陈法敬,陈静. 2019. 一种基于可预报性的暴雨预报评分新方法Ⅱ:暴雨检验评分模型及评估试验[J]. 气象学报, (0):-, doi:10.11676/qxxb2019.003
一种基于可预报性的暴雨预报评分新方法Ⅱ:暴雨检验评分模型及评估试验
A New Method for Heavy Rainfall Forecast Verification Based on PredictabilityⅡ:Verification Method and Test
投稿时间:2017-12-11  修订日期:2018-05-30
DOI:10.11676/qxxb2019.003
中文关键词:  暴雨评分基函数  暴雨评分模型 邻域匹配  距离权重 评估试验。
英文关键词:Kernal  function of  heavy rainfall  scoring, Scoring  model, Neighborhood  matching, Distance  weight, Evaluation  experments.
基金项目:中国气象局气象预报业务关键技术发展专项(YBGJXM(2017)06),国家科技支撑计划项目(2015BAC03B01)
作者单位E-mail
陈法敬 国家数值预报中心 chenfj@cma.gov.cn 
陈静 国家数值预报中心 chenj@cma.gov.cn 
摘要点击次数: 398
全文下载次数: 299
中文摘要:
      针对当前暴雨预报检验采用二分类事件检验方法存在的较严重的“空报”、“漏报”双重惩罚,没有考虑暴雨时空分布不均和预报评分可比性不够等问题,在分析预报员对暴雨预报评分期望值基础上,设计了一种基于可预报性的暴雨预报检验评分新方法和计算模型,分析了理想评分,并对2015—2016年4—10月中央气象台5km*5km定量降水格点预报和降水落区等级暴雨预报进行评分试验,获得了以下结果和结论: 1)预报员对暴雨预报评分期望值呈现梯级下降特征,与传统的TS评分存在显著差异;2)设计了一种基于可预报性的暴雨预报检验新方法,通过引入e指数函数构建暴雨预报评分基函数,进而构建暴雨评分模型,该模型可以较好地拟合预报员对暴雨预报评分的期望值,同时改善了评分在不同量级阈值处的断崖式突变情况;3)提出了预报与观测的邻域匹配方法,即一个预报点与所定义邻域中的一组观测相匹配,并利用距离加权最大值法确定暴雨评分值权重系数,预报与观测距离越近,距离权重系数越大,评分值权重越大,提高了评分的合理性,避免了距离较远的匹配站点得高分不利于鼓励预报员提高预报精度的问题;4)对中央气象台逐日5km*5km水平分辨率的定量降水格点预报产品和中央气象台定量降水落区等级预报产品进行了评分试验,暴雨预报准确率全国平均值大于60分。基于可预报性的暴雨预报检验新评分与传统暴雨预报TS评分逐日演变特征相似,但可以较好地解析TS为0的预报评分,解析后的新评分与预报员和公众的心理预期更为接近。
英文摘要:
      Abstract In order to solve the problem that the current verified scores (Thread Score, TS) for heavy rainfall forecasting are too low to to be qualified for forecasters,as well as ignoring the uneven temporal and spatial distribution of heavy rainfall in China.this paper designed a new verification method and computational model for heavy rainfall forecasting based on its predictability as well as analyzed the expectation scores based on forecaster, A new score modele is designed and tested by use of 5km*5km quantitative precipitation grid forecast and precipitation location forecast issued by Central Meteorological Observatory from April to October in 2015—2016.The results and conclusions are as follows: 1) expectation scores of forecastersof for heavy rainfall forecast score shows a descending characteristics, which is different from the traditional TS score. 2) A new forecast verifcationt method based on predictability of heavy rainfall was designed, constructing the rainstorm forecasting score basis function by introducing e exponential function, and then construct the storm grading model. The model can well fit the expection of forecaster score for the rainstorm, and improve the score at different levels of the threshold . 3)Propose a neighborhood matching method of forecasting and observation, that is, a forecasting point matches the set of observations in the defined neighborhood, and uses the distance weighted maximum method to define the weight coefficient of the rainstorm score, the closer distance between the forecast and the observation, the greater the distance weight coefficient is, the higher the accuracy of the score value weight is, which improve the rationality of the score,and avoid the problem that a high score from a distant matching station is not conducive to encouraging the forecaster to improve the accuracy of the forecast.4)Quantitative precipitation grid forecasting products with 5km * 5km horizontal resolution and the quantitative precipitation location forecast products of the Central Meteorological Observatory were tested. The accuracy of the heavy rainfall forecast is over 60 on average, which is similar to the daily evolution characteristics of the traditional TS score, but for the forecast whose TS score is 0, this new score and is more consistent with the psychological expectations of forecasters and the public.
查看全文   查看/发表评论  下载PDF阅读器
分享按钮