Sensitivity of Numerical Atmospheric River Precipitation Forecasts to Choice of Microphysics Parameterization

The Center for Western Weather and Water Extremes (CW3E) at Scripps Institution of Oceanography is tasked with providing accurate and reliable numerical forecasts of weather and hydrological extremes, such as Atmospheric Rivers (ARs), along the U.S. West Coast. As such, the CW3E has developed a high-resolution regional model, based on the Weather Research and Forecasting (WRF) model to provide such forecasts (West-WRF). As part of an ongoing effort to enhance the performance and efficiency of weather prediction systems utilized by CW3E, we leverage the 34-year West-WRF Reforecast (Cobb et al. 2023) to investigate the sensitivity of orographic quantitative precipitation forecasts (QPF) in AR forecasts to the choice of microphysics parameterization (MP). By varying only microphysics, we produce a 14-member 168 h West-WRF re-reforecast of a landfalling AR event occurring from 7-10 January 2017.

Recent advances in model microphysics have led to the development of multi-moment schemes (e.g., National Taiwan University) and perhaps even more notable; a shift in the treatment of ice phase hydrometeors [e.g., Predicted Particle Properties (P3)] where prescribing the evolution of a single ice class is favored in lieu of the traditional method, which prescribes separate ice categories (e.g., snow, graupel, hail) to represent riming characteristics. Although some MP schemes have been rigorously tested and deemed suitable for operational use (e.g., Thompson and Eta Ferrier microphysics), newer and more sophisticated schemes, such as those described above, may offer considerable improvements in operational NWP forecasts but are perhaps less well tested in operational settings. Here, we investigate the relative performance of these schemes. We seek to identify what benefits or drawbacks these newer schemes provide to an AR forecast relative to more traditional and commonly used MP schemes.

When compared to the Stage-IV gridded analysis dataset, results show that, in general, there is an ensemble-wide high bias in QPF along the Sierra Nevadas; however, the quality of observations in high terrain is uncertain. Nevertheless, analysis of 72-h precipitation accumulation fields indicate that the spatial distribution of QPF along the Sierra Nevada’s is highly sensitive to the choice of microphysics parameterization. We observe that many schemes produce similar spatial patterns of QPF at smaller watershed scales, such as in the Upper Yuba in California, where accurate QPF is critical for regional water management. Relative to the ensemble mean, we find some newer schemes produce less precipitation over the Sierra Nevadas and more on the leeside, but not all of them. We show that schemes with higher accumulation totals, such as the NTU scheme, are generally more efficient at producing model precipitation, likely due to differences in representations of hydrometeor species, which directly impacts the rate and quantity of hydrometeor fallout. Overall, schemes which predict the evolution of ice properties show promise, especially from a cost-analysis benefit perspective, where they not only show more reasonable QPF but do so with nearly the same or less computational expense relative to their traditional counterparts. We recognize that additional cases will be needed for more general results.