Atmospheric Correction of SeaWiFS Imagery over Turbid Coastal Waters: A Practical Method

Abstract

The current SeaWiFS algorithms frequently yield negative water-leaving radiance values in turbid Case II waters primarily because the water-column reflectance interferes with the atmospheric correction based on the 765-nm and 865-nm spectral bands. Here we present a simple, practical method to separate the water-column reflectance from the total reflectance at 765 nm and 865 nm. Assuming the type of aerosol does not vary much over relatively small spatial scales (∼50–100 km), we first define the aerosol type over less turbid waters. We then transfer it to the turbid area by using a “nearest neighbor” method. While the aerosol type is fixed, the concentration can vary. This way, both the aerosol reflectance and the water-column reflectance at 765 nm and 865 nm may be derived. The default NASA atmospheric correction scheme subsequently is used to obtain the aerosol scattering components at shorter wavelengths. This simple method was tested under various atmospheric conditions over the Gulf of Mexico, and it proved effective in reducing the errors of both the water-leaving radiance and the chlorophyll concentration estimates. In addition, in areas where the default NASA algorithms created a mask due to atmospheric correction failure, water-leaving radiance and chlorophyll concentrations were recovered. This method, in comparison with field data and other turbid water algorithms, was tested for the Gulf of Maine and turbid, posthurricane Gulf of Mexico waters. In the Gulf of Maine it provided more accurate retrievals with fewer failures of the atmospheric correction algorithms. In the Gulf of Mexico it provided far fewer pixels with atmospheric failure than the other methods, did not overestimate chlorophyll as severely, and provided fewer negative water-leaving radiance values. Background Since the launch of the Sea-Viewing Wide Field-of-View Sensor (SeaWiFS) aboard the SeaStar satellite in August 1997, global ocean color data are available to the science community on a regular basis. SeaWiFS is superior to the original Coastal Zone Color Scanner (CZCS, Hovis et al., 1980). It has much higher radiometric sensitivity (10 bits versus 8 bits) and additional spectral bands to aid in atmospheric correction and bio-optical applications. Specifications called for uncertainties less than ±5% in retrieved water-leaving radiance and less than ±35% in chlorophyll a concentration ([chl a]) in Case I waters ( Hooker et al., 1992; “Case I” defined in Morel and Prieur, 1977). SeaWiFS, however, generally fails to deliver such fidelity in turbid or shallow coastal waters (Case II waters). Turbid water constituents (suspended sediments, bubbles, etc., or bottom reflection) can contribute significant amounts of radiance to the atmospheric correction bands (765 nm and 865 nm). This will induce large errors when applying the standard atmospheric correction scheme since in that scheme the water-leaving radiance is assumed to be negligible in the near-infrared (IR) part of the spectrum (Gordon and Wang, 1994). Also, in turbid Case II waters, the current band ratio bio-optical algorithm does not work well, simply because there are often other constituents (e.g., colored dissolved organic matter, or CDOM) whose optical properties may not covary with phytoplankton pigment concentrations (Morel and Prieur, 1977; Carder et al., 1991; Carder et al., 1999; Muller-Karger et al., 1991).