【作者】 刘琳；

【导师】 朱克云； 陈静；

【作者基本信息】 成都信息工程学院 ， 大气科学， 2013， 硕士

【摘要】 极端降水天气属于小概率事件，它的发生具有很多不确定的因素，预报难度很大。本文首先利用中国756个观测站点降水资料，采用百分位法定义了不同台站的极端降水阈值，进而分析全国八个区域近50年的极端降水特征。然后根据Anderson-Darling检验原理研究基于集合预报资料的极端强降水天气预报方法，建立基于集合预报与模式历史预报累积概率密度分布函数连续差异的数学模型——极端降水天气预报指数（Extreme Precipitation Forecast Index,简称EPFI），并对2011年7月中国极端强降水天气进行预报试验。最后根据EPFI的预报原理研究持续性强降水的预报方法，对江淮区域持续性强降水过程进行预报和评估，并与T213控制预报的预报技巧进行对比，主要结论如下：(1)近50年我国极端降水事件趋于增多，但是极端降水事件的时空变化特征存在明显的区域差异。东北和华北地区极端降水事件有减弱、减少趋势，西北西部、长江中下游、华南以及青藏高原这四个区域，极端降水事件有增多，增强趋势，西北东部极端降水值和极端降水量比例减少，但是极端降水频数和平均强度有所增加，西南地区变化不明显。北京和武汉的极端降水较区域平均的更强，变化更快。(2)模式预报的极端强降水分布特征与实况非常接近，但是模式极端强降水明显偏弱，并且随着预报时效的延长，极端强降水逐渐减小。模式极端强降水的大值中心也存在一定的偏差：广东沿海地区的大值中心偏向了江苏、浙江和福建一带；西藏南部则出现了虚假的大值中心。(3)极端降水天气预报指数可以充分利用集合降水累积概率密度分布的尾端信息，为极端强降水提供科学合理的预报，基于CMA T213集合预报的极端降水天气预报指数可提前3-7天发出极端强降水预警信号，随着预报时效的延长，预报技巧逐渐降低。(4)通过对两种模式气候累积分布方案试验的对比可知：模式气候累积概率包含6-8月模式历史资料的预报技巧高于仅包含7月模式历史预报资料，这是因为中国的雨季具有一定的连续性（主要分布在6-8月），6-8月模式历史资料包含的降水信息比7月模式历史资料更全面。所以气候累积概率分布要尽可能全面的反应气候的真实状况。(5)综合多种持续性强降水的定义，给出江淮区域发生极端强降水的评定标准，利用T213控制预报对持续性强降水的预报标准对5次持续性强降水的预报能力进行评估，结果表明T213控制预报能提前2-7天预报出持续性强降水，平均预报能力为5.4天。但是预报能力与具体过程相关，差别很大。预报能力与降水量有密切关系。过程降水量大，则预报能力强；过程降水量小，则预报能力弱。(6) EPFI对持续性强降水的预报能力明显高于控制预报，对5次过程的预报能力都有提高，尤其对较弱的持续性强降水过程的预报改进较多。对2011年6月1日-30日进行持续性强降水预报实验，四个检验指标的评估检验表明EPFI对极端强降水的预报技巧高于T213控制预报，但是也存在空报稍多的情况。更多还原

【Abstract】 As extreme heavy precipitation is a small probability event and occurs with agreat deal of uncertainties, it is very difficult for accurate prediction. On the basis ofdaily precipitation data of756stations in china, the extreme precipitation thresholdvalue were defined in different stations according to percentile method, variationcharacteristics of extreme precipitation events in different8areas of china werestudied.Then, according to the principle of the Anderson-Darling test, a method forextreme heavy precipitation forecast based on the ensemble forecast data wasstudied,and the mathematical model(Extreme precipitation Forecast Index) based onthe Continuous differences of the cumulative distribution functions (CDF) betweenmodel climate and the EPS forecast was established. Extreme precipitation ForecastIndex （EPFI）was applied to forecast experiment for the several extreme heavyprecipitation events happened during17-31July2011, the effect of forecast wasevaluated. Finally, the method for continuous heavy rain forecast was studied accordingto the principle of EPFI forecast research.Then, five continuous heavy rain processhapped in jianghuai area were assessed, and the results were compared with T213control forecast, the main conclusions are as follows:(1)The extreme precipitation frequency, intensity and extreme precipitation in the totalprecipitation show increase trends in the last50years. However, it is found that thetrend change of EPE showed remarkable regional differences. The extremeprecipitation frequency and intensity show decreased trends In Northeast China andnorth China. However, there is an opposite trend in western part of Northwestchina, the Mid-low Reaches of the Yangtze river south china and Tibet plateau. Theprecipitation extremes and extreme precipitation in the total precipitation showdecreased trend in eastern part of northwest china while the extreme precipitationfrequency and intensity increased, the extreme precipitation changed little insouthwest china.The extream precipitation of Beijing and Wuhan was stronger thanthe regional average ones.(2) The extreme heavy precipitation distribution of model is very close to the observational one, while the model extreme heavy precipitation is weaker,and theextreme heavy precipitation decreases with the extension of forecasting time. Therewas a certain deviation to the observational one: the high of extreme heavyprecipitation of Guangdong was partial to the Jiangsu, Zhejiang and Fujian;Southern Tibet had a false value center.(3) The result shows that EPFI can take advantage of the tail information of the CDFbased on the EPS, providing scientific and rational forecast for extreme heavyprecipitation. The EPFI based on CMA T213EPS can give early warning onextreme heavy precipitation3to7days in advance. With the forecast time extending,EPFI was less skillful.(4) By comparing the CDF of two model climate datasets, we find that the forecast skillof the CDF that contains historical model precipitation information from June toAugust is better than the CDF that contains only the July precipitation information.China’s rainy season has certain continuity (mainly distributed in June-August),Thus, it is better for the model climatic CDF to include objective precipitationinformation as much as possible.(5) The evaluation criteria of the regional extremely heavy rainfall in Jianghuai wasgiven by Comprehensiving variety of continuous heavy rain definition. the resultssuggest that T213control forecast can forecast continuous heavy rain the2-7days inadvance, the average forecast ability was5.4days. But forecasting ability isassociated with different processes. the forecast is skillful in high rainfall process,vice versa.(6) EPFI prediction ability is significantly higher than the control forecast, forecastability have improved the process of five times, especially for weaker in the processof continuous heavy rain forecast to improve more. On June1,2011-30, persistentheavy rain forecast experiments, four inspection index assessment shows that EPFIforecast skill is higher than T213control of extreme heavy precipitation forecast,but EPFI forecast has a higher false alarm rate.更多还原