AI, citizen science, disaster response combine to help Hurricane Irma’s victims

Rebekah Yore, Operations Manager at Rescue Global, commented: “By the morning of Friday 15th September, we were told by the Zooniverse team that roughly 300,000 classifications from 7,500 people had taken place through the platform. This extraordinary effort is the equivalent to the output of one person working full-time for just over a year, or that same person working 24/7 without breaks for around three months. And the number of volunteers and classifications are increasing daily. This input is already having a direct effect on the ground, helping to provide situational awareness for all deployed teams.”

The sheer volume of images would take an individual months to sort through, but can be analyzed in a matter of hours by the “crowd.” Because the images are often of poor quality, human observers are much better placed to perform this part of the task than computers.

For the next step, however, computers are essential, and Oxford engineering researchers have developed a suite of sophisticated artificial intelligence tools which can process the resulting data. Machine learning approaches quickly reconcile inconsistent responses, aggregate the data and integrate information derived from other crowd-sourced mapping materials, such as the Humanitarian Open Street Maps and Tomnod. This approach generates the best information possible to inform relief efforts. This analysis enables the team to build impact ‘heat maps’ that identify the areas in need of urgent assistance. Oxford has considerable expertise in this area: the tools have been refined over several years and were previously used to assist Rescue Global in its response to the 2015 Nepal and 2016 Ecuador earthquakes.

The “heat maps” enable Rescue Global to decide where to send its own small reconnaissance planes to conduct detailed aerial assessments, and to share critical information with a multitude of governmental and humanitarian partners [see]. Working closely with Airlink, which flew in aid to a central location, Rescue Global has been using information gathered through the Zooniverse platform and its own needs assessments to coordinate the onward delivery of aid through a network of boats and planes, ensuring that it gets to those who need it most.

Oxford says that this new technology offers an evidence-based, rational approach to disaster management. Through collaboration with crisis responders like Rescue Global, Oxford researchers are making a unique and significant difference to victims of Hurricane Irma.

Dr. Steven Reece, Machine Learning Research Fellow and mapping lead at Oxford University, said: “As always we are extremely grateful to our friends in the satellite industry for providing data and, of course, the crowd for their amazing work interpreting the imagery so quickly. This has been a sustained campaign and we’ve now produced heat maps for all the Virgin Islands. With Hurricane Maria increasing in strength and bearing down on the same area, we will have a lot more work ahead of us.”

Volunteers are urged to help in classifying new images and to assist the ongoing response efforts at:; follow Rescue Global’s deployment to the Caribbean through regular updates (SITREPs) on their website: