EOF和CCA方法在台风路径预报试验中的比较
COMPARISIONS OF EOF AND CCA METHODS IN TYPHOON TRACK FORECAST TEST
-
摘要: 利用气候持续性因子,并分别考虑500hPa高度场的EOF因子和CCA因子,分别建立了基于EOF因子的和基于CCA因子的南海台风24-120小时路径预报模式。经过非独立样本和独立样本的预报检验、比较。将两种方法运用于500hPa高度场的分析表明,CCA方法得出的典型变量权重系数分布虽不能象EOF方法的特征向量一样表示出明显的而且平滑的空间场形式,但CCA方法浓缩了更多的与台风路径整体相关性最好的原变量场信息,因而更能有效地减小预报误差。预报检验表明,无论是基于EOF因子的路径预报模式,还是基于CCA因子的路径预报模式,都具有一定的预报能力,其中以后者的预报效果为更好,从5年预报来看,24、48、72、96和120小时的平均矢量误差分别为159.27、314.84、524.12、813.03和987.12km.该路径预报模式为南海台风短、中期路径预报提供了一种客观方法。Abstract: Based on both the CLIPER predictors and the EOF predictors,CCA predictors of 500 hPa height fields respectively, two sets of track forecast models for typhoons over the South China Sea at intervals from 24h to 120 h were proposed in this paper. The models were verified and compared using the dependent and the independent samples. The applied analyses of the two methods performed on the height fields or allohypsic fields at 500 hPa have been shown that the canonical weighted coefficient distribution derived by the CCA method did not exhibit the apparant and smooth space patterns of the eigenvectors derived by the EOF method. However, the ensemble information of original variable fields correlative to the tracks of typhoon were much more enriched by the CCA method. Consequently, the forecast errors were reduced effectively. The verifications indicated that the track forecast models either based on EOF predictors or based on CCA predictors show skill over CLIPER models, whereas the latter forecast performance is better. The 5-year-averaged vector errors are 86, 170, 283, 436 and 533 nautical miles at 24,48, 72, 96 and 120 h respectively. The track forecast models can be taken as effective objective prediction techniques for predicting the short and mid range tracks of typhoon over the South China Sea.
点击查看大图
计量
- 文章访问数: 864
- HTML全文浏览量: 0
- PDF下载量: 1179
- 被引次数: 0