Abstract
A general function field methodology for estimating ocean color variables from space is applied to the retrieval of spectral marine reflectance from Global Imager (GLI) data. The top-of-atmosphere GLI reflectance vectors, after correction for molecular effects, are considered as explanatory variables conditioned by the angular geometry. The inverse problem, therefore, is viewed as a collection of similar inverse problems, continuously indexed by the angular variables. The solution is in the form of a field of nonlinear regression models over the set of permitted values for the angular variables. The selected models, for reasons of approximation theory, are fields of shifted ridge functions. The fields constructed on synthetic GLI data for Case 1 waters are robust to noise, they handle well situations of weakly and strongly absorbing aerosols, and the retrievals are accurate in both oligotrophic and productive waters. In the presence of 1% noise, the RMS error is 0.0006 (4.2%) at 380 nm, 0.0003 (2.8%) at 460nm, and 0.0001 (1.5%) at 545nm, i.e., well within the acceptable limits for quantitative biology applications. The theoretical results, and the possible extensions, show the potential of the function field methodology for operational estimation of marine reflectance from GLI data, even in the near ultraviolet.
