Determination of optimum aerosol optical thickness ratios for atmospheric correction of coastal zone color scanner data in the Georges Bank-Gulf of Maine Region

Abstract

The influence of varying aerosol optical thickness ratios (AOTR) on the precision and accuracy of satellite-derived CZCS pigment was evaluated using 92 full-resolution CZCS scenes collected along the U.S. northeast coast (NEC) from 1979 to 1986. CZCS pigment was estimated using the standard CZCS two channel bio-optical algorithm and seven sets of AOTR values including individual ratios computed interactively for each NEC image using the “clear water” radiance technique (INDIVID); means of the individual NEC ratios (NECAVE, N = 92); means of individual ratios computed from Middle Atlantic Bight (MAB) and NEC scenes (NECMAB, N = 276); means of individual ratios computed from South Atlantic Bight (SAB), MAB and NEC scenes (ECOAST, N = 1091); “default” values for global marine aerosols (GLOBAL); and time-varying ratios derived from weak annual signals present in the individual ratios used to compute NECMAB and ECOAST means. Precision was estimated using geometric mean CZCS pigment values computed from the seven sets of images for three 70 × 70 km sub-domains located in the western Gulf of Maine, on southern Georges Bank and over the continental slope. The smallest root mean square difference (RMSD) for geometric mean CZCS pigment (<0.05 log 10 units) was obtained between images processed using means of individual ratios and time-varying ratios, while the largest RMSD (generally >0.1 log 10 units) was obtained between images processed using GLOBAL and INDIVID ratios. Accuracy was estimated using geometric mean CZCS pigment values computed from the seven sets of images and a sea truth data set. After removal of suspect CZCS pigment values, agreement between CZCS pigment and sea truth data was highest using ECOAST mean ratios and a spatial window diameter (temporal window) of 7 km (±0.5 days) for N = 546, displaying RMSD (log 10 units), linear regression slope and correlation coefficient of 0.19, 1.00 and 0.88, respectively. ECOAST “mean” AOTR values resulted in a superior atmospheric correction for computation of CZCS pigment values from the Georges Bank-Gulf of Maine region when compared with values computed using “default” global or individual “interactively determined” ratios. ECOAST mean AOTR values should provide a consistent regional atmospheric correction for all CZCS scenes from the Georges Bank-Gulf of Maine region, including scenes which lack low pigment water and therefore cannot make use of the “clear-water” radiance technique.