In this paper, data acquired by artificial observation and precipitation phenomena observation instruments at 20 observation stations of Beijing from September 2017 to Augusta 2018 are used to compare and assess the precipitation phenomena observation instruments. The results show that the lack of measurement rate of precipitation phenomenon instrument is 0.01%, false prediction rate is 0.6%, the capture rate is 93.5%, the missing report rate is 27.9%, and misreporting rate is 10.0%. From the above statistical parameters, we can see that the lack of measurement rate and false prediction rate of the instruments are low and precipitation phenomena capture rate is high, which shows the good performance of the instruments. The missing report rate and misreporting rate are high, indicating missing and misidentification casetend to happen frequently. Missing report mainly happens in the process of light precipitation and the misreporting mainly in the process of drizzle identification. The time consistency of distinguishing the beginning of precipitation processes is better, but poor at the ending time. The existing problems of the instruments are that raindrop overlap misidentification is easy to appear during heavy precipitation, and the identification rate is reduced during light precipitation. The quality control process is lacking in the case of wrong outputs of precipitation when no precipitation is seen.