Optimization of pilot points distribution in inverse modelling: A case study from Qatar Aquifer

Two approaches are commonly used for model calibration, namely the zonation and the pilot-points method. While the zonation approach assumes an abrupt spatial change in parameter values, the pilot-points method produces smoothly distributed parameters, however, the number and placement of the pilot-points can be challenging. The objective of this study is to explore the effect of pilot-points number and locations on the calibrated parameters. The northern aquifer of Qatar is used as a case study, using a 3D groundwater flow model. A numerical model of this aquifer was developed using MODFLOW and the model was calibrated using PEST package using historical data of groundwater levels. The model was run several times using a variable number and distribution of pilot points. The Root Mean Square Error for all the runs (corresponding to different configurations of pilot-points) was maintained under a certain threshold. A statistical analysis of the calibrated parameters was then performed to evaluate how far these parameters are impacted by the pilot-point locations. Finally, a new methodology for Optimisation of pilot points locations was proposed using recharge and piezometric maps. The results showed that the pilot-points number, locations, and configurations have a substantial impact on the calibrated parameter, especially in the high permeable regions. The outcome of this study may help focus on areas of high uncertainty where more field data should be collected to improve model calibration. It also helps placement of pilot points for a robust calibration.