【作者】 杨东；

【导师】 马敏劲； 陈静；

【作者基本信息】 兰州大学 ， 大气科学·气象学， 2017， 硕士

【摘要】 暴雨天气预报的检验和评估非常有必要,需要建立一套自上而下完整的预报检验业务流程,制定科学可行、操作性强的质量检验评估办法。为了满足社会公众对暴雨预报准确率的需求,完善暴雨预报准确率检验方法,本文分析了我国现行业务执行的暴雨预报检验方法,参考借鉴“模糊检验”思路,采用落区邻域对比原则和梯级递减评分原则,对2013-2014年中央气象台降水落区预报产品进行评分试验,设计了新的暴雨公众预报准确率检验的评分方案。同时引入指数衰减函数和距离权重系数进行敏感性分析,对该方案进行科学合理优化。分析结果表明:(1)对我国现行业务执行的暴雨预报检验方法进行了对比分析,发现如果采用严格“点对点”的TS评分方法,北方地区会出现较多的空报和漏报,使TS评分相对较低。暴雨预警信号检验方法中,相对于严格“点对点”一一对应的TS检验方法,暴雨预警信号采用一定程度的“模糊检验”,扩大了实况站点选取范围,避免了预报偏差较大造成的低评分或0评分,给出了有效评分,扩展了预报检验范围。(2)采用梯级递减原则,设计了五种不同的评分方案进行试验,选取不同的领域半径范围进行对比分析。制定了逐站暴雨预报梯级递减评分表,设计了新的检验规则。(3)梯级递减原则综合考虑了预报与实况量级的对应情况,实况出现相差一个量级时也可得分;落区邻域对比原则考虑了预报位置出现偏差的情况,实况出现的位置有偏差时也可得分。相对于TS评分,本方案可使暴雨公众预报准确率提高到60分以上。产生的预报效果对于社会公众是基本准确的,是可以接受的,预报评分相对会较高,这对预报员做好暴雨预警、提高预警准确率非常有利。(4)引入指数衰减函数对梯级递减评分进行了合理优化,避免了不同的量级得分出现断层和突变。衰减常数σ2越大,函数的衰减程度就越小,评分就越大。指数衰减函数的引入,使得预报与实况越接近,评分就会越高,同时不同量级预报和实况得分会有较好的连续性。(5)引入距离权重系数对落区邻域对比方法进行了合理优化。实况站点离预报检验站点越近,距离权重系数就会越高,使此站的得分相应提高。距离权重系数的引入,避免了在30公里邻域范围内,距离较远实况站的得分也会采信成为预报站点较高得分,提高了评分的合理性。(6)得出了优化后的暴雨公众预报准确率评分公式,为预报员和管理决策者提供了新的评分策略和思路。新的检验方法是对我国现有暴雨预报检验办法的补充、发展、丰富和完善,对于提高检验结果对于预报技术的改进、暴雨预报准确率和预警能力的提高亦有积极作用。更多还原

【Abstract】 It is necessary to test and evaluate the rainstorm forecast,and it is need to establish a top-down integrity of the forecast inspection business processes,to formulate a scientific and feasible method for quality inspection and evaluation.In order to satisfy the needs of the public for the accuracy in rainstorm forecast,and improve the method of calibration,in this paper,the method of rainstorm forecast test in our country is analyzed,and reference to the idea of fuzzy test,based on the principle of falling area neighborhood contrast and the principle of descending grade,the central meteorological station precipitation forecast products from 2013 to 2014 is tested.A new scheme for checking the accuracy in public rainstorm forecast is designed.The exponential attenuation function and the distance weight coefficient are introduced to carry out sensitivity analysis.The results show that:(1)The method of rainstorm forecast test in our country is analyzed,the empty and missing will be more frequently appear in northern area if using strict “point to point” method of TS score.The result is TS score is relatively low.Relative to the "point to point" one-to-one correspondence TS test method,the method of rainstorm warning signal using a certain degree of fuzzy test.The site selection is extended.The low score or 0 score caused by the larger deviation of prediction is avoided.Effective scoring is given,and the scope of the forecast test is expanded.(2)Five different scoring schemes were designed and tested by the principle of descending gradient,and different field radius range is analyzed.The descending gradient rainstorm scoring form is put forward,and a new inspection rule is designed.(3)The principle of descending gradient takes into account the correspondence between the forecast and the actual magnitude.It can also have score if there is an order of magnitude difference.The comparison principle of the falling zone neighborhood considers the deviation of the predicted position.It can also have score if there is a deviation of the actual occurrence.Compared with the TS score,this scheme can improve the accuracy of the rainstorm forecast to more than 60 points.The forecast effect is accurate and acceptable to the public,the forecast score will be relatively high,which is good for forecasters to do a good job of rainstorm warning,and it is favorable to improve the accuracy of early warning.(4)The exponential decay function is used to optimize the principle of descending gradient,which avoids the occurrence of fault and mutation in different order of magnitude.The greater the decay constant,the smaller the attenuation degree of the function,the larger the score will be.With the introduction of exponential decay function,the closer the prediction and the actual situation,the higher the score will be.At the same time,there is a good continuity in the different orders of magnitude prediction and the actual results.(5)The distance weight coefficient is introduced to optimize the comparison method of the falling area.The closer to the site,the higher the weight coefficient is,and it makes the score of this station increase accordingly.The introduction of the distance weighting coefficient can avoid the fact that the score of the remote station in the distance of 30 km can also be regarded as the higher score of the forecast site,which improves the rationality of the score.(6)The optimized formula of the public rainstorm forecast accuracy is obtained,which provides a new scoring strategy for the forecaster and the manager.The new scheme method is complementary to our existing rainstorm forecast verification scheme,it has a positive effect on the improvement of testing results and improved forecasting technology,it also has a positive effect on the improvement of rainstorm forecast accuracy and early warning ability.更多还原