The efficacy of band weighting schemes for improving the accuracy and precision of water quality parameters estimated from MERIS and MODIS image data

Abstract

[Abstract]: Optical remote sensing has been used to map and monitor water quality parameters such as the concentrations of hydrosols (chlorophyll and other pigments, total suspended material, and coloured dissolved organic matter). One approach to estimate hydrosol concentrations is to apply a Matrix Inversion Method (MIM) to the reflectance in each band, creating a system of linear equations, and then apply a least squares method to solve for the hydrosol concentrations. The accuracy and precision of this method depends on the width, position and inherent noise in the spectral bands of the sensor being employed, as well as the radiometric corrections applied to images to calculate the subsurface reflectance. It has been suggested that by differentially weighting the band equations in over-determined systems the reliance of the solution on any one band can adjusted. The aim of this work was to establish if this is true and if so,determine the optimal weighting regime for each sensor. The Hydrolight® radiative transfer model and typical hydrosol concentrations from Wivenhoe Dam, a large freshwater storage in South East Queensland, were used to simulate 1089 reflectance spectra for MERIS and MODIS images acquired at two sun positions. The accuracy and precision of hydrosol concentrations derived from each weighting regime were evaluated after errors associated with the air-water interface correction, atmospheric correction and the IOP measurement were modelled and applied to the simulated reflectance spectra. The technique showed the ability of a weighting regime to alleviate the effect of these errors and was used as a measure of a regime’s efficacy. The results of this study will be used to improve an algorithm for the remote sensing of water quality for freshwater impoundments.