Perda de solos em estradas rurais de leito natural na Bacia do Córrego Moeda (MS): Modelagem EUPS e a influência da Conectividade Hidrológica Retilínea Induzida

Abstract

Soil loss dynamics on unpaved rural roads (conceptualized here as natural surface roads) represent a key challenge for environmental management in watersheds, especially in regions of agro-industrial expansion. This thesis aimed to assess the relative influence of rainfall erosivity, slope, and management practices on soil loss dynamics on unpaved rural roads and their margins in the Córrego Moeda Watershed, in Três Lagoas, Mato Grosso do Sul, Brazil. The central hypothesis was that the erosive contribution of these roads is influenced by the synergistic interaction of these three factors. The methodology was based on integrating multi-scalar approaches: characterization of the rainfall regime by classifying the historical series (1983-2017) into quantile scenarios ("Very Dry", "Dry", "Normal", "Rainy", and "Very Rainy"); in situ collection of variables related to the factors of the Universal Soil Loss Equation (USLE) in 23 previously established Sample Units; and high-resolution recording of rainfall events (2023) for calculating erosivity (EI30). The results demonstrated that the roads constitute the primary source of sediments in the watershed, with soil losses up to 600 times greater than those in vegetated margins. The slope proved to be a decisive amplifier of the process, while seasonality and the occurrence of extreme events defined critical risk windows. The integrated analysis allowed for the proposition of the conceptual framework of Induced Rectilinear Hydrological Connectivity to explain erosive dynamics, in which roads, as rectilinear anthropogenic elements, connect surface hydraulic flow directly to the drainage network. It is concluded that the hypothesis was confirmed and that the proposed management guidelines, based on the selective disconnection of these flows, offer a viable path for erosion mitigation, providing an analytical model applicable to watersheds with analogous contexts.