Key points are not available for this paper at this time.
ABSTRACT Forecasts on subseasonal (SS) timescales have become increasingly vital in bridging the gap between medium‐range weather and long‐range seasonal forecasts. The East African region is a ‘sweetspot’ for SS predictability, offering strong potential for these forecasts to inform critical agricultural decisions, such as determining the optimal planting times based on the onset of the rainy season. However, knowledge of rainy season onset predictability at this timescale is currently limited. In this paper, we evaluate the predictability of weekly rainfall and rainy season onset over East Africa (EA) at SS timescales in the ECMWF extended range forecast model. The results show that weekly rainfall forecasts have strong predictive skill up to 4 weeks ahead, with the ‘long rains’ March–May (MAM) season having higher skill than the ‘short rains’ October–December (OND). The latest ECMWF model cycle (C48) shows improvements in skill over the previous versions. Subsetting SS weekly forecasts by MJO phases does not produce markedly different results, suggesting that other drivers play a stronger role in shaping SS predictability. SS forecasts of onset dates have notable skill at 2–3 weeks' lead time, with high correlation and ‘hit rates’ (HRs) for forecasts of anomalously early and late‐onset years. However, mean absolute errors in onset date forecasts remain high, typically ~6–12 days. This underscores the challenges of predicting specific onset dates and the need for less stringent metrics, such as broader onset ‘windows’. Broader onset windows improve forecast accuracy, though a ±3‐day window achieves HRs < 0.5. Analysis of case study seasons of anomalous onset illustrates the operational value of SS onset forecasts in informing agricultural decisions. The study highlights the potential of integrating SS forecasts into a seamless seasonal‐to‐subseasonal (S2S) framework to support agricultural decision‐making and resilience through a ‘Ready‐Set‐Go’ system. It also reinforces the need to enhance co‐production with agricultural stakeholders in determining and evaluating appropriate onset metrics for actionable information to take advantage of the high SS predictability.
Mwangi et al. (Wed,) studied this question.