Ocean color satellite is mainly designed for oceanic waters, but most of their bands are easily saturated in highly turbid waters. It's a great challenge to monitor water quality using satellites, because there are no operational algorithm to estimate chlorophyll-a concentration (Chl.a) in inland waters. In this paper, a novel approach based on the Empirical Orthogonal Function (EOF) analysis was developed using the spectral variance from the Rayleigh-corrected reflectance data (R_rc) at 469, 555, 645 and 859 nm, and Chl.a was estimated by considering the regression relationships between the spectral variance and 55 concurrent field measurements. The validation was then performed using independent data (N=40) from other laboratory, with R²=0.63, URMSE=85.46%. Application of the algorithm to MODIS images showed that spatial distribution patterns and seasonal changes are reasonable. Besides, this algorithm is immune to perturbations from thick aerosols and different sediments. Hence, we suggest that EOF algorithm could be applied to estimate Chl.a concentrations in inland waters.