A modular atmospheric correction algorithm is proposed that uses atmospheric and water contents models to predict the visible and near-infrared reflectances observed by a satellite over water. These predicted values are compared with the satellite reflectances at each pixel, and the model parameters changed iteratively with an error minimization algorithm. The default atmospheric model uses single-scattering theory with a correction for multiple scattering based on lookup tables. With this model we used parameters of the proportions of three tropospheric aerosol types. For the default water content model we need the parameters of the concentrations of chlorophyll, inorganic sediment, and gelbstoff. The diffuse attenuation and backscatter coefficients attributed to these constituents are calculated and used to derive the water-leaving reflectance. Products include water-leaving reflectance, concentrations of water constituents, and aerosol optical depth and type. We demonstrate the application of the method to sea-viewing wide field-of-view sensor by using model data.