB RELATIVE HUMIDITY VALUES LESS THAN 20% OR GREATER THAN 100%, AND THE UNCERTAINTY IN THE RADIOSONDE READINGS FOR PRESSURE, TEMPERATURE AND RELATIVE HUMIDITY......................................................................................... 37 FIGURE 2.10 PERCENTAGE ERROR IN THE IMPROVED DOWNWELLING TB FOR THE RADIOSONDE PROFILES USED BY THE ESTIMATION ALGORITHM. MODEL ERRORS ARE DUE TO BIAS AND RANDOM
S. Cruz-Pol
viii
MEASUREMENT UNCERTAINTIES IN RADIOMETER TB, THE CORRECTION FOR RAOB RELATIVE HUMIDITY VALUES LESS THAN 20% OR GREATER THAN 100%, AND THE UNCERTAINTY IN THE RADIOSONDE READINGS FOR PRESSURE, TEMPERATURE AND RELATIVE HUMIDITY. THE ERROR IN THE PREDICTED BRIGHTNESS LIES BETWEEN 1.5 % AND 2.0 %. ERROR BARS IN THE GRAPH REPRESENT THE STANDARD DEVIATION OF THE PERCENTAGE ERROR FOR ALL THE PROFILES USED IN THE ANALYSIS. ................................................................................................................................. 38 FIGURE 3.1 MECHANISMS RESPONSIBLE FOR THE MICROWAVE EMISSION OF A WIND-ROUGHENED SEA SURFACE INCLUDE LARGE GRAVITY WAVES, SMALL CAPILLARY WAVES AND SEA FOAM. ................ 46
FIGURE 3.2 WIND SPEED MODEL RELATING 0 TO WIND SPEED FOR THE MCW ALGORITHM AS CALIBRATED FOR TOPEX ALTIMETER. .............................................................................................. 51 FIGURE 3.3 LOCATION OF THE RADIOSONDE LAUNCH SITES. (SEE TABLE FOR COORDINATES). ............. 53 FIGURE 3.4 HISTOGRAM OF THE RANGE OF PATH DELAY VALUES FOR THE DATA USED IN THIS WORK. .. 54 FIGURE 3.5 AVERAGE SEA SURFACE TEMPERATURES VARIATION PER MONTH FOR STATION 9, LOCATED IN THE NORTH HEMISPHERE (BLUE), AND FOR STATION 29, LOCATED IN THE SOUTH HEMISPHERE (PINK). THE ERROR BARS REPRESENT THE STANDARD DEVIATIONS FOR EACH MONTH. ................... 56 FIGURE 3.6 AVERAGE SALINITY VARIATION PER MONTH FOR STATION 9, LOCATED IN THE NORTH HEMISPHERE (BLUE), AND FOR STATION 29, LOCATED IN THE SOUTH HEMISPHERE (PINK). THE ERROR BARS REPRESENT THE STANDARD DEVIATIONS FOR EACH MONTH. ....................................... 57 FIGURE 3.7 HISTOGRAMS OF THE RANGE OF (A) SALINITY AND (B) SEA SURFACE TEMPERATURE VALUES FOR THE DATA USED IN THIS WORK. ................................................................................................. 58 FIGURE 3.8 VARIATION OF THE NUMBER OF RAOB PROFILES USED DEPENDING ON THE LIMITS IN SPACE AND TIME SEPARATION IMPOSED ON THE DATA. ............................................................................... 60 FIGURE 3.9 VARIATION OF THE RMS DIFFERENCE BETWEEN DATA AND MODEL DEPENDING ON THE LIMITS IN SPACE AND TIME SEPARATION IMPOSED ON THE DATA. .................................................... 61 FIGURE 3.10. PLOT OF THE MODEL ERROR (TBTMR - TBMODEL) VERSUS THE SEA SURFACE TEMPERATURE 2 FOR E96. THE R VALUE OF THE LINEAR FIT IS SHOWN TO BE SMALL, DENOTING A SMALL DEPENDENCE OF THE ERROR IN THIS MODEL ON THE SEA SURFACE TEMPERATURE.. ........................ 63 FIGURE 3.11 THE MODIFIED AND NOMINAL OCEAN DIELECTRIC PERMITTIVITY MODELS, MODKS AND KS77(IN PINK) AND MODE AND E96 (IN BLUE). THE PLOTS SHOW THE VARIATION IN BOTH THE REAL AND IMAGINARY PARTS OF THE PERMITTIVITY VERSUS FREQUENCY. THE ERROR BARS DENOTE THE STANDARD DEVIATIONS IN THE MODIFIED MODELS. ALL PLOTS ARE FOR TSEA=280K AND S= 35. ................................................................................................................................... 71 FIGURE 3.12 ERROR IN THE MODIFIED OCEAN EMISSIVITY, MOD KS (IN PINK) AND MODE (IN BLUE) VERSUS FREQUENCY. THE ERROR BARS DENOTE THE STANDARD DEVIATIONS AT EACH POINT. PLOT IS FOR TSEA=280K AND S=35. ........................................................................................................ 72
S. Cruz-Pol
ix
LIST OF TABLES
TABLE 2.1 NEW RETRIEVED ATMOSPHERIC ABSORPTION PARAMETERS ............................................... 31 TABLE 2.2 STANDARD DEVIATIONS AND CORRELATION MATRIX FOR THE FOUR ESTIMATED PARAMETERS TAKING INTO ACCOUNT ERRORS IN THE RAOB PROFILES AND WVR BRIGHTNESS TEMPERATURE MEASUREMENTS. ...................................................................................................... 35 TABLE 2.3 UNCERTAINTIES INTRODUCED TO THE CALCULATED BRIGHTNESS TEMPERATURES BY THE L87R93 (NOMINAL) [JOHANSSON ET AL., 1987; ENGLAND ET AL., 1993] AND BY THE NEW ATMOSPHERIC ATTENUATION MODEL. .............................................................................................. 38 TABLE 3.1 COORDINATES OF THE RAOB STATIONS DEPICTED IN THE MAP ON FIGURE 3.3. .................. 53 TABLE 3.2 COMPARISON OF OVERALL PERFORMANCE OF SEVERAL OCEAN EMISSIVITY MODELS WITH RESPECT TO TMR DATA. .................................................................................................................. 64 TABLE 3.3 COMPARISON OF THE OVERALL PERFORMANCE OF K77 AND E96 OCEAN EMISSIVITY MODELS WITH TWO MODIFIED PARAMETER. ..................................................................................... 66 TABLE 3.4. COMPARISON AMONG OCEAN EMISSIVITY MODELS. ............................................................ 67 TABLE 4.1. ERROR BUDGET FOR THE PATH DELAY ALGORITHM ............................................................ 75 TABLE 4.2. TOTAL ERROR BUDGET FOR TOPEX MICROWAVE RADIOMETER (TMR) WET TROPOSPHERE RANGE CORRECTION. [KEIHM ET AL, 1995] ..................................................................................... 76 TABLE 4.3. RMS ERRORS OF INDIVIDUAL SEA SURFACE TOPOGRAPHY ERROR (UNITS IN CENTIMETERS [TSAOUSSI AND KOBLINSKY, 1994; FU ET AL., 1994] ......................................................................... 77 TABLE A-1. OXYGEN MICROWAVE SPECTRUM PARAMETERS [ROSENKRANZ, 1993] ............................ 88
S. Cruz-Pol
ACKNOWLEDGEMENTS
First of all, I want to thank God, for without Him there would be nothing. And thanks to my parents for raising me with love and for teaching me about Reincarnation and Allan Kardec. I admire you Papi for being so humble and for always being ready with a joke. I admire you Mami for your immense faith and noble example. The knowledge that our suffering and obstacles in life are a direct consequence of our acts in this or past lives, strengthens my Christian faith. Without that knowledge, so many questions about our purpose in life and what seems to be injustice sometimes would remain unanswered. I know I had to go through many situations good and bad, but in each I learned something. I know I had to be in this time and place, and thanks to that, I met a great human being, a great friend, Justin Bobak. Thanks JPB for answering all the hundred questions I always had, for being yourself, and for your friendship, good humor and afternoon coffee. I would also like to thanks all my other lab partners for making our working environment a fun place to come to every morning; to Hans Rosenberger for coffee every morning and for his good nature; and to Jude Giampaolo for being so nice and for the innumerable answers to my PC-networking, NT-registry, etc. questions. Thanks to Rafael Rodrguez Sols, because approximately 77.14% of what I know about computers, I learned from him. I would like to thank my advisor, Dr. Chris Ruf, for his valuable intellectual guidance through these last three and a half years of research. Thanks also go to all the members of my committee; Dr. Kultegin Aydin, Dr. Toby Carlson, Dr. Lynn Carpenter, Dr. Charles Croskey, Dr. Jenni Evans and Dr. Charles Kilgus for their suggestions and revisions to the final document. Thanks are due to Prof. Kwang Y. Lee for his well-meant advice and teaching, and to the sponsors behind the Fellowships from GEE and GEM, and the University of Puerto Rico. I will like to thank my daughters, Natal and Adriana, or as I like to call them my two most challenging and rewarding projects, for giving me inspiration and bringing me joy as I watch them bloom every day in front of me. I dont know who have learned more, you from me, or me from you. To my dear and wise mother-in-law, Vilma, for coming from Puerto Rico for a month, so we could work 12-15 hours a day, 6.3 days a week to finish this work in time, and for being a friend and providing advice every time I asked for some. And finally, I would like to thank my best friend, Jos Colom Ustriz, for being there all the time for me, for sincerely enjoying and sharing every one of my accomplishments. You are a great man, husband and father and I love you.
S. Cruz-Pol
xi
The vanity of some men, who imagine that they know everything, and are bent on explaining everything in their own way, will give rise to opposing opinions; but all who have in view the grand principle of Jesus will be united in the same love of goodness, and in a bond of brotherhood that will embrace the entire world. Putting aside all vain disputes about words, they will devote their energies to matters of practical importance, in regard to which, whatever their doctrinal belief, the convictions of all who receive the communications of the higher spirits will be the same. "Perseverance will render your labour fruitful. The pleasure you will feel in witnessing the spread of our doctrine and its right appreciation will be for you a rich reward, though perhaps rather in the future than in the present. Be not troubled by the thorns and stones that the incredulous and the evil-minded will place in your path; hold fast your confidence, for your confidence will ensure our help, and, through it, you will reach the goal. "Remember that good spirits only give their aid to those who serve God with humility and disinterestedness; they disown all who use heavenly things as a stepping-stone to earthly advancement, and withdraw from the proud and the ambitious. Pride and ambition are a barrier between man and God; for they blind man to the splendours of celestial existence, and God cannot employ the blind to make known the light." "JOHN THE EVANGELIST, ST AUGUSTINE, ST VICENT DE PAUL, ST LOUIS, THE SPIRIT OF TRUTH, SOCRATES, PLATO, FENELON, FRANKLIN, SWEDENBORG, etc., etc." from Spirits' Book by Allan Kardec, 1857 (translated from French), [www.GEAE.org]
Mi patria es el mundo, sin fronteras, SLCP
S. Cruz-Pol
INTRODUCTION
Knowledge of the state of the ocean plays a vital role in weather and ocean wave forecasting models [Wilheit, 1979a] as well as in ocean-circulation models [Dobson et al., 1987]. One approach to measuring the state of the ocean is by remote sensing of the oceans surface emission. Microwave radiometers on satellites can completely cover the earths oceans. Satellite radiometry offers numerous advantages over ship and buoy data. Some of these advantages include the vast coverage of global seas, including locations where radiosonde or buoys cannot be afforded, relatively low power consumption, no maintenance and continuous operation under a wide range of weather conditions. Measurements of the microwave brightness seen from the sea are used in the retrieval of physical parameters such as wind speed, cloud liquid water and path delay. A suitable model for these measurements includes contributions from atmospheric emission, mainly water vapor and oxygen, and from ocean emission. Seasat was the first satellite designed for remote sensing of the Earth's oceans. It was launched in 1978 by the National Aeronautic and Space Administration (NASA). The mission was designed to demonstrate the feasibility of global satellite monitoring of oceanographic phenomena and to help determine the requirements for an operational ocean remote sensing satellite system. It
included the Scanning Multichannel Microwave Radiometer (SMMR) which measured vertical and horizontal linearly polarized brightness temperatures at 6.6, 10.7, 18, 21 and 37 GHz. The SMMR was used to retrieve surface wind speed, ocean surface temperature, atmospheric water vapor content, rain rate, and ice coverage. Unfortunately, the mission only lasted approximately 100 days due to a failure of the vehicle's electric power system [Njoku et al.,1980]. The Defense Meteorological Satellite Program (DMSP) launched the first Special Sensor Microwave Imager (SSM/I) in 1987 on a near polar orbiting, sun synchronous weather satellite. It was the first of a series of several identical sensors launched to provide world-wide meteorological, oceanographic and solar-terrestrial physics measurements on a twice-daily basis [ Petty and Katsaros,
Ch. 1 Theoretical Background
S. Cruz-Pol
1992].
The SSM/I is a seven-channel, four frequency, linearly-polarized, passive microwave
radiometric system which operates at 19.35, 22.235, 37.0 and 85.5 GHz. It is used to measure ice coverage, precipitation areas and intensities, cloud water content, and ocean surface wind speeds [Hollinger et al., 1990]. In 1991 the European Space Agency launched The ERS-1 satellite. The primary mission of ERS-1 was to perform remote sensing of the Earth's oceans, ice caps, and coastal regions by providing global measurements of wind speed and direction, wave height, surface temperatures, surface altitude, cloud cover, and atmospheric water vapor levels. The mission included a nadir viewing radiometer operating at 23.8 and 36.5 GHz and co-aligned with the altimeter to provide range corrections with 2 cm accuracy [Gnther et al., 1993]. In 1992 the TOPEX/POSEIDON satellite was launched as a joint venture between NASA and Centre National d'Etudes Spatiale (CNES) to provide high-accuracy global sea level measurements. Data from TOPEX/Poseidon is used to map ocean circulation patterns, help understand how the oceans interact with the atmosphere, and improve our ability to predict the global climate [Stewart, 1986]. It includes a three channel nadir viewing microwave radiometer (TMR) at 18, 21 and 37 GHz designed to measure the water vapor along the path viewed by the altimeter to correct the altimeter data for pulse delay due to water vapor. It has a claimed accuracy of 1.2 cm [Keihm et al., 1995]. In 1998 the US Navy launched the GEOSAT Follow On (GFO), designed to provide real-time ocean topography data. It includes a radar altimeter with 3.5 cm height measurement precision. In addition, a dual frequency (22 and 37 GHz) water vapor radiometer is included to provide path delay correction with an accuracy of 1.9 cm [Ruf et al., 1996]. The need to improve the calibration of existing models for atmospheric and ocean emission is motivated by several current and upcoming satellite remote sensing missions. In the case of TMR, an improved atmospheric model would enhance the inversion algorithm used to retrieve path delay information. Another case is the JASON satellite, a joint NASA/CNES radiometer and altimeter scheduled to be launched in 2000 [JPL, 1998]. For JASON, absolute calibration is performed by occasionally looking at calm water. This type of calibration reduces the cost in hardware, complexity,
S. Cruz-Pol
size and power. However, the quality of the calibration depends strongly on the accuracy of a model for the calm water emission. In contrast, for the TMR an absolute calibration is performed using hot and cold references carried by the satellite [Ruf et al., 1995]. Errors in the modeling of microwave brightness temperature, TB, seen from orbit over the sea include errors in the models for vapor and oxygen absorption and sea surface emissivity. Conversely, errors in the measurement of the microwave TB include errors in the antenna temperature calibration, and beam pattern correction. Currently, the dominant error source when modeling the ocean In the case of the TOPEX/POSEIDON
brightness temperature is the vapor absorption model.
microwave radiometer, this uncertainty is approximately 35% higher than the radiometers TB measurement error [Keihm et al., 1995]. Precise microwave radiometry equipment such as this
demands more accurate models for the retrieval of the oceans parameters. The accuracy of these models must be consistent with the level of the errors introduced by the microwave sensor, otherwise the model uncertainties dominate the error budget. The improvement and revision of two models needed to achieve a higher accuracy in the ocean TB modeling are addressed in this work. The first model accounts for atmospheric absorption. The second accounts for the sea surface emissivity. In this document, a section is devoted to each of these models. In Part I, the development of an improved microwave atmospheric absorption model is presented. Part II is dedicated to ocean microwave emission. In both cases, a model is developed and iteratively adjusted to fit a carefully calibrated set of measurements. For the atmospheric absorption model, ground-based radiometric experiments were conducted at two locations of contrasting humidity conditions; San Diego, CA and West Palm Beach, FL. In addition, radiosonde profile data at each site were collected for comparison purposes in the retrieval of the atmospheric model parameters. Advantages over previous such experiments include the use of three independent radiometers for absolute calibration verification, sampling at eight distinct frequencies across the 22 GHz absorption line, and filtering of the raob data to minimize the effects of errors in the relative humidity readings.
S. Cruz-Pol
Uncertainties in the improved model for atmospheric emission are significantly improved over previous published models. The line-strength and width parameters' uncertainties are reduced to 1% and 1.6%, respectively. The overall uncertainty in the new absorption model is conservatively estimated to be 3% in the vicinity of 22GHz and approaching 8% at 32 GHz. The RMS difference between modeled and measured thermal emission by the atmosphere, in terms of the brightness temperature, is reduced by 23%, from 1.36 K to 1.05 K, compared to one of the most currently used atmospheric models. For the ocean emission study, satellite-based radiometric measurements from the TOPEX/Poseidon project are employed. In addition, altimeter (active remote sensor) data from the same satellite is utilized for the purpose of wind speed estimation and specular emissivity corroboration. We investigate the contribution from the specular ocean emission by employing the altimeter to pinpoint the exact times when the wind is calm, in order to relax the dependence of the correction to the specular model on the accuracy of the wind model. The modified ocean dielectric models exhibit significant improvements in the estimate of TB. Of the two, the modified Ellison et al.[1977] model exhibits superior overall performance, including the lowest bias at both frequencies, which is a very important attribute indicative of the accuracy of the model. Its frequency dependence was decreased to 0.30K, which will allow for more reliable
extrapolation to higher frequencies. In addition, this modified model has the lowest dependence on sea surface temperature and the lowest RMS difference for both 18GHz and 37GHz. Consequently, this is the model that we recommend for future remote sensing applications involving microwave emissions from the ocean emissivity of the ocean. The average error in the modified emissivity model, over the range 18-40 GHz, is found to be 0.37%, which in terms of brightness temperatures, translates into a model error of approximately 1K. We first develop the necessary background theory in Chapter 1. Chapter 2 deals with the model theory, experiments and data analysis related to the atmospheric absorption model. The third chapter presents the model theory, data, statement of the problem, and analysis for the ocean emission model. Conclusions are presented in Chapter 4.
S. Cruz-Pol
S. Cruz-Pol
1. THEORETICAL BACKGROUND
1-1 Radiative Transfer Equations

The atmosphere receives most of its energy by means of solar electromagnetic radiation. Some of this energy is absorbed by the atmosphere and some reaches the surface of the Earth where it can also be absorbed or it can be reflected. Energy absorption implies a rise in thermal energy and, therefore, temperature of the object. Any object with a temperature above absolute zero emits
electromagnetic radiation. Electromagnetic emission implies a decrease in the objects temperature. These processes, i.e. absorption and emission, altogether help create a balance between the energy absorbed by the Earth and its atmosphere and the energy emitted by them. The study of these energy transformation processes is called radiative transfer. The Planck function for spectral brightness describes the radiation spectrum of a blackbody at thermal equilibrium. It is given by
(1.1)
where h is Plancks constant (6.63 x 10-34 J), f is frequency in Hz, k is Boltzmanns constant (1.38 x 1023
J/K), T is absolute temperature in K, and c is the velocity of light (3 x 10 8 m/s) [Planck, 1914]. At the low-frequency limit, i.e. hf << kT, the Rayleigh-Jeans approximation is valid and the
Planck function can be simplified to [Ulaby et al., 1981],
(1.2)
The above expression is very significant, since it shows a linear relationship between the Planck spectral brightness and the physical temperature of an object. This expression is valid for frequencies
T Antenna
S. Cruz-Pol
Radiometer Radiometer R Receiver Receiver smaller than 300 GHz. For frequencies less than 117GHz and blackbody at 300 K, the Rayleigh-Jeans T approximation yields values less than 1% different from the Plancks expression. At microwave frequencies, therefore, the power received by an antenna due to blackbody radiation is directly Blackbody enclosure proportional to the temperature of the object. Radiometers are used to detect the electromagnetic radiation naturally emitted by the atmosphere, oceans, terrain, celestial bodies, etc. These devices remotely sense the microwave noise power in terms of the apparent temperature seen by the antenna. The apparent temperature is defined as the equivalent blackbody temperature at which a matched resistor would have to be in order to deliver the same amount of power at the antenna terminals (see Fig 1.1) [Chandrasekhar, 1960].
Figure 1.1 The power received by an antenna is equivalent to the noise power delivered by a matched resistor. If the blackbody enclosure is at temperature T, the brightness temperature, TB, measured by the antenna is defined as being equal to T. [Chandrasekhar, 1960]
The amount of radiation received by a radiometer depends on its viewing configuration. A ground-based radiometer operates in an upward-looking position. Therefore, it measures the radiation coming down from our atmosphere (see Fig. 1.2). A satellite-based radiometer, on the other hand, looks down toward the surface of the Earth. It receives contributions from the upwelling atmospheric radiation, the surface emission, and the radiation reflected at the air-surface interface (see Fig. 1.4).
S. Cruz-Pol
Figure 1.2. Passive remote sensing with upward-looking radiometer .
1-1.1 Ground-Based Radiometer

Consider a non-scattering atmosphere in which the temperature and the absorption coefficient are functions of altitude. The brightness temperature of the atmosphere, observed from the ground, also known as downwelling temperature, is obtained by integrating the contributions of individual layers of the atmosphere. The emission from each layer is attenuated by a factor e- by the intervening
medium as it travels down towards the point of measurement. The downwelling temperature is the sum of these contributions and it is given by [Ulaby et al., 1981]
(1.3)
S. Cruz-Pol
where is the incidence angle of the radiation which is measured with respect to the normal to the surface, (f, z) is the atmospheric attenuation in Nepers/km at frequency f and height z, is the opacity of the atmosphere between altitude 0 and z , and T(z) is the air temperature at height z. The opacity measures the total amount of extinction suffered through the path and is given by
(1.4)
When the radiometer is looking directly up, the incidence angle is = 0o and sec reduces to unity. This simplifies (1.3) to the form used in this work,
(1.5)
To compute the apparent temperature measured by the antenna, Ta, the contributions from the cosmos and galaxy have to be added (1.6) In the above equation, TC is the cosmic background radiation incident on the atmosphere from the top. As it travels down toward the antenna, it is reduced by a factor of e- due to absorption by the
intervening atmosphere. This factor is known as the transmissivity or transmittance function. The cosmic radiation at microwave frequencies varies with frequency as (1.7) which has an average of 2.78 K for the 20-32 GHz range. The frequency dependence accounts for departures from the Rayleigh-Jeans approximation [Janssen, 1993]. The physical behavior of equation (1.3) depends on the atmospheric conditions. Examining (1.3) and (1.4), it is noted that Ta depends on the vertical profiles of temperature and absorption. For a lossless atmosphere, (f, z)=0, and therefore TDN = 0. At the other extreme, for a perfect blackbody, (f, z)=1, and the TDN is reduced to the
S. Cruz-Pol
physical temperature of the atmosphere. For a non-uniform temperature profile, Ta will never exceed the physical temperature. Figure 1.3 shows plots of Ta for different atmospheric conditions.
Figure 1.3. Downwelling brightness temperature as observed by an upward looking radiometer for U.S. Standard and dry atmospheric conditions and for the hypothetical case absolutely no water in the atmosphere. The contribution around 22.235 GHz are mainly due to the resonant water-vapor emission.
1-1.2 Satellite-Based Radiometer

Again consider a non-scattering atmosphere. The apparent temperature received by an
antenna from above the atmosphere has three components. The radiometer measures the radiation coming up from the atmosphere, also known as the upwelling temperature, the radiation emitted by the surface beneath and the radiation reflected at the air-surface interface (Fig. 1.4). This last component consists of the downwelling atmospheric emission plus the cosmic background radiation. The total apparent temperature is given by
S. Cruz-Pol
Figure 1.4. Passive remote sensing with downward-looking radiometer.
(1.8) where Ts is the thermodynamic temperature of the surface in Kelvin, is the emissivity of the surface,
is the reflectivity of the surface, H is the satellite height in km, TC is the cosmic radiation and TDN is given by (1.3). The upwelling brightness temperature in the zenith direction is given by,
(1.9)
The absorption coefficient, (f, z), includes both the spectral (water vapor and oxygen) line absorption and any other source of microwave opacity present in the atmosphere as clouds, fog and rain.
S. Cruz-Pol
Equation (1.8) contains all the quantities needed to compute the response of a satellite-based microwave radiometer to changes in atmospheric and surface variables. The upwelling and
downwelling temperature terms depend on the atmospheric absorption weighted by the temperature profile of the atmosphere. The background term, TC, is multiplied by the transmittance function, to account for the attenuation along the vertical path between the satellite and the surface. It is then added to TDN and reflected back toward the satellite. Hence, it is attenuated by the intervening atmosphere, according to the transmittance function, and multiplied by the reflectivity parameter, . The surface emission consists of the physical temperature of the surface multiplied by the surface emissivity. The emissivity is a function of the dielectric properties of the surface. In the case of the ocean, it varies with frequency, temperature, wind speed and, to a lesser extent, on salinity. The variation of emissivity with frequency, temperature, and salinity is shown in Figures 1.5(a)-(c) as described by the most recent ocean emissivity model [Ellison et al., 1996]. Variation with wind speed is shown in Figure 1.5(d) as described by Wilheit [1979b]. Note how the dependence on sea surface temperature and frequency are much larger than that on salinity and wind speed. In general, the emissivity increases with increasing frequency and decreasing temperature. The frequency, salinity and temperature dependence will be examined further in Chapter 3.
S. Cruz-Pol
S. Cruz-Pol
Figure 1.5 Effects on the sea surface emissivity due to (a) frequency, (b) sea surface temperature, (c) salinity and (d) wind speed. (Unless otherwise noted, the plots are given for f= 37 GHz, S= 35 and TS=280 K.)
S. Cruz-Pol
10
1-1.2.1 Radio path delay
Radio path delay is the path distortion that a radio signal undergoes when traversing the atmosphere of the Earth. This refraction introduces uncertainties in the time of arrival of the signal due to bending and retardation along the propagation path. The physical phenomena behind the delays are the dispersion due to the free electrons of the ionosphere, the induced dipole moments of neutral atmospheric molecules (principally nitrogen and oxygen), and the permanent dipole moments of watervapor and cloud liquid water molecules. The dry component of the delay due to the neutral
atmosphere contributes about 2.3 m in zenith direction at sea level. The wet component of the delay caused by water vapor and cloud liquid water is smaller, but much more variable. It can range from less than 1 cm to 40 cm or more in the zenith direction. When we mention path delay in the remainder of this work, we refer to the variable wet component. The wet delay can be inferred from microwave radiometer measurements. The electrical path length L of a signal propagating along S is defined as (1.10)
where n is the refractive index of the atmosphere, and S is the path along which the signal propagates. The signal will propagate along the path that gives the minimum value of L. S is larger than the geometrical straight line distance, defined as G. However, the electrical path length of the signal propagating along G is longer than that for the signal propagating along S. The difference between the electrical path length and the geometrical straight-line distance is called the excess propagation path or path delay, defined as (1.11) In the case of a horizontally stratified atmosphere, the two paths S and G are identical in the zenith direction. In this case, the path delay is (1.12) The index of refraction, n, is conveniently expressed in terms of the refractivity, N, defined as
S. Cruz-Pol
11
(1.13) The refractivity is expressed as the sum of a dry and wet (vapor and cloud induced) components. The wet refractivity component due to vapor is given by [Bourdouris, 1963; Hills et al., 1982] (1.14) from which the vapor induced path delay component in centimeters can be found as
(1.15)
where v is the water vapor density in g/m3, T is the atmosphere air temperature profile in K and z is height in meters.
S. Cruz-Pol
12
2. Microwave Atmospheric Absorption Model

An improved model for the absorption of the atmosphere near the 22 GHz water vapor line is presented. The Van-Vleck-Weisskopf line shape is used with a simple parameterized version of the model from Liebe for the water vapor absorption spectra and a scaling of the model from Rosenkranz for the 20-32 GHz oxygen absorption. Radiometric brightness temperature measurements from two sites of contrasting climatological properties San Diego, CA and West Palm Beach, FL were used as ground truth for comparison with in situ radiosonde derived brightness temperatures. The retrieval of the new models four parameters, related to water vapor line strength, line width and continuum absorption, and far-wing oxygen absorption, was performed using the Newton-Raphson inversion method. In addition, the Hill line shape asymmetry ratio was evaluated on several currently used models to show the agreement of the data with the new model, and rule out atmospheric vapor absorption models near 22 GHz given by Waters and Ulaby, Moore and Fung which are based on the Gross line shape. Results show a 23% improvement in the difference between modeled and measured brightness temperature from the atmosphere compared to current models.
Absorption and emission by atmospheric gases can significantly attenuate and delay the propagation of electromagnetic signals through the Earths atmosphere. Improved modeling of the emission spectra for the dominant contributing gases, mainly water vapor and oxygen, is needed for many applications in communications, remote sensing, and radioastronomy. More precise models of atmospheric absorption can improve corrections for atmospheric effects on satellite observations of land and ocean surfaces [McMillin, 1980], produce more accurate remote measurements of atmospheric water vapor burden and temperature profiles [Grody, 1980], refine predictions of global climate changes [NASA, 1993], enhance planetary radio science measurements [Pooley, 1976], improve the accuracy of continental plate motion estimation [Shapiro et al., 1974], and expedite the resolution of accumulated strain at fault zones [Shapiro, 1976]. Estimated uncertainties for current models of the 20-32 GHz water vapor absorption range over 4-10% [Keihm et al., 1995]. For applications such as water vapor radiometer (WVR) measurements of integrated vapor and cloud liquid used in meteorological and climate modeling, 10% accuracies are often adequate. However, for many applications, especially those requiring calibration of microwave signal delays in the troposphere (path delay), the vapor absorption model uncertainty often dominates experimental error budgets. Examples include the measurement of the vapor-induced path delay over the worlds oceans by the TOPEX Microwave Radiometer [Keihm et al., 1995] and GEOSAT Follow-on Water Vapor Radiometer [Ruf et al., 1996], VLBI geodetic measurements
S. Cruz-Pol
13
[Linfield et al., 1996], and the tropospheric calibration effort planned for the Cassini Gravitational Wave Experiment [Keihm and March, 1996]. Current models for atmospheric microwave absorption have been developed from both laboratory and field experiments. Multimode cavity measurements by Becker and Autler [1946] lacked diagnostics necessary to control systematic errors down to a desirable level; their estimated error is between 5% and 10% [Walter, 1995]. The field measurements of water vapor absorption calibration near the 22 GHz resonance feature generally involve comparisons between direct measurements at two or three selected frequencies by a water vapor radiometer and theoretical brightness temperatures calculated from radiosonde (raob) profiles of temperature, pressure and humidity. raob comparison measurements [e.g. Westwater, 1978; Hogg et al., 1980; Snider, 1995] are well known to be subject to inaccurate humidity readings for extreme (very dry or very humid) conditions [Wade, 1994; Nash et al., 1995]. Current oxygen model uncertainties range from 1.5% to 8% for the resonant oxygen lines cluster near 58 GHz [Liebe et al. 1993], yet these fractional uncertainties increase significantly for frequencies down in the 20-32 GHz range. In this work, radiometric measurements of brightness temperature, TB, are used in conjunction with in situ raob measurements to estimate parameters for a simplified model of the 20-32 GHz atmospheric vapor and oxygen absorption. The experiment covered ~70 days of near-continuous WVR measurements and twice per day raob launches at the San Diego and West Palm Beach National Weather Service stations. Advantages over previous such experiments include the use of three
independent radiometers for absolute calibration verification, sampling at eight distinct frequencies across the 22 GHz absorption line, and filtering of the raob data to minimize the effects of high and low end errors in the relative humidity measurements. The spectrum of atmospheric absorption versus frequency is shown for dry and typical conditions on Fig. 2.1
2-1 Atmospheric Absorption
Ch. 2 Microwave Atmospheric Water Vapor Absorption Model
S. Cruz-Pol
14
For a non-scattering atmosphere (rain-free conditions) the main contributors to atmospheric radiation at microwave frequencies are the emission by water vapor and oxygen molecules. The movement of molecules, through vibration, rotation and electronic spinning, determines the nature of the radiation that they emit or absorb. Uncoupled electron charges or electron spins are responsible for electric and magnetic dipoles, respectively, within the molecules.
Figure 2.1 Water vapor absorption versus frequency for at a pressure of 1013 mbars1, air temperature of 290 K for no water vapor in the atmosphere and for when there is water vapor (2g/cm3). [Ulaby et al., 1980]
Water vapor and oxygen molecules both radiate due to molecular rotation. In the case of the water vapor molecule, this rotation causes its permanent electric dipole to radiate. Oxygen has no permanent electric dipole moment, but its magnetic dipole allows it to emit. The magnetic dipole emission is much weaker than that of a molecular electric dipole, but the great abundance of oxygen in the atmosphere compensates for the intrinsic weakness of its absorption.
In this work pressure is expressed in mbars which is equal to 1 hPascal.
S. Cruz-Pol
15
2-1.1 Line Shapes

We are concerned with the absorption line spectra of the water vapor and oxygen molecules, since they are the major contributors to atmospheric absorption. Their absorption spectra have been most commonly described by the Van-Vleck Weisskopf (VVW) and the Gross line shapes. Both line shapes are derived from a molecular oscillator analogy. In this analogy, the molecule is treated as a classical oscillator with a fixed rotational frequency equal in value to the frequency of the resonant line. Collisions between molecules cause reorientation and rotational phase shifts of the molecule, which contribute to the broadening and shifting of the electromagnetic spectral lines. The basic difference between the two line shape theories is that VVW assumes the oscillations are in phase with the electric field after the collisions, whereas Gross assumes that the molecular oscillation phases stay undisturbed and the post-collision momenta are randomized [Ben-Reuven, 1969]. The general form for the Van-Vleck Weisskopt [1945] is given by
(2.1)
and the Gross Line shape is given by [Gross, 1955]
(2.2)
where f is the measurement frequency, f0 is the center resonant frequency for the given molecule, S is the line strength and is the linewidth parameter. The line shape used in the atmospheric absorption model employed in this work is the VVW and is discussed with further detail in section 2-2.1.
2-2 Current Models and their limitations

Currently used models for atmospheric absorption include those by Liebe and Layton [1987], Liebe et al. [1993], Waters [1976] and Ulaby et al. [1981]. These will be referred to as L87, L93, W76
L93 S. Cruz-Pol L87 UMF81 W76 and UMF81, respectively, in the remainder of this thesis. Figure 2.2 shows a plot of the water vapor absorption spectrum for each of these models under typical mid-latitude surface conditions. Note that significant differences are evident in both magnitude and shape of the spectra. Some of the reasons for these differences are discussed below. 16
0.045
Water Vapor Absorption vapor [Np/km]
0.040 0.035 0.030 0.025 0.020 0.015 0.010
Figure 2.2 Water vapor absorption versus frequency for several models at a pressure of 1013 mbars, air temperature of 290 K and relative humidity of 50%. (L87 = Liebe 87, L93 = Liebe 93, W76 = Waters 76 and UMF81= Ulaby, Moore and Fung 81). Each model includes an empirical continuum term to account for the discrepancy between theoretical and experimental absorption spectra in the window region. The physical phenomena
behind the excess absorption in the continuum might be due to inaccuracies in the far wing line shape of vapor resonances [Gebbie, 1980], the exclusion of the effects of water clusters [Bohlander, 1979] and/or forbidden transitions between energy levels on these line functions [Rosenkranz, 1990].
20
22
24
Frequency [GHz]
26
28
30
32
S. Cruz-Pol
17
Although this excess has yet to be understood, empirical modifications are needed to obtain more accurate agreement between measurements and theory. Uncertainties in attenuation for the L87 water vapor model over tropospheric pressures are estimated to be from 2.3 % to 21.2% [Liebe and Layton, 1987]. For the oxygen model used in L87R93 and L93, the uncertainties range from 1.5% to 8% at the nominal center frequency of 58 GHz, also for tropospheric pressures [Liebe et al., 1993]. For frequencies farther from the center, the fractional uncertainty increases significantly.
2-2.1 Atmospheric Absorption Line Shapes

The L87 and L93 models employ the Van Vleck-Weisskopf (VVW) line shape whereas the W76 model uses the Gross line shape. The UMF81 uses the Gross line shape for the water vapor and VVW for the oxygen absorption. Both line shapes are derived from a molecular oscillator analogy. In this analogy, the molecule is treated as a classical oscillator with a fixed rotational frequency equal in value to the frequency of the resonant line. Collisions between molecules cause reorientation and rotational phase shifts of the molecule, which contribute to the broadening and shifting of the electromagnetic spectral lines. The basic difference between the two line shape theories is that VVW assumes the oscillations are in phase with the electric field after the collisions, whereas Gross [1955] assumes that the molecular oscillation phases stay undisturbed and the post-collision momenta are randomized [Ben-Reuven, 1969].
2-2.1.1 The L87R93 Atmospheric Absorption Model

The baseline model which is refined in this work uses a simplified version of the L87 model for water vapor absorption, in which we have combined the effects of other individual water vapor absorption lines, above the 22.235 GHz line, into the continuum term for computational expediency. We have incorporated the simplified L87 model for water vapor absorption together with the improved
S. Cruz-Pol
18
oxygen absorption model by Rosenkranz [1993]. This model is referred to as L87R93. Refinements to the water vapor absorption model are accomplished by the addition of three adjustable parameters, CL, CW, and CC, which account for scaling of the line strength, line width, and continuum term, respectively. The oxygen absorption model is refined with the addition of an adjustable scaling factor C X. Equations for the L87R93 atmospheric absorption model, including all refinement parameters, are presented below. The water vapor absorption model is given by (2.3) where TL, TS, and TC refer to the line strength, line shape and continuum terms and are given by , (2.4)
(2.5)
and, (2.6)
where f is frequency in GHz, fz is the water vapor resonant frequency = 22.235 GHz, = 300/T, T is air temperature in Kelvin, P is air pressure, PH2O is water vapor partial pressure and Pdry = P - PH2O . The first term in equation (2.6) is due to collisions of the water vapor molecule with foreign molecules like oxygen or nitrogen, while the second term relates to the collision among water molecules. The width parameter, , is defined as . (2.7)
Equations (2.3)-(2.7) introduce the following parameters: water vapor line strength CL, line width CW, and continuum CC. The oxygen absorption model is given by
S. Cruz-Pol
19
(2.8)
where
, S(T) is the line strength (2.9)
and S(To) and fn are listed in Table A-1 on Appendix A, fn is the nth oxygen resonant frequency and Ln is proportional to the shape of the lines
(2.10)
The pressure-broadened line half-width is, (2.11)
The O2 resonant lines are very close to each other and troposphere pressures are high enough ( > 100 mbars) to cause the lines to broaden and overlap. This is called collisional broadening and is taken into account through the interference parameter, Yn, defined as . (2.12)
where the coefficients y and v are listed in Table A-1. For the non-resonant spectrum (n=0), the summation is given by
(2.13)
where (2.14)
S. Cruz-Pol
20
and where w is also listed in Table A-1 on Appendix A for the n resonant lines. In equations (2.8)(2.14), f is frequency in GHz, = 300/T, T is air temperature in Kelvin, w is water vapor density in g/ m3, P is air pressure, PH2O is water vapor partial pressure given by e = T/217 and Pdry = P - PH2O . A selection of CL =1.0, CW =1.0, CC =1.2 and CX =1.0 yields values within 0.5% of the exact L87 water vapor absorption and Rosenkranz [1993] oxygen absorption models over 20-32 GHz. (Note that the value of the parameter CC has been increased from 1.0 to 1.2 to account for the wings of the higher water vapor absorption lines.) We refer to these as the nominal values of the C parameters. We note here that the L93 model in the range 20-32 GHz can be reproduced from L87R93 by using the parameters CL =1.05, CW =1.0, CC =1.25 and CX =1.0. Liebes increase by 5% in the strength of the water vapor absorption parameter, CL, from L87 to L93 was in large part due to earlier ground based WVR inter-comparisons with radiosondes reported by Keihm [1991]. The line width parameter, CW , was not adjusted in L93 because the Keihm [1991] data set did not include sufficient spectral resolution of the line shape near 22 GHz. Our intent in this work is to reexamine the adjustment that was made to L87 using an improved intercomparison database. It is for this reason that we begin with L87R93 as our nominal model. Figure 2.3a shows the change in absorption spectrum when each of the parameters is raised by 10% above the nominal L87R93 model. The line strength parameter, CL, increases the absorption significantly, and this increase is accentuated for frequencies near resonance. The width parameter, CW, increases the width of the curve but at the same time decreases the absorption near the center
CL 10% S. Cruz-Pol
CL 10%
21
CC 10% CW 10%
L87R93
CX 10% CW 10% CC 10%
CX 10%
Figure 2.3 (a) Effect of line parameter variation by 10% on total atmospheric absorption. (b) Differential effect of line parameter variation with respect to nominal L87R93 model. The arrows at the bottom of the figure indicate the frequencies measured in this experiment. (Same atmospheric conditions as Fig. 2.2).
S. Cruz-Pol
22
(resonance) region. Both the continuum, CC, and the oxygen, CX, parameters, increase the absorption through the 20-32 GHz frequency span, because of their dependence on the square of frequency. These effects are also portrayed in Figure 2.3b, in which we have plotted the difference in absorption with respect to the nominal L87R93 model.
2-3 Experiment Description and Calibration

2-3.1 Radiometer Data
The experiment consisted of the collection of data at two National Weather Service radiosonde launch sites. These were chosen for their contrasting humidity conditions to provide constraints on both the 22.235 GHz vapor emission line and the level of oxygen emission in the 20-32 GHz interval. The sites were at San Diego, CA from 11 December 1991 through 3 February 1992 and West Palm Beach, FL from 8 through 21 March 1992. The overall range of humidity in terms of vapor burden varied from 0.6 - 2.9 g/cm2. Only data obtained under cloud free conditions were used (90% were from San Diego, CA and 10% from West Palm, FL). The instruments used included three independently calibrated WVRs built by JPL, designated J1, J2 and D2, which together provided measurements at 20.0, 20.3, 20.7, 21.5, 22.2, 22.8, 23.5, 24.0 and 31.4 GHz. The J1 unit operated in a continuous tip curve mode at preselected elevation angles from 10o to 165o (where 90o corresponds to the zenith direction) and at up to nine frequencies; 20.0, 20.3, 20.7, 21.5, 22.2, 22.8, 23.5, 24.0 and 31.4 GHz. The J2 WVR included 5 elevation angles and 3 frequencies at 20.7, 22.2, and 31.4 GHz. The D2 WVR operated at 20.7 and 31.4 GHz. The J2 and D2 units, which were located approximately 10 m from J1, were only used for the purpose of absolutecalibration verification. The overall range of measured TB varied for each frequency as follows; from 16 to 30K at 20.0 GHz, from 12 to 33K at 20.3 GHz, from 19K to 38K at 20.7GHz, from 29K to 48K at 21.5GHz, from 26 to 53K at 22.2GHz, from 20 to 35K at 22.8 GHz, from 18 to 50 K at 23.5 GHz, from 23 to 46 K at 24GHz and from 16 to 26 at 31.4 GHz.
S. Cruz-Pol
23
A tip gain calibration [Elgered, 1993] was performed on all three WVR units. The zenith brightness temperatures were obtained by combining the smoothed gains obtained from the tip curves with antenna and reference counts from longer zenith integrations obtained between tip curves. Only tip data for which the tip curve fit RMS residuals were less than 1.0 K were used for calibration. Poorer quality tip results were deemed unreliable, in terms of absolute calibration, for the purposes of this experiment and often indicated cloudy conditions. The processed high quality tip gains were corrected for beam smearing effects using an opacity- and channel-dependent term derived using the Jseries beam pattern. Inter-comparison of TB data at frequencies common to J1 and J2 instruments revealed absolute calibration accuracies of approximately 0.5 K or less as demonstrated in Figures 2.4(a)-(c). Only radiometer data from the J1 unit were actually used for comparison with raob-derived TB s in the absorption model analysis. The J1 radiometer TB data were averaged over one half hour for the times coinciding with the radiosonde balloon launch. This was used as the ground truth for comparison purposes with the radiosonde-derived TBs. The 22.8 GHz data was not used due to an unexplained bias in that channel. The eight well-calibrated channels across the full line width of the 22.235 GHz feature constitute an excellent radiometric data set for constraining the line shape model.
20.7 GHz channel S. Cruz-Pol 24
22.2 GHz channel
31.4 GHz channel
Figure 2.4 Zenith Brightness temperature intercomparison between radiometer units J1 and J2 for absolute calibration purposes. Measurements were conducted during December 1991 at San Diego, CA at (a) 20.7 GHz, (b) 22.2 GHz and (c) 31.4 GHz. The data have been smoothed by a 30 minutes running average.
S. Cruz-Pol
25
2-3.2 Radiosonde Data

Raob balloons at both sites were released approximately 15 m from the J1 WVR. Raob data were obtained from the National Climatic Data Center (NCDC). These provide height profiles of pressure, air temperature and dew point temperature. A reading was produced every few hundred meters at irregular intervals of height up to approximately 16 km. Raob balloons were launched at most twice a day. There were 130 profiles in our database. Of these, 23 corresponded to bad (invalid frequency tuning) wvr data, and 31 did not coincide with the times wvr units were in operation. Of the remaining 76 profiles, 36 had clouds present, leaving a total of 40 available raob profiles for this analysis.
2-3.2.1 Radiosonde Processing

The measured air pressure and dew-temperature-derived-vapor profiles are exponentially interpolated at regular intervals of height of 30 meters. The air temperature profile is linearly
interpolated for the same height intervals. The integration interval for the radiative transfer equation was chosen to be every 30 meters of altitude for adequate precision in the resulting TB. The relative humidity and water vapor content was derived from the temperature, dew point and air pressure information using the Goff-Gratch formulation for saturation water vapor density ws [Goff, 1949]. See Appendix B for a complete description of the formulation. Note that Goff-Gratch includes a pressure dependence on saturation vapor density which has been largely neglected in the past. The difference in TB when including the pressure dependence was found to be up to 1.4K for the raob profiles considered here. This effect was largest for profiles with dry climate conditions.
2-3.2.2 Radiosonde Selection and Screening

Between 1973 and October 1993, the National Weather Service routinely truncated their radiosonde humidity measurements at 20% RH, making the radiosonde hygristor appear to lose its sensitivity below 20% RH [Wade, 1994]. Any humidity below 20% was recorded as 19%. The relative humidities are still reported with a high bias. (See Wade [1994] for more details concerning
S. Cruz-Pol
26
NWS raob biases.). The data also suffered from the inability of the raob to properly report dew point temperature for levels of relative humidity higher than 95%. To reduce these effects, a correction factor was applied to relative humidity values outside the 20% - 95% range. The relative humidity was set to 11% whenever it was less than 22% (this usually occurred at low altitudes) and to 20% (typical high altitude value) whenever it was higher than 100%. The unphysical values of RH greater than 100% generally arose at altitudes greater than 8 km, where the effects of RH errors are negligible for the present analysis. The selection of the radiosonde profiles to be used was based on the degree to which they were affected by these RAOB humidity problems. Some profiles showed TB differences of up to 12.5K when the humidity correction factor was used. To study this effect, the four line parameters were estimated with and without the correction. First, only one profile was use for the estimation, then more profiles were added in order of less to greater change in TB due to the correction. In every case, the change in each of the 4 parameters resulting from the correction was computed. This change was then compared with the errors in the parameters due to 0.5K Gaussian noise on the WVR TBs. Both effects are plotted in Fig. 2.5 versus the number of profiles used in the estimation, for the case of the line width parameter, CW. The combined root sum square (RSS) error is also plotted. As seen in the figure, the error due to the RH correction is of comparable magnitude to the error introduced by the noise in the WVR TB when between 10 and 20 profiles are used. The highest number of profiles that can be used without significantly increasing the RSS error in the estimated parameters is found to be 21 profiles. (Each profile is compared in the estimation with WVR data, with up to 8 frequency channels, for a total of 108 data points). The error plots for the other three parameters look similar to Fig. 2.5 except for the line strength parameter, CL. In the CL plot, the error starts increasing at 16 profiles, but at 21 profiles the RSS error is still only 0.04, so this number of profiles was chosen as a compromise to accommodate the estimation of all 4 parameters simultaneously. This corresponded to profiles with differences less than 4.5K between the TB calculated with the corrected and uncorrected relative humidity values.
S. Cruz-Pol
27
0.05
0.04
Noise in CW Parameter
0.03
0.02
0.01
0 4 6 8 10 12 14 16 18 20 22 24 26 28 30 No. of profiles used RH correction 0.5K TB noise Total Noise
Figure 2.5 Total error in CW line parameter due to a 0.5 K uncertainty in measured TB and the correction of the raob relative humidity reading, versus number of raob profiles used in the estimation (see Section 3.2 for a complete discussion). A trade off between the amount of data used and the minimum error in parameter estimation yields an optimum value of 21 raob profiles, providing a total of 108 data points. Hoehne [1980] provides estimates of the functional precision for VIZ radiosonde packages as 0.7 hPa for barometric pressure and 0.84K for air temperature. England et al., [1993] suggest a value of 5% be used for relative humidity. Manufacturers specifications list 4% for the carbon hygristor humidity sensors used by the VIZ radiosonde, but independent investigations have shown the errors to be dependent on the particular manufactured lot. Therefore, as England et al. [1993] note, until more complete investigation of the uncertainties is undertaken, 5% is a reasonable estimate for the humidity measurements. In any case, the dominant contribution to the error in brightness
temperatures inferred from the raobs comes from the 0.84K temperature uncertainty and not from the 5% humidity [England et al., 1993].
S. Cruz-Pol
28
2-4 Analysis and Results
2-4.1 Hills Ratio Test
Any of the absorption models described above could, in principle, be adjusted to fit our data set. In order to select the appropriate absorption models and line shapes, we adopt an asymmetry ratio test formulated by Hill [1986]. This test provides a clear-cut means for assessing the validity of the VVW and Gross line shape models and is independent of line strength and insensitive to errors in line width, continuum water vapor absorption and oxygen absorption. Two brightness temperatures
corresponding to frequencies approximately symmetric about the line center are used to compute the ratio. In this work, 20.7 and 24.0 GHz are used. The ratio is defined as the difference of the two TBs divided by their averages,
(2.15)
The Hill ratio was computed from the raob-based TB data for all models mentioned above including the new model presented in this work, and for the water vapor radiometer data itself. Data obtained at both sites are plotted as a time series in Figure 2.6. This test shows the superiority of VVW versus Gross. The purpose of including this test here is to demonstrate that we are starting off with the more reliable line shape model currently available which, as shown here, is the VVW line shape.
S. Cruz-Pol
29
Figure 2.6 Hill ratio comparison between various atmospheric models showing agreement of the chosen water vapor absorption line shape with the radiometer data. (See text for explanation of models' acronyms).
Note that the Hill ratio for the W76 and UMF81 models, both of which use the Gross line shape for the water vapor absorption, yields a much lower value than the corresponding ratio from the WVR data. The W76 and UMF81 models have average ratios of 0.005 and 0.007, respectively, whereas the WVR data has an average ratio of 0.045. In contrast, the new and L87R93 models, both of which use the VVW line shape, yield average ratios of 0.044 and 0.043, respectively. These results strongly suggest that VVW is the preferred choice for vapor absorption line shape at 22 GHz. Note that the same finding was obtained by Hill [1986] when the ratio test was applied to the original Becker and Autler [1946] laboratory data.
S. Cruz-Pol
30
2-4.2 Parameter Estimation: Newton-Raphson Iterative Method

The modeled TBs were calculated using the radiative transfer integral and the parameterized absorption model described in appendix A applied to the balloon profiles. The parameters CL, CW, CC and CX were estimated using a Newton-Raphson iterative method [Kagiwada and Kalaba, 1969]. The Newton-Raphson method is a fast-converging iterative procedure for nonlinear models. The first derivative of the nonlinear equation is taken with respect to each of the variables that will be estimated, in this case CL, CW, CC and CX, which form a vector. The derivative is then evaluated using the initial values for the four parameters to form the Jacobian matrix given by
(2.16)
The number of rows in the Jacobian is equal to the number of data points (i.e. the number of raob profiles times the number of frequencies at each). The number of columns is equal to the number of parameters being estimated. In our case, derivatives were calculated numerically due to the
complexity of the radiative transfer integral equation used to compute the TB. The initial values for the four parameters were chosen to be the nominal values, i.e. CL =1.0, CW =1.0, CC =1.2 and CX =1.0. Then the new C parameters are found as
cnew = cinitial + c
(2.17)
where c is the correction for the parameters and is computed from the minimum square error inversion by
c = JtJ
( )
J t TB
(2.18)
S. Cruz-Pol
31
and TB is the difference between the TB modeled with the initial value parameters and the true (observed) TB. These new values for the Cs are used as initial values for the next iteration. The process is repeated until changes in each of the parameters are less than 0.001. All four parameters were estimated simultaneously. This is possible because of the number of frequency channels
employed and the range of humidity conditions (0.6 - 2.9 g/cm2 vapor burden) covered during the span of our experiments.
2-4.3 New Model Retrieved Parameters

The final retrieved parameters, CL, CW, CC and CX, are shown in Table 2.1. As the table indicates, the nominal parameters used in the L87R93 model are 3 to 7 percent lower. Figures 2.7a-c depict plots of the brightness temperature for three climatological conditions. Each graph has a plot corresponding to the L87R93 and new models. Also shown are the radiometer measured brightnesses. The new estimated parameters show better agreement with the WVR data. A better indication of this agreement can be seen in Figures 2.8a-c, where we have plotted the difference in brightness temperature, taking the L87R93 model as the reference (therefore, by definition the L87R93 model is a line at zero in Figures 2.8a-c). In these figures we have included the L93 model which, as explained above, is similar to L87R93 except that it has a higher water vapor line
TABLE 2.1 New retrieved atmospheric absorption parameters Parameter CL CW CC CX RMS difference Liebe87-Rosenkranz93 1.0 1.0 1.2 1.0 1.36 K New Model 1.064 1.066 1.234 1.074 1.05 K
Humid S. Cruz-Pol 32
Brightness Temperature [K]
55 50 45 40 35 30 25 20 15 18
Moderate
20 22 24 26 frequency [GHz] (a) 28 30 32
50
45 40 35 30 25 20 15
dry 10 18.0 20.0 22.0 24.0 26.0 28.0 30.0 32.0 frequency [GHz] (b)
30 28 26 24 22 20 18 16 14 18 20 22 24 26 28 30 32 L87R93 New WVR
frequency [GHz] (c)
Figure 2.7 Brightness temperature spectra comparison between radiometer data (WVR) and radiosonde-derived data with new and nominal parameters for a vapor burden of (a) 2.9 g/cm2 , (b) 2.3 g/cm2, and (c) 1.3 g/cm2. (Note that only five channels were in operation during the raob launches for conditions (b) and (c) ).
S. Cruz-Pol
33
2.5
Difference in T B [K]*
2 1.5 1 0.5 0 -0.5 -1 -1.5 -2 18.0 23.0 28.0 33.0
frequency [GHz] (a)
Difference in T B [K]*
2 1.5 1 0.5 0 -0.5 -1 -1.5 18 23 28 33
frequency [GHz] (b)

2 1.5
Difference in TB [K]*
L87R93
1 0.5 0 -0.5 18
L92 New WVR
frequency [GHz] (c)
23
28
33
Figure 2.8 Plot of the difference TB -TB L87R93 for (a) humid (West Palm Beach), (b) moderate (San Diego) and (c) dry (San Diego) conditions. Note that TBLiebe87 is equivalent to our nominal model (CL = CW = CX =1.0 and CC = 1.2). (Only five channels were in operation during the raob launches for conditions (b) and (c) ).
S. Cruz-Pol
34
strength parameter, (CL =1.05, CW =1.0, CC =1.2 and CX =1.0 ). The increase in line strength of the L93 model resulted from measurements at only 21 and 31 GHz [Keihm, 1991]. The extra frequencies over 20-32 GHz used to derive our new model can constrain both the shape and strength of the absorption model simultaneously. The result is an increase in both the line strength and width parameters. The new model shows the best agreement with the radiometric temperatures.
2-4.4 Error Analysis

The RMS difference between modeled and measured TB is reduced by 23%, from 1.36 K to 1.05 K, with the new parameters. A numerical sensitivity analysis was conducted to determine the level of uncertainty in the estimated parameters due to measurement noise by the radiometer and radiosondes. Independent realizations of the entire estimation process were simulated, in which Possible biases in the absolute
random perturbations were made to the actual measurements.
calibration of the radiometer were modeled as an additive constant brightness temperature. Independent biases are determined for each frequency channel, but the bias at a particular channel is assumed constant for all radiosonde launches. Realizations of the biases are selected from a zero mean, normally distributed random process with standard deviation of 0.5 K. Additive random noise in the radiometer data was also modeled. This noise is independent for every channel and radiosonde profile, and is normally distributed with zero mean and 0.1 K standard deviation. To simulate the uncertainties in radiosonde measurements, the estimates of precision described in Section 3.2 were used for air temperature, pressure and relative humidity. These raob errors are incorporated into our sensitivity analysis by assuming that both constant biases and random noise exist in the measurements. For each realization, bias values are selected for air temperature, pressure and relative humidity which are assumed constant over the entire experiment. The values are selected from zero mean, normal distributions with standard deviations of 0.707 times the errors suggested by Hoehne [1980] and England et al. [1993]. Random noise in each individual measurement is also modeled by zero mean, normal distributions with standard deviations of 0.707 times the published errors. (The effects of the random noise were found to be negligible relative to the bias errors.) Errors due to the raob relative
S. Cruz-Pol
35
humidity pinning problems (also discussed in Section 3.2) were modeled in the following manner: For intervals of the profile over which the RH was below 20%, a random humidity level was selected from a uniform distribution between 0 and 20%. For intervals over which the RH was above 100%, a random level was selected from a uniform distribution between 0 and 100%. These ranges bracket the possible uncertainty in our humidity corrections.
TABLE 2.2 Standard deviations and correlation matrix for the four estimated parameters taking into account errors in the raob profiles and WVR brightness temperature measurements.
Parameter
CL
CW
CC
CX
Std. Deviation 0.016 .0096 0.155 .252 Parameter CL CW CC CX CL 1 -0.085 0.045 -0.048 CW -0.085 1 -0.513 0.485 CC 0.045 -0.513 1 -0.989 CX -0.048 0.485 -0.989 1
The four C parameters were repeatedly estimated with independent errors added to the data, to obtain 2600 simulated noise realizations. A covariance matrix for the four C parameters was then computed as well as the variance of each of the parameters. The results show that the standard deviations in the CL and CW parameters are 1.6% and 0.9%, respectively, and are 16% and 25% for CX and CC (see Table 2.2). Since the standard deviations of the oxygen and continuum parameters, CX and CC, are larger than the change from their nominal values, we cannot statistically justify the correction for these two. However, our corrections to the line width and line strength parameters, CL and CW, can be considered statistically significant and render a better atmospheric model than the nominal values. Correlation analysis between coefficients shows a high negative correlation of -99% between the errors in the oxygen and continuum terms (CX and CC). This is to be expected since both parameters vary essentially as frequency-squared. The correlation among errors in the other parameters vary between 3% and 52%, as seen in Table 2.2.
S. Cruz-Pol
36
The effect that the errors in the parameters have on the atmospheric model was addressed by a second noise simulation. In this case, 2000 independent realizations were simulated in which the total atmospheric absorption spectrum between 20 and 32 GHz was computed using the surface values of the 1962 U.S. Standard Atmosphere (T=288.15K, P=1013.25 hPa) at a relative humidity of 50%. For each realization, the four C parameters were randomly perturbed according to the standard deviations given in Table 2.2. From these realizations, mean and standard deviation absorption spectra were computed. The ratio of standard deviation to mean gives the percentage error in the absorption model. The mean and percentage error spectra are shown in Figure 2.9. The error in the new atmospheric absorption model is approximately 3% in the near vicinity of the 22 GHz water vapor line, and rises to 8% near 32 GHz. The error in the modeled brightness temperature varies from 1.5% to 2% as shown in Figure 2.10. Table 2.3 shows a comparison of the uncertainty introduced by the atmospheric attenuation in the calculated brightness temperature by the nominal and new model [Johansson et al.,1987].
S. Cruz-Pol
37
%ERROR (std/ave) average Absorption
10 9
0.05 0.05
% Error in Atmospheric Absorption
7 6 5 4 3 2 1 0
P= 1013.25 HPa, T= 288.2K = 7.718 g/m3
0.04 0.03 0.03 0.02 0.02 0.01 0.01 0.00
Frequency [GHz]
Figure 2.9 Percentage error in the improved model for atmospheric absorption using the 1962 U.S. Standard Atmosphere at sea level with RH = 50%. The error in the improved model Model errors are due to bias and random measurement uncertainties in radiometer TB, the correction for raob relative humidity values less than 20% or greater than 100%, and the uncertainty in the radiosonde readings for pressure, temperature and relative humidity.
Ave Atmospheric Absorpiton
0.04
20
22
24
26
28
30
32
S. Cruz-Pol
38
Figure 2.10 Percentage error in the improved downwelling TB for the radiosonde profiles used by the estimation algorithm. Model errors are due to bias and random measurement uncertainties in radiometer TB, the correction for raob relative humidity values less than 20% or greater than 100%, and the uncertainty in the radiosonde readings for pressure, temperature and relative humidity. The error in the predicted brightness lies between 1.5 % and 2.0 %. Error bars in the graph represent the standard deviation of the percentage error for all the profiles used in the analysis.
TABLE 2.3 Uncertainties introduced to the calculated brightness temperatures by the L87R93 (nominal) [Johansson et al., 1987; England et al., 1993] and by the new atmospheric attenuation model.
f [GHz] 20 25 30 35
L87R93 8.96% 8.70% 8.97% 8.70%
New 1.5% 1.3% 1.8% 2.2%
S. Cruz-Pol
39
2-5 Conclusions
An improved model for the absorption of the clear atmosphere near the 22 GHz water vapor line is presented2. The Van-Vleck-Weisskopf line shape is used with a simplified version of the model by Liebe [1987] for the water vapor absorption spectra and the model by Rosenkranz [1993] for the oxygen absorption. Radiometric brightness temperature measurements from two sites of contrasting climatological properties, San Diego, CA and West Palm Beach, FL, were used as ground truth for comparison with in situ radiosonde derived brightness temperatures. Estimation of the new models four parameters, water vapor line strength, line width, and continuum absorption, and far-wing oxygen, was performed using the Newton-Raphson inversion method. In addition, the Hill line asymmetry ratio was evaluated for several currently used models, showing agreement of the radiometric data with the VVW line shape, and ruling out atmospheric absorption models using the Gross line shape near 22 GHz given by Waters [1976] and Ulaby et al. [1981]. The RMS difference between modeled and measured TB was reduced from 1.36 K to 1.05 K, with the new parameters. An error analysis shows that the standard deviations in CL and CW are 1.6% or less and, 16% and 25% for CX and CC, respectively. These errors assume 0.5K bias and 0.1K random errors in the radiometer TB data, 0.7 hPa bias and random error for pressure, 0.84K for air temperature, and 5% for humidity, and uniformly distributed noise for RH <20% or > 100%. This indicates that our new values for CL and CW represent a statistically significant improvement on previous atmospheric absorption models. Our corrections to CX and CC are not statistically significant, given the errors associated with the experimental data. The percentage error in the new absorption model is approximately 3% near the 22 GHz line, and rises to 8% near 32 GHz. The L93 absorption model [Liebe et al., 1993] included a 5% increase in the water vapor line strength above the L87 model [Liebe and Layton, 1987]. This increase was largely a result of earlier WVR inter-comparisons with raobs reported in Keihm [1991]. Our results here confirm this increase,
See Appendix D for a FORTRAN program listing of the improved atmospheric absorption model.
S. Cruz-Pol
40
by proposing a 6% increase (CL = 1.06) with a 1.6% margin of error. While our results validate the line strength correction in Liebe et al. [1993], they also improve on the line width parameter (CW = 1.07 0.01). Improvement in the model for line width is possible because our new WVR data set has a greater number of frequency channels across the 22.2 GHz water vapor line. The atmospheric attenuation is only one of the two major contributors to the total brightness temperature seen from a satellite above the ocean. The second major contributor is the ocean itself. A model for the ocean surface emissivity is presented in the next chapter.
S. Cruz-Pol
41
3. SEA SURFACE EMISSIVITY

The brightness temperature measured from the sea surface depends on the specular ocean emission and the excess emissivity induced by the wind. In this part of the work, we adjust a model for observed TB from a satellite-based radiometer over the ocean, by comparing it to the TOPEX/Poseidon Microwave Radiometer (TMR) data over a four- year period (1992-1997). In order to fully model the TB, we need to know the sea surface temperature and salinity, the upwelling and downwelling brightness temperatures, the atmosphere transmisivity and the wind speed. For this purpose, nearcoincident radiosonde profiles from fifteen (15) stations around the worlds oceans are used to find the upwelling, downwelling and transmissivity of the atmosphere. The dielectric properties of sea water are found from the modified Debye equation using salinity and sea surface temperature data from NODC ocean depth-profiles. The wind speed is estimated from the TOPEX/Poseidon dual-frequency altimeter. Adjustment to the model is accomplished by means of the Newton-Raphson method.
3-1 Current models and their limitations

A satellite-based radiometer looks down at the ocean surface and hence its brightness temperature depends upon the ocean emissivity. The ocean emissivity can be decomposed into a contribution from the specular emission of the sea surface and emissivity induced by the wind.
3-1.1 Specular sea surface emissivity model

The specular emissivity of the ocean is a function of the frequency of operation and the dielectric properties of the sea water. If the ocean surface fills a flat half-space, the emissivity at normal incidence, is given by [Born and Wolf, 1965]
(3.1)
Ch. 3 Sea Surface Emissivity Model
S. Cruz-Pol
42
where the second term on the right is the Fresnel reflection coefficient at nadir and is the dielectric coefficient of the sea water. The dielectric coefficient of sea water at microwave frequencies below 40 GHz can be represented by a simple Debye relaxation expression, given by [Debye, 1941] = (3.2)
where s and
are, respectively, the static and high frequency dielectric coefficients of the sea water,
o is the permittivity of free space ( = 8.85 x 10-12 F/m), is relaxation time in seconds, is the ionic
conductivity of the dissolved salts in mho/m, and f is frequency in Hertz. The real and imaginary parts of the permittivity are and , respectively. The parameters s, , , and are all functions of
the temperature, T, and salinity, S, of the sea water and are given by Klein and Swift [1977] and, more recently, by Ellison et al. [1996]. These two models are, respectively, referred to as KS77 and E96 in the remainder of this document.
3-1.1.1 Klein-Swift Ocean-Water Dielectric Model
The Klein and Swift [1977] model consists of a simple Debye expression for the sea water dielectric over a limited frequency range (f < 10 GHz). The model includes polynomial fits for the static dielectric coefficient, the ionic conductivity and the relaxation time as a function of temperature and salinity. The sea water dielectric coefficient model in KS77 was derived from dielectric
measurements of sea water and aqueous NaCl solutions conducted at 1.43 and 2.65 GHz for salinities3 in the range 4/oo < S < 35/oo. Their derivation is based on the assumption that has a constant
value of 4.9 with uncertainty of 20%. Typical values of this parameter vary from 4.6 to 8.5 for salinity values between 23 and 39 /oo and temperatures typically found at the oceans (0 - 30 C). This model is still widely used to calculate the sea water dielectric coefficient although the authors recommended care when using their model at frequencies above 10 GHz. They stated that as
Salinity is expressed in parts per thousand (/oo) on a weight basis, i. e., total mass of solid salts in grams dissolved in one kilogram of solution.
S. Cruz-Pol
43
the frequency increases to X-band [8-12 GHz], the error in maximized [Klein and Swift, 1977].
[the real part of
] is
3-1.1.2 Ellison Ocean-Water Dielectric Model
The E96 model was developed using water samples from the Mediterranean, Polar, Atlantic and Mid-Atlantic Oceans. Ellison et al.[1996] improved the frequency range and added a polynomial fit for the high frequency dielectric coefficient. Yet, it was developed using limited ground truth (only six sea water samples). They performed laboratory measurements of the dielectric coefficient for a wider range of frequencies (2 - 40 GHz), and at salinities (20 - 40 /oo) and temperatures (-2 - 30 C) found in the worlds oceans. Their claimed accuracy is 3% or better for frequencies of up to 40 GHz. They applied the radiative transfer theory to atmospheric conditions for comparison with Topex Microwave Radiometer (TMR) data from the North Atlantic during 5 weeks in the Fall of 1993. They did not filter data with respect to specific wind conditions, or use actual radiosonde profile data to model the atmosphere, as we do in our analysis. Instead, they modeled the wind-roughen sea, as well as the atmosphere, using the European Center for Medium range Weather Forecast (ECMWF) model predictions of 10m winds and atmospheric profiles. The ECMWF uses ship and buoy
measurements to generate a meteorological prediction every 6 hours. Its accuracy for monitoring water vapor variations in the atmosphere is approximately 9% for humid atmospheric conditions and lower for dry conditions [Stum, 1994]. In our investigation, we employ the co-located TOPEX
altimeter data to select only low wind conditions, in order to reduce the dependence of our analysis on the wind model. For their comparison, they used the oxygen and water-vapor atmospheric attenuation model by Liebe et al. [1993]. Instead, we use the improved atmospheric model presented in Chapter 2. Furthermore, our comparison is limited to low humidity conditions only (path delay < 15 cm), in order to reduce the dependence of our analysis on the atmospheric model. Our present investigation of the specular sea emission seen by the Topex/Poseidon satellite radiometer provides field verification of the sea water parameters over a broader range of regions in the oceans.
S. Cruz-Pol
44
3-1.2 Wind-roughened emissivity Model
When the wind blows across the surface of the oceans, it generates roughness.
This
roughness increases the emissivity of the ocean. There are three mechanisms by which the windinduced roughness increases the emission from the sea. The first one is the gravity waves. These are ocean waves with wavelength long compared to the radiation wavelength, and are modeled with the theory of geometric optics [Stogryn, 1967; Hollinger, 1971; Wilheit, 1979b]. The second mechanism is the capillary waves. These have wavelengths that are small compared to the radiation wavelength, and are modeled by small perturbation theory [Wentz, 1975]. The third is the sea foam coverage over the ocean waves.
3-1.2.1 Geometric Optics approach
The ocean surface can be described by a series of reflecting flat facets with various inclinations characterized by a slope distribution. The individual contribution of each facet to the upwelling brightness temperature is calculated from the Fresnel reflection relations [Cox and Munk, 1955]. This approach was employed by Stogryn [1971] for 20 and 35 GHz frequencies, and by Hollinger [1971] for frequencies between 1 and 20 GHz. The latter study reveals that brightness temperatures are underestimated close to nadir. Furthermore, only fair agreement was obtained between model and measurements at 20 GHz, with increased degradation at lower frequencies.
3-1.2.2 Two scale approach
To improve the agreement between theoretical predictions and low-frequency observations, a composite-surface model was developed [Semyonov, 1966; Wu and Fung, 1972; Wentz, 1975]. This two-scale model combines geometric-optics and small-scale perturbation theory by superimposing
S. Cruz-Pol
45
small capillary waves on the larger gravity waves (see Fig. 3.1). The scattering coefficients are expressed as the sum of two contributions. The first term accounts for the gravity waves and is given by the geometric optics solution, slightly modified by the presence of the ripples, which impose a modification to the Fresnel reflection coefficients. The second term results from the average of the scattering coefficients due to the small irregularities over the large-scale slope distributions. The twoscale scattering model includes multiple reflections and shadowing effects. The model shows greater wind dependence at incidence angles away from nadir. Since this work concentrates on nadir looking radiometry conditions, the two-scale model will not be considered. The total nadir emissivity of the ocean can be expressed as [Wilheit, 1979b],
where W is the wind speed at 20m above the sea surface. The first term,
, in the above
expression refers to the specular emission of the sea surface and the second term refers to the effect of the wind-induced roughness on the ocean emissivity. For winds higher than 7m/s an additional term is added, fS, to account for the effective fractional coverage of black body foam. This will be further explained in section 3.1.2.3.
S. Cruz-Pol
46
Figure 3.1 Mechanisms responsible for the microwave emission of a wind-roughened sea surface include large gravity waves, small capillary waves and sea foam.
3-1.2.3 Foam coverage
Foam cover increases the emissivity of the surface at a rate of about 1K/ m/s, for wind speeds above 7 m/s at 19.35 GHz in the nadir direction [Norberg et al., 1971; Stogryn, 1972]. A layer of air bubbles and water on the surface has an effective dielectric coefficient between those of air and sea water (see Fig. 3.1). This results in a lower surface reflectivity or higher transmittance for the radiation by the sea. The foam effect is modeled with a frequency dependent function for winds greater than 7m/s.
S. Cruz-Pol
47
(3.4)
As seen in equation (3.4), no foam effect is found for lower winds since foam starts to form at wind speeds greater than 7m/s. Since this work concentrates on calm to low wind speed conditions, the foam effect will not be considered.
3-1.3 Air-Sea Stability

The wind speed used for this analysis is referred to a 19.5m height. The wind varies with height near the surface of the ocean. This variation is affected by the temperature difference between the sea and the air on top. When the sea is warmer than the air, unstable conditions prevail and higher waves are generated for a given wind speed. When the air temperature is greater than the sea temperature, the condition is referred to as stable and the significant wave height is lower than that for neutral or unstable conditions. The wind speed profile over the sea surface for neutral conditions has a logarithmic height dependence [Paulson, 1967] (3.5) where U* is the friction velocity, zo is the roughness parameter of the ocean, K0.4 is the Krmn constant and z is the height at which the neutral wind W is referred. The friction velocity is a measure of the turbulent momentum flux and the roughness length reflects the loss of momentum to the sea surface [Charnock, 1955]. For non-neutral conditions, the wind speed is defined by the expression [Paulson, 1972] (3.6) where L is the modified stability length, and given by [Panofsky, 1964] is the integrated profile stability function
(3.7)
S. Cruz-Pol
48
and
is the non-dimensionalized wind shear gradient function given by [Monin, 1977]
(3.8)
The wind speed measured by the altimeter has to be corrected to take into account the atmospheric stability conditions of the ocean. For this purpose, the sea surface temperature, s, (see the next section on NODC CD ROMs), the air temperature, a, and thermometer height, za, (see the next section about raob balloons) are used to compute the friction velocity, U*, and roughness parameter, zo, of the ocean [Cardone, 1969]. The solution requires the use of an iterative procedure with and
as the initial guesses. The roughness parameter and the modified stability length, L, are found from (3.9) and
(3.10)
where, g is gravitational acceleration, given by
is the mean temperature, and the friction velocity, U*, is
(3.11)
where a and s are the air and sea surface temperatures, za is the height at which a is measured, zm is the height at which wind speed is measured, K is the Krmns constant, g is the gravitational acceleration. The iteration process is repeated until Ln+1 - Ln is less than 0.3 meters [Cardone, 1969]. The values found for U* and zo after the iterative procedure converges, define the surface boundary layer wind distribution, and are used to calculate the wind speed under neutral conditions at 19.5 m above the surface. This is the wind speed used in this analysis. It is referred to as 19.5m neutral stability wind.
S. Cruz-Pol
49
3-2 Data sets

The data used for this part of the work included measurements from a period comprising four and a half years, from December 1992 to May 1997, from three different sources. These sources are 1) the TOPEX/Poseidon satellite mission (altimeter and radiometer), 2) fifteen raob stations around the globe and 3) the National Oceanographic Data Center (NODC). Each data set is described below.
3-2.1 TOPEX/Poseidon :Altimeter and Radiometer data
The goal of the TOPEX/Poseidon (T/P) mission is to increase the understanding of ocean dynamics by precise measurement of the sea level over several years [Stewart et al., 1986] . One full cycle of the satellite covers 90% of the Earths ice-free oceans. Every cycle consists of 127 orbits around the globe and is completed approximately every 10 days (9.9155 days). Each half-orbit is called a pass, therefore every cycle has a total of 254 passes. From its orbit 1,336 km above the Earth's surface, T/P measures the sea level using a radar altimeter. Measurements from the TOPEX
Microwave Radiometer (TMR) provide estimates of the total water-vapor content in the atmosphere [Ruf et al., 1995], which are used to correct errors in the altimeter measurements [Stewart et al., 1986]. Data from both of these instruments were used in this work.
3-.2.1.1 Altimeter
The dual-frequency altimeter is used primarily for a path length measurement [Stewart et al., 1986]. The altimeter is a nadir looking instrument. It sends a pulse to the sea surface and measures the time required for the pulse to reflect back to the satellite. The measurement of the travel time of the reflected microwave pulse yields the position of the sea surface relative to the orbit of the satellite.
S. Cruz-Pol
50
This can be determined to within approximately 4 centimeters [Quinn and Wolff, 1993] after the correction for atmospheric disturbances and calibration. In addition, the altimeter measures surface wind speed from the shape and intensity of the backscattered pulse. A return pulse that has spread out in time is an indication of a rough ocean due to high winds. If the pulse comes back with high amplitude, it means there is calm sea. The relation between the wind speed and the backscattered cross section, , has been determined empirically and tested in numerous investigations for SEASAT,
GEOSAT, ERS-1 and TOPEX [Gnther et al., 1993; Fedor and Brown; 1982, Witter et al.,1991; and Brown et al., 1981]. The altimeter is very sensitive to low wind conditions. Therefore, it can be used as a very effective filter to isolate the calm sea emission data from the TMR. The T/P altimeter pinpoints low wind conditions (whenever the received pulse saturates the receiver) at which time the TMR measured brightness is due mainly to the specular sea emissions. The altimeter has two internal calibration modes to determine corrections for range, gain control and wave height. It uses C and Ku4 band pulses to measure the sea level approximately every half second. In addition, it provides the ocean surface radar backscatter coefficient per unit area, o, from which sea surface wind speed can be estimated. The altimeter return o at Ku band in the range of 10 to 20 dB were selected for the present analysis since these values correspond to low wind conditions. The modified Chelton-Wentz (MCW) [Witter and Chelton, 1991] table as calibrated for TOPEX [Callahan et al., 1994] was used to convert the o values to wind speed at a height of 19.5 m above the ocean surface. Figure 3.2 shows a plot of o versus wind speed for the algorithm used in this work.
Specifically, 13.6 GHz (=2.21 cm) in the Ku-band and 5.3 GHz (=5.66 cm) in the C-band.
S. Cruz-Pol
51
Figure 3.2 Wind speed model relating 0 to wind speed for the MCW algorithm as calibrated for Topex altimeter.
The use of the MCW algorithm results in an RMS error of
1.4m/s and a bias of -0.4m/s for
winds less than 23 m/s [Callahan et al., 1994]. Only data with winds below 7 m/s, at which speeds surface foaming is negligible, are utilized in order to isolate low wind conditions and relax the dependence of the correction to the specular model on the accuracy of the wind model.
3-2.1.2 TOPEX Microwave Radiometer (TMR)
Water vapor in the atmosphere can delay the return of the radar pulses to the satellite. This interferes with the accuracy of the sea level measurement. To correct this delay, the TMR is used. The TMR is a nadir-viewing radiometer that measures the water vapor content in the atmosphere, by measuring the brightness temperature from the ocean surface at 18, 21 and 37 GHz [Ruf et al., 1994]. Temperatures are measured once per second. Internal hot and cold calibrations are performed
alternately every 14 seconds. A correction of -0.28, -0.07 and -0.04 K/year was added to each of the three frequency channels, respectively, to correct for drifts in the receiver calibration [Keihm et al.,
S. Cruz-Pol
52
1997].
TMR global measurement accuracy for the ocean surface brightness temperature has an
instrument RMS precision of 0.3K and a bias error of 0.8K.
3-2.1.2.1 TMR Data selection and Screening All three TMR brightnesses are used to filter out data points that have liquid water content greater than 100 microns to ensure clear sky conditions. The liquid cloud content is computed from the following algorithm [Keihm, personal communication. 1998].
(3.12)
As mention above, only the TMR brightness temperatures at 18 and 37 GHz are used in this work since 21GHz is much more sensitive to humidity and introduces significantly larger errors in the estimation of ocean emissivity.
3-2.2 Radiosonde Data

Data from thirty (30) raob launch stations around the globe were compiled (see the location of each station in the map in Figure 3.3). At each station, a raob balloon was launched at most four times a day. The atmospheric profiles include air temperature, pressure and dew point temperature from which the relative humidity is computed as explained in the preceding chapter. In this data, no 22% pinning of the RH data was noted, so no profiles were discarded for this reason.
S. Cruz-Pol
1 5 9 10 13 11 12 6 8 17 16 15 18 19 20 30 28 29 7 26 14 23 25 27
53
21 22 3
24
Figure 3.3 Location of the radiosonde launch sites. (See Table for coordinates).
TABLE 3.1 Coordinates of the raob Stations depicted in the map on Figure 3.3. Legend Station No. Code Name 1 6011 2 8301 3 8522 4 16429 5 32618 6 43311 7 43333 8 43369 9 47678 10 47909 11 47936 12 47945 13 47971 14 47991 Latitude (N is +) 62.01 39.33 32.38 37.55 55.12 11.07 11.4 8.18 33.06 28.23 26.12 25.5 27.05 24.18 Longitude (E is +) -6.46 2.37 -16.54 12.3 165.59 72.44 92.43 73.09 139.47 129.3 127.41 131.14 142.11 153.58 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 61901 61902 61967 61996 68906 68994 71600 78016 91217 91245 91348 91413 91643 94299 94996 96996 -15.56 -7.58 -7.18 -37.48 -40.21 -46.53 43.56 32.22 13.33 19.17 6.58 9.29 -8.31 -16.18 -29.02 -12.11 -5.4 -14.24 72.24 77.32 -9.53 37.52 -60.01 -64.41 144.5 166.39 158.13 138.05 179.13 149.59 167.56 96.49
The uncertainties in the raob reading are the same as in the raob data set presented in Chapter 2, i.e., 0.7 mbar for barometric pressure, 0.84K for air temperature and 5% for relative humidity. To
S. Cruz-Pol
54
ensure that only clear (no clouds) atmosphere data was employed in the analysis, profiles with relative humidity values greater that 94% were filtered out, since this indicates the presence of clouds. The profiles were used to compute the upwelling and downwelling temperatures, the transmissivity and path delay of the atmosphere. Profiles with path delay greater than 15 cm were eliminated to reduce the sensitivity of the new ocean model to the accuracy of the atmosphere model. The range of path delay values in the final data set is portrayed in the histogram of Figure 3.4 below.
Figure 3.4 Histogram of the range of path delay values for the data used in this work.
The values for the remaining raob-derived variables range from 5.1K to 13.4K for Tup, from 7.7K to 16.0K for Tdn, and from 0.95 to 0.98 for e at 18 GHz, and range from 13.9K to 26.2K for Tup, from
16.6K to 28.8K for Tdn, and from 0.91 to 0.95 for e at 37 GHz.
The time and space separation between raob stations and TMR measurements was limited to 6 hours and 300 km, respectively, as further explained in section 3-3. After the data were filtered for clouds, low winds, path delay and time and space separation, only data from fifteen (15) raob stations
Ch. 4 Conclusions
S. Cruz-Pol
55
were actually utilized in the analysis, more specifically from stations with legend numbers 1, 6 to 14, 20, 24 and 28 to 30 (see Table 3.1), which yields a total of 263 raob profiles.
3-2.3 NODC Ocean Temperature and Salinity Profiles
Sea water dielectric properties are dependent, among other things, on the sea surface temperature and salinity. The National Oceanographic Data Center provides depth-profiles of ocean temperature and salinity measurements taken between 1900 and 1990. The data undergoes some degree of NODC quality check. This includes testing for valid, in-water, positions; observed depth not exceeding bathymetric depth; and reasonable vessel speed of advance. In addition, all the data were passed through a range verification for temperature (-3.00 o to 46.00 o Celsius) and for salinity (00.00 to 46.00 ). Values that were outside these ranges were eliminated [NODC, 1991]. Only the values of temperature and salinity at the surface (zero-depth) were employed since the surface emissivity and reflectivity depend on them. The data was averaged over the whole 90 year period for each month at every 10 degree latitude/longitude square, as identified by the World Meteorological Organization (WMO). Only the 10 degree squares corresponding to each of the 15 raob stations were used. In case of missing data, an average over the surrounding 10 degree squares was employed. NODC provides the precision for the temperature (hundredths of a degree) and salinity (thousandths of ) readings. For each station, there are 12 temperature and 12 salinity averages corresponding to each month of the year. The monthly standard deviations over the 90-year period were calculated for every averaged value. All the averaged values of sea surface temperature and salinity and their corresponding standard deviations for each month and raob-station are tabulated in Appendix C. These values are also plotted in Figure 3.5 and 3.6 for two of the stations. Station 9 is located in the Northern Hemisphere at 33o latitude, northeast of Japan. Station 29 is located in the Southern Hemisphere at -20o Latitude, southeast of Australia. For station 9 there was a total of 44,189 depth profiles averaged over the 90-year period, while for station 29 there were only 1,490 data points available over the same period.
Ch. 4 Conclusions
S. Cruz-Pol
56
303 301 299 297 295 293 291 289 287 285 283 281 Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec
Sea Surface Temperature

Figure 3.5 Average sea surface temperatures variation per month for station 9, located in the North Hemisphere (blue), and for station 29, located in the South hemisphere (pink). The error bars represent the standard deviations for each month.
Ch. 4 Conclusions
S. Cruz-Pol
57
37
36
35
34
33
32
31
30
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec
Salinity
Figure 3.6 Average salinity variation per month for station 9, located in the North Hemisphere (blue), and for station 29, located in the South hemisphere (pink). The error bars represent the standard deviations for each month.
As seen in Figure 3.5, sea temperatures are higher during the Summer months, July through September, in the Northern Hemisphere, while these months correspond to the winter season in the Southern Hemisphere, therefore exhibiting the lowest sea surface temperatures for that station. Lower salinity was usually recorded for the warmer months near Japan (see Figure 3.6). This could be due to melting of the ice which decreases the concentration of salts or to the effect of seasonal variations in the Kurishio Current [Gill, 1982]. The range of values for sea surface temperature and salinity used in this analysis after all the filtering was performed on the data is presented in the two histograms shown in Figure 3.7.
Ch. 4 Conclusions
S. Cruz-Pol
58
(a)
(b)
Figure 3.7 Histograms of the range of (a) salinity and (b) sea surface temperature values for the data used in this work.
Ch. 4 Conclusions
S. Cruz-Pol
59
Typical values for salinity range between 34 to 36 with a few showing up at around 30 . Sea surface temperature values ranges from 5 to 30 oC [278 to 303 K].
3-3 Analysis and Results
3-3.1 Model for TB using raob, NODC, and altimeter data

Raob profiles were used to compute the downwelling and upwelling brightness temperatures as well as the opacity at each of the two frequencies. These correspond to the terms, TDN, TUP and -(0,H) in equation (1.8). In addition, values for ocean salinity and surface temperature were taken from averages of the NODC data set. These are used to compute the sea surface emissivity, , according to E96 and
KS77 models. The wind was found from Topex altimeter data using the MCW model [Witter and Chelton,1991], with a -0.63dB correction to the 0 values prior to using the model [Callahan et al.,1994 ].
3-3.2 Selection of the Maximum Time and Space Separation

Only data close to the Topex ground track in time (less than 6 hours) and proximity (less than 300 km) were employed in the analysis. The closest data point from TMR for every raob station measurement was used for comparison. The temporal separation of 6 hours was chosen because the raob balloons are usually launched every 12 hours. The separation between two TOPEX passes near the equator ground track is approximately 150km, therefore, choosing 300km for spatial separation allows for more than one pass be near to a raob station in case the other satellite pass in farther than 6 hours in time.
Ch. 4 Conclusions
S. Cruz-Pol
60
The spatial and temporal separation filters were evaluated by computing the RMS5 difference and total number of available raob profiles corresponding to different sets of distance-time limits between the TMR and raob data sets. The resulting RMS and number of profiles are depicted in Figures 3.8 and 3.9 as 3-D surface plots in which the two horizontal axes are the limits set on time (hrs) and space (km). As shown in Figure 3.9, the number of available profiles increases with larger limits on time and space. The RMS error also increases with more profiles (Fig. 3.8), but grows slowly up to distance and time separations of 300 km and 6 hours, respectively. For this reason, the limit in time and distance was kept at 300 km and 6 hours, respectively, in order to maximize the number of data points used while maintaining the RMS at a reasonably low level.
Figure 3.8 Variation of the number of raob profiles used depending on the limits in space and time separation imposed on the data.
The RMS computed here is the root-mean square difference between the TMR measured TBs and the modeled TBs
Ch. 4 Conclusions
S. Cruz-Pol
61
Figure 3.9 Variation of the RMS difference between data and model depending on the limits in space and time separation imposed on the data.
After all the data were filtered for clouds, low wind speed, path delay, and space and time collocation, we are left with a total of 263 raob profiles available with corresponding Topex altimeter and radiometer data. The total number of data points is then 526 since we are using two frequency channels, 18 and 37 GHz.
Ch. 4 Conclusions
S. Cruz-Pol
62
3-3.3 Evaluation of the Model Performance
The performance of the dielectric models of the ocean was evaluated by considering the dependence of the errors in the models on frequency, sea surface temperature and salinity, as well as by computing their RMS difference and bias with respect to the TMR measured data. The frequency dependence was determined as, frequency dependence = (3.13)
where aveTf is the error in brightness (TBTMR - TBmodel) averaged over all 263 data points at the frequency f. This parameter is very significant, since it is an indication of the confidence with which the model can be extrapolated to higher frequencies. The sensitivity of errors in the ocean model to temperature and salinity is determined by considering the R2 value of a linear fit to the plot of TB error versus a particular variable, Tsea or S. In this context, the R2 is a measure of how much of the error is dependent on the variable. Therefore, the smaller this value is, the better, since errors in the model should not be sensitive to either of these two variables. Figure 3.10 shows a plot and linear fit of the ocean model error versus Tsea for E96. The R2 value is found to be small. This is an indication that this model is adequate in the sense that errors are not highly dependent on the sea surface temperature of the ocean. The R2 is calculated for all models later in this section.
Ch. 4 Conclusions
S. Cruz-Pol
63
Figure 3.10. Plot of the model error (TBTMR - TBmodel) versus the sea surface temperature for E96. The R2 value of the linear fit is shown to be small, denoting a small dependence of the error in this model on the sea surface temperature..
Another important indication of the proper performance of the models is the bias, since this is a major determinant of the accuracy of the model. As mentioned before, in the upcoming satellite mission, JASON, scheduled to be launched in 2000 [JPL, 1998], the absolute calibration is performed by occasionally looking at calm water. Consequently, the quality of the calibration depends strongly on the accuracy of a model for the calm water emission. The obtained values for the RMS difference, bias, and dependence on salinity, temperature and frequency are shown on Table 3.2 for both ocean emissivity models, E96 and KS77. Both models are shown first with the L93 atmospheric absorption model. The models are also shown using the new atmospheric model developed in Chapter 2.
Ch. 4 Conclusions
S. Cruz-Pol
64
TABLE 3.2 Comparison of overall performance of several Ocean emissivity models with respect to TMR data. Model
Ocean Atmospher e KS77 L93 E96 KS77 E96 L93 New New Overall RMS [K] 3.55 3.27 3.28 3.45 RMS [K] RMS [K] Bias [K] Bias [K] Salinity dependence T sea dependence frequency dependenc e
18GHz 37GHz 18GHz 37GHz 18GHz 37GHz 18GHz 37GHz 2.54 3.01 2.63 3.27 4.29 3.35 3.71 3.34 -0.16 -1.63 -.67 -2.14 2.71 0.66 1.62 -.41 0.090 0.099 0.082 0.101 0.009 0.022 0.011 0.024 0.004 .0001 0.002 .0008 0.026 .002 0.018 .0003 -2.88 -2.30 -2.30 -1.74
As seen in Table 3.2 above, all combinations of models have a negligible dependence on salinity and sea surface temperature. On the other hand, the frequency dependence of KS77-L93 is very large, 2.88K. This is not surprising, since this model was meant to be valid only for frequencies less than 10 GHz, although it is commonly used for higher frequencies. The E96-L93 model improves the
frequency dependence (down to -2.30K) as well as the RMS and bias. The RMS and bias shown in the first two entries of Table 3.2 agree with results previously presented by Ellison et al. [1996]. They show an improvement in the RMS with their E96 ocean model over KS77, as well as a lower bias, when using L93. However, when the new atmospheric model is applied, the RMS and bias for the KS77 model are superior. On the other hand, E96 maintains its superior frequency dependence. This is to be expected since their ocean dielectric model was
developed from measurements at frequencies of up to 40 GHz. For both surface models, the frequency dependence with the new atmospheric model shows a small decrease from the one exhibited when using L93 (2.30K and 1.74K), but this dependence is still quite large when one considers the potential error from extrapolating either model to much higher frequencies (e.g. the 85-90 GHz atmospheric window).
3-3.4 Modified Dielectric Model Parameter Estimation
Ch. 4 Conclusions
S. Cruz-Pol
65
In order to reduce the sensitivity of the error to frequency as well as reduce the RMS difference and bias, both ocean models are parameterized and adjusted to the well calibrated TMR data via the Newton-Raphson method. The performance of these modified models is then evaluated in the same manner as above (see Section 3-3.3). Both ocean dielectric models, E96 and KS77, evolve from a simple Debye equation with different polynomial functions for s, permittivity as , , and , which define the real and imaginary parts of the
(3.14)
and
(3.15)
We introduce two new parameters, which are scaling factors to the real and imaginary parts, i.e. cR, and cI, and are defined by equation (3.2) modified as, (3.16) We modified these parameters for both the KS77 and E96 dielectric models. For the retrieval of the new ocean models real and imaginary dielectric coefficient parameters, cR and cI, the Jacobian is computed as,
(3.17)
where the derivatives are evaluated numerically. The final estimated parameters are found from,
Ch. 4 Conclusions
S. Cruz-Pol
66
(3.18)
where
is the correction to the parameters and is computed from the minimum square error
inversion by
(3.19) The resulting RMS difference, bias, frequency dependence and number of iterations for the modified models are presented on Table 3.36. The first row in the table refers to modifications to nominal KS77, where KS77 is attained when both parameters are unity. The next row in the table corresponds to modifications to E96, where, again, nominal E96 is attained when both parameters are set to one. The modified models from KS77 and E96 are referred to as ModKS and ModE in the remainder of this work.
TABLE 3.3 Comparison of the overall performance of K77 and E96 ocean emissivity models with two modified parameter.
Model cR cI Overall RMS RMS [K] 18 GHz 37GHz 3.03 2.98 2.64 2.58 3.38 3.32 Bias 18 GHz -0.288 -0.161 37GHz 0.2675 0.1405 -0.556 -0.302 3 3 frequency Iterations dependence
ModKS ModE
1.12 1.15
0.961 1.001
The number of iterations indicates how easily the retrieval converges, and hence tell us about the ability of the available data to adequately perform the parameters retrieval. The maximum allowed number of iterations was chosen as ten, since when the iteration number is ten or larger no convergence is attained and the retrieved values are not reliable. Note from Table 3.3, that both of the modified models converge rapidly. For both models, the retrieved parameters are close to unity. Interestingly, both yield similar values of 12-15% higher real part parameter, cR, while the modification to the imaginary part parameter, cI, is smaller, 0.1- 4%.
6
The bias is significantly
The improved atmospheric model is used in the remainder of this work in order to attain a lower frequency dependence and reduce the errors in the ocean model due to the atmospheric uncertainty.
Ch. 4 Conclusions
S. Cruz-Pol
67
decreased to about -0.16K for ModE and to about -.3 for ModKS at both frequencies. The frequency dependence is lowered to -0.56K and -0.30K for ModKS and ModE, respectively. The overall RMS difference for both modified ocean models decreased to 3.03K and 2.98K, respectively, indicating a superior estimation of the ocean brightness. The final estimated parameters are again shown in Table 3.4 for the modified ocean models together with the unmodified KS77 and E96 models. The RMS error in brightness temperature is decreased for both modified models, and the average difference is at most -0.29K. In addition, the frequency dependence is decreased considerably for both modified models, and the temperature and salinity dependence are kept small for the new ocean emissivity models.
Table 3.4. Comparison among Ocean Emissivity Models.

Model Ocean KS77 E96 KS77 E96 RMS [K] Bias [K] Salinity Dependence 18 GHz 0.0901 0.0993 0.0822 0.1012 0.0368 0.0555 37GHz 0.0093 0.0218 0.0112 0.0243 0.0056 0.0154 Tsea Dependence 18 GHz 0.0037 0.0001 0.0017 0.0008 0.0564 0.0219 37GHz 0.0264 0.0021 0.0175 0.0003 0.0410 0.0078 Frequency dependence [K] -2.88 -2.30 -2.30 -1.74 -0.56 -0.30
Atmospher 18 GHz 37GHz 18 GHz 37GHz e L93 2.54 4.29 -0.16 2.72 L93 New New 3.01 2.63 3.33 2.64 2.58 3.35 3.71 3.34 3.38 3.32 -1.63 -0.67 -2.14 -0.29 -0.16 0.66 1.63 -0.41 0.27 0.14
ModKS New (cR=1.12, cI=.961) ModE New (cR=1.15, cI=1.001)
A comparison between the two modified models suggests that ModE has a superior overall performance to that of ModKS. It has the lowest bias, which is a very important attribute indicative of the accuracy of the model. Its frequency dependence is half of that exhibited by ModKS, which will allow for more reliable extrapolation to higher frequencies. For instance, a frequency dependence of .30K, which was computed for a frequency change of 37-18 =19 GHz , means that at 85GHz the expected error will be only -0.30 times (85-18)/(37-18)= 1.06 K. Using the ModKS model yields double this amount. In addition, ModE has a lower dependence on sea surface temperature and lower RMS difference. The salinity dependence is still acceptably small. For these reasons, ModE is the
Ch. 4 Conclusions
S. Cruz-Pol
68
model that we would recommend for future remote sensing applications involving microwave emissions from the ocean.
3-3.5 Error Analysis

A numerical sensitivity analysis was conducted to determine the level of uncertainty in the estimated parameters due to measurement noise by the radiometer, altimeter-derived wind, radiosondes and NODC data and decorrelation between the TMR and raob data. Independent realizations of the entire estimation process were simulated, in which random perturbations were made to the actual measurements. The spatial and time decorrelation between the TMR and raob stations measurements introduce an additional error. This error was estimated by Ruf et al. [1994]. It was found that for an average separation distance of 150 km there is a 2.3 cm spatial decorrelation error in the path delay measurement, and for a mean time separation of 2.9 hours, the time decorrelation error is 1.4 cm. In our data set, the average distance separation is 142 km and the mean time separation is 3.1 hrs. Therefore, these values for decorrelation errors can be used as a conservative estimate. These errors have to be translated in terms of TB. For this reason, the correspondence between path delay and brightness is found at all three frequencies from the slope of their respective scatter plots versus path delay. The slopes are found to be 0.985, 1.99, 0.985 K/cm, respectively at 18, 21, and 37 GHz. From this it can be deduced that the decorrelation error is the same for both the 18 and 37 GHz frequencies, and it approximately doubles at 21 GHz. Consequently, the spatial and time decorrelation errors are estimated at 2.27 K and 1.38 K, respectively, for both the 18 and 37GHz channels. The 21 GHz is not utilized since its enhanced sensitivity to water vapor introduces additional uncertainty in the retrieval of the ocean parameters. Possible biases in the absolute calibration of the radiometer were modeled as an additive constant brightness temperature. Realizations of the TMR biases are selected from a zero mean, normally distributed random process with standard deviation of 0.8 K. Additive random noise in the
Ch. 4 Conclusions
S. Cruz-Pol
69
radiometer data was also modeled. This noise is independent for every channel and radiosonde profile, and is normally distributed with zero mean and 2.95 K (2.65 decorrelation in time and space plus .3K instrument RMS) standard deviation. Radiosonde measurements uncertainties were simulated in the same way as for the radiosonde data set in Chapter 2. However, no pinning error was introduced in the humidity
measurement since it was absent from this data set. The uncertainties in the NODC salinity and temperature readings are modeled as a bias error plus an RMS error. The bias takes into account the spatial variability introduced by averaging data in different locations within a given 10 degree square. Its value was estimated by computing the standard deviation within a given square averaged over a whole year. For sea temperature readings, this bias is found to be 1.0K, hence it was simulated by a normally distributed error with 1.0 K standard deviation. For salinity readings, this value was found to be 0.7 , consequently it was simulated by a normally distributed error with 0.7 standard deviation. The NODC RMS error for both are normally
distributed with zero mean and standard deviations as given by Appendix C. They are varied for each individual reading and for every noise realization. For the altimeter-derived winds, the uncertainties are again modeled as a bias plus an RMS normally distributed error. The bias is the same for wind values at each noise realization and it is a normally distributed with a standard deviation of 0.4m/s (see section 3-.2.1.1). The RMS is also normal, with a standard deviation of 1.4m/s. The two parameters, cR and cI, were repeatedly estimated with independent errors added to the data, to obtain 1,000 simulated noise realizations. For ModKS, the standard deviations in the cR and cI parameters are found to be 0.031 and 0.022, respectively, with a correlation of 0.574. The standard deviations in the cR and cI parameters are 0.040 and 0.022, respectively, with a correlation of 0.303, for the modified E96 model. Note that the changes in significant. The 4% change in (of 12% and 15%) are, therefore, statistically for E96, is significant
for KS77, but not the 0.1% change in
relative to the 2.2% error in the change.
Ch. 4 Conclusions
S. Cruz-Pol
70
The largest sources of error in determining the retrieved parameters was found to be the uncertainties in the salinity and sea surface temperature readings from NODC, followed by TMR instrument calibration errors, and then by the spatial and temporal decorrelation uncertainty. Uncertainties in the NODC data added errors of 0.0448 and 0.0081 to cR and cI, respectively. The error contribution from TMR instrument calibration was found to be 0.0219 and 0.0185, whereas the TMR decorrelation error was 0.0165 and 0.0056, for each of the retrieved parameter. The contribution from the wind was 0.0073 and 0.0022, whereas, the raob measurements added errors of 0.0012 and 0.0012 to both parameters, cR and cI, respectively. The effect that the errors in the parameters have on the dielectric model was addressed by a second noise simulation. In this case, 1,000 independent realizations were simulated in which and
were computed for each of the 263 profiles used by the estimation algorithm versus frequency. At each realization, the two parameters were randomly perturbed according to the statistics given above. The resulting values for dielectric parameters are plotted versus frequency in Figure 3.11.
Ch. 4 Conclusions
S. Cruz-Pol
71
Figure 3.11 The modified and nominal ocean dielectric permittivity models, ModKS and KS77(in pink) and ModE and E96 (in blue). The plots show the variation in both the real and imaginary parts of the permittivity versus frequency. The error bars denote the standard deviations in the modified models. All plots are for Tsea=280K and S= 35. As seen in Figure 3.11, the real part of the ocean dielectric coefficient is increased for both modified models. The imaginary part of the modified KS77 is forced down toward the E96 model and the E96 is only slightly adjusted, showing the superiority of the original E96 over KS77. The effect that the errors in the modified parameters have on the total ocean emissivity model was addressed by a similar noise simulation. Once again, 1,000 independent realizations were
simulated in which the emissivity versus frequency were computed for each of the 263 profiles used by the estimation algorithm. At each realization, the real and imaginary part parameters were randomly perturbed according to the statistics given above. The resulting values for ocean emissivity are plotted versus frequency in Figure 3.12.
Ch. 4 Conclusions
S. Cruz-Pol
72
Figure 3.12 Error in the modified ocean emissivity, Mod KS (in pink) and ModE (in blue) versus frequency. The error bars denote the standard deviations at each point. Plot is for Tsea=280K and S=35. As seen in Figure 3.12, at low frequencies the emissivities according to KS77 and E96 are both modified so as to approach a value in between them and within the error bars. For higher frequencies, both modified models still agree with each other and with E96 within their error bars, but KS77 predicts a statistically significant lower emissivity. The average error in the modified emissivity models, over the range 18-40 GHz, is found to be 0.0037 and 0.0035, for ModE and ModKS, respectively. In terms of brightness temperature, this error translates into approximately, 0.0037 290K, or 1.07K.
3-4 Conclusion
Ch. 4 Conclusions
S. Cruz-Pol
73
Recent work to determine the sea water dielectric coefficient was based on laboratory measurements of sea water samples from different parts of the ocean. Although these measurements should render good understanding of the emission from a calm ocean surface, their accuracy in providing values of the ocean still needed to be examined. Our present investigation of the specular sea emission seen from space provides field verification of the sea water specular emissivity over broader regions of the oceans. In this work, we investigate and adjust two ocean dielectric models using well calibrated radiometer data from the TOPEX/Poseidon satellite mission, paying particular attention to reducing the frequency dependence of the model and the overall bias of the estimated brightness. In addition, we evaluate the performance of several models for their dependence on salinity and sea temperature. The modified models exhibit significant improvements in the estimate of TB. Of the two modified models, ModE exhibits superior overall performance. It has the lowest bias at both
frequencies (0.16 and 0.14K, respectively), which is indicative of the accuracy of the model. Its frequency dependence was decreased from -2.3 to 0.30K, which is half of that exhibited by ModKS, and which will allow for more reliable extrapolation to higher frequencies. In addition, ModE has the lowest dependence on sea surface temperature and the lowest RMS difference of 2.58K and 3.52K for 18GHz and 37GHz, respectively. For these reasons, we recommend this model7 for future remote sensing applications involving microwave emissions from the ocean.
See Appendix E for a FORTRAN program listing of the modified ocean surface specular emissivity model
Ch. 4 Conclusions
S. Cruz-Pol
74
4. CONCLUSIONS
4.1 Case Study: Relevance of this work to the TOPEX/Poseidon altimetry mission
The atmospheric and sea surface emissivity models are the two primary components of a total model for the brightness temperature seen from a satellite. Many other factors, both from theoretical models and instrumental errors, contribute to the error budget that determines the overall accuracy of a satellites measurements. Table 4.1 places the water vapor attenuation and sea surface emissivity model uncertainties into the context of the total error budget for the retrieved path delay algorithm used by the TOPEX Microwave Radiometer. The individual components of the error are described by Keihm et al. [1995] and paraphrased here: Inherent - This error is due to the fact that the relationship between TB and PD is not a one-toone correspondence. Instead, there are a multiple number of possible water vapor profiles which yield the same brightness temperature but different path delays. Vapor Absorption Model - This refers to the uncertainty in the water vapor absorption model which can produce both offset and scale errors in the path delay retrieval. Oxygen absorption model - The effect of the uncertainty in the oxygen absorption model was assessed by considering a simplified global average version of the path delay retrieval algorithm. Liquid absorption model - This is the uncertainty in the model for the cloud liquid water content. Specular sea surface emissivity model - This is the path delay retrieval error due to the uncertainty in the sea surface emissivity model.
Ch. 4 Conclusions
S. Cruz-Pol
75
Emissivity vs. Wind speed model - This is the uncertainty introduced by the wind speed retrieval model used by TMR. The path delay retrieval varies with the estimate of wind speed. Biases in the wind speed estimate will bias the path delay. The first column in Table 4.1 is the pre-launch error budget for the TMR path delay algorithm
as presented by Keihm et al. [1995]. In the second column, we present the errors using our improved models for the water vapor and sea surface emissivity. An improvement of 37% is attained in the overall PD error budget when the results from this work are applied.
Table 4.1. Error Budget for the Path Delay Algorithm
Error Source
Inherent Vapor abs. Model Oxy. Abs. Model Liq. Abs. Model Specular sea surface emissivity model Emissivity vs. wind speed model RSS algorithm Error
PD error [cm] Nominal 0.37 0.80 0.05 0.03 0.20 0.21 0.93 New 0.37 0.40 0.05 0.03 0.02 0.21 0.59
In addition to the error in the path delay algorithm, the overall error budget for the wet troposphere correction includes other uncertainties [Keihm et al., 1995]: Antenna Temperature Calibration and Beam Pattern correction - This takes into account the accuracy of the TMR brightness temperature measurements including stochastic noise, prelaunch calibration residuals, and the antenna pattern correction error. Decorrelation between TMR and Altimeter main beams - This takes into account the difference in the beamwidth of the TMR channels (tens of kilometers) and the assumed equivalence of the path delay in the smaller footprint of the altimeter (~3 km). Beam Size Differences for 3 TMR Channels - This takes into account the difference in the beamwidths of the individual TMR frequency channels (43.4 km at 18 GHz; 36.4 km at 21 GHz, and 22.9 km at 37 GHz)
Ch. 4 Conclusions
S. Cruz-Pol
76
Path Delay Retrieval Algorithm Error - This is the error in the path delay retrieval algorithm presented in Table 4.1. These error sources are presented in Table 4.2. Table 4.2. Total Error Budget for TOPEX Microwave Radiometer (TMR) Wet Troposphere Range Correction. [Keihm et al., 1995]
Error Source
Antenna Temperature Calibration and Beam Pattern correction Decorrelation Between TMR and Altimeter Main Beams Beam Size Differences for 3 TMR Channels Path Delay Retrieval Algorithm Error RSS Total Error
PD error (cm) 0.69 0.30 0.11 0.93 1.20
In the case of the TOPEX/Poseidon altimeter, we are interested in the reliability and accuracy of its sea surface height measurements, since it is used primarily for the global monitoring of the ocean topography. Factors such as the precise orbit determination, gravitational and ocean tidal forces, solar radiation effects, atmospheric drag, altimeter noise, etc. have to be accounted for when determining the accuracy of such measurements. A complete error covariance model of the data for the sea surface topography is presented by Tsaoussi and Koblinsky [1994] and briefly summarized here. The altimeter measures the distance between the satellite and the sea surface to obtain a detailed map of the global topography. The sea surface height is obtained by subtracting the altimeter range measurements from the altitude of the satellite above a reference ellipsoid. The uncertainty in this sea surface height measurement is therefore dependent on the accuracies of the altimeter and the precise knowledge of the position of the satellite in space. The position of the satellite is determined by three different systems: Satellite Laser Ranging (SLR); Doppler Orbitography and Radiopositioning Integrated by Spacecraft (DORIS); and Global Positioning System (GPS). SLR uses laser beams sent from the ground and reflected from a laser reflector array to determine the exact position of the spacecraft. DORIS uses a radio tracking system developed by CNES. The satellite also carries a GPS
Ch. 4 Conclusions
S. Cruz-Pol
77
receiver on board which tracks signals sent by an array of 21 satellites that orbit the earth to pinpoint the precise position of TOPEX/Poseidon in space. These systems provide the spacecrafts radial position with an accuracy of better than 3 cm. Table 4.3 presents a list of errors encountered in the retrieval of the sea surface height for the model, pre-launch, post-launch and post-verification phases [Nerem et al., 1994; Tsaoussi and Koblinsky, 1994; Fu et al.., 1994; Keihm et al., 1995]. Sources of error include;
Table 4.3. RMS Errors of Individual Sea Surface Topography Error (units in centimeters [Tsaoussi and Koblinsky, 1994; Fu et al., 1994]
Error Source
Altimeter Noise EM Bias Ionosphere Dry troposphere Wet troposphere Atmospheric Load Ocean Tides Solid Earth tides Radial orbit height Gravity field High-frequency geoid 8 Total Error Total time dependent Error
Model 0.2 0.7 0.8 1.0 1.8 1.1 1.7 0.3 2.3 10.9 4.8 11.5 3.5
Prelaunch 2.0 2.0 2.2 0.7 1.2 2.8 n/a n/a 12.8 n/a n/a n/a 13.4
Postlaunch 1.2 2.0 0.7 0.7 1.5 2.8 n/a n/a 8.0 n/a n/a n/a 8.6
Postverification 1.7 2.0 0.5 0.7 1.1 n/a n/a n/a 3.5 n/a n/a n/a 4.7
Altimeter noise- This include white noise in the instrument components and mispointing and skewness effects. These combined altimeter errors are found to be less than 1 cm [Fu et al., 1994].
EM bias- Another error in the sea surface height measurement is the electromagnetic (EM) bias. The EM bias refers to the fact that the radar backscatter cross section is larger at wave troughs than at wave crests [Walsh et al, 1989]. For a typical 2-m SWH (significant wave height) the residual EM bias is about 2 cm.
includes the gravity field (geoid error)
Ch. 4 Conclusions
S. Cruz-Pol
78
Ionosphere - The range delay caused by the ionospheric free electrons is retrieved by the dualfrequency altimeter (see Section 1-1.2.1). Error in the retrieval of the ionospheric range delay is about 0.5 cm [Imel, 1994].
Wet Troposphere - The water vapor in the atmosphere is responsible for the wet propagation delay of the radar signal. The TMR is used to determine this wet path delay. Comparisons of TMR observations with ground based water vapor radiometers and radiosondes yield an estimated accuracy of 1.2 cm [Ruf et al.,1994].
Dry Troposphere - The dry troposphere delay in the altimeter signal is caused by the dry air mass of the troposphere. This delay is corrected by using the sea level pressure estimates from ECMWF. The RMS accuracy of this correction is estimated to be 0.7 cm.
Atmospheric Drag - The acceleration of the spacecraft caused by its interaction with the Earths atmosphere causes a drag on the satellites orbit. This atmospheric drag is easily modeled at the relatively low atmospheric density at the corresponding high altitude (1336 km). Errors in the modeled atmospheric load account for 2.8 cm or less [Tsaoussi and Koblinsky, 1994].
Ocean Tides - The natural rise and fall of sea level due to the pull of gravity among the Moon, Earth and Sun change the orbit of artificial satellites such as TOPEX. The error in this model has been estimated to be approximately 1.7 cm [Casotto,1989].
Solid Earth Tides - Another force acting on the satellite is generated by the inhomogeneous mass distribution on and within the Earth. Errors in the modeled solid earth tides are estimated at 0.3 cm [ Rosborough, 1986]
Radial orbit height - The uncertainty in the radial component of the satellite orbit is the largest error source in satellite altimetry. The post launch gravity improvement activities, which include comprehensive tracking of the satellite by SLR and DORIS and improvements in the force modeling and reference systems and numerical methods, have resulted in an RMS accuracy of approximately 3.5 cm [Tapley et al., 1994].
Ch. 4 Conclusions
S. Cruz-Pol
79
Gravity field - This uncertainty refers to the error in the model for the gravity field effect. It is estimated at about 11 cm [Lerch et al., 1994]. Most of this error is random and can be reduced by time averaging [Fu et al., 1994].
High-frequency geoid - This error relates to the exact size and shape of the Earth and the determination of the exact satellite position with respect to the geoid9 [Tapley et al., 1994]. The total RSS error and the total time-dependent error for each phase are presented in the
bottom two rows of Table 4.3. Post-launch tuning of all the physical models mentioned allows the non-conservatives forces acting on TOPEX to be modeled to the required accuracy. Consequently, some of the errors at pre-launch show considerable improvement in the post launch and verification phases. As seen in Table 4.3, the gravity field (geoid) error dominates the error budget on the sea surface topography. However, this error cancels out when performing time-averaging for the data. For the post-verification phase, the total time-dependent error reduces to 4.7 cm, of which 1.1cm is due to the wet troposphere uncertainty. Comparisons of the TOPEX measured sea level variation to the Tropical Ocean and Global Atmosphere data set yield an average RMS difference of 4.6 cm after smoothing the tide gauge data for temporal averaging [Nerem et al., 1994]. These results corroborate the level of the error presented in Table 4.3s post- verification stage of 4.7 cm. At a first glance, a wet tropospheric path delay of 1.2 cm looks insignificant compared to a total (pre-launch) error of 13.4 cm. However, as seen in the post-launch and model columns of Table 4.3, the significance increases compared to a total error budget of 3 to 4.7 cm. Improvements in the accuracy of the wet troposphere propagation path delay render more accurate measurements from the TOPEX altimeter mission.
4.2 Conclusions and future work

The contributions of this work are the improved models for the atmospheric water vapor absorption and the sea surface emissivity. The improved model for the absorption of the clear
Average sea level of an ocean at rest.
Ch. 4 Conclusions
S. Cruz-Pol
80
atmosphere near the 22 GHz line is presented in Chapter 2. The Van-Vleck-Weisskopf line shape is used with a simplified version of the model by Liebe [1987] for the water vapor absorption spectra and the model by Rosenkranz [1993] for the oxygen absorption. Radiometric brightness temperature measurements from two sites of contrasting climatological properties, San Diego, CA and West Palm Beach, FL, were used as ground truth for comparison with in situ radiosonde derived brightness temperatures. Retrieval of the new models four parameters, water vapor line strength, line width, and continuum absorption, and far-wing oxygen, was performed using the Newton-Raphson inversion method. In addition, the Hill line asymmetry ratio was evaluated for several currently used models, showing agreement of the radiometric data with the VVW water vapor line shape, and ruling out atmospheric absorption models using the Gross line shape near 22 GHz given by Waters [1976] and Ulaby et al. [1981]. The RMS difference between modeled and measured TB was reduced by 23%, from 1.36 K to 1.05 K, with the new parameters. Sensitivity analysis shows that the standard
deviations in the CL, CW, CX parameters are 5% or less, and 8% for CC, assuming 0.5K RMS errors in the TB data. The extra frequencies over the 20-32 GHz range constrain the shape and level of the absorption model simultaneously, producing the highest agreement with the radiometric temperatures. In order to reduce the correlation in the retrieved atmospheric parameter for the continuum and the oxygen cluster parameters, CC and CX, future experiments should include more variation in the air pressure within the data set. In addition, to avoid the painstaking process of selecting the raob data less affected by the relative humidity problem, more accurate raob balloons should be launched close to the radiometer sites. In Chapter 3, an analysis is presented to examine and adjust two ocean dielectric models using well calibrated radiometer data from the TOPEX/Poseidon satellite mission together with NODC salinity and sea surface temperature depth-profiles, and atmospheric profiles from 15 raob stations around the world. Particular attention was paid to reducing the frequency dependence of the model and the overall bias of the estimated brightness. In addition, we evaluated the performance of several models for their dependence on salinity and sea temperature.
Ch. 4 Conclusions
S. Cruz-Pol
81
The modified models, ModE and ModKS, exhibit significant improvements in the estimate of TB. Of the two modified models, ModE exhibits superior overall performance, including the lowest bias at both frequencies, which is a very important attribute indicative of the accuracy of the model. Its frequency dependence was decreased to 0.30K, which will allow for more reliable extrapolation to higher frequencies. In addition, ModE has the lowest dependence on sea surface temperature and the lowest RMS difference for both 18GHz and 37GHz. Consequently, this is the model that we
recommend for future remote sensing applications involving microwave emissions from the ocean emissivity of the ocean. The average error in the modified emissivity model, over the range 18-40 GHz, is found to be 0.0037, compared to 0.003 for E96, which in terms of brightness temperatures, translates into a model error of approximately 1K. We found that the dominant source of errors in determining the modified ocean dielectric models were the uncertainty in the salinity and sea surface temperature data from NODC. For this reason, a future experiment should provide more accurate readings of sea surface salinity and temperature.
Ch. 4 Conclusions
S. Cruz-Pol
82
References
Becker, G. E., and S.H. Autler, Water Vapor Absorption of Electromagnetic Radiation in the Centimeter Wave-length Range, Phys. Rev. , Vol. 70, pp. 300-307, 1946. Ben-Reuven A., The Meaning of Collision Broadening of Spectral Lines: The Classical-Oscillator Analog, Adv. in Atm. and Molec. Physics, Ed. by D. R. Bates, I. Estermann, Vol. 5, pp. 201-235, 1969. Bohlander, R.A., Spectroscopy of Water Vapor, Ph.D. Thesis, Dept. of Physics, Imperial College, London , 1979. Born, M. and E. Wolf, Principles of Optics, Pergamon Press, Oxford, Chapter 8, 1965. Bourdouris, G. On the index of refraction of air, the absorption and dispersion of centimeter waves by gasses, J. Res. Natl. Bur. Stand., Sect. D, Vol. 67, pp. 631-684, 1963. Brown, G. S., H. R. Stanley, and N. A. Roy, The wind speed measurement capability of spaceborne radar altimeters, IEEE J. Oceanic Eng., OE-6m 59-63, 1981. Callahan, P. S., C. S. Morris, and S. V. Hsiao, Comparison of TOPEX/POSEIDON 0 and significant wave height distributions to Geosat, J. Geophys. Res., Vol. 99, No. C12, pp. 25,015-25,024, Dec. 1994. Cardone, V. J., Specification of the Wind Distribution in the Marine Boundary Layer for Wave Forecasting, Ph.D. Thesis, Dept. of Meteorology and Oceanography, New York University, New York, 1969. Casotto, S., Ocean tide models for TOPEX precise orbit determination, Ph. D. dissertation, Center for Space Research, The University of Texas at Austin, 1989. Charnock, H., Wind stress on a water surface, Q. J. Royal Meteorol. Soc. 81, pp. 639-640, 1955. Chandrasekhar, S., Radiative Transfer, Dover Publications, Inc., New York, 1960. Cox, C. and W. Munk, Some problems in optical oceanography, J. Mar. Res., 14, 63-78, 1955. Debye, P., Polar Molecules, Dover Reprint, originally pub. Reinhold Publishing Corp., 1929. Dobson, E., F. Monaldo, J. Goldhirsh and J. Wilkerson, Validation of GEOSAT Altimeter-Derived Wind Speeds and Significant Wave Heights Using Buoy Data, John Hopkins APL Technical Digest, Vol. 8, No. 2, 1987. Elgered, G., Tropospheric Radio Path Delay from Ground-based Microwave Radiometry, Atmospheric Remote Sensing by Microwave Radiometry, Ed. By Janssen, Wiley, New York, pp. 240-243, 1993. Ellison, W. J., A. Balana, G. Delbos, K. Lamkaouchi, L. Eymard, C. Guillou and C. Pringent, Study and Measurement of the Dielectric Properties of Sea Water, CETP/CNRS, France, ESTEC/Contract 11197/94/NL/CN, Dec. 1996. Fedor, L. A. and G. S. Brown, Wave height and wind speed measurements from SEASAT radar altimeter, J. Geophys. Res. C87, 3254-3260, 1982. Fu, L. L., E. J. Christensen, C. A. Yamarone Jr., M. Lefebvre, Y. Mnard, M. Dorrer, and P. Escudier, TOPEX/POSEIDON mission overview, J. Geophys. Res., Vol. 99, No. C12, pp.24,369-24,381, 1994.
S. Cruz-Pol
83
Gebbie, H.A., Observations of Anomalous Absorption in the Atmosphere, Atmospheric Water Vapor, pp. 133-141, Ed. by A. Deepak, Academic Press, 1980. Gill, A. E., Atmosphere-Ocean Dynamics, Academic Press, New York, pp. 513,522, 578, 1982. Goff, J. A., Final Report of the Working Subcommittee of the International Joint Committee on Psychrometric Data, Trans. on Amer. Society of Mechanical Engineers, vol. 71, pp. 903-913, 1949. Grody, N. C., Satellite-Based Microwave Retrievals of Temperature and Thermal Winds: Effects of Channel Selection and A Priori Mean of Retrieval Accuracy, Remote Sensing of Atmosphere and Oceans, Ed. by A. Deepak, Academic Press, pp. 381-410, 1980. Gross, E. P., Shape of Collision-Broadened Spectral Lines, Phys. Rev., Vol. 97, pp. 395-403, January 15, 1955. Gnther, H., P. Lionello and B. Hanssen, The impact of the ERS-1 altimeter on the wave analysis and forecast. Report No. GKSS 93/E/44, GKSS Forschungszentrum Geesthacht, Geesthacht, Germany, p. 56, 1993. Hill, R. J., Water vapor-absorption line shape comparison using the 22-GHz line : The Van VleckWeisskopf shape affirmed, Radio Science, Vol. 21, pp. 447-451, May-June 1986. Hill, R. J., R. S. Lawrence, and J. T. Priestly, Theoretical and calculational aspects of the radio refractive index of water vapor, Radio Science., vol. 17, pp. 1251-1257, 1982. Hoehne, W. E., Precision of National Weather Service upper air measurements, NOAA Tech. Memo. NWS T&ED-16, [NITS Pub. PB81-108316], Boulder, CO, 1980 Hogg D.C., F.O. Guiraud, C. G. Little, R.G. Strauch, M. T. Decker, and E. R. Westwater, Design of a Ground-Based Remote Sensing System Using Radio Wavelengths to Profile Lower Atmospheric Winds, Temperature, and Humidity, Remote Sensing of Atmospheres and Oceans, Ed. by A. Deepak, Academic Press, pp. 313-364, 1980. Hollinger, J. P., Passive microwave measurements of sea surface roughness, IEEE Trans. on Geosci. Electron., 9, pp. 165-196, 1971. Hollinger, J. P., J. L. Peirce, and G. A. Poe, SSM/I Instrument Evaluation, IEEE Trans. on Geosci. Rem. Sen., 9, Vol 28, No. 5, pp. 781-790, 1990. Imel, D. A., Evaluation of the TOPEX dual-frequency ionospheric correction, J. Geophys. Res., Vol. 99, No. C12, 1994. Janssen, M.A, An Introduction to the Passive Microwave Remote Sensing of Atmospheres, Atmospheric Remote Sensing by Microwave Radiometry, Ed. By Janssen, Wiley, New York, pp.7-12, 1993. Johansson, J. M., G. Elgered, and J. L. Davis, Geodesy by radio interferometry: Optimization of wet path delay algorithms using microwave radiometer data, Chalmers University of Technology Research Pep. No. 152, Chalmers University of Technology, Gteborg, Sweden, 1987. JPL, JASON Microwave Radiometer: Critical Design Review, Jet Propulsion Laboratory, internal document, January 15 1998. Kagiwada, H. and R. Kalaba, Newtons method for the nonlinear integral equation for the H function of Isotropic scattering, Santa Monica, Calif., Rand Corporation, RM- 6030-PR, 1969. Keihm, S. J., Water Vapor Radiometer Intercomparison Experiment: Platteville, Colorado, March 114, 1991, Jet Propulsion Laboratory, CA, JPL Task Plan 80-3289, July. 1991.
S. Cruz-Pol
84
Keihm S. J., M. A. Janssen, and C. S. Ruf, TOPEX/Poseidon Microwave Radiometer (TMR):III. Wet Troposphere Range Correction Algorithm and Pre-Launch Error Budget, IEEE Trans. on Geosci. Rem. Sens., Vol. 33, No. 1, 1995. Keihm, S. J. and March, Advanced Algorithm and System Development for Cassini Radio Science Tropospheric Calibration, TDA Progress Report 42-127, pp. 1-19, Jet Propulsion Laboratory, Pasadena, CA, November 10, 1996. Keihm, S. J., C. Ruf, V. Zlotnicki and B. Haines, TMR Drift Analysis, Jet Propulsion Laboratory, Internal Report, October 6, 1997. Klein, L. A., and C. T. Swift, An Improved Model for the Dielectric constant of Sea Water at Microwave Frequencies, IEEE Trans. on Antennas Propagation, Vol. AP-25, No. 1, 1977. Lerch, F. J., et al., A geopotential model from satellite tracking, altimeter and surface gravity data: GEM-T3, J. Geophys. Res., Vol 99C, 2,815-2,839, 1994. Liebe, H. J., and D. H. Layton, Millimeter-wave properties of the atmosphere: Laboratory studies and propagation modeling, Nat. Telecom. And Inform. Admin., Boulder, CO, NIT Rep. 87-24, 1987. Liebe, H. J., G. A. Hufford, and M. G. Cotton, Propagation Modeling of Moist Air and Suspended Water/Ice Particles at Frequencies Below 1000 GHz, in AGARD Conference Proceedings 542, Atmospheric Propagation Effects through Natural and Man-Made Obscurants for Visible to MMWave Radiation, May, 1993. Linfield, R. P., S. J. Keihm, L. P. Teitelbaum, S.J. Walter, M.J. Mahoney, R.N. Treuhaft, and L.J. Skjerve, A test of water vapor radiometer-based troposphere calibration using very long baseline interferometry observations on a 21-km baseline, Rad. Sci. 31, pp. 129-146, 1996 McMillin L. M., The Split Window Retrieval Algorithm for Sea Surface Temperature Derived From Satellite Measurements, Remote Sensing of Atmosphere and Oceans, Ed. By A. Deepak, Academic Press, pp. 437-452, 1980. Monin, A. S., V. M. Kamenkovich, and V. Kort, Variability of the oceans, New York, Wiley, 1977. NASA, NASAs Mission to Planet Earth; Earth Observing System, PAM-552, 1993. Nash, J., Elm J. B., and T. J. Oakley, Relative Humidity Sensor Performance Observed In Recent International Radiosonde Comparison, Ninth Symposium on Meteorological Observations and Instrumentation, March 27-31, 1995, Charlotte, NC, American Meteorological Society, pp. 4344, 1995. Nerem, R. S., B. H. Putney, J. A. Marshall, F. J. Lerch, E. C. Pavlis, S. M. Klosko, S. B. Luthcke, G. B. Patel, R. G. Williamson, and N. P. Zelensky, Expected Orbit Determination Performance For The TOPEX/Poseidon Mission, IEEE Trans on. Geosci. Rem. Sen., Vol. 31, No. 2, pp. 333-354, 1993. Njoku E. G, J. M. Stacey, and F. T. Barath, The Seasat Scanning Multichannel Microwave Radiometer (SMR): Instrument description and performance, IEEE J. Oceanic Eng., Vol. OE-5, pp. 100-115, 1980. NODC (National Oceanographic Data Center), Global ocean temperature and salinity profile CDROMs, Washington D. C., 1991. Norberg, W., J. Conaway, D. B. Ross, and T. Wilheit, Measurements of microwave emission from a foam-covered, wind-driven sea, J. Atmos. Sci., Vol. 28, pp. 429-435, 1971. Panofsky, H. A., and J. L. Lumley, The structure of atmospheric turbulence, New York, Interscience Publishers, 1964.
S. Cruz-Pol
85
Paulson, C. A., Profiles of wind speed, temperature and humidity over the sea, Sci. Rept. NSF GP2418, Dept. of Atmospheric Sciences, University of Washington, 1967. Paulson, C. A., F. I. Badgley, and M. Miyake, Profiles of wind, temperature, and humidity over the Arabian Sea, University Press of Hawaii, 1972. Petty G. W. and K. B. Katsaros, The Response of the SSM/I to the Marine Environment. Part I: An Analytic Model for the Atmospheric Component of Observed Brightness Temperatures, J. Atm. and Oceanic Tech., V0l. 9, pp. 746-761, Dec. 1992. Planck, M., The Theory of Heat Radiation; translated by M. Masius. Philadelphia, p. 168, 1914. Pooley, G., Connected-Element Interferometry, Methods of Experimental Physics: Astrophysics, Ed. by M. L. Meeks, Vol. 12C, Chapter 5.2 , pp. 158-173, Academic Press, New York, 1976. Quinn, J. R. and P. J. Wolff, TOPEX/POSEIDON Operational Orbit Determination Results Using Global Positioning Satellites, Sciences Proceedings of the AAS/AIAA Astrodynamics Conference. Part 1 (of 3), v 85, San Diego CA USA p 143-158, Aug 16-19 1993. Rosborough, G. W. Satellite orbit perturbations due to the geopotential, Center for Space Research, Tech. Memo. 86-1, 1986. Rosenkranz, P.W., Absorption of Microwaves by Atmospheric Gases, Atmospheric Remote Sensing by Microwave Radiometry, Chapter 2, Ed. By Janssen, Wiley, New York, 1993. Ruf C. S., S. J. Keihm, B Subramanya, and M. A. Janssen, TOPEX/POSEIDON microwave radiometer performance and in-flight calibration, J. Geophys. Res., Vol. 99, No. C12, pp. 24,91524,926, 1994. Ruf C. S., S. J. Keihm, and M. A. Janssen, TOPEX/POSEIDON Microwave Radiometer (TMR): I. Instrument description and antenna temperature calibration, IEEE Trans. on Geosci. Rem. Sen., Vol. 33, pp. 125-137, Jan. 1995. Ruf C. S., R. P. Dewan and B Subramanya, Combined Microwave Radiometer and Altimeter Retrieval of Wet Path Delay for the GEOSAT Follow-On, IEEE Trans. on Geosci. Rem. Sen., Vol. 34, No.4, July 1996. Semyonov, B. I., Approximate computation of scattering of electromagnetic waves by rough surface contours, Radio Eng. Electron Phys., Vol. 11, pp. 1179-1187, 1966. Shapiro, I.I., D.S. Robertson, C.A. Knight, C.C. Counselman III, A.E.E. Roger, H.F. Hinteregger, S. Lippincott, A. R. Whitney, T.A. Clark, A. E. Niell, and D. J. Spitzmesser, Transcontinental Baselines and the Rotation of the Earth Measured by Radio Interferometry, Science, Vol. 186, p 920, 1974. Shapiro, I.I., Estimation of Astrometric and Geodetic Parameters, Methods of Experimental Physics: Astrophysics, Ed. by M. L. Meeks, Vol. 12C, Chapter 5.6 , pp. 261-276, Academic Press, New York, 1976. Smithsonian Meteorological Tables. 6th rev. ed., prepared by Robert J. List. Washington, Smithsonian Institution, 1966. Snider, J. B., Observed and theorical atmospheric emission at 20, 30, and 90 GHz: Recent results from land- and ocean- based locations, Microwave Radiometry and Remote Sensing of the Environment, Ed. by D. Solimini, VSP, Zeist, The Netherlands, 1995. Stewart, R. L.-L. Fu, and M. Lefebvre, Science opportunities from the TOPEX/POSEIDON mission, JPL Publ. 86-18, Jet Propulsion Laboratory, Pasadena, Calif., 1986. Stogryn, A., The apparent temperature of the sea at microwave frequencies, IEEE Trans. on Antennas Propagation, Vol. AP-15, pp. 278-286, 1967.
S. Cruz-Pol
86
Stogryn, A., Equations for calculating the Dielectric Constant of Saline Water, IEEE Trans on Microwave Theory and Techniques, pp. 733-736, 1971. Stogryn, A., The emissivity of sea foam at microwave frequencies, J. Geophys. Res., pp. 1658-1666, 1972. Stum, J., A comparison between TOPEX microwave radiometer, ERS-1 microwave radiometer, and European Centre for Medium Range Weather Forecasts derived wet tropospheric corrections, J. Geophys. Res., Vol. 99, No. C12, 1994. Tapley, B. D., J. C. Ries, G. W. Davis, R. J. Eanes, B. E. Schutz, C. K. Shum, M. N. Watkins, J. A. Marshall, F. S. Nerem, B. H. Putney, S. M Klosko, S. B. Luthcke, D. Pavlis, R. G. Williamson, and N. P. Zelensky, Precision orbit determination for TOPEX/POSEIDON, J. Geophys. Res., Vol. 99, No. C12, pp.24,383-24,404, 1994. Tsaoussi, L. S., and C. J. Koblinsky, An Error Covariance Model for Sea Surface Topography and Velocity Derived from TOPEX/Poseidon Altimetry, J. Geophys. Res., Vol. 99, No. C12, pp.24,669-24,683, 1994. Ulaby, F. T., R.K. Moore and A. K. Fung, Microwave Remote Sensing: Active and Passive, Vol. 1, pp. 270-278, Addison-Wesley, 1981. Van Vleck, J. H., V. F. Weisskopf, On the Shape of Collisionally Broadened Lines, Reviews of Modern Physics, Vol. 17, pp. 227-236, April-July 1945. Wade, C. G., An Evaluation Of Problems Affecting The Measurements Of Low Relative Humidity On The United States Radiosonde, J. Atmospherical and Oceanic Technology, Vol. 11, No. 3, pp. 687-700, 1994. Walsh, E. J., F. C. Jackson, E. A. Uliana, and R. N. Swift, Observations on electromagnetic bias in radar altimeter sea surface measurements, J. Geophys. Res., Vol 94, 14,575-14,584, 1989. Walter, S. J., Measurement of the Microwave Propagation Delay Induced by Atmospheric Water Vapor, University of Colorado, Ph.D. Thesis, Dept. of Physics, 1990. Walter, S. J. and T. R. Spilker, Microwave Resonator Measurements of Atmospheric Absorption Coefficients: A Preliminary Design Study, JPL Publication 95-14, Jet Propulsion Laboratory, CA, 1995. Waters, J.W., Absorption and Emission by Atmospheric Gases, Methods of Experimental Physics: Astrophysics, Ed. by M. L. Meeks, Vol. 12B, Chapter 2.3 , pp. 142-176, Academic Press, New York, 1976. Wentz, F. J., A Two-Scale Scattering Model for Foam-Free Sea Microwave Brightness Temperature, J. Geophys. Res., Vol. 80, pp. 3441-3446, 1975. Westwater, E. R., The accuracy of water vapor and cloud liquid determination by dual-frequency ground-based microwave radiometry, Rad. Sci. 13, 677-685, 1978. Wilheit, T.T., The Effect of Wind on the Microwave Emission From the Oceans Surface at 37 GHz, J. Geophys. Res., Vol. 84, No. C8, pp. 244-249, 1979. Wilheit, T.T., A Model for the Microwave Emissivity of the Oceans Surface as a Function of Wind Wind Speed, IEEE Trans. on Geosci. Electron., Vol. GE-17, No. 4, pp. 960-972, 1979. Witter, D. L. and D. B. Chelton, A Geosat Altimeter Wind Speed Algorithm and a Method for Altimeter Wind Speed Algorithm Development, J. Geophys. Res, Vol. 96, No. C5, pp. 88538860, May, 1991. Wu, S. T., and A. K. Fung, A noncoherent model for microwave emissions and backscattering form the sea surface, J. Geophys. Res., Vol. 77, pp. 5917-5929, 1972.
S. Cruz-Pol
87
S. Cruz-Pol
88
APPENDIX A Oxygen Microwave Spectrum Parameters

TABLE A-1. Oxygen Microwave Spectrum Parameters [Rosenkranz, 1993] n -1 1 -3 3 -5 5 -7 7 -9 9 -11 11 -13 13 -15 15 -17 17 -19 19 -21 21 -23 23 -25 25 -27 27 -29 29 -31 31 -33 33 fn [GHz] 118.7503 56.2648 62.4863 58.4466 60.3061 59.591 59.1642 60.4348 58.3239 61.1506 57.6125 61.8002 56.9682 62.4112 56.3634 62.998 55.7838 63.5685 55.2214 64.1278 54.6712 64.6789 54.13 65.2241 53.5957 65.7648 53.0669 66.3021 52.5424 66.8368 52.0214 67.3696 51.5034 67.9009 S(To) [cm2/Hz] 2.9360E-15 8.0790E-16 2.4800E-15 2.2280E-15 3.3510E-15 3.2920E-15 3.7210E-15 3.8910E-15 3.6400E-15 4.0050E-15 3.2270E-15 3.7150E-15 2.6270E-15 3.1560E-15 1.9820E-15 2.4770E-15 1.3910E-15 1.8080E-15 9.1240E-16 1.2300E-15 5.6030E-16 7.8420E-16 3.2280E-16 4.6890E-16 1.7480E-16 2.6320E-16 8.8980E-17 1.3890E-16 4.2640E-17 6.8990E-17 1.9240E-17 3.2290E-17 8.1910E-18 1.4230E-17 w [GHz/bar] 1.630 1.646 1.468 1.449 1.382 1.360 1.319 1.297 1.266 1.248 1.221 1.207 1.181 1.171 1.144 1.139 1.110 1.108 1.079 1.078 1.050 1.050 1.020 1.020 1.000 1.000 0.970 0.970 0.940 0.940 0.920 0.920 0.890 0.890 y [bar -1] -0.0233 0.2408 -0.3486 0.5227 -0.543 0.5877 -0.397 0.3237 -0.1348 0.0311 0.0725 -0.1663 0.2832 -0.3629 0.397 -0.4599 0.4695 -0.5199 0.5187 -0.5597 0.5903 -0.6246 0.6656 -0.6942 0.7086 -0.7325 0.7348 -0.7546 0.7702 -0.7864 0.8083 -0.821 0.8439 -0.8529 v [bar -1] 0.0079 -0.0978 0.0844 -0.1273 0.0699 -0.0776 0.2309 -0.2825 0.0436 -0.0584 0.6056 -0.6619 0.6451 -0.6759 0.6547 -0.6675 0.6135 -0.6139 0.2952 -0.2895 0.2654 -0.259 0.375 -0.368 0.5085 -0.5002 0.6206 -0.6091 0.6526 -0.6393 0.664 -0.6475 0.6729 -0.6545
S. Cruz-Pol
89
APPENDIX B Goff-Gratch Formulation For Water Vapor Density As A Function Of Temperature And Pressure.
The water vapor density, w, is a function of both temperature and pressure, [Goff, 1949] is as follows.
(B.1)
where air is the air density in g/m3 and is given by
(B.2)
where 348.38 is the reciprocal of the gas constant for dry air (i.e., 1/R=1/[287.0410-3] = 348.38 for air in g/m3) and Tv is the virtual temperature is Kelvins
(B.3)
where E is the apparent molecular weight of dry air (28.966 g) over the molecular weight of water (18.016 g), i.e. E = 1.607795. The saturation mixing ratio over water, Rw, is unitless [g/g], and given by
(B.4)
where Fw is defined in equation (B.9), Ew is the saturation vapor pressure of water as given by the Goff-Gratch formulation
(B.5) with (B.6)
S. Cruz-Pol
90
(B.7)
and
(B.8)
In the above equations, P is air pressure in hPa (1hPa = 1mbar), T is air or dew point temperature in Kelvins (see paragraph below), sea, hPa is the U.S. Standard Atmospheric pressure near
K is the boiling point of water, and
(B.9)
Fw is a linear fit to the correction factor for the departure of the mixture of air and water from the ideal gas law [ Smithsonian Meteorological Tables , 1966]. In the above formulation, the air temperature is used for T to find the saturation vapor density,
ws, whereas the dew point temperature is used to find the actual vapor density, w. The relative
humidity is found as
(B.10)
where w, and ws, are as defined above.
S. Cruz-Pol
91
APPENDIX C Table of Mean, Standard Deviation and Counts of Sea Surface Temperature and Salinity per month per raob Station for the period of 1900 to 1990.
Jan Station 6 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 7 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 8 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 9 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 10 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 11 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 12 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts
Feb
Mar
Apr
May
Jun
Jul
Aug
Sept
Oct
Nov
Dec
300.8 300.75 0.6 1.04 34.05 33.96 1.13 0.69 56 18
302.1 303.46 303.66 302.24 301.56 301.39 301.01 301.34 301.83 301.56 0.87 1.01 0.82 0.67 0.72 0.43 0.98 0.99 0.62 0.54 34.49 34.54 35.14 35.36 35.34 33.98 34.98 35.22 34.65 34.56 0.66 0.51 0.49 0.92 2.01 3.46 1.41 0.98 1.24 1.26 87 108 49 36 72 29 31 43 111 93
299.42 300.26 301.41 302.46 303.17 301.89 302.07 301.87 301.11 301.31 301.05 299.98 0.45 1.08 0.75 1.17 0.12 0.25 0.37 0.72 0.32 0.47 0.22 0.86 30.53 31.89 32.55 31.96 32.39 33.09 33.1 32.74 32.7 32.59 31.55 28.15 1.6 1.28 0.49 0.77 0.28 0.32 0.33 0.57 0.91 0.78 0.06 2.88 21 18 135 9 4 2 68 4 41 57 2 31 301.48 301.55 302.41 303.05 302.66 0.36 0.58 0.56 0.56 1.12 33.56 33.94 34.15 34.52 34.65 1.1 0.59 0.54 0.41 0.94 38 100 31 107 39 301.9 301.22 300.91 300.12 300.76 0.72 0.95 1.32 1.32 0.98 34.87 34.68 34.46 34.22 35.09 0.94 1.81 2.54 2.75 0.88 46 55 62 92 98 301.9 301.62 0.73 0.43 34.69 33.65 1.47 1.16 83 53
287.64 286.99 286.91 288.17 291.64 293.89 297.53 299.98 298.69 295.89 293.98 290.74 3.75 4.28 4.29 3.64 3.56 2.88 2.59 1.7 1.91 2.12 2.36 2.83 33.8 34.08 33.9 33.87 34.05 33.75 33.27 33.09 32.79 33.34 33.75 33.77 1.51 1.31 1.68 1.5 1.38 1.68 1.69 1.59 1.81 1.38 1.25 1.26 1986 3670 3393 2460 4796 3097 4449 5826 2882 4608 4222 2800 293.77 293.34 294.52 295.15 297.43 298.73 301.55 302.03 301.07 298.94 2.85 2.95 2.79 4.06 2.58 1.88 1.09 0.81 1.04 1.5 34.69 34.66 34.76 34.44 34.43 33.97 33.95 34.05 34.14 34.23 0.28 0.27 0.24 0.61 0.59 0.99 0.72 0.63 0.81 0.38 348 599 340 428 686 560 513 795 401 664 293.77 293.34 294.52 295.15 297.43 298.73 301.55 302.03 301.07 298.94 2.85 2.95 2.79 4.06 2.58 1.88 1.09 0.81 1.04 1.5 34.69 34.66 34.76 34.44 34.43 33.97 33.95 34.05 34.14 34.23 0.28 0.27 0.24 0.61 0.59 0.99 0.72 0.63 0.81 0.38 348 599 340 428 686 560 513 795 401 664 297.4 296.16 1.4 2.49 34.4 34.55 0.4 0.3 453 137 297.4 296.16 1.4 2.49 34.4 34.55 0.4 0.3 453 137 295.9 1.54 34.67 0.16 160
294.71 293.52 293.65 294.89 296.58 298.94 301.15 301.73 301.53 299.77 297.93 1.63 1.62 1.83 1.81 1.71 1.97 1.41 0.86 1.01 1.26 1.41 34.81 34.86 34.83 34.82 34.77 34.61 34.5 34.47 34.54 34.63 34.63 0.14 0.16 0.16 0.14 0.17 0.23 0.28 0.25 0.19 0.2 0.2 469 306 269 256 365 427 379 443 310 382 344
S. Cruz-Pol
Jan Station 13 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 14 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 20 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 24 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 28 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 29 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts Station 30 Mean Tsea Tsea-Std. dev. Mean Salinity Salinity-Std. dev. Counts 294.98 2.31 34.88 0.13 49 Feb Mar Apr May Jun Jul Aug Sept Oct Nov
92
Dec
295.4 294.99 295.28 297.47 299.87 2.75 2.42 2.73 2.31 1.72 34.89 35 34.94 34.88 34.85 0.26 0.26 0.18 0.17 0.18 89 143 93 58 66
301.9 301.39 301.55 300.68 299.23 297.63 1.05 0.84 0.59 1.09 1.87 1.63 35.05 34.69 34.9 34.8 34.8 34.78 0.36 0.27 0.26 0.22 0.13 0.18 36 191 83 73 60 22
295.01 296.22 295.05 297.22 298.98 299.81 300.88 300.31 300.54 301.64 299.86 297.24 2.08 1.98 2.5 2.75 1.62 1.66 1.46 1.2 1.1 0.4 1.8 2.19 34.75 35.06 35.07 35.07 35.02 34.97 34.91 34.83 35 34.73 34.95 34.79 0.31 0.19 0.15 0.18 0.2 0.22 0.36 0.23 0.06 0.06 0.22 0.2 19 26 38 38 31 74 25 82 2 2 24 26 278.52 281.22 281.27 281.62 281.73 279.96 280.52 1.48 4.17 3.01 4.57 11.83 2.62 4.5 34.01 34.01 34.09 34.17 34.1 33.99 34.23 0.09 0.5 0.39 0.67 1.32 0.38 0.7 2 31 4 11 3 5 2 300.69 300.34 299.69 300.23 300.76 301.03 301.02 0.57 0.45 0.72 0.44 0.35 0.35 1.23 34.45 34.64 34.63 34.8 34.69 34.69 34.62 0.25 0.29 0.23 0.24 0.22 0.31 0.23 23 9 20 9 95 78 38 277.6 285.54 286.57 288.24 279.22 2.38 0.79 0 0 3.43 33.87 34.94 35.17 35.44 34.2 0.07 0.11 0 0 0.44 2 2 1 1 3 301.5 301.95 302.03 301.52 300.52 0.42 0.23 0.17 0.14 0.77 34.8 34.75 34.67 34.38 34.52 0.22 0.18 0.15 0.1 0.29 57 35 9 16 25
301.37 301.14 301.52 300.89 299.92 299.12 298.64 298.32 298.64 299.68 301.44 301.32 0.57 0 0.48 0.79 0.65 0.75 1.27 1.01 1.19 0.72 1 1 35 36.11 34.03 34.57 34.61 34.93 35.5 35.26 35.31 35.42 35.28 35.49 0.72 0 0.93 0.9 0.3 0.17 0.47 0.21 0.29 0.27 0.35 0.53 7 1 14 18 54 16 11 9 27 21 46 22 299.61 299.41 299.43 299.65 298.41 296.74 295.23 0.78 0.69 1.47 1.17 1.06 1.15 1.35 35.32 35.53 35.49 35.37 35.5 35.53 35.61 0.26 0.27 0.26 0.16 0.21 0.2 0.12 78 236 30 144 286 73 33 294.9 296.78 295.99 297.97 298.63 1.5 0.86 0.68 1.04 1.16 35.54 35.56 35.64 35.8 35.15 0.21 0.25 0.13 0.23 0.24 43 275 19 161 112 298.5 298.77 299.81 299.67 0.52 0.95 0.46 0.87 34.42 34.53 34.1 34.64 0.2 0.23 0.36 0.22 7 14 5 8
300.51 300.83 301.44 300.83 300.35 299.11 298.58 298.12 0.89 0.25 0.67 0.93 0.48 1.46 1.99 1.89 34.52 33.85 34.37 34.21 34.18 34.59 34.43 34.39 0.39 0.19 0.34 0.26 0.07 0.29 0.38 0.34 7 2 4 10 7 11 6 8
S. Cruz-Pol
93
APPENDIX D FORTRAN Program for the New Atmospheric Model

C C C C C C C C PROGRAM ATMOSPHERE COMPUTES THE ATMOSPHERIC ABSORPTION FOR THE 18-32 GHZ FREQUENCY RANGE INPUTS: FREQ= FREQUENCY [GHZ], P= AIR PRESSURE IN MBAR [HPA] T= AIR TEMPERATURE [K] V= VAPOR DENSITY [ G/M3] OUTPUT: AtmAbsorption= ATMOSPHERIC ABSORTION [NP/KM] REAL CL,CW,CC,CX REAL P,V,T,FREQ REAL W300(34),FR(34),Y300(34),S300(34),VP(34) COMMON FN(34), SN(34), WN(34), YN(34),VN(34) CALL OXYTABLE() CL= 1.0586 CW= 1.0665 CL= 1.2870 CL= 1.0441 Avapor=WATER(FREQ,V,P,T,CL,CW,CC) Aoxygen= OXYGEN(FREQ,V,P,T,CX) AtmAbsorption= Avapor + Aoxygen RETURN END * FUNCTION WATER(FREQ,V,P,T,CL,CW,CC) ************************************ * COMPUTE VAPOR ABSORPTION AS A FUNCTION OF FREQUENCY (F), * * VAPOR DENSITY (V), PRESURE (P), TEMPERATURE (T), LINE * * STRENGTH FACTOR (CL), CONTINUUM STRENGTH FACTOR (CC), * * AND LINE WIDTH FACTOR (CW). * * RESULT IN NEPERS/KM * *********************************** REAL V,P,T,WATER,FREQ,CL,CW,CC,FZ,TP,PVAP,PDRY REAL TERM1,TERM2,TERM3 FZ=22.235 TP=300.0/T PVAP=V/0.7223/TP PDRY=P-PVAP W=CW*0.002784*(PDRY*TP**0.6+4.80*PVAP*TP**1.1) w2=w*w TERM1=CL*0.0109*PVAP*TP**3.5*EXP(2.143*(1.0-TP)) TERM2=(W/FZ)*(1.0/((FZ-FREQ)*(FZ-FREQ)+W2)+1.0/((FZ+FREQ)* 1 (FZ+FREQ)+W2)) TERM3=CC*0.1*(1.13E-07*PDRY*TP**0.5+3.57E-06*PVAP*TP**8.0)* 1 PVAP*TP**2.5 WATER=(1.0/4.34)*0.1820*FREQ*FREQ*(TERM1*TERM2+TERM3) RETURN END * FUNCTION OXYGEN(FREQ,V,P,T,CX)
S. Cruz-Pol ********************************* * ATMOSPHERIC ABSORPTION DUE TO OXYGEN WITH DEPENDENCE * * ON TEMPERATURE, PRESURE, VAPOR DENSITY AND FREQUENCY. * * RESULT IN NEPERS/KM ********************************* REAL V,P,T,CX,OXYGEN,FREQ,TH,TH1,B,X,Y,WB300 REAL PRESWV,PRESDA,DEN,DFNR,SUM,BFAC,DF,SF1,SF2,STR REAL FN(34), SN(34), WN(34), YN(34),VN(34) COMMON W300,FR,Y300,S300,VP WB300=0.56 X=0.8 BFAC=0.0 TH=300.0/T TH1=TH-1.0 B=TH**X PRESWV=V*T/217.0 PRESDA=P-PRESWV DEN=0.001*(PRESDA*B+1.1*PRESWV*TH) DFNR=WB300*DEN f2= FREQ*FREQ SUM=1.6E-17*F2*DFNR/(TH*(F2+DFNR*DFNR)) DO 200 K=1,34 FreDiff= FREQ-FR(K) FreSum=FREQ+FR(K) IF (MOD(K,2) .EQ. 0) GOTO 100 BFAC=EXP(-6.89526E-3*K*(K+1)*TH1) 100 DF=W300(K)*DEN df2=df*df Y=0.001*P*B*(Y300(K)+VP(K)*TH1) STR=S300(K)*BFAC SF1=(DF+FreDiff*Y)/(FreDiff*FreDiff+DF2) SF2=(DF-FreSum*Y)/(FreSum*FreSum+DF2) 200 SUM=SUM+STR*(SF1+SF2)*F2/(FR(K)*FR(K)) OXYGEN=CX*0.5034E12*SUM*PRESDA*TH*th*th/3.14159 RETURN END * SUBROUTINE OXYTABLE() ************************************* * Initialize arrays used in the oxygen absorption routine * ************************************ C THE FILE oxytable.txt CONTAINS THE LIST OF OXYGEN LINE C PARAMETERS IN THE ORDER GIVEN BY THE TABLE OF APPENDIX A C FOR INSTANCE, A CORRECT READING RESULTS IS C FN(1)= 118.7503 C FN(2)= 56.2648 C FN(3)= 62.4863, ETC. C REAL FN(34), SN(34), WN(34), YN(34),VN(34) COMMON FN, SN, WN, YN,VN open(unit=3,file=oxytable.txt, status=old) do I=1,34 read *, FN(I), SN(I), WN(I), YN(I),VN(I) end do close (3) END
94
S. Cruz-Pol
95
APPENDIX E FORTRAN Program for the Modified Ocean Surface Emissivity Model
C C C C C C C PROGRAM OceanEmissivity COMPUTES THE Ocean Specular Emissivity at 0 angle of incidence FOR THE 20-40 GHZ FREQUENCY RANGE INPUTS: FRE= FREQUENCY [GHZ], S= SALINITY [PPT] TS= SEA SURFACE TEMPERATURE [K] OUTPUT: Eocean= Ocean Specular Emissivity [UNITLESS] REAL CR,CI REAL S,TS,FRE COMPLEX PER, PERMITIVITY C PER = PERMITIVITY(TS, S, FRE) efactor= abs( (1-sqrt(per))/(1+sqrt(per)) ) Eocean = 1- efactor *efactor END C-------------------------------------MODIFIED ELLISON MODEL---------------------FUNCTION PERMITIVITY(Tk,S,fre) COMPLEX PERMITIVITY REAL T,S,gg,f,tk,fre REAL ER,EI,E0,einf,es,q,tar,t2,t3,t4,t5 C CR=1.147 CI=1.001 pi = 3.141592654 f=fre*1e9 e0 = 8.8419e-12 *fre in [GHz], f in [Hz], Tk in [k], T in [C], S in [ppm] *covert temperature from Kelvin to Celsius t=tk-273.15 t2=t*t t3=t*t2 t4=t*t3 t5=t*t4 *einf= high-frequency dielectric coefficient in F/m ggg= 6.492e-4 einf=6.4587 -.04203*t-.006588*t2 +ggg*t3-1.2328e-5*t4+5.043e-8*t5 *es= static dielectric coefficient a1=81.82-0.060503*t-.031661*t2+3.1097e-3*t3 !-1.1791e-4*t4+1.4838e-6*t5 gg=4.713e-8 a2=.12544+9.4037e-3*t-9.5551e-4*t2+9.0888e-5*t3-3.6011e-6*t4+gg*t5 es=(a1-S*a2) *tar = relaxation time in psec c1=17.303-.66651*t+5.1482e-3*t2+1.2145e-3*t3-5.0325e-5*t4 !+5.8272e-7*t5 cc=-6.272e-3 c2=cc+2.357e-4*t+5.075e-4*t2-6.3983e-5*t3+2.463e-6*t4-3.0676e-8*t5 tar= c1+s*c2
S. Cruz-Pol tar = tar*10**(-12.) *Q= ionic conductivity in mho/m d1 = 0.086374+0.030606*t - 4.121e-4*t2 d2 = 0.077454 + 1.687e-3*t + 1.937 e-5*t2 Q= d1 + s*d2 * the Debye relaxation equation to find complex permittivity-dielectric coefficient pft=pi*f*tar pft2= pft*pft ER= EINF +(ES-EINF)/(1.+4*pft2) EI=((ES-EINF)*2*pft)/(1.+4*pft2)+Q/(2*PI*E0*F) ER= CR*ER EI= CI*EI PERMITIVITY = CMPLX(ER,EI) RETURN END
96

Improved Model For Microwave Brightness Temperature Seen From Space OVer Calm Ocean

Transféré par

Informations du document

Description originale:

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Improved Model For Microwave Brightness Temperature Seen From Space OVer Calm Ocean

Transféré par

Droits d'auteur :

Formats disponibles

The Pennsylvania State University The Graduate School College of Engineering

Mi patria es el mundo, sin fronteras, SLCP

Ch. 1 Theoretical Background

The SSM/I is a seven-channel, four frequency, linearly-polarized, passive microwave

Ch. 1 Theoretical Background

brightness temperature is the vapor absorption model.

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

1-1 Radiative Transfer Equations

Planck function can be simplified to [Ulaby et al., 1981],

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

Figure 1.2. Passive remote sensing with upward-looking radiometer .

1-1.1 Ground-Based Radiometer

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

1-1.2 Satellite-Based Radiometer

Ch. 1 Theoretical Background

Figure 1.4. Passive remote sensing with downward-looking radiometer.

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

1-1.2.1 Radio path delay

Ch. 1 Theoretical Background

Ch. 1 Theoretical Background

2. Microwave Atmospheric Absorption Model

Ch. 1 Theoretical Background

2-1 Atmospheric Absorption

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

In this work pressure is expressed in mbars which is equal to 1 hPascal.

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

2-1.1 Line Shapes

and the Gross Line shape is given by [Gross, 1955]

2-2 Current Models and their limitations

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

Water Vapor Absorption vapor [Np/km]

0.040 0.035 0.030 0.025 0.020 0.015 0.010

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

2-2.1 Atmospheric Absorption Line Shapes

2-2.1.1 The L87R93 Atmospheric Absorption Model

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

, S(T) is the line strength (2.9)

The pressure-broadened line half-width is, (2.11)

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

CX 10% CW 10% CC 10%

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

2-3 Experiment Description and Calibration

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

20.7 GHz channel S. Cruz-Pol 24

22.2 GHz channel

31.4 GHz channel

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

2-3.2 Radiosonde Data

2-3.2.1 Radiosonde Processing

2-3.2.2 Radiosonde Selection and Screening

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model

Ch. 2 Microwave Atmospheric Water Vapor Absorption Model