Estimation of geostatistical interpolation methods on the spatial representation of the piezometric level of the El Hadaiek plain in northeastern Algeria

ABSTRACT This study compares three interpolation methods, Kriging universal (KU), inverse distance weighting (IDW), and radial basis function (RBF), to model the piezometric surface of the El Hadaiek aquifer. The evaluation showed moderate to strong correlations between observed and interpolated values: R2 = 0.63 for Kriging, R2 = 0.68 for IDW, and R2 = 0.71 for RBF. RBF yielded the lowest RMSE (6.20 m) and minimal residual variance (R2 = 0.012), demonstrating its superior accuracy in capturing localized groundwater fluctuations. IDW was computationally efficient with a slightly higher RMSE of 7.13 m, but less accurate in heterogeneous environments. Kriging, while effective for spatial autocorrelation, had the highest RMSE of 7.21 m and struggled due to sparse data and an inappropriate variogram. The results suggest that RBF is the most suitable method for detailed groundwater modeling in regions with high spatial variability. IDW remains a viable option for larger-scale studies, while Kriging should be used with a denser data network. These findings emphasize the need for seasonal validation and the integration of interpolation models into GIS-based decision-support systems for improved water resource management.