Distributed Modeling of Spencer Creek Watershed and Assessment of Future Changes in Hydrological Processes

Abstract

<p>The main purpose of this research is to implement and evaluate the effectiveness of a coupled model (MIKE SHE / MIKE 11) for Spencer Creek watershed (Ontario), and later to use this model for climate change impact study using Canadian Global Climate Model (CGCM 3.1) data and Canadian Regional Climate Model (CRCM 4.2) data. Both the CRCM and the CGCM data are downscaled using a Statistical Downscaling Method (SDSM) and a Time Lagged Feedforward Neural Network (TLFN).</p> <p>The hydrologic modeling results show that the coupled model captured the snow storage quite well with a correlation coefficient of 0.5-0.8. It also provided a good representation of evapotranspiration (ET) in the catchment with higher values in late spring and early summer months. The simulated streamflows are consistent with the observed flows at different sites with a Nash-Sutcliffe coefficient of around 0.4-0.5. The model couldn't capture the extreme or mixed events such as freezing rain in winter and rain on snow processes in early spring. Using a conservative climate change scenario, downscaled RCM with TLFN and SDSM yields smaller changes than raw RCM projections, but the downscaling with SDSM produces smaller changes than TLFN. With downscaled GCM scenarios, the coupled MIKE SHE/MIKE 11 model predicted 1-5% annual decrease in snow storage for 2050s and 5-22% increase with RCM scenarios. Similarly, with downscaled GCM scenarios, the coupled model predicted 1-10% increase in annual ET for 2050s and 2-22% increase with TLFN downscaled RCM scenario. But with SDSM downscaled RCM scenario, the model showed around 10% decrease in annual ET. Those results are consistent with the downscaled results for maximum and minimum temperatures. The coupled model predicted 10-25% increase in annual streamflows for all the stations with downscaled GCM scenarios- which is consistent with the predicted changes in the snow storage and ET. With raw RCM scenarios, the model predicted 5- 12% increase in annual streamflow, and 3-30% decrease with downscaled RCM results showing consistency with predicted increase in ET and the negative to small increase in precipitation. Overall, the wide range of projected future changes in hydrologic processes predicted by this study can be useful for understanding the integrated effect of climate change in this complex catchment.</p>