资源环境科技发展态势分析平台

The ‘short rains’ in East Africa from October to December have significant year-to-year variability. Their abundance or deficiency is often associated with floods or droughts for which early warning is crucial, though even in normal seasons skillful forecasts facilitate planning and preparedness. Here we study the relationship between initial-state sea surface temperatures and subseasonal rainfall forecast errors in the ECMWF model in the region. We demonstrate that the initial mode of the Indian Ocean Dipole (IOD) is a partial control on the rainfall error in weeks 3–4. This relationship is also clear on the seasonal scale, exemplified by too-wet forecasts during the 2015 season when the IOD was positive, and too dry forecasts in 2010 when it was negative. Our results provide an entry point for model improvement, and we show that a priori forecast corrections based on the initial IOD index are feasible.