40 
ing surfaces), its calibration is valid only for 
scenes characterized by unpolarized radiance. 
However, in the ocean-atmosphere system, the 
largest signal contributor (and largest signal vari- 
ation) is Rayleigh scattering radiance, which is 
very strongly polarized. Moreover, the polariza- 
tion varies greatly across the range of relevant 
scan angles. Since the instrument's error due to 
polarization sensitivity may also vary with scan 
angle, the problem is compounded. These errors 
would be very difficult to correct, even using 
completely polarized radiative-transfer models. 
While the problem might be possible to solve at a 
satisfactory level of accuracy, the expense and 
uncertainty are far greater than would be the 
case with an instrument designed with a low pola- 
rization sensitivity. The CZCS specification was 
2%, and this level of polarization sensitivity was 
deemed satisfactory. 
Solid Angle Resolution of IFOVs 
If a radiometer were designed with an optics 
train that varied the solid angle subtense of the 
instrument's field of view, e.g., to correct for foot- 
print distortion, that variation could lead to a loss 
of relative precision in the cross-scan direction or 
between tilt configurations. If the instrument's 
optical configuration is modified to vary the angu- 
lar IFOV for any reason, then the calibration of 
each distinct angular resolution configuration be- 
comes essentially independent. Aside from the 
complexity inherent in calibrating such a system, 
the loss of relative precision would cause the er- 
rors in achievable atmospheric corrections to 
reach totally unacceptable levels for any quantit- 
ative use of the data. In addition to these unten- 
able results, a variable-resolution design would 
be extremely difficult to monitor for degradation of 
the overall radiometric sensitivity. 
These effects can be avoided only if the instru- 
ment is designed in such a way that the product 
of the detector's solid angle and the area of the 
final optics aperture (or the area of the detector 
and the final optics solid angle) are constant for 
all instrument fields of view. 
Dynamic Range 
Both panels emphasized that the dynamic 
range of the SeaWiFS instrument must be such 
that subtle variations in reflectance and tempera- 
ture in open-ocean scenes of interest can be de- 
tected as well as major variations in scenes with 
high entropy, such as coastal waters. Within a 
fixed number of available quantizing bits these 
two requirements might conflict. Therefore, on- 
board adjustment of dynamic range is desirable. 
Possible solutions include programmable gain 
changes and nonlinear encoding. Careful con- 
sideration must also be given to the quantization 
of the points of the on-board calibration curve 
(i.e., from space and on-board calibration tar- 
get(s)). It is anticipated that 10 bits of digitization 
should provide sufficient range for detection of 
the radiance from typical open-ocean scenes. 
However, additional studies should be conducted 
to firmly establish the required range and to de- 
termine whether gain changes, additional bits, 
and/or nonlinear digitization are desirable. Ad- 
justable gain could also be used to enhance the 
signal at the reduced light levels occurring near 
the twilight portion of the orbit and to reduce the 
signal when employing the diffuser plate for in- 
flight instrument calibration. 
Bright Target Saturation 
The CZCS instrument experienced saturation- 
induced errors immediately after scanning over 
bright clouds or land. Following saturation, the 
data became unusable for distances up to 100 
km, depending on the brightness of the clouds 
and their spatial extent. The SeaWiFS design 
should incorporate protective circuitry to minimize 
or, if possible, eliminate this type of instrument arti- 
fact. 
Locational Accuracy 
The location of data from all pixels in latitude- 
longitude coordinates is important for quantitative, 
scientific use of ocean-color imagery. In order to 
make these data useful, scientists must be able to 
