Hydrologic response to climatic variability in a Great Lakes Watershed: A case study with the SWAT model

Process-based hydrologic models are usually calibrated prior to application to ensure that they closely match reality. However, different hydrologic response to varied climatic conditions might affect model calibration and validation. A case study was conducted for a 901 km(2) watershed of northern Michigan to compare the effects of calibrating the Soil and Water Assessment Tool (SWAT) watershed model with different climatic data-sets representing drought (1948-1949) versus average (1969-1970) conditions. The effects of the different climatic conditions on parameter response and sensitivity were evaluated, and performance of the two calibration versions was compared using a common validation period, 1950-1965. For the drought- and average-calibration periods, models were well calibrated, as indicated by high Nash-Sutcliffe efficiency coefficients (E = 0.8 and 0.9), and Low deviation of discharge values (D = 2.9% and 3.4%). Evapotranspiration parameters differed under the two sets of climatic conditions. The plant water uptake compensation factor (EPCO), appropriately reflected plant water uptake patterns in varied climatic conditions. Snow melting parameters differed between the two scenarios. A comparison of baseflow values simulated by SWAT versus those computed by a hydrograph separation method showed that the SWAT method treated most snowmelt as surface runoff, whereas the tatter method treated much of it as baseflow. The drought-calibrated version of the model performed much better during the validation period (1950-1965) (E = 0.8, D = 2.6%) than did the average-calibrated version (E = 0.4, D = 41.4%). (c) 2007 Elsevier B.V. All rights reserved.