Evaluation of the impact of land use/land cover changes on water balance in the C81F quaternary catchment of the Eastern Free State, South Africa

Abstract

South Africa is amongst the countries grappling with water scarcity issues. This means the assessment of this resource and its management is very important. Water balance components need to be assessed in a catchment and this can be done by understanding catchment-level hydrologic processes requiring hydrological models. This study aims to evaluate the impact of land use/land cover (LULC) changes on the water balance in the study area, namely the C81F quaternary catchment of the Eastern Free State, South Africa, under the Thabo Mofutsanyane District Municipality. A SWAT model was used for the simulation of streamflow. The model inputs were digital elevation model (DEM), LULC maps for 1990 and 2018, and soil maps and slope data created using the ArcGIS spatial analysis tool. A warm-up period of three years was used. The analysed LULC changes in 1990 showed that the least dominant classes were water and barren land, the most dominant was pasture and urban, followed by mixed forest which accounted for 71.44%, 16.73%, and 5.50% of the total area, respectively. However, the analysed LULC changes for 2018 showed the most dominant to be pasture, followed by urban, and generic agricultural land, comprising 65.08%, 18.82%, and 7.96%, respectively. Generic agricultural land, urban and pasture are the major LULC classes that impacted the water balance components. SWAT-CUP was used for calibration and validation, and a program called Sequential Uncertainty Fitting Version 2 (SUFI2) was chosen as one of the stochastic calibration programs in SWAT-CUP. The calibration flow data from 1993 to 2007 was used and the validation flow data was from 2008 to 2018. The effectiveness of the SWAT model was assessed using two widely used modelling evaluation techniques, namely quantitative statistics and graphical techniques. The model's effectiveness was assessed using Nash-Sutcliffe criteria (NSE), Coefficient of Determination (R2), and percentage of bias (PBIAS). Calibration resulted in R2 and PBIAS and NSE being 0.79, 0.72, and -3.6, respectively, and these results were deemed very good. Validation results for the same parameters were also found to be good at 0.72, 0.69, and 10, respectively. The model performance for calibration was very good and good at the validation stage, according to monthly ratings of model performance. The decrease of evapotranspiration by 0.7 mm was steady and may be linked to the increase in generic agricultural land. More results showed a decrease in percolation and lateral flow of 0.13 mm and 3.39 mm, respectively. Furthermore, surface runoff increased by 4.24 mm, and the adjusted SWAT parameters reflect and quantify the increased built-up area's effects, which increased surface runoff and decreased percolation in the basin. The increase in stream flow is related to an increase in urban areas, and it is recommended that more studies should be done on this catchment to assess the potential impacts of climate change on groundwater and water balance