Robust remote sensing algorithms to derive the diffuse attenuation coefficient for lakes and coastal waters

In this study, empirical and semianalytical algorithms are developed and compared for optically complex waters to retrieve the diffuse attenuation coefficient of downwelling irradiance (Kd(lambda)) from satellite data. In the first approach, a band ratio algorithm was used. Various sets of MERIS band ratios were tested to achieve the best estimates for K-d(490) based on the in situ dataset which was measured in Nordic lakes (oligotrophic to eutrophic conditions). In the second approach, K-d(490) was expressed as a function of inherent optical properties which were retrieved from MERIS standard products. The algorithms from both approaches were tested against an independent data set and validated in optically complex coastal waters in the Baltic Sea and in Nordic lakes with high concentrations of coloured dissolved organic matter (0.3 < a(cdom)(442) m(-1) < 4.5), chlorophyll a (Chl a) (0.7< C-Chl a(mg m(-3))< 67.5) and total suspended matter (TSM) (0.5 < C-TSM(g m(-3)) < 26.4). MERIS-derived K-d(490) values showed reliable estimates in case of both methods. The results indicate that for band ratio algorithms, the root mean square error (RMSE) decreases and the coefficient of determination (R-2) increases when using longer wavelengths in the visible spectrum as a reference band. It was found that the best estimates were retrieved from MERIS data when using the ratio of R-rs(490)/R-rs(709) for coastal waters (K-d(490) < 2.5 m(-1)) and the ratio R-rs(560)/R-rs(709) for more turbid inland waters (Kd(490) > 2.5 m(-1)). As a result, a combined band ratio algorithm was developed, which provides a promising approach R-2 = 0.98, RMSE= 17%, N = 34, p < 0.05) for estimating K-d(490) over a wide range of values (0.3-6.1 m(-1)).