Y为二级X为多级的相关量度研究

针对小概率事件中常采用的Ｙ取二值,Ｘ取多级的数据环境,采用将因子Ｘ的等级序列Ｘｉｔ转换成以条件概率为参数的序列（ｆ（Ｘｕ）的方法,导出一个与原因子序列（Ｘｕ）与预报序列Ｙｔ是成线性还是非线性相关无关的相关量度公式。它可为建立小概率事件预报模式提供更合理的可比性好的因子挑选准则。     

热带气旋登陆华东的客观预报方案

热带所旋登陆华东属小概率事件。构造适当划分区间的阶梯函数能较好地反映登陆事件条件概率的非线性分布。应用阶梯函数变换的因子对登陆事件概率的预报性能优于线性相关的标准化因子。预报方案应用阶梯函数变换因子，通过非线性相关比筛选和REEP分析构造热带气旋登陆华东的预报模式。应用初始场资料与数值预报产品分别构造了统计－天气学、统计－气候学和统计－动力学的分类预报模式。应用分类预报模式的预报结果进行综合的预报集成提高了预报技巧。预报检验与试验表明本方案对热带气旋登陆华东具有一定的预报能力。     

北京地面O3的集合预报试验

基于蒙特卡罗集合预报方法建立了一个北京地面臭氧的集合预报系统,该系统对包括排放源、气象场在内的154个模式输入和参数进行了扰动,共包含了50个集合成员。利用该集合预报系统模拟了北京奥运会期间一次臭氧过程(2008年8月11日～8月13日),分析了集合预报系统在概率预报,最优集合子集预报以及表征预报可靠性方面的特点。结果表明:与原确定性预报相比,集合预报能为用户提供更加丰富的预报信息,通过概率预报可以提供不同事件发生的概率,对臭氧事件的预报准确率更高。采用最优集合子集预报方式时的臭氧预报均方根误差比原确定性预报低了10%以上。     

动力-统计方法在24小时雷暴预报的应用

利用T213数值预报产品,计算了多个对流参数,应用事件概率回归方法分别建立了全国690个基本站4-9月的24h雷暴潜势预报方程,并根据TS评分值最大的原则确定了雷暴发生预报的临界概率,针对2010年4-9月进行试预报。研究结果表明：（1）基于T213输出产品计算得到的对流参数物理意义明确,与雷暴有无事件相关系数较高,对雷暴潜势预报方程贡献很大。（2）雷暴潜势预报方程对区域性雷暴的预报指示性较强,尤其对于雷电频发的地区效果更好。（3）690个站TS评分平均值为0．24。     

日较大降水的地方MOS预报程序

较大降水是一种小概率事件,它受局地天气变化的影响较大。用一般的统计方法不容易预报出来。我们首先利用日本气象传真输出的数值预告产品,对原始数据进行处理,用消零法消除样本中大部份非小概率事件的个例,从而使小概率事件的比重增大,再加上县站非模式输出因子和地面14时形势实况,建立了日较大降水的地方MOS预报程序。该预报程序拟合了我站79～81年5～6月166个样本中日降水量≥15毫米的全部个例13个(拟合率100％),试报了我站82年5～6月56个样本,4个≥15毫米的降水个例全部报出,无空报。     

一次短期集合预报试验

数值模式中物理过程描述不准确是造成数值预报误差的主要来源之一，采用数值模式中不同物理过程的集合预报方案是减小模式预报误差的有效方法。本文结合一个典型的夏季东北冷涡型降水天气过程，选用不同的积云对流参数化方案、显式降水方案、辐射方案以及模式采用不同的地表分辨率构成10集合预报成员，进行了48h的预报。结果显示集合预报在总体预报效果上比各个集合成员的预报效果好，简单的集合平均就可以提高模式形势场和降水要素场的预报准确率。在集合平均提高预报准确率的同时，采用天气系统移动路径图、面条图以及降水概率预报等方法，可以增强小概率事件的预报能力。     

基于模糊聚类分型的数值产品暴雨预报释用方法

本文阐述了一种基于数值产品的预报信息场的暴雨预报释用方法,提出了对暴雨历史样本形势场使用模糊聚类的客观分型技术,用条件概率让小概率事件成为相对的大概率,更有利于暴雨指标的提炼;该释用方法包含的信息量全面,以历史暴雨样本同期的包含大气三维空间压、温、湿、风及其释演的物理量,构建出25个数值产品参数指标作为鉴别暴雨的基础,同时也勾划出各暴雨型发生期大气高温、高湿、对流不稳定等机理结构特征,且均有具体物理量数据表述.预报方法的研究过程中探索了某些物理量场预报暴雨落区的规律;建立了暴雨预报工具,充分利用了现有业务中的质量较好的预报产品设立预报工具的起报条件,减少了暴雨空报率.     

500 hPa位势高度场极端天气事件的NCEP集合概率预报效果分析

基于百分位方法,首先通过构建气候等概率区间得到了位势高度场极端天气事件区分方法,并给出了相应的预报结果检验评价方案;以此为基础,对2003年夏季亚欧区域的500hPa位势高度场极端天气事件的NCEP集合概率预报效果进行了深入分析,得到如下结论：1）分析时段和区域内的极端事件的发生频率比气候平均状况略偏高。NCEP集合预报系统对异常偏低的极端事件的预报易偏高,但相对比较可靠;对相对异常偏高极端事件的预报频率随着预报时效的增大而减小,时效小于3d时预报频率偏高,时效大于5d时显著偏低,预报可靠性相对差一些。2）EV（economic value,经济价值）分析表明,NCEP集合预报系统对偏高（低）极端事件的概率预报效果明显高于正常天气事件。3）预报命中率分析表明,在高概率阈值下,预报命中率较低,这与预报系统在部分个例中的不发散有关。随着概率阈值的降低,预报命中率稳步提高,说明集合成员的发散使之可以覆盖实况可能出现的区间,从而得到更好的预报效果。     

基于对流参数的雷暴预报方法研究

通过对2006年6~9月(雨季)大量雷暴对流参数进行计算,选取离西昌发射场最近的单点同化资料。利用相关系数法选取相关性好且稳定的预报因子,进行事件概率回归,得到雷暴预报方程。并对2007年雨季西昌雷暴进行预报,取得了较好的效果。分析表明,基于对流参数的雷暴预报方法对场区雷暴的预报具有明显的效果。      

逐步消空法在上海雷电潜势预报中的应用

利用上海地区1999--2007年的探空资料、地面观测资料和历史天气图资料,从雷暴发生的形势场入手,分析上海地区雷暴主要类型.选取合理、适当的对流参数对于短时局地强雷雨天气的潜势预测、强度判别是有指示意义的.利用天气型、K指数、沙氏指数以及对流有效位能作为预报因子,采用逐步消空法找出雷暴识别指标集合,变小概率事件为条件概率下的大概率事件,进而提高雷电潜势预报准确率.     

两类天气预报评分问题研究及一种新评分方法

探讨了预报评价的意义及应遵循的原则, 对常用的几种两类预报评分方法进行分析, 指出其应用的局限性, 得到一个判定所作预报水平是否高于随机预报、具有预报技巧的简易判别式; 提出评分权重的概念, 指出以往评分存在问题的根源是评分权重分配不当, 使评分结果的真实性受到影响, 评分无可比性, 进而提出一种考虑了评分权重的新评分方法。新评分方法满足预报评价的原则, 侧重于对两类事件中事件概率较小一方预报效果的评估, 评分结果不受事件概率影响, 具有可比性。对比分析表明:新方法比其他方法优越, 能更准确地反映预报水平, 使不同季节、不同地域的预报评分可进行比较, 是一个通用的评分方法。     

Scenarios of extreme temperature events

Based on the principles of the probability theory a statistical model has been developed assessing the likelihood of occurrence of extreme temperature events from the knowledge of the statistical characteristics of the daily temperature extremes. It is demonstrated that the probability of such events is more sensitive to changes in the variability of climate than to changes in its average. Further, this sensitivity increases at a nonlinear rate the more extreme the event. The applicability of the model has been verified by comparing the simulated frequencies of a large spectrum of temperature events with the observed numbers derived from a long time series of daily temperature extremes at Potsdam. Accordingly, the relative simulation errors increase significantly as the events become more extreme. A correction is possible, because most of these errors are systematic rather than random. Moreover, in accordance with the climate observations the simulations reveal statistically significant linear trends in the number of extreme events since the end of the last century. Local scenarios of extreme temperature events have been derived for the city of Berlin by considering both hypothetical new climate states and climate changes simulated by a General Circulation Model (GCM). As a consequence of an increase in the atmospheric concentration of greenhouse gases up to the end of the next century according to the IPCC Scenario A the repetition rate of extreme events in summer (e.g., hot days) is expected to rise considerably relative to the current climate. Moreover, in the winter season cold days will become extremely rare.     

Conditions of generation of extremely high water content of Primorye rivers in autumn–winter 2012

Analyzed are the synoptic and hydrological conditions of the generation of extremely high water content of Primorye rivers in the autumn of 2012. Revealed are the general features of atmospheric circulation, precipitation, water content, and hydrological regime of rivers. Presented are quantitative parameters characterizing the extremity of observed events. The expert assessment of the probability of such combination of events demonstrates that its return period is about 500–1000 years if there is assumption on the stationarity of processes. Taking into account the revealed facts and available assessments of climate changes it can be supposed that the analyzed event rather indicates real changes in the hydrological regime of the region than represents a rare random phenomenon.     

Simulation of the probable maximum flood in the area of the Leningrad Nuclear Power Plant with account of wind waves

At the designing of nuclear power facilities at the coastal sites the risk of their flooding caused by the combinations of adverse hydrometeorological events should be assessed with the probability of exceedance to 0.01%. According to the IAEA recommendations, the combination of statistical and deterministic methods was used to calculate the flood level of such rare occurrence. The level of flooding caused by the storm surge and reiated wind waves were computed with the probability of 0.01% for the coastal part of the Koporye Bay of the Gulf of Finland in the area of the Leningrad Nuclear Power Plant 2 (LNPP 2) construction; the results are presented. The calculations are based on the CARDINAL and SWAN software and four nested numerical models (for the Baltic Sea, the eastern part of the Gulf of Finland, the Koporye Bay, and a part of the bay in the area of LNPP). The decrease in sea-surface drag coefficient at hurricane winds is taken into account.     

广义帕雷托分布在重庆暴雨强降水研究中的应用

引进广义帕雷托分布(GPD),借助于现代L-矩估计方法,模拟重庆地区极端降水事件,推算一定重现期的极端降水量分位数。模拟试验表明,基于超门限峰值法(POT)的GPD不但计算简便,而且基本不受原始序列样本量的影响,具有全部取值域的高精度稳定拟合(包括高端厚尾部),与GEV模拟结果相比,GPD具有更高精度和稳定性,更为实用。     

联合概率方法在北京灾害天气预报中的应用研究

基于ECMWF模式的集合预报数据,利用联合概率方法,针对北京地区冬季影响最大的寒潮和夏季强对流两类灾害性天气,建立了适用于本地区的两种集合预报业务产品。选取2 m温度和10 m平均风速制作寒潮预警信号联合概率预报产品,选取对流有效位能和0-6 km垂直风切变制作强对流潜势联合概率预报产品。通过对北京地区近年寒潮和强对流天气的预报检验表明：寒潮预警信号联合概率方法,当预报概率达到10%及以上时,实况就有可能达寒潮蓝色预警信号的级别;此方法对北京西北部的预报性能较好,其次为北京的东南部地区;对达到蓝色预警信号标准的区域具有较高的预报命中率,但对达黄色预警信号级别的区域,漏报率较高。强对流潜势联合概率方法的空报率较高,当预报概率达90%-100%时,实况才有可能出现强对流;与局地强对流相比,全市性强对流天气的高概率预报区域较为集中。     

气候变化影响下极端水文事件的多变量统计模型研究

以黄河流域太原气象站和淮河流域鲁台子水文站为研究对象,利用Copula函数构建气候要素（降水）同极端水文事件（干旱和洪水）之间的多元统计模型,分析不同降水条件下不同等级干旱和洪水的发生概率变化。结果表明,Gumbel Copula能够较好地描述太原站7月份的前期累加降水量和帕默尔干旱指数（PDSI）的相关结构。随着降水量的增加,极端干旱的发生概率逐渐减小,重旱、中旱和轻旱的发生概率则先增加后减小。Clayton Copula能够较好地描述鲁台子水文站前期累加降水量和洪峰流量之间的相关结构。当前期累加降水量大于等于某一定值时,随着年最大洪峰x的增大,发生洪峰≥x的极端洪水事件的概率逐渐减小。在同一个极端洪水发生概率下,前期累加降水量越大,洪峰流量出现大值的可能性越大。     

Changes in daily precipitation under enhanced greenhouse conditions

 An increase in global average precipitation of about 10% is simulated by two global climate models with mixed layer oceans in response to an equilibrium doubling of carbon dioxide. The UKHI model was developed in the United Kingdom at the Hadley Centre for Climate Prediction and Research and the CSIRO9 model was developed in Australia by the CSIRO Division of Atmospheric Research. Regional changes in daily precipitation simulated by these models have been compared. Both models simulate fewer wet days in middle latitudes, and more wet days in high latitudes. At middle and low latitudes, there is a shift in the precipitation type toward more intense convective events, and fewer moderate non-convective events. At high latitudes, the precipitation type remains non-convective and all events simply get heavier, resulting in fewer light events and more moderate and heavy events. The probability of heavy daily precipitation increases by more than 50% in many locations. Extreme events with a probability of 1% or less were considered in terms of return periods (the average period between events of the same magnitude). For a given return period of at least 1 y, precipitation intensity in Europe, USA, Australia and India increases by 10 to 25%. For a given precipitation intensity, the average return period becomes shorter by a factor of 2 to 5. Given that larger changes in frequency occur for heavier simulated events, changes may be even greater for more-extreme events not resolved by models. Received: 1 July 1996 / Accepted: 21 March 1997     

Breaking recent global temperature records

Global surface temperature was a record in 1988. What is the probability that this record will be surpassed in the next few years? Answers are provided given a variety of simple statistical models for temperature. The answers illustrate how record breaking is influenced by alternative model specifications. Estimates for the probability of a record are shown to range widely. If annual temperature is independent and identically distributed then a new record is unlikely. But probabilities increase rapidly if there is a trend or autocorrelation. Estimates of the probability of a record using data on global temperature suggest that another record in the next few years would not be a rare event.