P. Jeremy Werdell,1,*
Sean W. Bailey,1
Bryan A. Franz,1
André Morel,2
and Charles R. McClain1
1NASA Goddard Space Flight Center, 614.2, Greenbelt, Maryland 20771, USA
2Laboratoire d'Océanographie de Villefranche, Université Pierre et Marie Curie and Centre National de la Recherche Scientifique, Villefranche-sur-Mer, France
P. Jeremy Werdell, Sean W. Bailey, Bryan A. Franz, André Morel, and Charles R. McClain, "On-orbit vicarious calibration of ocean color sensors using an ocean surface reflectance model," Appl. Opt. 46, 5649-5666 (2007)
Recent advances in global biogeochemical research demonstrate a critical need for long-term ocean color satellite data records of consistent high quality. To achieve that quality, spaceborne instruments require on-orbit vicarious calibration, where the integrated instrument and atmospheric correction system is adjusted using in situ normalized water-leaving radiances, such as those collected by the marine optical buoy (MOBY). Unfortunately, well-characterized time-series of in situ data are scarce for many historical satellite missions, in particular, the NASA coastal zone color scanner (CZCS) and the ocean color and temperature scanner (OCTS). Ocean surface reflectance models (ORMs)
accurately reproduce spectra observed in clear marine waters, using only chlorophyll
as input, a measurement for which long-term in situ time series exist. Before recalibrating CZCS and OCTS using modeled radiances, however, we evaluate the approach with the Sea-viewing Wide-Field-of-view Sensor (SeaWiFS). Using annual
climatologies as input into an ORM, we derive SeaWiFS vicarious gains that differ from the operational MOBY gains by less than
spectrally. In the context of generating decadal
climate data records, we quantify the downstream effects of using these modeled gains by generating satellite-to-in situ data product validation statistics for comparison with the operational SeaWiFS results. Finally, we apply these methods to the CZCS and OCTS ocean color time series.
Robert E. Eplee, Wayne D. Robinson, Sean W. Bailey, Dennis K. Clark, P. Jeremy Werdell, Menghua Wang, Robert A. Barnes, and Charles R. McClain Appl. Opt. 40(36) 6701-6718 (2001)
You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.
SeaWiFS and Standard Deviationsa Calculated for MOBY and the ORM at the BATS and HOTS Sites
N
412
443
490
510
555
670
MOBY
150 (42)
1.0377 (0.009)
1.0140 (0.009)
0.9927 (0.008)
0.9993 (0.009)
1.0002 (0.008)
0.9738 (0.007)
BATS
241 (45)
1.0345 (0.018)
1.0020 (0.016)
0.9814 (0.013)
0.9941 (0.011)
1.0016 (0.011)
0.9731 (0.006)
HOT
176 (45)
1.0300 (0.015)
1.0086 (0.012)
0.9879 (0.009)
0.9979 (0.008)
1.0046 (0.009)
0.9718 (0.006)
BATS + HOT
417 (90)
1.0323 (0.017)
1.0053 (0.015)
0.9847 (0.012)
0.9960 (0.010)
1.0031 (0.010)
0.9725 (0.006)
In parentheses, with the exception of N, where we report the number of samples remaining after application of the semi-interquartile filter. Only these remaining samples are used to calculate the combined BATS + HOT _bar.
Table 2
SeaWiFS Calibration Verification Statistics for the Scenes Used to Derive
The median satellite-to-in situ ratio.
The absolute median percent difference (relative to the in situ observations).
The root mean square (standard deviation).
The average signed difference between the satellite and in situ observations (= Σ(satellite − in situ)∕N).
The median satellite-to-in situ ratio (with standard deviation).
The absolute median percent difference (relative to the in situ observations).
The slope of the reduced major axis linear regression (with standard error).
The root of the residual mean square (in units equal to those of the observations).
The Ca data were transformed prior to the regression analysis to account for their lognormal distribution.
*Indicates the slope and intercept (not shown) are statistically equal to 1 and 0, respectively, via a Student's t analysis at α = 0.05.
Table 6
SeaWiFS Validation Statistics for the Data Set Presented in Table 5 Using the ORM-Derived
N
Ratio (±SD)
MPD
Slope (±SE)
r2
RMSE
Lwn(412)
197
0.924 (0.25)
12.4
1.08 (0.02)
0.92
0.213
Lwn(443)
332
0.843 (0.22)
18.4
1.02 (0.02)*
0.86
0.229
Lwn(490)
332
0.863 (0.17)
15.3
0.93 (0.03)
0.80
0.157
Lwn(510)
172
0.931 (0.16)
12.3
1.14 (0.08)*
0.55
0.104
Lwn(555)
332
1.018 (0.24)
15.6
0.72 (0.02)
0.82
0.061
Lwn(670)
320
1.325 (1.86)
69.0
1.07 (0.05)
0.59
0.022
Ca
161
1.232 (0.90)
30.1
0.96 (0.03)
0.87
0.475
*Indicates the slope and intercept (not shown) are statistically equal to 1 and 0, respectively, via a Student's t analysis at .
Table 7
Nine-Year Means for the SeaWiFS Deep-Water and Trophic Time-Series Generated Using the MOBY and ORM
CZCS and Standard Deviationsa (in Parentheses) Calculated Using the ORM at the BATS Site
N
443
520
550
670
7
1.0094 (0.031)
0.9525 (0.019)
0.9543 (0.024)
1.008 (NA)
With the exception of 670 nm, as the EG94 value was adopted for our analysis.
Table 9
OCTS and Standard Deviations (in Parentheses) Calculated Using the ORM at the BATS and HOT Sites
N
412
443
490
520
565
670
45
1.1684 (0.016)
1.0453 (0.014)
0.9867 (0.013)
1.0294 (0.011)
1.0370 (0.010)
1.0567 (0.015)
Tables (9)
Table 1
SeaWiFS and Standard Deviationsa Calculated for MOBY and the ORM at the BATS and HOTS Sites
N
412
443
490
510
555
670
MOBY
150 (42)
1.0377 (0.009)
1.0140 (0.009)
0.9927 (0.008)
0.9993 (0.009)
1.0002 (0.008)
0.9738 (0.007)
BATS
241 (45)
1.0345 (0.018)
1.0020 (0.016)
0.9814 (0.013)
0.9941 (0.011)
1.0016 (0.011)
0.9731 (0.006)
HOT
176 (45)
1.0300 (0.015)
1.0086 (0.012)
0.9879 (0.009)
0.9979 (0.008)
1.0046 (0.009)
0.9718 (0.006)
BATS + HOT
417 (90)
1.0323 (0.017)
1.0053 (0.015)
0.9847 (0.012)
0.9960 (0.010)
1.0031 (0.010)
0.9725 (0.006)
In parentheses, with the exception of N, where we report the number of samples remaining after application of the semi-interquartile filter. Only these remaining samples are used to calculate the combined BATS + HOT _bar.
Table 2
SeaWiFS Calibration Verification Statistics for the Scenes Used to Derive
The median satellite-to-in situ ratio.
The absolute median percent difference (relative to the in situ observations).
The root mean square (standard deviation).
The average signed difference between the satellite and in situ observations (= Σ(satellite − in situ)∕N).
The median satellite-to-in situ ratio (with standard deviation).
The absolute median percent difference (relative to the in situ observations).
The slope of the reduced major axis linear regression (with standard error).
The root of the residual mean square (in units equal to those of the observations).
The Ca data were transformed prior to the regression analysis to account for their lognormal distribution.
*Indicates the slope and intercept (not shown) are statistically equal to 1 and 0, respectively, via a Student's t analysis at α = 0.05.
Table 6
SeaWiFS Validation Statistics for the Data Set Presented in Table 5 Using the ORM-Derived
N
Ratio (±SD)
MPD
Slope (±SE)
r2
RMSE
Lwn(412)
197
0.924 (0.25)
12.4
1.08 (0.02)
0.92
0.213
Lwn(443)
332
0.843 (0.22)
18.4
1.02 (0.02)*
0.86
0.229
Lwn(490)
332
0.863 (0.17)
15.3
0.93 (0.03)
0.80
0.157
Lwn(510)
172
0.931 (0.16)
12.3
1.14 (0.08)*
0.55
0.104
Lwn(555)
332
1.018 (0.24)
15.6
0.72 (0.02)
0.82
0.061
Lwn(670)
320
1.325 (1.86)
69.0
1.07 (0.05)
0.59
0.022
Ca
161
1.232 (0.90)
30.1
0.96 (0.03)
0.87
0.475
*Indicates the slope and intercept (not shown) are statistically equal to 1 and 0, respectively, via a Student's t analysis at .
Table 7
Nine-Year Means for the SeaWiFS Deep-Water and Trophic Time-Series Generated Using the MOBY and ORM