Case Study 7.1: How OpenStreetMap Used Humans and Machines to Map Affected Areas After Typhoon Haiyan
OpenStreetMap is a map database, built on the crowd-edited and copyleft model that many will recognize from Wikipedia. It provides some of the most detailed maps publicly available — particularly for many developing countries.
When Typhoon Haiyan struck the Philippines in 2013, a group of volunteer mappers came together to map and validate the damage experienced in the area. This was coordinated by the Humanitarian OpenStreetMap Team (HOT), which responds to humanitarian incidents by “activating” volunteers to map affected areas with fast turnaround. The work combines human validation with automated analysis to get results that are relied on by the Red Cross, Médecins Sans Frontières and others to guide their teams on the ground.
The HOT maintains a network of volunteers coordinated via a mailing list and other routes. Twenty-four hours before the typhoon struck, members discussed the areas likely to be hit and assessed the quality of existing data, preparing for a rapid response.
Once the typhoon reached the Philippines and was confirmed as a humanitarian incident, the HOT team called for the network of volunteers to contribute to mapping the area, including specific mapping priorities requested by aid agencies. There were two main goals. The first was to provide a detailed general basemap of populated areas and roads. The second was to provide a picture of what things looked like on the ground, postdisaster. Where had buildings been damaged or destroyed? Which bridges were down?
Work was coordinated and prioritized through the HOT Tasking Manager website (pictured below), which is a microtasking platform for map-making. It allows HOT administrators to specify a number of “jobs” to be done — such as mapping the roads and buildings within a defined area — and divides each job into small square “tasks,” each manageable by one volunteer mapper by tracing from aerial imagery.
During the Haiyan response, more than 1,500 mappers contributed, with up to 100 using the Tasking Manager at the same time. Dividing each job was crucial in order to make the best use of this surge of effort.
After claiming a task, a user edits their area of OpenStreetMap and can then mark their task square as “Done” (the red squares in the picture). However, the Tasking Manager requires that a second, more experienced person survey the work done before the task can be marked as “Validated” (green). (If the task was not completed properly, the “Done” status is removed by the second person.) Mappers can leave comments on the task’s page, explaining reasons for unvalidating or highlighting any issues encountered in mapping.
Aerial imagery is crucial to enable "armchair mappers" to contribute remotely by tracing roads, buildings and other infrastructure. Microsoft provides global Bing imagery for the use of OpenStreetMap editors, and this was used during Haiyan.
Representatives of HOT also liased with the State Department Humanitarian Information Unit through the Imagery to the Crowd program and other agencies and companies, to obtain high-resolution aerial imagery 1. Once that became available, the HOT team created new jobs in the Tasking Manager, asking volunteers to further validate and improve the basemap of the Philippines.
The Tasking Manager is the most visible validation step, but the OpenStreetMap ecosystem also crucially features a lot of automatic (machine-driven) validation. Map editing software (“JOSM”) automatically validates a user’s edits before upload, warning about improbable data, such as buildings overlapping, or rivers crossing without meeting.
Other automated tools regularly scan the OpenStreetMap database and highlight potential problems. Experienced mappers often use these for post-moderation: They can fix or revert problematic edits, or contact the user directly.
This workflow (combined with ongoing coordination and communication via email lists, blogs and wikis) provides a validation structure on top of OpenStreetMap’s human-driven community model.
The model remains highly open, with no pre-moderation and a semiformal hierarchy of validators; yet it rapidly produces highly detailed maps that international response agencies find very valuable.
Since the data are open, agencies responding to needs in the aftermath of Typhoon Haiyan have been able to use it in many different ways: They printed it out as maps; downloaded it to response teams’ SatNav units; used it to locate population centers such as villages; and analyzed it to understand patterns of disease outbreak.
This rapidly updated map data can also be used by journalists with a bit of geodata know-how; for example, to provide geolocated contextual information for data coming in from other sources such as tweets, to help validate claims about the relative impacts on different areas, or to produce infographics of the impact and spread of a disaster.
The original sentence was "Representatives of HOT also liaised with agencies/companies such as NASA, USGS and DigitalGlobe to obtain high-resolution aerial imagery." The correction was made in order to give credit to the main organisation which provided the imagery.↩