Downscaling is the collective term for the methods used to regionalize information from global climate models and create fine-spatial-scale projections of climate change. Our group is active in the development, evaluation, and application of downscaling techniques.
Until recently, there were two main types of downscaling methods: dynamical methods, which involve the use of high-resolution regional climate models, and statistical methods, which use mathematical relationships between local climate variables and their large-scale predictors. Our group has pioneered a third type of downscaling technique, which we call hybrid downscaling. As the name implies, it combines aspects of dynamical and statistical downscaling. We perform a limited set of dynamical downscaling simulations, analyze them in order to diagnose physical relationships between large-scale and fine-scale climate variables, and then build a statistical model that incorporates those physical relationships. With this statistical model, we are able to enlarge our set of simulations. This work enables us to assess climate change impacts on the scales that matter most to those making decisions about climate change adaptation.
In recent years, our group has also produced an extensive suite of dynamical downscaling simulations. Here, we make use of global climate model output from CMIP6 to drive high-resolution future Weather Research and Forecasting (WRF) model simulations out to the year 2100 across the western US. These simulations are predominantly executed at 9km resolution but can be as fine as 3km resolution in some cases. More details on the Western US Dynamical Downscaling Dataset can be found here: https://dept.atmos.ucla.edu/alexhall/downscaling-cmip6.