(essentially uncertainty for biological purposes) in the relationship of the water-
leaving spectral signal to variations in chlorophyll concentrations. It must be pointed
out that not all of the variation comes from the satellite estimates. Chlorophyll is not
measured by water sampling without error or variation, and modest variation is
expected across the range of pixels included in the patches averaged for comparison
to the ship data. Furthermore, roughly equal amounts of in-water light attenuation,
and thus water-leaving spectral variation, derive from chlorophyll and from accessory
pigments of phytoplankton. At any given concentration, there is significant variation
in the ratio of their concentrations and thus their effects, but the ratios examined (on
log–log scales) over the entire range from severe oligotrophy to pea-soup green
waters are well fitted by a 1 : 1 relationship (Trees et al. 2000). Those two aspects of
chlorophyll and accessory pigment variation contribute both to the overall excellence
of chlorophyll-a (Ca) estimates relative to in situ estimates and to the substantial
variation around the general trend. While the satellite estimates of Ca are not always
accurate for any given pixel, with sufficient areal averaging the data are very useful
because the estimator shows no obvious bias. Regional and global totals benefit from
massive averaging and should be reasonably accurate.
Fig. 2.16 Comparison of OC4v4 SeaWiFS chlorophyll estimates (text Eqn. 3)
(ordinate) to the NOMAD shipboard measures (abscissa) of sea-surface chlorophyll
from a wide range of ocean sites. The heavy line represents a 1 : 1 relationship, and
the good fit of the central tendency implies biases are small. The thin lines represent
offsets from one-third to three-fold, encompassing most of the comparisons.
(^) (After Bailey & Werdell 2006.)