Modeling aquifer hydrograph: performance review of conceptual MODFLOW and simulator models

Abstract

A decline in water levels in plains has been caused by reductions in precipitation, restriction of water resources, and ever-increasing demands for groundwater exploitation. This study utilized monthly data on the piezometric level for 5 years (2009–2010 to 2013–2014) derived from 8 piezometers in the Loor-Andimeshk plain aquifer. Using the Thiessen method, the weighted average of each piezometer level was first obtained, and then, the time series of the groundwater level of the plain was calculated to illustrate the aquifer hydrograph of the studied area. Then, the calculations ran on MODFLOW, groundwater conceptual, gene expression programming (GEP), and Adaptive Neural Fuzzy Inference System (ANFIS) models so as to model the aquifer hydrograph before comparing the results. Where the Root mean squared error (RMSE) = 0.449 m and correlation coefficient (R) = 0.883, it was shown that the MODFLOW model performed better than the GEP and the ANFIS models where RMSE = 0.122 m, R = 0.856, RMSE = 0.148 m, and R = 0.797, respectively. Also, the execution speed of the GEP model was faster than that of the ANFIS model and both models performed better than the conceptual model in terms of time efficiency. The results also showed that the MODFLOW model and the GEP model performed almost identically in modeling the aquifer hydrograph. Thus, in the case of data deficiency, this model simulator could be an accurate alternative to conceptual MODFLOW modelling.