A Physical Method for Optical Characterization of Pollution in Industrial Wastewater Ponds Using Imaging Spectroscopy

Investigating the application of remote sensing to water pollution in industrial ponds is of great interest for rapid and cost-effective pollution monitoring. This article presents a method to detect pollutants and map their spatial distribution in industrial ponds using the water inherent optical properties (IOPs), namely the absorption and backscattering coefficients, derived from imaging spectroscopy data. The IOPs of industrial water pollutants remain poorly known. Current remote sensing methods are site-specific and require in situ measurements to calibrate empirically based models. Here, a generic approach is proposed based on the semianalytical radiative transfer model adapted to take into account both the absorption and backscattering coefficients of pollutant particles. The model is then inverted using an alternating multipixel method, named industrial wastewater optical characterization (IWOC), to map the spatial distribution of the pollutants. The performances of IWOC are evaluated using noise-free and noisy simulated datasets for an absorption-dominated water case and a backscattering-dominated water case. The water reflectance spectra (R-rs) for noise-free synthetic datasets are satisfactorily retrieved by the IWOC method. The optical properties of the pollutants are also well retrieved, with maximum root mean square error (RMSE) values of 2.43 x 10(-3) m(-1) for the absorption-dominated case and fairly zero for the backscattering-dominated case. A sensitivity study shows that the impact of noise is the highest on the estimates of the spectral slope exponent of the backscattering coefficient. The performances of the IWOC method are also examined through hyperspectral airborne images acquired over relevant study areas. The reflectance R-rs is well retrieved with RMSE values ranging from 7.5 x 10(-5) sr(-1) to 5.82 x 10(-4) sr(-1). The a priori knowledge of the properties of the study areas is consistent with the spatial distribution of the effluents within the ponds as derived from the remote sensing observations. The approach conducted in this article is a first step toward a generic inversion method for the optical characterization of pollution sources in water, which could further lead to an operational method.