Using geospatial analysis to detect soil loss in pasture in the Brazilian savanna (Cerrado)

Abstract

Well-managed pastures support watershed recharge by conserving soil, increasing organic matter, and enhancing water and nutrient retention, while also reducing erosion, nutrient leaching and surface run-off. This study used remote sensing to assess vegetation cover (Vc) in degraded pasture areas and examine its relationship with soil loss in the Brazilian savanna, employing vegetation indices and the Universal Soil Loss Equation (USLE). Given its socioeconomic importance, the Ribeirão Serra Negra Watershed in Piracanjuba, Goiás, Brazil was the focus of the research. Soil loss was estimated from freely available geographic data on climate, while soil class and relevance were assessed through a digital elevation model (DEM). The use of images from the Sentinel-2A orbital sensor alongside the Normalized Difference Vegetation Index (NDVI) also made it possible to quantify and classify Vc according to its condition. The findings reveal that pastures in this region exhibit varying levels of degradation, with intermediate degradation classes being most prevalent: ‘moderately degraded’ areas covered 3,919.25 hectares (ha) (56.87% of the total area), while ‘severely degraded’ areas accounted for 1,984.30 ha (28.79%). The most common index values ranged from 0.6 to 0.7, aligning with the higher prevalence of the moderately degraded class. Soil losses were considerable in these intermediate classes, affecting 5,683.19 ha (82.46%); the most frequent loss rate was estimated to lie between 20 and 40 Mg ha-1 year-1, covering 4271.34 ha (61.98%). Higher Vc indices were more frequently observed in areas with lower soil losses, demonstrating a significant inverse correlation (-0.80, p-value ≤ 0.05) between these variables, with a determination coefficient (R²) of 0.64.