Empirical modelling of suspended solids in a subtropical lagoon (Brazil) using linear spectral mixing algorithm

Abstract The Linear Spectral Mixing Model (LSMM) technique has been used in remote sensing applications and allows the estimation of the proportional contribution of different constituents in the pixel reflectance. Despite the great potential for vegetation and land cover studies, there are few studies examining the LSMM efficiency for optically complex inland waters. In addition, LSMM-derived images have shown a promising application in water quality mapping. In this context, Mirim lagoon (Brazil) presents seasonal variability of suspended solids (SS) influenced by sediment yield from tributaries and surrounding areas (e.g.: croplands). While negative impact on water quality makes relevant the SS mapping for management of this ecosystem, there is no studies using remote sensors in this lagoon. This study develops an empirical model for SS concentration in the Northern Mirim lagoon using Sentinel-2 MSI images. The framework applies the linear spectral mixing algorithm in two MSI images. The empirical model was developed using the multiple linear regression and we validated the model performance using Monte Carlo simulation method. The mapping of SS using the empirical model was satisfactory in both images: the estimated concentrations are consistent with in-situ measurements. The results show that LSMM application supports the mapping of SS and its spatial distribution. Finally, this study is the first application of Sentinel-2 MSI images in SS mapping in the Mirim lagoon and the results present the potential of this dataset and empirical modelling for sediment mapping. The spatial information supports the local managers to understand the sources of degradation of water quality and protect this aquatic ecosystem.