Minimizing Errors in the Prediction of Water Levels Using Kriging Technique in Residuals of the Groundwater Model

Groundwater monitoring and water level predictions have been a challenging issue due to the complexity of groundwater movement. Simplified numerical simulation models used represent system; these however only provide conservative approximation system may not always capture local variations. Several other efforts such as coupling with hydrological using geostatistical methods are being practiced accurately predict levels. In this study, we present novel application tool on residuals model. The kriging method was applied model (MODFLOW) generated by TWDB (Texas Water Development Board) for Edwards–Trinity (Plateau) aquifer. study done years 1995 through 2000 where 90% observation data followed cross-validation remaining 10% observations. reduced average absolute error approximately 31 m (for MODFLOW simulation) less than 5 m. Furthermore, residuals’ standard from 9.7 4.7. This implies that mean value over entire period can be good estimation each year separately. use technique thus improved levels resulting in more accurate potentiometric surface maps.