From forecasting to real-time alerting to post-disaster response, right here’s how we’re collaborating with the UN, governments and worldwide organizations on world disaster resilience.
Forecasting and getting ready for hazards
Well timed nowcasts and forecasts allow governments, humanitarian organizations and communities to take action earlier than disasters strike. Throughout the 2025 hurricane season, the U.S. Nationwide Hurricane Middle used Google’s WeatherNext mannequin. It predicted Hurricane Melissa’s historic Jamaican landfall 5 days upfront, enabling the Met Service in Jamaica to inform the general public. In Nigeria’s Adamawa state, UN OCHA launched a Floods Anticipatory Motion Programme utilizing Google’s river flood forecasts. When forecasts point out a excessive threat of great flooding, it prompts a set of early interventions akin to shelter preparation. The NGO GiveDirectly employed the same strategy in Nigeria’s Kogi State, utilizing Google’s forecasts to ship money transfers earlier than flooding. This enabled households to evacuate safely and buy gear like sandbags to guard their property.
Our forecasts can be found on Flood Hub, masking 2 billion individuals throughout greater than 150 international locations, in areas in danger for vital flood occasions. We’re repeatedly bettering forecasting capabilities along with our companions. The World Meteorological Group (WMO) and nationwide hydrological businesses in Czechia, Nigeria, Uruguay and Vietnam launched a pilot with Google to guage how native knowledge impacts AI forecasting in regional river basins. We discovered that incorporating native streamflow knowledge into world AI fashions considerably improves forecasts in ungauged areas. The outcomes of the examine, to be revealed within the coming weeks, spotlight the worth of mixing world AI fashions with localized experience, providing a blueprint for the way AI can higher assist nationwide forecasting efforts.
To additional advance analysis, we lately open-sourced our Groundsource dataset for city flash floods, and our hydrology modeling framework, serving to specialists construct new approaches whereas retaining full management of their very own native knowledge. We examined the hydrology framework with the Czech Hydrometeorological Institute (CHMI), who developed an adapter enabling them and different hydrological companies worldwide to make use of the mannequin of their normal workflows.
For wildfires, we leverage satellite tv for pc imagery to track fire boundaries in Search and Maps, with protection in 34 international locations. In collaboration with the Earth Fireplace Alliance and Muon Area, we developed the purpose-built FireSat constellation, which goals to supply an unprecedented, wildfire dataset and assist fireplace businesses detect wildfires extra shortly earlier than they unfold, anyplace on earth. Earlier at the moment, three new FireSat satellites launched from Vandenberg Area Pressure Base in California.
Offering life-saving alerts in occasions of disaster
In occasions of disaster, entry to dependable, authoritative data is important for these affected. In 2025 alone, Google helped join individuals with disaster data over 10 million occasions per day, on common.
Among the many alerts we distribute are Public Alerts — surfacing content material from alerting authorities via Widespread Alerting Protocol (CAP) feeds. These Public Alerts characteristic knowledge from authorities in over 90 international locations up to now, from companions just like the US Nationwide Climate Service, the UK’s Met Workplace and Brazil’s Centro Nacional de Gerenciamento de Riscos e Desastres (CENAD). We encourage extra nations to publish CAP alert feeds.
When authorities concern these alerts, they will seem throughout Search, Maps and as Android notifications. This ensures that public security data, akin to extreme climate warnings and flood updates, and the sensible data individuals want to remain protected, reaches individuals shortly and straight.
Whereas warning individuals successfully about earthquakes stays a important problem, we’ve got made progress in alerting these outdoors the epicenter. When devastating earthquakes hit Venezuela final month, Google’s Android Earthquake alerting system, leveraging a community of Android telephones as mini-seismometers, alerted tens of millions of customers outdoors the epicenter enabling them to take cowl seconds earlier than the shaking started.
Utilizing satellite tv for pc imagery to speed up catastrophe response
Put up-disaster, the core problem is getting life-saving support to the individuals who want it as shortly as potential. AI-powered insights will help governments and organizations reply extra effectively.
Knowledge Insights for Social and Humanitarian Motion (DISHA) developed a injury evaluation workflow in collaboration with Google, applied in collaboration with the UN Satellite tv for pc Centre (UNOSAT). It makes use of Open Buildings and Building Damage Assessment fashions to research satellite tv for pc imagery, and was lately enhanced with a brand new interface, marking a new phase of operational impact. To this point, it has been deployed 11 occasions, supporting the response to disasters like earthquakes, floods and cyclones. It permits high-precision evaluation of a whole bunch of 1000’s of buildings in very quick timeframes, saving UNOSAT specialists weeks of labor per activation.
When Hurricane Melissa devastated Jamaica in October 2025, this AI-based analysis assigned preliminary injury scores to over 385,000 buildings to tell restoration efforts. Extra lately, following the February 2026 floods in Colombia, UNOSAT quickly assessed broken infrastructure by cross-referencing AI-derived constructing maps with radar imagery of the flooding. The analysis knowledgeable the response planning of UN humanitarian businesses and the nationwide authorities.
