When Does Choice of Downscaling Method Matter in Decision Making? A Case Study with Maple Syrup Production
Abstract
When planning and preparing for climate change, practitioners rely on climate models to help them make informed decisions. All climate change model data ultimately derive from global-scale models, which are typically too coarse for local-scale decision making; thus, these models are often "downscaled" in order to resolve finer details within the decision space. A few downscaling approaches exist, each with a unique set of strengths and limitations, yet their implications on any particular decision are not always clear to decision makers. Using maple syrup production as a case study, we demonstrate a possible method of evaluating the sensitivity of a specific decision to downscaling method selection. We compare two downscaling techniques (dynamical and statistical) and two training methods within the statistical downscaling approach (bias-corrected spatial disaggregation, or BCSD, and bias corrected constructed analogs, or BCCA) with respect to their ability to capture daily freeze- thaw cycles, the driver of sapflow in maple syrup production. For each downscaling approach, we evaluate simulations of historical freeze-thaw patterns using gridded temperature observations, and compare projected changes in freeze-thaw patterns by mid-century. We discuss the implications of our results on the decision of "when to tap" faced by maple syrup producers, as well as similar decisions in other industries. Our results reveal which downscaling technique(s) is (are) best suited for helping maple syrup producers make plans toward adapting their tapping practices for climate change. In addition, our results provide producers with a plausible range of optimal tapping dates by mid-century, based on the most skilled downscaling approach. Finally, we highlight insights relevant to the climate modeling community, and lessons learned toward making climate science actionable.