一个ENSO集合数值预测系统的历史后报试验

Hindcasting and Verification of an ENSO Ensemble Prediction System

对一个热带太平洋海气耦合ENSO集合数值预测系统进行历史后报检验。该系统模式为一个中等复杂程度的耦合模式,其中大气部分为统计模式,海洋部分为动力模式。扰动初始场通过集合卡尔曼滤波同化获得,并引入了模式误差,由一阶马尔科夫随机微分方程生成。集合预报样本数为100个。系统从1993—2011年共19 a的每个月都开始起报,每个预报样本都分别向后预报12个月。给出了El Nino指数时间序列预报结果检验、个例预报检验,以及El Nino指数和热带太平洋SSTA的相关系数和均方根误差检验,计算了模式的平均系统误差,并分析了其不同季节的预报性能。结果表明,系统具有较强的ENSO预测能力,能够提前6~12个月给出有参考价值的预报。

The hindcasting and verification is conducted to assess the performance of a tropical Pacific ocean-atmosphere coupled ENSO ensemble prediction system. As a coupled model, the system is characterized by medium complexity, in which the atmosphere part functions a s the statistical mode and the ocean part as the dynamic mode. The initial disturbance field is obtained through an ensemble Kalman filter(EnKF) data assimilation experiment. Model errors are also introduced in this paper, which are generated from a first-order Markov linear stochastic differential equation. The ensemble size is 100. The system released prediction in every month of the 19 years from 1993 to 2011, with each sample respectively being predicted forward for 12 months. This paper provides the examinations of the forecasting results of El Nino index time series, of some typical cases, as well as of the RMSE and correlational coefficients of the El Nino index and tropical Pacific SSTA. In addition, the average systematic error of the model is calculated and the seasonal forecasting performances are analyzed in this paper. The findings indicate that the system has a good capability of ENSO prediction, so as to present forecasts with reference value 6 to 12 months in advance.