Retrieval of Physical Properties of Mudflat Sediments from Hyperspectral Data Using the Modified Gaussian Model and Spectral Curve Fitting

The objective of this work is to retrieve parameters relevant to the understanding of sediment dynamics in a bay, Bourgneuf Bay (France) from hyperspectral remote sensing data. The erodability of mudflats is strongly influenced by the sediments bio-physical characteristics. These parameters include sediment grain size, moisture content and mineralogy as well as presence or absence of biofilm, all having an influence on sediment cohesiveness. The surface spectral signature gives access to mineralogy through specific absorption features. Sediment grain size influences the scattering / absorption relationship which translates into band shape, average surface reflectance and the general shape of the spectrum also called the continuum. Moisture content also influences the shape of the continuum as well as the strength of the specific liquid water absorption features observable in the NIR. Finally chlorophyll-a concentration derived from the strength of the absorption at 675 nm, is a good proxy for biofilm biomass. In order to access all these characteristics, we used the Modified Gaussian Model [MGM] to deconvolve the reflectance spectra into constituent absorptions and a continuum curve. This inversion process was semiautomatic. In this paper, an automated method for hyperspectral data is proposed, using as a first step spectroscopic derivative analysis to calculate an initial solution for the MGM input parameters.