Bayesian estimation of optical properties of nearshore estuarine waters: A gibbs sampling approach

A novel approach is developed for the retrieval of inherent optical properties of coastal water, from which waterquality constituent concentrations can be obtained. The technique combines an analytical bio-optical model with statistical modeling for the formulation of posterior probability distributions of phytoplankton absorption, backscattering, and colored dissolved organic matter absorption; a Gibbs Sampler is employed for optimization. In contrast to other methods that typically provide point estimates of the unknown parameters, the proposed method estimates posterior distributions of the parameters, quantifying the uncertainty present in the problem and revealing correlation patterns. The method is tested successfully on synthetic reflectance data and real datameasured in situ in the Hudson/Raritan Estuary of New York-New Jersey.