Seasonal and annual impacts of climate change on watershed response using ensemble of global climate models

Abstract

Climate change impacts watershed hydrology and contributes to alteration of hydrologic regimes in streams. However, global climate models (GCMs) operate at spatial and temporal scales that are too large to capture important watershed‐scale hydrologic shifts. A method of disaggregating monthly ensemble GCM data into temperature and precipitation data series for daily, watershed‐specific hydrologic simulations with SWAT was developed and assessed in the Soldier Creek watershed in northeast Kansas. A stochastic weather generator (WINDS) was employed to produce a baseline scenario (no changes from late 20th century conditions) and two scenarios based on ensemble means of 15 GCMs representing future conditions (A2 storyline) referred to as the 2050 and 2100 scenarios. Future hydrologic regimes exhibited non‐linear annual and monthly responses in hydrologic budget components, such as surface runoff, baseflow, and soil moisture, to temperature and precipitation changes. For the 2050 scenario, the combination of higher temperatures along with decreased annual precipitation and increased spring precipitation resulted in higher surface runoff, baseflow, and streamflow in May and June with a longer drought season later in summer. The significant decrease in streamflow, runoff, and baseflow for the 2100 scenario reflected an increase in monthly and annual temperature rather than a direct result of precipitation decline. The 2100 scenario also produced a reduction in low‐flow duration, an increase in the number of drought occurrences, and a decrease in flood frequency and duration. In retrospect, use of the stochastic weather generator to temporally downscale monthly GCM results while incorporating site‐specific climate variability (such as occurs with convective storms often missed in coarse‐resolution GCM data) produced more meaningful analysis of hydrological impacts, which is critical to predicting and understanding the impacts of climate change. Although this method allowed simulation of future‐climate shifts based on GCM‐simulated monthly shifts, it could not simulate potential shifts in climate patterns within a month, such as changes in transitional probabilities that govern the intensity and distribution of storms with months. In future work, translation of regional climate model responses into WINDS stochastic parameter adjustments will allow more accurate and efficient simulation. The severity of the increased drought and decreased flood responses simulated in this study would not be anticipated by review of precipitation trends alone nor by analysis of annual hydrologic responses alone. Similarly, many critical hydrologic responses reflected interactions between climate variables (e.g., precipitation and temperature) at sub‐annual temporal scales, which highlights the need to consider climatic interactions in future studies of climate change impacts.