Application of Doppler Radar Data to Nowcasting of Heavy Rainfall

Severe rainstorm, typhoon heavy rain and so on are the main precipitation system causing disastrous flood. In order to reduce the occurrence of meteorological disasters, timely forecasting and warning of severe weather are needed. The development on rainstorm nowcasting with tracking radar echoes by correlation (TREC) is described, which is also applied in Hong Kong rainstorm nowcasting system SWIRLS (Short-range Warning of Intense Rainstorms in Localized System) and NCAR's ANC (Auto-NowCast). TREC technique is applied to two successive CAPPI reflectivity fields. The first field is divided into a number of equally sized two-dimensional arrays of pixels. The arrays of reflectivity values are then cross-correlated with the arrays in the second field. The correlation coefficient R is calculated repeatedly for all possible arrays found at the second field to determine which array results in the highest correlation, and the center of this second array is the end point of TREC vector. Based on the Doppler radar data from the mosaic by Guangzhou and Meizhou radar, Wenzhou single radar, effects on the tracking result are discussed which are produced by the parameters variation including interval between two CAPPI reflectivity fields, the size of the boxes, the threshold of rain intensity etc. In order to correct noisy TREC vectors and improve the consistency of the resulting vector field, a two-step procedure is used:The purpose of the first step is to minimize the influence of apparently incorrect TREC vectors. Vectors with zero velocity (often caused by ground cluster) are replaced by vectors that represent the average of the neighboring vectors. Objective analysis is used in the second step to produce a continuous gridded vector field, which is used in rainstorm nowcasting. The nowcasting result is validated by making use of observational data of radar. The results show that the TREC vectors calculated from rain intensity CAPPI, reflectivity CAPPI and CR are consistent with each other, and they are capable of indicating the directory of rainstorm displacement. The result of extrapolation forecast from CR is slightly below the two others, and forecast accuracy of the three kinds of data are decreasing with forecast lead time, which can be improved by fitting the successive five TREC vectors. Though TREC technique segments radar imagery lacks of clear physical meaning, it is still valuable for forecasting of storm rainfall and typhoon rainstorm with complex structure and identified difficultly. Both single radar data and mosaic data are used in the research, which is helpful for forecasting severe weather utilizing Doppler weather radar observation net.