Understanding how models represent sub-seasonal rainfall variations and what influences model skill is essential for improving sub-seasonal forecasts and their applications. Here, empirical orthogonal function (EOF) analysis is employed to investigate weekly Eastern Africa short rains variability from October to December. The observed leading EOF modes are identified as (i) a monopole-like rainfall pattern with anomalies impacting southern Ethiopia, Kenya, and northern Tanzania; and (ii) a dipole-like rainfall pattern with contrasting anomalies between Tanzania and the northeastern sector of Eastern Africa. An examination of the links between the leading modes and specific climate drivers, namely, the Madden–Julian Oscillation (MJO), El Niño–Southern Oscillation, and Indian Ocean Dipole (IOD), shows that the MJO and IOD have the highest correlations with the two rainfall modes and indicates that the monopole (dipole)-like rainfall pattern is associated with MJO convective anomalies in the tropical Indian Ocean and western Pacific (Maritime Continent and Western Hemisphere). Assessments of model ability to capture and predict the leading modes show that the European Centre for Medium-Range Weather Forecasts (ECMWF) and the UK Met Office models outperform the National Centers for Environmental Prediction model at forecast horizons from one to four weeks ahead. Amongst the drivers examined, the MJO has the largest impact on the forecast skill of rainfall modes within the ECMWF model. If MJO-related variability is reliably represented, the ECMWF model is more skilful at predicting the main modes of weekly rainfall variability over the region. Our findings can support model developments and enhance anticipatory planning efforts in several sectors, such as agriculture, food security, and energy.

了解模型如何表示次季节降雨变化以及影响模型技能的因素对于改进次季节预报及其应用至关重要。在这里，采用经验正交函数 (EOF) 分析来调查 10 月至 12 月东非每周短暂降雨的变化。观测到的主要 EOF 模式被确定为 (i) 类单极子降雨模式，异常影响埃塞俄比亚南部、肯尼亚和坦桑尼亚北部； (ii) 坦桑尼亚和东非东北部地区出现类似偶极子的降雨模式，且异常情况对比鲜明。对主要模式和特定气候驱动因素（即马登-朱利安涛动 (MJO)、厄尔尼诺-南方涛动和印度洋偶极子 (IOD)）之间联系的检查表明，MJO 和 IOD 与两种降雨模式并表明类单极（偶极）降雨模式与热带印度洋和西太平洋（海洋大陆和西半球）的MJO对流异常有关。对模型捕捉和预测主要模式能力的评估表明，欧洲中期天气预报中心 (ECMWF) 和英国气象局模型在未来一到四周的预测范围内优于国家环境预测中心模型。在所检查的驱动因素中，MJO 对 ECMWF 模型中降雨模式的预报技能影响最大。如果可靠地表示与 MJO 相关的变化，ECMWF 模型就可以更熟练地预测该地区每周降雨变化的主要模式。我们的研究结果可以支持模型开发并加强农业、粮食安全和能源等多个领域的预期规划工作。