Stochastic Generation of Future Hydroclimate Using Temperature as a Climate Change Covariate

Abstract

AbstractIt is now standard practice for water supply agencies to use stochastic models to generate synthetic hydroclimate sequences that preserve the key statistics contained in the observed/instrumental hydroclimate data (usually rainfall, potential evapotranspiration, and streamflow) for use as inputs when determining water security. However, this approach is only adequate for determining current and future water supply risks if nonstationarity in past and future climate is accounted for. In this study, we develop an approach for stochastically generating future seasonal (monthly to annual) hydroclimatic conditions at multiple sites for water supply security assessment that capitalizes on an Australia‐wide relationship between annual average daily maximum temperature and annual rain (and flow). This approach is practical as it (i) avoids the extra time and additional uncertainties introduced by downscaling and bias correction of climate model produced rainfall information and (ii) takes advantage of the fact that climate model projections for temperature change are more realistic than climate model projections for rainfall. The approach is applied to a case study for catchments supplying Sydney, Australia's largest city. Two future scenarios were evaluated, 1 and 2°C warming over the next 50 years. Large reductions in streamflow were simulated, particularly for the 2°C scenario, resulting in major impacts on water security in the absence of any intervention. This is consistent with recent warming since 1990 being associated with more than a 40% decrease in average annual streamflow when compared with the average annual streamflow over the ∼110 years historical record.