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

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 mm/h. The correlation coefficient is 0.82 between the estimated snowfall from the stations 50 to 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 to 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.