Influence of dust and sulfate aerosols on ocean color spectra and chlorophyll <i>a</i> concentrations derived from SeaWiFS off the U.S. east coast

Abstract

The influence of two dominant aerosol species on satellite ocean color retrievals is examined off the U.S. East Coast in the western Sargasso subtropical gyre. Waters of very low chlorophyll concentration have normalized water‐leaving radiance ( nLw ) spectra highest in blue and decreasing monotonically with increasing wavelength. For water with chlorophyll concentrations less than 0.13 mg m −3 we compared the Sea‐viewing Wide Field‐of‐view Sensor (SeaWiFS) nLw spectra for different aerosol conditions (clear, dust‐dominated, and sulfate‐dominated) over 1 year between March 1999 and March 2000. With appropriate atmospheric correction, satellite‐derived nLw spectra should be insensitive to the presence of atmospheric aerosols, but we found the SeaWiFS spectra to be sensitive to the species and optical thickness of aerosols. The SeaWiFS bio‐optical chlorophyll algorithms use nLw (λ)/ nLw (555) band ratios (where λ = 443, 490, or 510 nm): any biases in the band ratios result in incorrect chlorophyll concentration estimates. When aerosols were negligible, nLw spectra were consistent with in situ reference data between 443 and 510 nm. With increasing aerosol optical thickness during dust events, most common during summer, the nLw spectra were lowered between 412 and 510 nm, decreasing nLw (λ)/ nLw (555) band ratios and resulting in artificially high chlorophyll a estimates. Sulfate‐dominated pixels were associated with elevated nLw spectra between 412 and 555 nm. Increasing sulfate optical thickness corresponded to decreases of the bio‐optical band ratios because of the increase at 555 nm, which also biased the chlorophyll a estimates high. The effect of sulfate upon ocean color retrievals is more problematic than dust off the U.S. east coast because of the nearly constant presence of sulfate along with commonly colocated pollutant aerosols that confound atmospheric correction algorithms.