Rainfall timing and forecast verification for agricultural climate services in West Africa: a critical review

Seasonal rainfall accumulation, planting onset, and dry-spell sequences are different verification targets for agricultural climate services. These targets have different seasonal frequencies, forecast horizons, and observational requirements. This focused critical review examines how those differences constrain agricultural interpretations of forecast skill in West Africa. It brings together rainfall-climatology studies, forecast evaluations, and reference-data assessments, and relates them to established verification and climate-service frameworks. The synthesis identifies four requirements: define the rainfall target and eligible calendar window; match forecast horizon and spatial support to that target; compare with independently fitted, seasonally appropriate benchmarks; and test sensitivity to rainfall references and justified event definitions. The Agricultural Rainfall Forecast Evaluation Framework (ARFEF) records these requirements alongside the action deadline and the evidence supporting an agricultural claim. A reproducible constructed example gives lower daily rainfall root mean square error for one forecast, 0.246 versus 1.279 mm, but a higher error fraction for a specified planting event, 0.50 versus 0.00. Threshold and reference changes also alter the ranking. The example isolates rainfall-threshold classification; identical subsequent dry-spell sequences do not test dry-spell prediction skill. The review provides a rainfall-specific reporting synthesis for applied climatology, while distinguishing meteorological skill from conditional loss estimates and observed agricultural outcomes. Operational archive comparisons are still required to test its practical benefit.