Full report

Original page 1

Thismicrofichewas producedaccordingto ANSI/ AIlMStandards and m,eet,,the II, ' qualil /specifications ; containedtherein.A poorblowbackimage !s theresultof the characteristicsof the originaldocument.

Original page 2

IIII :_ 'l'i;il[lll(l II_ =" v_.......... NASA Technical Memorandum 104566, Vol. 10 /jv-J/{_ /v: ,j_/ SeaWiFS Technical Report Series Stanford B. Hooker and Elaine R. Firestone, Editors Volume 10, Modeling of the SeaWiFS Solar andLunar Observations Robert H. Woodward, Robert A. Barnes, Charles R. McClain, Wayne E. Esaias, William L. Barnes, and Ann T. Mecherikunnel (NASA-TM-|O4566-VoI-[O) TECHNICAL REPORT MODFLING _F THe LUNAR OBSERVATIONS May 1993 N93-3185q $e_WiFS SERIES. VOLUME IO: $eaWir$ SOLAR AND (NASA) 29 D Unclas Hl/48 017668[

Original page 3

I' I I I I IIlll lln I I I I I i ii i NASA Technical Memorandum 104566, Vol. 10 SeaWiFS Technical Report Series Stanford B. Hooker, Editor NASA Goddard Space Flight Center Greenbelt, Maryland Elaine R. Firestone, Technical Editor General Sciences Corporation Laurel, Maryland Volume 10, Modeling of the SeaWiFS Solar and Lunar Observations Robert H. Woodward General Sciences Corporation Laurel, Maryland Robert A. Barnes CHEMAL, Inc. Wallops Island, Virginia Charles R. McClain, Wayne E. Esaias, William L. Barnes, and Ann T. Mecherikunnel NASA Goddard Space Flight Center Greenbelt, Maryland National Aeronautics and Space Administration Goddard Space Flight Center Greenbelt, Maryland 20771 1993

Original page 4

I II IIIIII I I II II I I I I • i li R. Woodward. R. Barnes. C. McClain, W. EsMas. W. Barnes. and A. Mecherikunnel ABSTRACT Post-launch stability monitoring of the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) will include periodic sweeps of both an onboard solar diffuser plate and the moon. The diffuser views will provide short-term checks and the lunar views will monitor long-term trends in the instrument's radiometric stability. Models of the expected sensor response to these observations were created on the SeaWiFS computer at the National Aeronautics and Space Administration's (NASA) Goodard Data Language (IDL) utility with a graphical user interface Space Flight Center (GSFC) using the Interactive (GUI). The solar model uses the area of intersecting circles to simulate the ramping of sensor response while viewing the diffuser. This model is compared with preflight laboratory scans of the solar diffuser. The lunar model reads a high resolution lunar image as input. The observations of the moon are simulated with a bright target recoveu" algorithm that includes ramping and ringing functions. Tests using the lunar model indicate that the integrated radiance of the entire lunar surface provides a more stable quantity than the mean of radiances from centralized pixels. The lunar model is compared to ground-based scans by" the SeaWiFS instrument of a full moon in December 1992. Quality assurance and trend analyses routines for calibration and for telemetry data are also discussed. 1. INTRODUCTION the data related to this task will be ingested and analyzed as images. IDL also provides a graphical user interface This document describes analytical prototypes for post- (GUI) allowing easy manipulation of models. launch lunar and solar diffuser observations, plus quality control and trend analysis routines for data related to in- 2.00NBOARD CALIBRATIONS strument calibrations and engineering. The lunar and solar diffuser observations do not provide an absolute calibration Lunar and solar on-orbit measurements are intended traceable to National Institute of Standards and Technol- for the monitoring of changes in the radiometric sensiogy (NIST) standards and are correctly viewed only as tivity of the SeaWiFS instrument. These measurements stability checks. However. they are closely coupled to an will complement interchannel gain and time delay integraunderstanding of the temporal changes in instrument cal- tion (TDI) checks, which also track changes in the sensor. ibration relative to the sun. and thus the terms "'calibra- For solar measurements, the instrument has two diffuser tion" and "'observations" are nearly synonymous. surfaces--the solar flight diffuser and a nearly identical All the software routines were developed on the Sea- surface on the back of the solar flight diffuser cover. The viewing Wide Field-of-view Sensor (SeaWiFS) Calibration/ sequence of measurements will include the monitoring of Validation (CALVAL) SGI 440 wo:kstation. It is expected changes in the reflectance of the diffuser cover and the exthat these modeling tasks will add to the understanding of posure of the flight diffuser with the removal of the cover. the calibration operations as well as to the understanding Early in the mission, there will be more calibrations to of potential problems. Many of the routines developed for provide an accurate baseline for further assessment. this task may either be directly ported to the SeaWiFS Many of the onboard calibration related activities will operational environment or may provide the groundwork occur while the platform is over the Scuth Pole or in backfor future software development. orbit, minimizing interference with ocean data collection. The observation models and simulated input data are Lunar measurements will involve scanning the moon at 1designed to be as realistic as possible. Solar irradiances are 3 month intervals when the moon is within 7 ° of full. The integrated to the SeaWiFS bands and are read as input for solar diffuser will be scanned at shorter intervals between both the diffuser and lunar models. A high resolution lunar lunar measurements to provide complete temporal coverimage is also read as input in the lunar model. Labora- age. tory simulations of the diffuser observations are compared It is expected that the diffuser will degrade with time to the solar model, and ground-based lunar scans by the as contaminants condense on the surface. In contrast, the SeaWiFS instrument are compared to the lunar model. lunar surface is assumed to maintain stable reflective prop- Most of the software routines discussed in this docu- erties. Lunar observations will therefore provide anchor ment were implemented using the Interactive Data Lan- points for calculating trends in the calibrations. The caliguage {IDL) software package. This utility is an inter- bration and instrument telemetry data will undergo qt, alpreted/compiled language which is designed for statistical it)" control and time series analyses prior to being stored in analysis and display of large multidimensional arrays or the CALVAL archiw. These analyses will include screenimages. IDL is used for the calibration task since much of ing for outliers and discerning trends in the measurements.

Original page 5

I II ilIIIliI I " I ill" I IIll| I' II i r II i i )] I Modeling of the SeaWiFS Solar and Lunar Observations For a complete description of the calibration activities McClain et al. (1992). 2.1 Calibration Background Lunar measurements are the single source of infocmation for monitoring the long-term drift in the radiometric sensitivity of the SeaWiFS instrument. The sun and the surface of the moon are considered as stable, non-changing light sources. Variations can be accounted for in the incident solar flux due to changes in the Earth-sun distance by using (Gordon et al. 1983): f0=o [ l+ecos 2rr(D- 3)] 2 see calibration measurements. This is primarily the endpoint for the time series of changes in the radiometric sensitivity of the sensor. The drafting of this schedule at a point approximately six months before the launch of the instrument reduces the schedule to a "strawman." since on-orbit factors--primarily the rate of change of the reflectance of the diffuser cover--may force a modification of the schedule of e_ents. The following sequence provides a SChedule of the calibration initialization phase of instrument operations. 1. With the solar diffuser cover on: • Take a solar diffuser measurement at the first. safe opportunity. (1) • Take an interchannel measurement immediately where D is the sequential day of the year. e is orbit eccen- thereafter. (This can be done at any point on tricity (0.016). and F0 is the mean solar irradiance. similar manner, variations in the reflectance of the moon. due to small differences in the lunar phase angle and due to small changes resulting from lunar libration, can also be removed by" calculation. With these corrections, changes in measurements of the lunar reflectance can be used directly to detect changes in the sensitivity of the instrument. For solar measurements with the diffuser, it is not possible to separate changes in instrument sensitivity from changes in the reflectance of the diffuser plate in an a primeasure- measurements as a baseline for instrument opori fashion. All that can be derived from these ments is the product of the change in the instrument assump- large changes in the reflectance of the solar difthe change in the diffuser. However. there is an together. fuser cover. (Large-scale changes in the diffuser tion that can tie diffuser and lunar measurements is as- measurements will show degradation of the dif- Basically. the change in the reflectance of the diffuser month. fuser surface. Large-scale changes in instrument sumed to be essentially linear over periods of one several sensitivity during initial operation on orbit will or so. Over longer periods, e.g., periods of one to years, the change may turn out to be exponential, In a orbit. ) • If possible, take a TDI measurement at the next available safe opportunity on the orbit following the diffuser measurement. Operationally. the TDI check will be about equal to a second solar diffuser measurement (see the section on TDI checks, below). • Repeat this sequence for eyeD" downlink (twice per day') for one week. This gives 14 sets of eration. This also gives a data set to check for and with be shown in measurements of the oceans.) gradually decreasing changes over time. However. this ex- 2. At the first full moon: ponential change can be treated as a series of many linear segments. Experience with diffusers on previous satellite instruments has led to the assumption that diffuser degradation has been caused by" the coating of the diffuser with solarized organic materials that have outgassed the spacecraft. This accumulation process does not cause sharp step changes in the diffuser's reflectance. Using nearly simultaneous lunar and solar diffuser sensitiv- portunity for a second measurement.) surements, it is possible to separate changes in the ity of the instrument (from the lunar measurements) • Take a lunar measurement at a lunar phase angle of about 7 ° , going into the full moon. • If lunar phase angles and instrument power permit, take a second lunar measurement at the same from lunar phase angle coming out of the full moon. (Power restrictions may limit the measurements mea- to one per full moon, without regard to the opfrom • Repeat this sequence monthly for a minimum of changes in the reflectance of the diffuser. The time se- 3-4 months. From these data, a minimum pracries of diffuser values will be normalized by lunar measure- tical interval (in months) for lunar measurements ments. It is then possible to use the assumption of a linear for the remainder of the mission can be deterchanges in mined . change in diffuser reflectivity to identify step instrument sensitivity between lunar measurements. changes **'ere found in measurements with the Coastal Color Scanner (CZCS), SeaWiFS" predecessor. The initialization phase of on-orbit calibration operainstru- • Take a solar diffuser measurement on the next ortions, approximately the first two months after the Step 3. When the solar diffuser cover is removed: Zone • Take a solar diffuser measurement, using the diffuser cover, on one orbit. ment is turned on, has been designed to provide a statis- bit. Remove the diffuser cover during the middle of SeaWiFS of this measurement. tically significant endpoint for the time series 2

Original page 6

Ill I II I I I II R. Woodward. R. Barnes, C. McClain, W. Esaias. W. Barnes. and A. Mecherikunnel • Using the flight diffuser, take a solar diffuser mea- measurements of its diffuser. However, corrections to the surement on the next orbit, followed by an inter- design of the NOAA-11 SBUV/2 have allowed long-term channel gain check and a TDI check. measurements of these diffuser changes. A similar, but • Make a lunar measurement at the next available more complicated scheme for monitoring diffuser changes opportunity. It is anticipated that the solar dif- has been proposed for the Moderate Resolution Imaging fuser cover will be removed one day before a lunar Spectrometer (MODIS) instrument on the Earth Observmeasurement. ing Satellite (EOS) AM platform. • Take a solar diffuser measurement once per down- Star platform such that it faces opposite to the usual velink (t_'ice per day) for one week to monitor any locity vector. Sunlight will reflect off the plate for about changes in the freshly exposed flight diffuser. 80 seconds as the platform pitches 5° while passing over 4. For routine operations: the South Pole. For the duration of the calibration, the The SeaWiFS diffuser plate will be aiftxed to the Sea- • With the sequence of one eve_ downlink, take instrument assembly will be tilted 20 ° aft and the sensor solar diffuser measurements. TDI checks, and in- will scan the diffuser plate for about 10-20 pixels along the terchannel gain checks. This sequence can be back-scan (Fig. 1). changed to once per week. if the stability of the The solar diffuser response can be characterized by the instrument permits. bidirectional reflectance distribution function (BRDF) of • Take lunar measurements monthly. This can be the diffuser assembly. This function ranges above and belog" the nominal overall reflectance of the diffuser and is changed to every other month or quarterly, if the stability of the instrument permits. used to describe the angular dependent reflective response of the diffuser plate to incident light. The SeaWiFS dif- 2.2 Solar Diffuser Calibration fuser BRDF can be translated to a three-dimensional table representing the spacecraft pitch, spacecraft yaw, and the Solar diffusers, such as the one designed for SeaWiFS. instrument scan angle. It is assumed that the solar flux at have been used several times on previous remote sensing the input of the diffuser assembly remains constant. The satellite instruments. The series of diffuser measurements BRDF of the diffuser assembly will probably change with with the Solar Backscatter Ultraviolet (SBUV) spectrome- time as contaminants condense on the reflecting surface. ter on the NIMBUS-7 satellite, and with the SBUV/2 spec- To account for this change, lunar measurements are r_ trometers on the National Oceanic and Atmospheric Ad- quired. ministration's NOAA-9 and NOAA-11 satellites, a history The SeaWiFS instrument has eight bands with each of the development of this technique is provided. Changes band having four detectors. Each of the 32 detectors has in the reflectance of the diffuser on the NIMBUS-7 SBUV its own photodiode, current-to-voltage converter, gain sewere calculated using an on-orbit experiment, based on lection, and analog-to-digital converter. For each detector, changes of the exposure rate of the diffuser for extended there will be an electronic calibration pulse at the outperiods on orbit. This experiment gave an empirical model put of the current-to-voltage converter. The pulse will be that separated changes in the SBUV spectrometer from recorded for about 100 pixels as part of the data stream changes in the instrument's diffuser (Cebula et al. 1988). following the diffuser plate portion of the scan. The elec- After several years of apparently successful observations tronic calibration pulse will have an amplitude near the by the SBUV. it became apparent that most of the long- maximum number of counts (1.024 counts for the 10-bit term changes in stratospheric ozone indicated by the in- SeaWiFS data). The pulse provides a measure of interstrument were not geophysical, but were instead a calibra- channel gains and eliminates the need for scanning Earth tion drift within the instrument (World Meteorological Or- targets for gain measurements. The calibration voltage ganization. 1990). The drift was determined to result from for the pulse will be summed with the output from the the empirical model for the diffuser reflectance. It was also detector's current-to-voltage converter. Since the pulse is found that alternate models for diffuser plate degradation applied while the instrument is viewing the black, inner could be applied to the results of the on-orbit experiment surface of the instrument housing, the output from the (Herman et al. 1990). with different models giving different detector should be zero. However, small changes in the long-term changes in stratospheric ozone. detector output can be removed by referencing the out- The SBUV/2 instruments on NOAA-9 and NOAA-11 put from the calibration pulse to "zero" values from the were equipped with a lamp that could be viewed either di- readings that immediately precede and follow the pulse. rectly by the spectrometer or indirectly using the surface It is possible to control combinations of the detectors of the diffuser assembly (Frederick et al. 1986). By taking used in each SeaWiFS band, i.e., from a single detector. a series of alternating measurements of these two sources. to pairs of detectors, to a combination of all four. For it is possible to obtain a ratio that gives the reflectance test purposes on orbit, each detector for individual bands of the diffuser. Problems with the design of the lamp on can be viewed independently. Table 1 gives the five pos- NOAA-9 prohibited the collection of long-term reflectance sible TDI configurations. During the initial gain checks. 3

Original page 7

IIIII III I IIII i II I II III I ii II I lllllll I I I II I II Modeling of the SeaWiFS Solar and Lunar Observations . + 175 ° Solar Data End I +222 ° Dark DC Restore End I | -95 ° Te sco Aperture Fully Closed (West) .85 ° Edge of Earth Aperture °° °°.°° -58.29 ° Earth Data Start Aperture 0° Nadir West Fig. 1. Scanning angles of SeaWiFS the stability of the calibration voltage can be checked by cycling through the gain sequence (gt) shown in Table 2. The set of g: values before and after each other gain gives a check of the constancy of the calibration voltage over the measurement period. For this type of check of relative values. a constant input value (or a well-characterized, slowly changing value) is necessary. Knowledge of the absolute value of the input is not required. Table 1. TDI configurations with the corresponding detector arrangements. Configuration Detector Arrangement Detl + Det2 + Det3 + Det4 Detl summed with itself 4 times Det2 summed with itaelf 4 times Det3 summed with itself 4 times Det4 summed with itself 4 times .k.LJliBi +140 o Dark DC Restore Start Calibration Pulse +107 ° Solar ApertureObscured +99" Edge of Solar Aperture Solar V'mw +85 ° Edge Of Earlh Apedure +81 ° Edge of Sotar Aperture °°° 60 ° Solar Data Start +58.29 ° Earth Data End East instrument for solar calibrations. Table 2. Gain sequences for TDI configurations. Start and stop times are in seconds. Configuration Gain Sequence Start Stop g:,g2.gl,ga,gl.g4,gl 0 7 gl,g2.g:,ga.gl,g4,gl 7 14 gt,g2,gt,g3,gl,g4,gt 14 21 gl,g2,gl,g3,g:.g4,gl 21 28 91,g2,g:,gs.gs,g4,91 28 35 gl,g2,gl,ga.gl,g4,gl 35 42 g:,ga.gl,ga,g_,g4,gl 42 49 g_.g2,g:,g3.g:.g4,g: 49 56 gl,g_,gl.ga,gl,g4,gl 56 6,3 gl,g2,g:,ga,gl.g4,gl 63 70 g:,g_,g:,g3,gl,g4.g: 70 77 The scan in the solar calibration mode will produce a data stream identical in form to a typical SeaWiFS (high resolution) Local Area Coverage (LAC) scan. Fig. 2 shows

Original page 8

R. Woodward. R. Barnes, C. McClain, W. Esaias, W. Barnes. and A. Mecherikunnel 1ooo Bandwidth Reduction .=_ 9O0 8O0 70O Counts 500 Scanning Diffuser 40O Band 5 3OO 200 I00 I " " " " I ' " " " I "" " " I .... I'''" 1''''1''''1''''1 0 IO0 200 300 400 500 nl_nd 6 Calibration Pulse Bandwidth Reduction .... I '''" I" '', I, ,,,I-,T_ 600 700 800 900 lO00 1100 1200 Pixel Fig. 2. Response of SeaWiFS channels 5 and 6 during an experimental solar calibration. the response of two of the SeaWiFS channels for an ex- WiFS bands using a full-width at half-maximum (FWHM) perimental solar calibration. The diffuser is viewed in the approximation. The model simulates a diffuser scan by sensor response for about 400 pixels from start to finish. considering the intersection area of two circles which are and the calibration pulse follows about 100 pixels down- in relative movement The circles represent the instantastream. The anomalously high readings in the first few neous field-of-view (IFOV) of the sensor and of the aperhundred pirels (before scanning the diffuser) are a result ture covering the solar diffuser plate. In the algorithm, the of possible solar reflectance contamination from the small circles are aligned such that the y locations of the circle gap between the diffuser and the telescope housings. The centers are identical, and the initial x locations of the cirportion of the scan that views the diffuser exhibits a nearly cle centers are set by" the user. The IFOV circle is then linear ramping to and from a maximum plateau. The ramp moved in a stepwise fashion in the x direction relative to occurs as the instrument scans from deep space, onto the the aperture circle and the intersection area is computed diffuser and then onto the black, inner-surface of the back for each step. The intersection area A, is defined as follows of the instrument. As can also be seen in Fig. 2. the cali- (see also Fig. 3): bration pulse produces a curved signal as it transitions to the maximum value around pixel number 700 and again as it transitions to the background value around pixel 840. This cur'¢ed response is a result of a bandwidth reduction r_(O_ sin 0_) At = rl_(01 - sin01) _- - (2) 2 ' 2 which is applied to reduce high frequency noise in the sig- where r I is the radius of circle one. r2 is radius of circle nal. two. 0l is the intersection angle of circle one. and 02 is intersection angle of circle two. 2.2.1 Diffuser Calibration Model ing At by the area of the smaller circle, which defines the The normalized intersection area is obtained by divid- The solar calibration model is designed to simulate the maximum intersection area. The normalized areas of in- LAC data stream of a solar calibration for various test scetersection for each step are translated to ramping coeffinarios. Solar irradiances from Neckel and Labs (1984) for cients, i.e.. the ratio of the sampled value to the actual selected channels are read into the model from a database value, which are then applied as multiplicative factors to on CALVAL. The irradiance data are integrated to the Sea- each pixel in succession along the scan. This intersection

Original page 9

I I - " I fill -- " II II I I i I I IIII IIIII II IIII III ........... .Modeling of the SeaWiFS Solar and Lunar Observations 2 Fig. 3. The area of intersecting circles. A,. m I0 .==,.,,-=J. × E_7 OF !_ OF IFFL 43: X COff'_ 0F .q_CR/_ I0 i t ,,, , J E]'ER FAL'T : ¢.3 FACTOR: ¢. OCf_ FRCTOR: ; c._ Drt_ FRL'rOR: !,. :it OF IF'Jr SOL¢#¢AZ_'tlIT_ _ IFFII X00 { ......... Em'E_ Fro'mR : i1.0 -7...J L Fl_tIl"Lt_ OF OtLIJI_TI_ _ (g'} 1 • i STARTOF LI_i_TIO T;.'.2.: ! .J .... 700 O(T10¢FAL'?OR: ! 1.0 i , i_NTI_F_I : 1.0 SSO 0(11_ FiCTOR : !1.0 L IOIE . i Fig. 4. The initial GUI for the solar calibration model.

Original page 10

I Ill I IIll I Illlll I I II Ill II I II I IIIIll 'll I R. V_'oodward. R. Barnes. C. McClain. W. Esaias. W. Barnes. and A. Mecherikunnel algorithm produces a symmetric pattern for a scan with imum and back for the diffuser portion of a single scan quasi-linear ramp-up and ramp-down intervals bracketing agreeing in general with the laboratory results shown in a period of fixed maximum output. A simulated calibra- Fig. 2. Also included in this figure is the simulated calition pulse may also be inserted in the modeled signal fol- bration pulse. lowing the above diffuser pattern. The simulated pulse pro- In Fig. 6. a two-dimensional representation of the calduces a stepwise jump in the signal and. therefore, does not ibration is shown with the x coordinate representing the accurately model the curved response produced by band- instrument scan and the y coordinate representing time. A width reduction in the transition to maximum and again three-dimensional representation of the same output is proto background (see Fig. 2). vided in Fig. 7. Fig. 8 shows the same calibration, except Fig. 4 shows the solar calibration model GUI. These with a random BRDF field applied as a three-dimensional input values may be adjusted to simulate various calibra- lookup table. The three dimensions on the diffuser reption scenarios for selected SeaWiFS channels (CltANNEL). resent the incident solar zenith angle, the incident solar For instance, the lengths and amplitudes of both the sen- azimuth angle, and the pixel number (viewing angle of sor response to the diffuser and the calibration pulse can the sensor). The prelaunch BRDF measurements from the be adjusted by modi_ing the input values on the GUI. The Hughes/Santa Barbara Research Center (SBRC) will be center of the intersecting circles (X CENTER OF IFOV and substituted as input to the solar model when testing is X CENTER OF APERTURE). the sizes of the circles (RADIUS completed. The addition of this information to the model OF IFOV and RADIUS OF APERTU_). and the scan move- will prove useful in tracking post-launch diffuser and sensor ment (INCREMENT OF IFOV) determine the lengths of the performance. ramp-up, maximum, and ramp-down intervals for the simulated diffuser measurement. ENTER FACTOR modifies the 2.3 Lunar Calibration ramping coefficients by applying nmltiplicative factors to the coefficients determined by the above circle intersecto monitor long-term changes in the instrument's radiotion algorithm. In addition, the value entered in ALBED0 metric sensitivity. Rather. for this purpose, the instru- 0F DIFFUSER is applied as a multiplicative factor and can ment will make a series of measurements of the surface of therefore be used to simulate diffuser degradation. the moon during the lifetime of the SeaWiFS mission. It is ZENITH ON DIFFUSER and AZIMUTH 0N DIFFUSER refer assumed that the lunar surface has stabilized over geologic to the angle of the incident sunlight on the diffuser and are time and that the reflectance of the surface will not change used for retrieving BRDF values from a three-dimensional over the 5-year mission of SeaWiFS. It is also assumed that lookup table. The third dimension in this table is linked the solar irradiance will be known from sources outside of to the scan angle of the sensor. This value is a function the SeaWiFS measurement. Solar diffuser measurements onh" of pixel number and is computed automatically in will be used to fill in the gaps between lunar measurements. the algorithm. The calibration pulse can be modeled uswhich cannot be made more often than once per month. ing MAGNITUDE 0F CALIBRATION PULSE Z which sets the The diffuser measurements will be normalized to the lunar Observations of the diffuser assembly will not be used maximum value for a pulse by applying a percentage of the values at the measurement times. Neckel and Labs (1984) irradiance for the given channel. START OF CALIBRATION PULSE and END OF CALIBRATION only about once a month when the lunar phase angle is PULSE assigns start and end pixel numbers for the pulse within 7 ° of full phase. When this criterion is satisfied within the scan. on selected orbits, the SeaStar platform will undergo a 360 ° tilt maneuver on the back-orbit allowing the sensor to 2.2.2 Diffuser Calibration Model Results produce a standard LAC data stream of the lunar surface. To maintain consistency, lunar calibrations can occur Figs. 5-7 show three representations of the modeled Given a nominal Sea.Star pitch rate of about 0.15 ° s- t. it is instrument response of a solar calibration for the criteria expected that the moon will encompass about 7 SeaWiFS listed in Fig. 4 and a fiat BRDF field, i.e.. all BILDF val- pixels at the lunar equator and about 20 SeaWiFS scan ues set to unity. In this example. X CF__ITER OF IF0V. X lines from lunar pole-to-lunar pole. CENTER OF APERTURE. and INCREMENT OF IF0V are chosen in such a manner as to produce about a 375 pixel startto-finish response of the diffuser induced signal. The sim- 2.3.1 Lunar Calibration Model Description The lunar calibration model simulates the sensor reulated calibration pulse is modeled to be 150 pixels wide sponse of a SeaWiFS lunar viewing for vaD'ing instrument with a maximum amplitude equal to the saturation radi- and spacecraft conditions. Table 3 provides a synopsis of ance value. The total calibration is simulated to span 82.4 the user interfaces for the model. The means and stanseconds which corresponds to 494 SeaWiFS scans. Addi- dard deviations for both the integrated scanned moon and tional ramping to phase in and out of the lunar sweep is the scanned central nine pixels are written to a specified also modeled using intersecting circles producing a total of output file which is used for subsequent statistical analy- 592 scans. Fig..5 shows a nearly linear ramping to max- sis. An output display is also produced, consisting of the 7

Original page 11

......... I I IIII I I IIIII ! I II I I I I I III I I I Modeling of the SeaWiFS Solar and Lunar Observations 15 OI =E 10 D & E Z < 5 0 500 _ 1000 500 PIXEL Fig. 5. A single scan of the solar diffuser measurement (including a calibration pulse) produced by the solar calibration model. Fig. 6. A two-dimensional representation of the solar diffuspr measurement (including a calibration pulse) produced by the solar calibration model.

Original page 12

Illl Ill I II I I II I II I I I I I II iJlJ ii I II I IlI I I R. Woodward. R. Barnes. C. McClain. W. Esaias. W. Barnes. and A..Mecherikunnel Fig. 7. A three-dimensional representation of the solar diffuser measurement {including a calibration pulse) produced by the solar calibration model. i' Fig. 8. The same as Fig. 7 except with a random BRDF field read as input for the solar calibration model.

Original page 13

• • [ • IIII I I I Ul • I II I I l m, .. m |, . .... Modeling of the SeaWiFS Solar and Lunar Observations original lunar image and the output following each model has a damped periodic function specified in EvrEx PERIOD step listed in Table 3 (scanned image, scanned image with and AMPLITUDE. For example, the following events occur if bright-to-dark response, and scanned image with bright- the values shown in Fig. 10 are chosen: to-dark and dark-to-bright response). Table 3. User interfaces for the lunar calibration model. Interface Function Inputs SC_'M00N Simulates the • Image size, image of the • Scan resolution. moon recorded • Scan position, for a lunar • Channel. calibration. • Sensor degradation. • Gain. and • Output filename. DRK2BRT Simulates the • Ringing and dark-to-bright • Ramp-up factors. response of the sensor. BRT2DRE Simulates the • Ringing and bright-to-dark • Ramp-down factors. response of the sensor. Fig. 9 shows the initial GUI of the lunar calibration 1. If the difference in successive pixels exceeds the value of THRESHOLD, ramping occurs over the subsequent five pixels. These pixels are factored by" 0.3. 0.85. 0.95. 0.98. and 1.00 (ENTER FACTOR). respectively, simulating the response to a bright target. If T_ESH012 is not exceeded, then the five pixels are factored by the quantity: FACTOR/(AP/TttRESHOLD), where :XP is the difference in successive pixels. If this value exceeds 1.00 it is set equal to 1.00. 2. If the difference in successive pixels exceeds the value of THRESHOLD. a de!aved ringing response is triggered. The ringing commences when successive downstream pixels remain equal or decrease in magnitude. The ringing occurs for 6 pixels (N PERIOD) employing a periodic wave of 8r, rad. (ENTER PERIOD) with wave amplitude specified as 1 mWcm -2 _rn -1 sr -1 (AMPLITUDE). The wave function is characterized by a damped sine wave using a sin(a')/z function, where _r is the pixel number within the scan line. The user can adjust the frequency, amplitude, and dumodel. The size of the input image is specified with X ration of the ringing by altering the input values for the IMAGE SIZE and g IMAGE SIZE. respectively. The scan lunar model. Fig. 12 shows the modeled ringing of a single pixel size and scan line sampling are adjustable by" setting scan for four selected scenarios. High amplitude and low SCANSIZE and SCANSTEP. respectively. Initial pixel and amplitude ringing are shown in panels a) and b). respecline position are determined by START X SCAN and START tively': variations in the ringing periodicity are shown in V SCAN. The simulated SeaWiFS pixels are constructed by panels c) and d). computing the mean of all input image pixels within a square of size SCANSIZE x SCANSIZE. The corner of each square is determined by pointers incremented in the x di- 2.3.2 Lunar Calibration Model Results A high resolution {16 bit) lunar image at 3 ° phase rection by SCANSIZE and in the y direction by SCANSTEP. (Kieffer. pers. comm.) was read as input for the lunar cal- This separation of x and y pointer movement allows over- ibration model. Fig. 13 shows an example of the display sampling or subsampling in the y direction, thus permit- output obtained from the model. The upper left panel ting the modeling of platform rotation rates. shows the original Kieffer moon. the lower left shows the Variance in pLxel registration can be tested using START scanned moon. the upper right shows the scanned moon X SCAN and START Y SCaN to initialize the pointer to a with the dark-to-bright response applied, and the lower pixel and line on the input image. Loss of instrument right shows the scanned moon with both the dark-to-bright sensitivity can be simulated using SENSOR DEGRADATION. and the bright-to-dark responses applied. In this example, which applies the specified percent decrease in range to the model input values were adjusted such that the scanned the input image. Selecting CHANNEL initiates a query to moon was 7 pixels wide and 22 scan lines high. The total the CALVAL database to obtain the wavelength depen- integrated mean of the moon and the mean of the central dent solar irradiance from Neckel and Labs (1984), A gain nine pixels are also displayed for each model step. factor can be input to the model by entering a value in the A test was performed using the lunar calibration model ENTER GAIN prompt. to determine the effects on the calibration of variable pixel The user interfaces shown in Figs. lO and 11 provide registration of the lunar surface. A total of 15 runs were inputs to routines for simulating the dark-to-bright and produced in which z pixel positions (along scan lines) were bright-to-dark target responses of the instrument. The moved at one-fourth SeaWiFS pixel increments while holdramp-up and ramp-down factors are specified b.v the ENTER ing the y pixel position (scan lines) constant: and y pixel FACTOR prompts. A ringing response is triggered when the positions were then moved at one-twelfth pixel increments absolute value of the difference in _uccessive values exceeds while holding the z pixel position constant. For this series THRF._HOLD,This ringing will last for N PERIOD pixels and of runs. an extreme case of ringing was applied. The initial 10

Original page 14

II I I I II II II II I I I I I I II I I R. Woodward, R. Barnes, C. McClain. W. Esaias. W Barnes. and A. Mecherikunnel 30 I STARTX S3_N a : _ DE_TIO_ ENT'_ IN : 2.0 E TITLE : ENT_ 8UT_ FILE : st_ts 85.1. I 1,j Fig. 9. Initial GUI for the lunar calibration model. 11

Original page 15

I I I I II I I li II l I I I i I - - II I II Illl IIIIIIII III II |11 I II Modeling of the SeaVs'iFS Sc,lar and Lunar Observations S • II IBm I I I II I I E]TERFETOR : 0.3 Eh'rE]RFET[)R : O.8S ENTERF: 0._ Eh_ FI_,OR : 1.C ENTER FETOR : 0.0 ENTERF_KT.,OR: 0.0 BITER FRCTDR : c.o E]f FI_'TOR: , 0.0 FACTDR : 0.0 ,, J I I II -"T I " ,, , IT Fig. 10. GU[ for the dark-to-bright instrument response in the solar calibration model. 12

Original page 16

R. Woodward. R. Barnes, C. McClain. W. Esaiaz. W. Barnes. and A. Mecherikunnel ,,J ._ ! Ek.----_ ._'R-I0: : , it, t,0, d ; : ' r i! _ FRCTOR: !0o7 ENTERFRCTOR: ',0.15 FACTOR: 0.05 T ENTERFRCTDR: , 0.0 i ENTERF_CTOR: io.o I ENTERFRCIOR: I 0.0 ENTERFRCTDR: i0.0 II I I "11"1 IIII II II I I I I "1 i'" n -I'T - Fig. 11. GUI for the bright-to-dark instrument response in the solar calibration model. 13

Original page 17

, ,,,, II 11 i , • '":'" " ' : III II II IIIIII I " il li III I Ill 'II ! Modeling of the SeaWiFS Solar aud Lunar Observations a) 20 ......... , ......... , ......... , ......... , ......... 15 "g" / ' E 04 10- 0 v E I q) (J ¢- 0 I "0 0 no___/ -1 0 10 20 30 a0 50 O;xel Number b) 15 ......... I ' ' ' ...... I ' ' ' ' ..... l ......... l ..... ' ' ' ' q- 10 E ("4 O 5 E I ;/ O £- O "O O ne 0 --5 ......... I ......... 1 ......... 1 , , , , i .... I ......... 0 I 0 20 30 40 50 Pixel Number Fig. 12. Ringing from the lunar calibration model for a) high amplitude ringing, b) low amplitude ringing. Note: the dashed line is the target and the solid line is the response. 14

Original page 18

I I I I II I I I I I i i i - R. Woodward. R. Barnes, C. McClain, W. Esaias, W. Barnes, and A. Mecherikunnel c) 15 • ' ' ' ' ' " ' ' I ' ' ' ' ' " ' ' ' | ' ' ' ' ' ' ' ' ' | ' ' ' ' ' ' ' ' ' J ' ' ' ' ' ' ' ' ' I I I I "L" 10 I ¢) I E I I l i u I I 5 II E I I I C o O rw 0 --5 ......... I ......... I ......... I , , , , , , , , , I , ,, , , .... . 0 10 20 30 40 50 Pixel Number d) 15 '' .... '''l ......... I'''' .... '1'''''''''|'''''''''-- I0 I/) I E l l I ]= 5 E I i/ L) ,I u t- O 10 _A 0 0 0 I 0 20 30 40 50 Pl)tel Number Fig. 12. (cont.) Ringing from the lunar calibration model for c) high frequency ringing, and d) low frequency ringing. 15

Original page 19

r 2. Modeling of the SeaWiFS Solar and Lunar Observations X,=eV=WI 1 II 1 II i '1 :l ,! 1, ! i { ,. ro_,=,:,,,,t.s:3.=s,,._ l1 3°1 4°717_ BRI{}{T TO _ 9: t _ I Lii I Fig. 13. Display output from the lunar calibration model 18

Original page 20

II I Ill ..... II I IIIII I I I | I III II R. Woodward. R. Barnes. C. McClain, W. Esaias, W, Barnes, and A. Mecherikunnel 6 I ' ' 5 4 U _ 3 E I u - 2 0 o O 0 i , , _ I- _ _ 0 5 Model Run ! ' ' ' _ L i , , _ I0 15 Fig. 14. Stability of the lunar mea.surements as a function of pixel registration for the integrated lunar mean (solid line) and the central 9-plxel mean (dashed , , I s s \ I \ I / I "C" U v I U 2 ID 0 line). 1 • • , • • i • A | • A i A i I i i . | • • • L • • , • • 0 10 20 30 Number of Scon Lines Fig. 15. Stability of the lunar measurements as a function of the number of scan lines for the integrated lunar mean (solid line) and the central 9-pixei mean (dashed line). 17

Original page 21

I II II II II I II I _ " I I I II II II] I Ill /111 I ill III III II I II Modeling of the SeaWiFS Solar and Lunar Observations amplitude of the ringing was set to 1 mW cm -2 #m-1 sr-1 and the ringing period was set to 8r, rad. The results were tabulated for each step of the model for both the integrated be changes in the side-to-side characteristics of the halfangle mirror during the course of the mission. The change in the average for the two mirror sides cannot be deterimage mean and the central 9-pixel (3 x 3) mean (Table 4). mined independently on orbit and must be part of the Standard deviations (a) of the model output over the runs were computed for each model step as a measure of stability. Fig. 14 shows the retrieved mean radiances for both the mean integrated image and the mean central 9 pixels as a function of pixel registration. These results indicate the retrieved radiances from the lunar calibration are more long-term sensitivity of the instrument. The average is an inherent part in the 20 scans that cover the surface of the moon and in the 480 scan lines of a diffuser measurement. The magnitude of the side-to-side differences in the half-angle mirror, however, can be tracked during standard measurements of the instrument's diffuser. Durstable, as indicated by" the standard deviations, for the in- ing each diffuser measurement, there are 240 sidel-side2 tegrated mean than the central 9-pixel mean. The relative instability of the central 9-pixel mean among calibrations is most iike]y a result of variable lunar reflectance as the 3 × 3 sampling box changes position with regard to the lunar surface. pairs of half-angle differences to be used for calculations. These differences transform into scan line-to-scan line differences in the ocean measurements. The magnitude of the scan-to-scan differences will be tracked in the onboard calibration information although, strictly speaking, these Another test, using the lunar calibration model, was differences are not part of the radiometric calibration for conducted to determine the effects of varying the number of scan lines across the moon. This test illustrates the impact of variable platform rotation rates upon lunar calibrations. The results are tabulated in Table 5 and displayed in Fig. 15. The integrated mean is again found to be more stable than the central 9-pixel mean in tracking lunar reflectance. These modeling studies can be compared to a groundbased sweep of the moon bv the SeaWiFS instrument at Goleta. California on December 9. 1992. On this date. the moon was sampled shortly after a lunar eclipse giving a very small lunar phase angle. Fig. 16 shows the Kieffer moon. the modeled lunar scan. and four observed sweeps of the moon. The observed lunar scans were not corrected for atmospheric interference, which possibly accounts for the discrepancies between the observations and the modeled scan. The scan at 1922 Pacific Standard Time (PST) was scanned from the top to the bottom producing an upsidedown view in comparison to the other images in the figure. An important consideration for the solar and lunar observation data is the potential for ghost images in the optical path of the instrument. These artifacts occur around bright targets and are a result of reflections off of the polarization scrambler, which is the third component in the SeaWiFS optical path (Fig. 17). The light reflected off the ei- moon and sun-moon distances. In addition to the varipolarization scrambler exhibits two artificial images on ther side of pixels in the direction perpendicular to the scan (Holmes, pets. comm.) as shown in Fig. 18. These artifacts contain about 3_ of the signal and are confined to about 4 pixels of the target edge. In Fig. 16, the observed lunar scans exhibit little or no evidence of ghost images. Fig. 19 shows a three-dimensional representation of the lunar sweep from opposite perspectives at 1922 PST for channel 6. In this figure, the natural logarithm of Sea- WiFS counts ha' been plotted as the vertical coordinate. There appears to be some elevated values in the direction perpendicular to the scan from pixeis 9-15. the instrument. 2.3.3 Additional Considerations Several factors are likely" to impact the measured lunar reflectance with the SeaWiFS instrument. One of these factors is the eccentricity of the lunar orbit. The lunar orbit places the moon 3,56,410 km from the Earth at perigee. and 406,697 km at apogee. Over the course of a lunar orbit. the Earth-moon distance (,EM) can be expressed a (Duffett-Smith 1979): a(1 - e ") EM = 1 -- ecoslM" - 6.289sin(M')] (3) where a is the semi major axis. e is the eccentricity-, and AI is the corrected mean anomaly, which is a function of date. and refers to an imaginary moon in a circular orbit. Assuming an inverse square distance relationship for the obselWed lunar reflectance, the amount of lunar radiance measured by SeaWiFS can vary by"as much as 30_ over allpossiblefullmoon orbitalpositions.Observed lunar radiance may" alsobe dependent, to a lesserextent,on the moon-sun distance. Lunar observationswilltherefore require normalization to account for the variableEarthableradiance,pixelregistrationwillvary as a resultof the change in distancebetween the sensor and the moon. This latterconsiderationmay be allayedby using the integrated lunar surface for the SeaWiFS observations,as discussed in Section 2.3.2. Additional functionalitywill be added to the lunar calibrationmodel to simulate the potential effectsof an ellipticalorbit. Libration isanother possiblesource of uncertainty for the SeaWiFS lunar observations. The major motions of the Earth-moon system allow observers on Earth. over a period of time. to view 59_ of the lunar surface,even Side-to-side differences in the half-angle mirror (Fig. 17) though the sane side of the moon is always facing the will be characterized at SBRC before launch. There may 18 Earth. Longitudinal librationoccurs because the moon

Original page 22

',r ' "" _ z '1 _l " I II I I I " '"" f _ ........ " < ..... R. Woodward. R. Barnes, C. McClain, W. Esaias, W. Barnes, and A. Mecherikunnel Table 4. Lunar observation model results for testing calibration stability as a function of variable pixel registration. All units are in mW cm -2 pm-l sr-l. As a measure of stability, standard deviations (a) of the model output over the runs are reported. x Pixel y Pixel Integrated Offset Offset Scan Dark-Bright 0 0 3.71 2.83 1/4 0 3.76 2.86 2/4 0 3.81 2.86 3/4 0 3.76 2.74 0 1/12 3.76 2.89 0 2/12 3.76 2.87 0 3/12 3.71 2.84 0 4/'12 3.71 2.81 0 5/12 3.76 2.88 0 6/12 3.76 2.87 0 7/12 3.72 2.84 0 8/'12 3.76 ".86 0 9,/12 3.71 2.83 0 10/12 3.69 2.90 0 11/12 3.76 2.89 Column a 0.03 0.04 Central Nine Bright-Dark Scan Dark-Bright Bright-Dark 2.33 5.02 5.19 4.17 2.31 5.16 5.25 3.80 2.30 5.01 4.45 5.26 2.26 4.88 5.21 4.71 2.37 5.00 5.24 4.14 2.24 4.98 4.96 4.23 2.32 4.45 4.89 4.15 2.32 5.03 5.22 4.18 2.37 5.01 5.15 4.15 2.36 4.98 4.99 4.24 2.33 4.96 4.97 4.17 2.32 4.93 4.84 4.18 2.33 5.01 5.17 4.16 2.36 4.99 5.00 4.25 2.36 4.96 4.92 2.21 0.04 0.06 I 0.14 0.33 Table 5. Lunar observation model results for testing calibration stability as a function of a variable number of scan lines. All units are in mlVcm -2 pm -1 sr -l. As a measure over the runs are reported. Number of Integrated of stability, standard deviations (a) of the model output Central :'ine Scan Lines Scan Dark-Bright Bright-Dark Dark-Bright Bright-Dark Scan 1 3.60 3.12 3.22 4.36 4.52 4.52 2 4.24 3.47 3.29 4.96 5.34 5.34 4 4.18 3.21 3.16 5.31 5.42 5.51 6 3.60 2.93 2.86 5.43 5.71 5.75 8 3.81 3.05 3.02 4.90 5.05 5.08 10 3.64 3.05 3.02 4.96 4.96 5.18 12 3.64 3.04 3.02 4.10 5.26 5.27 14 3.82 3.14 3.09 4.91 5.35 5.37 16 3.73 3.08 3.06 4.98 5.08 5.10 18 3.71 3.12 3.05 5.01 5.22 5.26 2O 3.70 3.10 3.03 5.02 5.26 5.30 22 3.75 3.13 3.06 4.96 5.08 5.13 24 3.72 3.12 3.06 4.98 5.12 5.16 26 3.72 3.12 3.O8 5.01 5.22 5.30 28 3.77 3.14 3.08 4.98 5.03 5.05 3O 3.70 3.05 3.01 4.97 5.01 5.04 Column a 0.18 0.11 0.10 0.22 0.26 0.26 19

Original page 23

II 11111 II I II - III II IIII I t lill Ili I I1 I III I I IIII Ill I I t ntmt. :i -Iodeling of the SeaWiFS Solar and Lunar Observations I I Lmm'Sam Fig. 16. Comparison of modeled lunar scan to ground-based measurements 20

Original page 24

I I t t I I I II I I I t tl R. Woodward. R. Barnes. C. McClain. W. Esaias. W. Barnes. and A. Mecherikunnel Primary Mirror Half Angle I _'2" MirrJr Telescope Motor \ ul Solar Calibration I Polarization Scrambler I Collimating Mirror Bands 3 & 4 Fold Bands 1 & 2 Mirror Bands 7 & 8 Bands 5 & 6 Fig. 17. A schematic of the SeaWiFS optical layout. 1 10-1 10-2 10-3 lO-5 X Fig. 18. Ghost images induced by the polarization scrambler. 21

Original page 25

I I II I III I I I I II II I _ I II I Modeling of the SeaWiFS Solar and Lunar Observations _5 _4 b) Fig. 19. A three-dimensional representation of a) the lunar sweep at 1922 PST for channel 6 and b) the same scene from the opposite perspective ( 180 ° difference). 22

Original page 26

I II I I I II I I I I Iili I IIII I I I I . II • R. Woodward. R. Barnes, C. McClain. W. Esaias. W. Barnes. and A. Mecherikunnel O_t_Ul_ e01t'11110¢ - DISPLAY HEJM v" DISPLAY ,_ DISPLAY I_DIN4 ,r DISR.AY $TN_N_D DEVIATION PlOT HIST DISPL.AY N_IHUI'I I:iHDNINIHLI4 v OEI_ USER SIq_I:IFIVn RN4[;_ 25O ! 2OO 0 | 0 2 EXIT ! 4 6 Radiance /mW/(crrv,2 uM st) Fig. 20. GUI for the calibration data and instrument telemetry quality control routines. 23

Original page 27

III I III I 21 I I I I I I I| " - I Ii I II Modeling of the SeaWiFS Solar and Lunar Obser_"ations STATIS'II_ TI_ /" PLOT VALUEI_ FUNCTIONOF TIME v PLOT STAI_AI_ De_IATIOHAS FUH_I_ TIE v I_ - I_llmlCI'P/ v TION _ _ I_IALYSES ,.,-POLYN3HIALFIT _ TRE REMOVAl. v" REGI_SSIONI_lqAl.c_s v CI_SS--133RRELRTIONS AI_Y - EXIT MEAN CAL VALUE VERSUS TIME ill- ® 2.5 • * i i i • • * i | • _ 2.0 E V i.i 0.5 Z 0.0 . • i .... i .... i,i 0 5 ]0 WI_S i %i • • • • • • I i i • i 15 2O Fig. 21. GUI for the calibration data and instrument telemetry trend analysis routines. 24

Original page 28

• i il ' "' " ' "= ' I II I [ III 111 I R. Woodward R. Barnes, C. McClain, W. Esaias. rotates on its axis at a constant rate but travels around 5.0 the Earth at varying velocities due to the elliptical orbit. I III II I I I I II II I I II W. Barnes, and A. Mecherikunnel RECOMMENDATIONS Latitudinal libration is a result of the 6.7 ° tilt in the lunar Modeling results from the lunar calibration indicate the integrated mean of the entire lunar surface is a more staaxis with regards to the Earth ecliptic. The obvious effect ble measure than the mean of a central grouping of pixels. of libration is that the sensor will not always be viewing This conclusion was reached by considering possible unthe same complete face of the moon. To correct for this certainties in the pixel registration and platform rotation effect, a database of lunar observations can be populated rates. with respect to libration. cal artifacts, i.e., ghost images, in the SeaWiFS signals. The magnitude of these artifacts is small in comparison to 3.0 QUALITY CONTROL the lunar signals, but the impact on calibrations remains In addition, there may be some evidence of opti- Quality control routines will be essential for en_uring uncertain. More analysis of ghost images may be required. the integrity of the calibration data. It is anticipated that Another consideration is to use the prelaunch sensor most of this activity will be activated in an automated characterizations, when available from the SBRC. to modfashion. However. user intervention routines can be imple- iL" the solar and lunar models, The models can be made mented using interfaces such as the prototype presented in to conform to the new data and to test the transfer of Fig. 20. Statistics are generated by ,sing this interface to prelaunch data to orbit. An additional enhancement to check for outliers or unusual patterns in the data. The user the models will include the input of the actual BRDF tacan select from three types of data average: mean, mode, ble from SBRC. Following launch, the models can be used and median. The standard deviation can be generated to to evaluate the actual performance of the sensor-diffuser check for unexpected variance in the data, Outliers can be system. determined by selecting for the minimum and maximum value. The user can also choose to specify a valid range of The authors would like to thank X_atson Gregg for use of the ACKNOWLEDGMENTS data. Unusual or unexpected patterns in the data can be solar irradiance data integrated to SeaWiFS bands. checked by selecting the histogram option. The example in Fig. 20 shows a histogram applied to radiances from a simulated lunar calibration. BRDF Bidirectional Reflectance Distribution Function CALVAL Calibration/Validation 4.0 TREND ANALYSIS CZCS Coastal Zone Color Scanner Trends in both the calibration and instrument telemetry data will require analysis, since patterns in these data GLOSSARY Earth Observing Satellite EOS may display drift or periodicity. An accurate resolution FWHM Full-Width at Half-Maximum of these effects will allow corrective techniques to be applied to the data. For example, polynomial or periodic GSFC functions may be implemented to compensate for any observed trends, IFOV Instantaneous Field-Of-View Fig. 21 shows a prototype interface for trend analysis. The selections activate routines that are designed to both characterize and compensate for trends in the instrument MEM Maximum Entropy Method calibration and telemetry data. In the example from this MODIS Moderate R_olution Imaging Spectrometer figure, a simulated degradation in the lunar calibration Goddard Space Flight Center GUI Graphical User Interface IDL Interactive Data Language LAC Local Area Coverage has been plotted as a function of day. Since a calibration NASA National Aeronautics and Space Administratio, reading may be composed of the mean of several pixels. NIMBUS Not an acronym--a series of NASA experimental the user can also track the trend of calibration scatter by plotting the standard deviation as a function of time. Periodicity in the data can be analyzed by selecting NOAA National Oceanic and Atmospheric Administration the AUTOCOKRF_TION AND SPECTKAL ANALYSIS and MEM PERIODICITY options. The maximum entropy method (ME.X,I) is used to deduce periodicity when data are not SBRC (Hughes) Santa Barbara Research Center evenly spaced in time. Regressions and cross-corrt _tions can be generated to check for relationships among calibra- SBUV SeaWiFS Sea-viewing Wide Field-of-view Sensor tion or telemetry parameters. Finally. trends in the data can be removed through the use of polynomial fitting algorithms. weather satellites containing a wide variety of atmosphere, ice. and ocean sensors. NIST National Institute of Standards and Technology PST Pacific Standard Time Solar Backscatter Ultraviolet Sensor SGI Silicon Graphics. Incorporated TDI Time Delay Integration 25

Original page 29

II loll -- I III I I III I i I II I I i llil llI I II .Modeling of the SeaVv'iFS Solar and Lunar Observations SYMBOLS Duffett-Smith, P., 1979: Practical astronomy with your calcua Semi-major axis of the Earth's orbit. lator, Cambridge University Press, New York. NY, 129 pp. ,4_ The intersection area. Frederick. J.E.. R.P. Cebula, and D.F. Heath. 1986: Instru- D Sequential day of the year. e Orbit eccentricity of the Earth, F0 Solar irradiance corrected for the Earth-sun distance. Gordon. H.R., D.K. Clark. J.W. Brown, O.B. Brown. R.H. To Mean solar irradiance. M" The corrected mean orbit anomaly of the Earth. which is a function of date, and refers to an imaginary moon in a circular orbit. r_ The radius of circle one. Herman, J.R., R.D. Hudson, and G.N. Serafino. 1990: An r2 The radius of circle two. z The pixel number within a scan line. /xp The difference in successive pixels. ment cha-acterization for detection of long-term changes in stratospheric ozone: An analysis of the SBUV/2 radiometer. d. Atmos. Ocean. Technol.. 3, 472-480. Ex'ans, and W.W. Broenkow, 1983: Phytoplankton pigment concentrations in the Middle Atlantic Bight: comparison of ship determination and CZCS estimates. Appl. Opt.. 22, 20-36. analysis of the 8 year trend in ozone depletion from alternate models of SBUV instrument degradation, d. Geophys. Res.. 95, 7,403-7.416. 0] The intersection angle of circle one. McCiain. C.R., W.E. Esaias, W. Barnes, P_, Guenther, D. En- 02 The intersecuon angle of circle two. s'EM The distance between the Earth and the moon. Standard deviation of a set of data values. Neckel, H. and D. Labs. !984: The solar radiation between 3300 REFERENCES dres. S. Hooker. G..Mitchell. and R. Bar,es. 1992: Calibration and validation plan for SeaWiFS. NASA Tech. Memo. 10_566. VoL 3. S.B. Hooker and E.R. Firestone. Eds.. 41 pp. and 12500 A. Solar Phys.. 90, 205-258. ('ebula, R.P.. H. Park, and D.F. Heath. 1988: Characterizat;on World Meteorological Organization. 1990: Report of the Inof the Nimbus-7 SBUV radiometer for the long term monitoring of stratospheric ozone d. Atmos. Ocean. Technol.. 5, 215-227. 26 ternational Ozone Trends Panel. 1988: World Meteorologzcal Organ:zatzon Global Ozone Research and Monztorin 9 Pro3ect. Repo_ No. 18. 2 Vols.. Geneva. Switzerland.

Original page 30

• " III I I I I I I I I III II IIII II I III I II II I I'II ill .i i I I II II I I II Form At)proved REPORT DOCUMr ITATION PAGE OMBNo.O -O, S PulSe rel>ling burden for tins on Ol mlormat,on 4 IsI,m=IIKI tO =,vwage 1 _ _ response. ,ncJudmg the hme tot rev_wmg instructions. Searching existing aala soorols, ga,,mrlng an<:l mamt,11,nmg tl 041,1 neecs4c, and ¢Omlt31et.'l ln<l Pewewn the <:l)llCt+o Ot dormatcO Sen<: oo*ments regatchng th=S ioutOen r.,tlmate or any otl_e aspect o! thes oolLect_, o+ ,lIomlaI,, mdudarKJ {I<JIIGIN)nS for re_Jc this C.JrOen. to Wash,on I-lea0l;luartets . r_kmctorate ,'or Informat)n OI0e¢IIK_S and Relxxts. 121. c letforson [lv H_llway. Su4e 1204. Addition. VA 2-430". ano to the OII¢1, ol Idat_l_rr_nl and Bu¢_. Plo4twotlt RIOuChon PR, pcl (0704-0188i. Wash_jton. DC 20503 1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE May 1993 4. TITLE AND SUBTITLE SeaWiFS Technical Report Series 3. REPORT TYPE AND DATES COVERED Technical Memoc mdum 5. FUNDING NUMBERS Volume 10--Modeling of the SeaWiFS Solar and Lunar Observations 970.2 6. AUTHOR(S) Robert H. Woodward, Robert A. Barnes, Charles R. McClain, Wayne E. Esaias, Wihiam L. Barnes, and Ann T. Mecherikunnel Series Editors: Stanford B. Hooker and Elaile R. Firestone 7. PERFORMING ORGAI_I'ZATION NAME(S) AND ADDRESS(ES) Laboratory for Hydrospheric Processes Goddard Space Flight Center Greenbelt, Maryland 20771 8. PERFORMING ORGANIZATION REPORT NUMBER 93B00085 9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING National Aeronautics and Space Administration Washington, D.C. 20546-0001 11. SUPPLEMENTARY NOTES AGENCY REPORT NUMBER TM-104566, Vol. 10 Robert H. Woodward: General Sciences Corporatit, n, Laurel, Maryland: Robert A. Barnes: CHEMAL, Inc., Wallops Island, Virgima; Elaine R. Firestone: General Sciences Corporation, Laurel, Maryland. 12a. DISTRIBUTION/AVAILABILITY STATEMENT Unclassified-Unlimited Subject Category 48 12b. DISTRIBUTION CODE Report is available from the National Technical Information Service, U.S. Dept. of Commerce, 5285 Port Royal Road, Springfield, VA 22151: (703) 557--4650. 13. ABSTRACT (Meuomum200wordsJ Post-launch stability monitonng of the Sea-viewing Wide Field--of-view Sensor (SeaWiFS) will include periodic sweeps o! both an onboard solar diffuser plate and the moon. The diffuser views will provide short-term checks and the lunar views will monitor long-term trends in the instrument's radiometric stability. Models of the expected sensor response to these observations were created on the SeaWiFS computer at the Nauona! Aeronautics and Space Administration's (NASA) Goddard Space Flight Center (GSFC) using the interactive Data Language (IDL) utility with a graphical user interface (GUD. The solar model uses the area of intersecting circles to simulate the ramping of sensor response while viewing the diffuser. This model is compared with preflight laboratory scans of the solar diffuser. The lunar model reads a high-resolution lunar image as input. The observations of the moon are simulated with a bright target recovery algorithm that includes ramping and ringing functions. Tests using the lunar model indicate that the integrated radiance of the entire lunar surface provides a more stable quantity than the mean of radiances from central ,zed pixels. The lunar model is compared to ground-based scans by the SeaWiFS instrument of a full moon in December 1992. Quality assurance and trend analyses routines Ior calibration and for telemetry data are also discussed. 14. SUBJECT TERMS Lunar Observations, 26 SeaWiFS, Oceanography, Modeling, Solar Observations, Calibration is. NUMBER OF PAGES 111. PRICE CODE 17. SECURITY CLASSIFICATION 111.SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION OF REPORT OF THIS PAGE Unc lassified Unc lassi fled NSN7540-01-2110-SS00 20. LIMITATION OF ABSTRACT OF ABSTRACT Unclassified Unlimited Standard Form 2911(Rev 2-119) Prelcrll_d by ANSI 1111. _$1-11, 2N.I02