EXPLORING THE ISSUE OF BOTTOM REFLECTANCE PARAMETERISATION IN SHALLOW WATER INVERSION MODELS
A key processing step in the optical classification of benthic substrates from hyperspectral imagery of coral reef environments is correcting for the radiometric effects of the water column to derive bottom reflectance. Shallow water inversion models are a proven methodology able to simultaneously retrieve the water column's inherent optical properties (IOPs), geometric depth and bottom reflectance from remote sensing reflectance with little or no in situ data. However the accuracy of the retrieved IOPs and depth is influenced partly by the spectral shape of the benthic endmember(s) that are input to the inversion model. Current inversion schemes parameterise the bottom reflectance as a single or a linear mixture of two or three benthic endmembers. However, the efficacy of these parameterisations remains untested given the narrow water column penetrating bands and the exponential attenuation of light with depth. Here, we analyse three parameterisations of bottom reflectance previously published: (1) a single endmember scaled by its brightness; (2) the fractional coverage of two endmembers, and; (3) a linear mixture of three endmembers. Here each parameterisation utilises the same IOP model and optimisation algorithm allowing for more readily comparable results. Issues such as over-parameterisation and the ability to detect benthic mixtures from above water reflectance will also be addressed. We evaluate these approaches using both in situ radiometry and airborne imagery from NASA's Portable Remote Imaging SpectroMeter (PRISM) of coral reefs.
Garcia, R. A., University of Massachusetts Boston, USA, email@example.com
Lee, Z., University of Massachusetts Boston, USA, Zhongping.Lee@umb.edu
Presentation is given by student: No