A University of Bath School of Management academic has developed an algorithm to help charities and aid organisations improve the way they help victims of storms, floods, earthquakes and other natural disasters.
Current planning tends to focus on the short-term challenge of supplying aid to victims in the immediate aftermath of a disaster – a single-pronged approach. But our model is two-pronged, integrating the issue of how to restore distribution networks to get aid to as many survivors as possible, said Dr. Ece Sanci.
“The incidence of natural disasters has increased dramatically in recent years. Since the beginning of 2021 alone, we’ve had extreme winter storms in the US, the Sulawesi earthquake in Indonesia, and severe flooding in Australia. Relief organisations must prepare for effective disaster response. If we are to deal with this ever increasing threat, scholars must help improve the way we respond to disasters,” she said.
Sanci and co-researcher Professor Mark Daskin of the University of Michigan have developed a planning model using an algorithm that takes into account the established practice and benefits of pre-positioning items needed for immediate aid relief in disaster-prone areas, such as water, food and medicines, but also addresses the need for access to equipment needed to restore distribution networks.
“By pre-positioning aid relief items responders need to rely less on local suppliers, who may have insufficient capacity to cope with the sudden surge in demand right after the disaster. Also, procuring these items from global suppliers may be expensive and time-consuming in the aftermath of the disaster. So finding the best location for these emergency response facilities is a key part of disaster relief planning,” Sanci said.
“But too much emphasis has been put on the location of these centres alone - even if the emergency response facilities are well located, people will suffer if damaged roads hinder the timely distribution of relief items. What’s worse, damage to the transportation network can leave victims entirely cut off,” she said.
Sanci said her and Daskin’s model combines the location and network restoration decisions, helping relief workers prepare better by determing both the locations of emergency response facilities and the equipment needed to restore distribution networks prior to a disaster.
“Our findings emphasised the importance of network restoration – we found that more, or larger, relief facilities have a limited impact when network restoration is ignored. However, if restoration resources are located at critical points, along with the required emergency response facilities, it is possible to satisfy the total demand for relief items quicker and at lower cost,” Sanci said.
Sanci said the algorithm offered a way to find a useful result in significantly less time than usual. She noted that, generally, algorithms take longer to find a result if they are evaluating a large number of scenarios needed for an accurate representation of uncertainty. The time, for example, taken by IBM’s high-performance solver CPLEX increases cubically with the number of scenarios it is considering.
“By contrast, the benefit of our algorithm is that the time taken to find a solution increases only linearly as the number of scenarios increase. In this way, our contribution is significant as it enables considerably faster decision making. In disaster relief work, this is important as time-sensitive and life-saving decisions will need to be made in response to short term forecasts,” she said.
“If we can find a solution to this issue, it may provide an alternative way to maintain connectivity with regions that are inaccessible due to damaged infrastructure,” she said.