An Overview of the Western United States Dynamically Downscaled Dataset (WUS-D3)
Ensemble mean future changes in (a) seasonal surface air temperature [K/K] and (b) precipitation per degree of global warming [mm/d/K] from 16 downscaled GCMs. Hatching indicates statistical significance to the 95 % confidence interval when grid point distributions are subjected to a two-sided Student’s t test. Stippling is not included for temperature because every grid point returns a p value smaller than 0.05.
Image by Stefan Rahimi, University of California Los Angeles
High-resolution physics-based climate change ensembles remain challenging to create due to limited compute (job execution costs and storage) along with Global Climate Model (GCM) data availability/standardization issues, and because GCMs contain biases that may distort a dataset’s fitness for purpose. This study presents a candid take on the challenges of producing dynamically downscaled datasets using the western United States as a laboratory to overview (i) a rigorous GCM selection process, (ii) historical GCM bias profiles, and (iii) document base climate change signals in mean and extreme precipitation and temperature, to ultimately assess the addition (or subtraction) of value by dynamic downscaling.
High-resolution physics-based climate change ensembles remain challenging to create due to limited compute (job execution costs and storage) along with Global Climate Model (GCM) data availability/standardization issues, and because GCMs contain biases that may distort a dataset’s fitness for purpose. This study presents a candid take on the challenges of producing dynamically downscaled datasets using the western United States as a laboratory to overview (i) a rigorous GCM selection process, (ii) historical GCM bias profiles, and (iii) document base climate change signals in mean and extreme precipitation and temperature, to ultimately assess the addition (or subtraction) of value by dynamic downscaling.
Challenges of producing a decision-relevant, physics-based climate change dataset are presented. However, just because a dataset is physics-based, that does not mean the dataset is physically credible. We present instances of physical skill and a lack thereof, laying the framework for other simulations which employ bias correction of the GCM boundary conditions to improve the regional climate modeling solution.