Dynamic panel-data-based groundwater level prediction and decomposition in an arid hardrock–alluvium aquifer

Abstract

The comprehensive recognition of groundwater level, especially in arid and semiarid areas, is important in many hydrogeological and hydrological studies. This research aims at predicting and decomposing groundwater level using dynamic panel-data (D-PD) method in an arid hardrock–alluvium Al-Buraimi region, Oman–UAE border. The D-PD method incorporates temporal and spatial variations to predict groundwater level and detects effects that are not simply computable in pure cross-sectional or pure time-series methods. Monthly rainfall, potential evapotranspiration and artificial abstraction, as independent variables, were calculated for 39 Thiessen polygons. The D-PD model was developed to predict groundwater level for the calibration (October 1996 to September 2008) and validation (October 2008 to September 2012) periods. The calibrated D-PD model was then used to decompose the groundwater level hydrograph and to quantify the influence of climate, artificial abstraction, “unobservable spatial-specific effect” and “unobservable time-specific effect” components. The results show significant effect of rainfall on the second time lag, while artificial abstraction and potential evapotranspiration as surrogate of natural groundwater abstraction have influence on the fourth and eighth time lags on the groundwater level prediction. It is found that “unobservable spatial-specific effects” and “unobservable time-specific effects” components of the D-PD model highly influence groundwater level prediction in the study area. The proper match obtained from D-PD method between computed and observed groundwater levels in the study area indicates that the method is efficient and robust and proposes its application to predict and decompose groundwater level variations in regions with complex hydrogeology.