Analysis of the correlation between land use and surface runoff in a Brazilian savanna basin

Abstract

– Since the 1980s, the Brazilian savanna (Cerrado) has been suppressed to make space for agricultural production, mainly soybean, corn, and cotton. This savanna strategically contains important South American river sources – Paraná, Paraguai, São Francisco, Araguaia, Tocantins, and tributaries of the Amazon River. Land use and cover are conditioning factors in generating the hydrological cycle components, notably surface runoff. Thus, this study aimed to evaluate spatial and temporal land use changes and their correlations with surface runoff in the Grande River basin. The study area is a relevant São Francisco River sub-basin, covering part of the Urucuia Aquifer System, encompassing the dry forest and savanna biomes. The inputs included landscape ecology metrics, land use time series from 1990 to 2020, and water flow values near the basin outlet. The landscape metrics were modeled with the R software landscapemetrics package. Aside from the time series trend, the correlation between land use, land cover, and flow variations was also evaluated through statistical tests – Mann-Kendall, Sen's slope, Pittitt, and rho Spearman. Changes were observed in areas covered by forest, savanna, and grassland formations. The areas presented reduction from 22,121 km2, 27,889 km2, and 15,366 km2 to 9916 km2, 27,870 km2, and 5666 km2, respectively. On the other hand, there was an increase, especially in rainfed agriculture (from 5241 km2 to 20,856 km2) and, more recently (2005), in irrigated agriculture (from 94 km2 to 1362 km2). Results show a reduction in the fragments of native vegetation (from 2,235 km2 to 1794 km2) and an increase in the fragments of agricultural activities (from 678 km2 to 1081 km2). For most classes, the change point was identified as the year 2005. A strong direct correlation was observed between the average and minimum flows (Q90) with forest and grassland formations. In comparison, an inverse correlation was noticed between pasture, urban infrastructure, rainfed agriculture, and irrigated agriculture classes.