Non-precipitation Identification Technique for CINRAD/SAD Dual Polarimetric Weather Radar

In China, the operational upgrade of dual polarimetric weather radar is being promoted. CINRAD/SAD dual polarimetric weather radar in some provinces such as Guangdong, Jiangsu, Shandong and Zhejiang has been upgraded in operation. By June of 2021, there are 69 dual polarimetric weather radars in national radar network, and it will increase to more than 100 in the future. The dual polarimetric radar is an important detection equipment for studying the microphysical process of precipitation, which can provide multiple polarizations including raindrop spectrum information, and thus better describe the microphysical characteristics of precipitation. The technical upgrade will bring revolutionary changes for data quality control, hydrogel classification and quantitative precipitation estimation. With the measurement parameters such as correlation coefficient or differential reflectivity, the dual-polarimetric weather radar can effectively remove non-precipitation echoes such as ground clutter, anomalous propagation, electromagnetic interference, sea waves, clear air clutter and so on. Based on the non-precipitation identification technique on S-band WSR-88D dual polarization weather radar, the distribution characteristics of correlation coefficient and differential reflectivity in precipitation echo and clutter are analyzed. The CINRAD/SAD dual-polarimetric weather radar data are used to test and improve the algorithm to adapt domestic weather radar, the differential reflectivity texture feature is added in the improved algorithm and the distribution characteristics of differential reflectivity horizontal texture on precipitation echo and clutter are analyzed, to better remove non-precipitation echo. During the evaluation of algorithm, several cases such as hail, melting layer, typhoon and different types of clutters during May-October in 2019 and 2020 are investigated. The results show that the improved algorithm can identify 95.2% of non-precipitation echoes, and the error rate of precipitation is 2.6%. For the large area clear air clutter, after adding the differential reflectivity texture feature, combining with the correlation coefficient texture feature, the accuracy of the algorithm is improved from 68.6% to 96.8% for one case, but the overall accuracy is less than 90% for many cases, and it needs to be improved by deep learning method in the future. Non-precipitation identification algorithm on CINRAD/SAD is applied in mosaic image, showing great application prospect in the future for precipitation classification and quantitative precipitation estimation. It can provide high quality data and play an important role in real-time operation.