Quimiometria e sensoriamento remoto como ferramentas de estimação de alguns parâmetros de qualidade da água em lagos intermitentes do baixo rio Doce

Abstract

In this work, chemometric prediction models based on limnological parameters of total phosphorus, transparency, and chlorophyll a, integrated with a trophic state index, and remote sensing imagery, were developed to assess the water quality of Lower Doce River Valley lakes’ (Southeastern Brazil). Were used as it is an area conditioned to the effects of the environmental disaster of the dam of iron ore tailings from Fundão (Minas Gerais - Brazil). The images refer to the field days of collecting water samples in the ponds. After being selected and organized, they were submitted to atmospheric correction and reflectance data extraction processes. Subsequently, the information from the limnological environmental monitoring was associated with the reflectance data of the images by the support vector regression, a multivariate calibration method, in order to adequately estimate the results of the investigated parameters. The models showed accuracy levels with prediction R 2 between 0.705 and 0.984, allowing for reliable estimates with RMSEP results of 0.457 and 7.305. The application of this technique enables the analysis of limnological parameters remotely, helping environmental monitoring and equipping managers for more efficient decision-making in actions for the conservation of water resources.