Evaluation of a Reanalysis‐Driven Configuration of WRF4 Over the Western United States From 1980 to 2020

• Published in: Journal of geophysical research. Atmospheres Vol. 127; no. 4
• Main Authors: Rahimi, Stefan, Krantz, Will, Lin, Yen‐Heng, Bass, Benjamin, Goldenson, Naomi, Hall, Alex, Lebo, Zachary J., Norris, Jesse
• Format: Journal Article
• Language: English
• Published: 27.02.2022
• Subjects: dynamical downscaling; reanalysis; regional climate; regional modeling
• ISSN: 2169-897X; 2169-8996
• DOI: 10.1029/2021JD035699

Dynamical downscaling remains a powerful tool for studying regional climate processes, and the genesis of high‐resolution historical and future climate data. This technique is particularly important over areas of complex terrain, such as the western United States (WUS), where global models are especially limited in representing regional climate. After identifying a suite of WRF options that best simulate snow and precipitation for an average water year (2010) over the WUS, we evaluate the performance of the dynamically downscaled European Centre for Medium‐range Weather Forecasting's fifth Reanalysis (ERA5) from 1980 to 2020 on 45‐km, 9‐km, and two 3‐km grids. We find that by decreasing the horizontal grid spacing within WRF, improvements to Sierra Nevada and Northern Rocky Mountain snow, Santa Ana and Diablo winds, and coastal meteorology occur. For landfalling atmospheric rivers (ARs), the downscaled reanalysis simulates greater upstream integrated vapor transport (IVT) than ERA5. However, WRF skillfully simulates the positioning of the IVT and the timing and magnitude of AR precipitation. This potential IVT bias, in conjunction with increasing resolution, leads to a wet precipitation bias across the Sierra Nevada in the 3‐km experiment. This conclusion is supported by streamflow analysis, although we note that the bias in the 3‐km experiment can also be explained by in situ undercatch issues. Meanwhile, the 9‐km experiment is more biased than the 3‐km experiment across the Northern Rocky Mountains compared to in situ measured SWE and precipitation, indicating a geographic sensitivity to biases. Plain Language Summary Atmospheric reanalyses are often used to evaluate the performance of global climate models (GCMs). Reanalyses have a model component, but they also incorporate observations in post‐processing to orient the final product about reality. The horizontal resolution of reanalyses is too low to resolve small landscape features that control localized meteorology and climate. Thus, scientists are in need of an even higher resolution modeling framework. It is therefore necessary to embed a regional climate model (RCM), a tool that uses physics to simulate weather in higher resolution, within the reanalysis to generate kilometer‐scale weather and climate information. This process is called dynamical downscaling. Using this process, we show that, in spite of increased resolution, selecting the wrong RCM configurations can lead to poor RCM performance compared to observations, and we identify a superior set of RCM options for simulating weather and climate across the western United States. Following our tests, we find that the RCM adequately reproduces the historical weather and climate (1980–2020), and RCM accuracy improves with increasing model resolution. In creating this product, we now have a benchmark by which to evaluate the fidelity of dynamically downscaled GCMs, as we seek to extend this modeling framework to future climate. Key Points A tractable method for dynamically downscaling the historical record across the western United States is presented Thorough testing of regional climate model (RCM) options is imperative before dynamically downscaling over climate time scales Higher‐resolutions in RCMs should generally add value when considering a diverse array of simulated atmospheric phenomena