郎咸梅. 2012. 中国东部冬季降水的动力结合统计预测方法研究[J]. 气象学报, 70(2):174-182, doi:10.11676/qxxb2012.017
中国东部冬季降水的动力结合统计预测方法研究
A hybrid dynamical statistical approach for predicting winter precipitation over eastern China
投稿时间:2010-04-12  修订日期:2011-07-08
DOI:10.11676/qxxb2012.017
中文关键词:  冬季降水,多元线性回归分析,季节预测模型,动力结合统计预测
英文关键词:Winter precipitation, Multivariate linear regression analysis, Seasonal prediction model, Hybrid dynamical and statistical approach
基金项目:中国科学院知识创新工程重要方向项目(KZCX2-YW-Q03-3)、国家重点基础研究发展计划项目(2009CB421406)、国家公益性行业(气象)科研专项(GYHY200906018)和国家自然科学基金项目(40875048)
作者单位
郎咸梅 中国科学院大气物理研究所国际气候与环境科学中心北京100029 
摘要点击次数: 3053
全文下载次数: 3240
中文摘要:
      针对中国东部6个气候关键区,首先,通过相关分析指出,冬季降水既与前期气候因子有关,又受同期大气环流的影响。因此,有必要采用动力与统计相结合的方法进行气候预测研究。然后,从实时预测的角度出发,综合考虑前期预测因子的观测信息和具有数值可预测性的同期气候因子的数值模式结果,使用多元线性回归分析方法就各区域平均冬季降水逐一建立了短期气候预测模型,并在预测模型中考虑了模型结果中系统误差的订正。交叉检验分析结果表明,所建立的各区域预测模型普遍具有较好的预测效果,预测优势主要表现在对冬季降水的变化趋势、年际变化、以及异常符号的预测准确率上。就6个区域平均而言,1982—2008年交叉检验结果与实况间的相关系数和距平同号率分别为0.69和78%,表明该预测思想具有可行性。
英文摘要:
      Correlation analysis revealed that winter precipitation in six regions of eastern China is closely related not only to preceding climate signals but also to synchronous atmospheric general circulation fields. It is therefore necessary to use a method that combines both dynamical and statistical predictions of winter precipitation over eastern China (hereinafter called the hybrid approach). In this connection, seasonal real-time prediction models for winter precipitation were established for the six regions. The models use both the preceding observations and synchronous numerical predictions through a multivariate linear regression analysis. To improve the prediction accuracy, the systematic error between the original regression model result and the corresponding observation was corrected. Cross validation analysis and real-time prediction experiments indicate that the prediction models using the hybrid approach can reliably predict the trend, sign, and interannual variation of regionally averaged winter precipitation in the six regions of concern. Averaged over the six target regions, the anomaly correlation coefficient and the rate with the same sign of anomaly between the cross-validation analysis and observation during 1982-2008 are 0.69 and 78%, respectively.
HTML   查看全文   查看/发表评论  下载PDF阅读器
分享按钮