AI-Augmented Early Warning System — Mozambique's INGD/INAM Pilot for Cyclone and Flood Forecasting

Cyclone Idai struck central Mozambique in March 2019, killing 603 people and causing an estimated US$3 billion in damage. In response, Mozambique's National Institute for Disaster Risk Management and Reduction (INGD) and National Meteorological Institute (INAM) built out a multi-hazard early warning system with World Bank support (the US$265 million Disaster Risk Management and Resilience Program): satellite- and radar-based forecasts from INAM are disseminated through 70 community radio stations and trained community brigades that evacuate at-risk households ahead of a storm.

Cyclone Freddy — the longest-lasting tropical cyclone on record — battered central Mozambique twice in 2023 with winds of up to 230 km/h, stronger than Idai. According to the World Meteorological Organization, Freddy killed 183 people and caused US$176 million in economic losses, a large reduction relative to Idai despite the storm's greater intensity, which WMO and UNDRR reporting attributes in part to the intervening early-warning investment.

Since approximately 2023-2025, INGD and INAM have layered a machine-learning and satellite-data-analytics forecasting pilot onto this system, in partnership with UNDP, the World Bank, the Norwegian Meteorological Institute and the Green Climate Fund, aiming to shorten the time between hazard detection and public alert for floods, cyclones and droughts. Independent reporting by FurtherAfrica states plainly that "the use of AI remains at pilot level," with promising but not yet fully proven gains in flood-forecast accuracy, and that donor-funded forecasting improvements "have not always translated into consistent warning dissemination" given institutional-coordination and financing-continuity constraints.

The practice is included as an honest, still-unfolding case: a proven, largely non-AI early-warning system with a strong track record is being incrementally augmented with AI, but the AI-specific contribution has no independent before/after measurement yet, and the country has committed to a 2027 target for full 'Early Warnings for All' coverage.

Read the full analysis: https://furtherafrica.com/2025/11/10/ai-meets-climate-action-in-mozambique-building-smarter-early-warning-systems/