Numerical modelling of the saline interface in coastal karstic aquifers within a conceptual model uncertainty framework

Numerical modelling is increasingly used as a tool for improving management strategies in aquifers and to support the design of comprehensive projects considering natural and anthropogenic processes. Overall, numerical simulation in karstic aquifers poses a major scientific challenge due to the non-Darcian groundwater flow dynamics. In specific cases, the equivalent porous medium approach has shown acceptable results, particularly in poorly karstified aquifers with regional/subregional scales such as this case. The Yucatan coastal karstic aquifer (Mexico) has been defined as a complex regional heterogeneous system, partially confined, thus allowing the discussion of multiple conceptual models. In this research, a two-dimensional numerical model of flow and transport was implemented using SEAWAT for the NW Yucatan aquifer. Four likely conceptual models were audited, calibrated and verified using hydrogeological field data, to select the best one, considering their fit and complexity. The numerical model accuracy was evaluated using the root-mean-square error, Nash Sutcliffe efficiency and the Pearson coefficient. The Akaike information criterion and Bayesian information criterion were included for evaluating the complexity of the numerical models. In addition, the signal of tide propagation into the aquifer was assessed as a proxy to improve the numerical calibration process. Results show that the most complex numerical model has a better calibration than the simpler models, but the model accuracy is worse when compared to less complex numerical models in the verification exercise. This research offers enhancement in the knowledge of numerical modelling in heterogeneous coastal aquifers within a conceptual-model uncertainty setting.