Lending Eyes for Moore Oklahoma

As the tornados touched down in Oklahoma a few weeks ago, first responders jumped into action. In support of these brave and motivated people and organizations, DigitalGlobe tasked its satellite constellation to capture imagery of the area as part of its FirstLook service, collecting high resolution panchromatic WorldView-1 satellite imagery as well as color GeoEye-1 imagery.

Upon collection, DigitalGlobe launched its recently acquired Tomnod Crowdsourcing System (TCS) to help extract the information from the image. The Tomnod approach is most powerful in situations where rapid insight is required in order to enable fast decision making. For Moore, Oklahoma we immediately deployed the Tomnod system on the imagery to help convert the plethora of pixels to information.

An email rallied our Tomnod crowd to the new campaign. In addition we sent out the call on Facebook and Twitter and to dedicated groups, such as CrisisMappers. Users that came to the site were given a short tutorial and then asked to view imagery and identify destroyed buildings, tarped roofs, and fallen trees.

Tutorial screen seen by visitors to the deployment

On the backend, we constantly analyze the CrowdRank score of each location and each member of our crowd. CrowdRank is our statistical reliability algorithm that combines the crowd’s inputs to zero-in on the most accurate results. Within 60 minutes, we had collected over 15,000 points of interest and we published our crowdsourced damage assessment map. Based on the crowd inputs, CrowdRank identified the damage areas in Moore within 1 hour!

The damage map below immediately highlights the trail of total destruction left by the tornado [orange]. Just off of the main path of the tornado, we also clearly see the tarped roofs that had been identified [blue] where buildings were partially damaged by high winds or flying debris.

The CrowdRank consensus points showing the clear path of the tornado as well as damage off of the main path. Orange points indicate completely destroyed buildings and blue points indicate damaged/tarped roofs

Our crowdsourcing process is meant to reduce the time between data collection and decisions. Therefore, it is critical to distribute these results quickly and efficiently. Our damage assessment map is immediately available in a variety of formats such as SHP, WFS, and KMZ. In the produced KMZ file we include an image chip of the location that had been damaged accessible in a lightweight package.

Want to get involved? Please sign-up for our latest project – help us find burned buildings for Colorado’s Black Forest Fire.