Caribbean AI Disaster Response: What Hurricane Melissa's Drone Data Proved

Hurricane Melissa came ashore in southwest Jamaica on 28 October 2025 as a Category 5 storm, the strongest system ever recorded to make a direct hit on the island, with sustained winds of 185 miles per hour. The World Bank and the Inter-American Development Bank later put the physical damage at US$8.8 billion, 41 percent of Jamaica's 2024 GDP and the costliest hurricane in the country's history. Within days, a locally led drone response run by Jamaica Flying Labs, with WeRobotics and Esri, had flown roughly 300 missions across 900 square kilometres, mapping 320 communities in six parishes and generating close to three terabytes of imagery precise enough to count damaged rooftops from the air.

That response is now the working template for how the wider Caribbean plans to handle the next major storm. A regional disaster risk management conference met in Montego Bay in May 2026 to turn what Jamaica improvised into standing capability, and CDEMA convened seventeen countries and territories the same month to build the early warning systems meant to buy that response more time before landfall.

TLDR

• Hurricane Melissa caused US$8.8 billion in physical damage to Jamaica, 41 percent of 2024 GDP, according to a joint World Bank and IDB assessment published on 19 November 2025, the costliest storm in the country's recorded history.
• Jamaica Flying Labs, WeRobotics and Esri flew roughly 300 drone missions across 900 square kilometres in six parishes and 320 communities within weeks of landfall, the largest drone deployment yet mounted after a single disaster.
• Over 2,000 Jamaicans used ChatMap, a WhatsApp-based reporting tool built with the Humanitarian OpenStreetMap Team, to report flooding, landslides and blocked roads while much of the power grid was down.
• CDEMA, the WMO and UNDRR brought 25 representatives from 17 Caribbean countries and territories to Port of Spain in May 2026 to build country-specific roadmaps for impact-based forecasting.
• CIMH forecasts roughly 12 named storms, 7 hurricanes and 2 major hurricanes for the 2026 season, now underway, with CDEMA urging vigilance regardless of storm count.
• CAIA's reading: the region proved that emergency-built AI and drone capacity works. The task now is making it permanent and pre-positioned in every CDEMA member state, not rebuilt from scratch after the next storm.

A Category 5 Storm and an $8.8 Billion Bill

Melissa made landfall near New Hope in St Elizabeth parish with winds strong enough that the US National Hurricane Center classified it as the most powerful storm on record to strike Jamaica directly. Six parishes absorbed the worst of it: Westmoreland, St Elizabeth, St James, St Ann, Trelawny and Manchester. Housing, tourism, education, health, agriculture, energy, transport and telecommunications infrastructure were all affected within the same forty-eight hours.

The World Bank and the IDB put physical damage at US$8.8 billion. The Jamaica Information Service, counting broader economic losses alongside the physical damage, arrived at a total loss and damage figure of J$1.952 trillion, close to US$12.2 billion. Neither number was available quickly, and neither could have been produced by satellite imagery alone. Satellites gave planners a rough picture of which regions were hit. They could not resolve individual roofs, tell a repairable structure from a total loss, or reach communities cut off by landslides and downed bridges. That gap, between what satellites could see and what recovery planners needed to know, is what the drone response was built to close.

Three Hundred Flights, Six Parishes, Three Terabytes

CDEMA and the Government of Jamaica handed the drone response to Jamaica Flying Labs, which brought in WeRobotics and Esri as technical partners. More than 30 local responders, government agencies, regional organisations and technology partners worked the operation, flying roughly 300 missions across 75 grid blocks of three kilometres by three kilometres. The result covered 900 square kilometres and 320 communities, and produced close to three terabytes of processed imagery, orthomosaics with enough resolution to count damaged rooftops street by street, a level of detail satellite passes simply could not deliver on the timeline recovery agencies needed.

"Within our communities, we know where the damage is and we use a bottom-up approach," said Valrie Grant, founder of Jamaica Flying Labs, who has described the wider effort as a local initiative that was globally supported rather than the reverse.

The bottom-up approach extended past the drones. Jamaica Flying Labs partnered with the Humanitarian OpenStreetMap Team to build ChatMap, a tool that converts WhatsApp messages and voice notes into geolocated data. More than 2,000 Jamaicans used it to report flooding, landslides and blocked roads in close to real time, even as a large share of the island's power grid stayed down. Where the drones supplied a systematic aerial record, ChatMap supplied the granular, human account of conditions on streets no flight plan had reached yet.

From Emergency Improvisation to a Regional Playbook

Building directly on the Melissa response, Jamaica Flying Labs, WeRobotics and Esri convened the first Drones, AI and GIS for Disaster Risk Management Conference for the Caribbean and Central America in Montego Bay from 5 to 7 May 2026, with CDEMA as strategic partner. The goal was explicit: take what a single island improvised under emergency conditions and turn it into a standing regional capability, expressed as a 2026 to 2030 Roadmap for Geospatially Enabled Disaster Management aligned with CDEMA's own UAS for CDM programme. A second edition is planned for 2027 in Panama, co-organised with Panama Flying Labs.

What Melissa proved is that a trained local team, given a clear mandate and the right technical partners, can produce emergency-grade geospatial intelligence within days rather than weeks. What it did not prove is that every Caribbean state has an equivalent team ready to activate the moment a storm passes. That gap, between Jamaica's demonstrated capability and the rest of the region's, is precisely what the Montego Bay roadmap now has to close, and it will take budget commitments rather than another conference to do it.

Seventeen Governments Building the Warning Layer

A response, however fast, still starts after landfall. CDEMA, working with the World Meteorological Organization and the UN Office for Disaster Risk Reduction, convened 25 representatives from 17 Caribbean countries and territories in Port of Spain from 26 to 29 May 2026 for a Regional Workshop on Impact-Based Forecasting and Warning Services and the Common Alerting Protocol, funded through the CREWS Caribbean Project. Participants ran table-top exercises against simulated high-impact weather scenarios and left with country-specific roadmaps for strengthening warning dissemination and coordination between meteorological and disaster management agencies at home.

"One of the clearest lessons from Hurricane Melissa was that regional coordination mechanisms remain strong and effective, particularly in the early activation of the Regional Response mechanism," said Elizabeth Riley, CDEMA's Executive Director, who has also argued that Caribbean readiness "cannot be confined" to the official June to November hurricane season given the range of hazards the region faces year-round.

Forecasting and drone-based response sit at opposite ends of the same problem. One is meant to buy time before a storm makes landfall. The other compresses the weeks a recovery would otherwise take into days once it has. Both depend on data infrastructure, trained people and standing agreements that have to exist before a storm forms, not get assembled while it is already spinning up in the Atlantic.

A Season Already Underway

Today, the region sits in the middle of the season these systems now have to prove themselves against. The Caribbean Institute for Meteorology and Hydrology forecasts roughly 12 named storms, 7 hurricanes and 2 major hurricanes for 2026. CDEMA has described the season as potentially less active than 2025 by storm count, while still urging the region to stay on guard. That caution is worth taking seriously on its own terms: Melissa's US$8.8 billion bill came from a single storm, not from a busy season. A quieter year on paper says nothing about what one storm, on the wrong track, at the wrong intensity, can still do to a single island's economy.

The Governance Question Behind the Footage

Every layer of this system runs on data that did not exist as a coordinated resource before Melissa: rooftop-level imagery precise enough to identify individual structures, WhatsApp reports tied to GPS coordinates and sent by named residents, and forecasting models trained on Caribbean-specific storm behaviour. None of it was collected for a bad purpose. All of it is exactly the kind of information a regulator, an insurer, or an actor with worse intentions would also find useful.

CAIA has argued, alongside the Caribbean AI Risk Management Council at caribbeanairisk.com, that disaster-response data deserves the same governance attention as financial or health data, not less, precisely because it is gathered under conditions where consent is the first process suspended. A resident who reports a flooded street over WhatsApp because the alternative is no help at all has not necessarily agreed to have that report retained indefinitely, resold to an insurer, or merged into a dataset for a purpose nobody explained at the time. The Melissa response worked because it was locally led and fast. The next one has to be locally led, fast and governed, or the region will have built excellent emergency infrastructure without the safeguards that make it safe to reuse.

What the Caribbean Should Do Before the Next Storm

• Fund a standing drone and geospatial capability in every CDEMA member state, not only Jamaica, resourced ahead of hurricane season rather than assembled once a storm has already made landfall.
• Attach budget lines to the Port of Spain forecasting roadmaps, since a country-specific plan without money behind it becomes a document nobody executes.
• Write regional data governance rules for disaster-collected geospatial and community data now, while the Melissa dataset is still available as a working test case rather than a hypothetical one.
• Fund trained local response teams directly, the way CAIA has argued Caribbean boards should fund director education: as the actual bottleneck, since every part of the Melissa response that worked depended on people who knew what to do with the data, not on the sensors that collected it.

Adrian Dunkley, founder of StarApple AI, the Caribbean's first AI company, and widely regarded as the region's leading AI voice, has made a related argument in boardrooms and governance briefings across the Caribbean: the bottleneck is rarely the sensor or the model, it is whether the people receiving the output know what to do with it. The Caribbean AI Association tracks this work at caribbeanaiassociation.com, StarApple AI's own record is at starappleai.org, and the Caribbean AI Risk Management Council covers governance for data like this at caribbeanairisk.com.

The 2026 season runs to 30 November, and CIMH is still counting on seven hurricanes, two of them major. Jamaica now knows what a locally led drone team can do inside a week of landfall. The rest of the Caribbean has not yet had to find out.