[Journal Article] Location of disaster assessment UAVs using historical tornado data
Geomatics, Natural Hazard, and Risk
Citation : Grogan, S., Perrier, N., Gamache, M. & Pellerin, R. Location of disaster assessment UAVs using historical tornado data. Geomatics, Natural Hazards and Risk (2021). https://doi.org/10.1080/19475705.2022.2115407
Read more here : https://doi.org/10.1080/19475705.2022.2115407 (Open Access!)
Preparing and responding to disasters is a complicated task. one must balance coverage of SAR resources versus preparation cost. This paper presents a method and solution to prepositioning UAV damage assessment and search teams in Oklahoma using historical tornado data. The approach is based on set covering and multi-station vehicle routing models. It also presents a method to robustify the solution in the event a UAV team cannot be activated to respond to the disaster. This can simulate a team unable to respond. Results show 70% more stations and teams being required when chance of a depot failure goes from 0% to 5% and 90% more stations required when 0 to 10%. We find that when trying to use a solution that does not account for depot failure, the system of UAVs cannot meet search completion targets in 3-4% of cases. These results demonstrate accounting for the chance of teams not being able to respond to domestic disasters is important and failing to do so means an increased chance of not being able to respond adequately to disasters and incorporating the chance of station failure has a profound impact on the number of stations needed.
What does this mean for practitioners?
For OR practitioners, while we might find “optimal solutions”, these solutions might be sensitive to unknown elements that can disrupt the effectiveness of the optimal solution. To use an example, a city planner may find the best locations to build fire stations; however, what happens when a fire station is on fire? To extend this idea to this paper, there is a non-negligible chance that the UAV search tool would be not able to adequately respond to a disaster. This can be due to unplanned maintenance, flight crew not being able to make it to the launch station, or the station itself being struck by a tornado. This paper contributes to the growing body of literature on robustifying solutions against events that have a reasonable chance of occurrence yet are unknown in their time or severity. If we plan for 100% uptime and ability to respond from search and rescue teams, there would be almost a 1 in 20 chance of not being able to adequately respond to a tornado.
For disaster response practitioners, we hope this poses some creative use of well-known and recorded data into their decision-making toolkit. We also pose a framework for trading off the ability to adequately respond to disasters versus the number of stations. Depending on the assumptions made surrounding the chance a station cannot respond to a disaster, the number of stations required could be anywhere from 30% to 80% more stations.