Effects of Finite Spectral Bandwidth in Multispectral Optical Bathymetry

摘要

Optical, multispectral remote sensing data from satellite and aircraft can beused to determine depths in shallow, clear waters. The physical principle used to calculate depths from received irradiances is based on the exponential attenuation of the light as a function of distance travelled through the water and the reflectivity of the bottom, both as a function of wavelength. An underlying, often unstated, assumption made is that the variation of attenuation and bottom reflectivity across a sensor spectral band (typically 60 to 80 nanometers for Landsat Thematic Mapper) can be ignored and that the irradiance measurement is valid at the band center. A more realistic simple model for a finite bandwidth sensor is derived in this paper using a piece-wise linear approximation. Using this model the limitations of conventional log-linear regression models for multispectral bathymetry are investigated. The primary conclusion is that for typical expected variations in attenuation and reflectivity as a function of wave-length, a multiband, log-linear regression model can be used for bathymetry calculations. However, straight-forward identification of the physical parameters (attenuation coefficient, bottom reflectivity) is invalid.