雷达定量降雪估测方法的设计和订正

THE DESIGN AND CORRECTION OF A QUANTITATIVE METHOD OF SNOW ESTIMATE BY RADAR

从降雪预警业务实际出发,设计了基于最优化法的雷达估测降雪方法,对2007年3月4日特大暴雪过程开展雷达降雪估测试验,并分析估测结果的误差。针对温度变化、雪花末速度、与雷达的距离和计算方法等方面的误差因素制定了3种改进方案。改进后的估测降雪量与实况降雪量的相关系数提高到0.6(6超过99%信度检验),平均相对误差降低至48.74%,对于0.3 mm/h的较弱降雪和5 mm/h以上的强降雪均具有估测能力。其中距离雷达50～100 km的样本估测降雪量与实况降雪量的相关系数达到0.82。在3种改进方案中,考虑降雪末速度影响的改进效果不明显,这可能与本次暴雪过程的回波较均匀有关;按雷达与样本距离分类进行雷达降雪估测的效果最明显,不仅可以增加相似程度,还减小了雷达近距离高估和远距离低估的误差;而算法的改进进一步提高估测精度。本次雷达降雪估测对于1.6～2.5 mm/h的较强降雪和2.6 mm/h以上的强降雪平均相对误差较小,分别为31%和27%,但雷达降雪估测高估了1.5 mm/h以下的降雪而低估2.6 mm/h以上的强降雪。一方面说明雷达回波对于降雪强弱变化不是很敏感,另一方面在业务实际工作中有可能利用这种一致性的误差进行订正,以提高降雪估测精度。

An optimization method is based to design a snowfall estimate method by radar for operational snow warning, and error estimation is analyzed through a case of heavy snow on March 4, 2007. Three modified schemes are developed for errors caused by temperature changes, snowflake terminal velocity, the distance from the radar and calculation methods. Due to the improvements, the correlation coefficient between the estimated snowfall and the observation is 0.66 （exceeding the 99% confidence level）, the average relative error is reduced to 48.74%, and the method is able to estimate weak snowfall of 0.3 mm/h and heavy snowfall above 5 mrn/h.The correlation coefficient is 0.82 between the estimated snowfall from the stations 50 ~ 100 km from the radar and the observation. The improved effect is weak when the influence of the snowflake terminal velocity is considered in those three improvement programs, which may be related to the uniform echo. The radar estimate of snow, which is classified by the distance between the sample and the radar, has the most obvious effect： it can not only increase the degree of similarity, but also reduce the overestimate and the undervaluation of the error caused by the distance between the sample and the radar. The improved algorithm further improves the accuracy of the estimate. The average relative errors are 31% and 27% for the heavy snowfall of 1.6 ~ 2.5 mm/h and above 2.6 mm/h, respectively, but the radar overestimates the snowfall under 1.5 mm/h and underestimates the snowfall above 2.6 mm/h. Radar echo may not be sensitive to the intensity of snowfall, and the consistency shown by the error can be exploited to revise and improve the estimation accuracy of snow forecast in the operational work.