Quality Control Method for Land Surface Hourly Precipitation Data in China

Abstract

High spatial-temporal resolution observations of precipitation from automatic weather stations (AWSs)serve as a vital data source, extensively utilized in research activities such as severe weather monitoring, model evaluation, and forecast analysis. Influenced by factors such as observation environments and equipment performance, precipitation observations inevitably contain various forms of random and systematic errors. A quality control method (multi-source data collaborative quality control, MDC) has been established for hourly precipitation data from AWSs in China, based on high spatial-temporal resolution radar data and weather phenomena. The MDC includes three modules: Precipitation self-detection, multi-source data collaborative detection, and dynamic blacklisting. The MDC has been applied to quality control of hourly precipitation data from AWSs from 2021 to 2023. A comprehensive effectiveness assessment of the method has been conducted using a combination of quantitative indicators and case analyses of detection effects on various types of erroneous data. Results indicate that the correct identification rate of the MDC reaches 99.92%, with a false exclusion rate of 0.08%. The majority of falsely excluded data consists of weak precipitation amounts ranging from 0.1-1 mm, accounting for 60.72%. While ensuring a high correct rate, the MDC also demonstrates a high capability in identifying erroneous data. The average error data hit rate of the MDC in China is 39.8%, which represents an improvement of 39.3% over the existing Meteorological Ddata Operatioin System (MDOS) real-time quality control system. The ability MDC to identify erroneous data between 0-50 mm is approximately 40%, and this hit rate significantly increases with higher precipitation values. When precipitation amounts exceed 100 mm, the hit rate achieves 100%. MDOS real-time quality control system has an almost zero hit rate for erroneous data with precipitation amounts less than 20 mm but possesses some identification capability for abnormal precipitation of more than 20 mm.The hit rate of the MDC shows significant spatial variation due to the coverage of radar and national station observations. In the eastern region, where observation stations are densely distributed, most stations have an error data hit rate of over 90%. However, in the western and northeastern regions, where observation stations are sparse and do not meet the conditions for multi-source collaborative detection, the hit rate of the MDC decreases significantly, approaching that of the MDOS real-time quality control. Case analyses of the quality control effects on different types of erroneous data reveal that the MDC significantly the identification ability of abnormal data such as clear sky precipitation, snowmelt precipitation, and false zero value precipitation, effectively making up for the deficiencies of traditional methods.

Keywords

• automatic weather stations
• hourly precipitation data
• multi-source data collaborative detection
• quality control