A Novel Method of Error Evaluation for Radar Nowcasting Based on Shape Matching
The relatively conventional pattern forecast method of local shorttime rainfall is mainly based on satellite images and radar echo extrapolation, but it is also worth finding a better method which can make good use of ground observation data to effectively test the radar forecast products. In the actual weather forecast, rainfall is often characteristiced regional and planar spatial distribution. So the test about shape of rainfall area is more important and more significant. In order to solve the problem in shape test with variety of difficulties and characteristics. This paper proposed a comprehensive evaluating method of 0-3 h QPF (quantitative precipitation forecast) radar rainfall forecast and quantitative indexes of shape test. We also did some experiments including grading test and error analysis on a typical radar forecast of continuous raining process in Guangdong Province in 08:30-11:24 BT 22 April 2016. To some extent, this method also has a good solution of poor data quality and accuracy control in temporal and spatial scale. The TSshape, PODshape, FARshape, Ratiop, Ratiot, Jaccard of a half hour and 1 h radar rainfall forecast shape test are above 40%, above 40%, below 30%, above 40%, above 80%, above 40% respectively. The experimental results reflect the effect of radar rainfall forecast well and show that the indexes of shape test are basically consistent with conventional quantitative indexes.