Time-series analysis of Landsat-MSS/TM/OLI images over Amazonian waters impacted by gold mining activities

Abstract

Water siltation caused by artisanal gold mining has impacted the Tapajós River Basin in Brazil for the past 40years, however spatial-temporal information about changes in water quality and consequences to the aquatic environment is lacking. To address this, the Landsat satellite family sensors were used to retrieve total suspended solids (TSS) of the water of the Tapajós River from 1973 to 2013. An image processing approach that includes atmospheric correction, based on the 6S model, and glint removing, based on shortwave infrared correction, was applied and validated with in situ radiometric data. An optimization of the atmospheric correction having dark dense forest spectra as reference was applied and allowed a robust correction of MSS, TM and OLI signal to surface reflectance values. Sediment concentration was estimated based on a non-linear empirical regression between measured TSS and satellite surface reflectance at red band. The multi-temporal analysis of TSS showed that the sediment load in the Tapajós aquatic system is in synchrony with mining activities, and a constant seasonal variation of water siltation is observed throughout the time frame of this study. At the end of the rainy season, mining activities intensify and, coupled with low water flow, TSS increases. During the high water level, TSS concentrations were consistently lower because of high water dilution and low mining activity. In a decadal analysis, a peak of sediment concentration coincides with a peak of gold production in all sites analyzed during early 1990s. More recently, due to the currently gold rush, an increase in suspended solids has been observed mainly in the Novo and Tocantins rivers where industrial mining has been installed.