MODELING THE SPATIOTEMPORAL IMPACTS OF LAND USE LAND COVER ON RIVER WATER QUALITY USING A SPATIAL REGRESSION MODEL: A CASE OF AWATA WATERSHED, GENALE DAWA BASIN, ETHIOPIA

Abstract

Land-use and land-cover changes can have a significant impact on the quality of river water, affecting hydrological processes and nutrient cycles and that can impact environmental health and human well-being. The study aimed to model the spatiotemporal impacts of land use land cover on river water quality parameters using the Ordinary Least Squares (OLS) and Geographically Weighted Regression (GWR) model in the Awata watershed of Genale dawa basin Ethiopia. GWR extended traditional regression models, such as OLS to address the spatial variations among variables. Landsat images from 2000 and 2022 were obtained to identify the land use land cover types that most affecting water quality. Physicochemical water quality data were collected from eight sampling points in 2022 twice during both seasons and analyzed using APHA to identify river water quality variation. Model calibration and validation based on R-Squre(R²) and Akaike Information Corrected Criterion (AICc) and Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) respectively. The land use land cover change result show that, reductions of agricultural land by (37.03%) while water bodies increased by (5.22%) and barren land by (36.31%). Most water quality parameters were higher during the dry season, except PO4-P (0.48 mg/L) and Mg++ (14.8mg/L) which were higher during the wet season. The Ordinary least square model determined that agricultural Ph (-0.02), DO (-0.12), NO3-N (-0.05) and NH4-N(-0.44) forested areas are the primary land-use types affecting river water quality during both seasons. The spatially varying relationship results revealed that agricultural land had a substantial positive impact on NH4-N (0.013-0.036), and PO4-P (0.0-0.003), but a negative effect on NO3-N (-0.014- -0.005) in wet season. Model validation of pH and EC R² values OLS to GWR account 0.91 to 0.99 and 0.34 to 0.99 and 0.66 to 0.67, and 0.99 to 0.09 explained by each LULC, with higher and lower AICc value in the wet and dry season respectively. Therefore, GWR better performed than OLS model. To overall, this study offers insights into how land use agriculture and forest affect river water quality and validates the effectiveness of spatial regression models in understanding these relationships. Policymakers and resource managers can use these findings to develop effective strategies for protecting and improving river water quality.