Surface Water Interaction Modelling Using Visual MODFLOW and GIS

PHYLARESModel calibration / ch.7__________________________________________________________________________________________46Chapter 7Model calibrationModel calibration determines how well the model results simulate field measurements, such aswater level in wells and river gains and losses. In this iterative process, model value of inputparameters such as rain recharge and hydraulic conductivity are adjusted by trial-and-error tobest and reasonably match measured data.Minimum and maximum values possible for the conceptual model should not be exceeded incalibrating the ground-water model. Calibration is evaluated through the analysis of residualsdesigned to bring the mean of the residuals close to zero, and to minimize the standard deviationof the residuals. The calibration procedure may lead to different solutions for different parametercombination.The degree of fit between model simulations and field measurements can be quantified bystatistical means. Statistical analysis for calibration includes the Mean Error (ME), Meanabsolute Error (MAE); Root Mean squared Error (RMSE), and Normalized Root Mean SquaredResidual (Normalized RMS). Among these, the RMSE is generally the best calibration indicator(Anderson and Woesner, 1992). The error criteria are defined as the error between measuredhead and the simulated head.Mean Error is the mean difference between the observed head ( h0) and the simulated head (s)h .1n∑ME = ( h − )0n i=1h s(7.1)MAE is the mean of the absolute value of the differences in observed and simulated head.MAE =n1∑|h0 − h s|n 1i=(7.2)______________________________________________________________________________________Groundwater – Surface water interaction modeling using visual MODFLOW and GIS