Interpolation methods for groundwater quality assessment in tank cascade landscape: A study of ulagalla cascade, Sri Lanka

Abstract

Interpolation methods are extensively used to map the spatial distribution of water quality parameters. However, the selection of the most appropriate method is a critical issue in environmental studies. The relative performances of deterministic and geostatistical methods in explaining the spatiotemporal variation of water quality parameters/indices in a tank cascade landscape were assessed. Inverse distance weighted (IDW), global polynomial interpolation (GPI), local polynomial interpolation (LPI), radial basis function (RBF), kriging (KR), and empirical Bayesian kriging (EBK) methods were evaluated using root mean square error (RMSE) in a leave-one-out cross-validation. Coefficient of variance, normality, level of autocorrelation, and extreme values near boundaries of the dataset showed a clear relationship with the relative performances of the different interpolation methods. Therefore, a clear understanding of the quality of the dataset is required in order to select the appropriate method to interpolate water quality parameters. EBK performed well for most parameters throughout the study period and is recommended as the best method to interpolate water quality parameters/indices in the Ulagalla cascade and other tank cascade landscapes in Sri Lanka and similar environments.