Kilometer-scale multi-physics simulations of heavy precipitation events in Northeast China

• Published in: Climate dynamics 2024-09, Vol.62 (9), p.9207-9231
• Main Authors: Yu, Hongyong, Prein, Andreas F., Qi, Dan, Wang, Kaicun
• Format: Article
• Language: English
• Subjects: Boundary layer stability; Boundary layers; Climatology; Computer applications; Computing costs; Control stability; Convective instability; Earth and Environmental Science; Earth Sciences; Estimation; Gauges; Geophysics/Geodesy; Heavy precipitation; Hurricanes; Maximum precipitation; Microphysics; Oceanography; Original Article; Parameterization; Physics; Planetary boundary layer; Precipitation; Satellite observation; Simulation; Snowstorms; Typhoons
• ISSN: 0930-7575; 1432-0894
• DOI: 10.1007/s00382-024-07386-w

Despite the fatal impact of heavy precipitation on people’s lives and the social economy, its accurate estimating remains challenging. In this study, we show how to address this issue by kilometer-scale simulations and how to reduce computational costs in Northeast China with the complex terrain and distribution of land and sea. Three typical heavy precipitation events are simulated at 3 km horizontal resolution, and each event is simulated with 24 combinations of schemes (with or without a scale-aware cumulus scheme, three microphysics schemes, and four planetary boundary layer schemes), which are evaluated against gauge observations. Compared to gauge observations, the ensemble mean of simulations of hourly maximum precipitation and average accumulated precipitation outperforms three widely accepted satellite products in the cold vortex and the snowstorm case, and is of comparable accuracy in the typhoon case. Overall, the microphysics scheme significantly impacts the maximum hourly precipitation, whereas the planetary boundary layer scheme has a strong control over the accumulated precipitation. The similarity among different simulations is linked to the level of convective instability's impact on heavy precipitation in each case, which also indicates that conducting 24 simulations can be not necessary. This study uses an ensemble performance estimation technique assuming the impact of different schemes is additive and finds that performing 13 rather than 24 simulations allows finding the best-performing combination of parameterization schemes, which allows for saving almost 50% of computational costs.