Evaluation of future flow variability in ungauged basins: Validation of combined methods

Predicting future flow variability and changes in ungauged basins remains a challenging task not only because of data limitation, but also the various sources of uncertainty involved. However, for future water resources development projects, reliable estimate of future flow variability is essential. In this study, future flow variability and changes are estimated in 90 selected basins across the Province of Ontario using a multi-model approach combining bias correction and regional climate models (RCMs), downscaling and global climate models (GCMs), dual regionalization method and hydrologic model. Raw RCM outputs, bias corrected RCM data, and downscaled GCM outputs are compared to identify the most accurate predictor (precipitation, temperature) series when compared to the observed data for the current period. It is found that the bias correction approach can reduce monthly precipitation and daily temperature errors of regional climate models by about 90% and provides the most accurate precipitation and temperature time series. The bias-corrected RCM data are used with a rainfall-runoff model and a coupled regionalization method to simulate continuous streamflow in both gauged and ungauged basins. It is found that consistent streamflow and baseflow estimate can be obtained whatever the RCM data used. In general, positive (increasing) changes in streamflow are observed for all seasons except for March, April and May where negative (decreasing) trends are dominant in some basins. Future changes in baseflow are mostly positive with smaller amplitude whatever the season for almost all the 90 selected basins. Overall, it is shown that for northern, central and southern regions of Ontario, the amplitude of the variability and changes of future streamflow and baseflow in ungauged basins should be in the range of −10 to +10mm per month based on most recent RCM simulations.