| Hyperspectral imagery has been shown to be a powerful technology for quantitative monitoring of shallow water coastal environments. In coastal remote sensing, to estimate sea bottom properties from hyperspectral imagery, we need to remove the effects of the atmosphere and the water column from the measured spectral signature. In our work, we use a standard algorithm available from NRL to correct for atmospheric effects in hyperspectral imagery to retrieve the water leaving remote sensing reflectance, R rs, from which the subsurface remote sensing reflectance, rrs, is retrieved. Here, we present results in the development of an algorithm combining inversion and unmixing models to retrieve bottom reflectance, water column optical properties, bathymetry, and benthic composition from subsurface remote sensing reflectance. A bio-optical model developed by Z. P. Lee in 1998 and 1999 relates Rrs , and rrs, to the water optical properties (OP's), depth, and bottom reflectance. We employ an iterative algorithm to retrieve the parameters of interest. As in Goodman (2004), Lee's original model is enhanced by adding a linear mixing model for approximating bottom composition, which is used to extract subpixel information in low spatial resolution satellite and airborne hyperspectral sensors. Results using both simulated data and AVIRIS imagery from Hawaii are presented. |
