Abstract:Radar malfunctions and other factors often lead to abnormal echoes, directly affecting data quality. The abnormal data entering the data sharing system have a very serious impact on the quality of observation data in the network of China’s new generation weather radar (CINRAD), as well as quantitative applications such as radar data assimilation and precipitation estimation, reducing the accuracy of short-term forecasting systems and numerical forecasting models. At present, the meteorological department carries out stream transmission, data sharing, and business monitoring based on radar radial data. However, data quality control mainly relies on the analysis and processing of volume scan data, using the whole or regional image features for analysis and processing. The data analysis is not precise enough, the identification of erroneous data is not accurate enough, and the timeliness is low, which cannot provide fast and refined real-time radar data quality control. At the same time, abnormal data have randomness and diversity, but existing data quality control methods mainly target a certain factor that affects the quality of radar data, such as sea waves, bird flocks, electromagnetic interference, radar fault echo data, etc. Therefore, it is urgent to study a universal method for rapid identification of abnormal echo data, improve the timeliness and quality of radar data quality control, and meet the business needs of refined forecasting services. This article proposes a method for identifying abnormal data based on radar radial data feature analysis. By analysing the number of radial echoes, the maximum distance of radial echoes, the average intensity of radial echoes, and the correlation of radial reflectivity factors, a membership function and an identification equation for abnormal echo data are established to determine the threshold for identifying abnormal echo data. This method more accurately and finely identifies multiple types of radar abnormal echoes. Under clear weather conditions, the position and intensity of ground clutter are relatively fixed. By statistically analysing the dynamic range of the number of normal radial echoes, the maximum distance of radial echoes, and the average intensity of radial echoes, the membership functions and thresholds of each parameter of abnormal radial data are determined, and a comprehensive identification equation is established. Under non-clear weather conditions, there are meteorological echoes, and the evolution of normal meteorological echoes has reasonable limits, with a certain degree of continuity in time and space. The radial reflectivity of adjacent volume scans has good correlation in normal echo data, while radar abnormal echoes often have suddenness, manifested as abnormal distributions of the radial echo numbers, the maximum radial echo distances, and the average radial echo intensity, and reduce the correlation between radial reflectivity of adjacent volume scans. Therefore, normal radar echo data are taken as a sample, and a certain time interval is selected. Through dynamic range statistical analysis of four parameters: the ratio of the number of radial echoes, the ratio of the maximum distance of radial echoes, the ratio of the average intensity of radial echoes, and the correlation of radial reflectivity factors in the same time interval and elevation angle adjacent azimuth radial data of normal echoes, the membership function, identification equation, and judgment threshold of abnormal echo data are determined to identify abnormal radial data. Using the above identification methods, the hardware faults, as well as abnormal echo data caused by calibration errors, super refraction, and electromagnetic interference of Shijiazhuang CINRAD/SA new generation weather radar, are identified. The overall recognition rate of abnormal data is over 90%. The abnormal data recognition method based on radial data using fuzzy logic can be applied to identify various abnormal echo data caused by hardware failures, as well as non-meteorological echoes such as super-refraction and electromagnetic interference, and has wide applicability. It quickly and finely identifies abnormal radar echo data, supplements and improves existing radar data quality control methods, and provides technical support for providing high-quality radar detection data to meet the precise forecasting and fine service meteorological business needs.
