Pilot point versus zonation approach for groundwater model calibration of karst aquifers

The groundwater inverse problem is a big challenge, especially in highly parameterised aquifers, because it is ill-posed and the solution is non-unique. The problem is even more complicated for karst aquifers, where variability of hydraulic properties is high. This study uses two approaches to calibrate a highly variable groundwater flow model for the north aquifer of Qatar. The aquifer comprises karstified limestone from Eocene and Miocene ages, and it is highly heterogeneous as many conduits, fractures and sinkholes occur in the formation. The two approaches followed are the zonation and the pilot point approach. In the zonation approach, the model domain is divided into several zones, and each is assigned a one value of hydraulic conductivity. The number of the calibrated variables in this case is limited to the number of zones. The second approach is based on pilot points, where calibration is done at a finite number of points in the model domain, and then the calibrated values are interpolated over the entire model domain using geostatistics (i.e. kriging). The parameter estimation and uncertainty analysis model (PEST) was used with a finite difference MODFLOW model. The model area is 4340 km2, and it was discretized into 222 rows and 148 columns, with a cell size of 500 by 500 meters. Results show both approaches succeeded to calibrate the model within a pre-defined error threshold value. The calibrated values of hydraulic conductivity in both cases were found to vary between 0.01 to more than 200 m/day, which reflects the high heterogeneity of the aquifer. While the zonation approach produced abrupt changes of the calibrated parameter at the zone boundaries, the pilot points approach produced a smoother map of the parameter.