RAL/CGD Joint Seminar - Dynamical downscaling & climate change: what the heck are we looking at?

Dynamical downscaling remains the most physically based method by which high-resolution climate change data are created. Here, an ongoing dynamical downscaling effort within a multi-institutional collaboration is described. The overarching goal is to use dynamical downscaling in concert with a statistical downscaling approach to create a high-resolution hybrid downscaling procedure that will result in the entirety of the CMIP6 ensemble downscaled to landscape scales across the western United States.
This presentation focuses on the dynamical downscaling leg of our mission. Amongst items explored, we present (i) the testing and historical (1980-2020) performance of a reanalysis-driven simulation, (ii) a rigorous GCM selection procedure, (iii) the impacts of a priori bias correction, and (iv) how a single direct downscaling simulation compares to a simulation subjected to the commonly applied “pseudo global warming” (PGW) technique. We discuss these topics in context of our grander research scope. In doing so, we stress the importance of extensive model testing and bias characterization before climate-scale simulations are launched. We also find that, despite the excellent performance of GCMs in simulating hemispheric, process-based, and even local climate features within an extensive performance evaluation construct, the regional climate model can produce unphysical results. However, we find that pre-downscaling bias correction can avoid some of these unphysical regional-scale behaviors for simulations driven by better-performing GCMs. Finally, we note important similarities and key differences between direct and PGW downscaling techniques.