用典型相关分析作副高的统计动力预报模式可预报性研究

本文提出一个关于西太平洋副热带高压（简称副高）的统计动力预报模式，利用它和典型相关分析方法对冬、春和夏季逐月副高预报的可行性进行研究。结果表明，模式的可预报性依赖于预报量场和因子场所提取的分量数，模式的差分形式及预报落后步长。对逐月和不同步长所作的可预报性分析发现步长为1个月有较高的可预报性，不同月份可预报性有所不同，一般夏季较冬季和春季要差。虽然如此，用该模式作夏季副高预报还是具有一定的可能性。在独立样本中所作的预报试验表明，月际预报符号相关系数一般均接近或超过0.60。

A Study of Predictability of Statistical-Dynamic Model for Subhigh by Using Canonical Correlation Analysis

A simple statistical-dynamic model for forecasting the subtropic high pressure (subhigh) is presented in this paper. The predictability for its forecasting in winter, spring and summer is studied by the model using canonical correlation analysis. The results show that the predictability depends on the number of the principal components, Which are extracted from the predictand and predictor fields using principal component analysis, the difference form and forecasting step of lag months. It is found, from the experiments in different step of lag months, that the highest predictability presents with the step of lag of one month. The predictability is different in various seasons. In general, the predictability in winter and spring are better than summer. Nevertheless, the experimental results show that it is possible for subhigh forecasting. The anomalous sign correlation coeffients betWeen observation and forecasting can reach or surpass 0.60 for most months.