VARIABILITY IN NEAR-SURFACE PARTICULATE ABSORPTION-SPECTRA - WHAT CANA SATELLITE OCEAN COLOR IMAGER SEE

An extensive database of approximately 400 in situ particulate absorpt ion spectra [a(p)(lambda)] is analyzed to assess the potential of usin g ocean color imagers to examine variability in the structure of the n ear-surface ocean planktonic ecosystem. This application of a(p)(lambd a) data is appropriate, as particulate absorption variations are the d ominant source of ocean color variation and are attributable to change s in the phytoplankton community structure. Empirical orthogonal funct ion (EOF) analyses are used to estimate the contribution of each stati stical mode to the total variance. The EOF analyses showed that >99% o f the variance found in the a(p)(lambda) data set can be simply attrib uted to the total amount of particulate material. When this source of variability is removed, two significant modes of variability can be id entified which comprise 79 and 18% of the normalized variance. These m odes are interpreted as representing the relative contribution of chlo rophyll-containing biomass and detrital materials, verifying the use o f two-component phytoplankton-detritus models to partition a(p)(lambda ). Only a small amount of the total a(p)(lambda) variability (<0.5% of the total) can be attributed to absorption features caused by accesso ry pigment groups. Thus, variability in a(p)(lambda) is almost entirel y associated with the quantity of the absorbing materials rather than their spectral quality (or normalized spectral shape). These results s uggest that remotely sensed ocean color spectra will reflect only thre e statistically significant components: the total amount of particulat e material, the relative amounts of chlorophyll-containing biomass, an d detrital materials. For most typical conditions it is unlikely that robust global algorithms for determining particular phytoplankton grou ps can be developed from remotely sensed ocean color spectra.