Application of geographically weighted regression models to predict spatial characteristics of nitrate contamination: Implications for an effective groundwater management strategy

Abstract

Groundwater nitrate contamination has been the main water quality problem threatening the sustainable utilization of water resources in Jeju Island, South Korea. The spatially varying distribution of nitrate levels associated with complex environmental and anthropogenic factors has been a major challenge restricting improved groundwater management. In this study, we applied ordinary least squares (OLS) regression and geographically weighted regression (GWR) models to determine the relationships between the NO3–N concentration and various parameters (topography, hydrology and land use) across the island. A comparison between the OLS regression and GWR prediction models showed that the GWR models outperformed the OLS regression models, with a higher R2 and a lower corrected Akaike Information Criterion (AICc) value than the OLS regression models. Interestingly, the GWR model was able to provide undiscovered information that was not revealed in the OLS regression models. For example, the GWR model found that orchards (OR) and urban (UR) variables significantly contributed to nitrate enrichment in the certain parts of the island, whereas these variables were ignored as a statistically insignificant factor in the OLS regression model. Our study highlighted that GWR models are a useful tool for investigating spatially varying relationships between groundwater quality and environmental factors; therefore, it can be applied to establish advanced groundwater management plans by reflecting the spatial heterogeneity associated with environmental and anthropogenic conditions.