Modelling Hydrological Response To Dynamic Land Use Land Cover Changes In Big Akaki Watershed, Ethiopia

Abstract

This study investigates the hydrological responses to dynamic land use land cover (LULC) changes in the big Akaki watershed, upper Awash River basin, Ethiopia. The population growth and socio-economic flux have forced people to clear forests for agriculture and settlement which causes the LULC changes. Thus, these LULC changes have a negative impact on water balance components. Quantifying and assessing the state of hydrological responses to dynamic LULC changes is very important for sustainable water resources development and management. The main objective of this study was to model hydrological response to dynamic LULC changes at basin and sub-basin levels in the study area of the big Akaki watershed. The Soil and Water Assessment Tool (SWAT) model was used to quantify the effects of LULC changes on hydrological responses and sediment yield. The LULC trends and dynamics were assessed using Landsat imagery of 2000, 2010, and 2020 in ERDAS Imagine 2015 with a supervised classification technique of maximum likelihood algorithm. The overall accuracy and kappa coefficient values were 90.50%, 90.00%, and 89.10% and 0.83, 0.82, and 0.81, respectively for the periods of 2000, 2010, and 2020. The classified LULC of the three land-use periods showed that absolute change of built-up, forest land, shrub land, and bare land areas were increased by 3.47%, 0.28%, 0.47%, and 0.06% whereas the agricultural land, grass land, and water body were decreased by 2.65%, 1.43%, and 0.19%, respectively. The SWAT model runs were performed using static and dynamic LULC inputs to assess the impacts of LULC change on hydrological responses and sediment yield between 2000 and 2020. The static model setup used only the 2000 land use map, whereas the dynamic model setup used 2000, 2010, and 2020 land use data. Sensitivities of input parameters including auto-calibration and validation have been performed using the SUFI-2 algorithm in SWAT Calibration Uncertainties Program (SWAT CUP). The two model performance was found to be satisfactory to predict the spatio-temporal water balance components with R2 of 0.8 (0.82), NSE of 0.73 (0.61), and PBias of 21.4 (37.7), respectively for streamflow and R2 of 0.84 (0.68), NSE of 0.81 (0.54) and PBias of 17.1 (39.1), respectively for sediment yield. Due to the expansion of built-up and reduction of agricultural land and grassland, the basin level average annual surface runoff (SurfQ), water Yield (WYLD), and sediment yield (SYLD) were increased by 2.51mm, 0.67mm, and 10.83t/ha, respectively whereas lateral flow (LatQ), groundwater flow (GWQ), groundwater recharge (GWR) and evapotranspiration (ET) were decreased by 1.18mm, 0.63mm, 0.56mm and 2.9mm, respectively. However, at the sub-basin level, the rapid expansion of built-up resulted in an increase of SurfQ and WYLD by 10.15mm and 8.78mm in sub-basin 13, respectively. Whereas the GWQ, ET, and SYLD were decreased by 2.58mm, 32.34mm, and 4.74t/ha in sub-basins 5, 9 and 6, respectively. The outcome of this study work revealed that incorporating the dynamic LULC approach into the SWAT model simulation gives a better representation of spatio-temporal variability of LULC changes. Thereby improving the accuracy of spatio-temporal estimation of water balance components by integrating better LULC evolution than the static LULC model approach.