Ocean color chlorophyll a algorithms for SeaWiFS, OC2, and OC4: Version 4

Full text

NASA Technical Memorandum 2000–206892, Volume 11
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 11, SeaWiFS Postlaunch Calibration and
Validation Analyses, Part 3
John E. O’Reilly
NOAA, National Marine Fisheries Service, Narragansett, Rhode Island
St´ephane Maritorena, Margaret C. O’Brien, David A. Siegel, Dierdre Toole,
David Menzies, and Raymond C. Smith
University of California at Santa Barbara, Santa Barbara, California
James L. Mueller
CHORS/San Diego State University, San Diego, California
B. Greg Mitchell and Mati Kahru
Scripps Institution of Oceanography, San Diego, California
Francisco P. Chavez and P. Strutton
Monterey Bay Aquarium Research Institute, Moss Landing, California
Glenn F. Cota
Old Dominion University, Norfolk, Virginia
Stanford B. Hooker and Charles R. McClain
NASA Goddard Space Flight Center, Greenbelt, Maryland
Kendall L. Carder and Frank M¨uller-Karger
University of South Florida, St. Petersburg, Florida
Larry Harding and Andrea Magnuson
Horn Point Laboratory, Cambridge, Maryland
David Phinney
Bigelow Laboratory for Ocean Sciences, West Boothbay Harbor, Maine
Gerald F. Moore and James Aiken
Plymouth Marine Laboratory, Plymouth, United Kingdom
Kevin R. Arrigo
Stanford University, Stanford, California
Ricardo Letelier
Oregon State University, Corvallis, Oregon
Mary Culver
NOAA, Coastal Services Center, Charleston, South Carolina

O’Reilly et al.
Table of Contents
Prologue .................................................................................................1
1. OC2v2: Update on the Initial Operational SeaWiFS Chlorophyll aAlgorithm ..........................3
1.1 Introduction .........................................................................................3
1.2 The SeaBAM Data Set ..............................................................................3
1.3 OC2v2 Algorithm ....................................................................................7
1.4 Conclusions ..........................................................................................8
2. Ocean Color Chlorophyll aAlgorithms for SeaWiFS, OC2, and OC4: Version 4 ........................9
2.1 Introduction ........................................................................................10
2.2 The In Situ Data Set ...............................................................................10
2.3 OC2 and OC4 ......................................................................................15
2.4 Conclusions ........................................................................................19
3. SeaWiFS Algorithm for the Diffuse Attenuation Coefficient, K(490), Using Water-Leaving ...........24
Radiances at 490 and 555 nm
3.1 Introduction ........................................................................................24
3.2 Data and Methods ................................................................................. 25
3.3 Results .............................................................................................25
3.4 Discussion ..........................................................................................25
4. Long-Term Calibration History of Several Marine Environmental Radiometers (MERs) ..............28
4.1 Introduction ........................................................................................28
4.2 ICESS Facility and Methods ........................................................................28
4.3 Results .............................................................................................33
4.4 Long-Term Averages ............................................................................... 41
4.5 Other Issues ........................................................................................43
4.6 Conclusions ........................................................................................45
Glossary ...............................................................................................46
Symbols ................................................................................................46
References ............................................................................................47
The SeaWiFS Postlaunch Technical Report Series ..............................................48
i

O’Reilly et al.
Abstract
Volume 11 continues the sequential presentation of postlaunch data analysis and algorithm descriptions begun
in Volume 9. Chapters 1 and 2 present the OC2 (version 2) and OC4 (version 4) chlorophyll aalgorithms used in
the SeaWiFS data second and third reprocessings, August 1998 and May 2000, respectively. Chapter 3 describes
a revision of the K(490) algorithm designed to use water-leaving radiances at 490nm which was implemented for
the third reprocessing. Finally, Chapter 4 is an analysis of in situ radiometer calibration data over several years
at the University of California, Santa Barbara (UCSB) to establish the temporal consistency of their in-water
optical measurements.
PROLOGUE
The SeaWiFS Project Calibration and Validation Team
(CVT) is responsible for the overall quality of the data
products and for verifying the processing code. The pre-
launch quality control strategy was outlined in Volume
38 of the SeaWiFS Technical Report Series (Prelaunch).
Since SeaWiFS began routine data processing in Septem-
ber 1997, the CVT has constantly worked to resolve data
quality issues and improve on the initial data evaluation
methodologies. These evaluations resulted in three major
reprocessings of the entire data set (February 1998, Au-
gust 1998, and May 2000). Each reprocessing addressed
the data quality issues that could be identified up to the
time of each reprocessing.
The number of chapters (21) needed to document this
extensive work in the SeaWiFS Postlaunch Technical Re-
port Series requires three volumes: Volumes 9, 10, and 11.
Volume 11 continues the sequential presentation of post-
launch data analysis and algorithm descriptions, begun
in Volume 9, by describing the algorithm improvements
to two versions of the chlorophyll aalgorithm and the re-
vised diffuse attenuation coefficient algorithm at 490 nm,
K(490), developed for the third reprocessing. In addition,
an analysis of radiometer calibration data at the University
of California Santa Barbara (UCSB) is described, which es-
tablishes the temporal consistency of their in-water optical
measurements.
It is expected that other improvements, including new
geophysical data products, and updated algorithms will
be developed in the future which will require additional
reprocessings. The SeaWiFS Project Office will remain
dedicated to providing better products and to the docu-
mentation of future analysis and algorithm improvement
studies.
A short synopsis of each chapter in this volume is given
below.
1. OC2v2: Update on the Initial Operational
SeaWiFS Chlorophyll aAlgorithm
The original at-launch SeaWiFS algorithm (OC2 for
Ocean Chlorophyll 2-band algorithm) was derived from the
SeaWiFS Bio-optical Algorithm Mini-workshop (SeaBAM)
data set (the number of data sets, N= 919) which con-
tains coincident in situ remote sensing reflectance, ˜
Rrs,
and in situ chlorophyll a,˜
Ca, measurements from a vari-
ety of oceanic provinces. Following the SeaWiFS launch,
the accuracy of SeaWiFS chlorophyllaestimates using the
OC2 algorithm was evaluated against new in situ measure-
ments. These new data indicated that OC2 was perform-
ing generally well in Case-1 waters with ˜
Caconcentration,
between 0.03–1 mg m−3, but tended to overestimate ˜
Caat
higher concentrations. To strengthen the SeaBAM data
set at ˜
Ca>1mgm
−3, 255 new stations were added to the
original data set. These new data generally showed lower
Rrs(490)/Rrs (555) band ratios at ˜
Ca>4mgm
−3than in
the original SeaBAM data set, which would explain some of
the overestimations observed with OC2. The new SeaBAM
data set was used to refine the coefficients for the OC2
modified cubic polynomial (MCP) function. The updated
algorithm (OC2v2) is presented along with its statistical
performance and a comparison with the original version of
the algorithm.
2. Ocean Color Chlorophyll aAlgorithms for SeaWiFS,
OC2, and OC4: Version 4
This chapter describes the revisions (version 4) to the
ocean chlorophyll two- and four-band algorithms as well as
the very large in situ data set used to update these algo-
rithms for use in the third reprocessing of SeaWiFS data.
The in situ data set is substantially larger (N=2,853)
than was used to develop earlier versions of OC2 and OC4.
The data set includes samples from a greater variety of bio-
optical provinces, and better represents oligotrophic and
eutrophic waters. The correlation between chlorophylla
concentration, Ca, estimated using OC4 and in situ Ca
(˜
Ca) estimated from fluorometric and high performance
liquid chromatography (HPLC) analyses was slightly high-
er than that for OC2. OC4 would be expected to perform
better than OC2, when applied to satellite-derived, water-
leaving radiances retrieved from oligotrophic and eutrophic
areas. Variations of the OC4 algorithm are provided for
other ocean color sensors to facilitate comparisons with
SeaWiFS.
1

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
3. SeaWiFS Algorithm for the Diffuse Attenuation
Coefficient, K(490), Using Water-Leaving
Radiances at 490 and 555 nm
A new algorithm has been developed using the ratio
of water-leaving radiances at 490 and 555nm to estimate
K(490), the diffuse attenuation coefficient of seawater at
490 nm. The standard uncertainty of prediction for the
new algorithm is statistically identical to that of the Sea-
WiFS prelaunch K(490) algorithm, which uses the ratio
of water-leaving radiances at 443 and 490nm. The new
algorithm should be used whenever the uncertainty of the
SeaWiFS determination of water-leaving radiance at 443
is larger than that at 490 nm.
4. Long-Term Calibration History of Several
Marine Environmental Radiometers (MERs)
The accuracy of upper ocean apparent optical prop-
erties (AOPs) for the vicarious calibration of ocean color
satellites ultimately depends on accurate and consistent in
situ radiometric data. The Sensor Intercomparison and
Merger for Biological and Interdisciplinary Oceanic Stud-
ies (SIMBIOS) project is charged with providing estimates
of normalized water-leaving radiance for the SeaWiFS in-
strument to within 5%. This, in turn, demands that the ra-
diometric stability of in situ instruments be within 1% with
an absolute accuracy of 3%. This chapter is a report on
the analysis and reconciliation of the laboratory calibration
history for several Biospherical Instruments (BSI) marine
environmental radiometers (MERs), models MER-2040
and -2041, three of which participate in the SeaWiFS Cal-
ibration and Validation Program. This analysis includes
data using four different FEL calibration lamps, as well
as calibrations performed at three SeaWiFS Intercalibra-
tion Round-Robin Experiments (SIRREXs). Barring a few
spectral detectors with known deteriorating responses, the
radiometers used by the University of California, Santa
Barbara (UCSB) during the Bermuda Bio-Optics Project
(BBOP) have been remarkably stable during more than
five years of intense data collection. Coefficients of varia-
tion for long-term averages of calibration slopes, for most
detectors in the profiling instrument, were less than 1%.
Long-term averages can be applied to most channels, with
deviations only after major instrument upgrades. The
methods used here to examine stability accommodate the
addition of new calibration data as they become available.
This enables researchers to closely track any changes in the
performance of these instruments and to adjust the cali-
bration coefficients accordingly. This analysis may serve as
a template for radiometer histories which will be cataloged
by the SIMBIOS Project.
2

O’Reilly et al.
Chapter 1
OC2v2: Update on the Initial Operational
SeaWiFS Chlorophyll aAlgorithm
St´
ephane Maritorena
ICESS/University of California at Santa Barbara
Santa Barbara, California
John E. O’Reilly
NOAA National Marine Fisheries Service
Narragansett, Rhode Island
Abstract
The original at-launch SeaWiFS algorithm (OC2 for Ocean Chlorophyll 2-band algorithm) was derived from
the SeaBAM data set (N= 919) which contains coincident remote sensing reflectance, ˜
Rrs, and in situ chlo-
rophyll a,˜
Ca, measurements from a variety of oceanic provinces. Following the SeaWiFS launch, the accuracy
of SeaWiFS chlorophyllaestimates using the OC2 algorithm was evaluated against new in situ measurements.
These new data indicated that OC2 was performing generally well in Case-1 waters with ˜
Caconcentration,
between 0.03–1 mg m−3, but tended to overestimate ˜
Caat higher concentrations. To strengthen the SeaBAM
data set at ˜
Ca>1mgm
−3, 255 new stations were added to the original data set. These new data generally
showed lower Rrs(490)/Rrs(555) band ratios at ˜
Ca>4mgm
−3than in the original SeaBAM data set, which
would explain some of the overestimations observed with OC2. The new SeaBAM data set was used to refine the
coefficients for the OC2 MCP function. The updated algorithm (OC2v2) is presented along with its statistical
performance and a comparison with the original version of the algorithm.
1.1 INTRODUCTION
The at-launch SeaWiFS chlorophyll aalgorithm, named
OC2 for Ocean Chlorophyll 2-band algorithm, is an em-
pirical equation relating remote sensing reflectances, Rrs,
in the 490 and 555 nm bands to chlorophyll aconcentra-
tion, Ca(O’Reilly et al. 1998). OC2 was derived from
a large data set (N= 919) of coincident in situ remote
sensing reflectance and chlorophyllaconcentration mea-
surements, ˜
Rrs(λ) and ˜
Ca, respectively. This large data set
covered a ˜
Carange of 0.02–32 mg m−3from a variety of oce-
anic provinces, and was assembled during SeaBAM. The
main SeaBAM objective was to evaluate a variety of bio-
optical algorithms and produce an at-launch operational
algorithm suitable for producing chlorophyll aimages at
global scales from SeaWiFS data (Firestone and Hooker
1998). The OC2 algorithm was chosen by the SeaBAM
participants, because it represented a good compromise
between simplicity and performance over a wide range of
Ca.
The formulation of the OC2 algorithm is an MCP:
Ca=10
(a0+a1R2+a2R2
2+a3R3
2)+a4,(1)
where R2= log10R490
555and Rλi
λjis a compact notation for
the Rrs(λi)/Rrs (λj) band ratio.
1.2 The SeaBAM DATA SET
While the SeaBAM data set (Fig. 1) is a large, quality-
controlled data set, it has several known limitations:
1. It is mostly representative of Case-1, nonpolar wa-
ters;
2. Data from very oligotrophic ( ˜
Ca<0.05 mg m−3)
and eutrophic ( ˜
Ca>3mgm
−3) areas are underrep-
resented;
3. The chlorophyll aconcentration data are determined
from both fluorometric and HPLC techniques; and
4. Because some of the ˜
Rrs(λ) measurements were not
exactly centered on the SeaWiFS wavelengths, ra-
diometric adjustments were necessary (O’Reilly et
al. 1998).
Additionally, even though the SeaBAM data were qual-
ity controlled, there was still significant variability in the
radiometric data (i.e., variations perpendicular to the x-
axis in Fig. 1). This variability is partly natural, caused
3

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
OC2
˜
Ca
(mg m-3)
R
490
555
˜
Fig. 1. A scatterplot of ˜
R490
555 versus ˜
Cafor the original SeaBAM data set (N= 919). The curve
represents the OC2 algorithm.
OC2
OC2v2
˜
C
a
(mg m-3)
R
490
555
˜
Fig. 2. A scatterplot of ˜
R490
555 versus ˜
Cafor the original SeaBAM (crosses) and the new data (circles).
The dotted curve represents the original OC2 algorithm; the solid curve represents OC2v2, the new
algorithm.
4

O’Reilly et al.
0.01 0.1 1 10 100
0.01
0.1
1
10
100 TYPE: REDUCED MAJOR AXIS
N
: 1174
INT: -0.0000
SLOPE: 1.0000
R
2: 0.9044
RMS: 0.1956
BIAS: -0.0000
0.01 0.1 1 10 100
0.01
0.1
1
10
100
-1.00 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75 1.00
0
17
34
51
68
85
102
119
136
153
170
N
: 1174
MIN: -0.753
MAX: 0.872
MED: -0.003
MEAN: -0.000
STD: 0.196
SKEW: 0.034
KURT: 1.511
0.01 0.1 1 10 100
0.0
0.2
0.4
0.6
0.8
1.0
= 1174
0.1 1.0 10.0
0.01
0.1
1
10
100
Frequency
c)
log(
C
a
/
˜
C
a
)
Relative Frequency
d)
C
a
(µg l-1)
Ca
˜
Ca
N
R
490
555
b)
C
a
Quantiles
(µg l-1)
˜
C
a
Quantiles
(µg l-1)
C
a
(µg l-1)
a)
˜
C
a
(µg l-1)
e)
˜
C
a
(µg l-1)
˜
Fig. 3. Comparisons between OC2v2 (modeled) Cavalues and (in situ)˜
Cadata: a) scatterplot of
Caversus ˜
Ca;b) quantile–quantile plot of Caversus ˜
Ca;c) frequency distribution of log(Ca/˜
Ca); d)
relative frequency of Ca(thin solid curve) and ˜
Ca(thick gray curve); e) R490
555 versus Ca. Also shown is
the OC2v2 model (solid curve).
5

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Table 1. Data sources and characteristics of the combined SeaBAM data set. The BBOP data sets were taken
monthly, and the California Cooperative Fisheries Institute (CalCOFI) data sets were taken quarterly. NFis the
number fluorometric chlorophyll asets, and NHis the number of HPLC chlorophyllasets. The last six subsets are the
new data that were added to the original SeaBAM data.
Data Set Provider Location Date(s) N NFNHWavelengths
BBOP92-93 D. Siegel Sargasso Sea 1992–1993 72 72 72 410 441 488 520 565 665
BBOP94-95 D. Siegel Sargasso Sea 1994–1995 67 61 67 410 441 488 510 555 665
WOCE J. Marra 50◦S–13◦N, Mar93 70 70 410 441 488 520 565 665
88–91◦W
10◦S–30◦N, Apr94
18–37◦W
EqPac C. Davis 0◦, 140◦W Mar92, Sept92 126 126 410 441 488 520 550 683
NABE C. Trees 46–59◦N, May89 72 72 412 441 488 521 550
17–20◦W
NABE C. Davis 46◦N, 19◦W Apr89 40 40 410 441 488 520 550 683
Carder K. Carder N. Atlantic Aug91 87 87 412 443 490 510 555 670
Pacific Jul92
Gulf of Mexico Apr93
Arabian Sea Nov94, Jun95
CalCOFI G. Mitchell Calif. Current Aug93–Sept96 303 303 412 443 490 510 555 665
MOCE-1 D. Clark Monterey Bay Sept92 8 88412 443 490 510 555
MOCE-2 D. Clark Gulf of Calif. Apr93 5 55412 443 490 510 555
North Sea R. Doerffer 55–52◦N, 0–8◦E Jul94 10 10 412 443 490 510 555 670
Chesapeake Bay L. Harding 37◦N, 75◦W Apr95 and 9 9 412 443 490 510 555 671
Jul95
Canadian Arctic G. Cota 74.38◦N, 95◦W Aug96 8 8 412 443 490 509 555 665
AMT-1 S. Hooker 50◦N–50◦S, Sept95 and 42 42 33 412 443 490 510 555
AMT-2 G. Moore 0–60◦W Apr96
MBARI F. Chavez 9◦N–9◦S, Oct97–Dec97 34 34 412 443 490 510 555 670
120–180◦W
CoASTS G. Zibordi 45.3◦N, 12.5◦E Sept97–Jan98 35 35 412 443 490 510 555 670
CARIACO F. M¨uller- 10.3◦N, May96–Aug97 14 14 5 412 443 490 510 555 670
Karger 64.4◦W
AMT-5 S. Hooker 50◦N–50◦S, Sept97–Jun98 82 82 412 443 490 510 555 670
AMT-6 S. Hooker 20◦E–60◦W
ROAVERRS 96–97 K. Arrigo Ross Sea Dec97–Jan98 67 67 412 443 490 510 555 670
CSC M. Culver 30–35◦N, May97–Nov97 23 22 15 412 443 490 510 555
76–82◦W
Total 1174 759 613
1. BBOP: Bermuda Bio-Optical Prfiler
2. WOCE: World Ocean Circulation Experiment
3. EqPac: Equatorial Pacific (Process Study)
4. NABE: North Atlantic Bloom Experiment
5. CalCOFI: California Cooperative Fisheries Institute
6. MOCE: Marine Optical Characterization Experiment
7. AMT: Atlantic Meridional Transect
8. MBARI: Monterey Bay Aquarium Research Institute
9. CoASTS: Coastal Atmosphere and Sea Time Series
10. ROAVERRS: Research on Ocean–Atmosphere Variability and Ecosystem Response in the Ross Sea
11. CSC: Coastal Services Center (NOAA)
6

O’Reilly et al.
Initial Algorithm
Ca
(mg m-3)
New Algorithm
Ca
(mg m-3)
Fig. 4. Statistical and graphical comparisons of the OC2 and OC2v2 algorithms. The thick solid curve
illustrates how both algorithms compare in the 0.01–100 mg m−3Caconcentration range (when the same
R490
555 ratios are used for both equations). The 1:1 (center), 1:5 (bottom), and 5:1 (top) lines are also
plotted.
by the bio-optical variability among the different oceanic
provinces sampled (e.g., variation in phytoplankton species,
relative concentration, and the influence of accessory pig-
ments or the physiological state of phytoplanktonic cells,
etc.), but some of this radiometric variability results from
differences in methodologies, instrument designs, calibra-
tions, data processing, and environmental conditions (sea
and sky state).
1.3 OC2v2 ALGORITHM
Since the SeaBAM workshop, new in situ measure-
ments have become available and were used to test the
accuracy of the OC2 algorithm. These new data indi-
cated that OC2 was performing generally well (within the
±35% accuracy) in Case-1 waters with ˜
Cabetween 0.03–
1mgm
−3, but at chlorophyllaconcentrations exceeding 2–
3mgm
−3, OC2 tended to overestimate ˜
Ca. This tendency
was also apparent in SeaWiFS chlorophyll retrievals from
some offshore, chlorophyll-rich waters, where ample histor-
ical sea-truth data suggest that the frequency of these high
SeaWiFS chlorophyll retrievals are improbable.
As indicated above, the SeaBAM data set contains rel-
atively few chlorophyll ameasurements above 2 mg m−3.
Moreover, those above 2 mg m−3are from a limited num-
ber of regions and may not adequately represent the full
range of bio-optical variability expected in chlorophyll-rich
waters. To strengthen the SeaBAM data set at ˜
Ca>
1mgm
−3, 255 new measurements of ˜
R490
555 and ˜
Cawere
added to the original SeaBAM data set. Characteristics of
the combined data (original SeaBAM data and new data)
are illustrated in Table 1 and Fig. 2. Note that not all
new data are from chlorophyll-rich waters. Nevertheless,
all available new data were used to form the combined set,
because these new sources increase the bio-optical diver-
sity of the data set. Among these new data, the high-
est chlorophyll concentrations come from the ROAVERRS
96–97 and AMT-6 surveys. It is important to note the
˜
R490
555 band ratios measured during these two surveys, at
˜
Ca>4mgm
−3, are substantially lower than the lowest
band ratios present in the original SeaBAM data (Fig. 2).
The dispersion of the combined data is greater than in
the original SeaBAM data set, particularly at chlorophyll
values exceeding 2 mg m−3. This increased dispersion is
expected at high ˜
Cavalues, because some of these data
come from near-shore coastal locations and may be influ-
enced by various optically active components other than
phytoplankton [colored dissolved organic matter (CDOM),
sediments, nonbiogenous detrital substances, etc].
It is clear from Fig. 2 that the original SeaBAM data
set did not adequately encompass the range of variabil-
ity in ˜
R490
555 band ratios present in chlorophyll-rich waters,
and that OC2 derived from SeaBAM would overestimate
chlorophyll afor many of the new observations with ˜
Ca>
2mgm
−3. Assuming the combined data shown in Fig. 2
are an improved representation of the natural variability
present in productive oceanic and coastal zones, the com-
bined data set was used to refine the OC2 functional coeffi-
cients. Because the underlying assumptions and appropri-
ateness for using the MCP function remain valid (O’Reilly
et al. 1998), other formulations were not explored. The
7

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
updated algorithm (OC2v2) is as follows:
Ca=10
(0.2974 −2.2429R2+0.8358R2
2−0.0077R3
2)−0.0929 (2)
where R2is defined as in (1).
Statistical and graphical comparisons between chloro-
phyll aconcentrations derived from OC2v2 versus ˜
Caare
presented in Fig. 3. A comparison of the output from
OC2 and OC2v2 is illustrated in Fig. 4. OC2 and OC2v2
yield very similar results for Caranging between 0.03–
1.5 mg m−3. At chlorophyll avalues exceeding 3 mg m−3,
OC2v2 estimates are substantially lower than OC2. At
very low chlorophyll aconcentrations, 0.01–0.02 mg m−3,
OC2v2 produces slightly higher concentrations than OC2.
1.4 CONCLUSIONS
While, on average, OC2v2, should result in an improve-
ment over OC2 in chlorophyll-rich areas, the uncertainties
remain large for ˜
Ca>3–4 mg m−3. It must be emphasized
that because the variability of the data increases at high
concentrations, the precision of the SeaWiFS retrievals is
inevitably degraded. It must also be kept in mind that the
limitations of the original SeaBAM data set remain valid
for the new combined data. More good quality ˜
Rrs(λ)
and ˜
Cadata are needed from regions with chlorophylla
concentrations above 3.0 and below 0.04mg m−3to better
characterize the bio-optical variability of these waters and,
thus, to identify potential strategies to achieve reasonable
satellite chlorophyll aretrievals.
Acknowledgments
The authors wish to thank all the participants of the SeaBAM
workshop for their help and contribution to the data set: K.L.
Carder, S.A. Garver, S.K. Hawes, M. Kahru, C.R. McClain,
B.G. Mitchell, G.F. Moore, J.L. Mueller, B.D. Schieber, and
D.A. Siegel. We also would like to acknowledge J. Aiken, K.R.
Arrigo, F.P. Chavez, D.K. Clark, G.F. Cota, M.E. Culver, C.O.
Davis, R. Doerffer, L.W. Harding, S.B. Hooker, J. Marra, F.E.
M¨uller-Karger, A. Subramaniam, C.C. Trees, and G. Zibordi
who kindly provided some of their data.
8

O’Reilly et al.
Chapter 2
Ocean Color Chlorophyll aAlgorithms for
SeaWiFS, OC2, and OC4: Version 4
John E. O’Reilly
NOAA, National Marine Fisheries Service, Narragansett, Rhode Island
St´
ephane Maritorena, David A. Siegel, Margaret C. O’Brien, and Dierdre Toole
ICESS/University of California Santa Barbara, Santa Barbara, California
B. Greg Mitchell and Mati Kahru
Scripps Institution of Oceanography, University of California, San Diego, California
Francisco P. Chavez and P. Strutton
Monterey Bay Aquarium Research Institute, Moss Landing, California
Glenn F. Cota
Old Dominion University, Norfolk, Virginia
Stanford B. Hooker and Charles R. McClain
NASA Goddard Space Flight Center, Greenbelt, Maryland
Kendall L. Carder and Frank M¨
uller-Karger
University of South Florida, St. Petersburg, Florida
Larry Harding and Andrea Magnuson
Horn Point Laboratory, University of Maryland, Cambridge, Maryland
David Phinney
Bigelow Laboratory for Ocean Sciences, West Boothbay Harbor, Maine
Gerald F. Moore and James Aiken
Plymouth Marine Laboratory, Plymouth, United Kingdom
Kevin R. Arrigo
Department of Geophysics, Stanford University, Stanford, California
Ricardo Letelier
College of Oceanic and Atmospheric Sciences, Oregon State University
Mary Culver
NOAA, Coastal Services Center, Charleston, South Carolina
Abstract
This chapter describes the revisions (version 4) to the ocean chlorophyll two- and four-band algorithms, as well
as the very large in situ data set used to update these algorithms for use in the third reprocessing of SeaWiFS
data. The in situ data set is substantially larger (N=2,853) than was used to develop earlier versions of OC2
and OC4. The data set includes samples from a greater variety of bio-optical provinces, and better represents
oligotrophic and eutrophic waters. The correlation between chlorophyll aconcentration, Ca, estimated using
OC4 and in situ Ca(˜
Ca) estimated from fluorometric and HPLC analyses was slightly higher than that for
OC2. OC4 would be expected to perform better than OC2, when applied to satellite-derived, water-leaving
radiances retrieved from oligotrophic and eutrophic areas. Variations of the OC4 algorithm are provided for
other ocean color sensors to facilitate comparisons with SeaWiFS.
9

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
2.1 INTRODUCTION
The accuracy, precision, and utility of an empirical
ocean color algorithm for estimating global chlorophyll a
distributions depends on the characteristics of the algo-
rithm and the in situ observations used to develop it. The
empirical pigment and chlorophyll algorithm widely used
in the processing of the global Coastal Zone Color Scan-
ner (CZCS) data set was developed using fewer than 60
in situ radiance and chlorophyll apigment observations
(Evans and Gordon 1994). Since the CZCS period, a num-
ber of investigators have measured in situ remote sensing
reflectance, ˜
Rrs(λ), and in situ chlorophyll aconcentra-
tion, ˜
Ca, from a variety of oceanic provinces. In 1997, the
SeaBAM group (Firestone and Hooker 1998) assembled a
large ˜
Rrs(λ) and ˜
Cadata set containing 919 observations.
This data set was used to evaluate the statistical perfor-
mance of chlorophyllaalgorithms and to develop the ocean
chlorophyll 2-band (OC2) and ocean chlorophyll 4-band
(OC4) algorithms (O’Reilly et al. 1998).
OC2 predicts Cafrom the Rrs(490)/Rrs (555) band ratio
using an MCP formulation. Hereafter, the Rrs ratio con-
structed from band Adivided by band Bis indicated by
RA
B, i.e., the Rrs(490)/Rrs(555) band ratio is represented
by R490
555. OC4 also relates band ratios to chlorophyll awith
a single polynomial function, but it uses the maximum
band ratio (MBR) determined as the greater of the R443
555,
R490
555,orR510
555 values. OC2 was employed as the standard
chlorophyll aalgorithm by the SeaWiFS Project following
the launch of SeaWiFS in September 1997. Although the
statistical characteristics of OC4 were superior to those of
OC2, the SeaBAM group recommended using the simpler
2-band OC2 at launch.
With the goal of improving estimates in chlorophyll-
rich waters, OC2 was revised (version 2) based on an ex-
panded data set of 1,174 in situ observations (Maritorena
and O’Reilly 2000) and applied by the SeaWiFS Project in
the second data reprocessing (McClain 2000). Additional
in situ data have become available as the result of new pro-
grams (e.g., SIMBIOS) and the continuation and expan-
sion of ongoing field campaigns. These new data increase
the variety of bio-optical provinces represented in the origi-
nal data set and fill in regions of the Rrs(λ) and Cadomain
which were not previously well represented. Also, results
from over 2.5 years of SeaWiFS data are now available to
assess the overall performance of the SeaWiFS instrument
and identify areas where improvements are needed in the
processing of satellite ocean color data (McClain 2000).
An update to the OC2 and OC4 chlorophyll algorithms
for SeaWiFS are presented in this chapter, along with a
description of the major features of the very large in situ
data set used to refine these models, and a comparison of
the updated algorithms with earlier versions. MBR chloro-
phyll algorithms for several other satellite ocean color sen-
sors are also provided to facilitate intercomparisons with
SeaWiFS.
2.2 THE IN SITU DATA SET
A very large data set of ˜
Rrs(λ) and ˜
Cameasurements
were assembled for the purpose of updating ocean color
chlorophyll algorithms for SeaWiFS calibration and vali-
dation activities. The data sets and the principal investi-
gators responsible for collecting the data are provided in
Table 2. Table 3 gives the location and acquisition time pe-
riods of the data, along with an indication of the number
of observations, how the chlorophyll aconcentration was
determined (fluorometry or HPLC), and how the radio-
metric observations were made (above- or in-water). The
wavelengths of the latter are presented in Table 4.
The data set has a total of 2,853 in situ observations. It
is the largest ever assembled for algorithm refinement, and
represents a large diversity of bio-optical provinces. The
Cadata are derived from a mixture of HPLC and fluoro-
metric measurements from surface samples: 28% and 72%
of the data, respectively (Table 3). The Cavalues range
from 0.008–90 mg m−3. The relative frequency distribution
of Cahas a primary and secondary peak at 0.2 mg m−3
and approximately 1 mg m−3, respectively (Fig. 5). Oce-
anic regions with Cabetween 0.08–3 mg m−3are relatively
well represented. There are 238 observations of Caex-
ceeding 5 mg m−3and 116 observations with Caless than
0.05 mg m−3. A comparison of the Cafrequency distri-
bution with those from previous versions (O’Reilly et al.
1998 and Maritorena and O’Reilly 2000) shows that olig-
otrophic and eutrophic waters are relatively better repre-
sented in the current data set. The present data set also
has a more equitable distribution over a broader range of
Ca(i.e., 0.08–3 mg m−3).
Measurements of Rrs(λ) were made using both above-
and in-water radiometers: 88% and 12% of the data, re-
spectively (Table 3). In several subsets, multiple Rrs mea-
surements were taken at stations where only a single Ca
measurement was made. For these subsets (BBOP9293,
WOCE, EqPac, NABE, GoA97, Ber96, Ber95, Lab97,
Lab96, Res96, Res95–2, Res94), the median Rrs value was
paired with the solitary Caobservation and added to the
data set.
Except in a limited number of circumstances, band ra-
tios determined from the median Rrs values agreed well
with the individual band ratios. Several subsets, however,
required adjustments to the ˜
Rrs(λ) values to conform with
the SeaWiFS band set. The ˜
Rrs(555) value was estimated
from the ˜
Rrs(565) measurement for the BBOP9293 and
WOCE data using an equation derived from concurrent
measurements of ˜
Rrs(555) and ˜
Rrs(565) from 1994–1995
BBOP surveys (equation 2 from O’Reilly et al. 1998). The
˜
Rrs(555) value for the CB-MAB subset was computed by
averaging the ˜
Rrs(550) and ˜
Rrs(560) values. The ˜
Rrs(510)
value was estimated from the ˜
Rrs(520) values for the Eq-
Pac, WOCE, NABE, and BBOP9293 data sets using the
following conversion equation based on Morel and Maritor-
10

O’Reilly et al.
Table 2. The data sets and the investigators responsible for the data collection activity.
No. Data Set Investigators
1 ROAVERRS 96–97 Arrigo, K.
2 CARDER Carder, K.
3 CARDER Carder, K.
4 CARDER Carder, K.
5 CARDER Carder, K.
6 MF0796 Carder, K.
7 TOTO Carder, K.
8 CoBOP Carder, K., J. Patch
9 EcoHAB Carder, K., J. Patch
10 Global Chavez, F.
11 MBARI EqPac Chavez, F., P. Strutton
12 MOCE-1 Clark, D.
13 MOCE-2 Clark, D.
14 MOCE-4 Clark, D., C. Trees
15 GoA97 Cota, G.
16 Ber95 Cota, G., S. Saitoh
17 Ber96 Cota, G., S. Saitoh
18 Lab96 Cota, G., G. Harrison
19 Lab97 Cota, G., G. Harrison
20 Res94 Cota, G.
21 Res95-2 Cota, G.
22 Res96 Cota, G.
23 Res98 Cota, G.
24 CSC Culver, M., A. Subramaniam
25 CSC Culver, M., A. Subramaniam
26 CSC Culver, M., A. Subramaniam
27 EqPac Davis, C.
28 NABE Davis, C.
29 CB-MAB Harding, L., A. Magnuson
30 AMT-1 Hooker, S., G. Moore
31 AMT-2 Moore, G., S. Hooker
32 AMT-3 Hooker, S., J. Aiken, S. Maritorena
33 AMT-4 Hooker, S., S. Maritorena
34 AMT-5 Hooker, S., S. Maritorena
35 AMT-6B Moore, G., S. Hooker, S. Maritorena
36 AMT-6 Hooker, S., S. Maritorena
37 AMT-7 Hooker, S., S. Maritorena
38 AMT-8 Hooker, S., S. Maritorena
39 HOT Letelier, R., R. Bidigare, D. Karl
40 WOCE Marra, J.
41 WOCE Marra, J.
42 CalCOFI Mitchell, G., M. Kahru
43 CalCOFI Mitchell, G., M. Kahru
44 RED9503 Mitchell, G., M. Kahru
45 AI9901 Mitchell, G., M. Kahru
46 JES9906 Mitchell, G., M. Kahru
47 CARIACO M¨uller-Karger, F., R. Varela, J. Akl, A. Rondon, G. Arias
48 NEGOM M¨uller-Karger, F., C. Hu, D. Biggs, B. Nababan, D. Nadeau, J. Vanderbloemen
49 ORINOCO M¨uller-Karger, F., R. Varela, J. Akl, A. Rondon, G. Arias
50 GOM Phinney, D., C. Yentch
51 Arabian Sea Phinney, D., C. Yentch
52 FL-Cuba Phinney, D., C. Yentch
11

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Table 2. (cont.) The data sets and the investigators responsible for the data collection activity.
No. Data Set Investigators
53 BBOP 9293 Siegel, D., M. O’Brien, N. Nelson, T. Michaels
54 BBOP 9499 Siegel, D., M. O’Brien, N. Nelson, T. Michaels
55 Plumes & Blooms Siegel, D., D. Toole, L. Mertes, R. Smith, L. Washburn, M. Brzezinski
56 NABE Trees, C.
57 CoASTS Zibordi, G.
Table 3. Data sources, locations, and acquisition dates (summarized by the three-letter month abbreviation and the
two-digit year) of the global data set. Nis the number of samples, the Cacolumn indicates the method(s) used for
chlorophyll adetermination (F for fluorometry and H for HPLC), and the Rrs column indicates the type of radiometric
used for measuring remote sensing reflectance (A for above water and B for below water).
No. Data Set Location Dates NC
aRrs
1 ROAVERRS 96–97 Ross Sea Dec97–Jan98 73 H B
2 CARDER North Atlantic Aug91 87 F A
3 CARDER Pacific Jul92 FA
4 CARDER Gulf of Mexico Apr93 FA
5 CARDER Arabian Sea Nov94, Jun95 FA
6 MF0796 Bering Sea Apr96 22 F A
7 TOTO Bahamas Apr98, Apr99 26 F A
8 CoBOP Bahamas May98, May–Jun99 43 F A
9 EcoHAB W. Florida Shelf Mar99–Mar00 (6 Surveys) 57 F A
10 Global Global Nov93–Jul98 (18 Surveys) 284 F B
11 MBARI EqPac Equatorial Pacific Oct97–Nov99 (6 Surveys) 89 F B
12 MOCE-1 Monterey Bay Sep92 8H B
13 MOCE-2 Gulf of California Apr93 5H B
14 MOCE-4 Hawaii Jan–Feb98 20 F B
15 GoA97 Gulf of Alaska Oct97 10 F B
16 Ber95 Bering Sea Jul95 17 F B
17 Ber96 Bering Sea Jul96 16 F B
18 Lab96 Labrador Sea Oct–Nov96 68 F B
19 Lab97 Labrador Sea May–Jun97 71 F B
20 Res94 Resolute Aug94 9F B
21 Res95-2 Resolute Aug95 14 F B
22 Res96 Resolute Aug96 11 F B
23 Res98 Resolute Aug98 91 F B
24 CSC Onslow Bay and S. MAB May97 12 F B
25 CSC S. Mid-Atlantic Bight Sep97, Nov97, Apr98, Feb99 45 F B
26 CSC Gulf of Mexico Apr99 6F B
27 EqPac 0◦N,140◦W Mar92, Sep92 36 H B
28 NABE 46◦N,19◦W Apr89 6H B
29 CB-MAB Chesapeake Bay and MAB Apr96–Oct98 (9 Surveys) 197 H B
30 AMT-1 E. North Atlantic and W. South Atlantic Sep–Oct95 23 F B
31 AMT-2 E. North Atlantic and W. South Atlantic Apr–May96 19 F B
32 AMT-3 E. North Atlantic and W. South Atlantic Sep–Oct96 20 H B
33 AMT-4 E. North Atlantic and W. South Atlantic Apr–May97 21 H B
34 AMT-5 E. North Atlantic and W. South Atlantic Sep–Oct97 45 H B
35 AMT-6B E. North Atlantic and W. South Atlantic Apr–May98 62 H B
36 AMT-6 E. North Atlantic and E. South Atlantic May–Jun98 35 H B
37 AMT-7 E. North Atlantic and W. South Atlantic Sep–Oct98 52 H B
38 AMT-8 E. North Atlantic and W. South Atlantic May–Jun99 46 H B
39 HOT N. Pacific Subtropical Gyre (ALOHA) Feb98–May99 50 H,F B
40 WOCE 50◦S–13◦N,88–91◦W Mar–Apr93 15 F B
12

O’Reilly et al.
Table 3. (cont.) The data sources, locations, and acquisition dates of the global data set.
No. Data Set Location Dates NC
aRrs
41 WOCE 10◦S–30◦N,18–37◦W Apr–May94 27 F B
42 CalCOFI California Coast 93–97 (16 Surveys) 299 F B
43 CalCOFI California Coast 97–99 (6 Surveys) 100 F B
44 RED9503 California Coast (Red Tide) Mar95 9F B
45 AI9901 Subtrop. Atlantic, Indian Ocean Jan–Mar99 36 F B
46 JES9906 E. Japan Sea Jun–Jul99 37 F B
47 CARIACO Cariaco Basin May96–Aug99 35 F A
48 NEGOM NE Gulf of Mexico Jul–Aug98 13 F A
49 ORINOCO Orinoco Delta, Paria Gulf, Orinoco Plume Jun98, Oct98, Feb99, Oct99 48 F A
50 GOM Gulf of Maine Mar95–Apr99 (11 Surveys) 92 F C
51 Arabian Sea Arabian Sea Jul95, Sep95, Oct95 15 F C
52 FL-Cuba Florida–Cuba Mar99 13 F C
53 BBOP 9293 Sargasso Sea (BATS) 92–93 30 H B
54 BBOP 9499 Sargasso Sea (BATS) Jan94–Aug99 83 H B
55 Plumes & Blooms Santa Barbara Channel Aug96–June99 251 F B
56 NABE 46–59◦N,17–20◦W May89 19 H B
57 CoASTS N. Adriatic Sea Sep97–Jan98 35 H B
Table 4. The wavelengths of the radiometer data.
No. Data Set Nominal Center Wavelengths [nm]
1 ROAVERRS 96–97 412 443 490 510 555 655
2 CARDER 412 443 490 510 555 670
3 CARDER 412 443 490 510 555 670
4 CARDER 412 443 490 510 555 670
5 CARDER 412 443 490 510 555 670
6 MF0796 412 443 490 510 555 670
7 TOTO 412 443 490 510 555 670
8 CoBOP 412 443 490 510 555 670
9 EcoHAB 412 443 490 510 555 670
10 Global 412 443 490 510 555 656 665
11 MBARI EqPac 412 443 490 510 555 670
12 MOCE-1 412 443 490 510 555
13 MOCE-2 412 443 490 510 555
14 MOCE-4 412 443 490 510 555 670
15 GoA97 405 412 443 490 510 520 532 555 565 619 665 683 700
16 Ber95 412 443 490 510 555 665 683
17 Ber96 405 412 443 490 510 520 532 555 565 619 665 683 700
18 Lab96 405 412 443 490 510 520 532 555 565 619 665 683 700
19 Lab97 405 412 443 490 510 520 532 555 565 619 665 683 700
20 Res94 412 443 490 510 555 665 683
21 Res95-2 412 443 490 510 555 665 683
22 Res96 405 412 443 490 510 520 532 555 565 619 665 683 700
23 Res98 405 412 443 490 510 520 532 555 565 619 665 683 700
24 CSC 380 412 443 490 510 555 683
25 CSC 380 412 443 490 510 555 683
26 CSC 380 412 443 490 510 555 683
27 EqPac 410 441 488 520 550 683
28 NABE 410 441 488 520 550 683
29 CB-MAB 412 443 455 490 510 532 550 560 589 625 671 683 700
30 AMT-1 412 443 490 510 555 665
31 AMT-2 412 443 490 510 555 665
13

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Table 4. (cont.) The wavelengths of the radiometer data.
No. Data Set Nominal Center Wavelengths [nm]
32 AMT-3 412 443 490 510 555 665
33 AMT-4 412 443 490 510 555 665
34 AMT-5 412 443 490 510 555 665
35 AMT-6B 412 443 490 510 555 665
36 AMT-6 412 443 490 510 555 665
37 AMT-7 412 443 490 510 555 665
38 AMT-8 412 443 490 510 555 665
39 HOT 412 443 490 510 555 670
40 WOCE 410 441 488 520 565 665
41 WOCE 410 441 488 520 565 665
42 CalCOFI 340 380 395 412 443 455 490 510 532 555 570 665
43 CalCOFI 412 443 490 510 555 665
44 RED9503 340 380 395 412 443 455 490 510 532 555 570 665
45 AI9901 412 443 490 510 555 665
46 JES9906 412 443 490 510 555 665
47 CARIACO 412 443 490 510 555 656
48 NEGOM 412 443 490 510 555 670
49 ORINOCO 410 443 490 510 555 670
50 GOM 412 443 490 510 555 665
51 Arabian Sea 412 443 490 510 555 665
52 FL-Cuba 412 443 490 510 555 665
53 BBOP 9293 410 441 488 520 565 665
54 BBOP 9499 410 441 465 488 510 520 555 565 589 625 665 683
55 Plumes & 412 443 490 510 555 656
Blooms
56 NABE 412 441 488 521 550
57 CoASTS 412 443 490 510 555 655 683
Global Data Set
0.01 0.10 1.00 10.00 100.00
Ca
(mg m-3 )
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
Relative Frequency
0.0
0.2
0.4
0.6
0.8
1.0
Cumulative Frequency
V4: Cumulative
V1: (
N
=
919)
V2: (
N
=
1174)
V4: (
N
=
2853)
Fig. 5. The relative frequency distribution of Caconcentration in the in situ data used to develop
versions 4 and earlier versions of the ocean chlorophyll algorithms (V1 is version 1, V2 is version 2,
and V4 is version 4). The version 3 data set, an intermediate test set, is not described here). Relative
frequency is the observed frequency normalized to the maximum frequency.
14

O’Reilly et al.
ena (2000):
Rrs(510) = Rrs (520)1.0605321 −0.1721619 γa+0.0295192 γ2
a+0.0150622 γ3
a−0.004133924 γ4
a(3)
where γa= log(Ca).
The Chesapeake Bay and Mid-Atlantic Bight (CB-
MAB) ˜
Rrs(λ) measurements were corrected for the influ-
ence of radiometer self-shading (Gordon and Ding 1992,
and Zibordi and Ferrari 1995) using equations provided
by G. Zibordi. Corrections for radiometer shading by the
Acqua Alta Oceanographic Tower were also applied to the
CoASTS ˜
Rrs(λ) data (Zibordi et al. 1999). The CalCOFI,
RED9503, and AI9901 data sets were also corrected for
radiometer self-shading (Kahru and Mitchell 1998a and
1998b.)
Interpolated estimates of Rrs were also generated for
non-SeaWiFS wavelengths, which were not consistently
present in the global data set, to develop chlorophyll al-
gorithms similar to OC4 for use by other ocean color sen-
sors. The interpolation–extrapolation method consisted of
two steps. A cubic spline interpolation method [using the
Interactive Data Language (IDL)†, version 5.3] and four
measured adjacent Rrs values were used to derive the inter-
polated Rrs estimate ( ˆ
Rrs). The interpolated values were
then regressed against those measured Rrs values present
in the global data set; the resulting regresssion equation
(Table 5) was applied in the second step to remove bias
in the interpolated values. This scheme resulted in good
agreement between interpolated and measured Rrs over a
wide range of chlorophyll concentration (Fig. 6).
The characteristics of the Rrs data most relevant to
bio-optical algorithms are illustrated in Fig. 7. An impor-
tant feature revealed by these plots is the dispersion of
the data (variability is orthogonal to the major axis of the
data). A pattern common to these plots is the progressive
increase in dispersion with increasing chlorophyll concen-
tration and decreasing band ratio. This is most evident
in the plots of R412
555 and R443
555 versus Ca. In addition to
bio-optical variability, some of the scatter is caused by a
variety of methodological errors (for example, surface ef-
fects, ship shadow, and lower radiometric precision and
extrapolation errors associated with measurements made
in turbid waters).
Considering only the degree of scatter evident in these
plots, the R443
555 provide the most precise (lowest disper-
sion) Caestimates at concentrations approximately less
than 0.4 mg m−3, whereas, the R510
555 and R490
555 band ra-
tios would provide relatively more precise estimates of Ca
in chlorophyll-rich waters. Over the entire data domain,
R490
555 yields the highest correlation with Ca,R2=0.862
(Fig. 7), followed by R443
555,R2=0.847. It must be kept in
mind, however, that R2is an index of the degree of linear
†IDL is a software product of Research Systems, Inc., Boulder,
Colorado.
association and a simple linear model is generally not the
best model to describe the band ratio Carelationships over
the entire range of the data.
2.3 OC2 AND OC4
The Rrs and Cadata (N=2,853) were used to revise
the OC2 and OC4 Caalgorithms. Four observations, with
˜
Cagreater than 64 mg m−3, were widely scattered in plots
of band ratios versus Caand were not used. A test ver-
sion of the OC4 MBR model revealed 45 observations had
log(Ca)/log( ˜
Ca) values exceeding three standard devia-
tions, so these data were also discarded. The final model
coefficients were derived using the remaining 2,804 Rrs and
˜
Cacombinations. Algorithm refinement involved the de-
termination of model coefficients using iterative minimiza-
tion routines (using IDL) to achieve a slope of 1.000, an
intercept of 0.000, minimum root mean square (RMS) er-
ror, and maximum R2between model and measured ˜
Ca
concentration. The first version of OC4 (O’Reilly et al.
1998) was formulated as a modified cubic polynomial (i.e.,
a third order polynomial plus an extra coefficient), how-
ever, the current version of OC4 uses a fourth order poly-
nomial (five coefficients), because this yielded better statis-
tical agreement between model (Ca) and ˜
Cathan an MCP
formulation. An MCP equation was used to refine OC2 to
the same set of values (N=2,804) used to update OC4.
The fourth order polynomial equation for OC4 version
4 (OC4v4), is:
Ca=10.00.366 −3.067R4S +1.930R2
4S
+0.649R3
4S −1.532R4
4S(4)
where R4S = log10 R443
555 >R
490
555 >R
510
555, where the argu-
ment of the logarithm is a shorthand representation for the
maximum of the three values. Hereafter, in an expression
such a R4S, the numerical part of the subscript refers to the
number of bands used, and the letter denotes a code for the
specific satellite sensors [S is SeaWiFS, M is the Moderate
Resolution Imaging Spectroradiometer (MODIS), O is the
Ocean Color and Temperature Scanner (OCTS), E is the
Medium Resolution Imaging Spectrometer (MERIS), and
C is CZCS]. The modified cubic polynomial equation for
OC2 version 4, hereafter referred to as OC2v4, is:
Ca=10.00.319 −2.336R2S +0.879R2
2S
−0.135R3
2S−0.071
(5)
where R2S = log10 R490
555.
The statistical and graphical characteristics of these
two algorithms are illustrated in Figs. 8 and 9. The R2
15

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
0.01 0.10 1.00 10.00 100.00
C
a
(mg m-3)
0.80
0.90
1.00
1.10
1.20
510 :
R
510
N
= 853
a)
0.01 0.10 1.00 10.00 100.00
C
a
(mg m-3)
0.80
0.90
1.00
1.10
1.20
520 :
R
520
N
= 350
b)
0.01 0.10 1.00 10.00 100.00
C
a
(mg m-3)
0.80
0.90
1.00
1.10
1.20
531 :
R
531
N
= 770
c)
0.01 0.10 1.00 10.00 100.00
C
a
(mg m-3)
0.80
0.90
1.00
1.10
1.20
550 :
R
550
N
= 258
d)
0.01 0.10 1.00 10.00 100.00
C
a
(mg m-3)
0.80
0.90
1.00
1.10
1.20
555 :
R
555
N
= 914
e)
0.01 0.10 1.00 10.00 100.00
C
a
(mg m-3)
0.80
0.90
1.00
1.10
1.20
560
:
R
560
N
= 197
f)
R
ˆ
0.01 0.10 1.00 10.00 100.00
C
a
(mg m
-3)
0.80
0.90
1.00
1.10
1.20
565 :
R
565
N
= 350
g)
R
ˆ
R
ˆ
R
ˆ
R
ˆ
R
ˆ
R
ˆ
Fig. 6. The ratio of Rrs based on interpolated Rrs (ˆ
R) to measured Rrs (R) versus chlorophyll concen-
tration (Ca): a) ˆ
R510:R510 ;b) ˆ
R520:R520 ;c) ˆ
R531:R531 ;d) ˆ
R550:R550 ;e) ˆ
R555:R555 ;f) ˆ
R560:R560 ; and g)
ˆ
R565:R565 .
0.1 1.0 10.0
0.01
0.1
1
10
R
2: 0.799
C
a
(mg m-3)
R
412
555
1 10
0.01
0.1
1
10
R
2: 0.847
C
a
(mg m-3)
R
443
555
1
0.01
0.1
1
10
R
2: 0.862
C
a
(mg m-3)
R
490
555
1
0.01
0.1
1
10
R
2: 0.836
C
a
(mg m-3)
R
510
555
Fig. 7. The relationship between R412
555,R443
555,R490
555, and R510
555 band ratios and chlorophyll concentrations
less than 64 mg m−3(N=2,849, except for R412
555 where N=2,813).
16

O’Reilly et al.
0.01 0.1 1 10 100
0.01
0.1
1
10
100
N
: 2804
TYPE: REDUCED MAJOR AXIS
INT: 0.000
SLOPE: 1.000
R
2: 0.883
RMS: 0.231
BIAS: 0.000
0.01 0.1 1 10 100
0.01
0.1
1
10
100
-1.00 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75 1.00
0
32
64
96
128
160
192
224
256
288
320
Frequency
N
: 2804
MIN: -0.759
MAX: 0.783
MED: -0.009
MEAN: 0.000
STD: 0.231
SKEW: 0.208
KURT: 0.506
0.01 0.1 1 10 100
0.0
0.2
0.4
0.6
0.8
1.0
Relative Frequency
d)
0.1 1.0 10.0
0.001
0.01
0.1
1
10
100
e)
C
a
(mg m-3)
C
a
(mg m-3)
C
a
(mg m-3)
a)
c)
b)
C
a
Quantiles
(mg m-3)
˜
C
a
)
log(
C
a
/
R
490
555
˜
C
a
Quantiles
(mg m-3)
˜
C
a
(mg m-3)
Ca
˜
Ca
N
˜
Fig. 8. Comparisons between OC2v4 modeled values (Ca) and in situ data ( ˜
Ca): a) Scatterplot of
Caversus ˜
Ca;b) Quantile–quantile plot of Caversus ˜
Ca;c) Frequency distribution of log(Ca/˜
Ca); d)
Relative frequency of Ca(thin solid curve) and ˜
Ca;e) R490
555 versus ˜
Ca. Also shown is the OC2v4 model
(solid curve).
17

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
0.01 0.1 1 10 100
0.01
0.1
1
10
100
0.01 0.1 1 10 100
0.01
0.1
1
10
100
b)
-1.00 -0.75 -0.50 -0.25 0.00 0.25 0.50 0.75 1.00
0
34
68
102
136
170
204
238
272
306
340
N
: 2804
MIN: -0.749
MAX: 0.742
MED: -0.010
MEAN: 0.000
STD: 0.222
SKEW: 0.295
KURT: 0.666
0.01 0.1 1 10 100
0.0
0.2
0.4
0.6
0.8
1.0
0.1 1.0 10.0
0.001
0.01
0.1
1
10
100
TYPE: REDUCED MAJOR AXIS
INT: 0.000
SLOPE: 1.000
R
2: 0.892
RMS: 0.222
BIAS: 0.000
N
: 2804
C
a
(mg m-3)
˜
C
a
(mg m-3)
C
a
Quantiles
(mg m-3)
˜
C
a
Quantiles
(mg m-3)
Frequency
c)
log(
C
a
/
˜
C
a
)
a)
Relative Frequency
d)
C
a
(mg m-3)
Ca
˜
Ca
N
C
a
(mg m-3)
R
443
555
R
490
555
R
510
555
>>
e)
˜
Fig. 9. Comparisons between OC4v4 modeled values (Ca) and in situ data ( ˜
Ca): a) Scatterplot of
Caversus ˜
Ca;b) Quantile–quantile plot of Caversus ˜
Ca;c) Frequency distribution of log(Ca/˜
Ca); d)
Relative frequency of Ca(thin solid curve) and ˜
Ca;e) R490
555 versus ˜
Ca. Also shown is the OC4v4 model
(solid curve).
18

O’Reilly et al.
value between ˜
Caand (model) Cais slightly higher with
OC4 (0.892) than OC2 (0.883). Both models yield a rela-
tive frequency distribution that is approximately congru-
ent with the ˜
Cadistribution. The OC2 and OC4 models
are extrapolated to a Cavalue of 0.001, well below the
lowest concentration (0.008 mg m−3) present in the in situ
data (Figs. 8e and 9e). If clear (clearest) water is oper-
ationally defined as Ca=0.001 mg m−3, then the clear
water reflectance ratio (R443
555) predicted by OC4 is within
the theoretical range given in Table 6, whereas the extrap-
olated clear water R490
555 reflectance ratio for OC2 is greater
than the theoretical clear water estimates.
Because the OC2v4 and OC4v4 algorithms were tuned
to the same data set, their Caestimates should be very
highly correlated and internally consistent, with a slope
of 1 and an intercept of 0. This is illustrated in Fig. 10.
The reduced scatter (orthogonal to the 1:1 line), centered
at about 1 mg m−3, indicates the region where both algo-
rithms use the 490 nm band.
Additional noteworthy characteristics of OC4 are illus-
trated in Figs. 11 and 12. The R443
555 ratio dominates (50%)
at MBRs above approximately 2.2, R490
555 between 2.2 and
1.1, and R510
555 at MBRs below 1.1 (Fig. 11). With respect to
chlorophyll concentration, the R443
555 ratio dominates (50%)
when Cais below approximately 0.33 mg m−3,R490
555 for Ca
between 0.33–1.4 mg m−3, and R510
555 when Caexceeds ap-
proximately 1.4 mg m−3(Fig. 12).
Relative to OC2v2, OC2v4 predicts slightly higher Ca
above concentrations of 3 mg m−3(Fig. 13), while OC4v4
generates slightly lower Caestimates at very high concen-
trations (Fig. 14). At Cabelow 0.03 mg m−3, OC2v4 esti-
mates are very similar to OC2v2, while OC4v4 estimates
are slightly higher than those from OC4v2, particularly so
when Cais below 0.01 mg m−3. (Version 3 equations were
preliminary and provided to the SeaWiFS Project for test-
ing and evaluation and are not described here.)
There is considerable interest and benefit from com-
paring and merging data from various ocean color sensors
(Gregg and Woodward 1998). This is one of the major ob-
jectives of SIMBIOS (McClain and Fargion 1999). In the
particular case of ocean color data merging, one method-
ological issue to be resolved is how data from satellite
sensors having different center band wavelengths can be
merged to generate seamless maps of chlorophylladistribu-
tion. Among several possible approaches, one is to develop
internally consistent, sensor-specific variations of empirical
chlorophyll aalgorithms tuned to the same data set. This
implies a comprehensive suite of in situ measurements at
wavelengths matching the various satellite spectrometers
or perhaps hyperspectral in situ data. To facilitate com-
parisons with SeaWiFS chlorophylla, MBR algorithms for
several ocean color sensors are presented in Table 7. These
algorithms must be considered as an approximation, be-
cause the in situ data set is biased to SeaWiFS channels
and a number of radiometric adjustments were made to
the Rrs(λ) data to compensate for wavelength differences
among the sensors (Table 4).
2.4 CONCLUSIONS
A large data set of ˜
Rrs and ˜
Cameasurements was com-
piled and used to update the OC2 and OC4 bio-optical
chlorophyll aalgorithms. The present data set, which is
substantially larger (N=2,853) than that used to develop
the version 2 algorithms (N=1,174), includes samples from
a greater variety of bio-optical provinces, and better rep-
resents oligotrophic and eutrophic waters.
Over the four-decade range in chlorophyll aconcentra-
tion encompassed in the data set (0.008–90 mg m−3), the
R490
555 band ratio is the best overall single band ratio index of
chlorophyll aconcentration. In oligotrophic waters, how-
ever, the R443
555 ratio yields the best correlation with Caand
lowest RMS error, while in waters with chlorophyll concen-
trations exceeding approximately 3 mg m−3,theR510
555 ratio
is the best-correlated index. OC4 takes advantage of this
band-related shift in precision, and the well-known shift
of the maximum of Rrs(λ) spectra towards higher wave-
lengths with increasing Ca. Dispersion between the OC2
model and ˜
Catended to increase with increasing chloro-
phyll concentrations above 1 mg m−3, whereas dispersion
using OC4 remained relatively low and uniform through-
out the range of in situ data. Consequently, OC4 yields a
slightly higher R2and lower RMS error than OC2.
Statistical comparisons of algorithm performance with
respect to in situ data, however, provide only partial infor-
mation about their performance when applied to satellite-
derived water-leaving radiances. Operationally, OC4 would
be expected to generate more accurate Caestimates than
OC2 for several reasons. In oligotrophic water, OC4 would
be expected to provide more accurate Caestimates than
OC2, because the signal-to-noise ratio (SNR) is greater
in the 443 nm band than the 490 nm band. In eutrophic
waters, strong absorption in the blue region of the spec-
trum results in lower SNR for water-leaving radiances re-
trieved in the 412 nm and 443 nm bands relative to the
490 nm and 510 nm bands. Furthermore, the influence
of the atmospheric correction scheme on the accuracy of
derived water-leaving radiances used in band-ratio algo-
rithms must be considered. The SeaWiFS atmospheric cor-
rection algorithm (Gordon and Wang 1994 and Wang 2000)
uses the near infrared bands (765 and 865 nm) to char-
acterize aerosol optical properties and estimates aerosol
contribution to total radiance in the visible spectrum by
extrapolation. The 510 nm band, being closer to the near
infrared bands, is less prone to extrapolation errors than
the 490 nm and 443 nm bands. In chlorophyll-rich water,
therefore, OC4 would be expected to provide more accu-
rate estimates of Cathan OC2.
The present version of the ˜
Rrs(λ) and ˜
Cadata set rep-
resents a significant improvement in size, quality, and bio-
optical diversity when compared with earlier versions, but
19

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Table 5. Regression statistics (reduced major axis) for the linear relationship between log (measured Rrs) and log
(interpolated Rrs), where mis the slope and bis the intercept.
Rrs NR
2mb
510 853 0.995 0.9948 0.00299
520 350 0.990 1.0328 0.06280
531 770 0.995 0.9614 −0.1005
550 258 0.999 0.9827 −0.0425
555 914 0.998 1.0032 0.01141
560 197 0.998 1.0178 0.02361
565 350 0.989 1.0487 0.11512
Table 6. Comparison between theoretical and extrapolated clear water reflectance ratios using OC2 and OC4 algo-
rithms, where ais the absorption per meter, bbis the backward scattering coefficient per meter, and fis the function
(unspecified). The theoretical reflectance ratios are based on the absorption and backscattering values from Pope and
Fry (1997) and Morel (1974).
Rrs Band Ratio Rrs =fbb
a+bbRrs =fbb
aAlgorithm
443:555 16.53 21.78 18.21 (OC4)
490:555 6.13 6.66 7.502 (OC2)
Table 7. The maximum band ratio algorithms for the SeaWiFS, CZCS, OCTS, MODIS, and MERIS sensors. As
with the OC4, OC4O, and OC4E algorithms, the argument of the logarithms for OC3M and OC3C is a shorthand
representation for the maximum of the indicated values.
Sensor Name Equation
SeaWiFS OC4 Ca=10.00.366 −3.067R4S +1.930R2
4S +0.649R3
4S −1.532R4
4S
where R4S = log10 R443
555 >R
490
555 >R
510
555
MODIS OC3M Ca=10.00.2830 −2.753R3M +1.457R2
3M +0.659R3
3M −1.403R4
3M
where R3M = log10 R443
550 >R
490
550
OCTS OC4O Ca=10.00.405 −2.900R4O +1.690R2
4O +0.530R3
4O −1.144R4
4O
where R4O = log10 R443
565 >R
490
565 >R
520
565
CZCS OC3C Ca=10.00.362 −4.066R3C +5.125R2
3C −2.645R3
3C −0.597R4
3C
where R3C = log10 R443
550 >R
520
550
MERIS OC4E Ca=10.00.368 −2.814R4E +1.456R2
4E +0.768R3
4E −1.292R4
4E
where R4E = log10 R443
560 >R
490
560 >R
510
560
20

O’Reilly et al.
0.01 0.10 1.00 10.00 100.00
OC2v4 (mg m-3)
0.01
0.10
1.00
10.00
100.00
OC4v4 (mg m-3)
TYPE: REDUCED MAJOR AXIS
N
: 2804
INT: 0.000
SLOPE: 1.000
R
2: 0.992
RMS: 0.058
BIAS: 0.000
Fig. 10. Comparisons of Cafrom OC2 and OC4 when using ˜
Rrs from the in situ data set.
1 10
0
10
20
30
40
50
60
70
80
90
100
Relative Frequency (%)
R
443
555
R
490
555
R
510
555
>>
R
443
555
R
490
555
R
510
555
Fig. 11. The relative frequency of band ratios used in the OC4 model versus the maximum band ratio.
21

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
0.01 0.10 1.00 10.00 100.00
0
10
20
30
40
50
60
70
80
90
100
Relative Frequency (%)
R
443
555
R
490
555
R
510
555
Ca
(mg m-3)
Fig. 12. The relative frequency of band ratios used in the OC4 model versus chlorophyll concentration.
0.1 1.0 10.0
0.001
0.01
0.1
1
10
100
Ca
(mg m-3)
R
490
555
OC2v2
OC2v4
Fig. 13. The comparison of Caestimates from OC2v4 with OC2v2.
22

O’Reilly et al.
1 10
0.001
0.01
0.1
1
10
100
Ca
(mg m-3)
OC4v2
OC4v4
R
443
555 >>
R
490
555
R
510
555
Fig. 14. The comparison of Caestimates from OC4v4 and OC4v2 models.
it still lacks observations from the clearest oceanic waters.
These observations are required to resolve the asymptotic
relationship expected between Rrs(λ) and Caas chloro-
phyll aconcentration diminishes below 0.01mg m−3, and
reflectance band ratios approach the theoretical values for
pure sea water. They are also needed to determine if the
OC2 and OC4 extrapolations beyond the lowest ˜
Caare ac-
curate. Given the spatially and temporally comprehensive
time series achieved by the SeaWiFS mission, these clear-
est water regions and optimal sampling times may now be
easily identified and targeted for special shipboard surveys.
Although clearest waters encompass a relatively small frac-
tion of the global ocean, these and highly eutrophic areas
represent bio-optical and ecological extremes and changes
in their magnitude or areal distribution may provide very
sensitive indicators of global change.
Acknowledgments
The authors would like to acknowledge the following individuals
for their significant contribution of in situ data and ideas: J.
Marra, C Davis, D. Clark, G. Zibordi, C. Trees, R. Bidigare,
D. Karl, J. Patch, R. Varela, J. Akl, C. Hu, A. Subramaniam,
N. Nelson, T. Michaels, R. Smith, and A. Morel.
23

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Chapter 3
SeaWiFS Algorithm for the Diffuse Attenuation
Coefficient, K(490), Using Water-Leaving
Radiances at 490 and 555nm
James L. Mueller
CHORS/San Diego State University
San Diego, California
Abstract
A new algorithm has been developed using the ratio of water-leaving radiances at 490 and 555 nm to estimate
K(490), the diffuse attenuation coefficient of seawater at 490nm. The standard uncertainty of prediction for the
new algorithm is statistically identical to that of the SeaWiFS prelaunch K(490) algorithm, which uses the ratio
of water-leaving radiances at 443 and 490nm. The new algorithm should be used whenever the uncertainty of
the SeaWiFS determination of water-leaving radiance at 443 is larger than that at 490nm.
3.1 INTRODUCTION
The attenuation over depth z(in meters), of the spec-
tral downwelling irradiance, Ed(λ, z ) (in units of mWcm−2
nm−1at wavelength λ), is governed by the Beer–Lambert
Law:
Ed(λ, z)=Ed(λ, 0-)e−K(λ,z)z,(6)
where K(λ, z) is the diffuse attenuation coefficient in per
unit meters, averaged over the depth range from just be-
neath the sea surface (z=0
-) to depth zin meters. Gor-
don and McCluney (1975) showed that 90% of the remotely
sensed ocean color radiance is reflected from the upper
layer, of depth z90, corresponding to the first irradiance
attenuation length, thus satisfying the condition
Ed(λ, z90)
Ed(λ, 0-)=e−1.(7)
The depth z90 is found from an irradiance profile, by in-
spection, as the depth where condition (7) is satisfied.
From (6), the remote sensing diffuse attenuation coefficient
at wavelength λcan be found as K(λ)=z−1
90 m−1.
Austin and Petzold (1981) applied simple linear regres-
sion to a sample of spectral irradiance and radiance profiles
to derive a K(490) algorithm of the form
K(490) = Kw(490) + ALW(λ1)
LW(λ2)B
,(8)
where Kw(490) is the diffuse attenuation coefficient for
pure water, LW(λ1) and LW(λ2) are water-leaving radi-
ances at the respective wavelengths of λ1and λ2, and
Aand Bare coefficients derived from linear regression
analysis of the data expressed as ln[K(490) −Kw(490)]
and ln[LW(λ1)/LW(λ2)]. In Austin and Petzold (1981),
Kw(490) = 0.022 m−1was taken from Smith and Baker
(1981), and because the algorithm was derived for CZCS,
λ1= 443 nm and λ2= 550 nm.
The SeaWiFS ocean color instrument has channels at
443 and 555 nm. Mueller and Trees (1997) found a dif-
ferent set of coefficients for (8) using wavelengths λ1=
443 nm and λ2= 555 nm, and also used the ratio of nor-
malized water-leaving radiances. The substitution of nor-
malized water-leaving radiances in (8) had no significant
effect, but the change in λ2yielded small, but statisti-
cally significant different coefficients Aand B. Following
Austin and Petzold (1981), Mueller and Trees (1997) also
assumed Kw(490) = 0.022 m−1(Smith and Baker 1981).
The Mueller and Trees (1997) result was adopted for the
SeaWiFS prelaunch K(490) algorithm.
SeaWiFS determinations of LW(443) are persistently
lower than water-leaving radiances that are determined
from matched in situ validation measurements. The seri-
ous underestimates of SeaWiFS LW(443) yield correspond-
ingly poor agreement between SeaWiFS and in situ K(490)
determinations. On the other hand, SeaWiFS determina-
tions of LW(490) and LW(555) agree much more closely
with validation measurements.
This chapter is the report of an algorithm based on
(8) using 490 and 555 nm, which should yield improved
uncertainty in SeaWiFS K(490) estimates. The algorithm
also adopts a reduced value of Kw(490) based on recently
published values of pure water absorption (Pope and Fry
1997).
24

O’Reilly et al.
3.2 DATA AND METHODS
Two samples of K(490) and normalized water-leaving
radiances are used in the present analysis. Sample 1 is
used for a regression analysis to derive coefficients Aand
Bfor (8) with λ1= 490 nm and λ2= 555 nm. The data
in Sample 2 are entirely independent from Sample 1 and
are used to determine standard uncertainties of predic-
tion in K(490) calculated using the algorithm derived here
from Sample 1, and using the prelaunch K(490) algorithm
(Mueller and Trees 1997).
The data comprising Sample 1 were drawn from spec-
tral irradiance and radiance profiles locally archived at
the San Diego State University (SDSU) Center for Hydro-
Optics and Remote Sensing (CHORS). Each Sample 1 pro-
file was analyzed to determine K(490) and water-leaving
radiance using the integral method of Mueller (1995a).
Sample 1 includes the data analyzed by Mueller and Trees
(1997), but excludes two cruises for which reliable upwelled
spectral radiance profile [Lu(490,z)] measurements were
not available. Data from two additional cruises in the Gulf
of California were added to Sample 1, bringing the total
sample size to 319 data pairs.
Sample 2 was provided from the SeaBASS archives by
the SIMBIOS Project Office at GSFC, and consists of 293
sets of K(490), water-leaving radiances and incident sur-
face irradiances (443, 490, and 555 nm) which are indepen-
dent of Sample 1. Water-leaving radiances in Sample 1
were determined by the SIMBIOS Project using the stan-
dard methods employed at GSFC for SeaWiFS match-up
validation analysis.
K(490) and normalized water-leaving radiance ratio
pairs were determined for each sample using the methods
described in Mueller and Trees (1997). A linear regres-
sion analysis was performed on the Sample 1 data pairs
to determine the values of coefficients Aand Bin (8),
with λ1= 490 nm and λ2= 555 nm. Based on Pope and
Fry’s (1997) recent determination of absorption for pure
water aw(490) = 0.015 m−1, and the pure water backscat-
tering coefficient bw(490) = 0.008 m−1reported by Smith
and Baker (1981), the backscattering fraction is heuris-
tically assumed to be less than 0.5 and performed three
regressions assuming values of 0.018, 0.017, and 0.016 m−1
for Kw(490). Finally, standard uncertainties of prediction
were calculated, both for the present (490 and 555 nm) and
the prelaunch (443 and 555 nm) algorithms, as the RMS
differences between the measured and predicted K(490) in
Sample 2.
3.3 RESULTS
Three regression analyses were performed on Sample
1 using successive values of 0.018, 0.017, and 0.016 m−1
for Kw(490). The scatter between ln[K(490) −0.016] and
ln[LW(490)/LW(555)], in per unit meters is illustrated in
Fig. 15a, together with the logarithmic regression line cor-
responding to the algorithm
K(490) = 0.016 + 0.15645LWN (490)
LWN (555)−1.5401
.(9)
In log space, the squared correlation coefficient R2in-
creased monotonically from 0.931–0.937, and the standard
error decreased from 0.186–0.167, as Kw(490) decreased
from 0.018–0.016 m−1. On this basis, the appropriate al-
gorithm selected for use with SeaWiFS was the Kw(490)
=0.016 m−1case.
In linear space, the standard uncertainty of the esti-
mate, calculated as the RMS discrepancy between pre-
dicted and measured K(490) for Sample 1, is 0.012 m−1.
The scatter between predicted and measured K(490), rel-
ative to the one-to-one line, is illustrated in Fig. 15b.
Figures 16a and 16b illustrate the scatter about the
one-to-one line when K(490) predictions using (9) are com-
pared to measurements from Sample 2. The standard un-
certainty of prediction in K(490) using (9) is estimated
from these data to be 0.018 m−1in the range of K(490) <
0.25 m−1(which is the range fit with Sample 1) and cor-
responds to 26% of the mean for this subsample of 249
pairs. When the algorithm of (9) is extrapolated into
the range K(490) >0.25 m−1, the standard uncertainty
of prediction increases to 0.193 m−1(48% of the mean for
this subsample of 31 pairs). The mean biases in predic-
tions are −0.002 m−1for measured K(490) <0.25 m−1,
and −0.130 m−1for measured K(490) >0.25 m−1.
The standard uncertainties and mean biases of predic-
tion for K(490) calculated with the SeaWiFS prelaunch al-
gorithm (Mueller and Trees 1997) are 0.020 and 0.000 m−1,
respectively, for the subsample of Sample 2 with measured
K(490) <0.25−1, and 0.196 and −0.085 m−1for the sub-
sample with measured K(490) >0.25 m−1.
3.4 DISCUSSION
There is little to choose between the in situ perfor-
mances and uncertainties of the (9) K(490) algorithm, us-
ing the ratio of water-leaving radiances at 490 and 555nm,
and the SeaWiFS prelaunch algorithm (Mueller and Trees
1997), using the ratio of water-leaving radiances at 443
and 555 nm. When used with SeaWiFS data, however, (9)
may be expected to yield more accurate K(490) estimates
as long as the uncertainties in estimated LWN (490) are
much lower than those for LWN (443). It is recommended,
therefore, that (9) be substituted for the Mueller and Trees
(1997) K(490) algorithm for processing SeaWiFS data, at
least until future improvements in atmospheric corrections
may produce equivalent uncertainties in water-leaving ra-
diances at 490 and 443 nm.
Neither algorithm performs well in water masses where
K(490) >0.25 m−1. In part, this may be due to extrapo-
lating a regression equation beyond the range of the data
25

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
➀
➀
➀➀
➀
➀
➀
➀
➀
➀
➀
➀➀
➁
➁
➁
➁➁➁
➁➁
➁➁
➁
➁
➁
➁
➁
➁
➁
➁
➂
➂
➂➂
➂➂
➂
➂
➂➂
➂
➂➂➂
➂
➂
➂
➂
➂➂
➂
➂
➂➂
➂
➂
➂
➂
➂
➂
➂
➂
➂
➂
➂
➃
➃
➃
➃
➃
➃
➃
➃
➃➃
➃
➃
➃➃ ➃➃
➃
➃➃
➃➃
➃
➃
➃
➃
➃➃
➃➃
➃
➃➃
➃
➃
➃
➃
➃
➃
➃➃
➃
➃
➃
➃
➃➃
➃➃
➃
➃
➃➃
➃
➃
➃
➃➃
➃
➃
➃
➃
➃
➃
➃
➃
➃
➃➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄➄
➄
➄
➄
➄
➄➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄➄➄
➄
➄➄
➄
➄➄
➄
➄
➄
➄
➄➅➅➅
➅
➅
➅
➅
➅
➆
➆
➆
➆
➆➆
➆
➆
➆
➆
➆
➆
➆
➆➆
➆
➆➆
➆
➆
➆
➆
➇
➇
➇
➇
➇➇
➇➇
➇
➇
➇
➇
➇
➇
➇➇➇➇
➇
➇➇
➇➇
➇
➇
➇
➇
➇➇
➇
➇➇
➇
➇
➇
➇➇
➇
➇
➇
➇
➇➇
➇
➇
➇
➇
➇
➇➇➇
➇
➇
➇➇
➇➇
➇➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇➇
➈➈
➈
➈
➈➈
➈
➈➈➈
➈
➈
➈
➈➈
➈➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈➈➈
➈
➈➈➈
➈
➈➈
-5.5
-4.5
-3.5
-2.5
-1.5
-0.5
0.5
-0.5 0.0 0.5 1.0 1.5 2.0 2.5
ln[
WN
L
(490)/ (555)]
a
Predicted
K
(490) [m-1]
➀
➀
➀
➀➀
➀
➀
➀➀➀
➀
➀
➀
➁➁
➁
➁
➁
➁
➁
➁➁
➁
➁
➁
➁
➁
➁
➁➁
➁➂
➂
➂
➂
➂
➂
➂
➂➂
➂
➂➂
➂
➂
➂
➂
➂
➂
➂
➂
➂
➂➂
➂
➂
➂
➂
➂
➂
➂
➂➂
➂➂➂
➃➃
➃➃
➃
➃
➃
➃
➃
➃➃
➃➃
➃
➃
➃➃
➃
➃
➃
➃
➃
➃➃ ➃
➃
➃➃
➃
➃➃➃
➃
➃➃➃➃
➃➃
➃
➃➃
➃
➃
➃
➃
➃
➃➃
➃
➃
➃
➃
➃
➃
➃
➃
➃➃
➃➃
➃➃
➃
➃➃
➃
➄
➄
➄
➄➄
➄
➄
➄➄➄➄➄
➄
➄
➄
➄
➄
➄
➄
➄➄➄ ➄➄
➄
➄
➄
➄
➄
➄➄
➄
➄
➄
➄
➄➄
➄
➄
➄
➄
➄
➄
➄
➄
➄
➄➄➄➄➄
➆
➆➆
➆
➆
➆➆➆
➆
➆
➆
➆
➆➆
➆➆
➆
➆➆
➆
➆➆
➇
➇➇
➇
➇➇
➇
➇➇➇➇
➇
➇➇
➇
➇➇
➇
➇
➇
➇
➇
➇
➇➇
➇
➇
➇➇➇
➇➇➇
➇➇
➇
➇➇➇
➇➇➇➇
➇➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇
➇➇➇
➇➇ ➇➇
➇
➇
➇
➇
➇
➇➇➈
➈
➈
➈
➈➈
➈
➈
➈➈
➈
➈➈
➈
➈
➈
➈
➈➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈
➈➈➈
➈
➈➈➈
➈
➈➈
➈➈
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Measured
K
(490) [m-1]
b
WN
L
ln[
K
(490)-0.016]
Fig. 15. Scatter comparisons of K(490). a) Logarithmic scatter comparison of K(490) versus the ratio
of water-leaving radiances at 490 and 555nm. The solid line is the least squares regression fit to the data
(excluding the GoCal98A red tide data) given by (9). b) Linear scatter in measured K(490) compared with
predictions using (9) with the ratio of water-leaving radiances at 490 and 555nm. The data are from panel
a). The key for these panels are: 1. Siegel: Sargasso Sea 1994; 2. Mitchell: CalCOFI 1994; 3. GoCal 1995;
4. GoCal 1997; 5. GoCal 1998A (with Red Tide Station); 6. GoCal 1998A Red Tide Data; 7. Trees, Arabian
Sea, JGOFS Proc. 2; 8. Trees, Arabian Sea, JGOFS Proc. 6; 9. Trees, Arabian Sea, JGOFS Proc. 7.
Predicted
K
(490) [m-1]
a
➀➀➀➀
➀➀
➀➀
➀➀➀
➀➀➀➀➀➀
➀➀➀➀
➀➀➀➀➀➀➀➀
➀➀
➀
➀
➀➀➀➀➀➀➀➀➀➀➀➀➀➀
➀
➀
➀➀➀➀➀➀➀➀
➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀➀
➀➀➀➀
➀➀➀➀➀➀➀
➀➁
➁
➁
➁➁
➁
➁➁
➁
➁
➁
➁
➁
➁
➁
➁
➁➁
➂
➂
➂
➂➂➂
➂
➂
➂
➂➂
➂
➂
➂➂
➂
➂➂➂➂
➂
➃➃
➃
➃
➃➃
➃➃
➃
➄
➄➄
➄
➄➄➄
➄
➄
➄➄
➄➄
➄
➄
➄
➄➄
➅
➅
➅➅➅
➅➅
➅
➅➅➅
➅
➅
➅
➅
➅➅➆
➆➆
➆
➆➆
➆
➆➆
➆
➆➆
➆➆
➆➆
➆➆
➆
➆
➆
➆➆
➆
➆➆
➆
➆➇
➇
➇➇➇➇
➇➇ ➇
➈
➈➈
➈
➈
➈
➈
➈
➈➈
➈➈
➈
➈➈➈
➈
➈
➈
➈
➈
➈➈
➈
➈
➈
➈
➈
➈
➉➉
➉
➉
➉
➉
➉
➉➉
➉
➉
➉
➉
➉
➉
➉
➉➉
➉➉
➉
➉
0.0
0.2
0.4
0.6
0.8
1.0
0.0 0.2 0.4 0.6 0.8 1.0
Measured
K
(490) [m-1]
Predicted (490) [m-1]
➀
➀➀➀
➀➀
➀➀
➀➀
➀
➀➀
➀
➀➀
➀
➀➀➀
➀
➀➀➀➀
➀
➀➀➀
➀➀ ➀
➀
➀
➀➀➀➀➀➀➀
➀➀➀
➀➀➀
➀
➀
➀➀➀ ➀➀➀ ➀➀
➀➀➀➀
➀➀
➀➀➀➀➀➀➀ ➀
➀➀➀
➀➀➀
➀
➀
➀
➀➀➀ ➀➀
➀➀➀➀➀➀➀
➀
➁➁
➁
➁
➁➁
➁➁
➁
➁
➁
➁
➁
➁➁
➁
➁
➂
➂
➂
➂➂
➂
➂
➂
➂➂
➂
➂
➂
➂
➂
➂
➂➂➂
➂
➃
➃
➃
➃
➃
➃➃
➃
➃
➄
➄
➄➄
➄➄
➄
➄
➄
➄
➄➄
➄
➄
➄
➄➄➄
➅
➅➅
➅
➅
➅
➅
➆
➆
➆
➆➆
➆
➆➆
➆
➆➆
➆
➆➆
➆
➆➆
➆➆
➆
➆
➇
➇
➇
➇
➇
➇
➇➇➇
➈➈
➈
➈
➈
➈
➈➈
➈➈
➈➈➈➈
➈
➈➈
➈
➈
➈
➈
➈
➉
➉
➉
➉➉
➉
➉
➉➉
➉➉
➉
➉
➉
➉
➉
➉➉
➉➉
➉
➉
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Measured (490) [m-1]
b
K
K
Fig. 16. Scatter comparisons of K(490). a) Same as Fig. 15b, but for an independent sample of K(490)
and water-leaving radiances at 490 and 555nm. The solid line corresponds to a one-to-one agreement. b) A
subset of panel a), where the area of greatest concentration of data points is enlarged for better viewing. The
key for these panels are: 1. BATS 1998; 2–5. CalCOFI -9802, -9804, -9807, and -9809, respectively; 6. April
1998 SMAB; 7. November 1998 SMAB; 8. Feb 1999 SMAB; 9. CARIACO 1998; 10. GoA97; and HOTS
1998.
26

O’Reilly et al.
used to fit its coefficients. In the present circumstances,
however, it is at least equally likely that the poor pre-
dictions result from extremely large uncertainties in both
K(490) and water-leaving radiances derived from radio-
metric measurements near the sea surface in extremely
turbid water masses. In such cases, instrument self shad-
ing, wave focusing, uncertainty in instrument depth de-
termination, and uncertainty in extrapolating Lu(λ, z)to
the sea surface (especially when the linear slope estima-
tion method of analysis is employed) contribute large and
poorly understood uncertainties to measured K(490) and
water-leaving radiances alike. For the near term, the best
policy is to regard SeaWiFS K(490) data with values of
greater than 0.25 m−1with caution and skepticism.
27

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Chapter 4
Long-Term Calibration History of Several
Marine Environmental Radiometers (MERs)
Margaret C. O’Brien, David W. Menzies,
David A. Siegel, and Raymond C. Smith
ICESS, University of California, Santa Barbara
Santa Barbara, California
Abstract
The accuracy of upper ocean AOPs for the vicarious calibration of ocean color satellites ultimately depends on
accurate and consistent in situ radiometric data. The SIMBIOS Project is charged with providing estimates
of normalized water-leaving radiance for the SeaWiFS instrument to within 5%. This, in turn, demands that
the radiometric stability of in situ instruments be within 1% with an absolute accuracy of 3%. This chapter
reports on the analysis and reconciliation of the laboratory calibration history for several BSI MERs, models
MER-2040 and -2041, three of which participate in the SeaWiFS Calibration and Validation Program. This
analysis includes data using four different FEL calibration lamps, as well as calibrations performed at three
SIRREXs. Barring a few spectral detectors with known deteriorating responses, the radiometers used by UCSB
during the BBOP have been remarkably stable during more than five years of intense data collection. Coefficients
of variation for long-term averages of calibration slopes, for most detectors in the profiling instrument, were less
than 1%. Long-term averages can be applied to most channels, with deviations only after major instrument
upgrades. The methods used here to examine stability accommodate the addition of new calibration data as they
become available; this enables researchers to closely track any changes in the performance of these instruments
and to adjust the calibration coefficients accordingly. This analysis may serve as a template for radiometer
histories which will be cataloged by the SIMBIOS Project.
4.1 INTRODUCTION
The accuracy of upper ocean AOPs, which are needed
for the vicarious calibration of ocean color satellites, ulti-
mately depends on accurate and consistent in situ radio-
metric data. Accurate validation of SeaWiFS demands an
in-water radiometric stability within 1%, with an accuracy
of 3% (Mueller and Austin 1995). Considerable energy
has been spent refining calibration protocols for profiling
radiometers. The SeaWiFS Project Office has sponsored
several workshops through its Calibration and Validation
Program, which have yielded significant improvements in
the research community’s ability to provide accurate AOP
estimates. These include the SIRREXs, conducted annu-
ally since 1992 (Mueller 1993, Mueller et al. 1994, Mueller
et al. 1996, and Johnson et al. 1996), as well as the Data
Analysis Round-Robin (DARR) workshop in 1994 (Siegel
et al. 1995).
At the Institute for Computational Earth System Sci-
ence (ICESS) at UCSB, several research projects provide
validation data for ocean color satellites. These include
BBOP in the Sargasso Sea, the Plumes and Blooms Project
in the Santa Barbara Channel, and the Palmer Area Long
Term Ecological Research (LTER) site on the Antarctic
Peninsula. The ICESS Calibration Laboratory and BBOP
have participated in all of the workshops held by the Cal-
ibration and Validation Program.
This report presents an analysis of the multiyear lab-
oratory calibration history for several BSI MERs, models
MER-2040 and -2041. This analysis includes data using
four different FEL calibration lamps, as well as calibra-
tions performed at three SIRREX exercises. This report
will show that, barring a few sensors with known deteri-
orating responses, the radiometers used by UCSB during
BBOP have been remarkably stable during six years of
intense data collection.
4.2 ICESS FACILITY AND METHODS
The ICESS optical calibration facility is housed in a
climate-controlled room. A 1.2×1.8 m (4×6 ft) optical ta-
ble, with threaded holes arranged in a 2.5 cm (1 in) grid,
supports one end of a 2.4 m (8ft) long optical rail. A black,
wooden baffle with a 25.4 cm (10 in) diameter hole strad-
28

O’Reilly et al.
dles the rail 61.0 cm (2 ft) from the illumination end of the
bench and extends the full 1.2 m (4 ft) width by a height of
1.5 m (5 ft). A 30.5 cm (12 in) square plate can be bolted
over the baffle hole to hold a 7.6cm (3 in) adjustable iris
if necessary. An alignment beam, consisting of a helium–
neon (He:Ne) laser with two adjustable mirrors, is centered
on the hole, parallel to the rail, and is mounted on a plat-
form at the distal end of the optical rail. The table and
rail assembly are surrounded by a black, pleated curtain
suspended from a track mounted on the ceiling. When the
room lights are off and the curtains drawn, no detectable
light reaches the instrument except through the hole in the
baffle. Shadow forms can be inserted between the lamp and
instrument to block direct light during the measurement
of stray light.
The lamp holder array consists of a sliding platform on
the optical rail supporting two horizontal vernier stages at
right angles, a rotary stage, a vertically adjustable post,
and an FEL lamp holder. An alignment jig replaces the
lamp in the holder to properly position the lamp holder
to the alignment laser beam. The lamp holder array can
be easily slid along the rail to provide calibration distances
from 50 cm to over 2 m. The standard lamps are purchased
from, and calibrated by, Optronic Laboratories, Inc.†(Or-
lando, Florida) and calibrations are traceable to the Na-
tional Institute of Standards and Technology (NIST). An
83-DS power supply with a 0.02 Ω shunt provides power for
the FEL lamp. A 4.5 digit voltmeter is used to monitor
the current and voltage during calibrations. The lamp is
allowed to warm up for 10 min before each calibration. The
current is maintained at 8A (±1 mA) and is reproducible
to 0.03%.
The mounting platform for radiometers consists of a
large scissor jack, which can support instruments up to
22.7 kg (50 lbs.) and 20.3 cm (8 in) in diameter. The jack
has independent height and crossbeam adjustments to cen-
ter the instrument on the optical axis. It is attached to a
45.7 cm (18 in), square platform which in turn, can be fas-
tened to the optical bench at any location with 15.2 cm
(6 in) tall aluminum posts which are 43.2 cm (17 in) apart.
Because the hypotenuse of a 30.5 cm (12 in) right triangle
is 43.1 cm (17 in), the platform can be easily positioned
at the 45◦angle desired for radiance calibrations with a
reflective plaque.
During irradiance calibrations, the test instrument is
positioned so that its cosine collector is centered on the
alignment beam and normal to it. Calibrations are usu-
ally performed at a distance of 50 cm, which is measured
through the baffle iris using a 50 cm measuring rod. The
†Certain commercial equipment, instruments, or materials are
identified in this technical memorandum to foster under-
standing. Such identification does not imply recommenda-
tion or endorsement by NASA, NIST, or ICESS, nor does it
imply that the materials or equipment identified are neces-
sarily the best available for the purpose.
lamp holder’s vernier stage is used to make the final dis-
tance adjustment.
Beginning in July 1992, radiance calibrations were per-
formed using a 50.8 cm (20 in) diameter Labsphere, Inc.,
integrating sphere with a variable (2.54–10 cm, 1–4 in di-
ameter) entrance aperture and 15.2 cm (6 in) diameter exit
aperture located 90◦from the entrance. It is illuminated
externally by the same FEL lamp used for irradiance cal-
ibrations. The sphere is positioned on the bench at the
end of the optical rail and the lamp is positioned 50 cm
from the sphere’s 5 cm (2 in) diameter entrance aperture.
The raised platform with the scissor jack is positioned to
hold the test instrument a few centimeters from the exit
aperture and the wooden baffle; black felt is used to block
all stray light.
Beginning in August 1994, radiance calibrations were
also performed using a 60.1 cm (24 in) Spectralon
R‡re-
flectance plaque. At the extreme end of the optical bench,
a vertical bracket at the plaque’s center supports it at a
position normal to the laser alignment beam. The lamp
holder is positioned at a distance of 200 cm from the plaque,
and a baffle with a 25.4 cm (10 in) diameter hole between
the lamp and the plaque allows the lamp to illuminate only
the plaque. The scissor jack and its platform are moved
to align the radiance collector at 45◦to, and 33 cm (13 in)
from, the plaque. From 1994–1996, radiance calibrations
were performed routinely using both the sphere and the
plaque.
4.2.1 Calibration Lamp History
Three NIST-traceable FEL lamps were used for cali-
brating the ICESS radiometers—F219, F303, and F304—
all of which were purchased from Optronic Laboratories.
There are manufacturer’s calibrations for all three lamps.
At UCSB, lamp F219 was used for all calibrations from
1989–1991 (Fig. 17). Since that time, it has been used only
during lamp intercalibration experiments, so there should
have been no further significant aging of this lamp. Lamp
F303 was purchased in June 1992 and was used for all
routine calibrations at UCSB from July 1992–July 1995.
Lamp F303 was recalibrated by Optronic Laboratories in
July 1995 after approximately 50 h of service. Lamp F304
was a seasoned, uncalibrated FEL lamp, purchased in June
1992, and used only a few hours until July 1995 when it was
calibrated by Optronic Laboratories and put into use for
routine calibrations at UCSB. A fourth lamp, F305, was
not calibrated by the manufacturer, but has been used for
comparisons between other lamps.
All four lamps were intercalibrated during at least two
of the three SIRREX calibration workshops in July 1992,
June 1993, and September 1994. F219 was examined at
SIRREX-1 and again at SIRREX-2 in June 1993. Lamp
‡Spectralon is a registered trademark of Labsphere, Inc., in
North Sutton, New Hampshire.
29

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
1989
1990
1991
1992
1993
1994
1995
1996
1997
SIRREX
Lamp Calibrations
at Optronics
xfer cals vs day of exp
F219 F303 F304
F219
F303
F303, F304
F219, F303
F219, F303, F304, F305
F303, F304, F305
F304 F304 to F219, F303, F305
F304 to F303, F305
Fig. 17. A timeline of FEL lamp calibrations at Optronic Laboratories, transfer calibrations at UCSB,
and SIRREX experiments.
30

O’Reilly et al.
Table 8. Wavelength centers and bandwidths measured at full-width at half-maximum (FWHM) power (in
parentheses). All values are in nanometers and the upwelled spectral radiance is denoted as Eu(z, λ). The
channels marked with 94 were added in 1994. If a detector has been replaced, the wavelength given is the one
most recently measured. The column headings denote the MER model and serial number, and the collector
type.
2040 S/N 8728 2041 S/N 8729 2041 S/N 8734 2040 S/N 8714
λ Ed(z,λ)Lu(z,λ)Ed(0+,λ)Ed(0+,λ)Ed(z,λ)Eu(z,λ)
[nm] Center FWHM Center FWHM Center FWHM Center FWHM Center FWHM Center FWHM
340 340.3 8.4
380 378.4 10.0
410 410.2 9.6 411.2 9.4 410.4 9.5 410.4 10.4 410.3 11.4 410.8 11.4
441 441.6 10.8 441.7 10.7 441.7 10.9 442.2 10.4 441.5 11.1 442.0 12.1
465 465.4 10.0 465.8 9.7 464.0 9.5
488 488.0 9.9 486.7 9.6 487.8 10.1 488.5 11.4 483.6 10.7 487.9 10.8
510 510.2 94 9.3 511.9 94 8.2 507.4 11.1 507.5 11.4
520 518.6 11.5 519.4 11.4 519.5 10.3 518.5 10.0 520.1 8.5 518.0 8.3
540 537.8 9.5
555 555.2 94 9.8 555.7 94 9.8
565 564.8 11.1 565.5 10.9 564.6 11.7 563.4 10.8 563.4 10.8 563.3 9.5
587 587.1 10.4 586.0 9.9 585.4 10.2 587.0 9.6 586.5 10.5
625 623.4 94 10.7 624.8 94 10.7 622.6 10.6 624.0 12.1 624.2 12.0
665 664.4 9.5 664.8 9.4 662.9 11.2 663.0 9.7 662.4 9.3
683 680.3 94 9.0 681.1 10.0 680.8 13.9
F303, which has been used extensively for calibrations at
UCSB, was examined at all three SIRREXs; F304 and
F305 were both tested at SIRREX-2 and -3. In addition,
a BSI Profiling Reflectance Radiometer (PRR) with the
same type of photodiodes as the MERs, which was cali-
brated with lamp F303, was used for one of the training
sessions at SIRREX-4 at NIST in May 1995.
In addition to the SIRREX comparisons, one other
comparison between lamps was performed at UCSB. Be-
fore F304 replaced F303 as the lamp used for routine cal-
ibrations, the 1995 Optronic Laboratories calibration for
lamp F304 was transferred onto F303, F219, and F305
using a third MER-2040 instrument (S/N 8733) with 13
irradiance channels between 340–683nm. As mentioned
above, F219, F304, and F305 were used only during the
SIRREXs and had not aged between 1992 and 1995. The
transfer from F304 to F303 and F305 was repeated in May
1996.
4.2.2 Radiometers
At the UCSB optical calibration facility, there are cal-
ibration histories for five BSI spectroradiometers [serial
numbers (S/N) 8728, 8729, 8733, 8734, and 8714] span-
ning up to seven years (Fig. 18). The MER-2040 series of
spectroradiometers is composed of discrete, sealed photo-
diodes, each with triple cavity interference filters giving a
nominal full-width at half-maximum bandwidth of 10 nm.
The wavelength centers range throughout the visible and
ultraviolet-A (UVA) spectrum from 340–683 nm. The ra-
diance detectors are identical 3-cavity filtered photodiodes
mounted in a Gershun tube array. The half-angle field of
view is 10.2◦in air and 13.7◦in water. Instrument 8714
is known as the Bio-Optical Profiling System II (BOPSII)
and was described in Smith et al. (1997).
Radiometers 8728 (MER-2040) and 8729/8734 (MER-
2041) are used routinely in the BBOP at the Bermuda
Atlantic Time Series (BATS) station and have been cali-
brated three or four times per year since July 1992 (Ta-
ble 8 and Fig. 18). The BBOP profiling instrument (S/N
8728) was designed originally with eight downwelling ir-
radiance channels (410–665 nm) and nine upwelling radi-
ance channels (410–683 nm). In January 1994, it was mod-
ified to meet the SeaWiFS protocols (Mueller and Austin
1995) and the number of channels was increased to 12 each
of downwelling and upwelling channels (410–683nm), plus
upwelling natural fluorescence. The gains of all channels
were also adjusted at this time. The original deck sensor
(S/N 8729) has six downwelled channels (410–665 nm) and
the optics have not been modified. In August 1994, it was
replaced by S/N 8734, which has 13 downwelled channels
(340–683 nm).
Radiometer 8733 (MER-2040) was used intensively in
the field from 1992–1993 during the Tropical Ocean Global
Atmosphere (TOGA) Coupled Ocean Atmosphere Re-
sponse Experiment (COARE) and is now used occasion-
ally on BBOP. It has been calibrated approximately once
per year. Instrument 8733 has 13 each of downwelled and
upwelled irradiance channels between 340–683 nm. Radi-
ometer 8714, the BOPSII (MER-2040), has been used for
profiling on all of the Palmer Area LTER and Ice Colors
31

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Add detectors to 8728,
increase gain
First Plaque Calibration
Add detectors to 8714
8728: Ed(555), Lu(555)
8728: Ed(665), Lu(665)
8728: Ed(510), Lu(510)
8734: Ed(340)
INSTRUMENT UPGRADES
1989
1990
1991
1992
1993
1994
1995
1996
1997
BBOP: S/N8728
BOPS: S/N8714
Instrument Calibrations
F219F303F304
Fig. 18. A timeline of calibrations and upgrades for BBOP (S/N 8728, 8729, and 8734) and BOPSII
(S/N 8714) radiometers.
32

O’Reilly et al.
Table 9. Comparison of lamp irradiances between 400 and 700 nm obtained from SIRREX experiments, Op-
tronics, and the transfer of lamp F304 to F305. The data values are the ratios of lamp output measured during
the SIRREX experiments to those provided by Optronics or to the transfer calibration. The standard deviation
is indicated by σ.
SIRREX-1 SIRREX-2 SIRREX-3
Lamp Mean σMean σMean σ
F303-92 1.0087 0.0083 1.0178 0.0087 1.0332 0.0084
F303-95 0.9710 0.0104 0.9788 0.0037 0.9932 0.0028
F304-95 0.9916 0.0045 0.9919 0.0029
F305†0.9895 0.0041 0.9905 0.0027
F219-89 1.0332 0.0079 1.0414 0.0107
F219†0.9854 0.0096 0.9937 0.0038
†Transfer calibration from F304.
cruises in Antarctic waters, as well as many open ocean
projects, and has been calibrated once or twice yearly since
February 1989 (Fig. 18). This instrument had 8 channels
of downwelled irradiance (increased to 13 for 410–665 nm
in November 1994), and 8 channels of upwelled irradiance
(410–624 nm, Table 9). Both 8728 and 8714 had individual
detectors replaced.
This report is primarily concerned with the BBOP in-
struments, because they contribute data to the SeaWiFS
Calibration and Validation Program. Data from the BOP-
SII and TOGA–COARE instruments (S/N 8714 and 8733)
are used primarily to corroborate the conclusions, because
their calibration histories are at least as long as that of the
BBOP instrument and their calibrations involve the same
lamps.
The wavelength properties of each detector were mea-
sured using a double-grating monochrometer. An uncal-
ibrated FEL lamp and condensing lens were used as the
illumination source for the entrance slit, and the output
spot was centered on the radiometer’s cosine collector or
on an individual radiance detector. The wavelength pro-
ducing the maximum signal was determined, followed by
the wavelengths on each side of the peak producing 50% of
the maximum signal. The reported wavelength for a detec-
tor is the average of the two half-maximum wavelengths; its
bandwidth is the difference between these two wavelengths
(Table 8). The wavelength response of the monochrome-
ter was calibrated by observing the visible spectral lines
of a mercury pen lamp. Repeat determinations for any
detector have agreed to within 0.5nm.
4.3 RESULTS
Because this report is concerned with accuracy, as well
as radiometer stability, significant attention has been given
to the calibrations of the lamps. The following discussion
will illustrate:
a) Calibration lamp output must be examined closely;
2) The two profiling instruments (BBOP S/N 8728,
and BOPSII S/N 8714) appear to be stable over
several years; and
3) Long-term averages of calibration coefficients should
be calculated whenever possible.
4.3.1 Lamps
FEL lamp F303 was used continuously for all calibra-
tions from 1992–1995 for a total of approximately 50 h.
Optronic Laboratories specifies that lamp irradiances are
accurate and stable to within approximately 1% for 50 h
or 1 year of use when the supplied current is maintained to
within 0.1%. The two manufacturers’ calibrations in 1992
and 1995 for F303 indicated that its output had changed by
up to 5% and that it should not be used for further radiom-
eter calibrations. The most extreme changes were noticed
at wavelengths less than 500 nm†. Currently, F303 is used
only for monitoring the performance of its replacement,
F304. Given the possible change in the performance of the
primary lamp, the response histories of two radiometers
were examined with both lamp calibrations for evidence
supporting the validity of one or both calibrations.
Because there is no long interruption between the cal-
ibrations of BBOP instruments (S/N 8728 and 8729/8734
are calibrated every 3–4 months), the slopes of radiome-
ter 8728, calibrated with lamp F303 between January 1994
and August 1995, were calculated using both the 1992 and
1995 Optronic Laboratories irradiance calibrations for this
lamp. To compare the relative changes over time, each
slope was normalized to that determined on 9 August 1995,
the date on which lamp F304 replaced F303. The normal-
ized slopes calculated with both lamp calibrations were
examined for drift or step changes, which might indicate
when the calibration lamp’s output had changed. While
using a single FEL lamp and calibration, most channels
on the BBOP profiling radiometer (S/N 8728) showed a
constant calibration response over time (Fig. 19). When a
different FEL lamp or a different calibration of the same
lamp was used, however, there were significant changes of
†The raw data are available at the following universal resource
locator (URL) address http://www.icess.ucsb.edu/bbop.
33

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Year
1992 1993 1994 1995 1996 1997 1998
Normalized Calibration Coefficient
0.96
0.98
1.00
1.02
1.04
1.06
1.08
Ed(410)
Ed(488)
Ed(565)
Ed(683)
F303
1992 1995 F304
1995
norm
Fig. 19. The normalized calibration coefficients for four downwelling irradiance channels (radiometer
8728) using Optronic Laboratories calibrations of lamp F303 in 1992 and 1995. Data were normalized
to the F304 values measured on 9 August 1995, the first date for which lamp F304 was used.
2–6% in the coefficients for most of the channels. Agree-
ment was best between lamps or calibrations employing
the July 1995 Optronic Laboratories calibration for lamps
F303 and F304. In fact, the agreement was excellent when
the 1995 Optronic Laboratories calibration was used for
all radiometer calibrations with lamp F303 back to 1992
(below).
The response of the BOPSII radiometer (S/N 8714) to
new lamps, or different calibrations of a lamp, was similar
to that of radiometer 8728. There were marked steps in
the slopes for each irradiance channel when lamps were
replaced and their original calibrations were applied (data
not shown). The calibration coefficients obtained for the
BOPSII instrument using F219 with the manufacturer’s
lamp calibration compared poorly to later data. When the
coefficients for instrument 8714 were recalculated using the
1995 transfer calibration from lamp F304 to F219, there
was better agreement between early (pre-1992) and later
calibrations. Similar to the BBOP instruments, when the
July 1995 Optronic Laboratories calibration of F303 was
used for all radiometer calibrations from 1992–1995, the
agreement was much better (see below); in fact, 19 out of
21 channels varied less than 1%.
The requirement that radiometer calibrations be accu-
rate, as well as stable, and the problems of reconciling dif-
ferent lamp irradiances led to a more in-depth analysis of
the lamp calibrations. Simply computing coefficients using
the current lamp irradiances provided by Optronic Labo-
ratories apparently was not possible. It was implausible
that the responses of two different radiometers had drifted
with the same rate and magnitude as did the output of
lamp F303.
4.3.2 SIRREX Data
The data from the three SIRREX activities were exam-
ined to determine whether or not they supported the two
differing Optronic Laboratories calibrations for lamp F303.
There are caveats accompanying each SIRREX data set,
which must be taken into account when interpreting SIR-
REX data. During SIRREX-1 (July 1992), the required
uncertainty of 1% was not achieved when transferring the
NIST scale of spectral irradiance from the Goddard Space
Flight Center (GSFC) standard lamp (F267) to the other
lamps (Johnson et al. 1996). The data are, however, in-
cluded here for completeness, and this goal was achieved
during SIRREX-2 and -3. During SIRREX-3, a recent
NIST calibration of the standard lamps (GSFC lamps F268
and F269) became available and indicated that the output
of lamp F269 had drifted by approximately 1.5% some-
time during the previous year, likely as early as SIRREX-2
34

O’Reilly et al.
(Mueller et al. 1996). This necessitated a recalculation of
the SIRREX-2 data sets, and increased their combined un-
certainty. The SIRREX-3 results should be the most reli-
able because many of the procedural problems experienced
during the first two experiments were rectified and the
standard lamps were calibrated by NIST only one month
previously.
Each lamp’s irradiance, at wavelengths used during the
SIRREXs, was computed using the recommended Lagran-
gian interpolation procedure†. When the SIRREX-2 and
-3 data for lamps F304 and F305 were first compared,
it appeared the SIRREX-2 data were an average of 1.6%
lower than those from SIRREX-3 between 400–1,000 nm,
which supports the drift observed in the SIRREX standard
lamps described by Mueller et al. (1996). To correct for
this drift, the average ratios of SIRREX-2 to SIRREX-3
data were computed for both lamps F304 and F305 be-
tween 400 and 1,000nm and this factor applied to the
SIRREX-2 data for all lamps and all wavelengths. The
ratio between SIRREX-2 and -3 can be computed only
for lamps F304 and F305 because they had not been used
between these two experiments. Each calibrated lamp’s
output was compared to the irradiance measured at each
SIRREX experiment (Fig. 20 and Table 9). To examine the
performance of lamps for which there was no current man-
ufacturer’s calibration, lamp irradiances were computed
from the transfer calibrations (performed at UCSB) from
F304 to F219 and F305. These were confined to wave-
lengths between 380 and 665 nm.
Lamp F303: In general, there was best agreement be-
tween each SIRREX experiment and the closest Optronic
Laboratories calibration for lamp F303 (Fig. 20a). The
color shift between 1992 and 1995 suggested by the Op-
tronic Laboratories calibrations, however, was not con-
firmed at either SIRREX-2 or SIRREX-3. The SIRREX-2
(1993) data for F303 between 400 and 700 nm did not agree
well with either the original 1992 data or the 1995 data
(1.6 versus 2.1%, Table 9). There is good agreement above
400 nm between the SIRREX-3 (1994) and the 1995 Op-
tronic Laboratories calibration of lamp F303 (0.7%, Ta-
ble 9). It should be noted that during SIRREX-3, F303 was
calibrated against the standard lamp F268, which was re-
cently calibrated by NIST, rather than F269 which was ob-
served to shift in irradiance during the experiment (Mueller
et al. 1996). A change in the output of F303 could not be
inferred from the SIRREX results.
Lamps F304 and F305: The Optronic Laboratories cal-
ibration of lamp F304 (1995) and the derived calibration
of F305 compared well with both the SIRREX-2 and the
SIRREX-3 data for these lamps, 0.8 and 1.0%, respectively
(Fig. 20b and Table 9).
†This was from an internal Optronic Laboratories report titled
“Report of Calibration of One Standard of Spectral Irradi-
ance OL FEL-C, S/N: F-304,” Project No. 903-479, 28 July
1995.
Lamp F219: There was very poor agreement between
the original Optronic Laboratories F219 calibration and
the earliest SIRREX data (SIRREX-1), but these dates
were three years apart. The agreement was better between
the derived F219 calibrations and SIRREX-2 in the visible
region (Table 9).
In general, the SIRREX data supported the same con-
clusions as the radiometer comparisons: namely, that F303
(1995 calibration) and F304 agree and that the 1995 trans-
fer calibration of F304 to F305 and F219 was reliable (data
shown below). The SIRREX data do not support well the
1992 Optronic Laboratories calibration of lamp F303.
Because the 1995 Optronic Laboratories F304 calibra-
tion seems to be the touchstone for the other lamp cali-
brations, one additional comparison was examined to con-
firm its absolute values. A BSI PRR (S/N 9626) with
the same type of photodiodes as the MERs was used for
one of the training sessions at SIRREX-4 (May 1995) at
NIST. Although the setup was not optimal, readings were
taken with two FEL lamps, F423 and F422 (owned and
calibrated by NIST). Using the F304 Optronic Laborato-
ries calibrations to compute irradiances for the two NIST
lamps, the agreement was within 1% for all but one value
at 665 nm, where the disagreement was most likely due
to reflected stray light from the dark color of the baffling
on the calibration bench (Table 10). This instrument had
been calibrated in March, May, and August 1995 at UCSB,
and in May 1995 at BSI, and these calibrations also agreed
within 1% at all wavelengths (data not shown).
The Optronic Laboratories F304 and F303 calibrations
in July 1995 were in good agreement with the data from
two SIRREX experiments in 1993 and 1994, as well as with
data from two NIST calibrated lamps in 1995. Based on
the radiometer responses, it appears that the irradiance of
F303 had not changed since 1992. The excellent consis-
tency between the early radiometer calibrations calculated
with the 1995 Optronic Laboratories calibration of lamp
F303 and those done with F304, since July 1995, support
the hypothesis that the 1992 Optronic Laboratories cali-
bration for lamp F303 was not accurate and should not be
used. Rather, it appeared that the 1995 Optronic Labo-
ratories F303 calibration should be used for all radiometer
calibrations with this lamp. For all subsequent discussions,
the 1995 Optronic Laboratories irradiances for lamps F303
and F304 will be used for the BBOP radiometers’ calibra-
tions from 1992 to mid-1995 (F303) and from mid-1995 to
the present (F304).
4.3.3 Irradiance History
Once it was established which lamps and calibrations
were most reliable, all of the calibrations of two UCSB pro-
filing radiometers (S/N 8728 and 8714) could be examined
in detail. Because of their different use and calibration
timelines, the irradiance histories of these two radiometers
will be discussed separately.
35

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
SIRREX Irradiance/Optronics Irradiance
0.95
1.00
1.05
0.95
1.00
1.05
SIRREX2:Optronics95
SIRREX3:Optronics95
Wavelength [nm]
400 500 600 700 800 900 1000
SIRREX/Transfer Calibration
0.98
0.99
1.00
1.01
1.02 SIRREX2:Transfer
SIRREX3:Transfer
a)
b)
c)
SIRREX1: Optronics92
SIRREX2: Optronics92
SIRREX3: Optronics92
SIRREX1: Optronics95
SIRREX2: Optronics95
SIRREX3: Optronics95
Fig. 20. Ratios of SIRREX results to Optronic Laboratories irradiances for three lamps used at the
ICESS Calibration Facility: a) Lamp F303, ratios of three SIRREX irradiances to Optronic Laboratories
calibrations from two dates, in 1992 and 1995; b) Lamp F304, ratios of two SIRREX irradiances to
Optronic Laboratories calibration from 1995; c) Results from SIRREX-2 and -3 compared to the transfer
calibration of F304 to F305, note scale change.
36

O’Reilly et al.
Table 10. Comparison of computed irradiances for lamps F422 and F423 measured during SIRREX-4[
ˆ
E(λ)]
using PRR S/N 9626 [ ˆ
E(λ)] as the transfer radiometer. The percent differences (PD) from the actual irradiances
are also given.
F422 Irradiance F423 Irradiance
λ[nm] ˆ
E(λ)ˆ
E(λ)PD [%] ˆ
E(λ)ˆ
E(λ)PD [%]
412 2.6004 2.5827 −0.69 2.7239 2.7132 −0.40
443 3.9873 3.9513 −0.91 4.1652 4.1490 −0.39
490 6.5817 6.5255 −0.86 6.8502 6.8224 −0.41
510 7.8034 7.7309 −0.94 8.1107 8.0559 −0.68
555 10.6600 10.5983 −0.58 11.0500 11.0274 −0.21
665 17.0185 16.7927 −1.34 17.5592 17.4974 −0.35
4.3.3.1 BBOP Radiometers
To compare the differences among irradiance calibra-
tions over the entire project, each detector’s coefficient was
normalized to that from the first calibration in July 1992.
The gains of underwater instrument 8728 were adjusted in
early 1994 and so later data were renormalized to the first
calibration after this date. From 1992–1996, most of the
channels of the profiling radiometer (S/N 8728) were very
stable with no appreciable trends (Fig. 21a). The scat-
ter of all channels, however, increased from January 1994
to December 1996, up to about 2%. The differences be-
tween slopes calculated on any two consecutive dates were
small, about 0.2%. On two dates (9 August 1994 and 19
December 1996), calibrations were performed using both
lamps F303 and F304 and the calibration coefficients were
nearly identical. Those detectors, which did not remain
stable to within 2% during four years, had shown marked
deterioration (up to 5% in three months) and have been
replaced (Fig. 22 and Sect. 4.4). There was a slight drift
downwards in some of the blue channels, indicating that
these detectors may have begun to deteriorate.
The original BBOP surface sensor (S/N 8729) was very
stable from 1992–1994, but drifted towards increasing sen-
sitivity during 1994–1995 (Fig. 21b). The last calibration
before instrument 8729 was taken out of service agreed well
with the calibration performed in early 1997, about two
years later. Except for the most recent calibration (1997),
lamp F303 was used for all calibrations of instrument 8729.
There were no repairs or physical events that would explain
the drift. Its replacement (S/N 8734) also showed a similar
drift upward during 1995–1996, but has been stable since
mid-1996 (Fig. 21c). In May 1996, a smudge of O-ring
grease was cleaned from under the Teflon
R†cosine collec-
tor, which was likely to have been the cause of the drift.
Calibrations after this event were renormalized to the May
1996 values; since then, the instrument 8734 calibrations
have been very stable (Fig. 21c). The two UV channels,
Ed(340) and Ed(380), have shown marked deterioration
and have been replaced (data not shown).
†Teflon is a registered trademark of E.I. du Pont de Nemours,
Wilmington, Delaware.
4.3.3.2 BOPSII Radiometer
The calibration history of the BOPSII profiling radiom-
eter (S/N 8714) began in February 1989 using lamp F219
(Fig. 17). In October 1992, lamp F303 replaced F219,
which was in turn replaced by F304 in October 1995. For
calculations with lamp F219, the transfer calibration from
F304 to F219 was used. The 1995 Optronic Laboratories
calibration for lamp F303 was used for all calibrations per-
formed with that lamp. The data were treated similarly
to that for the BBOP radiometer. Slopes were normalized
to the January 1994 value. Most downwelling irradiance
detectors showed similar variations to those in instrument
8728 (Fig. 23). During seven years, most calibration slopes
for each channel were within 2%; in fact, 19 out of 21
channels varied less than 1%. Two detectors, Ed(510) and
Ed(520), however, still showed large, unexplained drifts
from 1989–1992.
4.3.4 Radiance History
The radiance calibration history is longer for the inte-
grating sphere than for the plaque, however, the reflectance
of this sphere has never been satisfactorily characterized
and an arbitrary reflectance value was used in the slope
calculations. From mid-1994 through 1996, the radiance
channels of instrument 8728 were calibrated with both the
sphere and the plaque, so the plaque calibrations could be
transferred onto the earlier sphere calibrations.
Plaque radiance values were computed from the lamp
irradiance corrected for the inverse square law, the man-
ufacturer’s determination of the plaque’s reflectivity (pro-
vided at the time of purchase), and the assumed Lam-
bertian distribution. Calibration coefficients for instru-
ment 8728 were normalized to those from August 1994,
the first date on which plaque calibrations were performed
(Fig. 24a). As with the irradiance history, most of the ra-
diance channels showed very little change. On instrument
8728, 10 out of 12 channels varied less than 1% from the
normalized value during 10 calibrations (the range for all
channels was 0.26–1.55%). Figure 24a also shows a slight
decrease in the blue channels during 1995 and 1996, al-
though the overall decrease is still less than 2%.
37

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Year
1992 1993 1994 1995 1996 1997 1998
0.99
1.00
1.01
1.02
1.03
1.04
Normalized Calibration Coefficient
0.98
0.99
1.00
1.01
1.02
1.03
0.98
0.99
1.00
1.01
1.02
1.03
Ed(410)
Ed(441)
Ed(488)
Ed(520)
Ed(565)
Ed(683)
a)
norm
norm
b)
c)
norm
Fig. 21. The normalized calibration coefficients for selected downwelling irradiance channels on three
BSI MER-2040 and -2041 radiometers used during BBOP computed using the 1995 Optronic Laborato-
ries calibrations of F303 and F304: a) underwater instrument (S/N 8728), b) surface sensor (S/N 8729),
and c) surface sensor (S/N 8734).
38

O’Reilly et al.
0.7
0.8
0.9
1.0
E
d
(510)
E
d
(555)
E
d
(665)
Year
1992 1993 1994 1995 1996 1997 1998
Normalized Calibration Coefficient
0.6
0.7
0.8
0.9
1.0
L
u
(510)
L
u
(555)
L
u
(665)
a)
b)
665 all 665 510
510
665
all
665
555
Fig. 22. Deteriorating detectors on the BBOP radiometers (S/N 8728 and 8734) from 1992–1997. The
arrows indicate when detectors were replaced.
Year
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998
Normalized Calibration Coefficient
0.95
1.00
1.05
1.10
1.15
Ed(410)
Ed(441)
Ed(488)
Ed(520)
Ed(565)
Ed(665)
norm
Fig. 23. The normalized calibration coefficients for selected downwelling irradiance channels on the
BOPSII profiling radiometer (S/N 8714) calculated with the transfer calibration of F304 to F219 and
the 1995 Optronic Laboratories calibrations of lamps F303 and F304.
39

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Year
1992 1993 1994 1995 1996 1997 1998
0.92
0.94
0.96
0.98
1.00
1.02
1.04
1.06
1.08
1.10
Normalized Calibration Coefficient
0.97
0.98
0.99
1.00
1.01
1.02
u
u
u
u
u
Lu(410)
L (441)
L (488)
L (520)
L (565)
L (683)
a)
b)
norm
norm
Fig. 24. The normalized calibration coefficients for selected radiance channels on the BBOP radiometer
(S/N 8728) from 1992–1997: a) reflectance plaque, and b) integrating sphere.
Sphere coefficients were normalized twice, because of
the gain change in January 1994. Coefficients from cal-
ibrations before the gain change were normalized to the
December 1992 calibration, and those from later dates to
the August 1994 data—the same date as plaque normal-
izations (Fig. 24b). Over the long term, calibrations with
the sphere were more variable than those with the plaque.
From August 1994 to September 1996, the same 10 chan-
nels discussed above had an average coefficient of varia-
tion (CV) of 1.4% when calibrated with the sphere. The
plaque and sphere were illuminated by the same lamp, so
the greater variation in the sphere calibrations may have
been due to changes in the back loading of the sphere when
the instrument was positioned close to the exit aperture, or
to changes in the reflectivity of the sphere coating. These
possibilities were examined before the plaque calibrations
were transferred onto the earlier sphere calibrations.
Because the latter plaque data are very stable, two
plaque calibrations were used to examine possible shifts
in sphere reflectance. For two dates, 17 May 1995 and 18
September 1996, it was assumed that the nominal plaque
reflectances were correct. These were used to compute the
sphere reflectances that would yield the same slopes (in
units of V mW−1cm nm sr) for the measured sphere volt-
ages as were measured with the plaque (Fig. 25). The
difference between the nominal and calculated sphere re-
flectances was clearly spectral and ranged from 0.34–0.83%
in May 1995, and from 0.24–0.68% in September 1996.
The differences between the 1995 and 1996 calculations
were generally less than 0.2%. Although the differences be-
tween the two computed reflectances were greatest in the
blue region, the changes were not large enough to suggest
that the reflectivity of the sphere had changed during this
time. When the two estimated sphere reflectances were
used to calculate calibration slopes, their differences were
magnified to approximately 2% (Table 11). Because the
differences between the two computed sphere reflectances
were small and within the reproducibility of the sphere
calibrations, and any differences would be magnified if a
single estimate of sphere reflectance was used, the aver-
age ratio of plaque-to-sphere coefficients was computed for
40

O’Reilly et al.
Wavelength [nm]
400 450 500 550 600 650 700
Reflectance
0.9750
0.9775
0.9800
0.9825
0.9850
Measured May 95
Measured Sep 96
Fig. 25. Computed reflectance of the ICESS integrating sphere on two dates, 17 May 1995 and 18
September 1996. The dashed line shows the manufacturer’s nominal reflectance.
each channel. These factors were applied to the average
calibration slopes measured with the sphere in 1992 and
1993.
Table 11. Radiance calibration slopes for radiome-
ter 8728 measured on 18 September 1996 computed
from sphere reflectances estimated on two dates,
May 1995 and September 1996. The units are in
VµW−1cm2nm sr.
λ[nm] May 95 Sept. 96 PD [%]
410 0.90070 0.87252 3.13
441 0.93630 0.90027 3.85
465 0.86910 0.84426 2.86
490 0.88819 0.86418 2.70
510 0.85202 0.83449 2.06
520 0.87318 0.85510 2.07
555 1.03165 1.01033 2.07
565 0.90756 0.88888 2.06
589 0.86213 0.84250 2.28
625 0.90776 0.89353 1.57
665 1.08689 1.08205 0.45
683 0.99982 0.98244 1.74
This examination of the calibration histories of these
radiometers demonstrated that most of the detectors were
stable over the course of these instruments’ 6–8 year his-
tories. Furthermore, the stability of the calibration co-
efficients also implies stability of the amplifiers, analog-
to-digital converters and optical windows of the MER in-
struments, as well as, reproducibility of calibration lamp
geometry and the lamp power supply. A change in any
of these components would have been evident in the cali-
bration coefficients. Their absolute calibrations, however,
are tied to just one lamp calibration by Optronic Labora-
tories (F304 in May 1995), which itself, is guaranteed to
about 1%. The variety and number of comparisons of lamp
F304’s irradiance to other lamps lend confidence to these
values.
4.4 LONG-TERM AVERAGES
Long-term averages of calibration slopes can be com-
puted with confidence, because the radiometers used with
BBOP appear to be very stable. These long-term averages
should be used whenever possible and recomputed after
major upgrades. Calibration coefficients for deteriorating
channels should be interpolated. For these purposes, sta-
bility has been defined by a CV less than 1%. When the
CV exceeded these limits, the calibration data were ex-
amined closely for trends or shifts and, in most cases, a
physical reason for the change was evident which justified
computing a new long-term average.
Tables 12–14 summarize the slopes that will be used
for most channels on the three BBOP radiometers. For
the profiling radiometer (Table 12), there are two main
time periods: 1992–1993 during which there were a total of
15 channels, and 1994–1996 after the upgrade to 12 down-
and 13 upwelling channels. In most cases, one slope can be
used for each channel for each time period. At the end of
1994, both 555 nm detectors were replaced and the slopes
of several other channels were affected as well: Ed(488),
41

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Table 12. The average calibration slope (ACS) values used for MER-2040 (S/N 8728) from 1992–1996. The
CV and the number of observations (n) are also given. For most channels, only two slopes are necessary: before
and after the gain change in January 1994. The top half of the table gives the average calibration slopes for the
downwelled irradiance channels, Ed(λ), in units of V µW−1cm2nm. For the channels affected by the repairs
in January 1995 [Ed(410), Ed(488) and Ed(520)], it was necessary to compute separate averages for 1994 and
1995–1996. Before plaque calibrations were available, the average slopes for the upwelled radiance channels,
Lu(λ) (in units of V µW−1cm2nm sr) were determined using the mean slope from the sphere calibrations
and the plaque-to-sphere ratios (given in the bottom half of the table). Their CVs were calculated from the
uncorrected sphere calibrations.
Measurement 1992–1993 1994–1996 1994 1995–1996
Channel ACS CV nACS CV nACS CV nACS CV n
Ed(410) 0.04796 0.21 5 0.03256 0.70 7 0.03211 0.90 3
Ed(441) 0.04568 0.23 5 0.03517 0.94 15
Ed(465) 0.04855 0.14 5 0.03248 0.68 15
Ed(488) 0.05436 0.42 5 0.03395 0.51 4 0.03450 0.46 11
Ed(510) 10.03411 0.37 6 2
Ed(520) 0.04894 0.27 5 0.03314 0.47 4 0.03370 0.56 11
Ed(555) 1 2 0.03633 0.97 11
Ed(565) 0.05456 0.20 5 0.03409 0.67 15
Ed(587) 0.05556 0.16 5 0.03548 0.80 15
Ed(625) 10.03626 0.52 4 0.03684 0.48 11
Ed(665) 20.03449 0.59 4 2
Ed(683) 10.03498 0.48 4 0.03563 0.56 11
Lu(410) 0.3482 0.38 3 0.8917 1.47 11
Lu(441) 0.2800 0.69 3 0.9111 0.65 11
Lu(465) 0.3071 0.42 3 0.8301 0.38 11
Lu(488) 0.2804 0.52 3 0.8561 0.43 11
Lu(510) 10.8843 0.41 4 2
Lu(520) 0.2523 0.59 3 0.8560 0.30 11
Lu(555) 10.8921 0.01 2 1.0198 0.96 8
Lu(565) 0.2476 0.59 3 0.8898 0.26 11
Lu(587) 0.2440 1.04 3 0.8505 0.40 11
Lu(625) 10.8967 0.29 11
Lu(665) 20.8814 0.87 4 2
Lu(683) 0.2379 0.83 3 0.9819 0.28 11
1 Not applicable.
2 Indicates that a detector was deteriorating, e.g., Ed(665), and no average could be computed for that time period.
Table 13. The ACS values used for MER-2041 (S/N 8729) from 1992 to August 1995, in units of V µW−1cm2
nm.
Measurement Sept. 1992–Aug. 1994 Aug. 1994–1995
Channel ACS CV nACS CV n
Ed(410) 0.02564 .91 7 0.02566 .68 3
Ed(441) 0.02587 .28 7 0.02635 .42 3
Ed(488) 0.02934 .22 7 0.03002 .45 3
Ed(520) 0.03037 .28 7 0.03094 .44 3
Ed(565) 0.03163 .42 7 0.03213 .55 3
Ed(665) 0.03492 .55 7 0.03539 .49 3
42

O’Reilly et al.
Ed(520), Ed(625), and Ed(683). For these channels, new
long-term averages were computed for 1995–1996. Sev-
eral detectors—Ed(510), Ed(555), Ed(665), Lu(510), and
Lu(665)—deteriorated (Fig. 22), and it was not possible to
calculate their average slopes for many months. For these
channels, the deterioration was assumed to be linear and
a slope was calculated for each cruise using a least-squares
regression. During 1996 and 1997, all of these aging de-
tectors were replaced and new long-term averages must be
determined.
Table 14. The ACS values used for MER-2041
(S/N 8734) from August 1995–1997 (in units of V
µW−1cm2nm). No average could be computed for
the time period of May 1995–1996 because of the
drift of the instrument’s response. For the Ed(340)
channel, the mean is for the time period of Decem-
ber 1996–August 1997.
May 1996–Aug. 1997
Channel ACS CV n
Ed(340) 0.01496 2.12 3
Ed(380) 0.00538 0.31 6
Ed(410) 0.02477 0.40 6
Ed(441) 0.02888 0.41 6
Ed(465) 0.02049 0.38 6
Ed(488) 0.01534 0.51 6
Ed(520) 0.01617 0.40 6
Ed(540) 0.01278 0.41 6
Ed(565) 0.01108 0.47 6
Ed(587) 0.01257 0.36 6
Ed(625) 0.01411 0.36 6
Ed(665) 0.01581 0.36 6
Ed(683) 0.01499 0.32 6
The calibration coefficients of the two surface sensors
(S/N 8729 and S/N 8734) appear to have drifted since 1992
(Figs. 21b–c, Sect. 4.3.2.1). For these, long-term averages
will be used only for time periods when the responses for
these instruments were stable. For instrument 8729, the
overall drift (1992–1995) was about 2% (Fig. 21b), but the
CV of the average slopes can be reduced to less than 1%
by dividing the data into two time periods and calculating
means for each (Table 13). There was no obvious physical
reason for this drift. Radiometer 8734 drifted about 3–4%
from May 1995 to May 1996, because of the grease accu-
mulating under the cosine collector (Fig. 21c). Because the
CVs during its first year of use were well over 1%, it was
assumed that the drift was linear and the calibration slopes
were interpolated as for deteriorating detectors. For cali-
brations after May 1996, the slopes were very steady and
long-term averages can be used (Table 14).
For most detectors, the average slope calculated for
years 1994–1996 (instrument 8728) or for 1995–1996 (in-
struments 8729 and 8734) will also be applicable to future
data. The methods used here to examine past stability will
accommodate the addition of new calibration data as it is
available; it will be possible to closely track any changes
in the performance of these instruments and adjust the
calibration coefficients accordingly.
4.5 OTHER ISSUES
Although calibration lamp behavior was the first con-
sideration when examining differences between calibra-
tions, several other factors are involved in determining the
final calibration coefficents. These may affect all of the co-
efficients calculated for a particular instrument (e.g., the
effect of immersion on the cosine collector), or, like the
lamp calibrations, may change over time (e.g., aging of the
reflectance plaque).
4.5.1 Immersion Effects
The effect of immersion in water on acrylic cosine col-
lectors is to decrease the irradiance responsivity of the ra-
diometer compared to that measured in air. It has been
determined experimentally that the immersion effects of
different cosine collectors of the same design and material
may differ by as much as 10% (Mueller 1995b). This makes
questionable the practice of applying one immersion coeffi-
cient, which is based on material and design specifications,
to all collectors in a class. The immersion coefficients for
the BOPSII and BBOP profiling instruments were mea-
sured at SDSU CHORS during 1994 and 1995, respectively
(Mueller 1995b, and Mueller 1996).
The final immersion coefficients for instrument 8728
were predicted from the linear regression (441–625 nm) or
were the average of two measurements of immersion (for
410, 665, and 683 nm). The final immersion coefficients for
instrument 8714 were predicted from the linear regression.
Nominal immersion coefficients (provided by the manufac-
turer) and those determined at CHORS for these two in-
struments are presented in Table 15 and Fig. 26. Differ-
ences from the nominal values ranged from 3.5–10%. A
single, nominal immersion coefficient cannot be applied
to all instruments and possibly, several measurements of
the immersion effect must be performed on a single instru-
ment to determine an accurate coefficient. Uncertainties of
10% in accuracy or reproducibility are unacceptable for vi-
carious calibration. The measured immersion coefficients
reported in Table 15 have been used for all calculations
of calibration slope during BBOP. It is likely that when
the immersion coefficients for the BBOP instrument have
been more clearly defined by additional immersion tests
that these calibration slopes will be recalculated.
4.5.2 Possible Plaque Aging
Figure 24a showed a slight decrease in the slopes of the
blue channels on instrument 8728 during 1996, although
the decrease was small—less than 2%. It is possible that
the plaque is yellowing or becoming soiled, or that the
43

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Table 15. Immersion coefficients (1/Fi) for radiometers used in this study (from Mueller 1995b and Mueller
1996). The nominal values were provided by the manufacturer (BSI) at the initial calibration of instrument
8728. The column headings denote the MER model and serial number, and the collector type, either Edor
Eu(upwelled irradiance).
λ[nm] Nominal 2040 (S/N 8728) 2040 (S/N 8714)
(Nominal) Immersion EdEdEu
410 0.705 0.7856 0.71311 0.71592
441 0.694 0.7637 0.71911 0.72351
465 0.691 0.7688
488 0.691 0.7736 0.72796 0.73498
510 0.694 0.7785 0.73998
520 0.695 0.7803 0.73578 0.74275
555 0.705 0.7885
565 0.708 0.7907 0.74528 0.75464
587 0.715 0.7958 0.75050 0.76090
625 0.726 0.8042 0.75915 0.77129
665 0.736 0.7642
683 0.739 0.7870
Wavelength [nm]
400 450 500 550 600 650 700
Immersion Factor
0.68
0.70
0.72
0.74
0.76
0.78
0.80
0.82
8728 Ed
8714 Ed
8714 Eu
Nominal
Fig. 26. The nominal and measured immersion coefficients for three irradiance heads on two radiometers.
44

O’Reilly et al.
Wavelength [nm]
400 450 500 550 600 650 700
1996 Transfer/1995 Transfer
0.990
0.995
1.000
1.005
1.010
F304 to F303
F304 to F305
Fig. 27. The ratio of transfer calibrations performed in May 1996 to those from May 1995 from lamp
F304, to lamps F303 and F305.
two blue detectors, Lu(410) and Lu(441), are deteriorat-
ing. Because calibrations of other recently purchased in-
struments are being monitored at ICESS, this question will
be clarified. In addition, the plaque is scheduled to be re-
calibrated by the manufacturer in the near future.
4.5.3 Quality Control Measures
These results precipitated refinements in calibration
methods and record keeping at ICESS. First, to avoid pos-
sible confusion when standard lamps are replaced, lamps
dedicated to each project were purchased so that the con-
sistency of calibrations can be monitored easily for many
years. Second, annual in-house cross-checks between lamps
were initiated in 1995.
The transfer calibration performed from lamp F304 to
F305, and F303 in May 1995 was repeated in May 1996
(Fig. 27). With one exception, lamp F305 at 465 nm, the
differences between the two transfers one year apart were
less than 0.5%, implying that the output of these three
lamps had not changed during the year following the 1995
lamp calibration at Optronic Laboratories. The agreement
was particularly good between lamps F303 and F304. In
fact, this 465 nm detector [S/N 8733, Ed(465)] was recently
replaced after it was determined to be unstable. Lamp
F303 was retired from routine calibrations in 1995 and now
serves as a standard to which any other lamp’s performance
can be compared. Transfer calibrations such as these will
be continued annually to closely monitor performance of
lamps between routine calibrations at the manufacturer
and future SIRREX workshops.
4.6 CONCLUSIONS
The variations of calibration slopes for most channels
of radiometer 8728 were less than 1% between 1992 and
1997. Because the radiometers used for BBOP appear to
be very stable, there can be excellent confidence in the
long-term averages of calibration slopes and consequently,
in the AOPs produced from profile data. These long-term
averages should be used whenever possible.
When the 1995 calibrations of lamps F303 and F304
were used to calculate slopes, there was an almost seamless
transition when F304 replaced F303 as the primary lamp
for radiometer calibrations. The methods used here to
examine past stability, accommodate the addition of new
calibration data as they become available, enabling close
monitoring of changes in instrument performance, and the
necessary adjustment of calibration coefficients.
These results show that there can be confidence in the
calibration of the MER-2040 series radiometers at the 1%
level. It appears, however, that the calibration responses
of these instruments may have been more stable than the
irradiance of the lamps used as calibration standards. Ul-
timately, the absolute calibrations of the radiometers dis-
cussed here are tied to one lamp calibration by Optronic
Laboratories at the midpoint of this time series, which it-
self, is guaranteed to about 1%. That calibration has been
compared with as many others as possible, and as a re-
sult, there can be confidence in its accuracy. It is essential
that comparisons such as these continue to ensure the high
quality of radiometer data.
45

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
Glossary
ALOHA A Long-term Oligotrophic Habitat Assessment
[the Hawaii Ocean Time-series (HOT) deep-
water station located about 100 km north of
Oahu, Hawaii].
ACS Average Calibration Slope
AI9901 Atlantic–Indian Ocean Cruise, 1999
AMT Atlantic Meridional Transect
AOP Apparent Optical Properties
BATS Bermuda Atlantic Time Series
BBOP Bermuda BioOptics Project
Ber95 Bering Sea Cruise, 1995
Ber96 Bering Sea Cruise, 1996
BOPSII Bio-Optical Profiling System II (second gen-
eration)
BSI Biospherical Instruments, Inc.
CalCOFI California Cooperative Oceanic Fisheries In-
vestigation
CARIACO Carbon Retention in a Colored Ocean
CB-MAB Chesapeake Bay–Middle Atlantic Bight
CDOM Colored Dissolved Organic Matter
CHORS Center for Hydro-Optics and Remote Sensing
COARE Coupled Ocean Atmosphere Response Exper-
iment
CoASTS Coastal Atmosphere and Sea Time Series
CoBOP Coastal Benthic Optical Properties (Bahamas)
CSC Coastal Service Center, (NOAA, SC)
CV Coefficient of Variation
CVT Calibration and Validation Team
CZCS Coastal Zone Color Scanner
DARR Data Analysis Round-Robin (workshop)
EcoHAB Ecology of Harmful Algal Blooms
EqPac Equatorial Pacific
FEL Not an acronym, but a type of irradiance lamp
designator.
FL-Cuba Florida–Cuba cruise.
FWHM Full-Width at Half-Maximum
GOM Gulf of Maine
GoA97 Gulf of Alaska Cruise, 1997
GSFC Goddard Space Flight Center
HOT Hawaii Ocean Time-series
HPLC High Performance Liquid Chromatography
ICESS Institute for Computational Earth System Sci-
ence
IDL Interactive Data Language
JES9906 Japan East Sea Cruise, 1999-06
JGOFS Joint Global Ocean Flux Study
Lab96 Labrador Sea Cruise, 1996
Lab97 Labrador Sea Cruise, 1997
Lab98 Labrador Sea Cruise, 1998
LTER Long Term Ecological Research
MCP Modified Cubic Polynomial
MBARI Monterey Bay Aquarium Research Institute
MBR Maximum Band Ratio
MER Marine Environmental Radiometer
MERIS Medium Resolution Imaging Spectrometer
MF0796 R/V Miller Freeman Cruise, 1996-07
MOCE Marine Optical Characterization Experiment
MODIS Moderate Resolution Imaging Spectroradiom-
eter
NABE North Atlantic Bloom Experiment
NASA National Aeronautics and Space Administra-
tion
NEGOM Northeast Gulf of Mexico
NIST National Institute for Standards and Technol-
ogy
NOAA National Oceanic and Atmospheric Adminis-
tration
OC2 Ocean Chlorophyll 2 algorithm
OC2v2 OC2 version 2.
OC4 Ocean Chlorophyll 4 algorithm
OC4v2 OC4 version 2.
OC4v4 OC4 version 4.
OCTS Ocean Color and Temperature Scanner
ORINOCO Orinoco River Plume
PD Percent Difference
PRR Profiling Reflectance Radiometer
RED9503 Red Tide Cruise, 1995-03
Res94 Resolute Cruise, 1994
Res95-2 Resolute Cruise, 1995
Res96 Resolute Cruise, 1996
Res98 Resolute Cruise, 1998
RMS Root Mean Square
ROAVERRS Research on Ocean–Atmosphere Variability
and Ecosystem Response in the Ross Sea
SDSU San Diego State University
SeaBAM SeaWiFS Bio-optical Algorithm Mini-
workshop
SeaWiFS Sea-viewing Wide Field-of-view Sensor
SIMBIOS Sensor Intercomparison and Merger for Bio-
logical and Interdisciplinary Oceanic Studies
SIRREX SeaWiFS Intercalibration Round-Robin Ex-
periment
SIRREX-1 The first SIRREX, July 1992.
SIRREX-2 The second SIRREX, June 1993.
SIRREX-3 The third SIRREX, September 1994.
SIRREX-4 The fourth SIRREX, May 1995.
SMAB Southern Mid-Atlantic Bight
S/N Serial number
SNR Signal-to-Noise Ratio
SPO SeaWiFS Project Office
TOGA Tropical Ocean Global Atmosphere
TOTO Tongue of the Ocean study (Bahamas)
UCSB University of California, Santa Barbara
URL Universal Resource Locator
UVA Ultraviolet-A
WOCE World Ocean Circulation Experiment
Symbols
aAbsorption coefficient.
ACoefficient.
awAbsorption coefficient for pure water.
bIntercept.
BCoefficient.
bbBackscattering coefficient.
bwBackscattering coefficient for pure water.
CaChlorophyll aconcentration.
˜
CaIn situ chlorophyll aconcentration.
46

O’Reilly et al.
ˆ
E(λ) Irradiance measured during SIRREX-4.
ˆ
E(λ) Irradiance measured using a PRR as the transfer
radiometer.
Ed(λ) Downwelled spectral irradiance.
Eu(λ) Upwelled spectral irradiance.
fFunction.
FiImmersion coefficient.
K(490) Diffuse attenuation coefficient at 490 nm.
Kw(λ) Diffuse attenuation coefficient for pure water.
Lu(z, λ) Upwelled spectral radiance.
LW(λ) Spectral water-leaving radiance.
LW(λ1) Water-leaving radiance for wavelength λ1.
LW(λ2) Water-leaving radiance for wavelength λ2.
mSlope.
nNumber of observations.
NNumber of data sets.
NFNumber of fluorometric chlorophyllasets.
NHNumber of HPLC chlorophyllasets.
R2Squared correlation coefficient.
R2log10(R490
555).
R2S log10(R490
555), see R3C , where the argument of the log-
arithm is a shorthand representation for the maxi-
mum of the three values. In an expression such as
R2S, the numerical part of the subscript refers to the
number of bands used, and the letter denotes a code
for the specific satellite sensors [S is SeaWiFS, M is
the Moderate Resolution Imaging Spectroradiome-
ter (MODIS), O is the Ocean Color and Tempera-
ture Scanner (OCTS), E is the Medium Resolution
Imaging Spectrometer (MERIS), and C is CZCS].
R3C log10(R443
550 >R
520
550),
R4E log10(R443
560 >R
490
560 >R
510
560), see R3C .
R4O log10(R443
565 >R
490
565 >R
520
565), see R3C .
R4S log10(R443
555 >R
490
555 >R
510
555), see R3C .
RA
BRrs ratio constructed from band Adivided by band
B.
Rrs Remote sensing reflectance.
˜
Rrs In situ remote sensing reflectance.
ˆ
Rrs Interpolated remote sensing reflectance.
Rλi
λjA compact notation for the Rrs(λi)/Rrs (λj) band
ratio.
xThe abscissa.
yThe ordinate.
zDepth.
γalog(Ca).
λWavelength.
σStandard deviation.
References
Austin, R.W., and Petzold, T., 1981: The determination of
the diffuse attenuation coefficient of sea water using the
Coastal Zone Color Scanner. Oceanography from Space,
J.F.R. Gower, Ed., Plenum Press, 239–256.
Evans, R.H., and H.R. Gordon, 1994: Coastal zone color scan-
ner “system calibration”: A retrospective examination, J.
Geophys. Res., 99, 7,293–7,307.
Firestone, E.R., and S.B. Hooker, 1998: SeaWiFS Prelaunch
Technical Report Series Final Cumulative Index. NASA
Tech. Memo. 1998–104566, Vol. 43, S.B. Hooker and E.R.
Firestone, Eds., NASA Goddard Space Flight Center,
Greenbelt, Maryland, 4–8.
Gordon, H.R., and W.R. McCluney, 1975: Estimation of the
depth of sunlight penetration in the sea for remote sensing.
Appl. Opt., 14, 413–416.
, and K. Ding, 1992: Self-shading of in-water optical in-
struments, Limnol. Oceanogr., 37, 491–500.
, and M. Wang, 1994: Retrieval of water-leaving radiance
and aerosol optical thickness over the oceans with Sea-
WiFS: A preliminary algorithm, Appl. Opt., 33, 443–452.
Gregg, W.W., and R.H. Woodward, 1998: Improvements in
high frequency ocean color observations: Combining data
from SeaWiFS and MODIS, IEEE Trans. Geosci. Remote
Sens., 36, 1,350–1,353.
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. 37, S.B. Hooker and E.R. Firestone, Eds.,
NASA Goddard Space Flight Center, Greenbelt, Mary-
land, 65 pp.
Kahru, M., and B.G. Mitchell, 1998a: Spectral reflectance and
absorption of a massive red tide off Southern California. J.
Geophys. Res., 103, 21,601–21,609.
, and , 1998b: Evaluation of instrument self-shading
and environmental errors on ocean color algorithms. Proc.
Ocean Optics XIV, Kona, Hawaii, S. Ackleson and J.
Campbell, Eds., [Available on CD-ROM.]
Maritorena, S., and J.E. O’Reilly, 2000: OC2v2: “Update on
the initial operational SeaWiFS chlorophyll aalgorithm.”
In: O’Reilly, J.E., and 24 Coauthors, 2000: SeaWiFS Post-
launch Calibration and Validation Analyses, Part 3. NASA
Tech. Memo. 2000–206892, Vol. 11, S.B. Hooker and E.R.
Firestone, Eds., NASA Goddard Space Flight Center,
Greenbelt, Maryland, 3–8.
McClain, C.R., 2000: “SeaWiFS postlaunch calibration and
validation overview.” In: McClain, C.R., E.J. Ainsworth,
R.A. Barnes, R.E. Eplee, Jr., F.S. Patt, W.D. Robinson,
M. Wang, and S.W. Bailey, SeaWiFS Postlaunch Cali-
bration and Validation Analyses, Part 1. NASA Tech.
Memo. 2000–206892, Vol. 9, S.B. Hooker and E.R. Fire-
stone, Eds., NASA Goddard Space Flight Center, Green-
belt, Maryland, 4–12.
47

SeaWiFS Postlaunch Calibration and Validation Analyses, Part 3
, and G.S. Fargion, 1999: SIMBIOS Project 1999 Annual
Report, NASA Tech. Memo. 1999–209486, NASA God-
dard Space Flight Center, Greenbelt, Maryland, 128 pp.
Morel, A., 1974: “Optical properties of pure water and pure sea-
water.” In: Optical Aspects of Oceanography, N.G. Jerlov
and E. Steemann Nielsen, Eds., Academic Press, San Diego,
California, 1–24.
, and S. Maritorena, 2000: Bio-optical properties of oceanic
waters: a reappraisal. J. Geophys. Res., (submitted).
Mueller, J.L., 1993: The First SeaWiFS Intercalibration Round-
Robin Experiment, SIRREX-1, July 1992. NASA Tech.
Memo. 104566, Vol. 14, S.B. Hooker and E.R. Firestone,
Eds., NASA Goddard Space Flight Center, Greenbelt,
Maryland, 60 pp.
, 1995a: “An integral method for analyzing irradiance and
radiance attenuation profiles.” In: Siegel, D.A., M.C.
O’Brien, J.C. Sorensen, D.A. Konnoff, E.A. Brody, J.L.
Mueller, C.O. Davis, W.J. Rhea, and S.B. Hooker, 1995:
Results of the SeaWiFS Data Analysis Round-Robin
(DARR-94), July 1994. NASA Tech. Memo. 104566, Vol.
26, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard
Space Flight Center, Greenbelt, Maryland, 44–52.
, 1995b: “Comparison of irradiance immersion coefficients
for several Marine Environmental Radiometers (MERs).”
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. 27, S.B. Hooker, E.R.
Firestone, and J.G. Acker, Eds., NASA Goddard Space
Flight Center, Greenbelt, Maryland, 3–15.
, 1996: MER-2040 SN 8728: Irradiance Immersion Factors,
CHORS Tech. Memo. 004–96, Center for Hydro-Optics
and Remote Sensing, San Diego State University, San Die-
go, California, 3 pp.
, B.C. Johnson, C.L. Cromer, J.W. Cooper, J.T. McLean,
S.B. Hooker, and T.L. Westphal, 1994: The Second Sea-
WiFS Intercalibration Round-Robin Experiment, SIR-
REX -2, June 1993. NASA Tech. Memo. 104566, Vol. 16,
S.B. Hooker and E.R. Firestone, Eds., NASA Goddard
Space Flight Center, Greenbelt, Maryland, 121 pp.
, and R.W. Austin, 1995: Ocean Optics Protocols for Sea-
WiFS Validation, Revision 1. NASA Tech. Memo. 104566,
Vol. 25, S.B. Hooker, E.R. Firestone, and J.G. Acker, Eds.,
NASA Goddard Space Flight Center, Greenbelt, Mary-
land, 67 pp.
, B.C. Johnson, C.F. Cromer, S.B. Hooker, J.T. McLean,
and S.F. Biggar, 1996: The Third SeaWiFS Intercalibra-
tion Round-Robin Experiment (SIR- REX -3), 19–30 Sep-
tember 1994. NASA Tech. Memo. 104566, Vol. 34, S.B.
Hooker, E.R. Firestone, and J.G. Acker, Eds., NASA God-
dard Space Flight Center, Greenbelt, Maryland, 78 pp.
, and C.C. Trees, 1997: “Revised SeaWiFS prelaunch al-
gorithm for the diffuse attenuation coefficient K(490).”
In: Yeh, E-n., R.A. Barnes, M. Darzi, L. Kumar, E.A.
Early, B.C. Johnson, and J.L. Mueller, 1997: Case Studies
for SeaWiFS Calibration and Validation, Part 4. NASA
Tech. Memo. 104566, Vol. 41, S.B. Hooker and E.R. Fire-
stone, Eds., NASA Goddard Space Flight Center, Green-
belt, Maryland, 18–21.
O’Reilly, J.E., S. Maritorena, B.G. Mitchell, D.A. Siegel, K.L.
Carder, S.A. Garver, M. Kahru, and C. McClain, 1998:
Ocean color chlorophyll algorithms for SeaWiFS, J. Geo-
phys. Res., 103, 24,937–24,953.
Pope, R.M., and E.S. Fry, 1997: Absorption spectrum (380–
700 nm) of pure water, II. Integrating cavity measurements,
Appl. Opt., 36, 8,710–8,723.
Siegel, D.A., M.C. O’Brien, J.C. Sorensen, D.A. Konnoff, E.A.
Brody, J.L. Mueller, C.O. Davis, W.J. Rhea, and S.B.
Hooker, 1995: Results of the SeaWiFS Data Analysis
Round-Robin (DARR-94), July 1994. NASA Tech. Memo.
104566, Vol. 26, S.B. Hooker and E.R. Firestone, Eds.,
NASA Goddard Space Flight Center, Greenbelt, Mary-
land, 58 pp.
Smith, R.C., and K.S. Baker, 1981: Optical properties of the
clearest natural waters (200–800 nm). Appl. Opt., 20, 177–
184.
, D.A. Menzies, and C.R. Booth, 1997: Oceanographic Bio-
Optical Profiling System II, Ocean Optics XIII, S.G. Ack-
elson and R. Frouin, Eds., Proc. SPIE, 2963, 777–789.
Wang, M., 2000: “The SeaWiFS atmospheric correction al-
gorithm updates.” In: McClain, C.R., E.J. Ainsworth,
R.A. Barnes, R.E. Eplee, Jr., F.S. Patt, W.D. Robinson,
M. Wang, and S.W. Bailey, SeaWiFS Postlaunch Cali-
bration and Validation Analyses, Part 1. NASA Tech.
Memo. 2000–206892, Vol. 9, S.B. Hooker and E.R. Fire-
stone, Eds., NASA Goddard Space Flight Center, Green-
belt, Maryland, 57–63.
Zibordi, G., and G.M. Ferrari, 1995: Instrument self-shading
in underwater optical measurements: experimental data,
Appl. Opt., 34, 2,750–2,754.
, J.P. Doyle, and S.B. Hooker, 1999: Offshore tower shad-
ing effects on in-water optical measurements. J. Atmos.
Ocean. Technol., 16, 1,767–1,779.
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 Tech. Memo.
1998–206892, Vol. 1, S.B. Hooker and E.R. Firestone, Eds.,
NASA Goddard Space Flight Center, Greenbelt, Mary-
land, 58 pp.
48

O’Reilly et al.
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´alez-Ben´
itez, I. Huskin, M. Quevedo, R.
Barciela-Fernandez, C. de Vargas, and C. McKee, 1998:
AMT-5 Cruise Report. NASA Tech. Memo. 1998–206892,
Vol. 2, S.B. Hooker and E.R. Firestone, Eds., NASA God-
dard 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, Mary-
land, 40 pp.
Vol. 4
Johnson, B.C., E.A. Early, R.E. Eplee, Jr., R.A. Barnes, and
R.T. Caffrey, 1999: The 1997 Prelaunch Radiometric Cal-
ibration of SeaWiFS. NASA Tech. Memo. 1999–206892,
Vol. 4, S.B. Hooker and E.R. Firestone, Eds., NASA God-
dard Space Flight Center, Greenbelt, Maryland, 51 pp.
Vol. 5
Barnes, R.A., R.E. Eplee, Jr., S.F. Biggar, K.J. Thome, E.F.
Zalewski, P.N. Slater, and A.W. Holmes 1999: The Sea-
WiFS Solar Radiation-Based Calibration and the Transfer-
to-Orbit Experiment. NASA Tech. Memo. 1999–206892,
Vol. 5, S.B. Hooker and E.R. Firestone, Eds., NASA God-
dard Space Flight Center, 28pp.
Vol. 6
Firestone, E.R., and S.B. Hooker, 2000: SeaWiFS Postlaunch
Technical Report Series Cumulative Index: Volumes 1–5.
NASA Tech. Memo. 2000–206892, Vol. 6, S.B. Hooker and
E.R. Firestone, Eds., NASA Goddard Space Flight Center,
Greenbelt, Maryland, 14 pp.
Vol. 7
Johnson, B.C., H.W. Yoon, S.S. Bruce, P-S. Shaw, A. Thomp-
son, S.B. Hooker, R.E. Eplee, Jr., R.A. Barnes, S. Mar-
itorena, and J.L. Mueller, 1999: The Fifth SeaWiFS In-
tercalibration Round-Robin Experiment (SIRREX-5), July
1996. NASA Tech. Memo. 1999–206892, Vol. 7, S.B.
Hooker and E.R. Firestone, Eds., NASA Goddard Space
Flight Center, 75 pp.
Vol. 8
Hooker, S.B., and G. Lazin, 2000: The SeaBOARR-99 Field
Campaign. NASA Tech. Memo. 2000–206892, Vol. 8, S.B.
Hooker and E.R. Firestone, Eds., NASA Goddard Space
Flight Center, 46 pp.
Vol. 9
McClain, C.R., E.J. Ainsworth, R.A. Barnes, R.E. Eplee, Jr.,
F.S. Patt, W.D. Robinson, M. Wang, and S.W. Bailey,
2000: SeaWiFS Postlaunch Calibration and Validation
Analyses, Part 1. NASA Tech. Memo. 2000–206892, Vol.
9, S.B. Hooker and E.R. Firestone, Eds., NASA Goddard
Space Flight Center, 82pp.
Vol. 10
McClain, C.R., R.A. Barnes, R.E. Eplee, Jr., B.A. Franz, N.C.
Hsu, F.S. Patt, C.M. Pietras, W.D. Robinson, B.D. Schie-
ber, G.M. Schmidt, M. Wang, S.W. Bailey, and P.J. Wer-
dell, 2000: SeaWiFS Postlaunch Calibration and Valida-
tion Analyses, Part 2. NASA Tech. Memo. 2000–206892,
Vol. 10, S.B. Hooker and E.R. Firestone, Eds., NASA God-
dard Space Flight Center, 57pp.
Vol. 11
O’Reilly, J.E., and 24 Coauthors, 2000: SeaWiFS Postlaunch
Calibration and Validation Analyses, Part 3. NASA Tech.
Memo. 2000–206892, Vol. 11, S.B. Hooker and E.R. Fire-
stone, Eds., NASA Goddard Space Flight Center, 49 pp.
49