【作者】 支蓉；

【导师】 丑纪范； 封国林；

【作者基本信息】 兰州大学 ， 气象学， 2010， 博士

【摘要】 本文基于NCEP/NCAR和ERA40再分析资料构建全球温度场和高度场关联矩阵,并以此为基础探讨气象要素场关联性的整体特征,主要结论如下：(1)分别构建温度关联矩阵和随机关联矩阵并进行比较,发现温度关联矩阵中既存在关联“噪声”又存在真实关联。采用信度为0.01的统计检验能够简单有效地滤除温度关联矩阵中的关联“噪声”。对日温度序列做step=5d,10d...的滑动平均继而构建关联矩阵发现：365-730d可能是温度场时间尺度上的一个转折点；全球平均关联系数CTglobal的总体变化趋势是随step的增加而增大,但趋势逐渐变缓。位于赤道东太平洋海域的正关联中心(区域Ⅰ)(162.5°E-102.5°W,7.5°N-12.5°S)和位于北太平洋海域的负关联中心(区域Ⅱ)(157.5°E-147.5°W,27.5°-47.5°N)是值得重点关注的两个区域,二者之间存在强负相关。取长度为10a滑动窗口研究CTglobal随时间的演变特征发现：CTglobal在1981-1987年之间发生了明显的跃变；区域Ⅰ和Ⅱ对这一跃变也有清晰的反映,区域Ⅱ的跃变略滞后于区域Ⅰ；此外,北太平洋海域负关联中心随时间演变还出现了明显的“位移”,据此提取出北太平洋海域负关联中心最强负关联值,定义为动态北太平洋指数(Moving North Pacific Index, MNPI)。(2)构建考虑时间延迟情况下的关联矩阵,分析时间延迟对全球温度场关联性时空特征的影响。结果表明：随着延迟时间的增加,全球温度场的关联性总体逐渐减弱,但对应不同的延迟时间其规律也不同,延迟1-30d的情况下,可根据CTglobal的下降快慢大体将其划分为3段：延迟1-7d左右、8-20d左右、21-30d左右。温度场关联系数的空间分布型没有随延迟时间的增加发生明显变化,但在数值上的变化总体呈沿纬向的带状分布,其中北半球中纬度的亚洲大陆大部以及赤道中东太平洋地区分别与同纬度的其他地区变化趋势相反。区域Ⅰ和Ⅱ的平均关联系数随延迟时间的演变规律与全球平均状况相比更加复杂。(3)讨论增/降温趋势和极端温度事件对全球温度场关联性的影响,由NCEP/NCAR再分析资料得到的结果表明：去除增/降温趋势之后,原本呈现降温趋势的地区关联性有所增强,而原本增温趋势较显著的区域与关联性减弱较显著的区域也有较好的对应关系。但由ERA40再分析资料得到的结论则并没有这种明显的对应关系。总体而言,两套资料的结论都显示增/降温趋势对全球温度场关联性的空间分布型没有太大影响。采用不同的极端温度值替代方案造成全球平均关联系数随时间的演变在90年代之后发生较大差异,究其原因,可能与不同方案下温度场关联系数的南、北半球空间分布在90年代前后发生较大改变有关。(4)分别基于本文定义的动态北太平洋指数和Trenberth等定义的北太平洋指数,针对长江中下游区域夏季汛期降水模式预报误差寻找降水预测误差场的4个相似年份,将相似年误差场的算术平均作为模式预测结果的误差场进行预测。结果表明,基于动态北太平洋指数选取相似误差场进行模式预测误差订正比利用北太平洋指数效果优越。在此基础上结合欧氏距离加权平均方法能进一步改善预报效果,提高预报技巧。2003-2009年共7年的夏季汛期降水独立样本回报结果表明,动态北太平洋指数对长江中下游夏季汛期降水的具有很高的预测技巧,尤其具有年代际变化特征,与北太平洋海温的演变特征相一致。最后利用动态北太平洋指数制作了2009年中国夏季汛期降水预测,经检验达到了很高的预报技巧。(5)基于200hPa、500hPa、700hPa高度场和地面气压场全球再分析资料构建关联矩阵。结果表明各层次高度场的关联性在中低纬度区域较好并向高纬度地区递减,呈准带状分布；垂直方向上,低层的关联性较弱,随着高度增加关联性逐渐增强；北太平洋区域从低层至高层均存在一个较强的负关联中心,体现出一定的特殊性。考虑时间延迟的情况下,随着延迟天数的增加关联性逐渐减弱,延迟天数达到15天左右其衰减趋势趋于平缓,达到60天左右发生显著转折。从空间角度来看,环流系统内部关联的衰减速度表现出低纬向中高纬加快,高层向低层加快的特征。各层次高度场和地面气压场全球平均关联系数随时间的变化特征基本保持一致,在1978-1982年和1996-1998年左右发生了两次明显的跃变。(6)进一步研究北太平洋负关联中心由低层到高层的空间分布特征及其与全球其他强关联中心之间的关系,结果表明：各层次高度场中与北太平洋区域之间存在显著关联的地区比较一致,主要分布在三个区域：赤道中太平洋(中心A)和白令海峡(中心B)(负关联)、北美大陆东南部(中心C)(正关联)。北太平洋区域格点与这三个区域格点间关联性的累加效果使得各层次高度场的北太平洋区域都存在一个明显的负值中心,该负关联中心从低层到高层存在先南后北、先西后东的移动过程,且方向的转变都发生在500hPa高度场。地面气压场中,中心C对北太平洋区域的影响明显弱于其他两个中心,而在700hPa、500hPa和200hPa高度场都是中心A的作用最强。更多还原

【Abstract】 Based on temperature and height field correlation matrixes constructed by NCEP/NCAR and ERA40 reanalysis data, we study the correlation characteristics of global meteorological elements. Main conclusions are as follows:(1) By comparison of temperature and random correlation matrixes, we find that noise and true correlations both exist in global temperature field, and the 99% confidence test is used in order to filter correlation’noise’. We construct correlation matrixes based on moving average of temperature series and find that 365-730d maybe a turning of time scale of temperature field. Global average correlation coefficient CTglobal increase along with time scale, but the rate is slower and slower. The positive correlation center (162.5°E-102.5°W,7.5°N-12.5°S) (AreaⅠ) and the negative correlation center (157.5°E-147.5°W,27.5°-47.5°N) (AreaⅡ) are the two areas we concern on, there is strong negative correlation between them. With the help of a 10a slipping window, we find that CTglobal had an abrupt change between 1981-1987a, so did AreaⅠandⅡ. And the position of AreaⅡchange with time. We defined a new index named Moving North Pacific Index (MNPI) based on the value of the strongest negative correlation of Area II.(2) We construct matrixes in case of considering time delay, the results show that: with the delay time increasing, CTglobal weakened. However, different delay time corresponding to different rule, take 1-30d for instance, generally based on decreasing speed can be divided into three sections:1-7d,8-20d and 21-30d. The spatial distribution pattern of temperature coefficient did not significantly changed with the increasing delay time, but the overall distribution of change in value was the strip along the latitudinal, and most of the Asian continent and the eastern equatorial Pacific are reverse to the trend of other areas in the same latitude.(Area I and II show more complex evolution rules than the average state.(3) We discussion the effection of warming/cooling trends of temperature and extreme temperature events on the temperature field correlation matrixes, the results show that:based on the NCEP/NCAR reanalysis data, by removing the warming/ cooling trends, regions that originally show cooling trends have enhanced regional correlation, and regions that originally show warming trends have weakened regional correlation, but by the ERA40 reanalysis data, there is no such correspondence. Overall, the conclusions of the two sets of data are shown that warming/cooling trend of temperature can not significantly affected the spatial distribution of the correlation of global temperature field. Different alternative methods of the extreme temperature make quite different reasons after 1990s, and it may be associated with the great change of correlation distribution of the southern and northern hemisphere before and after 1990s.(4) Aim at model prediction error of summer precipitation of Yangtze River region, we use the MNPI we defined and the NPI defined Trenberth et al. to find the 4 similar years of precipitation forecast error, and make the arithmetic average of them as the model prediction error. The results showed that MNPI perform better than NPI. On this basis, combined with the Euclidean distance weighted average method can further improve the forecast skill. The results of 2003-2009 summer precipitation independent sample return prediction show that MNPI perform good on prediction of summer precipitation of Yangtze River region. MNPI shows interdecadal variations, consistent with the evolution characteristics of North Pacific sea surface temperature. Finally, we give the forecast of summer precipitation in China in 2009 with MNPI, and reached a very high forecast skill.(5) Correlation matrixes are constructed with NCEP/NCAR global height field data and surface pressure data, furthermore, the characteristics of these two matrixes are analyzed. Research results shows that correlations at all levels of height field are better in middle-low latitudes and descend to the high latitudes, presenting characteristics of quasi-zonal distribution. In the vertical direction, correlations of low levels are weaker and become more and more stronger with the increase of height. There is always a negative correlation center in the north pacific at all levels of height, reflecting a certain degree of particularity. Results of delayed correlation shows that decreased with the increasing of delay time, the tendency became gently when the delay tine reach 15d and there is an obvious turning point when the delay tine reach 60d. From the angle of space, the decay rate of correlation increase with latitude and the decrease of height. Changes over time of CT global in three height fields and surface pressure are almost the same, two abrupt changes occurred during 1978-1982 and 1996-1998.(6) Further research on the spatial distribution of the North Pacific negative correlation center showes that:the strong correlation centers associated with the North Pacific negative correlation center consistent from low-to-high levels, mainly in three regions:the central equatorial Pacific (center A) and the Bering Strait (center B), and southeastern North American continent (center C). The cumulative effect of correlation among North Pacific and the other three centers are as follows:we can get clear negative centers in the North Pacific region on all levels. The negative center moves from low to high by the following rules:first eastward, then westward; first moved south, moving north after; and changes have occurred both on the 500hpa height field. On surface pressure field, the correlation between the center C and the North Pacific region is the weakest, on the other levers, the most significant center correlated to the North Pacific region is the center A.更多还原