Classification analysis of prediction skill among ensemble members in MJO subseasonal predictions—based on the results of the CAMS-CSM subseasonal prediction system

Abstract

Given the inherent imperfections in models and the inevitability of initial condition errors, subseasonal prediction ability faces ongoing limitations. Most international numerical models employ ensemble forecasts to enhance the accuracy of subseasonal prediction. The prediction skill for the Madden–Julian Oscillation (MJO), as a vital source of subseasonal predictability, depends on both model performance and the physical nature of the events. Based on the reforecast results of the CAMS-CSM subseasonal prediction system, the differences in MJO prediction skill among ensemble members are classified and compared together with the characteristics of different kinds of MJO events. In the category with generally high ensemble member prediction skill, MJO events often have extended durations, stronger intensities, and the intense convection primarily locates in the Indian Ocean, gradually shifting eastward to the western Pacific. In the category with mostly poor ensemble member prediction skill, the convection strength during MJO event propagation is weakest. In the category with mixed prediction skill among ensemble members, MJO events tend to have shorter durations and lower intensities, and the convection centers during subsequent propagation exhibit stationary characteristics over Maritime Continent regions.摘要由于模式误差和初始误差所致, 次季节-季节预报技巧整体偏低. 国际上多数模式都采用集合预报的方式来提高次季节预报的准确率. 热带大气季节内振荡 (MJO) 作为次季节尺度可预报性的重要来源, 其预测水平取决于模式性能和MJO事件本身的物理特性. 根据中国气象科学研究院气候系统模式次季节预测系统的回报结果, 结合不同类型MJO事件的特征, 对模式集合成员间的预报技巧进行了分类和比较. 在集合成员预报技巧普遍较高的一类MJO事件中, 对流异常信号持续时间较长, 强度较大, 强对流异常中心主要位于印度洋区域, 并逐渐东传至西太平洋. 在集合成员预报技巧多数较差的MJO事件中, 对流异常信号的强度最弱, 维持时间最短. 在集合成员预报技巧优劣参半的类别中, MJO往往持续时间较短, 强度较低, 在后续传播过程中, 对流异常中心多停驻在海洋性大陆区域.