OliWoods-Org/mama-disaster-response

Africa Early Warning System. Weather ingestion, alert generation, WhatsApp delivery. AGPL-3.0. mama-disaster-response Africa Early Warning System. Weather ingestion, alert generation, WhatsApp delivery. --- > *"Less than 40% of people in Africa have access to early warning systems for extreme weather. In 2023 alone, floods in East and West Africa killed thousands of people who had no advance warning. The technology to save them exists. It just wasn't pointed at them."* > — WMO State of Climate in Africa, 2023 ## Why This Exists Early warning systems save lives — but only if people receive the warning. In sub-Saharan Africa, where smartphone penetration averages 34% and grid power is unreliable, standard push-notification disaster systems reach almost no one. WhatsApp, however, reaches 500 million people across the continent. This system is built around that reality. - **Less than 40%** of Africans have access to early warning systems for extreme weather events (WMO, 2023) - **Africa faces 5x more climate disasters** per capita than other regions, yet receives less than 3% of global climate adaptation funding (UNEP, 2023) - **2,000+ deaths** from preventable flood and drought events in East Africa in 2023 alone, most in communities with no warning infrastructure (ReliefWeb) - **WhatsApp penetration** in sub-Saharan Africa exceeds 70% in urban areas and 40% in rural areas — making it the highest-reach alert channel on the continent This system turns weather data into WhatsApp messages in local languages, before the flood arrives. ## System Architecture ```mermaid flowchart TD A[Weather Data Ingestion\nECMWF + CHIRPS + GFS\n+ Regional Met Services] --> B[Data Normalization Layer\nStandardize formats\nacross 15+ data sources] B --> C[Hazard Detection Engine\nFlood / drought / cyclone\nheatwave / locust thresholds] C --> D{Severity Classification\nGreen / Yellow / Orange / Red} D -->|Yellow-Red| E[Alert Generation\nClaude API — plain-language\nalert in 12 African languages] E --> F[Community Targeting\nGeo-fence to affected\nvillage / district / r …