coverage uses, but also requested that the data 
be aggregated on board to provide global-area 
coverage at approximately 4.5 km resolution. 
The rationale for these requirements can be sum- 
marized as follows. 
The distribution of chlorophyll in the ocean is 
patchy on all scales down to less than a kilome- 
ter. To adequately map the variation in phyto- 
plankton concentration in high-concentration 
shelf areas (a major goal of the first of the MAR- 
EX studies), a sensor must be able to resolve 
about a kilometer of the ocean. This small foot- 
print size also allows measurements to be made 
close to the shore, permitting resolution of local 
outwelling and upwelling zones, which tend to be 
near-shore phenomena. The high data rate that 
will result from such a spatial resolution may be 
reduced somewhat for wide-area studies of 
open-ocean phytoplankton concentration, where 
statistical rather than process experiments are of 
more interest. For the latter case, a 4.5-km reso- 
lution is acceptable. Therefore, a system is re- 
quired that generates, stores, and transmits both 
resolutions of data. This type of data-processing 
system is analogous to the NOAA/AVHRR system, 
which has been used as a basis for defining the 
proposed SeaWiFS data products. 
The conversion of the higher resolution (1.13 
km) data to a global (4.5 km) data set should take 
place in the on-board data processing system 
prior to recording or transmitting the data. The 
algorithm for this conversion will be determined at 
a later date; however, the consensus was that 
averaging the 16 1.13-km pixels is not desirable 
and that selecting the output of a single cloud- 
free pixel to represent a 4.5-km area would be a 
better solution. 
Radiometric Accuracy and 
Relative Precision 
The scientific and operational utility of an 
ocean-color imaging system depends completely 
on the ability to measure radiance at the instru- 
ment's aperture with sufficient accuracy and res- 
olution to separate and remove from each meas- 
SYNTHESIS OF REQUIREMENTS 
urement the contributions from the atmosphere 
and, thus, to infer the amount of water-leaving ra- 
diance within, essentially, the limits in accuracy of 
the atmospheric-correction models. On this ba- 
sis, a radiometric accuracy of approximately 5% 
is required in each band, and the relative preci- 
sion between individual measurements within 
each band must be much less than 1%. The em- 
phasis in the expression of these requirements is 
on the importance of relative precision between 
measurements in a given image or band and be- 
tween images on different days when a given po- 
sition on the surface is viewed at different scan 
and solar angles. 
On small spatial scales, pixel-to-pixel varia- 
tions in aerosol radiance are typically of the same 
order of magnitude as variations in the amount of 
water-leaving radiance. Under hazy conditions, 
the fluctuation in the amplitudes of aerosol radi- 
ance may be 4 to 10 times larger than the varia- 
tions in water-leaving radiance associated with 
ocean fronts of similar spatial scales. As a result, 
the algorithms used to remove the aerosol effects 
require precision between bands to within 1%. 
Even limited atmospheric correction algorithms, 
sufficient to locate front and eddy boundaries, re- 
quire removal of these atmospheric effects, and 
all quantitative applications require good radio- 
metric accuracy and extremely good relative pre- 
cision. 
Polarization sensitivity and the constancy of 
the solid-angle field of view of the radiometer are 
two major factors affecting relative radiometric 
precision. 
Polarization Sensitivity 
Polarization sensitivity is defined as the ratio 
of the difference between maximum and minimum 
output to the sum of the maximum and minimum 
output obtained when the plane of incoming lin- 
early polarized radiation is rotated through 180°. 
Normally, a radiometer is calibrated using un- 
polarized input radiance. Under these circum- 
stances, if the radiometer's response is polariza- 
tion sensitive (usually because of internal reflect- 
39 
