Modelagem e análise da qualidade da água em reservatórios utilizando mapas auto-organizáveis.

Abstract

Water quality in reservoirs located in the Brazilian semiarid region is strongly influenced by climatic variability, multiple water uses, and anthropogenic pressures, which makes its analysis challenging due to the complexity and multidimensionality of environmental data. In this context, the present study aimed to apply artificial neural networks of the Self-Organizing Map (SOM) type to analyze water quality patterns in seven reservoirs located in the Paraíba River basin, in the state of Paraíba, Brazil. The data used were obtained from the PELD-RIPA project and include physical, chemical, and biological variables associated with water quality. Initially, exploratory and multivariate statistical techniques were applied, including descriptive statistics, correlation analysis, and Principal Component Analysis (PCA), in order to understand the relationships among variables and to select relevant indicators for modeling. Subsequently, SOM models were developed using two complementary approaches: an integrated model considering the combined data from the seven reservoirs and individual models for each system. The results demonstrated the ability of SOM to identify complex patterns and environmental clusters associated with water quality conditions. The integrated modeling provided a comparative view among the reservoirs, while the individual models allowed the identification of specific limnological characteristics of each system. Overall, the results indicate that Self-Organizing Maps constitute an efficient tool for the exploratory analysis of multivariate environmental data, contributing to monitoring efforts and supporting water resources management in semiarid regions.