Abstract:

Manual and automatic precipitation data (i.e., parallel observation data) collected at 2363 meteorological stations in China between August 2017 and August 2018 were used to analyze the integrity and accuracy of automatic observed data (singular value removed). Firstly, the comparison of the capture rate indices of rain, snow, sleet, hail, and drizzle obtained from parallel observations shows that, during the entire precipitation process, rain, snow, and drizzle had a higher capture rate than sleet and hail had, with values of 66.9%, 69.2%, and 50.1%, respectively; and the hourly capture rate of hail (51.8%) was the highest. Secondly, as for miss rate of phenomenon automatic observed, the highest value was recorded for drizzle (65.9%), the second-highest for snow (35.6%), and the lowest for hail (16.2%). Thirdly, the misreported rates of drizzle, rain, and snow were lower than those of sleet and hail: Analogous percentages of rain were misreported as drizzle, higher percentages of drizzle or snow were misreported as rain, and also higher percentages of snow were misreported as drizzle or rain. The last statistical result is that the index of empty report rate shows that drizzle and hail were not captured at a substantial number of meteorological stations through automatic observations. Compared with manual observations, a portion of precipitation phenomena was missed, misreported, or even not captured by automatic observations using disdrometers, although these instruments have a minute level temporal resolution. Brook no delay, quality control procedures should be carried out once the data are collected by disdrometers, the algorithm for phenomenon identification should be continuously optimized, and the comprehensive identification of precipitation phenomena should be carried out by combining other weather elements.

Key words: precipitation phenomenon, parallel observation, data accuracy, capture rate, miss rate, misreported rate