Better Hydrological Projections by Combining Outputs from Weather Generator and Downscaled Climate Models

The pressure on water resources is projected to increase dramatically in the future and turn into a global crisis unless bold actions are taken. While Healthy water resources are key to every nation’s wealth and well-being, stressors such as climate change, land-use shifts, and increased water consumption are unfortunately threatening water availability and access worldwide. Researchers and practitioners are therefore under great pressure to develop methodologies and tools that can streamline projected changes into adaptation decisions. The vast majority of climate change adaptation studies use a top-down approach, which essentially consists of using of a limited set of climate change scenarios to discover future risks. There is also no established way to evaluate the credibility of a given projection scenario, making it a challenge to include them in a decision or design framework. Such approaches also inherit significant portion of uncertainties comes from underlying physical and social assumptions by general circulation models (GCMs). Even when multi-model multi-scenario projections are used, not all possible future conditions are covered and therefore plausible risks may be overlooked. This paper puts forth a new avenue to explore the likelihood of some climate-induced risks in hydrological modeling. It involves employing a very large number (i.e. 250) of climate realizations generated by a reliable weather generator forced by several climate change models and scenarios. Each realization of the downscaled climate time series is used afterwards individually as an input to a calibrated SWAT model to obtain streamflow time series. The new set of projections provides a better coverage of the possible risk space. Emplying such technique consists some "data mining" in large databases and thus facilitating decision makers to make proactive, knowledge-drive decision by identifying risky situations using stochastically generated climate states then use projections to evaluate their likelihood in the future.

Application on the South Nation Watershed in Eastern Ontario, Canada; implementing four regional climate models (RCMs) with two emission scenarios (RCPs), is presented.