Suscetibilidade erosiva em bacia hidrográfica do litoral pernambucano: mapeamento e modelagem por meio de técnicas computacionais

Abstract

Water erosion is one of the main processes responsible for environmental degradation in tropical watersheds, affecting landform stability, soil fertility, and water quality. In view of this issue, this research aimed to map and analyze the erosive susceptibility of the União River watershed, located on the southern coast of Pernambuco, Brazil, an area characterized by high humidity, undulating relief, and increasing anthropogenic pressure. The study was developed through the integration of geotechnologies, remote sensing, and machine learning algorithms, with the purpose of improving the spatial prediction of areas vulnerable to linear erosion. Topographic variables (altimetry, slope, curvature, and aspect), hydrological variables (topographic wetness index and drainage density), geological variables (lithology and soils), and anthropogenic variables (land use and land cover) were used. The methodological steps included data processing in a GIS environment, the construction of an inventory of erosive features, and the modeling of susceptibility using Logistic Regression, Random Forest, Artificial Neural Networks, and XGBoost algorithms, statistically validated through metrics such as AUC and ROC curve. Additionally, small-format aerial photogrammetry with a Remotely Piloted Aircraft (RPA) was applied to generate high-resolution digital elevation models, enabling detailed analysis of critical areas and in situ verification of erosive features. The results showed that most of the basin presents low susceptibility, with high and very high susceptibility classes concentrated on steep slopes and estuarine margins. All algorithms showed satisfactory performance (AUC > 0.90), demonstrating high predictive accuracy and confirming the effectiveness of integrating predictive modeling and high-resolution spatial data. It is concluded that this approach constitutes a robust tool for environmental planning, conservation management, and sustainable governance of coastal watersheds