A regression technique for evaluation and quantification for water quality parameters from remote sensing data

The paper attempts to define optical physics and/or environmental conditions under which the linear multiple-regression should be applicable. It is reported that investigation of the signal response shows that the exact solution for a number of optical physics conditions is of the same form as a linearized multiple-regression equation, even if nonlinear contributions from surface reflections, atmospheric constituents, or other water pollutants are included. Limitations on achieving this type of solution are defined. Laboratory data are used to demonstrate that the technique is applicable to water mixtures which contain constituents with both linear and nonlinear radiance gradients. Finally, it is concluded that instrument noise, ground-truth placement, and time lapse between remote sensor overpass and water sample operations are serious barriers to successful use of the technique.