Unconditional uncertainties of historical and simulated river flows subjected to climate change

The objective of this study was to estimate the unconditional sample uncertainty of observed streamflows and streamflows simulated from a hydrologic model (MISBA) for the Fraser River Basin (FRB) and the Athabasca River Basin (ARB) of Canada under historic conditions and under future conditions simulated by General Circulation Models (GCMs). For each basin, the multifractal properties of 54 simulated hydrographs based on the predictions of seven GCMs and four climate scenarios over three 30-year periods of the 21st century were evaluated and used to generate extended artificial time series by the randomized generalized multifractal cascade model. Uncertainty estimates derived from this multifractal approach were compared with classical statistics, Hurst exponent, and autocorrelation methods. The multifractal approach resulted in greater departures from statistical independence, higher skewness, and a wider range of flows over a 30-year time scale than the other methods. Under climate change, the multifractal strength of streamflows of FRB increased with temperature as the snow fed character of the basin weakened, but in the colder ARB, a decrease in multifractal strength resulted under rising temperatures because of declined snow packs. The estimated unconditional sample uncertainty was compared with uncertainty associated with model and emissions scenario selection. Among four major sources of uncertainty, uncertainty associated with two different hydrologic models (MISBA and SAC-SMA) and global emission patterns in the 21st century were relatively small (20% or less) even though for the latter, the range of SRES climate scenarios may not be representative of future economy and the seven GCMs may not accurately represent actual physical processes. The uncertainties associated with multifractal variation were the largest, on the order of ±50%.