摘要: |
对上海最高气温1924~2001年期间的缺测资料进行了恢复试验,以期通过该个例的研究,探讨我国台站历史资料集中缺测或中断序列的恢复、插补的一般方法。选取使用上海气象观测站的相邻若干台站的最高气温资料,分别用回归分析法、判别分析法和偏最小二乘法,对上海站相应期间缺测资料进行逐年逐月的恢复,并利用交叉检验的均方误差作为恢复效果的评判。结果发现对缺测资料恢复的总效果以偏最小二乘法为最好,因此选用该方法作为资料恢复的主要方法。恢复过程为每次仅恢复某月份、某测站、某年份的缺测资料,恢复后的估计值作为观测值使用,再进行其它缺测资料恢复的新资料集,如此反复地进行,以近期资料开始,逐步向前期资料延拓。这样做的目的是为了使得和后续的观测资料直接连接。 |
关键词: 资料恢复 均一性 判别分析 回归分析 偏最小二乘回归 |
DOI: |
分类号: |
基金项目:国家自然科学基金课题(40605021);国家科技基础性条件平台工作(2005DKA31700-01)共同资助 |
|
EXPERIMENTAL STUDY ON RECONSTRUCTION OF MAXIMUM TEMPERATURE DATA IN SHANGHAI |
LI Qing-xiang1, HUANG Jia-you2, JU Xiao-hui1
|
1.National Meteorological Information Center, Beijing 100871, China;2.Department of Atmospheric Science, School of Physics, Peking University, Beijing 100871, China
|
Abstract: |
Experimental study on reconstruction of maximum temperature data in shanghai has been presented in this paper. Based on the data from neighboring stations, monthly missing data are reconstructed by the following methods: regression analysis, discrimination analysis and partial least square regression. The last method is proved the most effective of the three. The reconstruction process is described as the following. The missing value of a certain month, in a certain year, of a certain station was reconstructed firstly each time, and then the reconstructed value was taken as observational value in reconstructing of the other missing values in the dataset. Through the repetition of the above processes, the preceding missing data were reconstructed based on the data for the adjacent period. |
Key words: data reconstruction homogeneity discrimination analysis regression analysis partial least square regression |