Seasonal Prediction of the Yangtze River Runoff Using a Partial Least Squares Regression Model
Seasonal Prediction of the Yangtze River Runoff Using a Partial Least Squares Regression Model
复制标题
DOI：
10.1080/07055900.2018.1448751
复制
发表时间：
2018-03
期刊：
影响因子：
1.2
通讯作者：
X. Ye;Zhiwei Wu;Zhaomin Wang;Huying Shen;Jianming Xu
中科院分区：
文献类型：
--
作者：
X. Ye;Zhiwei Wu;Zhaomin Wang;Huying Shen;Jianming Xu
摘要长江是亚洲第一大河，也是世界第三大河，流经欧亚大陆的大片土地。长江径流的季节性预测是一个重要而又具有挑战性的课题。利用长江流域1950-2016年的月径流观测资料，建立了长江流域径流指数（YRI）。YRI不仅能够定量分析长江径流状况，而且能够评估东亚夏季风的强度。长江流域夏季降水量与年降水量呈显著正相关。它还可以捕获东亚夏季风环流系统的主要组成部分。为预测长江夏季径流量，采用偏最小二乘（PLS）回归方法，结合YRI时间序列，寻找前冬海表温度（SST）模态。研究结果表明，第一个SST模态与厄尔尼诺（或拉尼娜）的衰减阶段有很强的联系，而第二个SST模态与持续的大拉尼娜（或大厄尔尼诺）有关。这表明，厄尔尼诺-南方涛动（ENSO）或巨型ENSO可能是长江夏季径流可预测性的一个重要来源。经过47年的训练（1950-1996），建立了一个物理-经验PLS模型，然后使用1997 - 2016年的3个月预测来验证模型。PLS模型表现出有前途的预测能力，优于一些国家的最先进的再分析数据系统。
ABSTRACT As the longest river in Asia and the third-longest river in the world, the Yangtze River drains a large land area of the Eurasian continent. Seasonal prediction of the Yangtze River runoff is of crucial importance yet is a challenging issue. In this study, observed monthly runoff data are used to develop a new Yangtze River runoff index (YRI) for the 1950–2016 period. The YRI is not only able to quantify the runoff state of the Yangtze River but is also able to evaluate the intensity of the East Asian summer monsoon (EASM). The YRI is highly correlated with summer precipitation in the Yangtze River basin. It can also capture the principal components of the EASM circulation system. To predict the Yangtze River summer runoff, we employed a partial least squares (PLS) regression method to seek sea surface temperature (SST) modes in the previous winter associated with the YRI time series. The findings indicate that the first SST mode exhibits a strong link with the decaying phase of El Niño (or La Niña), while the second SST mode is related to a persistant mega-La Niña (or mega-El Niño). These suggest that an El Niño–Southern Oscillation (ENSO) or mega-ENSO may be an essential source of predictability for the Yangtze River summer runoff. After a 47-year training period (1950–1996), a physical-empirical PLS model is built then a 3-month-lead forecast is used to validate the model from 1997 to 2016. The PLS model exhibits promising predictive skill that is better than some state-of-the-art reanalysis data systems.