Mapping the coloured dissolved organic matter absorption coefficient in a eutrophic reservoir using remotely sensed images

Abstract

Dissolved organic matter (DOM) has 3 major sources: (1) autochthonous or surface water-derived of algal origin, (2) allochthonous or terrestrial material from soils, and (3) synthetic organic substances of man-made or industrial origin. Thus, its correlation with dissolved organic carbon (DOC) or chlorophyll a (Chl-a) depends on environmental dynamics and their related changes. Because coloured dissolved organic matter (CDOM) is the coloured fraction of DOM, historical measurements of DOM may be used as a suitable proxy for CDOM. In this research we (1) evaluated the performance of a set of bio-optical models, including empirical models and a quasi-analytical algorithm (QAA), by using simulated reflectance data based on the Thematic Mapper sensor parameters (TM); and (2) analyzed the possible sources of CDOM/DOM in a tropical eutrophic reservoir for a 15-year TM/Landsat-5 time series. Results showed that the literature-based models failed to retrieve the CDOM absorption coefficient (aCDOM) from simulated reflectance data. After correlation and collinearity analyses (2D colour correlation plot), an alternative wavelength (485 nm) and band ratio (B4/B1) gave the best relation to aCDOM. The proposed model had satisfactory performance (R2 = 0.91, p < 0.0001, RMSE = 7%, Nash-Sutcliffe = 0.91). Two significantly different aCDOM patterns were identified for the Funil hydroelectric reservoir (Brazil): one associated with land cover land use and rainfall/runoff occurrence and the other with elevated Chl-a associated with algal blooms.