In many algorithms designed to retrieve water constituent concentrations, an assumption of negligible water-leaving radiance is made in the near infrared (NIR) wavelengths (750-950 [nm]). This allows fairly accurate atmospheric correction to be applied to the oceanic imagery. Given this assumption, it is possible to derive model estimates of aerosol type and density and compensate for its effects in other regions of the spectrum. Unfortunately this assumption is only valid in areas of very low total suspended sediment (TSS) concentrations, such as open ocean waters. Increased TSS load causes significant back-scattering within the water, increasing the water leaving signal in the NIR region, which confounds the compensation algorithms and incorrectly attributes all the effects entirely to aerosols. A possible solution to this problem is to model suspended sediment inherent optical properties (IOPs) in the NIR using the Ocean Optical Phytoplankton Simulator (OOPS) and couple this data using atmospheric and hydrologic radiative transfer models (MODTRAN and HYDROLIGHT, respectively). These models can be combined for a given sensor geometry to predict sensor reaching radiance and match the resulting spectra with actual image spectra. This is done including the NIR which should allow us to better characterize the sediment load. Look-up tables (LUTs) of radiances from modeled atmospheres and reflectances from modeled water allow us to make this an iterative process which can be optimized to give us the best match for the pixel in question. The products of this algorithm are the water constituent concentrations and the atmospheric parameters. This process has been implemented for AVIRIS hyperspectral imagery captured over the Rochester, NY embayment with reasonable success in atmospheric and constituent retrieval.