Fuzzy Rule-based Methodology for Estimating Monthly Groundwater Recharge in a Temperate Watershed

Abstract

A fuzzy rule-based methodology is developed for estimating monthly groundwater recharge using the Toms River basin in New Jersey as a case study. As an alternative to a water budget approach, which depends upon difficult-to-quantify parameters, the fuzzy methodology simplifies model input by using air temperature as a surrogate for evapotranspiration and streamflow as a surrogate for both soil-moisture deficit and direct runoff. The accuracy of the fuzzy rule-based method is compared with that of ordinary linear regression. It is found that for the most accurate fuzzy rule-based model, the mean percent difference between its estimate and the “actual” recharge is 14.0%, slightly larger than the 12.5 mean % difference achieved with linear regression. However, as noise was synthetically introduced into the input data, the difference in predictive accuracy between fuzzy rule-based modeling and linear regression decreased further. In addition, unlike linear regression, neural networks, or physically based models, the transparent nature of the fuzzy rules provides explicit qualitative and quantitative insights into historic system behavior and may be used to forecast recharge under variable hydrologic and climatic conditions.