用前期大气环流指数预测新疆北部夏季降水的探讨

Study on Predicting the Summer Precipitation in Northern Xinjiang with Previous Circulation Index

用1961—2007年新疆北部29站夏季(6—8月)降水资料,计算得到北疆夏季降水指数。以前期的月环流指数为因子,考虑相关系数的不稳定性,用滑动相关法初选因子,采用两级逐步回归的集合方案,尝试对北疆夏季降水指数进行预测。总样本数为47时,北疆夏季降水指数的拟合(预测)序列与原始序列之间的相关系数为0.7248。另外建立了10种不同总样本数的预测模型。经综合分析比较,以最终的统计集合预测效果最好,预测2008年北疆夏季降水将偏少,降水指数为106.9。结果表明:(1)滑动相关—逐步回归—集合预测方法以及在此基础上建立的统计集合预测模型具有一定的预测能力;(2)相关分析与时间序列分析相结合的思路和方法,可在季节、月等不同时间尺度的短期气候预测业务中应用。

Based on the summer （from June to August） precipitation data trom 29 meteorological stations distributed over Northern Xinjiang, the precipitation index is calculated. Adopting assemble regression method constructed by two stepwise regressions, including the initial {actors constructed by sliding-correlation method on account of the correlation coefficient stability, it is attempted to predict the summer precipitation index along with the factors of the previous monthly atmospheric circulation indexes. While the sample number is 47, the correlation coefficient between the original data and the simulated data is as high as 0. 7248. There are 10 prediction models to be set up with different sample size respectively, and the independent test is analyzed. The summer precipitation is to be lower in 2008, of which the index is 106.9. The results show that the statistical predict model based on the integration of sliding correlation, stepwise regression and assemble prediction has a certain prediction ability. The analysis method synthesizing regression and time series could be applied widely to short-term climatic prediction operation.