中文摘要：

提出一种基于SVD的迭代对气象场序列缺测记录插补延长的技术方法,对长江流域20个测站1月份气温做插补试验,平均均方误差为0.25,插补精度明显优于迭代EOF,插补效果良好且性能稳定;而且插补站数所占比例越小效果越好.此研究表明,基于SVD迭代的插补方法是一种非常有效的插补途径.

英文摘要：

A new method of interpolating the missing meteorology data based on the singular value decomposition （SVD） iteration is designed in this paper. On the basis of SVD mathematics technique, making the former computation results being replaced to the next computation via the singular value decomposition iteration calculation, the precision of computation results increases little by little. In the course of interpolating missing meteorology data based on the singular value decomposition iteration, factors of missing meteorology data are introduced into calculation equation, so the missing meteorology data can be interpolated and the calculation precision increases gradually through iteration calculation. Consequently, interpolation of the missing data is realized. Interpolating experiments are done by SVD iteration method, （1） Based on the data of January mean air temperature at 48 stations during the period from 1955 to 2000 in the Yangtze River Basin, on the assumption that the data at 20 stations from 1955 to 1964 have been missed, interpolating experiments are done for the missing data of 20 stations. The results show that when iteration precision ε is 0. 05 and truncated rank K is 4, interpolation precisions are the best, and the average mean square error is only 0. 25. （2） On the same condition, i. e. , missing data of 20 stations and the same factors mentioned above, interpolating experiments are done using the method of EOF iteration, and the average mean square error is 0. 39. Compared experimentations show that interpolative precision by the singular value decomposition iteration method is better than that by EOF iteration method with good result and stable capacity. （3） On the other hand, other interpolation experimentations are done on the assumption that there are 10 stations＇ and 5 stations＇ meteorological data of 48 stations from 1955 to 1964 missed in the Yangtze River Basin. The interpolation results show that when iteration precision ε is still 0. 05 and truncated r