Satellite view of Earth at night with city lights and network links
Humanitarian

AI Assisted Mapping Is Putting Invisible Communities on the Disaster Map

7 min read|Updated November 2026
Share

At a glance

The Humanitarian OpenStreetMap Team, a nonprofit born after the 2010 Haiti earthquake, has mobilized roughly 468,000 volunteers to add more than 147 million buildings to free maps of vulnerable regions. Its open AI assisted mapping tool, fAIr, lets mappers run machine learning models over satellite imagery to detect buildings, roads, and infrastructure automatically, producing detailed maps in a fraction of the time manual tracing takes.

  • HOT was founded in 2010 after the Haiti earthquake, when responders arrived to find no usable maps of the country
  • The organization has supported more than 170 disaster activations and works across 94 countries
  • Around 468,000 volunteer mappers have made over 246 million map edits and mapped more than 147 million buildings as of 2024
  • fAIr is an open AI assisted mapping service where mappers run models over an area, validate the results, and fine tune the model on local examples
  • fAIr outputs are published as open data layers on the Humanitarian Data Exchange and Hugging Face

When a flood, earthquake, or storm hits an unmapped region, first responders are effectively working blind. In many parts of the world, vulnerable communities simply do not exist on digital maps. No roads. No buildings. No infrastructure. That invisibility can be fatal, because aid cannot be routed to places no one can find. The Humanitarian OpenStreetMap Team, known as HOT, is closing that gap by pairing a global volunteer mapping network with an open AI assisted mapping tool called fAIr, and the combination is changing how quickly crisis maps can be built.

Why it matters is simple arithmetic. HOT has supported more than 170 disaster activations since it was founded and works across 94 countries. Its goal, backed by The Audacious Project, is to map places home to one billion people so that no one is left out of life saving assistance. Manual mapping alone cannot keep pace with that ambition. AI that detects buildings and roads from satellite imagery can.

Born From an Earthquake

HOT originated in 2010 after the earthquake in Haiti. Hundreds of humanitarian organizations arrived ready to help, only to discover there were no updated maps of the country. Without reliable geographic information, coordinating resources, locating isolated communities, or planning safe routes was nearly impossible. Volunteers around the world began tracing satellite imagery into OpenStreetMap, the free editable map of the world, and a new model of crisis response was born.

Since then the model has scaled dramatically. As of 2024, HOT had mobilized around 468,000 mappers who together made more than 246 million map edits and added over 147 million buildings to OpenStreetMap, along with millions of kilometers of roads. Those maps are trusted and used by governments, NGOs, and humanitarian organizations responding to hurricanes, landslides, floods, and wildfires.

Mapping at the Speed of Disaster

The bottleneck has always been time. Crisis mapping was, in the words of HOT's tech partnerships lead Petya Kangalova, extremely labor intensive. Each building and road had to be traced by hand from satellite imagery. With AI assisted mapping tools, HOT can produce detailed, accurate maps in a fraction of the time by analyzing drone and satellite imagery to detect buildings, roads, and infrastructure automatically. That gives first responders critical information to take more informed, locally appropriate action before, during, and after a crisis.

The scale of recent emergencies makes this urgent. Nearly 240 million people worldwide are affected by conflict and disaster, and 118 million have been forced from their homes, according to the International Rescue Committee. When a typhoon crosses an island province or floodwaters rise over a dense informal settlement, knowing where every structure stands determines where teams search, what routes stay open, and how supplies are distributed.

How fAIr Works

fAIr, which HOT runs at fair.hotosm.org, is an open AI assisted mapping service released under the AGPL v3 license. It works as a shared collection of geo AI models. A mapper chooses a model, runs it over an area of interest, and the model proposes features such as building footprints. The mapper then validates what the model finds, correcting mistakes the way a human reviewer would. Crucially, the mapper can also fine tune the model on local examples, so the system steadily gets better at reading the specific building styles, terrain, and settlement patterns of that region.

The outputs are published as open data layers on the Humanitarian Data Exchange and on Hugging Face, so other organizations can build on them. HOT is deliberate about the boundary of the tool: fAIr keeps a human in the loop, and its roadmap has been through many model iterations and improvements, with a major shift in its approach underway as of 2026 according to the organization's own technology updates.

Validation and Local Knowledge

AI generated map features only matter if they are accurate, and the mapping community is testing that claim openly. A peer reviewed study published in the International Journal of Digital Earth in 2025 examined AI generated buildings in OpenStreetMap, comparing their frequency of use and their geometric characteristics against buildings mapped without AI, using real HOT Tasking Manager projects as the test case. That kind of independent scrutiny is part of why humanitarian agencies trust the data.

The other safeguard is local knowledge. In countries such as Sierra Leone, where dense settlements make mapping particularly complex, community members identify critical infrastructure like hospitals so maps reflect on the ground realities. HOT invests in localized training and education so that mapping is no longer something done to communities. It is done with them. Local mappers sustain the effort long after the emergency cameras leave.

Why Open Data Changes the Equation

Commercial satellite analytics exists, but its data is usually locked behind subscriptions and licenses. HOT's entire stack is open. The maps live in OpenStreetMap, the AI outputs land on public data exchanges, and the fAIr code is open for anyone to inspect or run. That means a small disaster agency in a low income country gets the same map quality as a large international organization, and every improvement made by one mapper or one model compounds for everyone.

A satellite image. A model that recognizes rooftops and roads. A volunteer who corrects what the machine misses. This is what modern preparedness looks like. The next earthquake will not wait for maps to be drawn by hand, and thanks to AI assisted mapping, it will not have to.

Common Questions

What is fAIr?

fAIr is an open AI assisted mapping service developed by the Humanitarian OpenStreetMap Team. A mapper chooses a machine learning model, runs it over an area of interest, validates the buildings and roads it detects, and can fine tune the model on local examples. It is licensed under AGPL v3 and its outputs are published as open data.

What is the Humanitarian OpenStreetMap Team?

HOT is a nonprofit founded in 2010 after the Haiti earthquake, when responders found no usable maps of the country. It coordinates a global volunteer network that adds missing geospatial data to OpenStreetMap, has supported more than 170 disaster activations, and works across 94 countries.

How many buildings has HOT mapped?

As of 2024, HOT had mobilized around 468,000 mappers who made more than 246 million map edits and mapped over 147 million buildings, according to published figures from the organization and its partners.

How does AI speed up crisis mapping?

Instead of tracing every building and road by hand from satellite imagery, AI models detect features automatically. A mapper validates and corrects the results. HOT says this produces detailed, accurate maps in a fraction of the time that manual mapping takes.

Are AI generated map features accurate enough for disaster response?

A 2025 peer reviewed study in the International Journal of Digital Earth analyzed AI generated buildings in OpenStreetMap and compared them with non AI mapped buildings, and HOT keeps human validators in the loop for every project. Local mappers also flag critical infrastructure such as hospitals so maps match on the ground realities.

Sources: Humanitarian OpenStreetMap Team website and fAIr documentation on GitHub; The Audacious Project profile of HOT; HOT Tasking Manager statistics; International Journal of Digital Earth, AI generated buildings in OpenStreetMap (2025); The Manila Times report on AI in disaster response (November 2026); International Rescue Committee global crisis figures (2026).