Modelos de estimativa da concentração de sedimentos suspensos via sensoriamento remoto para o rio Teles Pires

Abstract

Water erosion is a natural phenomenon intrinsic to the formation of soils, but it is potentiated by inappropriate anthropic actions of land use in watersheds. One of the consequences of inadequate management or lack of conservation actions in watersheds is the increase in the concentration of suspended sediments (Css) in rivers and lakes, causing changes in the morphology of channels and silting up of reservoirs. As demands on water resources increase, conflicts over water use may occur, implying the development and implementation of water monitoring methodologies, which are decisive for proper management of water resources. The determination of Css in watercourses can provide a diagnosis of the state of land use and occupation of a hydrographic basin, however the methods for obtaining this data are costly and the punctual character of the samples makes it difficult to extrapolate along the rivers and reservoirs. Remote sensing presents itself as a viable alternative to remedy these obstacles, since the variation in suspended sediment concentrations is noticeable in satellite images. In this way it is possible to use the reflectance of the satellite images to estimate the Css spatially and temporally. To accomplish this task, the Sentinel 2a and b platform products were employed, their spatial resolution of 10 meters combined with the temporal resolution of 5 days can provide relevant information on the dynamics of the Css in the basin. The images used in the study corresponded to the same day as the field collections, providing the creation of empirical models. The processing environment used was Google Earth Engine, a tool that allowed agility and flexibility in the application of the methodology. Access to several data sources and processing robustness show that the tool can be an ally in the study of water quality parameters via remote sensing. Among the studied bands and indices, the best Css estimator was the reflectance of the B4 band, which corresponds to the red range of the visible spectrum, and the exponential model presented the best fit and accuracy compared to the linear model.