AI to predict and manage natural disasters

New applications can refine systems in use today at various stages of natural disaster management, starting with mitigation strategies

By Emanuele Ballestracci

60,000 dead, 150,000 injured and as many displaced per year. These are not the atrocities resulting from international conflicts, civil wars or terrorist attacks but the consequences of recurring phenomena whose destructive force is too often underestimated: natural disasters. Between 1998 and 2017, climate and geophysical disasters killed more than 1.3 million people and injured 4.4 billion, often among the weakest segments of the world's population. All the more, global warming will only increase in number and intensity of these phenomena, as we are already experiencing in recent years. Certain regions will suffer more than others, and Southeast Asia is among those most at risk. 99 percent of its population is already exposed to the danger of flooding, and between 2004 and 2014 it recorded 50 percent of global deaths due to extreme weather events. The situation will only deteriorate unless there is a revolution in the world's commitment to combating global warming, which now seems less and less likely.

A beacon of hope, however, comes from technological innovations in the field of artificial intelligence (AI). Indeed, its use in the creation of predictive models makes it possible to analyze large data sets, identify trends and thus predict potential disasters. Its applications would thus refine the systems in use today in the various phases of natural disaster management: cataclysm prediction and detection; early warning systems; vulnerability and risk assessment; spatial modeling; and mitigation strategies. Not only that, new detection systems are being developed that will especially benefit less resilient areas of the planet, such as the “AI-SocialDisaster.” This is a decision support system for identifying and analyzing natural disasters such as earthquakes, floods and fires using data drawn from social media feeds. Thus, by using information produced in real time by each individual without relying on advanced -- and expensive -- detection equipment, government capabilities for crisis management in rural areas will increase exponentially. For example, the Japanese company Spectee is developing a natural disaster detection system adapted for the Philippines, using precisely the information from social media. The role of private individuals is generally critical to the advancement of these new technologies. Microsoft Azure can be used to improve earthquake warnings and virtual representations of physical spaces in disaster response, while Amazon Augmented AI can lend itself to building integrated models for disaster scene recognition from low-altitude disaster images. China and the United States are already collaborating with respective champions in high-tech, such as Xiaomi and Google, while in South Korea, the Seoul metropolitan government has announced the development of a “digital disaster response platform” in which AI will be instrumental. In addition, Japan, Singapore, and China have made great strides in developing early warning systems, leveraging advanced technologies such as IoT sensors, AI models, and geographic information systems.

In addition to multinational corporations and governments, international and regional organizations are also making contributions. In 2015, the United Nations adopted the “Sendai Framework for Disaster Risk Reduction,” which outlines goals and priorities for action to prevent new disaster risks and reduce existing ones. In contrast, among regional organizations, ASEAN is one of the most active on natural disasters, a reflection of its high exposure to such phenomena. In 2009, the ASEAN Agreement on Disaster Management and Emergency Response was signed, two years later the AHA Center was established to revive regional coordination, and at the 28th ASEAN Summit in Laos in 2016 the Joint Declaration “One ASEAN, One Response” was signed. Finally, the ASEAN Civil Alliance for Regional Countermeasures was established last August 19, and since 2022, the topic of AI use has been increasingly discussed among member country summits, especially at the annual Strategic Policy Dialogue on Natural Disaster Management. Even the International Telecommunication Union (ITU), the United Nations specialized agency for information and communication technologies, has launched a new AI-themed working group: the Focus Group on Artificial Intelligence for Disaster Management (FG-AI4NDM). 

However, despite the potential of AI, there is no shortage of issues. First and foremost is the inability of AI models to provide “accountability” and “explainability.” Put simply, AI models function like black boxes: given certain predictions and inputs, they provide outputs but do not explain the relationship between the variables. This is a serious failing when these are used for crisis management, where maximum transparency is critical. However, should recent attempts to develop “Explainable Artificial Intelligence” models succeed, AI would undoubtedly become an even more valuable resource for counterbalancing the effects of natural disasters.