Predictable patterns of the May–June rainfall anomaly over East Asia
Predictable patterns of the May–June rainfall anomaly over East Asia
复制标题
DOI：
10.1002/2016jd025856
复制
发表时间：
2017-02
期刊：
影响因子：
--
通讯作者：
W. Xing;Bin Wang;S. Yim;K. Ha
中科院分区：
文献类型：
--
作者：
W. Xing;Bin Wang;S. Yim;K. Ha
初夏（5 - 6月，MJ），东亚（EA）副热带锋是亚洲季风的一个定义性特征，它产生了全球副热带最突出的降水带。结果表明，1979-2010年4个耦合气候模式集合后报对初夏东亚（20°N-45°N，100°E-130°E）降水的动力学预报能力一般，不能用于可预报性的估计。本研究使用了一种替代的，基于经验正交函数（P-E）的物理经验（P-E）模型方法来预测降雨异常模式，并估计其潜在的可预测性。前三个主要模态是有物理意义的，分别归因于（a）异常的西北太平洋副热带高压与下伏印度洋-太平洋暖洋的相互作用，（B）与北太平洋海温（SST）异常相关的强迫，（c）赤道中太平洋SST异常的发展。建立了一套市盈率模型来预测前三个主成分。所有预测因子都比5月提前0个月，因此这里的预测被命名为0个月提前预测。交叉验证的后报结果表明，这些模式可以用显着的时间相关性技能（0.48-0.72）进行预测。利用预测的主成分和相应的模式，总MJ降水距平的1979-2015年期间后推。时间平均模式相关系数（PCC）得分达到0.38，显著高于动力学模型的多模型集成技能（0.21）。估计的潜在最大可达PCC约为0.65，这表明动力学预测模型可能有很大的改进空间。局限性和未来的工作进行了讨论。
During early summer (May–June, MJ), East Asia (EA) subtropical front is a defining feature of Asian monsoon, which produces the most prominent precipitation band in the global subtropics. Here we show that dynamical prediction of early summer EA (20°N–45°N, 100°E–130°E) rainfall made by four coupled climate models' ensemble hindcast (1979–2010) yields only a moderate skill and cannot be used to estimate predictability. The present study uses an alternative, empirical orthogonal function (EOF)‐based physical‐empirical (P‐E) model approach to predict rainfall anomaly pattern and estimate its potential predictability. The first three leading modes are physically meaningful and can be, respectively, attributed to (a) the interaction between the anomalous western North Pacific subtropical high and underlying Indo‐Pacific warm ocean, (b) the forcing associated with North Pacific sea surface temperature (SST) anomaly, and (c) the development of equatorial central Pacific SST anomalies. A suite of P‐E models is established to forecast the first three leading principal components. All predictors are 0 month ahead of May, so the prediction here is named as a 0 month lead prediction. The cross‐validated hindcast results demonstrate that these modes may be predicted with significant temporal correlation skills (0.48–0.72). Using the predicted principal components and the corresponding EOF patterns, the total MJ rainfall anomaly was hindcasted for the period of 1979–2015. The time‐mean pattern correlation coefficient (PCC) score reaches 0.38, which is significantly higher than dynamical models' multimodel ensemble skill (0.21). The estimated potential maximum attainable PCC is around 0.65, suggesting that the dynamical prediction models may have large rooms to improve. Limitations and future work are discussed.