An Approach to Model and Evaluate Stream Temperature Response to Climate Change in Wisconsin

Expected climatic changes in air temperature and precipitation patterns across the State of Wisconsin may alter future stream temperature and flow regimes and in turn influence fish distribution and assemblage composition. In an effort to gain a better understanding of how climatic changes may influence stream temperature, two models were integrated: an artificial neural network (ANN) stream temperature model and the US Geological Survey (USGS) Soil-Water Balance (SWB) model. The ANN stream temperature model was developed to predict daily mean summertime stream temperature (June 1 to August 31) for the 1:100,000 scale National Hydrography Dataset streams in Wisconsin. The ANN model used available landscape categorical data and climatic time series data, along with water temperature measurements collected at 371 wadeable stream sites during the summer time period 1990 to 2008 (1748 days). The SWB model was developed to provide estimates of groundwater recharge on a daily time step and provided a means to evaluate changes in groundwater recharge and air temperature, with links to downscaled regional climate models. The SWB model used a time series of daily values for precipitation and air temperature along with landscape characteristics that included land cover, soil type, and available water holding capacity. Estimates of mean annual recharge were derived from SWB for the time period 1989 to 2008 and used as input to calibrate the ANN stream temperature model. The ANN model and SWB recharge estimates were integrated and used to generate a water temperature time series for 38,532 confluence to confluence stream segments covering approximately 86,898 km of stream length within Wisconsin.  ANN model performance was assessed under current climate conditions, with provisions to simulate future conditions under various climate-change scenarios. The ANN model explained 74% of the variation in observed daily mean water temperatures over the 1748 day period for 67 independent sites. The approach to integrate the two models provides a means to estimate the temporal variability in groundwater recharge and mechanism to evaluate the influence of changing air temperature and precipitation amounts on groundwater recharge and soil moisture.  It also provides a means by which downscaled regional climate model results can be used to estimate the effect of climate change on stream temperature and in future studies, associated fish species distribution and assemblage composition.