【作者】 龚志强；

【导师】 丑纪范； 封国林；

【作者基本信息】 兰州大学 ， 气象学， 2009， 博士

【摘要】 本文结合NCEP/NCAR再分析资料和国家气候中心常规观测站日降水资料,运用关联矩阵分析方法,基于复杂网络理论构建温度场和环流系统的复杂网络。针对(1)传统的遥相关型研究仅考虑负相关,未考虑正相关,且多集中研究某一遥相关型,不能从整体上揭示了气候系统的空间结构特征;(2)遥相关指数主要通过SST或SLP的差值等进行定义,能够体现遥相关型信号的强弱,但不能体现遥相关型对气候系统作用的强弱;(3)对遥相关型的年代际配置特征及其与中国气候变化联系的研究相对较少等科学问题做了初步探讨。得到了以下一些主要结论:(1)概率统计角度揭示温度场内部的关联性及其时空演化特征。全球格点温度序列间既存在关联“噪声”,又存在真实关联;格点温度序列之间存在的关联既包括近邻或次近邻格点之间的短程关联,又包括如太平洋年代际涛动等远程格点之间的遥相关。关联系数的均值在1977年前后表现为两种不同变化态,对应了20世纪70年代中后期的温度突变;关联系数的空间分布在两种尺度下都表现为沿纬向呈带状分布,赤道太平洋、赤道印度洋和赤道大西洋及北太平洋是关联异常区域。关联性的研究为温度关联网络的构建奠定了理论基础。(2)温度场和环流系统内部关联作用的空间结构特征。构建了温度场和环流系统的关联网络,结构特征量的引入为气候系统空间结构复杂性提供了衡量指标。两种系统都具有典型的小世界效应,即系统中大量短程关联的存在保证了系统的结构稳定性,部分遥相关的存在提高了系统内部信息传递的效率。网络的顶点度分布都表明低纬度和中高纬度系统的空间结构存在显著差异,环流系统关联网络则明确界定了两种系统的分界为南北纬27.5°附近。低纬度环流子系统顶点度较高且空间分布一致性较好,主要体现了全局耦合的性质,格点间主要以直接关联作用为主;中高纬度环流子系统顶点度相对较小,遥相关型对应的局部区域顶点度较高,具有典型的小世界效应。换而言之,低纬度和中高纬度系统内格点间的作用机制存在很大差异。(3)遥相关型对环流系统空间结构的影响。建立气候系统空间结构复杂性衡量指标的基础上,研究了各种遥相关型对系统结构影响的主次关系。北半球中高纬度系统中,滤除AO和NAO后,系统的短程连边和长程连边均显著减少,网络的平均路径长度显著增加,网络信息传递的效率显著减小。相比较而言AO的作用较NAO更大一下。PNA在北半球环流系统中是一种相对独立的型,其作用主要体现了中高纬度和低纬度之间的联系。对于南半球中高纬度系统而言,AAO的作用最强,滤除AAO网络的平均顶点度减少了45%,平均道路长则增加了34%,因此在南半球环流系统中AAO是要首要关注的对象;南极偶极子在200hPA和800hPa高度场中均存在,因此对应了一种相对稳定的遥相关型,它的存在对AAO及南半球中高纬度环流系统均有一定的影响。(4)赤道中东太平洋海表温度和遥相关型作用强弱指标体系的建立。网络的顶点度直接体现顶点作用的强弱,因此采用顶点度区域平均的方法,定义了赤道中东太平洋温度关联指数(TPTI指数)和五种遥相关型作用指数,从而构建了一组新的指标体系。TPTI指数较以往通过SST距平值等的定义更深入的体现了该区域对全球气候变化作用的强弱。TPTI指数偏强的年份基本对应了强El nino年和La nina年,并将El nino年和La nina年分为强作用年和弱作用年,进一步说明气候系统中关键区域的信号强和作用强是未必等价的。春季和夏季TPTI指数与中国夏季降水EOF第一模态的相关系数分别为-0.34和-0.35,具有应用于夏季降水预测的潜在价值。五种遥相关型作用指数直接体现其在气候系统中作用的变化,并且根据各种指数的强弱变化确定中国夏季降水各个区域的主要相关因子。(5)遥相关型的年代际配置特征。PNA,NAO,AO,EUPA和WP五种显著的遥相关型年代际尺度的配置主要存在以下几个特征:(1)作用中心的移动,NAO,EUPA和WP的作用中心在年代际尺度存在显著的移动的特征,且70年代中后期及80年代末期90年代初存在共同的跳跃过程。AO型在1976年以前作用较弱,80年代末期以来作用则显著增强;PNA型模态的空间分布则相对稳定,作用中心移动的特征较弱,这也进一步体现了其在北半球中高纬度系统中是一种相对独立的模态。(2)根据各种指数确定了不同时段北半球中高纬度起主导作用的主要遥相关型:1976年以前是PNA,EUPA和WP三种模态的配置为主导,1977年以后NAO的作用显著增强,五种模态共同作用,1985年以后AO的作用显著增强,其它模态的作用则相对较弱。70年代中期及90年代初期的气候突变可能与行星际尺度遥相关型的调整有联系。(6)多因子物理统计模型的构建及其在夏季降水预测中的应用。从年代际尺度的角度分析了各种遥相关型指数与中国夏季降水的可能联系,将中国夏季降水分成10个不同区域,分别确定其主要的可能影响因子。从多因子配合作用的角度构建了夏季降水的多因子物理统计预测模型,基于模型,多因子结合,年代际尺度推导年际变率,通过历史经验信息判断相似等手段结合,提出了一种新的夏季降水预测的新方法,2005年夏季降水的独立样本检验表明该方法具有应用于中国夏季降水预测的潜在可能。更多还原

【Abstract】 Temperature and height correlation networks are constructed based on NCEP/NCAR reanalysis data,using correlation matrix analysis methods and complex networks theory.Face problems in climate system complexity:(1)only negative correlations are considered and focus on just one or two teleconnections in traditional research,which can’t give a whole view on the spatial characteristics of climate system;(2)Difference of SLP in different places is used in the definition of teleconnection index in traditional way,which can’t show the acting extent of teleconnection to the climate system;(3)Few works on decadal cooperation of teleconnections and their correlations with climate changes in China,we make some primary discussion.Main conclusions are as following:(1)Correlations of global temperature and its temporal and spatial characteristics is analyzed based in probability statistical way.Noise and true correlations both existed in global temperature field,which contains short and long distance correlations i.e.Pacific decadal oscillations(PDO).Mean correlation coefficients show different status near 1977,which correspond to the temperature abrupt change in 1970s.Zonal spatial distribution of correlations coefficients are presented both in 1d and 30d scales,and coefficients in tropical Pacific,Indian ocean, Atlantic and north Pacific are all abnormal correlation places.Correlation research in global temperature field established the foundations of the construction temperature correlation networks.(2)Characteristics of structure of spatial correlations in temperature and circumfluence systems.Correlation networks of temperature and circumfluence is constructed,both of these two kinds of network shows small world effect,that is to say large number of short distance connections ensure the structure stability of system, a few number of teleconnections improve the information transfer efficiency.Degree distribution differences show that spatial structure of middle and low latitude systems are quite different,and the exact boundary of circumfluence is at 27.5N and 27.5S. Average degree of low latitude circumfluence network is comparatively larger and synchronization is better than that of middle latitude,so spatial characteristics of low latitude circumfluence network presents the property of global coupling networks. Furthermore,Average degree of middle latitude circumfluence network is comparative smaller;however degree of some area is rather lager,which show that middle latitude network have typical small world property.(3)Definition of tropical pacific temperature correlation index(TPTI) and teleconnection action index(TAI).Degree denotes the action importance of point in the network,so TPTI and five kinds of TAI is defined by calculating the area average degree,and then a new index system for teleconnections and tropical pacific temperature is established.Obviously,physical mean of TPTI is better than that of traditional abnormal Sea Surface Temperature(ASST).Besides,Lager TPTI often correspond to the robust El nino or La nina signal which indicate that it is not equal between robust signal and action for the key area in the climate system.TPTI applied in the research of summer precipitation in China,results show that correlation coefficients between summer/spring TPTI and the temporal series of first mode of summer precipitation is-0.35 and-0.34,so TPTI might be used in prediction of China summer precipitation.Five kinds of TAI show the corresponding teleconnections’ action in the climate system,which is also helpful in confirming the leading mode of middle latitude circumfluence in different decadal.(4)Teleconnections’ influence on the structure of circumfluence system.In the middle latitude circumfluence system of north hemisphere,short and long distance connection all decreased by filtering out AO and NAO,and the average path length distinctly increased,and the influence of AO is comparatively lager than that of NAO. While PNA is rather a separated teleconnection in the system,its action most denotes the correlations between middle and low latitude system.In the middle latitude circumfluence system of south hemisphere,AAO is the most important mode, filtering out AAO the average degree of middle latitude circumfluence network of south hemisphere decreased 45%,and the average length increased 34%.We find a obvious dipole in Antarctic Arctic(AAD) with 500hpa height data,which is also existed in 200hpa and 800hpa.AAD might correspond to a new teleconnection haven’t been paid much attention,however it has definitely influence on AAO and the middle latitude circumfluence system of south hemisphere.(5)Research on characteristics of decadal leading teleconnections.Decadal characteristics for PNA,NAO,AO,EUPA and WP might be presented from two sides: (1)moving of action centre,action centre of NAO,EUPA and WP all moved during different periods in last 40 years and there was a common abrupt change process during 1970s;while the mode of PNA is comparatively stable and moving of action centre is also quite small,which further indicate that PNA is a separated teleconnection in middle latitude system in north hemisphere.(2)Leading modes of teleconnection is confirmed based on the TAI:the leading mode is PNA,EUPA and WP before 1976;action of NAO notably increased and all the five teleconnections are comparatively robust after 1977;action of AO notably increased and other modes are comparatively weaker after 1985.(6)The construction of physical statistical and factors coupling model and its application to China summer precipitation(CSP).Correlation between TAI and CSP base on the decadal scale,and based on the influence factors CSP been divided into 10 characteristic areas.Furthermore,physical and statistical model of CSP been established based on the factors coupling and a new method for CSP prediction is put forward.Factors coupling,using decadal data to estimate annual data,history similar information et al.is considered in this prediction method,numerical test show that this way can efficiently used in CSP prediction.更多还原