In a variety of remote sensing applications inverse modeling is used to retrieve parameters of ourenvironment from measurements of the radiance spectrum. As the dimensionality of the data increaseswith the number of spectral channels or as models become more complicated, inverse modeling becomes expensive in terms of computing time. We describe a technique to parameterize a given inverse model, which reduces the necessary computing time by orders of magnitude. An example that uses simulated data of SeaWiFS, an advanced ocean color sensor to be launched in 1993, is given, whereby the concentrations of suspended matter, phytoplankton, and gelbstoff (yellow substance) in water are determined.