Full report

Original page 1

NASA/TM-1999-206892, Vol. 4 SeaWiFS Postlaunch Technical Report Series Stanford B. Hooker and Elaine R. Firestone, Editors Volume 4, The 1997 Prelaunch Radiometric Calibration of SeaWiFS B. Carol Johnson, Edward E. Early, Robert E. Eplee, Jr., Robert A. Barnes, and Robert T. Caffrey April 1999

Original page 2

The NASA STI Program Office ... in Profile Since its founding, NASA has been dedicated to the advancement of aeronautics and space science. The NASA Scientific and Technical Information (STI) Program Office plays a key part in helping NASA maintain this important role. The NASA STI Program Office is operated by Langley Research Center, the lead center for NASA's scientific and technical information. The NASA STI Program Office provides access to the NASA STI Database, the largest collection of aeronautical and space science STI in the world. The Program Office is also NASA's institutional mechanism for disseminating the results of its research and development activities. These results are published by NASA in the NASA STI Report Series, which includes the following report types: • TECHNICAL PUBLICATION. Reports of completed research or a major significant phase of research that present the results of NASA programs and include extensive data or theoretical analysis. Includes compilations of significant scientific and technical data and information deemed to be of continuing reference value. NASA's counterpart of peer-reviewed formal professional papers but has less stringent limitations on manuscript length and extent of graphic presentations. • TECHNICAL MEMORANDUM. Scientific and technical findings that are preliminary or of specialized interest, e.g., quick release reports, working papers, and bibliographies that contain minimal annotation. Does not contain extensive analysis. • CONTRACTOR REPORT. Scientific and technical findings by NASA-sponsored contractors and grantees. CONFERENCE PUBLICATION. Collected papers from scientific and technical conferences, symposia, seminars, or other meetings sponsored or cosponsored by NASA. SPECIAL PUBLICATION. Scientific, technical, or historical information from NASA programs, projects, and mission, often concerned with subjects having substantial public interest. TECHNICAL TRANSLATION. English-language translations of foreign scientific and technical material pertinent to NASA's mission. Specialized services that complement the STI Program Office's diverse offerings include creating custom thesauri, building customized databases, organizing and publishing research results... even providing videos. For more information about the NASA STI Program Office, see the following: Access the NASA STI Program Home Page at http://www, sti.nasa.gov/STI-homepage.html E-mail your question via the Internet to help@sti.nasa.gov Fax your question to the NASA Access Help Desk at (301) 621-0134 • Telephone the NASA Access Help Desk at (301) 621-0390 Write to: NASA Access Help Desk NASA Center for AeroSpace Information 7121 Standard Drive Hanover, MD 21076-1320

Original page 3

NASA/TM-1999-206892, Vol. 4 SeaWiFS Postlaunch Technical Report Series Stanford B. Hooker, Editor NASA Goddard Space Flight Center, Greenbelt, Maryland Elaine R. Firestone, Senior Technical Editor SAIC General Sciences Corporation, Beltsville, Maryland Volume 4, The 1997 Prelaunch Radiometric Calibration of SeaWiFS B. Carol Johnson and Edward E. Early National Institute of Standards and Technology Gaithersburg, Maryland Robert E. Eplee, Jr., and Robert A. Barnes SAIC General Sciences Corporation Beltsville, Maryland Robert T. Caffrey NASA Goddard Space Flight Center Greenbelt, Maryland National Aeronautics and Space Administration Goddard Space Flight Center Greenbelt, Maryland 20771 April 1999

Original page 4

ISSN 1522-8789 Available from: NASA Center for AeroSpace Information 7121 Standard Drive Hanover, MD 21076-1320 Price Code: A17 National Technical Information Service 5285 Port Royal Road Springfield, VA 22161 Price Code: A10

Original page 5

B.C. Johnson, E.A. Early, R.E. Eplee, Jr., R.A. Barnes, and R.T. Caffrey ABSTRACT The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) was originally calibrated by the instrument's manufacturer, Santa Barbara Research Center (SBRC), in November 1993. In preparation for an August 1997 launch, the SeaWiFS Project and the National Institute of Standards and Technology (NIST) undertook a second calibration of SeaWiFS in January and April 1997 at the facility of the spacecraft integrator, Orbital Sciences Corporation (OSC). This calibration occurred in two phases, the first after the final thermal vacuum test, and the second after the final vibration test of the spacecraft. For the calibration, SeaWiFS observed an integrating sphere from the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC) at four radiance levels. The spectral radiance of the sphere at these radiance levels was also measured by the SeaWiFS Transfer Radiometer (SXR). In addition, during the calibration, SeaWiFS and the SXR observed the sphere at 16 radiance levels to determine the linearity of the SeaWiFS response. As part of the calibration analysis, the GSFC sphere was also characterized using a GSFC spectroradiometer. The 1997 calibration agrees with the initial 1993 calibration to within 4-4%. The new calibration coefficients, computed before and after the vibration test, agree to within 0.5%. The response of the SeaWiFS channels in each band is linear to better than 1%. In order to compare to previous and current methods, the SeaWiFS radiometric responses are presented in two ways: using the nominal center wavelengths for the eight bands; and using band-averaged spectral radiances. The band-averaged values are used in the flight calibration table. An uncertainty analysis for the calibration coefficients is also presented. 1. INTRODUCTION The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) is an optical sensor, launched on 1 August 1997, which measures the radiance of Earth from orbit. Measurements occur at eight spectral regions from about 410 870 nm; at each band, the bandwidths are between 20 nm and 40 nm (Barnes et al. 1994a). The purpose of the SeaWiFS Project is to obtain valid ocean color data, which is then used to determine biological parameters. The required relative standard uncertainty in the water-leaving spectral radiance from the SeaWiFS measurements is 5% (Hooker et al. 1992). SeaWiFS was originally calibrated by the instrument manufacturer, Raytheon Santa Barbara Research Center t (SBRC), in November 1993 at the SBRC facility in California. Since that time, the instrument was integrated onto the OrbView-2 spacecraft (formerly called SeaStar) by Orbital Sciences Corporation (OSC) and underwent several rounds of environmental testing. In preparation for an August 1997 launch of OrbView-2, the SeaWiFS Project, the National Institute of Standards and Technology (NIST), and OSC undertook a second calibration of SeaWiFS in January and April 1997 at OSC's Germantown, Maryland facility. In addition, limited electronic data were taken in June 1997 at Vandenberg Air Force Base (VAFB), California to ensure that the instrument had survived shipment from OSC to VAFB for launch. There were several reasons for recalibrating SeaWiFS prior to launch. First, more than four years had elapsed Formerly known as Hughes Santa Barbara Fl.emote Sensing, a subsidiary of Hughes Aircraft. from the initial calibration of the instrument, raising concerns about long-term changes in the optical, electrical, and mechanical components of the instrument. Second, the effects of spacecraft environmental testing on the instrument needed to be quantified. Third, the initial radiometric calibration of the instrument did not include a determination of the radiometric uncertainties in the calibration, and that analysis is presented here. For these reasons, SeaWiFS was recalibrated in January 1997 after the final thermal vacuum test of OrbView-2. The calibration was repeated in April 1997 after the final vibration test of the spacecraft to quantify possible changes in the response of the instrument. For the calibration at OSC, SeaWiFS observed the spectral radiance of a 1.07 m diameter (40 in) integrating sphere [from the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC)] at four radiance levels. Measurements were obtained with SeaWiFS at all 4 electronic gains and for the 32 channels in the 8 bands. The standard gain setting for ocean observations is gain 1, denoted g 1. The standard operating mode for ocean observations is to average the output of the four channels in each band. Alternatively, four sequential measurements with the same channel are averaged (Table 1 in Woodward et al. 1993). In Woodward et al. (1993), these modes are denoted as the 4:1 and 1:1 time delay and integration (TDI) setting, respectively$. The central results $ In some earlier references, the notation 1:1, 2:2, 3:3, and 4:4 is used, thereby denoting which of the four channels was involved. 1

Original page 6

The 1997 Prelaunch Radiometric Calibration of SeaWiFS from this work correspond to g 1 and the 1:1 TDI setting, but the other results are also reported. On orbit, SeaWiFS uses g 1 and the 4:1 TDI setting. A six channel filter radiometer, the SeaWiFS Transfer Radiometer (SXR)J was used as a transfer standard for the GSFC sphere in 1995 and 1997. The SXR is described in Johnson et al. (1998). Measurements of the sphere output were made with the SXR as part of the sphere's calibration at NIST (Early and Johnson 1997). It was expected that the SeaWiFS calibration would occur in 1995 and that SXR measurements would verify the stability of the sphere's output between the NIST calibration of the sphere, and the calibration of SeaWiFS using the sphere at OSC. There were various delays, however, and the 1997 SXR measurements indicated that the spectral radiance of the sphere source decreased by approximately 6% at 411 nm, and by lesser amounts at longer wavelengths. Consequently, the sphere radiance had to be determined using the 1995 and 1997 SXR measurements. The sb: SXR wavelengths range from 410 775 nm. For wavelengths outside of this region, measurements were made using a model 746 spectroradiometer (Optronic Laboratories, Inc.) with an integrating sphere irradiance collector (746/ISIC) to determine the source spectral shapeS. This system from GSFC is described in Johnson et al. (1996). The measurements described here were made using the SeaWiFS/OrbView-2 system on 10 separate days between 21 January 1997 and 16 June 1997. The events occurred in three major steps: 1) Measurements after the OrbView-2 final thermal vacuum test; 2) Measurements after the OrbView-2 final vibration test; and 3) Measurements after the shipment of OrbView-2 to VAFB. These instrument tests are summarized in Tables 1 3. The measurement of the sphere source with the 746/ISIC was a separate experiment, done at OSC, when SeaWiFS was involved in other tests. This document begins with a summary of the initial calibration of SeaWiFS by SBRC. This is followed by a discussion of the equipment and procedures used in the calibration of SeaWiFS at OSC in Germantown, Maryland. Then, the calibration of the GSFC sphere is presented using the SXR along with the conversion of the flux measured There is an ambiguity in the nomenclature for SeaWiFS and the SXR. SeaWiFS is an eight band instrument, each band having its own interference filter and electronic output. The SXR is a six channel instrument, each channel having its own interference filter and electronic output. In this document, an SXR channel is functionally equivalent to a SeaWiFS band. Identification of commercial equipment to adequately specify the experimental problem does not imply recommendation or endorsement by NASA or NIST, nor does it imply that the equipment identified is necessarily the best available for the purpose. by the SXR to the radiance observed by SeaWiFS. The calculation of the SeaWiFS calibration coefficients from the sphere radiances, including their uncertainties, is then derived. Finally, the 1993 and 1997 calibrations of SeaWiFS are compared. 2. 1993 SeaWiFS CALIBRATION The initial characterization and calibration of SeaWiFS was performed by the instrument's manufacturer, SBRC, in 1993. Those tests are described in the SeaWiFS Calibration and Acceptance Data Package (an SBRC internal document) and in the Sea WiFS Technical Report Series (Barnes et al. 1994a, Barnes et al. 1994b, and Barnes and Eplee 1997). The original tests at SBRC that relate to the radiometric calibration at OSC are summarized below. A comparison of the results of the 1993 SBRC calibration of SeaWiFS and the 1997 calibration of SeaWiFS is discussed in Sects. 8 and 9. 2.1 Channel Configuration Barnes and Holmes (1993) and Barnes et al. (1994b) described the SeaWiFS instrument in depth. SeaWiFS is an eight-band filter radiometer which measures the upwelling Earth radiance in the visible and near-infrared spectral regions. The specified wavelengths for these bands are 412 nm, 443 nm, 490 nm, 510 nm, 555 nm, 670 nm, 765 nm, and 865 nm (Barnes et al. 1994b). These values, also known as the nominM center wavelengths, are denoted A D and are part of the instrument specifications to the manufacturer. The actual center wavelengths differ from the nominal values by up to 2 or 3nm (Barnes and Yeh 1996). Each SeaWiFS band has four independent channels. A channel consists of a silicon photodiode, transimpedance amplifier, intermediate electronics, and an analog-to-digital (A/D) converter. The photodiodes for each band are etched from a single piece of silicon and are located behind a single interference filter. This gives each channel in that band a nearly identical spectral response. Three channels in each band have almost equal overall gains, and the other channel has a different gain. The overall gains are adjusted so the three high gain (ocean) channels are optimized for the low radiance levels associated with measurements of the Earth's oceans. The A/D converter (but not the photodiode, transimpedance amplifier, or the intermediate amplifiers) saturates when observing bright sources, such as clouds or bright calibration sources. Saturation occurs at 1,023 digital counts. When SeaWiFS is operated in the standard 4:1 TDI setting, the output of the four channels in each band are averaged (Sect. 1). Consequently, a knee in the digital output of the band, as a function of radiance, occurs when the ocean channels reach saturation; the band has high gain (sensitivity to radiance) below the knee and lower sensitivity above (Barnes et al. 1994b). This change in sensitivity of

Original page 7

B.C. Johnson, E.A. Early, R.E. Eplee, Jr., R.A. Barnes, and R.T. Caffrey Table 1. Calibration measurements following the final thermal vacuum test of OrbView-2. Date Measurements 3 January 1997 Initial characterization of the GSFC sphere output prior to moving the sphere from NIST in Gaithersburg, Maryland to OSC in Germantown, Maryland. The changes in the sphere radiance were first noted. 21 ,January 1997 Verification of the sphere output at OSC using the SXR. 22 ,January 1997 Preliminary measurements of the sphere with SeaWiFS. Data were obtained at the standard SeaWiFS gain and TDI settings for the sphere output with a lamp configuration of 16, 8, 1, and 0. 23 ,January 1997 Repeat measurements of the sphere with SeaWiFS and the SXR. Data were obtained for the sphere output with 16, 8, 4, 1, and 0 lamps in operation three times, once in the morning and twice in the afternoon. 24 ,January 1997 SeaWiFS prelaunch calibration, part 1. Measurements were made for all gain and TDI settings for the sphere output with 16, 8, 4, 1, and 0 lamps operating. Gain ratios for each detector were made using the electronic calibration pulse. 6 March 1997 Characterization of the spectral radiance of the sphere output with 16 lamps operating using the SXR and the GSFC 746/ISIC. FEL lamp F182 was used. 7 March 1997 Characterization of the spectral radiance of the sphere output with 16 lamps operating using the SXR and the GSFC 746/ISIC. FEL lamp F391 was used. Table 2. Calibration measurements following the final vibration test. Date Measurements 11 April 1997 SeaWiFS prelaunch calibration, part 2. Measurements were made for all gain and TDI settings for the sphere output with 16, 8, 4, 1, and 0 lamps operating. Measurements were also made at all 16 possible radiance levels from the sphere (the standard gain and TDI settings) to measure the linearity of the SeaWiFS responses. 23 April 1997 Postvibration calibration pulse test. Gain ratios were obtained for each detector. Table 3. Calibration measurements following the shipment of OrbView-2 to VAFB. Date Measurements 16 June 1997 Post-shipment calibration pulse test. Gain ratios were measured for each detector. the band is referred to as bilinear gain. It is possible to change the TDI mode to provide the average of four readings from one channel in a band (denoted 1:1 TDI in this document); other possible combinations of the individual channels in each band can also be implemented (Woodward et al. 1993). In the 1:1 TDI setting, the low sensitivity channel in each band does not saturate during the laboratory measurements, while the high sensitivity channels have constant digital output above the saturation radiance. For extremely bright sources, the low sensitivity channel would also saturate. SeaWiFS was designed so measurements of clouds and land surfaces would not saturate the low sensitivity channels; however, in some cases, sun glint from the ocean could result in saturation of these channels. The electronic gains for the high sensitivity channels in each band can be changed, while the gains for the low sensitivity (or cloud) channels are fixed. The standard gain for ocean measurements is g 1, and g 2 is used for unusually dark oceans, such as when the solar zenith angle is exceptionally large. The gain settings of g 3 and g 4 are used for measurements of the moon and the SeaWiFS solar diffuser, which are used as calibration targets. For each choice, the gain is the same for all of the high sensitivity ocean channels for all bands. The SeaWiFS calibration at OSC included measurements at g 1 for each band using the 4:1 and 1:1 integration modes (an average of the four channels in each band and an average of the four readings of each channel separately). This required five measurements for each band. In addition, measurements were made for g 2, g 3, and g 4, and for spectral radiances above and below the knees in the bilinear gains. 2.2 Initial Calibration SBRC performed the radiometric calibration of Sea- WiFS in November 1993. For that calibration, the 1:1 3

Original page 8

The1997PrelaunchRadiometricCalibrationofSeaWiFS Table4. SeaWiFScalibrationcoefficientsfromtheSBRCcalibrationforg 1 and the 1:1 TDI setting. Results are given using two methods: band-centered (columns 3 and 4) and band-averaged (columns 5 and 6). The wavelength is expressed in nanometers and the calibration coefficients are in units of mW cm 2 sr 1 #m i count ] Band Channel AD [] Calibration AB [] Calibration Number Number CoeNcients[-g] CoeNcients[-D-] 1 1 412 0.060039 415.5 0.062084 2 0.010903 0.011275 3 0.010994 0.011369 4 0.010878 0.011248 2 1 443 0.010546 444.7 0.010676 2 0.010533 0.010664 3 0.010582 0.010713 4 0.067570 0.068407 3 1 490 0.068075 493.0 0.069132 2 0.008228 0.008356 3 0.008163 0.008290 4 0.008163 0.008290 4 1 510 0.007150 511.0 0.007145 2 0.007150 0.007145 3 0.007133 0.007128 4 0.066333 0.066289 5 1 555 0.065167 558.2 0.065375 2 0.005736 0.005754 3 0.005757 0.005775 4 0.005785 0.005804 6 1 670 0.003241 668.8 0.003220 2 0.003228 0.003207 3 0.003292 0.003270 4 0.054820 0.054449 7 1 765 0.042978 767.5 0.042974 2 0.002298 0.002297 3 0.002306 0.002305 4 0.002298 0.002297 8 1 865 0.001633 865.0 0.001626 2 0.001646 0.001638 3 0.001628 0.001621 4 0.034280 0.034113 [] Table 5 in Barnes et al. (1994b). [] Table 23 in Barnes and Eplee (1997). TDI setting was used and SeaWiFS observed the SBRC integrating sphere, model SIS-100, at six radiance levels. These radiance levels, selected by operating subsets of the lamps in the SIS-100, resulted in a determination of the calibration coefficient for each channel at a single radiance value. Seven of the SeaWiFS bands were calibrated at g 1, while g 4 was used for band 8 to avoid saturation of the channels. Ancillary data on the gain ratios for each channel were used to convert the calibration coefficient to those at the other instrument gains. In October 1993, the spectral radiances of the SIS-100 were measured by SBRC at the six lamp levels by compar- 4 [] Table 21 in Barnes and Eplee (1997). [] Table 25 in Barnes and Eplee (1997). ing them with NIST-traceable standard irradiance lamps L The goal was to achieve a relative standard uncertainty of 5_ for the radiometric calibration of SeaWiFS. The 1993 calibration coefficients at the values of A D are given in Table 4. These results are derived from the original method used to analyze the calibration data; it is best suited to instruments with narrow spectral bandwidths. Once the actual relative spectral responsivities, R(5), for SeaWiFS were determined, it was realized a more thorough approach For a diagram of the sphere calibration facility at SBRC, see Appendix B in Mueller (1993).

Original page 9

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey andYeh1996).Table4 alsore- saturated for all eight radiance levels. The measurements wasnecessary(Barnes for the from band 2 were not saturated for seven of the radiance portsthe 1993SBRCresultsusingthis approach SeaWiFScalibration.In referencetodeterminingtheradi- levels (saturation occurred for the brightest level). The aretermed values from band 3 were unsaturated for sb: of the raanceofthecalibrationsource,thetwomethods "band-centered"and"band-averaged"(Sect.6.1). diance levels (saturation occurred for the two brightest The band-averagedmethodof determiningthe mea- levels). This sequence continued to band 8, where the surementwavelengthfor eachbandusesthe actualR(A) functions for SeaWiFS. The value for the band-averaged center wavelength, AB, is dependent on the relative spectral shape of the radiance of the SBRC sphere source: AB f )LSBR°()OR())dA fLsBRo(A) R(A) dA (1) The spectral radiance LSBRC(A) in (1) corresponded to the brightest of the sb: radiance levels from the SIS-100, as determined in 1993. The band-averaged radiances were calculated using the LSBRC(A) and the R(A) values and the equation for band-averaged radiance [Barnes 1996b or (5) below]. Table 4 gives these wavelengths and the corresponding band-averaged calibration coefficients from the 1993 measurements. The inclusion of both methods in this document is for historical reasons and for comparison purposes (Sect. 6.1). 2.3 Gain Ratios The gain ratios for the SeaWiFS channels, that is, the output of the individual channels at g 2, g 3, or g 4, normalized by the gain at g 1, were measured by the manufacturer in November 1993. A square-wave (voltage) calibration pulse was injected into the post-photodiode electronics during a portion of the SeaWiFS scan when the detectors were not illuminated. The output of the photodiode's transimpedance amplifier and the voltage pulse are summed and fed into the amplifier that controls the different gain settings (Woodward et al. 1993). The ratios of the channel output, relative to that at g 1, are the gain ratios. The gain ratios in the SeaWiFS performance specifications (Barnes et al. 1994a) are given for each of the eight bands for the standard 4:1 TDI setting. Because the instrument has bilinear gains, the gain ratios must be determined for each channel. The gain ratios, as determined in November 1993, are listed in Table 5. This test was repeated during the prelaunch calibration of SeaWiFS, and it will be repeated routinely on orbit. 2.4 Linearity The linearity of SeaWiFS was tested at SBRC in February 1993 and again in November 1993 (Barnes et al. 1994a). In the February test, SeaWiFS observed 8 radiance levels, with 25 individual measurements at each level. This linearity test was performed for each band at radiance levels where the instrument output was not saturated. During the February 1993 test, the values from band 1 were not values were unsaturated for only one radiance level (the lowest). Except for one radiance level where the output of the SBRC sphere was an outlier, the calibration coefficients deduced for SeaWiFS for each band at a particular radiance level agreed with the average calibration coefficient for that band to better than 1%. With the radiance level that was suspect included, the agreement was from 1.4 1.9%. Table 5. SeaWiFS gain ratios, relative to g 1, from the SBRC calibration [from Table 6 in Barnes et al. (1994b)]. Band Channel Gain Setting No. No. g 1 g 2 g 3 g 4 1 1.000 1.000 1.000 1.000 1.000 1.988 1.320 1.681 1.000 1.988 1.320 1.681 1.000 1.988 1.320 1.681 1.000 1.989 1.319 1.682 1.000 1.989 1.319 1.682 1.000 1.989 1.319 1.682 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.989 0.896 1.681 1.000 1.989 0.896 1.681 1.000 1.989 0.896 1.681 1.000 1.989 0.789 1.682 1.000 1.989 0.789 1.682 1.000 1.989 0.789 1.682 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.989 0.642 1.595 1.000 1.989 0.642 1.595 1.000 1.989 0.642 1.595 1.000 1.989 0.364 0.665 1.000 1.989 0.364 0.665 1.000 1.989 0.364 0.665 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.987 0.311 0.575 1.000 1.987 0.311 0.575 1.000 1.987 0.311 0.575 1.000 1.991 0.261 0.499 1.000 1.991 0.261 0.499 1.000 1.991 0.261 0.499 1.000 1.000 1.000 1.000 5

Original page 10

The1997PrelaunchRadiometricCalibrationofSeaWiFS In November1993,SeaWiFSobservedthreeradiance1997 to verify the operation of the sphere before shipment levelswith alleightbands.Thesereferenceradiancesare to OSC, and at OSC during the post-thermal vacuum calthoseexpectedat thetop ofthe atmosphere,andarealso ibration of SeaWiFS. Every measurement of the GSFC calledtypical radiancesin the SeaWiFSspecificationssphere by SeaWiFS was bracketed by measurements of the (Barneset al. 1994a).Theradiancelevelsfor theNovem-sphere with the SXR. For all measurements, the voltage ber1993testwerethetypicalradianceforeachband,two- gain was unity, the SXR was located about 135 cm from thirdsofthetypicalradiance,andone-thirdofthetypical the exit aperture of the integrating sphere, and the lens radiance.Forband1 andbandsg 8, the agreementbe- focal setting was between 1.2 m and 1.3 m. For the majortweenthecalibrationcoefficientsateachradianceleveland ity of the measurements, the SXR was aligned to view the the averagevalueforeachbandwasbetterthan1_. For center of the exit aperture of the sphere at normal inciband2, the agreementwasfrom1.2 1.g_. Again,there dence. For the SeaWiFS linearity tests at OSC, the SXR wasanoutlierin the SIS-100sphereoutputfor oneofthe was placed about 2 m from the sphere and aligned to view band2 measurements(Barneset al. 1994a). 3. TEST EQUIPMENT The equipment used during the 1997 calibration of the SeaWiFS instrument is described below, and references that include detailed descriptions of the equipment are cited. The equipment includes the SXR, the GSFC integrating sphere, and the 746/ISIC spectroradiometer. The experimental configuration at OSC and the method of data acquisition for SeaWiFS on the OrbView-2 system are also described. 3.1 SXR The SXR is a sLx-channel radiometer calibrated for spectral radiance over the wavelength range from 410 775nm (Johnson et al. 1998). Each measurement channel consists of a silicon photodiode and an interference filter, which are temperature stabilized at 26°C, and a transimpedance amplifier. Refractive and reflective optical components image the source onto the six filtered photodiodes. The full-angle field of view is 2.5 ° and an alignment system ensures proper focus on objects from 85 cm to infinity. At the minimum focus, the diameter of the source area imaged by the SXR is approximately 45 mm. The SXR was characterized and calibrated at NIST for relative spectral responsivity, linearity with radiant flux, point-spread responsivity, and stability. It was calibrated in 1994 and 1996 using a small integrating sphere which was calibrated for spectral radiance using the NIST Facility for Automated Spectroradiometric Calibrations (FAS- CAL). The SXR is controlled, and data are logged, using a Macintosh computer running custom software which sequentially records signals from each channel as selected by the operator. The recorded measurements from the SXR were made with a single digital multimeter (DMM), a Hewlett-Packard (HP) model 34401A. The DMM was calibrated in June 1995 and July 1996. The SXR electronics and temperature control system, as well as the DMM, were warmed up for about 12 h before any measurements were made. The SXR was used to measure the GSFC sphere during its calibration at NIST in April 1995, again in January the center of the exit aperture at an angle of about 30 ° from normal. The SXR can be rotated about the optical axis in increments of 90 °. During the NIST calibration of the sphere in 1995 and the measurements at NIST of the sphere in January 1997, the orientation was north (one of the connectors on the back of the SXt{ is used to reference the orientation), but at OSC the orientation was east. The difference in orientation is not expected to affect the SXR measurement uncertainty (Johnson et al. 1998). The SXR has a small, wavelength-dependent, size-ofsource effect (SSE) that results in an overestimate of the spectral radiance of sources that are larger than the source used to calibrate the SXR. Two methods were used to account for the SSE in the measurements of the integrating sphere. For the sphere calibration in 1995 and the study in January 1997, (both at NIST) a model incorporating the normalized point-spread responsivity data was used to determine the correction. For the measurements at OSC, an on-axis circular black disc was placed in front of the exit aperture of the sphere and centered on the optical axis of the SXR. Measurements of the disc were used for the SXR instrument background, thus accounting for the amplifier offset and the extraneous signal from the finite response outside the instrument's nominal field of view. The SXR parameters are given in Table 6. The SXR measurement wavelengths, AsxR, were calculated using a moment analysis which includes the relative spectral response (Johnson et al. 1998). The analysis does not include the spectral shape of the source that is measured: AsxR f AR(A)dA f R(A)dA • (9) Here, R(A) is the relative spectral responsivity for one of the six channels in the SXR. The SXR channels are narrow compared to the 20 nm and 40 nm bandwidths for the SeaWiFS bands, so the measurement wavelengths are not very sensitive to the relative spectral shape of the source (Johnson et al. 1998). In Table 6, the calibration coefficients are negative because the output voltage from the SXt{ is negative for measurements of optical sources if the interconnecting cable

Original page 11

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table 6. SXRparametersat a voltagegainof unity. As an exampleof the combinedrelativestandard uncertaintyforSXRmeas_rementsof spectralradiance,valuesaregivenwhichcorrespondto measurementsof the 1.07mdiameterGSFCs)heresourcewith 16lampsin operation. Channel AsxR Calibration Number Coefficient [nm] [Vcm2 sr #mmW 1] 1 411.222 -1.101185 2 441.495 -1.468061 3 486.938 -0.2442614 4 547.873 -0.2425734 5 661.718 -0.2604715 6 774.767 -0.03013285 is connected with the ground shield to the common input of the DMM. The SSE correction was calculated using the modeled point-spread measurements at NIST in April in referenced to the 16-1amp setting, because the dynamic 1995, and was measured using the black disc at OSC 1997. Both correction factors are given in Table 6; the agreement is within -t-0.3%. The relative standard uncertainty is the root-sum squared of the individual components of uncertainty which correspond to the SXR measurements of the GSFC sphere source with 16 lamps operating (,Johnson et al. 1998, and Tsai and ,Johnson 1998). These uncertainties are stated to illustrate the typical accuracy of the SXR for measurements of integrating sphere sources. 3.2 GSFC Integrating Sphere The GSFC integrating sphere is 107 cm in diameter and coated with barium sulfate. It was manufactured by Optronic Laboratories Inc. for the National Oceanic and Atmospheric Administration (NOAA) and later transferred to the Space Geodesy Network and Sensor Calibration Branch at GSFC. The exit aperture is 39.5cm in diam- 16 the sphere source. The SXR was also used to measure eter and the sphere is internally illuminated by up to baffled 45 W quartz halogen lamps. The lamps are positioned uniformly around the exit aperture. The lamps are divided into four sets of four, and each set is operated by a separate precision current source. Additional details can be found in ,Johnson et al. (1996) and Early and ,Johnson (1997). In anticipation of the radiometric calibration measurements at OSC prior to the launch of the OrbView-2 satellite, and in support of the Fourth SeaWiFS Intercomparison Round-Robin Experiment (SIRREX-4, Johnson et al. 1996), the GSFC integrating sphere was calibrated at NIST in April 1995. The calibration used a standard strip lamp calibrated for spectral radiance, L(A), and a prism-grating monochromator (Early and Johnson 1997). The calibration was done at four radiance levels, with first 16 lamps, then 8, 4, and 1 lamp in operation. The strip lamp, which was calibrated using FASCAL, was used to determine the spectral radiance responsivity of the monochromator. The Size-of-Source Correction Relative Standard GSFC Sphere Uncertainty Model Black Disc [%] 0.9954 0.9972 0.99 0.9987 0.9958 1.19 0.9988 0.9961 0.62 0.9975 0.9969 0.65 0.9948 0.9980 0.60 0.9914 0.9930 0.68 monochromator was used to determine the spectral radiance of the exit aperture of the sphere at the 16-1amp setring. Measurements at the other three radiance levels were range of the monochromator was insufficient for direct ratios to the strip lamp. The measured spectral radiance and the relative standard uncertainty were reported every 10nm from 380 1,100nm. The relative standard uncertainties, u nL, which is equal to u(L(A))/L(A), were from 0.3 2.6%, depending on the measurement wavelength and the number of lamps in operation. Table 7 summarizes the spectral radiance of the GSFC sphere as measured in April 1995 at NIST; more details are given in Early and Johnson (1997). During the April 1995 measurements, the SXR was used to measure the spectral radiance at the center of the exit aperture of the GSFC sphere. The SXR and the NIST strip lamp and spectroradiometer system agreed to within 1.2% at the six measurement wavelengths of the SXR and all four sphere levels (Early and Johnson 1997). The agreement was within the combined uncertainties of the two methods for determining the spectral radiance of the spatial uniformity of the spectral radiance in the exit aperture of the sphere source at the same four radiance levels. 3.3 GSFC Spectroradiometer The GSFC 746/ISIC spectroradiometer is a scanning single-grating monochromator which has a 10.2 cm diameter integrating sphere with a known entrance aperture as the collection optic. The 746/ISIC is calibrated prior to each measurement of an integrating sphere using a 1,000 W quartz-halogen standard irradiance lamp which is traceable to NIST. First, the irradiance responsivity is determined from measurements of the standard lamp. Then, the 746/ISIC is rotated to measure the spectral irradiance from the integrating sphere. The diameter of the exit aperture of the sphere, the diameter of the entrance aperture of the small sphere on the 746/ISIC, and the distance between the two apertures is used to convert the 7

Original page 12

The1997PrelaunchRadiometricCalibrationofSeaWiFS Table 7. GSFC sphere radiances and relative standard design wavelengths for the lamp configurations studied uncertainties, u Ln u(L(A))/L(A), at the SeaWiFS by Early and ,Johnson (1997). The spectral radiance L(A) is given in units of mW cm 2 sr i nm 1 and the relative standard uncertainty is in percent. A D 16 Lamps 8 Lamps [nm] L(A) uL6 L(A) u L 412 3.9096 0.51 1.8763 0.53 443 6.5185 0.47 3.1403 0.49 490 11.525 0.42 5.5805 0.42 510 13.960 0.40 6.7728 0.41 555 19.774 0.38 9.6299 0.38 670 33.762 0.33 16.548 0.33 765 41.605 0.31 20.433 0.31 865 45.046 0.30 22.188 0.30 measured spectral irradiance to the average spectral radiance for the exit aperture of the sphere. For the measurements at OSC on 6 March, the 746/ISIC measured from 400 900 nm with a bandwidth of about 4.8 nm and a step size of 10nm. A single grating with 1,200 lines per millimeter, and blazed at 500 nm, was used, along with two order sorting filters which eliminated second order effects. On 7 March, the measurements covered the spectral range from 400 1,150nm. The 746/ISIC system is sensitive to extraneous sources of radiation, since the field of view is the entire hemisphere. Table 8 indicates that scattered radiation is the major contributor to the uncertainty in the measurements of the 746/ISIC for wavelengths that are shorter than 440 nm. Table 8. Estimated uncertainties for the 746/ISIC near a subset of the SeaWiFS bands (,Johnson et al. 1997). Wavelength Relative Standard [nm] Uncertainty [%] 412 6.5 443 2.5 490 1.9 565 1.6 665 1.5 For the SeaWiFS calibration project, the 746/ISIC was used on 6 7 March to measure the spectral radiance of the GSFC sphere source in the OSC clean room. On 6 March, the SXR and the 746/ISIC measured the GSFC sphere with 16 lamps operating. Standard lamp F182, which was 4 Lamps 1 Lamp L(A) u L L(A) u L 0.93162 0.54 0.23594 0.75 1.5616 0.49 0.39759 0.56 2.7808 0.42 0.70939 0.44 3.3771 0.41 0.86132 0.42 4.8069 0.38 1.2253 0.39 8.2771 0.33 2.1060 0.33 10.233 0.31 2.5944 0.32 11.121 0.30 2.8096 0.32 3.4 SeaWiFS and OrbView-2 SeaWiFS was integrated with OrbView-2 in January 1994 and was shipped to VAFB in June 1997. During the January and April 1997 calibration, OrbView-2 was in the high-bay clean room at the OSC Germantown, Maryland facility. The instrument was controlled using the ground support equipment (GSE) at OSC, which was located just outside the clean room. In order to reduce sources of electronic noise, SeaWiFS and OrbView-2 were operated with battery power from the solar array simulators. In SeaWiFS, the scan mirror rotates 360 ° at 6Hz so that during portions of the scan, the view is through the instrument input aperture, while at other times, the inside of the instrument housing is observed. A second aperture provides a view of the SeaWiFS diffuse reflecting panel, which can be illuminated by the sun during the mission. Data can be archived for a selected portion of the complete scan, corresponding to 116.6 °. These portions of the rotation of the scan mirror through the full cycle are named for convenience: Earth mode is when the scan mirror views the Earth from -58.3 ° to +58.3 ° about the nadir position (at 0 ° and the center of the telescope aperture), and solar mode is when the scan mirror views the interior of the instrument and the diffuse panel from 60 175 ° (Woodward et al. 1993). The electronic calibration pulse occurs during a portion of the solar mode, after the view of the on-board diffuser. During the work reported here, data were acquired for both the Earth and solar modes. For both the Earth and solar modes, the zero offsets (or dark counts), for each channel are measured when the scan mirror views the interior of the instrument housing from 140 220 ° Woodward et al. (1993). The instantaneous calibrated using FASCAL in August 1994, was used on field of view of one channel is 1.6 × 1.6 mrad 2. The four 6 March to calibrate the 746/ISIC. Standard lamp F391, calibrated by Optronic Laboratories, was used on 7 March. On both days, the wavelength calibration of the 746/ISIC was determined using selected sources of standard emission lines. channels in each band are aligned along the direction of the moving image from the scan mirror. In either the 1:1 or the 4:1 TDI setting, the output of the four channels in each band are delayed before they are summed, so the resulting average corresponds to the same location on the

Original page 13

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Earth'ssurfacethis is termeda pixel. With the scan the calibration and the postvibration measurements were mirrorrotatingat 6Hz,thereare1,285pixelsin onescan at the discretion of OSC, because the general prelaunch line,foreithertheEarthor the solarmode.Forthedark activities were assigned the highest priority. count(DC)restoremeasurements,the averageoffsetfor ° is calibration of SeaWiFS at the end of January, a test prothe measurementsoverthe scananglesfrom140220 During December 1996 and January 1997, prior to the storedaspixel 0 andis usedto correctthe subsequentcedure was developed and documented in an internal Seapixelsin thatscanline. was uum Calibration Test Procedure. The document was de- Real-timedataacquisitionatOSCin Germantown WiFS publication: TP12001 SeaWiFS Post Thermal Vacusedto controlSeaWiFSandarchivethe data. In order signed as a compilation of the documents that resulted facilityto from the experiment, of the methods of data analysis to be to transferthedatafilesfromthe Germantown a secondemployed, and of the step-by-step experimental activities. the SeaWiFSProjectin Greenbelt,Maryland, Virginiawasused. TP12001 served as a pre-experiment planning document; computerat the OSCsite in Reston, facility once the experiment began, some of the procedures were Computersecurityprotocolsat the Germantown usu- modified and additional procedures were performed. Log preventedthedirecttransferofthefiles.Thetransfer afterthedailyex- sheets from the experiments are appendices in TP12001. allyoccurredduringtheeveninghours, perimentswerecompleted. An attempt was made to plan for possible contingenroomat OSCis cies, such as, failed equipment or schedule conflicts with Theenvironmentally-controlledclean conditionsin the NIST or SeaWiFS personnel. Spare parts, such as, insidethehigh-bayarea.Theenvironmental thetem- seasoned 45W lamps for the GSFC sphere source, were thecleanroomwererecordedon22January1997: humiditywas34%. procured and taken to OSC. Additional personnel were peraturewas21.1°Candthe relative Thefiltersforthecleanroomarelocatedin thesouthwall, made available in case the measurements at OSC came at andthe directionof air flowis fromthis wallto the op- inopportune times for the NIST or SeaWiFS personnel. posite(or north)wall,whichconsistsof manyhorizontal Prior to the calibration, special data acquisition software at the for OrbView-2 was developed so that the SeaWiFS outslats,or louvers.A metalroll-updooris located downstreamendof the cleanroomin the eastwall,so equipmentandmaterialcanbemovedfromthehigh-bay areainto the high-baycleanroom. A rowof windowsis alsosituatedin theeastwall;thesewindowswerecovered with blackpaperduringthe calibrationexperiment.The overheadlightsinsidethecleanroomwereturnedoffduringallmeasurements.Althoughlightcouldentertheclean roomthroughtheopenlouvers,it wasnotpossibleto close themandobstructtheairflow.Instead,alloftheoverhead lightsin thehigh-bayareathat illuminatedthenorthwall werealsoturnedoffduringall ofthemeasurements. 4. MEASUREMENT PROCEDURES The SeaWiFS calibration test occurred over a 4 month interval (Tables 1 3). The measurement procedures associated with the calibration of SeaWiFS can be divided into five functional areas: 1) Preparation and planning; 2) System verification; 3) Optical alignment and equipment operation; 4) Record keeping and archiving of the procedures and data; and 5) Executing measurements. 4.1 Preparation and Planning Preparation and planning were a critical part of the calibration of SeaWiFS. On 9 September 1996, personnel from the SeaWiFS Project, OSC, and NIST met to discuss the experiment. The OSC facility was evaluated, and tasks put counts could be reported in near-real time. Based on the November 1993 calibration of SeaWiFS (Tables 4 5) and the expected radiance of the GSFC sphere from the 1995 NIST calibration (Table 7), the anticipated output of SeaWiFS was determined. A spreadsheet was prepared so the new results could be compared to those based on the 1993 calibration. A spreadsheet was also prepared for the reduction of the SXR data, and the results from the April 1995 SXR measurements of the GSFC sphere were made available for rapid comparison. Once the calibration experiment was underway, daily meetings were held to discuss the results acquired to date, and to review the plans for the next sequence of measurements. These daily review meetings usually occurred in the morning, while SeaWiFS and the GSFC sphere source were warming up. Departures from TP12001, which were generally in the experimental procedures, were discussed and noted as corrections to that document. 4.2 System Verification Before a particular experiment was performed, and in some cases as part of the experiment, the instruments, data acquisition software, and laboratory environment were examined to determine the optimum experimental parameters to be used during the actual measurements. Usually, this consisted of performing additional measurements. These tests established the operating parameters of the instruments, allowed for the identification of sources of bias, and helped to familiarize personnel with the complete experimental system. The first measurement of this nature occurred on 3 Janwere assigned to various teams or individuals. The dates of uary 1997, at NIST. The GSFC sphere was operated with 9

Original page 14

The1997PrelaunchRadiometricCalibrationofSeaWiFS a lampconfigurationof16,8,4,andthen1;measurementswas checked using a bubble level. The support frame was weremadewith theSXR.Theresults,whencomparedto readjusted until the exit aperture was plumb. Then, the theexpectedspectralradianceoftheGSFCsphere(Early adjustable feet of the sphere support frame were locked andJohnson1997),indicatedthatforthesamelampcon- in place and the position of each of the feet was outlined figuration,theoutputhaddecreasedasa functionofwave- on the floor using masking tape. The separation distance length(Sect.1). It mustbe assumedthat the SXRhad between the sphere and SeaWiFS was adequate for placefrom19941997; ment of the SXR and passage of personnel from the SXR remainedin calibrationoverthe interval analysisof operation station on the east side of the clean room to the this assumptionis supportedby a preliminary This3 Jan- exit door on the west side. thesecondSXRcalibrationat NISTin 1996. equip- The SXR was mounted on its tripod, and positioned in uarytestalsodemonstratedthattherewerenogross or the its nominal position 1.35 m from the plane of the sphere's mentfailureswith eitherthe GSFCspheresource SXR(lampfailure,powersupplyinstabilities,etc.). exit aperture to the face plate of the SXR, with the SXR on 21 January aligned to view the center of the exit aperture at nor- The secondverificationtest occurred 1997in the OSCcleanroomin Germantown.The SXR mal incidence. This separation distance corresponded to andthe results the configuration used at NIST in April 1995, and was measuredtheoutputof the GSFCsphere Theresults chosen to achieve repeatability between the two measurewerecomparedto the 3 January1997values. had ments. With the aperture cover over the exit aperture of werein agreement,establishingthat the equipment by the transportfromGaithersburgto the sphere, the height, the location in the horizontal dinot beenaltered Germantown.Severalmeasurementsweremadewithvar- rection, and rotation about these two orthogonal axes of the statusof the the SXR were adjusted until the SXR was aligned coriousilluminationconditions(especially area)sothat ex- rectly to the GSFC sphere. Alignment was judged satisoverheadlightsin theexternalhigh-bay during factory when the SXR was level, at the correct distance traneoussourcesofradiantfluxwouldbeminimized Thefinalconfigura-from the sphere, and centered in the sphere's exit aperthe SeaWiFScalibrationexperiment. tion is describedin Sect.3.4. ture. At the same time, the white alignment circle in the before center of the aperture cover was centered in the SXR's field Thethird verificationwason22January1997, inspectedof view. (The SXR has an ocular alignment system that theSeaWiFScalibrationtests.SBRCpersonnel light is co-aligned with the optical axis of the sL,: measurement the opticsof SeaWiFSusingvisibleandultraviolet wereobservedchannels.) The position of the SXR tripod was recorded sources.A fewfibersanddust particles contaminationby outlining the position of the feet on the floor with tape. insidethehousingonabaffle,butnoserious by vacuuming.During measurements with SeaWiFS or the 746/ISIC, the wasapparent.Thefewfiberswereremoved pulsewasusedto determineSXR and its tripod were moved as a unit to an area away The voltagecalibration of Sea- from the ongoing measurements. gainratios(Sect.2.3)asanadditionalverification WiFS.TheSeaWiFSoutputcorrespondingto a darktarget,termedbackground, was another indicator of system stability. The voltage calibration pulse was used to verify SeaWiFS at OSC and at VAFB (Tables 1 3). 4.3 Alignment and Operation The calibration experiment required that the SXR, the 746/ISIC, and SeaWiFS measure concentric areas in the exit aperture of the GSFC sphere source. Because the measurements were to take place at different times (e.g., before and after the SeaWiFS vibration tests), the alignment procedures and operation of the equipment were designed with repeatability in mind. All of the equipment was cleaned using lint-free cloths and alcohol before they were placed in the clean room. On 21 January 1997, the GSFC sphere source was positioned so the separation between the sphere's exit aperture and SeaWiFS' entrance optic was about 2.5m. The sphere was raised to the highest position possible, and positioned so the SeaWiFS nadir sample measured the center of the sphere's exit aperture at approximately normal incidence. The support frame of the GSFC sphere was made level, and the orientation of the plane of the exit aperture 10 The black disc used for recording the background counts for the various instruments (SXR, 746/ISIC, and SeaWiFS) was mounted on a second tripod. Using the visual sighting system in the SXR, this tripod was aligned to the optical axis of the SXR GSFC sphere system and its location was also marked using tape. For most of the measurements, the disc tripod was left in place but the mount holding the disc was removed and the head of the tripod was lowered to a position below the lower edge of the exit aperture of the sphere. On 22 January 1997, SeaWiFS was aligned to the GSFC sphere after the optics were inspected and cleaned. Using the controls on the mounting platform, OrbView-2 was rotated so that the motion of the SeaWiFS scan mirror resulted in measurements in the horizontal plane; this was the configuration used at SBRC in November 1993. With OrbView-2 level, the optical axis of SeaWiFS was too high and would have resulted in scans that were not centered in the sphere's exit aperture. Because the sphere was at the maximum height, OrbView-2 was pitched by about 1.1 ° using the hydraulic controls on the OrbView-2 mounting platform. OrbView-2 was moved until SeaWiFS appeared to view the center of the sphere's exit aperture with the nadir scan mirror position.

Original page 15

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Becausethe alignmentusingtapemeasureswasdif- 24 January, a special warm-up test of SeaWiFS was done, ficult, the final alignmentinvolvedmeasurementsof the with SeaWiFS turned on after the SXR began its sequence sphereexit apertureunderconditionsof partialobscura-of data acquisition. tion. As onememberof the teamheld a large,opaque targetovera portionofthesphere'sexit aperture,another 4.4 Documentation member,situatedatthecontrolpaneloftheGSE,observed theoutputof aselectedpLxel.Thetargetwasmovedfrom Various instrument record (or log) sheets were designed andthenthever- for use during SeaWiFS calibration. In addition, a labotheedgeto thecenter,in thehorizontal to the ratory notebook suitable for use in the clean room was tical directions.Thefinalalignmentcorresponded of used to record the actual sequence of events. Various peropticalaxisof SeaWiFSabout2.5cmfromthe center with sonnel associated with the overall experiment kept other the exit apertureofthe sphere.Theblackdiscused An undersizedrecords in individual laboratory notebooks. These docu- SeaWiFSwasalignedin a similarfashion. untilthecounts ments, along with this technical memorandum, form the discwaschosenandthetargetwasmoved value.Then,the record of the experiment. in the nadirpLxelreacheda minimum properdiscwasput into place. the GSFC sphere log sheet was designed to record the output Thissequenceof alignmentproceduresestablished for current from each of the four GSFC sphere power supplies, alignmentoftheGSFCsphere,theSXR,andSeaWiFS Three test equipment log sheets were designed. The SeaWiFSas indicated by the digital meters and the voltage across the22 24Januarysetofmeasurements.Because wasmovedin andout ofthe cleanroomforthe vibration each lamp. This voltage is output to a fifth digital meter hadto bere- and a multiturn switch is used to select the lamp corretests,the alignmentproceduregivenabove peatedon11April 1997. sponding to the digital output. The lamps are connected in Johnsonin a four-wire configuration, and the displayed voltage cor- TheGSFCspherewasoperatedasdescribed Theprecisionresponds to the voltage across the lamp. The organization et al. (1996)andEarlyandJohnson(1997). the of the GSFC sphere log sheet was made to coincide with the currentsupplieshavea potentiometerfor adjusting current.On21January,the first time the GSFCspherenormal operation of the sphere source (Early and Johnson wasoperatedatOSC,thepotentiometerswereadjustedso 1997). This involves measurements with a lamp configuraread tion of 16, 8, 4, 1, and 0 lamps operating. The current outthedigitaldisplayonthefourseparatepowersupplies 6.500A 4-0.002 A. With 16 lamps operating, the sphere was allowed to warm up about 20 min before the final adjustment to the potentiometers. The power supplies contain an internal ramp-up circuitry so the current is varied from 0 6.5 A over an interval of about 2 min. This protects the filament in the lamp; however, there is no ramp-down circuitry. Because the position of the potentiometers can be difficult to reproduce, and because this was the standard operating practice for the GSFC sphere, the lamps were turned off by immediate cessation of the current. This may act to shorten the lamp lifetime, but it does result in enhanced repeatability of the GSFC sphere radiance. Compared to the operating values in April 1995 at NIST, the lamp currents for the sphere source at OSC in 1997 agreed within -t-1 mA. The lamp voltages were the same within -t-10 mV, except for lamp 10, which varied by up to -t-150 mV during the 1995 calibration of the sphere. For the January, March, and April set of measurements at OSC, the SXR was turned on and the filters and detectors were raised to 26°C using a thermoelectric element and a commercial temperature controller. Then the SXR was left in this configuration from 21 24 January, 6 7 March, and 10 11 April. SeaWiFS was turned on the day of the measurements, and generally, it was turned off during lunch. When SeaWiFS was on, there was power to the detector electronics and to the scan mirror motor. The batteries remained on trickle charge throughout. During the SeaWiFS mission, the detector electronics are turned off on the dark side of the orbit, about every 45 min. On put by each of the four power supplies, the voltage across each lamp, and the time and date of the measurement of these sphere parameters were recorded on the log sheet. Up to six separate measurements could be recorded on a single log sheet. The SXR log sheet was used to record file name, time of the measurements, and the raw signals at various intervals in the measurements of the GSFC sphere source. For each of the six channels, the three possible types of data include the SXR output voltage for measurements with the lens cap on (background), the lens cap off and the SXR black disc in place (ambient), and the lens cap off and the black disc removed (signal). This latter configuration corresponds to direct measurements of the radiance in the exit aperture. Four sets of these data could be recorded on a single log sheet. Other information on the experimental parameters was entered into the computer data file. The GSFC 746/ISIC log sheet followed the format developed at the SIRREX activities. Experimental parameters, such as the distance between the two apertures, the filename, the type of measurement, and the configuration of the 746/ISIC were recorded on the log sheets. 4.5 Measurements Each of the experiments listed in Tables 1 2 followed a general format (Table 9). The format was finalized by first developing the TP12001 through discussions, then through trial and error (on 21 January 1997). Steps 1 6 were 11

Original page 16

The1997PrelaunchRadiometricCalibrationofSeaWiFS doneat the beginningof a setof experiments.In step 8, 1, and 0 lamps in operation. A black disc 10 cm in diam- 7, OrbView-2andSeaWiFSwereturnedon, andhouse-eter was used for the SXR ambient data type and a black to de- disc 17.8 cm in diameter was used for SeaWiFS. For Seakeepingdatawererecorded.Thesedatawereused wereoperatingWiFS, background data were acquired during the Earth termineif the spacecraftandinstrument or voltswere and solar modes. The preliminary results indicated that normally.Step3,wherebackgroundcounts electronics,the agreement with the 1993 calibration coefficients was recorded,wasmeanttomonitortheinstrument suchasthe transimpedanceamplifiers,whichhavefinite within 5%, and extensive cleaning of the optics was not offsetvoltages. Thecords(step12)weresectionsof flat cord,acceptableforusein thecleanroom,heldacrossthecenterofthe exit aperture.Topositionthe cords,the aperturecover platescrewswereused thescrewswereinstalledin two pairsofopposingtappedholesaroundthecircumferenceof the exit aperture.Theintersectionof thecordsoccurred at thecenteroftheexit apertureandwasapparentin the SXR'sfieldof view. Steps9 27describemeasurementsof the spherewith 16lampsoperating;the brightestspheresettingwasalwaysmeasuredfirst. Thesubsequentmeasurementsatthe lowerlevels,however,didnot alternatebetweenthe SXR andSeaWiFS,asTable9 indicates,andtheSXRwasnot alwaysusedfirst. Rather,the moreefficientprocedure of "A-B-B-A"wasadopted,whereA andB representthe required. 4.5.3 Measurements on 23 January 1997 The objective of the measurements on 23 January was to determine the repeatability of the sphere SXR and the sphere SeaWiFS system by measuring the same illumination condition of the sphere in three separate measurements. The first warm-up interval for SeaWiFS was 1.5 h. The sphere was operated with a lamp configuration of 16, 8, 4, 1, and 0 (Table 9). SeaWiFS was operated with the standard gain and TDI setting. The first set was in the morning, and the black discs were used with both the SXR and SeaWiFS. Then, with the sphere off and the room lights off, SeaWiFS recorded background counts for the Earth and solar mode at two gain settings. SeaWiFS and the sphere were turned off during lunch, and the warm-up twoinstruments.Thefirst A-Bpaircorrespondedto one interval for both instruments after lunch was about 20 min. spheresetting,andthenextB-Apaircorrespondedto the The second and third set of measurements of the five sphere followingspheresetting.Whichinstrumentbeganthese- levels took place during the afternoon of 23 January 1997. quencedependedonwhichinstrumentwasalignedto the sphereat thattime.Thissequenceminimizedthenumber 4.5.4 Measurements on 24 January 1997 ofrealignmentsoftheSXRtothesphere.Finally,formeasurementswith the746/ISIC,thegeneralconceptlistedin Table9 wasfollowed,with the 746/ISICsubstitutingfor The objective of the measurements on 24 January was threefold: SeaWiFS;however,onlythe 16-1amplevelwasmeasured. 1) Quantify the warm-up interval for SeaWiFS; 4.5.1 Measurements on 21 January 1997 Theobjectiveon21Januarywastoverifytheoperation oftheGSFCsphereandtheSXR(Sect.4.2).Afteraligning thesphereandtheSXR,testswereperformedtodetermine whichoftheoverheadlightsin theexternalhighbayarea affectedthemeasurements(Sect.4.2).Datawereacquired with a lampoperationof 16, 8, 4, and1. A standard operatingmodefortheSXRwasselected:twosetsforthe backgrounddatatype,threesetsfortheambientdatatype, andfivesetsforthe signaldatatype. Eachsetconsisted of 10readingsoftheDMMforeachchannel,startingwith channel1 andendingwith channel6. It took about45s to acquireonesetofSXRdata. 4.5.2 Measurements on 22 January 1997 2) Perform the full calibration test for each of the 32 channels; and 3) Measure the linearity of SeaWiFS using the SXR as the reference instrument. These latter data, however, were lost, either from human error or a computer malfunction; therefore, the linearity test was repeated on 11 April. For the SeaWiFS warm-up test, the SXR was aligned to view the center of the exit aperture of the sphere at an angle of about 30 ° from the sphere SeaWiFS optical axis. Lamps 3 and 4 in the sphere were turned on and allowed to stabilize for 50 min [the designations for the lamps are given in Early and Johnson (1997)]. Next, a 20min SXR continuous measurement sequence was initiated, with measurements at all six channels. At the same time, SeaWiFS was turned on and measurements were made for about 20 min. Finally, both lamps in the sphere were turned off. Theobjectiveofthemeasurementson22Januarywas For the full SeaWiFS calibration, the sphere was operto determinepreliminaryvaluesfor the calibrationcoef- ated with a lamp configuration of 16, 8, 4, 1, and 0. The ficientsof SeaWiFS.Afterthe inspectionandcleaningof measurement sequence began with SeaWiFS and the black the SeaWiFSoptics(Sect.4.2),the SeaWiFSalignmentdiscs were used to record ambient signals for both Seawasverifiedasdescribedabove(Sect.4.3).SeaWiFSwas WiFS and the SXR. At each sphere level, SeaWiFS was operatedin thestandardgainandTDI combination,g 1 operated at all four gain settings and at the five TDI comand the 4:1 TDI setting. The sphere was operated with 16, binations. The measurements, therefore, included a repeat 12

Original page 17

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table9. GeneralformatoftheprocedureformeasurementsoftheGSFCspheresourceat OSCusingtheSXR or SeaWiFSduring1997. Step Data Type Procedure Align the GSFC sphere and secure its position. Align the SXR to the sphere and mark the location on the floor. Align the second tripod and its black disc to the optical axis and mark location on floor. 4 Move the SXR and its tripod out of the way. 5 Align SeaWiFS to the sphere and secure its position. 6 Remove the dust cover from SeaWiFS. 7 Turn on SeaWiFS; initialize electronics and software. 8 Remove the sphere exit aperture cover and turn off all of the room lights. 9 Move the SXR with its tripod to the measurement location to view the sphere. 10 Background Record the SXR or SeaWiFS (SXR) or turned off (SeaWiFS). 11 Remove the SXR lens cap. output with all sources of radiant flux blocked 12 Verify the alignment using two cords stretched across the exit aperture. 13 Install the SXR disc onto the second tripod. 14 Verify the alignment of the disc using a narrow visible object held across the disc center in the horizontal and vertical directions. 15 Replace the eyepiece cap on the SXR. 16 Turn on the sphere power supplies 1 4, for all 16 lamps operating. 17 Allow sphere to warm up at least 15 min. 18 Ambient Measure the spectral radiance of the black disc with the SXR. 19 Record the sphere lamp currents and voltages. 20 Remove the SXR disc and the lower head of the second tripod. 21 Signal Measure the spectral radiance of the sphere with the SXR. 22 Move the SXR out of the way. 23 Install the black disc for SeaWiFS on the second tripod; adjust the height. 24 Ambient Measure the spectral radiance of the black disc with SeaWiFS. 25 Remove the SeaWiFS disc and lower head of the second tripod. 26 Signal Measure the spectral radiance of the sphere with SeaWiFS. 27 Record the sphere lamp currents and voltages. 28 Turn off power supply 4 and 3, leaving 8 lamps operating. 29 Wait 5 min and continue measurements with 8, 4, 1, and 0 lamps. 30 Record the SeaWiFS background counts and then turn off SeaWiFS. of the standard gain and TDI setting that was measured on 23 January. 4.5.5 Measurements on 6-7 March 1997 The objective of the measurements of the GSFC sphere with the 746/ISIC on 6 7 March was to determine the spectral radiance of the source as a function of wavelength. This information was required to quantify the wavelengthdependent degradation in the spectral radiance (observed at the six SXR wavelengths) in regions other than those of the SXR measurement wavelengths. The measurements took place in the OSC clean room. SeaWiFS and OrbView- 2 were outside the room undergoing vibration testing. On the morning of 6 March, the SXR and the 746/ISIC were turned on and aligned to the GSFC sphere. The wavelength calibration of the 746/ISIC was measured using a helium emission-line source. In the afternoon, standard irradiance lamp F182 was used to calibrate the 746/ISIC, which then measured the GSFC sphere with 16 lamps operating. An ambient measurement of the sphere was made using the 17.8 cm diameter black disc centered on the optical axis between the sphere source and the 746/ISIC. Then, the lamp was used a second time to calibrate the 746/ISIC. The 746/ISIC was moved from in front of the sphere and additional measurements of mercury and neon emissionline sources were performed. During this time, the SXR was used to measure the sphere. On 7 March 1997, the measurements of the sphere with 16 lamps operating using the 746/ISIC and the SXR were repeated. First, the SXR measured the output of the sphere; then the 746/ISIC was aligned to the sphere and lamp F391 was used to calibrate the 746/ISIC. Next, the 746/ISIC measured the GSFC sphere, and then F391 was 13

Original page 18

The1997PrelaunchRadiometricCalibrationofSeaWiFS usedagainto calibratethe 746/ISIC. Ambient measurements, with the direct beam blocked, were taken for both the measurements of the standard lamp and the sphere source. After this, the SXR was aligned to the sphere and a second measurement set was acquired. Finally, the sphere was turned off and several additional measurements with the 746/ISIC and the mercury and neon emission-line sources were made. 4.5.6 Measurements on 11 April 1997 The objective of the measurements on 11 April was to determine if the calibration coefficients or other measurable parameters (offset counts) had changed during the post-thermal vacuum vibration test of OrbView-2. Beginning with the raising of the sphere to its maximum height, the alignment procedures followed the steps taken on 21 January (Sect. 4.5.1), which are explained in Sect. 4.2. The partial occultation method was used to optimize the alignment of SeaWiFS to the exit aperture of the sphere. The sphere source was operated with a lamp configuration of 16, 8, 4, 1, and 0. As on 24 January, SeaWiFS made measurements at all four gain settings and at five TDI settings. By the time the measurements began, SeaWiFS had been on for 4h and the SXR had been on for over 12h. The black disc was used to record the ambient signal for both SeaWiFS and the SXR. Only one sequence of the various sphere levels was performed, starting with the SXR. For the measurement of the SeaWiFS linearity, the SXR was relocated to the same off-axis position used during the warm-up test on 24 January. The sphere was operated at 16, 15, 14, ... 2, 1, and 0 lamps. The SXR recorded one data set (one set with the average of 10 readings for each channel) at each sphere level and SeaWiFS recorded data for 20s. SeaWiFS was operated at the 4:1 TDI setting and measurements were made of each sphere level at gain 1 and gain 3. 5. SXR AND 746/ISIC RESULTS The details of the analysis of the SXR and the 746/ISIC measurements of the GSFC sphere at OSC are described in this section. The analysis is necessary because, based on the SXR as a transfer radiometer, the GSFC sphere did not remain stable between April 1995 and January 1997, so the calibration by NIST in April 1995 was not valid in 1997. After the SXR and the 746/ISIC results are described, the method used to determine the spectral radiance of the sphere source for the calibration of SeaWiFS at OSC is explained. 5.1 SXR Measurements The SXR was used to measure the GSFC sphere source in conjunction with the calibration of this source at NIST in April of 1995 (Early and Johnson 1997). The agreement between the NIST calibration, which was based on 14 a standard lamp of spectral radiance and the SXR, was 1.2% or better (Table 6 in Early and Johnson 1997). As a function of the SXR measurement wavelength, the difference was very similar for each of the four lamp levels. The two largest differences were -1.1% at 442 nm, with the SXR giving a smaller value than the NIST calibration, and 1.2% at 548 nm, with the SXR giving a larger value than the NIST calibration. The spectral radiance of the GSFC sphere source measured by the SXR in 1995 is given in Table 10. The SXR measurements were corrected for the SSE using the model based on the measurements of the point spread response function (Johnson et al. 1998). The measurement wavelengths and calibration coefficients used to reduce the SXR data for this 1995 study coincided with those used in earlier publications (Mueller et al. 1996 and Johnson et al. 1996). The average of the SXR spectral radiance measurements of the GSFC sphere at OSC from 21 January to 11 April 1997 are also given in Table 10. As explained above, the black disc was used to record the excess signal arising from the SSE in the SXR. In addition, revised SXR effective wavelengths and calibration coefficients (Table 6) were used to reduce the 1997 values. Therefore, the 1997 spectral radiances normalized by the 1995 values include differences resulting from these changes in the SXR measurement wavelengths and calibration coefficients. The revised SXR measurement wavelengths were different from those used in 1995, because the definition in (2) was used instead of the previous definition, which was analogous to (1). The revised calibration coefficients were different, because the model used to interpolate the spectral radiance of the SXR calibration source was updated to one that gave smaller residuals. These changes are within the SXR uncertainties. The GSFC sphere spectral radiance measured by the SXR, normalized to the average value, from 21 January to 11 April are plotted in Fig. 1 for the 16 and 1 lamp levels. Because there is no obvious trend in these data, the average of all of the SXR results was determined. The standard deviation of the measurements at each wavelength and lamp level is one component in the SXR uncertainty budget (Tsai and Johnson 1998). Because the ratios of the SXR-measured spectral radiances in Table 10 depend on the method used to assign the measurement wavelengths and calibration coefficients, the ratio of the net signals from the SXR were determined. The average net signal, corrected for SSE, for each SXR channel determined in 1997 at OSC, was divided by the corresponding value from the 1995 NIST calibration experiment. These ratios are plotted in Fig. 2a. The difference in the ratio of the net voltages to the ratios of the spectral radiances for the 1995 and 1997 SXR measurements agree to within -t-0.5%, except for channel 2, where the two methods of determining the ratio differ by 1%. Figure 2a indicates that for the 16, 8, and 4 lamp configuration, the relative decrease in the radiance of the GSFC sphere was the same at each SXR wavelength, with

Original page 19

B.C. ,Johnson, E.A. Early, R.E. Eplee, 1.006 - a) 1.004. 1.002 • o[] 1.000. v _ fit 0.998 • 0 ¢ + N [] ¢ °,._ 0.996. © Z 0.994. [] 0.992. I I I I 400 450 500 550 Wavelength [nm] 1.006 - b) 1.004. 0 °,.._ V V V 1.002. <> <> V zx 1.000. A A C13 [] 0.998. t_ °_ [] 0.996. © @ Z ,Jr., R.A. Barnes, and R.T. Caffrey °zx o4_ lg [] ,¢ [] January21 [] January24 0 January 22 _ March 6 /k January23A _ March7A V January 23B _ March7B Jalmary23C @ Aprilll I I I I I 600 650 700 750 800 <> 3 @ [] @ [] [] January21 V January23B [] January24 0.994. 0 January 22 _ January 23C @ Aprilll /X January 23 A 0.992. I I I I 400 450 500 550 Wavelength [nm] Fig. 1. Spectral radiances of the GSFC integrating I I I I I 600 650 700 750 800 sphere measured at OSC in Germantown, Maryland using the SXR. The results are normalized by the average values at each SXR measurement wavelength. Each symbol corresponds to a different day, as indicated in the legend. The relative standard deviations of each SXR measurement are illustrated using vertical lines only because these values are larger than, or comparable to, the size of the plotted symbols: radiances for the 1 lamp configuration. a) radiances for the 16 lamp configuration, and b) 15

Original page 20

The 1997 Prelaunch I{adiometric Calibration of SeaWiFS 1.02 - a) l.Ol tt'3 1.00 0.99 - [--, 0.98 - 0.97 - © > 0.96 - v Z 0.95 - X 0.94 - 0.93 - 0.92 - I I I I 400 450 500 550 Wavelength [nm] 1.01 b) 1.00 tt') "'* 0.99n_g Z 0.98- [] t"-- 0 0.97- [ [] mm ---- 0.96- [] [] t'-- 0.95- 0.94 I I I 400 500 600 Wavelength [nm] [] 16 Lamps O 8 Lamps A 4 l,amps V 1 Lamp I I I I I 600 650 7OO 75O 800 [] [] [] B_ _mmB [] I I I 700 800 900 Fig. 2. The average net signals measured by the SXI{ and the 746/ISIC in 1997 for the GSFC integrating sphere at the different lamp settings normalized by the corresponding value from the NIST 1995 calibration of the sphere source. The vertical lines represent, for the 16 lamp configuration, the combined standard uncertainty for these ratios: a) measurements by the SXI{, and b) measurements by the 746/ISIC (16 lamps). 16

Original page 21

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table10. A comparisonoftheSXRspectralradiancesmeasuredfortheGSFCspheresourcein 1995and1997, andthe ratioofthesevalues.TheSXRspectralradiancesarein unitsof mWcm 2#m i sr 1. ASXR 16 Lamps 8 Lamps 4 Lamps 1 Lamp [nm] 1995 1997 97:95 1995 1997 97:95 1995 1997 97:95 1995 1997 97:95 411.2 3.853 3.646 0.946 1.848 1.746 0.945 0.9188 0.8690 0.946 0.2349 0.2202 0.937 441.5 6.317 6.074 0.962 3.040 2.920 0.961 1.512 1.454 0.962 0.3865 0.3689 0.954 487.0 11.22 10.84 0.966 5.422 5.244 0.967 2.702 2.614 0.967 0.6899 0.6631 0.961 547.9 19.08 18.57 0.973 9.269 9.034 0.975 4.626 4.511 0.975 1.180 1.143 0.969 661.7 32.81 32.40 0.988 16.05 15.88 0.989 8.026 7.947 0.990 2.040 2.009 0.985 774.8 42.36 42.60 1.006 20.82 20.97 1.007 10.42 10.51 1.009 2.642 2.651 1.003 the largest change observed at the shortest measurement wavelength. For the single lamp configuration, the spectral trend is the same, but the magnitude of the change is greater. Ancillary SXR data on the GSFC sphere, acquired at SIRREX-4 in May 1995, and during a special test in December 1995, were compared to the 1997 values and correlated with lamp operating hours and the total elapsed time. The observed change in the GSFC sphere depends on the total elapsed time, and not the total lamp operating hours. This indicates that a temporal change in the reflectance of the barium sulfate coating, which is a known of sources of uncertainty is from 1.4_?.4% (Fig. 2a). effect, is the primary reason for the change in the output the GSFC sphere between 1995 and 1997. The difference observed for the 16, 8, and 4 lamp configuration and the single lamp configuration may be due to the change in the color temperature of lamp 1, which is used for more hours than the other 15 lamps in the sphere. The uncertainties shown in Fig. 2a reflect the use of the SXR as a transfer radiometer for measurements of the same source separated in time. There are four components to the uncertainty: 1) The stability of the SXR between the measurements of the GSFC sphere source; 2) The measurement uncertainty in 1995; 3) The measurement uncertainty in 1997; and 4) The uncertainty in the correction for the SSE. To estimate the stability of the SXR from 1995 to 1997, the calibration coefficients that were used to reduce the 1997 OSC measurements (determined from the SXR calibration in September 1994) were compared to new values (preliminary results determined from the SXR calibration in the fall of 1996). For channels 1 and 3 5, the values agreed to within +0.7%; for channels 2 and 6, the agreement was within -1.7% and 1.3%, respectively. The standard deviations of the 1995 SXR measurements were not accessible for this work, so the SXR measurement uncertainties from the second set of measurements on 23 January 1997 at OSC were used to represent the SXR measurement uncertainties the day the sphere was calibrated at NIST (24 April 1995). The standard deviations of all of the SXR results at OSC (Fig. 1) were used 1997. These standard deviations, therefore, include the repeatabilities of the measurements for the GSFC sphere and the SXR, i.e., for the source detector system. The uncertainty associated with the correction for the SSE was determined using the larger of either the estimated uncertainty in the correction from the model (,Johnson et al. 1998) or the difference between the results predicted from the model and the results measured with the black disc. These uncertainties are from 0.1 0.3%, depending on the SXR measurement channel (Tsai and ,Johnson 1998). The combined relative standard uncertainty from these four 5.2 746/ISIC Measurements The spectral scans of the gas discharge lamps were used to estimate the accuracy of the wavelength scale of the 746/ISIC. The 692.947, 724.517, and 966.542nm lines in neon, and the 435.84, 546.07, and 1013.975nm lines in mercury were measured, along with the helium-neon laser line at 632.82nm. The center wavelength for each line shape was determined using the mean of the wavelengths for which the signal was 10% of the maximum value. The difference in these center wavelengths, plus the wavelength reading of the 746/ISIC, was fit to a second order polynomial and then used to correct the 746/ISIC measurements. The correction to the 746/ISIC wavelength scale, which read low, was between 0.52nm at 400nm and 0.89nm at 850 nm. For the measurements of the sphere, the net signals were determined using the scans of the sphere aperture with the black disc in place, and for the lamp, the net signals were determined with the direct beam blocked from the lamp to the 746/ISIC. The spectral radiances for the 6 March measurements with F182, where the lamp was measured before and after the sphere, were determined using the second lamp scan only, because the signals from the two lamp scans differed by about 0.5% and the second lamp scan was deemed more reliable (J. Cooper pers. comm.). The spectral radiances for the 7 March measurements using F391, where the lamp was measured before and after the two measurements of the sphere, resulted in consistent signals, so the average lamp signal from the three lamp to represent the measurement uncertainty for the SXR in scans and the average sphere signal were determined. 17

Original page 22

The1997PrelaunchRadiometricCalibrationofSeaWiFS 1.06- 1.04. 1.02 q') " 1.00 p.. 0.98 C...) 0.96 0.94. ! / 0.92 0.90 ' I ' I ' I ' _El - - I ' I ' I ' I ' I ' I 300 400 500 600 700 800 900 1000 1100 1200 Wavelength Inm] Fig. 3. Measurements of the relative change in the spectral radiance of the GSFC sphere source as measured in 1997 and normalized by the 1995 measurements at NIST. The SXR and 746/ISIC normalized results are plotted using the same symbols as in Fig. 2. The solid line is a third order polynomial fit for the 16, 8, and 4 lamp configuration, and the dashed line is a third order polynomial fit for the 1 lamp configuration. The vertical lines represent the combined uncertainties (Fig. 2), except for the two points at 1,000nm and 1,100nm. The 6 March 746/ISIC measured radiances were not consistent with the 7 March values; the 7 March values were about 4% greater over the spectral range where the measurements could be compared (400 900nm). A check of the records for F391 revealed more than 50 h of operation since its calibration. For this reason, it was concluded that F391 could not be used to reliably calibrate 746/ISIC. This is unfortunate, because F182 was only calibrated to 900nm. The 6 March 746/ISIC data are presented in Fig. 2b as the ratio of the 6 March 1997 746/ISIC measurements of the GSFC sphere source with 16 lamps operating, normalized by the April 1995 NIST calibration (Early and Johnson 1997). It is possible to directly compare these results because both were measured with a 10nm wavelength interval, nearly the same bandwidth, and at the same wavelengths. 5.3 Source Spectral Radiance Comparison of Figs. 2a and 2b indicates that the sphere radiance for wavelengths beyond about 750 nm did not de- 18 crease significantly between April 1995 and the measurements in 1997. In the spectral region from 400M50nm, however, the ratio of the SXR net voltages and the rario of the spectral radiances determined using the spectroradiometers indicate the sphere radiance decreased by _6%, with the 746/ISIC predicting the smaller change. These ratios, which indicate the change in the radiance of the sphere as measured using the SXR and the 746/ISIC, are plotted together in Fig. 3. This change is consistent with the aging of the barium sulfate coating; Johnson et al. (1996) report the measurements of the GSFC sphere before and after it was recoated in April 1994. The ratio of net signals is the correct factor to use for the SXR comparison of the 1995 and 1997 sphere radiance because this removes any sensitivity to the assigned values for the SXR calibration coefficients or measurement wavelengths. The SXR has demonstrated good stability, including transportation to a remote facility; measurements of a NIST sphere source before and after shipment to Japan resulted in an average difference of 0.05% (Johnson et al. 1997).

Original page 23

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey to the sphereand The ratioof spectralradianceassigned sourceis the correctfactorto usefor the 746/ISICmea- L_,97(A) F_ (A)Ln,95 (A), (4) surements.The746/ISICcannotbeusedtodeterminesignalratiosbetweenthe 1995calibrationandthe 1997Sea- where n corresponds to the number of lamps operating (the WiFScalibrationbecausemeasurementswerenotmadein 16, 8, 4 or 1 lamp configuration). The coefficients for (3) 1995usingthe746/ISIC.In addition,the 746/ISICis not are given in Table 11. The results for L_,97(A) are shown stable;it requirescalibrationwheneverit is used,sothe in Fig. 4. resultsarealwaysexpressedin radiometricunits. The uncertainties in the spectral radiance determined Thecorrectionfactorsweredeterminedbyusingonlya using the 1995 NIST calibration values and (4) were depart of themeasurementsfromthe twodeterminationsof termined by summing in quadrature the uncertainties in the relativechangein thesphereradiance.Themostre- the ratios used in the fit (Fig. 3) and the uncertainties in liabledeterminationderivesfromtheSXRmeasurements,the NIST calibration (Early and Johnson 1997). The unbut theresultsatthesLxSXRwavelengthsareinsufficient to determinethe relativechangein the sphereradiance overthe requiredspectralregionof 3801,100nm.The longestmeasurementwavelengthof the SXRis 775nm. The746/ISICmeasurementsaresensitivetoscatteredlight belowabout450nm (Johnsonetal. 1997),andtherelative standarduncertaintyiscomparableto theSXR-determined relativechangein thespherebelowabout550nm.Toestimatethe correctionfor wavelengthsgreaterthan775nm, twovaluesfromthe 746/ISCmeasurementsofthe16lamp configurationwereincluded,at 830and900nm(Fig.3). certainties in the ratios were estimated every 10 nm from 380 1,100 nm using a smoothing interpolation function in a standard data analysis package (KaleidaGraph from Synergy Software). The input values consisted of the uncertainties in the ratios at the sLx SXR wavelengths, as well as the values at 830, 900, 1,000, and 1,100nm. The uncertainties in the spectral radiance of the GSFC sphere at OSC are about 1.6% for band 1, less than 1% for bands 2 6, and 1.2% for bands 8. There is a weak dependence on lamp configuration (Fig. 5). the de- As a verification of the spectral radiance determined Thelinesin Fig.3 arethe resultof modeling asa third orderpolynomial.for the GSFC sphere using the method described above, a creasein thesphereradiance rel- cubic spline interpolation was performed at the SXR mea- Twofits wereperformed,oneto representtheaverage forthe surement wavelengths (Table 6) in these radiances (on a ativechangefromApril 1995to thespringof 1997 GSFCsphereradiancescorrespondingto 16,8,or 4 lamps 10nm grid). The results were compared to the average operating,andthesecondforthe1lampconfiguration.In spectral radiances determined with the SXR in 1997 at orderto extendthepolynomialfit to 1,100nmwithoutin- OSC (Table 10). The agreement was good, with average troducinganynonphysicalbehavior,theratiowasassigneddifferences of 0.9, 0.2, 0.7, 0.8, -0.7, and 1.3% for SXR to beunityat 1,000and1,100nm,with arelativestandardchannels 1 6, respectively. uncertaintyof5%.Thisis areasonableapproximation,becausethe reflectanceof the spherecoatingis expectedto changemorerapidlyat shorterwavelengthsratherthanat longerwavelengths,andsuchchangesarenot expectedto resultin increasedsphereradiances. Thecoefficientsforthetwofits, a0, al, a2, and a3, are given in Table 11. For the 16, 8, and 4 lamp configuration, the 18 results from the SXR were used, along with the two results from the 746/ISIC measurements and the two forced values. For the 1 lamp configuration, the six SXR results were used, along with the two results from the 746/ISIC measurements, even though they were for the 16 lamp configuration. The two forced values were also included. This was necessary to produce a reasonable fit over the desired spectral range of 380 1,100nm. For each fit, the uncertainties indicated by the vertical lines in Fig. 3 were used to weight the data. These polynomials represent the correction factor Fn(A), for the NIST 1995 spectral radiances used to determine the spectral radiance of the GSFC sphere source at OSC in 1997. The NIST 1995 spectral radiances are denoted Ln,95(A) and the final values used at OSC for the calibration of SeaWiFS in 1997 are denoted Ln,97(A): fn(k) ao @ alk @ a2k 2 @ a3 A3 (3) 5.4 Spectral Radiances for SeaWiFS The spectral radiances L,97(A) from (4) are given at the same wavelengths that were measured during the NIST calibration of the GSFC sphere source in April 1995, i.e., from 380 1,100nm every 10nm. Two methods of using these values to determine the SeaWiFS calibration coefficients are described. The first method is required to compare the results at OSC to the results obtained by the instrument manufacturer for their November 1993 measurements (Barnes et al. 1994b). For that work, the calibration coefficients were determined using the net counts from SeaWiFS and the corresponding spectral radiance at the nominal center wavelengths, AD (Table 4). To determine the spectral radiance at these wavelengths, a cubic spline interpolation in the Ln,97(A) values was used. The results are given in Table 12, along with the corresponding value for the standard uncertainty in these values, u_ (L(AD)) (Fig. 5). After the initial calibration of SeaWiFS in 1993 (at SBRC) was completed, the SeaWiFS Project concluded that band-averaged spectral radiances provided a better representation of the measurements (Barnes and Yeh 1996). 19

Original page 24

The 1997 Prelaunch Radiometric 50 4O 'E 30 (-q [] [] [] [] .................... EJ ....................................................... 20 [] [] :k [] [] [] i0 ' I ' I ' I ' I ' 300 "400 500 600 700 Wavelength Fig. 4. Spectral radiance of the GSFC sphere source Calibration of SeaWiFS 0.50 0.45 0.40 0.35 :z 0.30 0.25 ¢zz. 0.20 0.15 0.10 0.05 , 0.00 I ' I ' I ' I 800 900 1000 1100 1200 [nm] for the recalibration of SeaWiFS at OSC in 1997. The spectral radiance with 16 lamps operating is indicated with the open squares; the values correspond to the left ordinate. The spectral radiance with the 8, 4, and 1 lamp operation is illustrated using the solid, dot, and dash-dot-dot lines. These spectral radiances were normalized by the values for the 16 lamp configuration and these ratios correspond to the right ordinate. 6 35 .3 _4 _3 _2 °i _0 ,I ! /' 3(10 4O(1i 5 )0i 600i 7 _)(1 800 900 1000 1200 Wavelength [nm] Fig. 5. Relative standard uncertainties, expressed i i i 11'00 i in percent, of the spectral radiance of the GSFC sphere source at OSC in 1997 for the calibration of SeaWiFS. The solid line corresponds to the 16 lamp configuration, and the dash-dot line corresponds to the 1 lamp configuration. 2O

Original page 25

B.C. Johnson, E.A. Early, R.E. Eplee, Jr., R.A. Barnes, and R.T. Caffrey Table 11. Polynomial coefficients for the correction factors used to determine the spectral radiance of the GSFC sphere at OSC in 1997 from the NIST 1995 calibration values. Two sets of values were determined, one for the sphere levels corresponding to 16, 8, and 4 lamps operating, and the other for the sphere levels corresponding to 1 lamp operating. Coefficient Val ll es Unit 16, 8, 4 Lamps ao 0.630958 al 1.22001× 10 3 a2 -1.33729×10 6 g3 4.83255×10 lO Table 12. Spectral radiances and relative standard 1 Lamp 0.614468 Dimensionless 1.26463×10 3 nm 1 -1.40356×10 a nm 2 5.22699×10 10 nm a uncertainties, uS(As) un(L(AD))/Ln,97(AD) at the SeaWiFS nominal center wavelengths for the GSFC sphere source at OSC. The units of spectral radiance are W cm 2 sr 1 nm 1 and the relative standard uncertainty AD 16 Lamps 8 Lamps is in percent. 4 Lamps 1 Lamp [nm] L16,97(AD) uD6(AD) Ls,97(AD) uD(AD) L4,97(AD) u4D(AD) L1,97(AD) uD(AD) 412 3.668 1.61 1.760 1.61 0.8740 1.61 0.2201 1.72 443 6.188 0.98 2.981 0.99 490 11.10 0.61 5.376 0.61 510 13.52 0.57 6.559 0.58 555 19.33 0.59 9.416 0.59 670 33.51 0.78 16.42 0.78 1.482 0.99 0.3754 1.10 2.679 0.61 0.6800 0.71 3.270 0.58 0.8300 0.67 4.700 0.59 1.193 0.66 8.215 0.78 2.083 0.81 765 41.49 1.16 20.38 1.16 10.20 1.16 2.580 1.18 865 44.96 1.35 22.15 1.35 11.10 1.35 2.800 1.35 Table 13. Band-averaged radiances and relative standard uncertainties, uB(AB) un(LB(AB))/Ln,B,97(AB). The units of band-averaged radiance are #W cm 2 sr i nm 1. The values of AB are for the 16 lamp configuration. They vary by less than 0.05 nm for the three other lamp configurations. AB 16 Lamps 8 Lamps 4 Lamps 1 Lamp [nm] L16,B,97 uB6(AB) LS,B,97 uB(AB) L4,B,97 u4B(AB) L1,B,97 uB(AB) 415.55 3.837 1.86 1.842 1.87 0.9146 1.87 0.2301 1.98 444.79 6.303 0.98 3.037 0.99 1.511 0.99 0.3821 1.09 493.08 11.32 0.85 5.482 0.86 2.732 0.86 0.6927 0.96 511.02 13.55 0.58 6.573 0.58 3.277 0.58 0.8309 0.67 558.16 19.42 0.77 9.460 0.78 4.722 0.78 1.197 0.85 668.80 33.30 0.87 16.32 0.87 8.163 0.87 2.068 0.90 767.51 41.50 1.18 20.39 1.18 10.21 1.18 2.579 1.20 865.05 44.69 1.65 22.01 1.65 11.03 1.65 2.782 1.65 The band-averaged radiance is defined (Barnes 1996c) as /3°° L,97(A)R(A)dA 8O Ln,B,97 (5) RAdA 1°° ( ) ' 0 where /g(A) is the spectral responsivity for one SeaWiFS band (Barnes et al. 1994) and Ln,97(A) is the spectral radiance of the source. Note that the units of L_,B,97 are the same as those for spectral radiance, but this is due to the choice of normalization; the physical interpretation of L_,B,97 corresponds to radiance. The band-averaged spectral radiances calculated from (4) and (5) are given in Table 13. The calculation of the uncertainties in these values, u_(LB) , is described in Sect. 7. For the laboratory measurements, it is not necessary to associate a value for AB with Ln,B,97. For in-orbit measurements, it may be necessary to define AB for the algorithms that convert the SeaWiFS measurements to geophysical quantities. 6. SeaWiFS RESULTS The methods used to reduce and analyze the SeaWiFS data are described in this section. The emphasis is on the measurements made on 24 ,January and 11 April 1997 in 21

Original page 26

The1997PrelaunchRadiometricCalibrationofSeaWiFS the 1:1TDI setting;thesedataareusedto derivethecal- The quantity of interest to the SeaWiFS science team is ibrationcoefficientsfor SeaWiFS.Similarmethodswere the band-averaged radiance from orbit. From (9) and (5), usedto analyzethe SeaWiFSdatafromthe othermeasurements(Tables1 3). 6.1 Measurement Equation Thebasicmeasurementequation,in thelaboratoryduringtheprelaunchcalibration,or fromorbit,is 1 _ /1100 Sx(g,c) 7x(g,c)J38 ° L_(A) R(A)dA, (6) where S_ (g, c) are the corrected net signals that depend on the measurement condition (x "lab" or "orbit"), g is the gain, and c is the channel. L_(A) is the spectral radiance of the source for that measurement condition, and _ (g, c) is a constant that depends on the measurement conditions and instrument parameters. The limits of the integral correspond to the spectral region where the instrument has sensitivity. The center wavelength method used at SBRC approximates the integral in (6), effectively assuming that the functions R(A) for SeaWiFS are very narrow. Then, the measurement equation becomes Slab(g , C) Ln,97(AD) R(AD) AA _/lab (g, C) ' (7) where Sl_b (g, c) are the corrected net counts from SeaWiFS for the measurements of the GSFC sphere source (Sect. 6.2) and AA is a bandwidth parameter. For the purpose of comparing the OSC results to the SBRC results, the calibration coefficient K2(g, c) is Zl U(g, K (g, = R(AD)ZXA (S) Sla b (g, C) " The center wavelength method does not allow for the finite bandwidth of R(A) and the fact that the relative spectral shape of the source depends on the measurement condition. For example, (Barnes 1996a) estimated that the GSFC sphere source, as measured at NIST in April 1995, had the same relative spectral shape as a 2,850 K blackbody. The relative spectral shape for the upwelling Earth a al. (1997) described a method for correcting the measureradiance, however, was modeled in that document by 12,000 38,000 K blackbody. The finite bandwidths and the large differences in the relative spectral shape of the source used to calibrate SeaWiFS, compared to the source that it on coefficient, K2 (g, c), and several instrument factors are inmeasures from orbit, require that the analysis be based (6). The ratio of the corrected signals for the two measurement conditions is [1100 Lorbit (A) R(A) dA (9) and the corrected net signals is described. The presenta- Sorbic(g, Slab(g , C) "_orbit(g, C) [1100 Llab(A) R(A) dA d 380 22 _i 1°° Lorult(A) R(A) dA 0 Sorult(9, 7orU,(9, /lab(g,C) (10) _i _°°Ln,97(A) R(A) dA 0 x Slab (9, C) _(A) dA _31 I00 0 from which ooLn,97(A) R(A) dA 80 K2(9,c) Sl U(9, I°°oR(A) dA (11) Ln,B,97 Sl_b (9, C)" Either method (band-centered or band-averaged) results in a calibration coefficient with units of spectral radiance per count. The advantage of using the band-averaged radiances, both for the laboratory and on-orbit measurements, is that no error is introduced by approximating the integral, as is the case for the band-centered method [compare (9) with (8) and (10)]. If the R(A) functions are accurate, the band-averaged radiance and the signal will change by the same relative amount for sources with different spectral shapes; then K2 (g, c) for the band-averaged method is independent of the source spectral shape. For the SeaWiFS recalibration, measurements are expressed in terms of the ratio of the net counts obtained during the calibration to those obtained during an experiment. Also identified, through the ratio of the constants gx (g, c), are factors other than relative spectral shape that may be different between the calibration laboratory and the final measurement. Instrument factors that can be modeled using this approach include sensitivity to polarization, the spatial extent and variation of the source, or departures of gain ratios from nominal values. For example, Yeh et ments of oceans for the excess signal caused by nearby bright sources, such as clouds. For the general SeaWiFS algorithm, the gains are incorporated into the calibration cluded in determining the corrected signals (Sect. 6.2). 6.2 Corrected Net Signal In this section, the relationship between the raw counts tion follows from the prelaunch radiometric characterization and calibration equations (Barnes et al. 1994b).

Original page 27

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey 6.2.1 Model ThecorrectednetsignalsSx (g, c) are given by Sx(g, c) (Cout - Cdark) [1 + K3(T - Tref)] (12) × K4(PxZ) R_cG(_ - _0) + co. In (12) above Cout is the value of the total output counts from the sensor in digital counts; Cdark is the value of the dark counts (or zero offsets) from the sensor in digital counts; K3 is the correction factor for the temperature dependence of the photodiode responsivity in units of K 1; T is the focal plane temperature in Kelvin; Tref is a reference temperature, Tref 293 K; K4 is a (dimensionless) correction factor for the scan modulation; Pxl is the value of the pixel (dimensionless) along the scan line; /_i is a dimensionless correction factor for the difference in reflectance between the two sides of the scan mirror; / is the dimensionless value (I or 2) to denote the mirror side; CG is the time-dependent overall system gain in per unit time; t is the time of the SeaWiFS measurement; to is the time of the SeaWiFS launch; and CO is a dimensionless calibration system offset. The values for Cdark and Ri that are necessary to determine Slab(g, C) were determined using the data acquired at OSC. The details of the calculations are given below. The time-dependent overall system gain and offset, CG and CO, respectively, allow for adjustments to the laboratory calibration as a function of time on orbit. The overall system gain or offset are expected to vary slowly with time and could be caused by degradation of the mirrors or other system components. During the mission, periodic solar or lunar measurements will be used to assess changes in these parameters. For the laboratory work reported here, CG- 1 and CO- O. The temperature correction factor, K3, was derived by Eplee and Barnes (1997) for each of the eight bands. K3 is small, from 0.00008 0.00090K 1. During the calibration of SeaWiFS at OSC, the maximum focal plane temperatures, T, were about 302 K. The resulting corrections, applied to each scan line, ranged from 0.07 0.8%. The scan modulation correction factor, K4, accounts for variations in the instrument response as a function of scan mirror position, i.e., for each pixel in a scan line. These corrections were determined during the characterization of SeaWiFS by SBRC (Barnes et al. 1994b), resulting in a quadratic model. Compared to the nadir pbcel (in the center of the scan line), this correction can be up to 1.7% at the ends of the scan. Because only two pbcels around the nadir pixel were used for the determination of Slab(g, C), K4 is unity to within -t-0.0006%, so K4 is equal to unity for the measurements reported here. 6.2.2 Measurements During the measurements at OSC, SeaWiFS viewed the GSFC sphere at different radiance levels; for the radiometric calibration, there was a lamp configuration of 16, 8, 4, and 1. On 24 January and 11 April 1997, measurements were made for all gains for both TDI settings (4:1 and 1:1). This corresponded to five measurements at each radiance level, because in the 1:1 TDI setting, the output of Sea- WiFS corresponds to one channel for each band, requiring four measurements to include all four channels. At each radiance level, the total counts, Cout, were recorded for each channel and each gain for all eight bands. Each measurement for a fixed set of parameters consisted of acquiring 120 successive scan lines, which took about 20 s. The output of the channels in bands 2 8 saturated for some of the sphere radiance levels and gain combinations. Table 14 gives the radiance level and gain combinations that did not saturate SeaWiFS in the 1:1 TDI setring. From Table 14, it is clear that the number of independent measurements of K2 (g, c) for each band, channel, and gain combinations for this experiment varied between zero and four compare band 8 (c 1 and g 1) to the corresponding values for band 1. The ocean channels were saturated at all lamp levels for band 6 at g 2, and for bands 7 and 8 at g 1 and g 2, respectively. The calibration coefficients were, therefore, calculated using the gain ratios and the calibration coefficients for the unsaturated measurements. Determination of the gain ratios is described in Sect. 6.3.1. For the analysis, only the central portion of each scan line was used. Because SeaWiFS was aligned with the exit aperture of the sphere source, counts from the nadir (0 ° scan angle) pixel corresponded to the central position in the exit aperture. The average value for the net corrected signal was determined for the five pixels centered on the nadir pixel using (12), which becomes Slab(g, C) <(Cou t - Cdark)[1 +/_3(T- Tref)]/_i> , (13) because K4(Pxl) 1, CG 1, and CO 0. For each set of measurement parameters (sphere level, band, channel, and gain), the average net corrected signal was determined for 600 measurements (5 pixels and 120 scans). For each sphere level, the final values for Slab (g, C) were determined from the average of the 24 January and 11 April measurements. 6.2.3 Offset Counts The values for Cdark were determined for each scan line using the corresponding dark count (DC) restore value 23

Original page 28

The1997PrelaunchRadiometricCalibrationofSeaWiFS Table14. Sphereradiancelevels,denotedbythenumberoflampsoperating,andSeaWiFSgain,g 14, that did not result in saturated values for Cout for the measurements at OSC on 24 January and 11 April in the 1:1 TDI configuration. Combinations that resulted in the saturation of SeaWiFS are indicated by a dash (-). Band Channel N_zmber N_zmber g 1 1 16, 8, 4, 1 2 16, 8, 4, 1 3 16, 8, 4, 1 4 16, 8, 4, 1 1 16, 8, 4, 1 2 16, 8, 4, 1 3 16, 8, 4, 1 4 16, 8, 4, 1 1 16, 8, 4, 1 2 8, 4, 1 3 8, 4, 1 4 8, 4, 1 1 8, 4, 1 2 8, 4, 1 3 8, 4, 1 4 16, 8, 4, 1 1 16, 8, 4, 1 2 4, 1 3 4, 1 4 4, 1 1 1 2 1 3 1 4 16, 8, 4, 1 1 8, 4, 1 2 3 4 1 2 3 4 8, 4, 1 (Barnes et al. 1994a). This is the procedure used at SBRC and on orbit. In order to estimate the standard uncertainty in Cdark, the mean and standard deviation of the individual DC restore values acquired during the 120 DC restore scan lines was determined. In some cases, Cdark was very stable, resulting in zero for the standard deviation. In this case, the standard deviation was assigned the value 1/x/T2, which arises from a resolution of I count and the assumption of a uniform, rectangular probability distribution function (International Organization for Standardization 1993). The average values of Cd_k for the radiometric calibration analysis are given in Table 15, along Gain Setting g 2 g 3 g 4 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 8, 4, 1 16, 8, 4, 1 8, 4, 1 8, 4, 1 16, 8, 4, 1 8, 4, 1 8, 4, 1 16, 8, 4, 1 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 4, 1 8, 4, 1 4, 1 4, 1 8, 4, 1 4, 1 4, 1 8, 4, 1 4, 1 4, 1 8, 4, 1 4, 1 4, 1 8, 4, 1 4, 1 4, 1 8, 4, 1 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 1 4, 1 1 1 4, 1 1 1 4, 1 1 4, 1 4, 1 4, 1 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 16, 8, 4, 1 16, 8, 4, 1 16, 8, 4, 1 1 1 1 1 1 1 1 1 1 1 1 1 8, 4, 1 8, 4, 1 8, 4, 1 April and 24 January agreed to within 4-0.5%. As a separate investigation, the dark counts were measured using different techniques to compare them to the DC restore values. On 24 January and 11 April, following the measurements of the sphere with one lamp illuminated, this lamp was turned off and SeaWiFS measured the dark sphere. All four gains were measured, in the 1:1 and 4:1 TDI setting. As measured by the SXR, this produced a zero radiance source. For the 32 channels in the 1:1 TDI setting, the average of the difference between the counts from the dark sphere and the corresponding DC restore value was less than 4-0.03 counts with a standard deviawith the standard uncertainty. The values for Cd_k for 11 tion of about 0.1 counts, independent of gain setting. For 24

Original page 29

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table15. Averagevaluesof Cdark , in digital counts, from 24 January and 11 April for the SeaWiFS calibration analysis. The standard uncertainties u(Cdark ) are also given in digital counts. Band Channel Gain 1 Gain 2 Gain 3 Gain 4 Number Number Cdark u(Cdark ) Cdark 1 1 21.0 0.20 21.0 2 23.2 0.13 23.5 3 18.4 0.17 16.9 4 20.9 0.11 19.0 2 1 18.5 0.18 17.4 2 21.0 0.06 19.9 3 15.8 0.13 12.4 4 18.0 0.20 18.0 3 1 21.0 0.20 21.0 2 22.1 0.12 23.0 3 20.9 0.08 19.0 4 19.0 0.07 17.1 4 1 21.4 0.17 21.4 2 19.6 0.18 18.5 3 19.3 0.17 19.9 4 21.0 0.20 21.0 5 1 26.0 0.20 26.0 2 22.3 0.16 23.4 3 22.1 0.11 15.9 4 17.1 0.09 11.9 6 1 21.7 0.17 13.2 2 17.5 0.18 11.8 3 33.1 0.15 36.3 4 21.0 0.20 21.0 7 1 23.0 0.20 23.0 2 20.2 0.13 11.0 3 21.1 0.10 22.4 4 26.6 0.18 24.4 8 1 19.0 0.13 16.3 2 25.9 0.13 29.0 3 16.2 0.15 13.3 4 20.0 0.20 20.0 the 8 bands in the 4:1 TDI setting, the average difference u(Cdark ) Cdark u(Cdark ) Cdark u(Cdark ) 0.20 21.0 0.20 21.0 0.20 0.20 23.3 0.17 23.4 0.19 0.20 18.0 0.13 17.4 0.18 0.18 20.2 0.14 19.6 0.19 0.18 18.1 0.09 17.8 0.15 0.15 20.6 0.17 20.2 0.14 0.17 14.8 0.14 13.4 0.18 0.20 18.0 0.20 18.0 0.20 0.20 21.0 0.20 21.0 0.20 0.10 22.0 0.06 22.9 0.13 0.11 21.0 0.20 19.6 0.17 0.13 19.0 0.20 17.7 0.15 0.17 21.5 0.18 21.4 0.17 0.18 20.0 0.20 18.9 0.12 0.13 19.1 0.11 19.8 0.14 0.20 21.0 0.20 21.0 0.20 0.20 26.0 0.20 26.0 0.20 0.18 22.0 0.20 23.0 0.09 0.14 24.7 0.16 18.3 0.16 0.16 19.0 0.20 14.1 0.11 0.26 27.0 0.20 24.5 0.18 0.28 21.0 0.15 19.4 0.17 0.27 31.0 0.20 32.0 0.07 0.20 21.0 0.20 21.0 0.20 0.20 23.0 0.20 23.0 0.20 0.20 27.0 0.20 24.1 0.09 0.22 20.0 0.15 20.8 0.14 0.21 28.0 0.20 27.3 0.16 0.26 21.0 0.20 20.1 0.11 0.24 23.2 0.13 24.0 0.11 0.25 18.4 0.17 17.9 0.08 0.20 20.0 0.20 20.0 0.20 0 0.5_0, with the majority of the comparisons below 0.3_0. was less than 4-0.02 counts with a standard deviation of These differences are not statistically significant given the about 0.07 counts. Also on 24 January and 11 April, the black disc was placed over the center of the exit aperture of the sphere when it was at each of the four radiance levels, and Sea- WiFS measured the radiance of this dark source. A gain setting of g 1 and the 4:1 TDI setting were used. For each lamp level, the bands that had no saturated channels (Table 14) were analyzed by comparing the black disc measurements to the values of Cdark from the DC restore measurement. The counts produced by the black disc were no more than 1.6 counts higher than the corresponding values of Cdark, and corresponded to a relative bias of from values for u(S_b). Also, it should be noted that interpretation of these results in terms of the documented stray light effects in SeaWiFS (Barnes et al. 1995) is difficult because the black disc was in the near field of the optical system, which is focused for objects at infinity. 6.2.4 Mirror Side Correction For SeaWiFS, alternate scan lines correspond to measurements using the two sides of the half-angle mirror (Barnes et al. 1994b). Because the reflectance of the two sides may differ, the mirror side correction factor must be determined before calculating S_b(g, c) from (13). This is 25

Original page 30

The1997PrelaunchRadiometricCalibrationofSeaWiFS forthealter- 6.3.1 Gain Ratios possibleby usingonlytheaveragenetcounts natescanlines,becausethe GSFCspheresourceis stable anduniform. Thenet countsforthe alternatescanlinesareproportionalto the reflectanceof the mirror,Pl or P2. Defining the average reflectance of the mirror as p (D1 + p2)/2, the reflectance of each side is (14) D1 C1 + C2 19, where Ci are the net counts, Cout - Cdark, for mirror side i 1 or 2. The average corresponded to measurements of alternate scan lines for the data acquired on 24 January and 11 April. The average reflectance p is unity because its value is incorporated into the calibration coefficients K2(g, c). Therefore, the correction factors Ri are (15) Note that the sum of the two mirror side correction factors is unity. The correction factors are given in Table 16 and indicare the reflectance of the two sides of the half-angle mirror The ratios of the variable gains of the SeaWiFS ocean channels were determined on 24 ,January, using the voltage calibration pulse. As explained in Sects. 2.3 and 4.2, the gain ratios are measured by operating SeaWiFS in the solar calibration mode and recording the response, in counts, to a constant voltage which is input to each channel's voltage amplifier. During the voltage pulse, the gain for each channel is varied from g 1 to g 4. The input to the voltage amplifier is the sum of this calibration voltage and the output of the transimpedance amplifier. Because the scan mirror observes the dark interior housing of the SeaWiFS instrument during this measurement, the voltage from the channel's preamplifier circuit is small compared to the calibration voltage, but it was accounted for with the DC restore measurement. For each channel in each band, the gain ratios were determined from the ratio of the net counts for g 2, g 3, or g 4 relative to the counts for g 1. The final values, given in Table 17, are the average of 60 pixels and 120 scan lines. The nominal value for the gain ratios for all gain settings (g 2, g 3, or g 4) for the cloud channels (channel 1 for bands 1, 3, 5, and 7, and channel 4 for bands 2, 4, 6, and 8) is unity. The nominal value for the gain ratios for the ocean channels, g 2, is the same to less than 0.1% for bands 3 8; for bands 1 is 2. The other values are given in Table 17; g 3 was and 2, the difference is 0.14 and 0.1%, respectively. These values are comparable to those reported by Barnes et al. (1994b). Table 16. Values for the mirror side correction factors determined from the measurements by Sea- WiFS of the GSFC sphere source on 24 January and 11 April. Band Number R 1 R 2 1 1.0007079 0.9992921 2 1.0005341 0.9994659 3 0.9999844 1.0000156 4 1.0002266 0.9997734 5 0.9999985 1.0000015 6 1.0000016 0.9999984 7 1.0000002 0.9999998 8 1.0000085 0.9999915 6.3 Calibration In this section, the corrected net signals (from Sect. 6.2) and the two methods of describing the radiance of the GSFC sphere nominal center wavelength or bandaveraged center wavelength (Sect. 6.1) are used to determine the calibration coefficients for SeaWiFS. Because the ocean channels for three of the bands saturated for g 1 or g 2 for all four sphere levels measured, the measured gain ratios were required before determining the calibration coefficients. 26 designed for the on-orbit lunar measurements, and g 4 was designed for the on-orbit measurements of reflected sunlight using the diffuser panel. The relative standard uncertainty in the gain ratios is also given in Table 17 and explained in Sect. 7.3. 6.3.2 Calibration Coefficients The calibration coefficients for SeaWiFS were determined for each of the 32 channels using the data acquired in the 1:1 TDI configuration. Independent determinations of the values for K2 (g, c) result from measurements of the different radiance levels of the sphere source (Table 14). The procedure used at OSC was similar to the original calibration at SBRC, except that six sphere levels were used (Barnes and Eplee 1997). The OSC measurements for gains other than g 1 can be used to verify the gain ratio values determined from the electronic measurements. As explained above, all four levels of the GSFC sphere saturated the ocean channels in band 6 for g 2 and in bands 7 and 8 for g 1 and g 2; for these 15 values of K2(g, c), therefore, this verification was not possible. In the 4:1 TDI configuration, the output for each band is the average of the output of the four channels (this is the standard mode from orbit), so the data acquired with the 4:1 TDI should agree with the weighted average of the 1:1 data, using the K2(g, c) values for the weights. Thus, the 4:1 data acquired for each sphere radiance level on 24 January and 11 April can be used to verify the values of K_(g, c) (Sect. 8.3).

Original page 31

B.C. Johnson, E.A. Early, R.E. Eplee, Table 17. Results of the determination of the gain ratios at OSC in 1997. The gain at g 2, 3, or 4 is normalized Jr., R.A. Barnes, and R.T. Caffrey from the internal voltage calibration pulse for SeaWiFS to the gain at g 1. The relative standard uncertainty in the gain ratios for g 2, 3, and 4 is given in the fifth, seventh, and last columns. Band Channel Gain 1 Gain 2 Uncertainty Number Number [] 1 1 1.0156 0.27 2 1.9865 0.20 3 1.9862 0.21 4 1.9925 0.18 2 1 1.9916 0.16 2 1.9841 0.15 3 1.9889 0.17 4 1.0242 0.46 3 1 1.0509 0.67 2 1.9887 0.17 3 1.9886 0.13 4 1.9901 0.14 4 1 1.9901 0.15 2 1.9878 0.15 3 1.9882 0.15 4 1.0435 0.82 5 1 1.0441 0.83 2 1.9908 0.17 3 1.9894 0.16 4 1.9895 0.15 6 1 1.9910 0.19 2 1.9876 0.21 3 1.9905 0.19 4 1.0333 0.63 7 1 1.0492 0.92 2 1.9879 0.19 3 1.9886 0.17 4 1.9905 0.18 8 1 1.9917 0.17 2 1.9891 0.18 3 1.9913 0.18 4 1.0411 0.78 Using the 1:1 TDI configuration, K2(g,c) was determined for each channel using the band-centered and bandaveraged methods, see (8) and (10). Because the four channels in each band have the same R(A) (Sect. 2.1), the same sphere radiance applies to each channel within a band. As an example, for g 3 and c 4 for the 4 lamp configuration, there are eight equations (one for each band) with the form: K2(3, 4) L4'97 (AD) (16) S_ab (3, 4)' Gain 3 Uncertainty Gain 4 Uncertainty [] [] 1.0051 0.29 1.0105 0.27 1.3228 0.21 1.6818 0.20 1.3201 0.21 1.6806 0.21 1.3194 0.18 1.6796 0.18 1.3192 0.16 1.6819 0.16 1.3189 0.18 1.6810 0.16 1.3172 0.18 1.6828 0.17 1.0081 0.47 1.0161 0.46 0.9999 0.66 1.0309 0.88 0.8949 0.17 1.6811 0.17 0.8968 0.13 1.6811 0.14 0.8957 0.16 1.6799 0.14 0.7869 0.23 1.6814 0.15 0.7883 0.19 1.6818 0.15 0.7882 0.16 1.6836 0.15 0.9855 0.84 1.0290 0.83 0.9853 0.86 1.0294 0.84 0.6414 0.18 1.5964 0.17 0.6431 0.25 1.5964 0.17 0.6419 0.16 1.5942 0.15 0.3630 0.24 0.6629 0.28 0.3649 0.31 0.6650 0.27 0.3642 0.26 0.6652 0.25 0.9667 0.56 0.9794 0.69 0.9666 0.89 0.9682 0.92 0.3101 0.28 0.5755 0.23 0.3130 0.32 0.5758 0.29 0.3109 0.32 0.5769 0.28 0.2603 0.46 0.4989 0.21 0.2613 0.40 0.4995 0.20 0.2600 0.48 0.4982 0.22 0.9589 0.81 0.9726 0.80 (band-centered method) or [1100 L4,97(A)/(A) dA K2(3, 4) J3s0 &ab(3, 4) 1100 R(A) dA (17) ,/380 L4,B,97 &_b(3, 4)' (band-averaged method), where R(A) or AD depends on the band. In this case, there are four measurements of K2(3, 4) for bands 1, 2, 4, and 6 from the 16, 8, 4, and 1 lamp configuration; three measurements for bands 3 and 8 27

Original page 32

The 1997 Prelaunch Radiometric Calibration of SeaWiFS from the 8, 4, and I lamp configuration; two measurements for band 5 from the 4 and 1 lamp configuration, and one measurement for band 7 from the one lamp configuration. to the smallest of the four saturation radiances, and knee 2 corresponds to the next smallest. Knee 3 occurs between knee 2 and the radiance at which the band completely Some of the band-averaged results are illustrated in saturates. At each knee, the average of the predicted net Figs. 6 and 7. Figure 6 shows the 1422(9, c) values for g 1 and their combined standard uncertainty for the single cloud channel in each band. This corresponds to the first channel in bands 1, 3, 5, and 7 and the fourth channel in bands 2, 4, 6, and 8. The independent values counts for the four channels in each band corresponds to the counts at the knee when SeaWiFS is operated in the standard 4:1 TDI configuration. For all eight bands and each of the four gain settings, the radiance and net counts at the three knees and saturation are given in Table 20. obtained for each sphere radiance level are shown; some of Figure 8 illustrates the relationship between radiance and the data for the one-lamp configuration do not agree with the results at the other sphere levels. Figure 7 illustrates the 1422(9, c) values at g 1 for one of the ocean channels in each band (channel 2). The 1422(9, c) values for bands 7 and 8 are not shown because these were not measured directly. The agreement among the calibration coefficients from the multiple sphere levels is good. For the calculation of the K2(g, c) values for which the GSFC sphere source always saturated the channels (ocean channels at g 1 for bands 7 and 8; ocean channels for g 2 for bands 6, 7, and 8), the derived K2 (g, c) values for other gains and the four and one lamp level were used. If the gain ratios in Table 17 are written as G(g, c), then K2(1, c) K2(g, c) G(g, c), (18) because the gain ratios are with respect to g 1. The coefficients for the ocean channels in band 6, g 2, were 3, vector to the exit aperture. SeaWiFS was aligned to view derived from the one lamp measurements at g 1, g 3 the center of the exit aperture with the nadir pLxel corand g 4, and the four lamp measurements at g 1 responding to the central position. For each sphere level, (Table 14). The ocean channels in bands 7 and 8 at g and g 2 were derived from the one lamp measurements at g 3 and g 4. The final values are reported in Table 18 (band-centered method) and Table 19 (band-averaged method). In each case, the final value is the weighted average of the results obtained at the different sphere levels. The weighting factors used were the uncertainties u(K2) which were obtained for each sphere level. The determination of these uncertainties is discussed in Sect. 7. The uncertainties in the weighted averages were determined according to the procedures outlined for determining combined standard uncertainty for correlated input quantities (International Organization for Standardization 1993). As in Barnes et al. (1994b), the 32 values of K2(g, c), in conjunction with the value of Cdark, were used to predict the net signal at saturation, Ssat(g, c) this is simc) as described above. The uncertainty determined from the ply (12) with Cout 1,023 counts. Using the K2(g, values determined from the band-averaged radiances (Table 19), the saturation radiance, Ls_t(g, c), was calculated for each channel, Ls_t(g, c) Ss_t(g, c)K2(g, c). In each band, the saturation radiance for the cloud channel will be much greater than that for the ocean channels. The net corrected signals (or counts) for each channel in a band, evaluated at the radiance values where each channel saturates, are termed "knees." Knee I corresponds 28 SeaWiFS output for the 4:1 TDI setting for band 1 and gain 1. 6.4 Linearity In order to test the linearity of the net corrected signals from SeaWiFS with respect to radiant flux, the SXR and SeaWiFS measured the flux from the GSFC sphere at 17 different radiance levels, from 16 0 lamps, on 11 April 1997. The measurements started with 16 lamps illuminated; at each successive level one lamp was switched off. The 4:1 TDI setting and g 1 was used, because this is the typical measurement configuration for SeaWiFS ocean color measurements. Gain 3 (g 3) was included to provide additional radiance levels that produced fewer than the knee 1 counts in Table 20. For the test, the SXR was positioned to view the center of the exit aperture of the GSFC sphere from an angle of about 30 ° from the normal the SXR acquired one data set at all sLx channels, with 10 measurements at each channel. SeaWiFS recorded data for 120 scan lines. The data were analyzed by determining the net voltages for each SXR channel and the net corrected counts from SeaWiFS, plotting the SeaWiFS counts as a function of SXR voltage, and fitting the result to a straight line. For each SeaWiFS band, the slope of the line is inversely proportional to the calibration constant, which can be determined for the 4:1 TDI setting from the data in Table 20 (Barnes et al. 1994b). The SXR black disc was not used; instead, for each channel, the voltage offset measured with the lens cap in place was subtracted from the total signal and this result was corrected for the average SSE as measured over the course of the entire experiment at OSC. The SeaWiFS net corrected counts S_b(g, c) were determined standard deviations in the S_ab(g, c) values was used to weight the data in the linear regression. Because a one-to-one correspondence does not exist in measurement wavelengths between the SXR and SeaWiFS, the spectral variation in the sphere as a function of lamp configuration was investigated. For each channel in the SXR, and band in SeaWiFS, the ratio of the output signals for successive lamp levels was compared to the value

Original page 33

B.C.,Johnson,E.A.Early,R.E.Eplee,,Jr.,R.A.Barnes,andR.T.Caffrey 0.085 - 0.080 - 0.075 - g 0.070 - O o 0.065 - '_ ::i_ 0.060 - 0.055 - Z [] 16 l.amps 0.0502 O 8 Lamps A 4 Lamps 0.045- V 1 Lamp ---- 0.040- 0.035- _ 0.030- 0.025 - 0.020 I I I I 2 3 SeaWiFS Band [] T y S V V Kt V I I I I I I 4 5 6 7 8 9 Fig. 6. Band-averaged K2(9, c) values for the single cloud channel in each band at 9 1 as a function of the radiance level of the GSFC sphere source. The symbols correspond to the different lamp configurations. The vertical bars correspond to the combined standard uncertainty in the calibration coefficients as described in Sect. 7 (16 and 1 lamp level only which are not shown if they are comparable or smaller than the size of the plotted symbol). Similar results were obtained at the other gain settings for the cloud channels. For each set of results, the weighted 0.012- 0.011 0 0.010 o 0.009- ::L 0.008 '_ 0.007 0oo6 [] 16 Lamps 0.005 O 8 Lamps I A 4 Lamps 0.004 V Lamp 0.003 0.002 I I I 1 2 3 SeaWiFS Band averages gave the final values (Table 19). 0.0114q 0.0112 -] T 0.0110-] 0.0106 4 0.0108 q I 0.0104-1 0,0102 I O i _ ' V I I I 4 5 6 Fig. 7. Band-averaged 1422(9, c) values for one ocean channel in each band at g 1 as a function of the radiance level of the GSFC sphere source. The symbols correspond to the different lamp configurations. Similar results were obtained at other gain settings and ocean channels. The insert is an expanded view of the results for bands 1 and 2; the vertical bars correspond to the combined standard uncertainty in the calibration coefficients as described in Sect. 7 (16 and 1 lamp level only). 29

Original page 34

The 1997 Prelaunch Radiometric Calibration of SeaWiFS 70a) i ::L 60. i ¢?, 5o. E 40. {1,} 30. n¢ 20. {1.) Ok0 ¢D " 10- < -+ O. I I I I I 0 1O0 200 300 400 SeaWiFS 11.2 - i =k i ll.l ¢.) o,.. n¢ 11.0 {1.} Ca0 {1.} ;> < -+ 10.9 I 792 794 SeaWiFS J I I I I I I 500 600 700 800 900 1000 Signal [counts] I I 796 798 Signal [counts] Fig. 8. The relationship between band-averaged radiance and corrected net output from SeaWiFS band 1 in the standard configuration (9 1, 4:1 TDI setting): a) measurements over the entire dynamic range of the band, and b) measurements in the region of the three knees. The slopes of the lines are equal to the band 1 calibration coefficient/(2(1) in the various regions. 3O

Original page 35

B.C. Johnson, E.A. Early, R.E. Eplee, Jr., R.A. Barnes, and R.T. Caffrey Table 18. Values of K2(g, c) determined from Ln,97(AD) values: the band-center method. The units of K2(g, c) are mWcm 2 sr l#m i count 1 and the relative standard uncertainty is given in percent. Each result is the weighted average of the results at the different lamp levels. The uncertainties are the average of the uncertainties in K2 (g, c) obtained at the different sphere levels. Band Channel (c) Gain 1 Gain 2 Gain 3 Gain 4 N,,mber N,,mber K20, c) u(K2) K O, c) u(K ) K O, c) K20, c) 1 1 0.05760 3.12 0.05512 3.08 0.05664 3.15 0.05587 3.10 2 0.01050 2.88 0.005290 2.86 0.007947 2.88 0.006247 2.87 3 0.01060 2.90 0.005338 2.86 0.008031 2.86 0.006303 2.87 4 0.01049 2.87 0.005259 2.86 0.007943 2.87 0.006237 2.86 2 1 0.01038 2.01 0.005216 2.03 0.007872 1.98 0.006178 2.03 2 0.01038 1.99 0.005228 2.02 0.007865 2.00 0.006170 2.03 3 0.01042 2.01 0.005236 2.03 0.007905 1.99 0.006189 2.03 4 0.06720 2.17 0.06414 3 1 0.06799 1.55 0.06520 2.16 0.06636 2.16 0.06491 2.21 1.53 0.06843 1.55 0.06619 1.54 2 0.008227 1.44 0.004141 1.47 0.009194 1.44 0.004900 1.47 3 0.008168 1.44 0.004106 1.47 0.009106 1.44 0.004857 1.48 4 0.008164 1.44 0.004100 1.47 0.009113 1.44 0.004857 1.48 4 1 0.007146 1.31 0.003594 1.33 0.009085 1.32 0.004254 1.34 2 0.007154 1.31 0.003601 1.34 0.009088 1.31 0.004258 1.34 3 0.007129 1.31 0.003589 1.33 0.009044 1.31 0.004238 1.33 4 0.06681 1.40 0.06486 5 1 0.06707 1.20 0.06556 1.40 0.06764 1.42 0.06525 1.40 1.19 0.06792 1.19 0.06585 1.20 2 0.005840 1.17 0.002941 1.23 0.009104 1.17 0.003668 1.24 3 0.005856 1.17 0.002950 1.24 0.009119 1.18 0.003676 1.24 4 0.005859 1.17 0.002952 1.24 0.009133 1.17 0.003684 1.23 6 1 0.003347 1.15 0.001680 1.15 0.009205 1.08 0.005053 1.15 2 0.003328 1.15 0.001676 1.16 0.009138 1.08 0.005006 1.15 3 0.003400 1.14 0.001709 1.15 0.009340 1.09 0.005116 1.14 4 0.05717 1.08 0.05544 7 1 0.04343 1.37 0.04357 1.08 0.05847 1.07 0.05776 1.09 1.34 0.04508 1.34 0.04471 1.34 2 0.002373 1.41 0.001193 1.42 0.007655 1.38 0.004121 1.38 3 0.002385 1.42 0.001199 1.43 0.007639 1.38 0.004131 1.38 4 0.002371 1.41 0.001191 1.43 0.007622 1.38 0.004113 1.38 8 1 0.001699 1.57 0.0008531 1.57 0.006530 1.53 0.003405 1.52 2 0.001720 1.56 0.0008648 1.57 0.006591 1.53 0.003440 1.53 3 0.001698 1.57 0.0008528 1.58 0.006533 1.53 0.003408 1.53 4 0.03494 1.50 0.03495 1.50 0.03587 1.50 0.03552 1.50 Table 19. Values of K2(g, c) determined from (11), which is the band-averaged method. The units of K2(g, c) are mWcm 2 sr l#m i count 1 and the relative standard uncertainty is given in percent. Each result is the weighted average of the results at the different lamp levels. The uncertainties are the average of the uncertainties in K2 (g, c) obtained at the different sphere levels. Band Channel _) Gain 1 Gain 2 GMn 3 Gain 4 N.mbe N.mbe K O, c) K O, c) K20, c) u(K ) K O, c) 1 1 0.06025 3.28 0.05767 2 0.01098 3.04 0.005534 3 0.01109 3.05 0.005584 4 0.01098 3.03 0.005502 3.24 0.05925 3.32 0.05844 3.27 3.01 0.008314 3.03 0.006536 3.03 3.02 0.008402 3.02 0.006594 3.03 3.02 0.008310 3.02 0.006525 3.02 31

Original page 36

The1997PrelaunchRadiometricCalibrationofSeaWiFS Table 19. (cont.) Valuesof K2(g, c) determined from (11), which is the band-averaged method. The units of K2(g, c) are mWcm 2 sr l#m i count 1 and the relative standard uncertainty is given in percent. Each result is the weighted average of the results at the different lamp levels. The uncertainties are the average of the uncertainties in K2 (g, c) obtained at the different sphere levels. Band Channel (c) Gain 1 Gain 2 Gain 3 Gain 4 Number Number K2(g, c) u(K2) K (g, c) u(K ) K (g, c) K2(g, c) 2 1 0.01057 2.01 0.005313 2 0.01058 1.98 0.005325 3 0.01061 2.01 0.005333 4 0.06845 2.17 0.06534 3 1 0.06932 1.68 0.06649 2 0.008387 1.57 0.004221 3 0.008327 1.56 0.004185 4 0.008323 1.56 0.004179 4 1 0.007159 1.31 0.003600 2 0.007167 1.32 0.003607 3 0.007142 1.31 0.003595 4 0.06694 1.40 0.06499 5 1 0.06735 1.31 0.06581 2 0.005865 1.28 0.002952 3 0.005882 1.28 0.002961 4 0.005884 1.28 0.002963 6 1 0.003323 1.21 0.001668 2 0.003304 1.21 0.001664 3 0.003376 1.21 0.001697 4 0.05676 1.15 0.05499 7 1 0.04343 1.38 0.04357 2 0.002372 1.42 0.001193 3 0.002384 1.43 0.001199 4 0.002370 1.43 0.001191 2.02 0.008018 1.98 0.006293 2.03 2.02 0.008011 1.99 0.006284 2.03 2.02 0.008052 1.99 0.006304 2.03 2.15 0.06760 2.16 0.06613 2.20 1.66 0.06977 1.68 0.06749 1.67 1.59 0.009373 1.56 0.004994 1.60 1.60 0.009283 1.57 0.004951 1.60 1.60 0.009290 1.57 0.004951 1.60 1.34 0.009102 1.32 0.004261 1.34 1.34 0.009105 1.31 0.004265 1.34 1.34 0.009061 1.31 0.004245 1.34 1.41 0.06778 1.42 0.06538 1.41 1.31 0.06820 1.31 0.06609 1.32 1.35 0.009143 1.29 0.003681 1.35 1.35 0.009159 1.29 0.003689 1.35 1.35 0.009172 1.28 0.003698 1.34 1.22 0.009143 1.15 0.005017 1.21 1.23 0.009076 1.15 0.004969 1.21 1.22 0.009277 1.15 0.005079 1.21 1.15 0.05808 1.14 0.05737 1.16 1.36 0.04509 1.36 0.04472 1.36 1.43 0.007652 1.40 0.004119 1.40 1.44 0.007636 1.40 0.004130 1.40 1.44 0.007619 1.40 0.004111 1.40 8 1 0.001688 1.83 0.0008476 1.84 0.006487 1.79 0.003383 1.79 2 0.001709 1.82 0.0008592 1.83 0.006548 1.80 0.003418 1.79 3 0.001687 1.83 0.0008472 1.84 0.006491 1.80 0.003386 1.79 4 0.03470 1.77 0.03471 1.77 0.03564 1.77 0.03528 1.78 Table 20. The net counts from SeaWiFS bands 1 8, as a function of radiance for all four of the gain settings, determined using the values of K2(g, c) in Table 19. The net counts for each band are the average of the net counts for the four detectors in that band (the 4:1 TDI setting). At each knee, one of the detectors saturates. The radiance is given in units of mWcm 2 sr 1 #m 1 and the output of the SeaWiFS band is in net corrected counts. Band Gain (9) Knee 1 Knee 2 Knee 3 Saturation Number Setting Radiance Output Radiance 1 10.98 793.11 11.00 5.524 771.83 5.531 8.311 782.33 8.334 6.533 775.85 6.547 10.60 789.53 10.62 5.342 772.98 5.342 8.031 780.04 8.057 6.302 774.83 6.325 32 Output Radiance Output Lsat(g,c) Ssat(g,c) 794.13 11.14 797.87 60.37 1002.15 772.51 5.618 776.76 57.79 1002.91 783.77 8.444 787.52 59.37 1002.39 776.98 6.631 780.52 58.56 1002.66 790.68 10.69 792.45 68.80 1004.67 773.04 5.390 775.46 65.67 1006.09 781.80 8.118 783.91 67.94 1005.14 776.79 6.364 778.46 66.46 1005.65

Original page 37

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table20. (cont.) ThenetcountsfromSeaWiFSbands1 8, asa functionofradiancefor allfourof thegain settings,determinedusingthe valuesof K2 (9, c) in Table 19. The net counts for each band are the average of the net counts for the four detectors in that band (the 4:1 TDI setting). At each knee, one of the detectors saturates. The radiance is given in units of mW cm 2 sr 1 #m 1 and the output of the SeaWiFS band is in net corrected counts. Band Gain (g) Knee 1 Knee 2 Knee 3 Saturation Number Setting Radiance Output Radiance 3 1 8.345 780.00 8.356 2 4.202 767.01 4.203 3 9.302 782.27 9.327 4 4.968 768.75 4.977 4 1 7.169 778.07 7.171 2 3.606 764.87 3.607 3 9.096 784.11 9.115 4 4.259 766.61 4.268 5 1 5.869 770.79 5.887 2 2.951 759.18 2.982 3 9.143 782.29 9.152 4 3.681 762.27 3.707 6 1 3.322 761.97 3.327 2 1.674 756.69 1.683 3 9.094 783.38 9.107 4 4.987 766.66 5.009 7 1 2.362 759.40 2.378 2 1.189 753.65 1.199 3 7.581 786.61 7.622 4 4.094 768.13 4.115 8 1 1.695 762.30 1.699 2 0.8532 757.83 0.8540 3 6.500 794.62 6.521 4 3.393 773.47 3.403 expected if every lamp had identical output and identical spectral shape; the results are shown in Fig. 9. The values are close to unity, with variability increasing as the number of lamps decrease, as expected. Table 21. Number of lamps illuminated during the SeaWiFS linearity test with the 4:1 TDI setting using the GSFC sphere source that resulted in measurements below knee 1 of the bilinear gain curve Table 20 and Fig. 9). Band Lamp Configuration Number g 1 g 3 16 16 16 16 12 13 8 11 5 7 1 4 0 3 0 2 Output Radiance Output Lsat(g,c) Ssat(g,c) 780.74 8.394 782.02 69.46 1002.24 767.22 4.221 768.34 66.62 1002.96 783.72 9.382 785.36 69.91 1002.24 769.73 4.995 770.69 67.62 1002.69 778.21 7.192 779.03 67.08 1002.67 765.02 3.624 766.26 65.12 1002.82 785.25 9.133 785.79 67.92 1002.61 767.69 4.282 768.61 65.51 1002.74 772.36 5.919 773.85 67.15 1001.13 764.55 2.996 765.83 65.61 1003.69 782.82 9.209 784.58 67.99 1000.07 765.81 3.731 767.52 65.89 1002.66 762.79 3.342 763.89 56.87 999.67 759.32 1.685 759.57 55.10 1002.41 784.10 9.203 787.10 58.20 997.99 768.90 5.033 770.20 57.48 998.77 762.94 2.388 764.05 43.43 1000.30 758.02 1.207 759.72 43.57 1002.79 789.54 7.659 790.96 45.09 998.50 770.76 4.139 772.35 44.72 999.21 763.42 1.704 764.27 34.81 1002.74 758.30 0.8555 758.75 34.81 1003.34 796.36 6.547 797.55 35.74 1002.36 774.98 3.415 775.93 35.39 1002.48 The results are essentially the same, whether the SXR net voltages or the SeaWiFS net counts are used. For example, both instruments indicate the radiance change between the two-lamp and the three-lamp configuration was larger than the expected value of 1.5, with the bluer wavelengths showing the greatest change. This indicates the variation was caused by the lamps and not other factors, such as the uniformity of the spectral radiance in the exit aperture, he SeaWiFS linearity estimates for bands 4 and 8 (at 510 nm and 865 nm), were therefore, analyzed using the SXR channels 3 and 6 (at 487 nm and 775 nm). With the data presented this way, it is easy to see which SeaWiFS bands were below tile knee on the bilinear gain curve, i.e., none of the ocean channels were saturated. Above the knee, the SeaWiFS output for a band in the 4:1 TDI setting does not increase as rapidly with increasing radiance, so the normalized ratios are greater than unity. These points are off the scale in Figs. 9b and 9c. The number of lamps that resulted in measurements below the knee of the bilinear gain curves are given in Table 21 for g 1 33

Original page 38

The1997PrelaunchRadiometricCalibrationofSeaWiFS andg 3. The net corrected signals in digital counts for the two gain settings are given in Tables 22 23. For g 1, the net SeaWiFS counts for bands 1 5 were fit to a straight line. The independent variable for each curve fit is the net SXR voltages for the channel with the measurement wavelength closest to the SeaWiFS band. For g 3, the SeaWiFS counts for bands 1 7 were analyzed the same way. As an example, the SeaWiFS counts stable to better than 0.1% over the 20min interval. Sea- WiFS was configured for the 4:1 TDI setting with g 2 for bands 1 5, g 3 for bands 7 and 8, and g 4 for band 6. The counts were stable with time, and the standard deviations for the 20 min set for each band were less than 1 count. Repeatability is the level of agreement between results obtained for successive measurements carried out under and the fitted model for band 1 at g 1 are plotted as the same conditions (International Organization for Stana function of SXR channel 1 voltage in Fig. 10. The results of the linear fits were satisfactory in all cases. The dardization 1993). Reproducibility is the level of agreement between the results for changed conditions of meaconstant offset b0 was less than one count and the ratio of surements. The SeaWiFS Project has defined short-term the fitted slopes for g 3 and g 1 for the same band agreed to within 1.5% of the expected gain ratios. The linear correlation coefficients were nearly unity, indicating the probability that the data were not correlated was less than 0.001 for all but the band 7, g 3 case, where the probability was 0.0042. The standard deviations of the fits ranged from about 0.3 counts for band 1, to 4.1 counts for band 7. As an additional test, the data for bands 1 and 2 with g 1 were fit to a quadratic curve in the SXR data the net channel 1 voltage; the improvement in the fit was not statistically significant. To illustrate the degree of linearity, the slope was determined for each SeaWiFS count, Yn, and SXR voltage, xn, data pair: (y_ - bo)/X_. The subscript n refers to the number of lamps operating. The result was compared to the average slope for the linear regression for each fit. The results are shown in Fig. 11. The dispersion in the normalized slopes is less than 4-0.6% and the standard deviation is about 0.2%. Above knee 3, the data can also be analyzed using the simple linear regression model. The calibration constant for the 4:1 TDI setting above knee 3 is a constant for each band that depends on the spectral radiance and the net counts at saturation and knee 3 (Table 20 and Barnes et al. 1994b). Bands 3 8 with g 1 were analyzed using the SXR data for channels 3 6, again using the SXR channel that was closest in wavelength to the SeaWiFS band. The results were also satisfactory, with the probability that the data were uncorrelated less than 0.001 in all cases except for band 3, where the probability was 0.0033. The standard deviations of the fits varied from 0.6 1.8 counts. 6.5 Warm Up and Repeatability The temporal stability of the output from SeaWiFS was investigated on 24 January 1997 using the GSFC sphere source with two lamps illuminated. The SXR was aligned to view the center of the exit aperture of the sphere from an angle of about 30 ° from normal. The sphere was allowed to warm up for I h and then the SXR and SeaWiFS measured the sphere for about 20 min. The SXR was fully warmed up, but the power for the detector electronics and the scan motor in SeaWiFS were off until the measurements started. At 412nm, the relative SXR output was 34 repeatability in terms of measurements made within a two week interval (Barnes et al. 1994a), and long-term repeatability in terms of longer time intervals. The successive measurements of the same radiance levels in the 1:1 TDI setting on 24 January and 11 April were repeatable to between 0.1_?.2%, depending on the radiance level. These values were determined by calculating the standard deviation of the ratio of the S_ab (g, c) values for the January and April measurements. The standard deviation was determined from the ratios corresponding to all unsaturated bands at a particular gain setting for each lamp level. There was no dependence with gain, so the average standard deviation for each lamp level was calculated and used to estimate the uncertainty component associated with the long-term repeatability of SeaWiFS. The short-term repeatability for the K2(g, c) values was estimated from several measurements of the GSFC sphere source using the 4:1 TDI setting in January 1997. AIthough all gains were used, the majority of the data were for g 1, and these results were used for the short-term repeatability analysis. On 23 January, the four radiance levels were measured three times (Table 1), and on 24 January the four radiance levels were measured once. On 22 January, three radiance levels were measured once (corresponding to the 16, 8, and 1 lamp illumination). For each band, the relative standard deviation (that is, the standard deviation of the separate measurements normalized by the mean) was determined. The result was between 0.01% (band 8, all four radiance levels) and 0.6% (band 1, 1 lamp). Other measures of the repeatability of SeaWiFS are the result of a comparison of the three sets of gain ratios: 24 January and 23 April at OSC, and 16 June at VAFB. The gain ratios agreed to within 4-0.5%, or within 1 count. Finally, the stability of the ratios of the calibration coefficients from two bands was evaluated. The SeaWiFS bio-optical algorithms are a function of such band ratios. For example, band 2 divided by band 5, and band 3 divided by band 5 are used for the chlorophyll concentration algorithm; and band 7 divided by band 8 is used for the atmospheric algorithm. These three color ratios varied by less than 0.4% for the 24 January and 11 April measurements.

Original page 39

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table22. Valuesfor Slab(g, c) during the SeaWiFS linearity test for g 1 on 11 April 1997. Results that are below the knee of the bilinear gain curve are shown in the columns with the actual band numbers. Results that are above the curve are in columns marked with the prime symbol (1). Lamps Band 1 2 3 31 4 41 5 51 6 61 7 8 16 278 472 797 805 825 891 976 10041 15 255 433 792 802 82O 881 961 10041 14 237 404 790 799 816 872 947 10041 13 219 374 787 796 811 864 932 1000 12 201 343 776 793 807 855 918 981 11 185 315 711 790 803 846 904 962 10 167 284 643 786 798 838 890 942 9 150 256 580 783 794 829 877 925 8 132 226 512 713 7 116 199 450 627 6 99 170 385 536 5 81 139 316 441 4 66 113 256 356 3 48 83 188 263 2 33 57 129 180 1 16 28 65 90 Results where the band satur_es. 790 821 863 906 786 813 850 887 782 804 836 868 769 795 822 849 620 787 809 831 459 778 795 812 313 770 781 793 157 475 767 775 Table 23. Values for Slab(g, C) during the SeaWiFS linearity test for g 3 on 11 April 1997. Results that are below the knee of the bilinear gain curve are shown in the columns with the actual band numbers. Results that are above the curve are in columns marked with the prime symbol (1). Lamps 1 2 3 31 4 41 16 362 615 795 806 15 332 564 792 8O2 14 309 526 790 799 13 286 487 759 795 12 263 447 697 792 11 241 410 640 786 10 218 370 578 710 9 196 333 521 641 8 173 294 461 566 7 152 259 405 498 6 130 221 346 426 5 106 181 284 350 4 86 146 230 283 3 63 108 169 209 2 44 74 116 143 1 22 37 58 72 Results where the band satur_es. Band 5 51 6 61 7 71 8 81 824 889 975 1003 t 819 879 959 1003 t 815 870 946 1003 t 810 862 931 999 806 853 917 980 802 844 903 961 797 835 888 941 793 827 875 924 788 819 861 905 711 811 848 886 608 802 834 867 501 793 820 848 404 704 807 830 299 523 789 810 204 355 535 679 102 178 268 340 35

Original page 40

The 1997 Prelaunch Radiometric Calibration of SeaWiFS [] a) 1.04- 0 A V 1.03 - X 09 <> [] 1.02 - c]] © 1.01 t_o 1.00 - 09 [] 9 b 0.99 - © V Z 0.98 - A 0 [] 0.97 ' I ' I ' I ' 0 2 4 6 Number b) 1.04- _9 0 [] 1.03 - c_9 V 1.02 - © 1.01 1.00 - v 0 c_9 o _ 0.99 - b x o 0.98 - Z 0.97 I I I 0 2 4 6 Number of Lamps Fig. 9. The net signal, using the SXR and SeaWiFS, [] Channel 1 0 Channel 2 A Channel 3 V Channel 4 Channel 5 [] Channel 6 [] <> V A 0 [] I ' I ' I ' I ' I ' I 8 10 12 14 16 18 of Lamps i J i [] Band 1 r i O Band 2 i i A Band 3 r i )< Band 4 J i V Band 5 X Band 6 [] Band 7 )l( Band 8 _ _ o 0 _ a a 0 [] I I I I I I 8 10 12 14 16 18 for the lamp configuration with n - 1 lamps illuminated divided by the net signal for the lamp configuration with n lamps illuminated, multiplied by the factor n/(n- 1): a) Net signal ratios from the SXR. The symbols represent the SXR instrument channels, b) Net signal ratios from SeaWiFS at 9 1. The symbols represent the SeaWiFS bands. The dotted lines with arrows indicate the bands that crossed the knees in the bilinear gain curves (Table 22). 36

Original page 41

B.C. ,Johnson, E.A. Early, R.E. Eplee, ,Jr., R.A. Barnes, and R.T. Caffrey c) 1.04 ¸ i °_ 6 27 A Band 3 1.03 i 3 t _ 1.02- _5 m .0 1.01- Ill 1.00 0 _ 0.99- [] N 0 i 0.98- [] z 0.97 ; 6 Number of Lamps i / [] Band 1 O Band 2 X Band 4 V Band 5 Band 6 J i [] Band 7 t i Band 8 J i i [3 [] O [] 8 1'0 1'2 14 1'6 1'8 Fig. 9. (cont.) The net signal, using the SXR and SeaWiFS, for the lamp configuration with n - 1 lamps illuminated divided by the net signal for the lamp configuration with rz lamps illuminated, multiplied by the factor n/(n - 1): c) Net signal ratios from SeaWiFS at g 3. The symbols represent the SeaWiFS bands. The dotted lines with arrows indicate the bands that crossed the knees in the bilinear gain curves (Table 23). 300 - 250. o © 200. a <> <> 150. ;n 5" I00. o (D r*'3 50. 0. 0 0 0.5 1 0 1.5 2 0 1.00 0.75 0.50 :Z <> 0.25 =-- 0.00 Z <> <> <> <> 0.25 -0.50 -0.75 -1.00 2.5 3.0 3.5 4 0 4.5 SXR Channel 1 [V] Fig. 10. The net corrected counts from SeaWiFS for band 1 and 9 1 plotted as a function of the net voltage of channel 1 of the SXR (open squares). The data were acquired on 11 April 1997, as the sphere was operated with 16, 15, 14, ... 2, and 1 lamp illuminated. The uncertainties in the SeaWiFS counts cannot be illustrated clearly because they are small compared to the symbols. are the residuals to the fit, normalized by the uncertainties The solid line is a linear fit to the data. Also shown, (open diamonds). 37

Original page 42

The 1997 Prelaunch Radiometric 1.006 - a) 1.005 - A 1.004 - 1.003 - 1.002 - o O A 1.001 - V 1.000 - E 0.999 - O o Z 0.998 - O 0.997 - 0.9% - [] O 0.995 - A 0.994 - V 0.993 I I I 2 4 6 Lamps Illuminated 1.oo7- b) 1.006 - 1.005 1.004 1.003 O 8 '. 1.002 6 © V 1.001 t ) .. 1.000 v o Z [] 0.996 - O 0.995 0.994 O 0.993 I I I I 0 2 4 6 Lamps Illuminated Fig. 11. Normalized slopes for the SeaWiFS linearity Calibration of SeaWiFS _5 [] Band 1 0 Band 2 f Band 3 V Band 4 Band 5 I I I I I I 8 10 12 14 16 18 [] Band 1 0 Band 2 A Band 3 V Band 4 O Band 5 [] Band 6 @ Band 7 [] [] _ 0 [] v DO 0 0 I I I I I I 8 10 12 14 16 18 test for 9 1 and 9 3. The slope for each data pair, (Yn - bo)/xm where b0 was the fitted intercept, was normalized by the fitted slope. The symbols correspond to the SeaWiFS bands: a) slopes for 9 1, and b) slopes 38 for 9 3.

Original page 43

B.C. Johnson, E.A. Early, R.E. Eplee, Jr., R.A. Barnes, and R.T. Caffrey 7. UNCERTAINTY ANALYSIS The uncertainty analysis follows from the method of propagation of uncertainties applied to the measurement equations described in Sect. 6. The combined uncertainty in the independent calibration coefficients K2 (g, c), u(K2), is determined from J 2 J 1 V L j 1 j 1 (19) J 0 E -z;K2(g'c) o-_kK2(g'c)u(zJ'zk)] " k j+l J (Taylor and Kuyatt 1994) where K2(g, c) depends explicitly on the set of parameters zj, which have standard uncertainties u(zj), and where j and k are indicies of summation. There are J parameters in (19), and the estimated covariance for pairs of parameters is u(zj, zk). The partial derivative of K2(g, c) with respect to zj are termed sensitivity coefficients. The sensitivity coefficients can be de- 7.1.2 Band-Averaged Method The uncertainty in the band-averaged radiance for each band depends on the uncertainty in the spectral radiance of the sphere, u(L(A)), and the uncertainty in the relative spectral response for each SeaWiFS band, u(R(A)). The uncertainty in the band-averaged radiance, un(LB) , was evaluated by treating the uncertainties u(L(A)) and u(R(A)) as if the random portions were negligible. This is not really the case; Fig. 1 indicates that the random uncertainty in the spectral radiance can be up to 0.3%. This approximation, however, results in a conservative estimate for un(LB) and also simplifies the calculations. In the case of systematic uncertainties, the correlation coefficient (which is proportional to the covariance) is unity, and (19) is simplified to (21) j 1 "_zjLn'B'97U(ZJ) Evaluating (21) involves approximating the integral extermined analytically from the measurement equation [(8) pressions in (5) as sums and then determining the partial or (11), and (13)], or from experimental measurements. Inspection of the measurement equation, (8) or (11), indicates two obvious uncertainty components: the radiance (either the band-centered or band-averaged values) or the net corrected signals. Other components include the linearity and repeatability of SeaWiFS, as well as terms associated with the GSFC sphere source. 7.1 Radiance Let L(A) represent either the spectral radiance of the GSFC sphere source at the nominal SeaWiFS center wavelengths (band-centered spectral radiance) or the normalized integral of the spectral radiance of the sphere source, with the relative spectral responsivities of the SeaWiFS bands as weighting factors (band-averaged radiance). From (8) or (11) 0g2(g, c) 0L(A) u(L(A)) K2(g, c) u(L(A))L(A)' (20) so it is necessary to determine the relative standard uncertainty in the radiance, u(C(A)). 7.1.1 Band-Centered Method The method of determining the spectral radiance of the GSFC sphere source for the SeaWiFS measurements was described in Sect. 5. The sources of uncertainty are described in Sects. 5.1 and 5.3, as well as in Tsai and Johnson (1998). The relative standard uncertainties at the Sea- WiFS nominal center wavelengths are given in Table 12 and illustrated in Fig. 5. derivatives. The value of J was 2, corresponding to spectral radiance and relative spectral responsivity. The values for u(L(A)) were the same as for the bandcentered approach (Sect. 5.3). They were determined every I nm using cubic spline interpolation. Figure 5 illustrates these relative values in units of percent, 100 u (L(A))/L(A). The values for u (R(A)) were determined by comparing the results from the two independent measurements of these quantities at SBRC during the characterization of Sea- WiFS. One method involved measurements of the separate components in the SeaWiFS optical train: spectral reflectance of the mirrors, spectral transmittance of the filters and beamsplitters, and spectral responsivity of the channels. The results, which covered the spectral interval from 380 1,150 nm, were multiplied together to determine relative spectral responsivity for each band in SeaWiFS. The other method involved measurements of the system responsivity using a monochromatic source (Barnes et al. 1994b). Within -t-30 nm of the nominal center wavelengths, the extent of the system level measurements, the two methods differed by no more than 6%. As a conservative estimate, the values for u(R(A)) for the spectral region, bounded by the 1% response (referenced to the peak response), were assigned a relative standard uncertainty of 6%. For regions where this comparison could not be made, an uncertainty of 50% was assigned to u(R(A)). The final results for the relative standard uncertainty in the bandaveraged radiances are in Table 13. 7.2 Corrected Net Signal The basic equation for the net corrected signal is (13), which indicates that S_ab(g, c) depends on a number of 39

Original page 44

The1997PrelaunchRadiometricCalibrationofSeaWiFS parameters:Gout, Cdark, /_3, T, and Ri. Propagation of day, Slab (g, c) was determined for each set of experimental uncertainty for these parameters gives: oSlau (g, c) (Cout) slau(g, c) (Coud 0Cout AC (Coud, (22) 0Slab(g , C) g(Cdark) Slab(g , C) U(dark) OCdark AC U(Cdark), (23) oSlau(g, c) 0K3 u(K3) [Slau(g, c) - AC R] u(K)K (24) zXTSlau(g, c) (K) oSlau (g, c) Slau(g, c) - AC R OF _(Y) Ay (Y) (25) K slau(g, c) _(T), and OSlab(g, c) u(Ri) Slab(g, c) u(Ri), (26) ORi Ri where AC Cout -- _dark and AT T - Tref • In (22) (26), the standard uncertainty in Cout, Cdark, K3, T, and Ri are denoted u(Cout), u(Cdark), u(K3), u(T), and u(Ri), respectively. The values for Cout were not readily available for analysis because the quantity AC was determined for each scan line. The corrections for temperature and mirror side reflectivity are small, however, so the quantity Slab(g , c)/AC differs from unity by no more than -4-0.8% for focal plane temperatures up to 302 K. This sensitivity coefficient can, therefore, be ignored in (22) and (23) without causing the uncertainty component to be in error by more than 1%. The standard deviation of the 600 measurements of Slab(g, C) was used to estimate U(Cout). The standard deviation of the 120 measurements of Cdark, was used to estimate U(Cdark). The sensitivity coefficients in (24) and (25) involve the difference [Slab(g, c) -- AC Ri], which is/£3 ATRi AC _ K3 AT Slab(g, C) because the Ri is unity on average and _ab(g, c) _ AC. In (24), the relative standard uncertainty in K3 is assumed to be 10%. In (25), u(T) is determined assuming a uniform rectangular distribution function with a resolution of 0.267°C, resulting in a value of 0.077°C. The maximum temperature difference on 24 January was 8.6°C, and this value was used for estimating the uncertainty components for the measurements on 24 January and 11 April. Finally, in (26), the relative uncertainty in Ri, u(Ri)/Ri, was estimated assuming a uniform, rectangular distribution function for Ri bounded by /_1 and R2. The uncertainty components in always saturated by the GSFC sphere source (Table 14), arising from the uncertainty in K3, T, or Ri are given Table 24 for each SeaWiFS band. Most of the values are negligible, especially for bands 3 8. Two independent determinations of Slab(g,C ) were made: on 24 January and 11 April 1997. These are denoted Slab(g, c)Jan and Slab(g, C)Apr, respectively. On each 4O parameters (sphere radiance level, band, gain, and channel) from the average of 600 measurements. The average value of Slab(g, c) was used in (8) or (11) to determine the calibration coefficients K2 (g, c). The combined uncertainty in Slab(g, c), u(Slab), is the quadrature sum of the components in Table 25. The two independent measurements of Slab(g , C) reduce the uncertainty associated with these random components. Table 24. The values for the uncertainty components (in percent), in relative units, which are independent of gain and sphere level. These are conservative values, in that the maximum focal plane temperature of 8.6°C was used for evaluating (24). Band Uncertainty Components Number (K) _(T) (r) 0.077 0.007 0.041 0.050 0.005 0.031 0.036 0.003 0.001 0.034 0.003 0.013 0.034 0.003 0.000 0.013 0.001 0.000 0.009 0.001 0.000 0.007 0.001 0.000 7.3 Gain Ratios The uncertainties (Table 17) were determined from the standard deviations of the net counts for the two gains that comprise a gain ratio by the propagation of uncertainty method. The gain ratio for gain g is Slab (9, c) (27) G(g, c) Slab (1, c)' where Slab (g, C) are net corrected counts. For the channels that were not saturated by the GSFC sphere, the combined uncertainty is the quadrature sum of the components listed in Table 26. As above, the standard deviations reported for Slab(g, c) were used to estimate U(Cout). The effect of the uncertainty in Cdark was accounted for using the value for the measurement uncertainty. The terms associated with the uncertainty in the focal plane temperature and the temperature coefficient were estimated as explained above, with the maximum temperature difference of 8.6°C; these terms are small (Table 24). Because no correction was applied for the mirror side reflectivity, there is no component of uncertainty from this term. For the 15 channel and gain combinations that were the 1422(9, c) values were determined using the gain ratios (Sect. 6.3.2). Because of the uncertainty in the gain ratios (Table 17), there are additional uncertainty components in these calibration coefficients corresponding to these factors.

Original page 45

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table 25. Directuncertaintycomponentsfor thenet correctedsignalsfromSeaWiFS.Thecombineduncertaintyis thequadraturesumofthe separatecomponents.TypeA uncertaintycomponentsareestimatedusing statisticalmeans,andTypeB componentsareestimatedusingothermeans(TaylorandKuyatt1994). Source of Expression Uncertainty [counts] Type Experiment Variables _out 1 (U2 (S]ab (g, C) Jan) + U2 (S]ab (g, C)npr) ) 1 e Sphere level, band, 1 2 _dark (u (Cdark,Jan) + u2(Cdark,apr)) I(3 AT Slab(g, C) U(K3) T /3 Slab (g, C) u(T) Ri S_ab (g, c) u(Ri)/1 channel, gain A Band, channel, gain B Band g Band B Band Table 26. Direct uncertainty components for the gain ratios. The combined uncertainty in the gain ratios is the quadrature sum of the separate components. Source of Expression (Relative Type Experiment Uncertainty Standard Uncertainty) Variables Cout at g 2, 3, or 4 u(S]ab(g,c))/S]ab(g,c) A Band, channel, gain Cdark at g 2, 3, or 4 (Slab(g , 6)/2) 1 g Band, channel, gain Cout at 9 1 u(Slab(1, c))//Slab(1, C) A Band, channel, gain Cdark at g 1 (Slab(I , 6)/2) 1 g Band, channel, gain /£3 ATu(K3) T I 3 u(T) 7.4 Other Components There are several other factors that affect the uncertainty in the values for K2(g, c) which are implicit to the measurement equation. These include the linearity of response, repeatability, uniformity of the spectral radiance in the exit aperture of the sphere source, and the effect of wavelength errors in the relative spectral responsivities for each SeaWiFS band. 7.4.1 Linearity For the SeaWiFS linearity test, the number of radiance levels was limited, and it was not possible to cover the full range of the linear portion of the bilinear gain 3, ity distribution. All of these uncertainty components were curve for all bands and all gains. For g 1 and bands 4, and 5, the linearity measurements cover the full range of their responses below the knee; including the g 3 re- 11 SeaWiFS. This value was used for the uncertainty composults, the coverage is also good for bands 2 and 6. Figure shows the responses in each band to be almost linear at the increased radiance levels (more than approximately eight lamps illuminated). The increased scatter in the normalized slopes with decreasing radiance is due to fewer Sea- WiFS counts in the measurements. In general, however, B Band B Band the good agreement with the linear model using the SXR to measure the flux indicates that SeaWiFS performs as a linear instrument. The SXR was shown to be linear using an optical beamconjoiner, and the uncertainty associated with linearity is about 0.11% (Johnson et al. 1998). The relative standard uncertainty introduced by spectral variations in the sphere radiance that were difficult to quantify using the limited set of wavelengths from the SXR is small probably less than 0.4% (Fig. 9). Below the knee, the variation in the normalized slope with lamp level (Fig. 11), indicates that the linearity is bounded by 1.006 and 0.994, resulting in a relative standard uncertainty of 0.35%, assuming a uniform probabilcombined in quadrature, resulting in a relative standard uncertainty of 0.54% for the departure from linearity in nent associated with linearity for the K2(g, c) values for the 1:1 TDI setting. Above the knee, where all three ocean channels are saturated, the relative standard uncertainty associated with linearity is estimated to be 1.0%. SeaWiFS measurements 41

Original page 46

The1997PrelaunchRadiometricCalibrationofSeaWiFS Table 27. Thevaluesfor the relativestandarduncertaintycomponents,in unitsof percent,associatedwith linearityandrepeatability.Thenumbersin the secondrowreferto the numberof lampsilluminatedin the GSFCspheresource. Band Linearity Long-Term Repeatability Short-Term Repeatability Number 16 8 4 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 0.54 0.11 0.15 0.09 above the knee will be used primarily for stray light correction of ocean color data, and corrections for nonlinearity at this level of uncertainty are not significant. 7.4.2 Repeatability The short-term and long-term repeatability measurements were described in Sect. 6.4. The uncertainty component for long-term stability was determined using the ratio of the measurements on 24 January and 11 April for the 1:1 TDI setting. The average (over gain settings) standard deviation (all 32 channels at the same gain) for ratios at the same radiance was used to estimate the uncertainty component associated with the long-term repeatability of SeaWiFS (Table 27). More data were available for the short-term repeatability analysis, from the multiple sets of the measurements at g 1 and the 4:1 TDI setting. For each band, the uncertainty component associated with short-term repeatability was estimated using the standard deviation of the separate measurements at the same sphere level. The results are given in Table 27. 7.4.3 Sphere Uniformity During the calibration, SeaWiFS and the SXR viewed the center of the exit aperture of the GSFC sphere source. For the SXR, the imaged area was about 7.2 cm in diameter, and the SXR was offset no more than about 0.5 cm from the center of the exit aperture. For SeaWiFS, flux was collected from an area about 8.4 cm in diameter (the aperture stop was 7.62 cm in diameter, with 1.6 mrad field of view and a distance of 2.5 m). This region was offset from the center of the exit aperture by about 2.2 cm. The spatial uniformity of the spectral radiance of the GSFC sphere source's exit aperture was measured just before the 1995 NIST calibration using the SXR (Early and Johnson 1997). The spatial nonuniformity, expressed as the peak-to-valley variability in spectral radiance over the entire exit aperture, are given in Early and Johnson (1997) for the four sphere levels measured by SeaWiFS. 42 1 16 8 4 1 0.17 0.29 0.27 0.41 0.59 0.17 0.14 0.15 0.34 0.42 0.17 0.06 0.11 0.27 0.23 0.17 0.05 0.11 0.24 0.21 0.17 0.05 0.04 0.21 0.24 0.17 0.02 0.02 0.01 0.19 0.17 0.01 0.01 0.01 0.01 0.17 0.01 0.01 0.01 0.01 The variation of the radiance uniformity with wavelength was not significant compared to the variation in the uniformity with the number of lamps illuminated. The spatial mapping in 1995 at NIST was done at five SXR measurement wavelengths, corresponding to channels 1, 2, 3, 4, and 5. For the SeaWiFS calibration, with the SXR and SeaWiFS aligned to the center of the exit aperture and measuring nearly the same target area, the uncertainty caused by variations in spectral radiance is small compared to the uniformity values reported for the entire exit aperture. The peak-to-valley variability over the central 10 cm was used to estimate the uncertainty, assuming a uniform probability distribution function. For the 16, 8, 4, and 1 lamp configurations, this relative standard uncertainty is less than 0.4% (Table 28). It is assumed that the radiance uniformity of the GSFC sphere had not changed between the characterization in 1995 and the SeaWiFS calibration in 1997. 7.4.4 Wavelength The uncertainty in the wavelength accuracy, u(A), of the relative spectral responsivity for the SeaWiFS bands, R(A), affects the accuracy of measurements with SeaWiFS. For either method of determining the calibration coefficients (band-centered or band-averaged), the magnitude of the uncertainty depends on the local slope of the source's spectral radiance according to u(K2(g,c)) u(A) OL(A) A (28) K2(g,c) A OA L(A)" With an estimate of I nm for the uncertainty in the wavelength accuracy of the relative spectral response measurements, the uncertainty component for the calibration coefficients is from 0.02 2.03%, depending on the SeaWiFS band (Table 28). The values vary strongly with wavelength because the relative slope of the GSFC sphere's radiance varies strongly with wavelength.

Original page 47

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey Table 28. Thevaluesfor the relativestandarduncertaintycomponents,in unitsof percent,associatedwith radianceuniformityin theexit apertureoftheGSFCsphereandthepossiblewavelengtherrorin the SeaWiFS relativespectralresponsivityfunctions.Thenumbersin thesecondrowrefertothenumberoflampsilluminated in theGSFCspheresource. Band Source Uniformity Wavelength Uncertainty Number 16 8 4 1 16 8 4 1 0.09 0.12 0.2 0.4 1.98 1.99 2.00 2.03 0.09 0.12 0.2 0.4 1.46 1.48 1.48 1.49 0.09 0.12 0.2 0.4 1.05 1.06 1.07 1.07 0.09 0.12 0.2 0.4 0.92 0.93 0.93 0.93 0.09 0.12 0.2 0.4 0.69 0.69 0.70 0.70 0.09 0.12 0.2 0.4 0.31 0.32 0.32 0.32 0.09 0.12 0.2 0.4 0.15 0.15 0.15 0.15 0.09 0.12 0.2 0.4 0.02 0.03 0.03 0.03 7.4.5 Modeling Effects The estimation of the band-averaged radiance was done independently by personnel at GSFC and NIST. Slight differences were observed, so the ratio of the two results was included in the uncertainty budget to account for modeling errors; the values were small (Table 29). Table 29. The values for the relative standard uncertainty component, in units of percent, associated with modeling errors. The numbers in the second row refer to the number of lamps illuminated in the GSFC sphere source. Band Modeling Errors Number 16 8 4 1 0.207 0.207 0.212 0.060 0.166 0.165 0.170 0.041 0.121 0.121 0.122 0.009 0.110 0.112 0.112 0.016 0.098 0.096 0.098 0.002 0.086 0.086 0.086 0.005 0.072 0.072 0.073 0.002 0.027 0.027 0.027 0.000 7.5 Discussion Not tabulated are values for the uncertainty associated with the net corrected signal, u(S_ab), because there are 512 values one for each band, channel, gain, and sphere radiance level. The components of u(S_ab) that 24. mity of the radiance in the GSFC sphere's exit aperture only depend on the SeaWiFS band are given in Table The remaining components are associated with the output counts: u(Cout) and u(Cd_rk). These values were larger so for bands 3 6, where the radiance uncertainty component than the other three components given in Table 25, u(S_b) was determined by the measurement precision. As a function of experimental parameters, u(S_b) was generally greater for the cloud channels than for the ocean channels, but was not very sensitive to the gain setting. The values for u(S_b) increased as the sphere radiance decreased (Figs. 6 and 7). Finally, the u(S_ab) values were generally greater for the shorter SeaWiFS measurement wavelengths. The uncertainties in the weighted averages were determined according to the procedures outlined for determining combined standard uncertainty for correlated input quantities (International Organization for Standardization 1993). Figures 6 and 7 illustrate the combined standard uncertainty in K2(g, c) for g 1 for a portion of the bandaveraged results. As mentioned in Sect. 6.3.2, the values for u(K2) in Tables 18 19 were determined using the procedure for determining combined standard uncertainty for correlated input quantities (International Organization for Standardization 1993). As with un(Le) , this treats the uncertainties for the individual K2 (g, c) for each sphere level as systematic uncertainty. (Remember that the final values for each K2 (g, c) are weighted averages of the values corresponding to each sphere level.) This approach was taken because many of the uncertainty components arise from systematic sources. The more correct procedure of assigning a portion of the uncertainty to random and systematic sources before using (21) would be very tedious and time consuming, thus, this approach results in a conservative estimate. Comparing Tables 13 and 19 (or Tables 12 and 18) indicates that the uncertainty in the radiance is a significant contributor to the combined uncertainty; however, for the measurements below 600nm with one lamp illuminated, u(S_b) is greater than the radiance uncertainty component. The uncertainty associated with the uniforis important for the 1 lamp level (Table 28). The nonlinearity uncertainty component (Table 27) is important is smallest. The wavelength error uncertainty component increases with decreasing wavelength (Table 28). The components associated with the short-term repeatability and modeling errors are also largest at the shorter wavelengths, but they are not significant sources of uncertainty (Tables 27 28). 43

Original page 48

The1997PrelaunchRadiometricCalibrationofSeaWiFS 8. CALIBRATION COMPARISON In this section, the different calibrations of SeaWiFS are compared. This includes the clark counts, gain ratios, calibration coefficients, and the ratios of calibration coefficients among bands. The previous results are the manufacturer's calibration of SeaWiFS from the November 1993 measurements. 8.1 Dark Counts and Gain Ratios Table 9 in Barnes et al. (1994b) is analogous to Table 20 in this report, except that in the Barnes et al. paper, the radiance values were corrected for out-of-band response. Comparing the counts at saturation for g 1 shows that the agreement is better than 0.3 counts. This indicates that the dark counts in the standard setting (g 1 and 4:1 TDI setting) remained stable within +0.3 counts from November 1993 to April 1997. Barnes et al. (1994b) reported the same value of the counts at saturation, independent of gain, while this work reports a unique value at each gain. At g 2, g 3, or g 4, the difference between Barnes et al. 1994b and this work is within +1.3 counts for bands 1 4 and within +2.4 counts for bands 58. In Fig. 12, the gain ratios from Table 17 are compared to the previous values in Barnes et al. (1994b). The new gain ratios for g 2, g 3, and g 4 are divided by the previous values, with the cloud channels shown separately from the ocean channels. All of the new gain ratios for the ocean channels agree with the previous values to within +0.3%, which is within the estimated uncertainties stated in Table 17. The gain ratios for the cloud channels agree with the previous values to within about +4%, which is greater than the estimated uncertainty for the gain ratios in this work. The final values for the calibration coefficients determined in this work were used to calculate gain ratios, see (18), and were compared to the values determined from the voltage calibration pulse. For the ocean channels, the ratios of the K2(g, c) coefficients (Table 19) agreed with the corresponding gain ratios in Table 17 to within -t-0.3%. For from its initial calibration in 1993 until the recalibration 1997. Parameters that can change in SeaWiFS include the shape and wavelength of the relative spectral responsivities. The comparison is restricted to g 1. The values based on the SBRC calibration are given in Table 4, and the new values are in Tables 18 19. The new values, divided by the old values, are illustrated in Fig. 13 for the band-averaged method. For the ocean channels, the ratio varies between -3% to +4% as a smooth function of wavelength. The dispersion of the ratios for the three ocean channels in each band is small, from 0.14).5%. The cloud channels follow the same pattern, but with a greater dispersion. This is caused by the reduced sensitivities for these channels. A similar trend is observed if the band-centered calibration coefficients, K2(9, c), are compared, except that the difference is increased by about 1% for band 1 and 0.5% for band 2. Given that the relative standard uncertainty of the original calibration by SBRC was 5%, the results are in agreement within the combined uncertainty of the two methods of calibration. As a verification of the analysis in this work and the relationship between the predicted and observed output from SeaWiFS in the 1:1 and 4:1 TDI setting, a comparison was done for those channels that were below knee 1 for g 1. For each band, the output in the 4:1 TDI setting depends on the calibration coefficients for each channel, with an effective calibration coefficient K2(g) given by 1 1 1 0.25 -- @ -- K2(g) K2(g,l) K2(g, 2) (29) 1 1 @ + K_ (g, 3) K_ (g, 4) The agreement for bands 2 6 was better than +0.2%, and for band I, the agreement was better than 4-0.5% for the three brightest sphere levels (16, 8, and 4 lamps illuminated). The agreement for the 1 lamp configuration was 0.9%. The significance of the stability of the calibration coefficient ratios among the various bands, or color ratios, was the cloud channels, the agreement was worse. For g 3 mentioned in Sect. 6.5. Figure 13 indicates that these color and g 4, the differences are -t-2%, and for g 2, the differences are as large as -t-5% for bands 7 and 8. 8.2 Calibration Coefficients Assuming the sphere sources used at SBRC in 1993 and at OSC in 1997 were calibrated accurately for spectral radiance, then the two experimental calibrations are an indicator of the reproducibility of SeaWiFS for measurements of a laboratory calibration source. Discrepancies greater ratios are different for the 1993 SBRC calibration and the 1997 OSC calibration. For g 1 and the band-averaged method (Table 4), the band-2 to band-5 ratio decreased by about 2.6%, the band-3 to band-5 ratio decreased by about 1.3%, and the band-7 to, and-8 ratio decreased by about 0.8%. 9. DISCUSSION For the ocean channels, the gain ratios are in good than those expected from the estimated uncertainties in agreement with the values determined by SBRC. In Barnes the spectral radiances of the sources and the resolution of SeaWiFS, would indicate that there are differences between the sources, or that SeaWiFS did not remain stable 44 et al. (1994b), the three gain ratios for each ocean channel in a band were averaged, and a single value was reported (compare Tables 5 and 17). Also for the ocean channels,

Original page 49

B.C. ,Johnson, E.A. Early, R.E. Eplee, 1.06 - a) [] ].04' ¢¢3 A [] t'-- 1.02. [] A A @ o @ 1.00. @ 0.98 0.96 I ' I ' I ' 0 5 10 Channel Number 1.007 - b) 1.006 - [] g=2 1.005 @ g=3 ee3 A g=4 1.004 1.003 t--- 1.002 @ [] [] © 1.001 @ .,=o A A []D 1.000 ._= [] A A 0.999 - 0.998 - [] ,Jr., R.A. Barnes, and R.T. Caffrey [] DR [] AA [] g=2 @ _i=3 A g=4 @@ A A @ I ' I ' I ' I ' I 15 20 25 30 35 @ A @ @ [] A A _A [] AD _ [] [] [] [] @A @@ @ [] [] 0.997 - @ 2 @ @ 0.996 - 0.995 1 ' 1 ' 1 ' 0 5 10 Channel Number Fig. 12. Gain ratios for the cloud and ocean channels @ 1 ' 1 ' 1 ' 1 ' 1 15 20 25 30 35 for 9 2, 9 3, and 9 4 normalized by the previous values (Barnes et al. 1994b). The channels for each band are numbered consecutively 1 32, beginning with band 1, channel 1 and ending with band 8, channel 4: a) gain ratios for the cloud channels, and b) gain ratios for the ocean channels. 45

Original page 50

The 1997 Prelaunch Radiometric Calibration of SeaWiFS 1.06 - 1.04 ¢¢3 1.02 F'- 1.00 E 0.98 0.96 0.94 I ' I ' I A /X Cloud Channels [] Ocean Channels ' I ' I ' I ' 400 500 600 700 800 900 Wavelength [nm] Fig. 13. Comparison of the calibration coefficients determined using the band-averaged spectral radiances between this work and the 1993 SBRC calibration for 9 1. The new values are normalized by the previous results. The vertical bars are the uncertainties u(K2) for one of the ocean channels in each band. The relative standard uncertainty in the original calibration is not illustrated, but it was estimated to be about 5%. the values obtained in this work with the voltage calibration pulse agree with those derived from the calibration coefficients. For the cloud channels, the lack of agreement between the SBRC gain ratios and those determined in this work using the voltage calibration pulse is because Barnes et al. (1994b) set the gain ratios to be unity. The justification was the instrument design and the small signal-tonoise ratios, with values of about 75. It is now understood that the gain selection for the cloud channels involves different resistors, and that these resistors are imperfectly matched; therefore, the values for the gain ratios determined in this work are more accurate. The relative disagreement of 2 5% for the cloud channels with the gain ratios determined from the voltage calibration pulse and the band-averaged calibration coefficients is understood in terms of the signal-to-noise ratios of namic range and relative spectral shape of the laboratory (as small as 10 20) that resulted from measurements the GSFC sphere, especially for lamp level 1. The scatter in the calibration coefficients among sphere radiance levels shown in Fig. 6 is typical of the other gain settings. Even with this amount of scatter, the overall agreement in the gain ratios for the cloud channels (determined using the 46 two methods) is better than the agreement between the SBRC and the new values determined from the voltage calibration pulse. The linearity of SeaWiFS was measured on two occasions by SBRC (Sect. 2.4 and in Barnes et al. 1994a) in the standard configuration (9 1 and the 4:1 TDI setting). If the single radiance level that resulted in anomalous results is excluded, the agreement with the new linearity measurements is good; compare Tables 24 25 in Barnes et al. (1994a) with Fig. 11 in this document. The 1997 calibration described here also reports results for the linearity of SeaWiFS for the case where the ocean channels are saturated. For both the SBRC and the OSC measurements of the linearity and calibration coefficients of SeaWiFS, the dysources of spectral radiance limited the results. SeaWiFS was designed to determine the upwelling ocean radiance from measurements of the Earth's radiance from orbit. The difference in the relative spectral shape of the calibration source and the Earth's ocean atmosphere system

Original page 51

B.C.Johnson,E.A.Early,R.E.Eplee,Jr., R.A.Barnes,andR.T.Caffrey is equivalentto the differencebetweena 12,000K black- the SBRC integrating sphere source that was used to calbodyanda 2,850K blackbody(Barnes1996a,andBarnes ibrate SeaWiFS. Two sphere levels were studied; the first andYeh1996).Therefore,at OSC(aswellasSBRC),the was comparable to the GSFC sphere with 16 lamps illumithe knee, nated, and the second was more than twice the radiance. short-wavelengthbandswerenot studiedabove andat g 1, the ocean channels in bands 7 and 8 were not studied below the knee. The overall result is that the linearity of the longer wavelength bands must be inferred from the results for the shorter wavelength bands. The calibration coefficients for the longer wavelength bands must be inferred from measurements at g 4 and the lowest sphere radiance level. The agreement with the new calibration coefficients and the SBRC values can only be assessed for g 1 (Table 4). The new K2(g, c) values, divided by the SBRC values, are shown in Fig. 13 for the band-averaged results. Although the agreement is within the combined uncertainties of the two determinations, there is an obvious spectral dependence the new calibration coefficients for bands 5 8 are greater than the SBRC values, while those for bands 1 and 2 are less. As in this work, the SBRC determination of the calibration coefficients for the ocean channels in bands 7 and 8 involved radiometric measurements at g 4 and electronic measurements of the gain ratios. The difference in the gain ratios (Table 5 and Table 17), however, cannot explain the 4% difference in the calibration coefficients for the ocean channels. The measurement and modeling of the GSFC sphere source that resulted in the spectral radiance values used to determine the calibration coefficients in this report could be the cause of the discrepancy between this work and the SBRC measurements. The sphere was measured with low uncertainty in 1995 by direct comparison to a standard of spectral radiance (Early and Johnson 1997), and the SXR was used to relate the 1995 radiances to the 1997 values. The stability of the SXR over this time interval is, therefore, critical; it was established by comparing two separate calibrations of the SXR (Sect. 5.1). The additional measurements of the GSFC sphere using the SXR at SIRREX-4 in May 1995 and in December 1995 support the argument that the GSFC sphere changed spectrally with time because of changes in the reflectance of the inner surface. The independent measurements of the GSFC sphere at OSC using the 746/ISIC agree with the SXR results within the combined uncertainties (Fig. 3). Another source of the discrepancy is the spectral radiance of the sphere source used at SBRC to calibrate SeaWiFS in 1995. For example, if the spectral radiance at SeaWiFS bands 1 and 2 was overestimated, while for bands 6, 7, and 8 it was underestimated, the results would The agreement of the 746/ISIC results with the SBRC results was within 3% (Appendix B in Mueller 1993). At the same experiment, however, direct comparisons of lamp standards of spectral irradiance from SBRC and GSFC were not consistent given the combined uncertainties, and neither was an SBRC calibration of the GSFC integrating sphere source (Mueller 1993). In August 1996, the 746/ISIC, the SXR, and several other transfer radiometers measured the SBRC sphere at four radiance levels (Butler and Johnson 1996). The internal agreement among the various transfer radiometers was from 1 2%, but not all of the measurements could be compared to SBRC values because they had not calibrated the sphere prior to the intercomparison. During the postcomparison calibration at SBRC, some of the lamps failed, leaving only the lowest radiance level available for comparison. A second intercomparison was held in May 1998; this time, the Visible Transfer Radiometer, which was built by NIST for the NASA Earth Observing System (EOS) was used. A future paper will describe the results of both experiments. Another independent verification of the calibration of the SBRC sphere source is the prelaunch solar-based radiometric calibration using SeaWiFS and its onboard diffusely reflecting panel (Biggar et al. 1994 and Biggar et al. 1995). For these measurements, in October 1992 and November 1993, SeaWiFS was taken outdoors and aligned so the sun illuminated the pane, as it does on orbit. Using ancillary data for the exoatmospheric solar irradiance, and correcting for the effects of the atmosphere, calibration coefficients were determined. The solar-based calibration was compared to the SBRC calibration by calculating the Sea- WiFS output for the solar calibration during flight. The agreement was within -2% to +4%, except for band 8 in the 1992 experiment (Fig. 11 in Biggar et al. 1995). No corrections were made for the sensitivity of SeaWiFS to the relative spectral shape of the source (the out-of-band corrections). If the corrections given in Barnes et al. (1994b) are applied to the sphere-based values, then the agreement is within -1% to +2.5% for the 1993 measurements. The sign of the difference is positive if the sphere-based calibration resulted in increased predicted output from SeaWiFS, which would occur if the SBRC sphere radiance was underestimated. In general, there was no obvious dependence of the differences with wavelength. Several times since its launch in August 1997, Seabe as in Fig. 13. Aging of the SBRC sphere source is not a WiFS measured the radiance of the solar-illuminated onlikely source of the discrepancy because the sphere was calibrated just before the measurements with SeaWiFS. There have been independent assessments of the SBRC methods using the 746/ISIC, the SXR, and other portable transfer radiometers. In October 1992, the 746/ISIC measured board panel. The observed counts are compared to those predicted from the ground-based solar measurements performed in November 1993 (Biggar et al. 1995). Preliminary analysis indicates that the actual measurements agree with the predicted values to within 3%. 47

Original page 52

The 1997 Prelaunch Radiometric Calibration of SeaWiFS 10. SUMMARY An ocean color satellite, SeaWiFS, was calibrated prior to launch at the spacecraft integrator facility before and after the final vibration test. An integrating sphere source, which had been calibrated by NIST almost two years before, was used as the laboratory standard of spectral radiance. The output of the source was observed to have degraded, based on measurements using a portable filter radiometer. The sphere radiance was also measured using a portable monochromator that was calibrated in situ using a lamp standard of spectral irradiance. The measurement equation for SeaWiFS is presented and the sources of uncertainty for the laboratory calibration are clearly identified. For the on-orbit measurements, a similar analysis would apply, but in some cases, additional corrections and the uncertainties associated with them may have to be considered. An example is the sizeof-source correction. None of the new results were corrected for the relative spectral shape of the source; they are referenced to the laboratory calibration source, as in (9). Likewise, no atmospheric corrections were applied. For g 1 and the 1:1 TDI setting, the 1997 calibration coefficients reported in this work agree with the initial 1993 values to within +4%. This is within the mutual combined relative standard uncertainties, which are from 5.1 6%. The results of the 1997 calibration are being used for the SeaWiFS data processing algorithms. There appears to be a spectral dependence in comparing the SBRC and the new calibration coefficients. The source of this spectral dependence is not known. The response of the SeaWiFS channels in each band is linear to better than 1%. As well as could be determined, the associated radiometric parameters, such as, the mirror side correction, dark counts, and gain ratios agree with the manufacturer's calibration. Factors that could not be studied in the new work, which were quantified by SBRC, include the polarization sensitivity, the relative spectral responsivity, and the wavelength scale. ACKNOWLEDGMENTS Many people contributed to this effort. Alan Holmes (formerly of SBRC, now with Santa Barbara Instrument Group) provided overall advice and assistance during the SeaWiFS calibration experiment, and helped to align SeaWiFS to the sphere source. Steve Brown and Howard Yoon of NIST assisted with the data acquisition, documentation, and operation of the SXR; Brown also analyzed the 746/ISIC wavelength calibration data. Chris Eyerman directed the OSC operation of SeaWiFS, and Joe Nvotka and Tim Wiegand operated the SeaWiFS and OrbView- 2 systems. Several other OSC personnel provided valuable assistance, including James Eitnier, Steve Hood, John McCarthy, Howard Runge, and Eric Stucky. John Cooper (formerly of Hughes STX, currently with Kajax Engineering, Inc.) operated the 746/ISIC and provided preliminary results for that instrumerit. Steven Smith (Jackson and Tull, Inc.) and Jack Xiong (Science Systems Applications, Inc.) assisted with the data acquisition. Gene Feldman and Watson Gregg of the SeaWiFS 48 Project assisted with archiving the SeaWiFS results. NIST's participation was made possible by NASA Interagency Agreemerit S-64096-E. GLOSSARY A/D Analog-to-Digital DC Dark Current DMM Digital Multimeter GSE Ground Support Equipment GSFC Goddard Space Flight Center HP Hewlett-Packard FASCAL Facility for Automated Spectroradiometric Calibrations FEL Not an acronym, but a lamp designator. ISIC Integrating Sphere Irradiance Collector NASA National Aeronautics and Space Administration NIST National Institute of Standards and Technology NOAA National Oceanic and Atmospheric Administration OrbView-2 Not an acronym, but the current name for the SeaStar satellite. OSC Orbital Sciences Corporation SBRC Santa Barbara Research Center (Raytheon) SeaStar Not an acronym, but the former name of the satellite on which SeaWiFS was launched, now known as OrbView-2. SeaWiFS Sea-viewing Wide Field-of-view Sensor SIt/REX SeaWiFS Intercalibration Round-Robin Experimerit SIS Spherical Integrating Source SSE Size-of-Source Effect SXR SeaWiFS Transfer Radiometer VAFB Vandenberg Air Force Base SYMBOLS a0 Polynomial coefficient for the chage in the sphere radiance. al Polynomial coefficient for the change in the sphere radiance. a2 Polynomial coefficient for the change in the sphere radiance. a3 Polynomial coefficient for the change in the sphere radiance. b0 Constant for SeaWiFS linearity analysis. Cdark Output of SeaWiFS in counts used for the zero offsets. C,i Net counts for SeaWiFS for mirror side i. Cout Total output of SeaWiFS in counts. CG Time-dependent overall system gain for Sea- WiFS. CO Calibration system offset for SeaWiFS. c Integer to represent a SeaWiFS channel in a given band (values 1, 2, 3, or 4). Fn(a) Correction factor for GSFC sphere radiance for n lamps operating.

Original page 53

B.C. ,Johnson, E.A. Early, R.E. Eplee, ,Jr., R.A. Barnes, and R.T. Caffrey g Integer to represent SeaWiFS gain (values are either 1, 2, 3, or 4). G(g, ) Gain ratios (the gain at setting g normalized by the gain at g 1 and for channel c). i Integer to represent SeaWiFS scan mirror side (values 1 or 2). J Integer index for summation. J Total number of parameters, k Integer index for summation. /42 (g, ) Calibration coefficient for a SeaWiFS band, gain setting g and channel c for either the bandcentered or band-averaged radiance method, depending on context. K2(g) Calibration coefficient for a SeaWiFS band and gain setting g in the 4:1 TDI setting for either the band-centered or band-averaged radiance method, depending on context. /43 Correction factor for SeaWiFS focal plane temperature, Correction factor for SeaWiFS scan modulation. L(A) Spectral radiance, Spectral radiance for the laboratory measurement. Ln,95() Spectral radiance of the GSFC sphere source at NIST in 1995 for n lamps operating, Ln,97() Spectral radiance of the GSFC sphere source at OSC in 1997 for n lamps operating, Ln,97(B) Band-averaged spectral radiance of the GSFC sphere at OSC in 1997 for n lamps operating, Ln,97(AD) Band-centered spectral radiance of the GSFC sphere at OSC in 1997 for n lamps operating. nn,B,97 Band-averaged radiance of the GSFC sphere source at OSC in 1997 for n lamps operating, Lorbit(A) Spectral radiance for the measurement on orbit. L_t (g, c) Saturation radiance for a SeaWiFS configuration (band, gain, and channel). LsBRc () Spectral radiance of the sphere source at SBRC. r_() Spectral radiance for measurement condition denoted x (x lab or orbit). n Integer to represent number of lamps illuminated in the sphere source, Pxl PLxel number, i.e., the numerical designation of a pixel in a scan line. R() Spectral responsivity of an optical instrument. Ri Correction factor for the difference in the reflectance between the two sides of the SeaWiFS scan mirror, i 1 or 2. _lab (g, C)Apr Average corrected net counts for a SeaWiFS band, gain g, and channel c for 11 April 1997. &_b(g, c).j_n Average corrected net counts for a SeaWiFS band, gain g, and channel c for 24 January 1997. _orbit (g, C) Corrected net counts for a SeaWiFS band, gain, and channel on orbit. Corrected net counts for a SeaWiFS band, gain, and channel at saturation. & (g, _) Corrected net counts for a SeaWiFS band, gain g, and channel c for the measurement condition denoted x (x lab or orbit). Time of the SeaWiFS measurement (refers to on-orbit measurements). to Time of the SeaWiFS launch. T Focal plane temperature for SeaWiFS. Tref Reference value for SeaWiFS focal plane temperature. Standard uncertainty. u _(Coud Standard uncertainty in the SeaWiFS output counts when observing an illuminated source. u(Cd_rk) Standard uncertainty in the SeaWiFS output counts when observing a dark source. u(K2) Standard uncertainty in the SeaWiFS calibration coefficient. U(/3) Standard uncertainty in the SeaWiFS temperature correction coefficient. Standard uncertainty in spectral radiance. u(L(A)) un(LB) Standard uncertainty in the band-averaged spectral radiance for n lamps operating. u(kB) Relative standard uncertainty for Ln,97(_B). un(LD) Standard uncertainty in the band-averaged spectral radiance for n lamps operating. u D (AD) Relative standard uncertainty for Ln,97(D). uL Relative standard uncertainty for L(). u( Ri ) Standard uncertainty in the mirror side correction. u(R(A)) Standard uncertainty in the SeaWiFS relative spectral responsivity. u(&b) Standard uncertainty in the corrected net counts from SeaWiFS. u(T) Standard uncertainty in the SeaWiFS focal plane temperature. u(zj) Standard uncertainty in parameter zj. u(zj, zk) Covariance associated with parameters zj and Zk. u(A) Standard uncertainty in the wavelength accuracy of the R() values. xn Net voltage for the SXFI. for n lamps operating (for the SeaWiFS linearity analysis). Corrected net counts for SeaWiFS for n lamps yn operating (for the SeaWiFS linearity analysis). zj Parameter j in the SeaWiFS measurement equation. % (g, c) Measurement constant for a SeaWiFS band, gain g, and channel c for the measurement condition denoted x (x lab or orbit). AC Defined as Cout - Cdark. AT Defined as T - _ref. AA Bandwidth parameter. Wavelength. _B Band-averaged center wavelength for SeaWiFS. Asxa Measurement wavelength for the SXR. fi Average reflectance of the SeaWiFS scan mirror. pi Reflectance of the side i of the SeaWiFS scan mirror, where i equals 1 or 2. 49

Original page 54

The1997PrelaunchRadiometricCalibrationofSeaWiFS REFERENCES Barnes,R.A.,1994:SeaWiFS Data: Actual and Simulated. [World Wide Web page.] From UR.Ls: http://seawifs •gsf c.nasa• gov/SEAWIFS/IMAGES/spe ctral, dat and /spectra2.dat NASA Goddard Space Flight Center, Greenbelt, Maryland. --, 1996a: "Calculation of an equivalent blackbody temperature for the GSFC sphere." In: Barnes, R.A., E-n. Yeh, and R.E. Eplee, 1996: SeaWiFS Calibration Topics, Part 1. NASA Tech. Memo. 104566, Vol. 39, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 5 17. --, 1996b: "A comparison of the spectral responses of Sea- WiFS and the SeaWiFS Transfer Radiometer." In: Barnes, R..A., E-n. Yeh, and R..E. Eplee, 1996: SeaWiFS Calibration Topics, Part 1. NASA Tech. Memo. 104566, Vol. 39, S.B. Hooker and E.R.. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 39 48. --, 1996c: "SeaWiFS center wavelengths." In: Barnes, R.A., E-n. Yeh, and R..E. Eplee, 1996: SeaWiFS Calibration Topics, Part 1. NASA Tech. Memo. 104566, Vol. 39, S.B. Hooker and E.R.. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 49 53. --, and A.W. Holmes, 1993: Overview of the SeaWiFS ocean sensor. Proc. SPIE, 1939. 224 232. --, W.L. Barnes, W.E. Esaias, and C.R. McClain, 1994a: Prelaunch Acceptance Report for the SeaWiFS Radiometer. NASA Tech. Memo. 104566, Vol. 22, S.B. Hooker, E.R. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 32 pp. --, A.W. Holmes, W.L. Barnes, W.E. Esaias, C.R. McClain, and T. Svitek, 1994b: SeaWiFS Prelaunch Radiometric Calibration and Spectral Characterization. NASA Tech. Memo. 104566, Vol. 23, S.B. Hooker, E.R. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 55 pp. , , and W.E. Esaias, 1995: Stray Light in the SeaWiFS Radiometer. NASA Tech. Memo. 104566, Vol. 31, S.B. Hooker, E.R.. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 76 pp. --, and E-n. Yeh, 1996: "Effects of source spectral shape in SeaWiFS radiance measurements." In: Barnes, R.A., E-n. Yeh, and R.E. Eplee, 1996: SeaWiFS Calibration Topics, Part 1. NASA Tech. Memo. 104566, Vol. 39, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 18 38. --, and R.E. Eplee, 1997: "The 1993 SeaWiFS calibration using band-averaged spectral radiances." In: Barnes, R.A., R.E. Eplee, E-n. Yeh, and W.E. Esaias, 1997: SeaWiFS Calibration Topics, Part 2. NASA Tech. Memo. 104566, Vol. 40, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 39 46. 5O Biggar, S.F., P.N. Slater, K.J. Thome, A.W. Holmes, and R..A. Barnes, 1994: "Preflight solar-based calibration of Sea- WiFS." In: McClain, C.R.., R..S. Fraser, J.T. McLean, M. Darzi, J.K. Firestone, F.S. Patt, B.D. Schieber, R..H. Woodward, E-n. Yeh, S. Mattoo, S.F. Biggar, P.N. Slater, K.J. Thome, A.W. Holmes, R..A. Barnes, and K.J. Voss, 1994: Case Studies for SeaWiFS Calibration and Validation, Part 2. NASA Tech. Memo. 104566, Vol. 19, S.B. Hooker, E.R.. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 25 32. --, K.J. Thome, P.N. Slater, A.W. Holmes, and R.A. Barnes, 1995: "Second SeaWiFS preflight solar radiation-based calibration experiment." In: Mueller, J.L., R..S. Fraser, S.F. Biggar, K.J. Thome, P.N. Slater, A.W. Holmes, R..A. Barnes, C.T. Weir, D.A. Siegel, D.W. Menzies, A.F. Michaels, and G. Podesta, 1995: Case Studies for SeaWiFS Calibration and Validation, Part 3. NASA Tech. Memo. 104566, Vol. 2Z S.B. Hooker, E.R.. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 20 24. Butler, J.J., and B.C. Johnson, 1996: EOS radiometric measurement comparisons at Hughes Santa Barbara Remote Sensing and NASA's Jet Propulsion Laboratory. The Earth Observer, 8(5), 17 19. Early, E.A., and B.C. Johnson, 1997: "Calibration and characterization of the GSFC sphere." In: Yeh, E-n., R.A. Barnes, M. Darzi, L. Kumar, E.A. Early, B.C. Johnson, J.L. Mueller, and C.C. Trees, 1997: Case Studies for Sea- WiFS Calibration and Validation, Part 4. NASA Tech. Memo. 104566, Vol. 41, S.B. Hooker, and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt Maryland, 3 17. Eplee, R.E., and R.A. Barnes, 1997: "The SeaWiFS temperature calibration." In: Barnes, R.A., R.E. Eplee, E-n. Yeh, and W.E. Esaias, 1997: SeaWiFS Calibration Topics, Part 2. NASA Tech. Memo. 104566, Vol. 40, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 56 62. Hooker, S.B., W.E. Esaias, G.C. Feldman, W.W. Gregg, and C.R.. McClain, 1992: An Overview of SeaWiFS and Ocean Color. NASA Tech. Memo. 104566, Vol. 1, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 24 pp., plus color plates. International Organization for Standardization, 1993: Guide to the Expression of Uncertainty in Measurement, International Organization for Standardization, Geneva, Switzerland, 101 pp. Johnson, B.C., S.S. Bruce, E.A. Early, J.M. Houston, T.R.. O'Brian, A. Thompson, S.B. Hooker, and J.L. Mueller, 1996: The Fourth SeaWiFS Intercalibration Round-Robin Experiment (SIRREX-4), May 1995. NASA Tech. Memo. 104566, Vol. 3Z S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 65 pp.

Original page 55

B.C.,Johnson,E.A.Early,R.E.Eplee,,Jr.,R.A.Barnes,andR.T.Caffrey --, F. Sakuma,J.J. Butler,S.F.Biggar,J.W. Coop-Yeh, E-n., M. Darzi, and L. Kumar, 1997: "SeaWiFS stray er,J. Ishida,andK. Suzuki,1997:Radiometricmeasure- light correction algorithm." In Yeh, E-n., R..A. Barnes, M. mentcomparisonusingtheOceanColorTemperatureScan- Darzi, L. Kumar, E.A. Early, B.C. Johnson, J.L. Mueller, ner(OCTS)visibleandnearinfraredintegratingsphere.J. Res. Natl. Inst. Stand. Technol., 102, 62_646. --, J.B. Fowler, and C.L. Cromer, 1998: The SeaWiFS Transfer Radiometer (SXR). NASA Teeh. Memo. 1998 206892, Vol. 1, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 58 pp. Mueller, J.L., 1993: The First SeaWiFS Intercalibration Round- Robin Experiment, SIRREX-1, July 1992. NASA Teeh. Memo. 104566, Vol. 14, S.B. Hooker and E.R.. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 60 pp. --, and R.W. Austin, 1995: Ocean Optics Protocols for Sea- WiFS Validation, Revision 1. NASA Teeh. Memo. 104566, Vol. 25, S.B. Hooker, E.R. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 67 pp. --, B.C. Johnson, C.L. Cromer, S.B. Hooker, J.T. McLean, and S.F. Biggar, 1996: The Third SeaWiFS Intercalibration Round-Robin Experiment (SIR.R.EX-3), 19 30 September 1994. NASA Teeh. Memo. 104566, Vol. 34, S.B. Hooker, E.R. Firestone, and J.G. Acker, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 78 pp. Taylor, B.N., and C.E. Kuyatt, 1994: Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results. NIST Teeh. Note 1297, US Dept. of Commerce, National Institute of Standards and Technology, Gaithersburg, Maryland, 20 pp. Tsai, B.K., and B.C. Johnson, 1998: Radiometric Traceability for Fundamental Measurements: Estimation and Evaluation of Combined Standard Uncertainties, Metrologia, 35, 58_593. Woodward, R.H., R.A. Barnes, C.R. McClain, W.E. Esaias, W.L. Barnes, and A.T. Mecherikunnel, 1993: Modeling of the SeaWiFS Solar and Lunar Observations. NASA Teeh. Memo. 104566, Vol. 10, S.B. Hooker and E.R.. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 26 pp. and C.C. Trees, 1997: Case Studies for SeaWiFS Calibration and Validation, Part 4. NASA Tech. Memo. 104566, Vol. 41, S.B. Hooker, and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 24 30. THE SEAWIFS POSTLAUNCH TECHNICAL REPORT SERIES Vol. 1 Johnson, B.C., J.B. Fowler, and C.L. Cromer, 1998: The Sea- WiFS Transfer Radiometer (SXR). NASA Teeh. Memo. 1998 206892, Vol. 1, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 58 pp. Vol. 2 Aiken, J., D.G. Cummings, S.W. Gibb, N.W. Rees, R. Woodd- Walker, E.M.S. Woodward, J. Woolfenden, S.B. Hooker, J-F. Berthon, C.D. Dempsey, D.J. Suggett, P. Wood, C. Donlon, N. Gonz£1ez-Bemtez, I. Huskin, M. Quevedo, R. Barciela-Fernandez, C. de Vargas, and C. McKee, 1998: AMT-5 Cruise Report. NASA Tech. Memo. 199_206892, Vol. 2, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 113 pp. Vol. 3 Hooker, S.B., G. Zibordi, G. Lazin, and S. McLean, 1999: The SeaBOARR-98 Field Campaign. NASA Tech. Memo. 1999 206892, Vol. 3, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 40 pp. Vol. 4 Johnson, B.C., R.E. Eplee, Jr., R.A. Barnes, E.A. Early, and R.T. Caffrey, 1999: The 1997 Prelaunch Radiometric Calibration of SeaWiFS. NASA Tech. Memo. 1999 206892, Vol. 4, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard Space Flight Center, Greenbelt, Maryland, 51pp. 51

Original page 56

REPORT DOCUMENTATION Form Approved PAGE OMBNo.0704-0188 Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503. 1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE April 1999 4. TITLE AND SUBTITLE SeaWiFS Postlaunch Technical Report Series Volume 4: The 1997 Prelaunch Radiometric Calibration 6. AUTHOR(S) 3. REPORT TYPE AND DATES COVERED Technical Memorandum 5. FUNDING NUMBERS of SeaWiFS Code 970.2 B. Carol Johnson, Edward E. Early, Robert E. Eplee, Jr., Robert A. Barnes and Robert T. Caffrey Series Editors: Stanford B. Hooker and Elaine R. Firestone 7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS (ES) Laboratory for Hydrospheric Processes Goddard Space Flight Center Greenbelt, Maryland 20771 9. SPONSORING / MONITORING AGENCY NAME(S) AND ADDRESS National Aeronautics and Space Administration Washington, DC 20546-0001 11. SUPPLEMENTARY NOTES 8. PEFORMING ORGANIZATION REPORT NUMBER 99B00036 10. SPONSORING / MONITORING (ES) AGENCY REPORT NUMBER TM--1998-206892, Vol.4 E.R. Firestone, R.E. Eplee, Jr., and R.A. Barnes: SAIC General Sciences Corporation, Beltsville, Maryland; and B.C. Johnson and E.E. Early: National Institute of Standards and Technology, Gaithersburg, Maryland 12a. DISTRIBUTION / AVAILABILITY STATEMENT Unclassified-Unlimited Subject Category: 48 Report available from the NASA Center for AeroSpace 12b. DISTRIBUTION CODE Information, 7121 Standard Drive, Hanover, MD 21076-1320. (301) 621-0390. 13. ABSTRACT (Maximum 200 words) The Sea-viewing Wide Field-of-view Sensor (SeaWiFS) was originally calibrated by the instrument's manufacturer, Santa Barbara Research Center (SBRC), in November 1993. In preparation for an August 1997 launch, the SeaWiFS Project and the National Institute of Standards and Technology (NIST) undertook a second calibration of SeaWiFS in January and April 1997 at the facility of the spacecraft integrator, Orbital Sciences Corporation (OSC). This calibration occurred in two phases, the first after the final thermal vacuum test, and the second after the final vibration test of the spacecraft. For the calibration, SeaWiFS observed an integrating sphere from the National Aeronautics and Space Administration (NASA) Goddard Space Flight Center (GSFC) at four radiance levels. The spectral radiance of the sphere at these radiance levels was also measured by the SeaWiFS Transfer Radiometer (SXR). In addition, during the calibration, SeaWiFS and the SXR observed the sphere at 16 radiance levels to determine the linearity of the SeaWiFS response. As part of the calibration analysis, the GSFC sphere was also characterized using a GSFC spectroradiometer. The 1997 calibration agrees with the initial 1993 calibration to within +4%. The new calibration coefficients, computed before and after the vibration test, agree to within 0.5%. The response of the SeaWiFS channels in each band is linear to better than 1%. In order to compare to previous and current methods, the SeaWiFS radiometric responses are presented in two ways: using the nominal center wavelengths for the eight bands; and using band-averaged spectral radiances. The band-averaged values are used in the flight calibration table. An uncertainty analysis for the calibration coefficients is also presented. 14. SUBJECT TERMS 15. NUMBER OF PAGES SeaWiFS, Oceanography, Radiometric Calibration, SXR, Transfer Radiometer, Integrating Sphere Source 17. SECURITY CLASSIFICATION 18. SECURITY CLASSIFICATION OF REPORT OF THIS PAGE Unclassifted Unclassifted NSN 7540-01-280-5500 16. PRICE CODE 19. SECURITY CLASSIFICATION 20. LIMITATION OF ABSTRACT OF ABSTRACT Unclassifted UL Standard Form 298 (Rev. 2-89) Prescribed by ANSI Std. Z39.18 298-102