欧洲多个耦合气候模式对东亚冬季气候的预测性能研究

Study of East Asian winter climate predictability by using European multi model ensemble prediction

中文关键词:  短期气候预测, 多模式集合, 合成数据集, 可预测性

英文关键词:Short-term climate prediction, Multi model ensemble, Synthetic data sets, Predictability

中文摘要:

      在短期气候预测方法中，多模式集合预测作为一种实用方法得到了广泛的研究。利用DEMETER多模式集合预测系统1980—2001年的回报试验，研究了欧洲7个耦合模式对东亚地区(0°—60°N，70°—140°E)冬季大气环流和气候异常的预测效能。研究的气候要素是冬季500 hPa高度场、850 hPa风场、表面气温场和降水场。集合平均(EM)是最基本的多模式集合构建方法。为了进一步订正模式预测的误差，基于经验正交函数分解进行订正，产生“合成数据集”，并利用该数据集进行合成集合平均或合成超级集合(SEM/SSE)。研究表明，东亚地区冬季气候异常的模式预测效能热带高于中高纬度地区，海洋高于内陆。多模式集合平均和合成集合平均或合成超级集合均从整体上对东亚地区冬季气候异常的预测效能有一定程度的提高，体现了其相对于7个单一模式的优势。两类不同的多模式集合方法对预测结果也有一定的影响，其中，合成集合平均或合成超级集合对冬季500 hPa高度场、850 hPa风场和降水场异常的预测效能优于集合平均；但是对于冬季表面气温场异常的预测，集合平均优于合成集合平均或合成超级集合。

英文摘要:

      In short-term climate prediction, the multi-model ensemble prediction is widely used as a practical approach. In this paper, the predictability of anomalies for the winter atmospheric circulation and climate in East Asian area (0°-60°N, 70°-140°E) is evaluated by using the 1980-2001 hindcast data from the DEMETER multi-model ensemble prediction system. The climate variables used are 500 hPa geopotential height, 850 hPa wind, surface air temperature and precipitation. In this paper, the Ensemble Mean (EM) is used as the primary method to construct the multi-model ensemble prediction. To correct the model predictions, the modes in the prediction space are calibrated by using the Empirical Orthogonal Function. A group of new Synthetic Data Sets are generated and then used as inputs for the Synthetic Ensemble Mean or Synthetic Superensemble (SEM/SSE) method. The results show that, in East Asia, the winter climate anomalies predictability is larger in the tropics than in the middle-high latitudes; besides, the predictability is larger in oceans than in inland areas. Multi-model ensembles, both EM and SEM/SSE, could generally improve the predictability of winter climate anomalies in East Asia, suggesting the multi-model ensembles’ advantages against individual models used in the DEMETER project. The two types of method used for multi-model ensemble construction could also influence final prediction results. For geopotential height, wind, and precipitation anomalies, the prediction skill of the SEM/SSE method is better than that of the EM method; while for winter surface air temperature anomaly, the prediction skill of the EM method is better than that of the SEM/SSE method.