Simulation of daily and monthly stream discharge from small watersheds using the SWAT model

Abstract

The Soil and Water Assessment Tool (SWAT) was evaluated and parameter sensitivities were determined while modeling daily streamflows in a small central Kentucky watershed over a two-year period. Streamflow data from 1996 were used to calibrate the model and streamflow data from 1995 were used for evaluation. The model adequately predicted the trends in daily streamflow during this period although Nash-Sutcliffe R2 values were -0.04 and 0.19 for 1995 and 1996, respectively. The model poorly predicted the timing of some peak flow values and recession rates during the last half of 1995. Excluding daily peak flow values from August to December improved the daily R2 to 0.15, which was similar to the 1996 daily R2 value. The Nash-Sutcliffe R2 for monthly total flows were 0.58 for 1995 and 0.89 for 1996 which were similar to values found in the literature. Since very little information was available on the sensitivity of the SWAT model to various inputs, a sensitivity analysis/calibration procedure was designed to evaluate parameters that were thought to influence stream discharge predictions. These parameters included, drainage area, slope length, channel length, saturated hydraulic conductivity, and available water capacity. Minimization of the average absolute deviation between observed and simulated streamflows identified optimum values/ranges for each parameter. Saturated hydraulic conductivity, alpha baseflow factor, drainage area, channel length, and channel width were the most sensitive parameters in modeling the karst influenced watershed. The sensitivity analysis process confirmed die trace studies in the karst watershed that a much larger area contributes to streamflow than can be described by the topographic boundaries. Overall, the results indicate that the SWAT model can be an effective tool for describing monthly runoff from small watersheds in central Kentucky that have developed on karst hydrology however calibration data are necessary to account for solution channels draining into or out of the topographic watershed.