In the early afternoon of October 10, 2018, Hurricane Michael made landfall in the Florida Panhandle southeast of Panama City. At landfall, Hurricane Hunter data estimated sustained winds of 155 mph and a central pressure of 919 millibars, the third lowest pressure of any hurricane to make landfall in the United States. NOAA predicted storm surges up to 13 feet in areas to the east of the eye.
The day after the hurricane, NOAA collected rapid response aerial imagery of the Florida coast. The imagery was processed by NOAA using Radiant Solutions’ OSSIM Libraries. Radiant Solutions’ GeoHIVE team prepared the NOAA imagery for a crowdsource campaign. The team used DigitalGlobe’s Vivid 2.0 global basemap as pre-event imagery, and Ecopia Building Footprints powered by DigitalGlobe. For every building footprint in the AOI, pre-event imagery was chipped out and post-event imagery was chipped out and rasterized such that the footprint is burned into the JPEG image chip as a pink outline of the building. We launched the campaign to the crowd on Saturday afternoon.
Three hours later, we had accurate damage assessed building footprint polygons. The crowd classified damage as Destroyed (building is no longer present or has been pushed off foundation/stilts), Damaged (building is damaged but still present) or No Damage Visible (cannot see damage to the building). Due to the nature of overhead imagery, the crowd was only able to see the roof or at most one wall. Therefore, crowdsourced damage polygons using overhead imagery underestimate the true level of damage.
Here is a map of the Gulf County, Florida coastal areas impacted by Hurricane Michael. Red dots represent destroyed buildings, orange dots represent damaged buildings and green dots represent no damage. Of the 7,167 buildings along the Gulf County Florida coastal areas impacted by Hurricane Michael, 830 (12%) are visibly destroyed, 1,394 are visibly damaged (19%) and 4,943 (69%) have no visible damage. Mexico Beach sustained the greatest damage and destruction.
Here are sample images across the three levels of damage the GeoHIVE team assigned the crowd to classify.
Crowdsourced rapid response damage assessment offers a fast and scalable solution to characterizing damaged areas across broad AOIs after natural disasters. This campaign demonstrates our ability to use imagery and vectors from different sources and fuse them into a crowdsourcing campaign. Although it does not completely characterize damage, rapid damage assessment from GeoHIVE provides quick insights into the spatial distribution of damage at large scale.
A Radiant Solutions’ full report and images that may be published can be downloaded here.