A Novel Emergent Constraint Approach for Refining Regional Climate Model Projections of Peak Flow Timing
Global climate models (GCMs) are unable to produce detailed runoff conditions at the basin scale. Assumptions are commonly made that dynamical downscaling can resolve this issue. However, given the large magnitude of the biases in downscaled GCMs, it is unclear whether downscaled GCM projections are credible, especially for stakeholder-relevant land surface variables. Here, we use an ensemble of 12 GCMs (SSP3-7.0) that were dynamically downscaled by the Weather Research Forecasting (WRF) model across the western US to evaluate this question. We focus on evaluating changes in peak flow timing (defined as the annual maximum 1-day flow) across snowmelt-influenced basins in the Sierra‐Cascade Mountain range of the western US. We then apply the emergent constraint method, for the first time to dynamically downscaled data, to reduce uncertainty in projections of peak flow timing. Note, this stakeholder‐relevant variable, which requires dynamical downscaling, is typically obtained from subsequent hydrologic simulations with bias‐corrected data. However, whether or not the removal of GCM biases influences projected change signals in a physically realistic manner is a matter of ongoing debate. To further address this question, we evaluate if the removal of historical biases via two bookend bias correction techniques reinforce a particular outcome of the emergent constraint approach.
Understanding shifts in peak flow timing is important for adapting flood and water supply management, particularly in reservoirs that capture flow from snowmelt-influenced basins in mountainous regions. If peak flow is expected to occur earlier due to shifts towards precipitation falling as rain rather than snow, then this shortens the time-frame water/reservoir managers can capture and store water for later seasons when it is dry. As such, we focus on understanding why there is a large spread in dynamically downscaled projections of this variable, and evaluate if we can reduce this projected uncertainty.
We find that dynamically downscaled GCMs have a large spread in their historical representation of peak flow timing. This spread is driven by cold and warm biases in GCMs that influences the amount of precipitation that falls as snow or rain. For example, a GCM with a cold bias leads to more precipitation falling as snow, and thus peak flow conditions driven by snowmelt typically occur later when compared to observational data of peak flow timing. These cold biased GCMs typically have larger shifts in peak flow timing due to the greater sensitivity or larger room for change in precipitation falling as rain vs snow compared to a GCM that has a warm bias. Using the physical underpinning of biases in historical peak flow timing, and the emergent constraint method, we are able to reduce projections in peak flow timing uncertainty across basins in the Sierra and Cascade Mountain ranges by 39% (25 ± 34.75 days, 95% CI to − 28.25 ± 20.75 days, 95% CI). We additionally find that bias correction procedures sometimes agree with our emergent constraint approach; however, they can also lead to greater uncertainty.