Applicability Evaluation of Provincial Precipitation Real-time Analysis Product in Beijing

Abstract

The national precipitation real-time analysis product is a gridded product developed using probability density matching, Bayesian model averaging, multi grid variation, optimal interpolation and other technologies by National Meteorological Information Center. It has advantages of high accuracy, high quality, and spatiotemporal continuity, and is widely used in national nowcasting forecasting operations. In September 2022, National Meteorological Information Center issues a provincial multi-source fusion real-time analysis system to promote the collaborative application of precipitation analysis in different provinces. The same core multi-source fusion algorithms for real-time precipitation analysis products are applied in this system, allowing access to additional provincial local observations. The consistency and accuracy of the provincial precipitation analysis products in Beijing from September 2022 to August 2023 are evaluated by automatic weather station observations, error analysis, effective precipitation time proportion, and other methods. "23·7" extreme precipitation event is also analyzed in terms of cumulative precipitation, precipitation intensity, and hourly precipitation error. Results show that root mean square error of the provincial precipitation analysis product is less than 1 mm, and the average absolute deviation is below 0.16 mm, which closely aligns with observations from automatic weather stations. The bias of provincial precipitation real-time analysis product increases with magnitude of precipitation. The intensity of light rain exceeds the observation, while the spatial distribution difference of bias is evident. The maximum negative deviation occurs in both moderate rain and rainstorm magnitudes of Yanqing, while the maximum positive deviation is observed at Changping. During the extreme precipitation event of "23·7", the spatial distribution of provincial precipitation real-time analysis product is largely consistent with observations from automatic weather stations. The precipitation intensity is consistent with the trend of time variation observed by automatic weather stations, with an average root mean square error of 1.8 mm and an average absolute deviation of 0.806 mm, which more accurately reflects the trend of precipitation intensity variation. Overall, the provincial precipitation real-time analysis product has high accuracy in Beijing and can reflect the spatial distribution of precipitation, but the estimate is lower than the observation in the precipitation course of local heavy rainstorm.

Keywords

• real-time analysis product
• skill scoring
• multi-source fusion
• accuracy evaluation