Linear Spectral Mixing Model for estimating optically active components in estuarine waters of Patos Lagoon, Brazil

Abstract

ABSTRACTUnderstanding about type and concentration of components in large continental waterbodies is of great value to environmental studies. The synoptic view of multispectral remote-sensing images has the potential to systematically estimate important parameters like suspended solids (SS) and chlorophyll (Chl). However, measures derived directly from radiometric image data usually retrieve inaccurate estimations, preventing discrimination of the spectra and amount of specific components. This work proposes the application of Linear Spectral Mixing Model to estimate concentrations of SS and Chl in the Patos Lagoon estuary, Brazil, through Landsat-8 Operational Land Imager sensor data. The linear mixing model was applied to produce fraction images of SS, sandy bottom, and dissolved organic matter. Simple and multiple linear regressions were performed in order to produce empirical models confronting parameters measured in situ with traditional radiometric data from spectral bands and with the resulting fraction images. The proposed empirical models estimates reached concentration of SS (coefficient of determination, R2 = 0.84) and Chl (R2 = 0.77) in comparison to the data collected in situ. These results showed great potential for using Spectral Mixing Models to indirectly estimate water components by remote-sensing data.