基于土壤湿度和年际增量方法的我国夏季降水预测试验

选取欧亚大陆9个关键区的土壤湿度年际增量作为预测因子,采用变形的典型相关分析(BP-CCA)结合集合典型相关分析(ECC)方法建立集合预测模型,对我国东部夏季降水的年际增量进行预测,进而预测夏季降水。其中,1980~2004年的资料用于历史预测试验,而2005~2014年的资料用于独立样本预测试验。首先利用BP-CCA方法对9个因子分别建立单因子预测模型,然后采用ECC方法对9个预测因子按照不同的组合方式建立集合预测模型,并且对独立样本检验的效果进行了评估。结果表明,不同预测因子的组合对我国夏季降水均表现出一定的预测能力:东欧平原、贝加尔湖以北、我国河套地区及长江以南地区的土壤湿度对华北夏季降水预测效果较好;而巴尔喀什湖以北地区、我国西北地区、河套地区以及长江以南地区的土壤湿度对江淮夏季降水有较好预测效果;东欧平原、巴尔喀什湖以北地区以及我国河套地区的土壤湿度对华南降水预测技巧较高。这三组模型预测出的降水变化趋势与相应区域的观测结果较为一致,且预测评分(PS)均超过70分,距平相关系数(ACC)均为正值。研究表明土壤湿度因子中包含了对我国夏季降水有用的预测信号,可以考虑将土壤湿度应用于夏季降水的预测业务中。

Summer Precipitation Prediction in China Using Soil Moisture and the Year-to-Year Increment Approach

Using year-to-year increments of soil moisture in nine key regions over Eurasia as predictors, statistical prediction models were developed. These models were based on Barnett-Preisendorfer Canonical Correlation Analysis (BP-CCA) and were combined with the Ensemble Canonical Correlation analysis (ECC) method to predict the year-to-year increments of summer precipitation over eastern China, and thus obtain the prediction of summer precipitation. Specifically, data during the epochs of 1980-2004 and 2005-2014 were used to perform historical prediction and independent sample tests, respectively. First, single factor prediction models of the nine predictors were built using the BP-CCA method. Then, ensemble prediction models were developed using the ECC method, based on different combinations of the nine predictors, and the scores for the region with predictive skill in the independent sample test were also calculated. The results showed that the combinations of different predictors had different predictive skill for summer precipitation in China. The soil moisture in the eastern European Plain, in the area north of Lake Baikal, in Hetao region, and in the area south of the Yangtze River, had a good predictive effect for summer precipitation in North China. The soil moisture in the area north of Balkhash Lake, in Northwest China, in Hetao region, and in the area south of the Yangtze River, had a good predictive effect for summer precipitation in the Yangtze-Huaihe region. The soil moisture in the eastern European Plain, in the area north of Balkhash Lake, and in Hetao region, had high predictive skill for summer precipitation in South China. The precipitation change trends predicted by three models were consistent with observations in the corresponding regions. Prediction scores all exceeded 70 points, and anomaly correlation coefficients were all positive. The study shows that soil moisture contains useful signals for summer precipitation in China, and can be considered for application in summer precipitation prediction operations.