Multi Model Approach to Study the Effect of Conceptualization and Discretization on Groundwater Model

Abstract

The objective of the study is to develop multiple steady state conceptual and numerical groundwater flow models for Bhiwadi region of Alwar district of Rajasthan, India. Automated parameter estimation tool (PEST) is used for the calibration of the models. To evaluate and select the best suited conceptual flow model, several statistical criteria such as RMSE (Root Mean Square Error), R-squared (R^2), Nash-Sutcliffe efficiency (NSE), percent bias (PBIAS), multi model information criteria: Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) are used. The result of the study indicates that value R2, NSE and Pbias are almost similar for all the models, hence are insufficient to evaluate the models. Therefore, AIC and BIC statistical criteria are used for accurate evaluation. The result also indicates that grid size and orientation have immense impact on modeling output. For the study area coarse grid size has given better results as compared to finer grid size.