ANALYSIS OF UNCERTAINTY IN ANALYTIC ELEMENT GROUNDWATER MODELS BY THE MONTE CARLO METHOD

Among the biggest challenges in groundwater modeling are assessing and communicating the uncertainty in model predictions. Although the calibration process helps us choose appropriate parameter values, uncertainty in these values leads to some uncertainty in the model results. To develop a statistically rigorous method of evaluating the impact of uncertainty on model predictions, we applied the Monte Carlo method to an analytic element groundwater model. The Monte Carlo method entails running many simulations with different combinations of parameter values and summarizing the results with statistics and graphics. This capability has been developed recently for the finite-difference model MODFLOW, but, to our knowledge, this technique has not previously been applied to simpler analytic element models, such as GFLOW. We applied this technique to GFLOW models we developed as part of a Source Water Area Protection project to delineate zones of contribution for municipal wells in southern Wisconsin. These wells pump from two aquifers: Pleistocene sand and gravel and Cambrian sandstone. Zones of contribution are considerably larger in the sand and gravel, because it has a higher hydraulic conductivity and is much thinner than the sandstone. The Monte Carlo analysis indicated that the effect of parameter uncertainty on the size and shape of zones of contribution is also much greater in the sand and gravel aquifer, probably due to the difference in hydraulic properties between the aquifers. The Monte Carlo technique provides insight into the behavior of groundwater systems, and it allows modelers to represent uncertainty in their results in a way that is useful to those who must make decisions based upon them.