Case Study: Saving Lives with Crowd-Sourced Mapping in Haiti

From DevSummit
Jump to: navigation, search

Introductions

  • Schuyler - HOT, open data in humanitarian relief in recovery
  • Tom - CivicLab in Chicago[[1]], use tools like maps for activists
  • Will - March Hare with Ushahidi, software to support protesters
  • David - chat secure, POM+
  • Caroline - limited experience, want to learn more
  • Sarah - crowd sourcing experience
  • Lake - used to be with Architecture for Humanity, HOT website
  • Natalie - how this might work with things hit here
  • HIllary - OLPC in Haiti acquaintances
  • Wall Street and Occupy, mapping and crowd sourcing. Contribution levels
  • Willow just back from Sandy, how maps were being used

What can people do remotely?

Just contributing $10 doesn't give you the satisfaction of knowing you've actually helped

Open Street Map

Wikipedia but with GPS. Used to make maps etc

People go out in to the world and do a track log, annotate what they've done. Make roads and parks etc.

Places that places are better documented in OSM than any other method.

Commons-Based Peer Production.

Don't trust your sat nav device signs - when data couldn't be quickly updated.

Timely geographic data. Makes a difference for first responders.

Haiti

Presentation = [2]

The wealthier a place is, the more free time people have to map things. So Haiti was pretty sparse. Over half the buildings in t PAP were destroyed or at least damaged. 1 in 8 people died. Geography and no building codes. Earthquakes don't happen, poorly built structures falling on people in earthquakes kill people.

Groups with satellite imagery started publishing their data in a fit of social responsibility.

No more coordination than a wiki page and an IRC channel. No hierarchical dispersion of tasks. People just decided what they should be doing. Progression of detail. Haitian diaspora filling in name details.

For the first time ever, people were able to materially assist people in need while sitting on their couch. Access to imagery, traced.

Went into Ushahidi. Feed of text messages from 4636 with Digital. Damage repots, requests for help. All those went through a volunteer group translated, categorized, geolocated.Went into SOUTHCOM. People's lives were actually saved. Photos tagged on flickr to OSM. MapAction provided support to the UN and other groups, they used OSM, Every actor on the ground was using these maps.

Added into smart phones. Humanitarian Aid Team specifically asking for OSM.

The flooding in Pakistan barely got a mention. Haiti is in our back yard. Haiti has been a subject of philianthropic intervention for years. Floods went on for weeks, hard to maintain a news cycle. Volunteerism always correlates to media coverage.

Something like 10 Terabytes of data for Haiti. Pakistan had a few square kilometers come in. There is a politics to this.


Also rewarding. Not just fun. That people can help people remotely from the comfort of their own sofa is unprecedented.

Process

Satellite and aerial imagery Volunteers marshaled that imagery and made it useful Vols traced the imagery in SOM Made it useable to groups like MapAction

Hurricane Sandy

Great if you have no maps to begin with. What if you have an area which is already mapped? In OSM from census.

The need for situational awareness does not change. The First Responders don't know where is hardest hit. You have 72 hours to save people's lives - after that the survivor rate goes WAY down. Have to have access to most timely and accurate data possible.

Before Sandy was with Map Mill. Came out of Camp Roberts RELIEF. Army Cornels sitting next to hippies like Schuyler and Willow, testing out technologies. John Crowley had Civil Air Patrol and Tech at FEMA. CAP has about 5k members. Fly airplanes over areas affected by disaster right after the fact. Someone taking pictures. Got them into the same room, Geographic Info Officer at FEMA. If you take them in these ways, would actually be useful.

GPS is where the plane was, not what the photo is of. Oblique angle. Didn't care about the specifics, just where it's worst to get there quickly.

Basically Hot or Not for damage assessment. Did it look this way before the disaster or because of the disaster?

Multiple people looking at the same images so you aggregate value.

Split on the US Grid System, Able to see where all the troubles are. Close the loop of requests for assistance.

Using GIS info in a different way - don't have to be skilled. Aggregate picture.

Guts

The software was written before the disaster happened, worked before the disaster happened. If it isn't already working, the responders cannot afford to take chances. Can't just be reactive, implement tools ahead of time. This took us 6 hours to set up. Next time it'll be 30 minutes.

Call to action was just on social media - twitter and facebook. And then thanking coming in from FEMA with URL in video. This same sort of thing could be used in a Bay Area earthquake. Direct links to people in formal orgs to make getting things live quicker in the future.

Interannotator agreement. 0 5 10 . were mostly agreeing 60% of the time. 70% is better, aimed for 5 votes per image. Knowing when to quit pushing for participation - FEMA didn't need more feedback.

Standardizing how people take images. How do different groups rate images? GIS damaged trained, CAP people, volunteers, etc Maximize value of engagement.

When people are still working, maybe direct them to less well known issues.

140k assessments from asking on Twitter. That's awesome! Average damage assessment has a spike at the beginning, likely because CAP was taking images of places they thought would be damaged Most images were assessed at 6 seconds - what is it about images that make people spend that exact amount of time on them?

What things should be on the map?

Drones? Create a culture of imagery openness.