Answers are grounded in this paper and cite their sources.
Central Mozambique (Sofala, Manica, Zambézia and Tete) is exposed to tropical cyclones, extremeprecipitation and riverine floods in the Zambezi, Púnguè, Búzi and Licungo basins. Artificialintelligence (AI) models have substantially improved meteorological and hydrological prediction, yettheir contribution to reducing losses depends on conditions that the literature rarely analyses together:transferability to basins with sparse data, the translation of forecasts into territorial impact, and thecapacity to turn warnings into action. This article argues that the value of AI in Early WarningSystems (EWS) is set by the warning value chain, not by forecast skill alone. Based on a criticalnarrative review and documentary analysis of cyclones Idai and Freddy, three propositions areformulated: (P1) the skill of AI models for cyclones is not automatically transferable to theMozambican coast; (P2) in poorly monitored basins, hydrological gains depend more on data andregionalisation than on model architecture; and (P3) loss reduction is bounded by the weakest link ofthe chain, which in the region tends to lie in impact conversion and last-mile communication. A hybridintegration framework is proposed that assigns each link a function, an institutional responsibility anda validation criterion. AI should be treated as a complementary decision-support instrument, withphased adoption conditional on local validation.
No takes yet. Share an insight, caveat, or question.
Marcio Danilo Joaquim de Morais (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: