Development of a Fuzzy Inference System to Predict Runoff Using Only Rainfall Data for Row Crop Watershed in the Claypan Region

Estimation of surface runoff from a watershed is a pre-requisite to determine pollutant loss, develop regulatory measures such as Total Maximum Daily Loads (TMDL), and to forecast the water supply during spring and early summer (Mahabir et al., 2003). Regression models used for forecasting runoff for water supply often do not perform well especially in forecasting low and average runoff events, and when data are limited (Mahabir et al., 2003). These researchers emphasize that regression equations are site specific and must be developed for each site. Sen (2009) states that mathematical modeling which forms the basis for many hydrological modeling systems is based on a set of restrictive assumptions and often overlooks the fuzziness or variability in the problem. Sen (2009) also states that hydrologic events are often too complex to be described precisely using mathematical formulae. In hydrology, the two most valuable sources of information are: hydrologist’s expert views and measured data. In classical hydrologic modeling, there is no role for valuable experience of the expert. In most instances of hydrologic studies, numerical data may be limited, but the hydrologist’s observations provide a set of linguistic information that can lead to logical and rational thinking and formulation of a preliminary set of rules (Sen, 2009).