Comparing Deterministic and Stochastic Methods in Geospatial Analysis of Groundwater Fluoride Concentration

Dental and skeletal fluorosis caused by consuming high-fluoride groundwater has been reported over several decades globally. Prediction maps to estimate the fluoride contaminated area rely on interpolation methods. This study presents a comparison of accuracy nine spatial methods in predicting groundwater. Leave-one-out cross-validation (LOOCV), hold-out validation with an independent dataset were used assess precision is first large (N = 13,585) applied at regional level India. Our findings showed that inverse distance weighted (IDW) algorithm outperformed other terms less discrepancy between measured predicted fluoride. IDW local polynomial (LPI) only predict areas (fluoride > 1.5 mg/L). However, estimated typical assessment percentage unsuitable samples was much higher (6.1%) compared (0.2%) LPI (0.2%). LOOCV provided viable results than two Interpolation are accompanied uncertainty which regulated sample size, density, distribution, minimum maximum concentrations, smoothing border effects. Drawing among variegated capturing wide range prediction suggested rather relying one method exclusively. The identified this can be Government planning remediation actions.