【作者】 王伟；

【导师】 李洪平；

【作者基本信息】 中国海洋大学 ， 地图学与地理信息系统， 2012， 硕士

【摘要】 在全球变暖日趋加剧的今天，陆地的气温变化一直是大气学研究的一个重点。由于气温的变化原因复杂多样，不同区域的气温变化原因不尽相同，所以气温变化的可预测性研究和影响区域气温敏感因子的变化特征以及两者的相关性一直是该领域研究的一个重点和难点。青岛气温影响因素多种多样，与全球海表温度是否存在可作为预测依据的相关性，至今没有得出任何肯定的结论。研究发现，海气相互作用对青岛气温系统的影响绝不仅仅局限于厄尔尼诺区域。青岛气温异常与海表温度异常的关系也不仅仅局限于太平洋或其它某一特定海域。因此，十分有必要从全球范围的海表温度异常对青岛气温可能产生的影响作进一步的诊断研究。本文利用国家160站的青岛气温资料和ERSST海表温度资料对青岛气温异常与全球海表温度异常进行分析，采用奇异谱算法研究了青岛气温和全球海表温度的趋势和周期性变化，并利用AR模型算法对青岛气温异常进行预测。研究发现：在近30年青岛平均气温显著提高，利用奇异谱主成分分析法重构气温异常序列，拟合率达到0.7497，经分析得出青岛气温异常存在6年半、2年、1年和3、4个月的主要周期；利用3个不同长度的时间序列采用AR模型实验预测了2009年1月-2011年12月的青岛气温异常，实况序列与重构序列的最高相关系数达到0.9453，收到了很好的预测效果；并利用此种方法将未来3年的青岛气温异常进行了预测。通过研究全球海表温度的相关数据，发现全球平均海表温度在近61年呈波动上升趋势，利用奇异谱主成分分析法重构全球海表温度异常序列，拟合率达到0.9897，经分析得出青岛气温异常存在4年半、3年半、2年半、1年半、1年和9个月的主要周期；采用AR模型实验预测了2009年1月-2011年12月的全球海表温度异常，实况序列与重构序列的相关系数达到0.9306。同时，分不同的时间尺度（61年、30年、10年）研究青岛气温异常和全球海表温度的线性相关性和滑动相关性，并按自然时间序列和逐月序列两类时间尺度进行详细的分析研究。研究发现按自然时间序列的61年、30年和10年数据计算出来的线性相关和滑动相关都不足0.7，不足以作为预报青岛气温异常的参考因子，而按61年、30年和10年逐月时间序列得到的相关系数显著提高，30年数据的结果更佳，最大相关系数已达0.7880，与青岛气温异常存在很大的相关度，该点为8N、49W海域，滑动区间为17，其他月份也有大于0.7的结果生成。本文采用奇异谱主成分分析法和AR模型对青岛气温异常进行了预测，收到了很好的预测效果；采用线性相关和滑动相关两种方法，通过分不同时间尺度（61年、30年、10年）分不同时间排列顺序（自然时间顺序和逐月时间顺序）分别研究青岛气温异常与全球海表温度异常的相关性，其所得结果可以作为预报青岛气温异常的参考因子，为青岛气温异常的预报提供更加精确的参考。利用本文所得到的结论，分别具体分析相关海域对青岛气温异常的影响都将在以后的学习中做进一步研究。更多还原

【Abstract】 With the aggravating of global warming, the land temperature variation is always ahot topic in the field of atmospheric research. There are different reasons according totemperature variation in different regions because they are complex and various. So, it isalways a important and difficult point in this field to find out the correlation between thetemperature variation predictability and the change features of the sensitive factors whichinfluence the regional temperature.The influencing factors of Qingdao city temperature are very various. Until now thereare no explicit results about the correlation between the global SST (sea surfacetemperature) and the land temperature which can support the forecast of Qingdao citytemperature. The impact of sea-air interaction to the Qingdao city temperature is not onlylimited to the El Nino region based on previous researches. The relationship between airtemperature anomalies and SST anomalies is not also limited to the Pacific or other thespecific sea areas. Therefore, it is necessary to make further research about influencingfactors of Qingdao city temperature from the worldwide SST anomalies.In this dissertation, historical data of Qingdao city temperature from160nationalstations and ERSST data is used to analyze the trend and cyclical changes of Qingdao citytemperature anomalies and global SST anomalies. The methods of SSA (singularspectrum analysis) and AR model are used to predict the Qingdao city temperature.Research found that the Qingdao city average temperature is significantly enhanced innearly30years, the series of temperature anomaly are reconstituted with the method of theSSA principal component analysis that the coincidence rate has reached0.7497, andreceive the main cycles of Qingdao city temperature anomaly are6.5years,2years,1yearand3or4months. The time series of three different lengths is used to forecast theQingdao city temperature anomaly with the method of AR model that the maximumcorrelation coefficient between live series and reconstitute series has reached0.9453which received a good prediction effect, and forecast the Qingdao city temperature anomaly of next three years with this method. Meanwhile, Research found that theaverage global SST appears rising trend in nearly61years through the research of theglobal SST data, the series of global SST anomaly are reconstituted with the method of theSSA principal component analysis that the coincidence rate has reached0.9897, andreceive the cycles of global SST anomaly are4.5years,2.5years,1.5years,1year and9months; The time series of three different lengths is used to forecast the global SSTanomaly with the method of AR model that the maximum correlation coefficient betweenlive series and reconstitute series has reached0.9306.Meanwhile, linear correlation and sliding correlation are used to find out thecorrelation of the Qingdao temperature anomalies and global SST anomaly according tothree kinds of time scales (61Years,30Years,10Years) and two kinds of time series(natural time series and monthly series). Research found that both of the linear correlationcoefficient and sliding correlation coefficient are not enough0.7according to the data ofnature time series (61Y,30Y,10Y) which is not enough to be a reference factor toforecast the Qingdao city temperature anomaly. The correlation coefficients according tothe monthly time series (61Y,30Y,10Y) have improved significantly, and the results of30years is better than the others, the maximum correlation coefficient has reached0.7880which prove there are a big correlation between them, the point in8N,49W and thesliding region is17, meanwhile, other region also have the correlation which above0.7.In this dissertation, two kinds of methods of SSA and AR model is used to forecastthe Qingdao city temperature anomaly which received a good prediction effect, andresearch the correlation of Qingdao temperature anomaly and global SST anomaly withthree kinds of time scales (61Y,30Y,10Y) and two kinds of time series (natural timeseries and monthly series), the results can be used as the reference factors to forecast theQingdao temperature anomaly, and provide more accurate reference for the Qingdaotemperature forecasting. The Concrete analysis about the influence on Qingdao citytemperature will be researched in the future study with the Conclusions from thisdissertation.更多还原