How AI Will Help to Manage the Rising Risk of Natural Disasters
Climate change is a major driver of natural disasters. According to a 2019 report by the World Meteorological Organization (WMO), the number of natural disasters globally has increased by a factor of five over the last 50 years. Natural disaster reporting is now an expected feature in the global news cycle, with a surge of catastrophic weather-related events — from wildfires in Turkey, to floods in Germany — leaving many feeling like the climate crisis is reaching a boiling point.
Yet, some have long suffered the impacts of climate-related disasters like severe drought and coastal flooding in relative silence. These events disproportionately affect populations in small island developing states (SIDS) and least developed countries (LDC), where people have less access to resources, less developed infrastructure, and rely more heavily on agricultural production. Vulnerable populations have been battling through or fleeing the impacts of climate-related disasters for decades.
There is a glimmer of hope, however. Though the prevalence of natural disasters is on the rise, deaths from these major events have decreased — by almost three-fold, according to data from the WMO report. The drop in deaths is largely thanks to improved early warning and disaster management systems, demonstrating the importance of being able to predict natural disasters and manage the chaos that ensues.
Artificial intelligence (AI) could be a game-changer in natural disaster risk management. AI tools can boost our ability to predict natural disasters, communicate risks, and manage the aftermath. Our last post explored some of the many ways AI technology can be applied to push the energy transition ahead, but this is only a small corner of the climate change AI landscape. While there are numerous ways AI may help to mitigate or reverse some of the impacts of climate change, some of the most important use cases will actually be in helping us to deal with its inevitability.
Finding Patterns to Make Predictions
AI’s ability to analyze large datasets makes it a fantastic tool for making sense of data collected on seismic activity, rainfall records, satellite imagery and more. Over a period of time, these data create patterns, from which labels and categories can be constructed using machine learning (ML) systems. This information can then be used to detect anomalies or changes in order to aid researchers in understanding the potential risk and magnitude of natural disasters, as well as help governments to plan infrastructure in disaster-prone areas.
Japan is a nation at the forefront of AI-powered climate change technology for natural disasters. The 2011 earthquake and ensuing tsunami — which took the lives of approximately 18,500 people, destroyed thousands of homes and critical infrastructure to the tune of nearly $235 billion, and led to the worst nuclear disaster since Chernobyl — provides a clear example of the destruction that natural disasters can cause. As an island nation, Japan is extremely vulnerable to climate change: it weathers more earthquakes than most other countries around the world due to its location atop four tectonic plates and many active volcanoes.
This challenging reality has led the Japanese government to invest in finding innovative technological solutions for problems that will continue to worsen as climate change advances. Japan currently employs an AI-based system to predict earthquakes and tsunamis using satellite imagery, and machine learning (ML) hybrid systems are also under development for monitoring ground motion to predict seismic activity. Though the technology is in the early stages of deployment, research published in Nature demonstrates the promise of putting AI on the job.
Satellite-powered AI systems are also used in Japan to detect signs or landslides as well as to monitor aging infrastructure like roads and bridges. These systems help to detect and repair vulnerabilities before natural disasters hit. Satellites are not the only contributing technology assisting AI in managing the risk of natural disasters. The IoT also helps to fill the gaps by feeding ML systems a constant flow of data. Sensors are used to monitor both volcanic activity to predict eruptions and water levels to predict flooding.
It’s important to note that these types of tools currently have limited reach due to the high costs associated with their installation and maintenance. Unfortunately, many countries at the highest level of risk for natural disasters can’t afford the technology, as it currently stands. There is hope that as these tools become more common (and better) they will become cheaper, expanding their potential for impact to the places they are needed most.
Managing the ‘Right-Before’ and the ‘Just-After’
Being able to model, simulate, and predict natural hazards is only a part of the whole picture. Natural disasters will continue to occur regardless of how well they can be predicted. AI can also play an important role in helping governments and communities to communicate risks and notifications, and organize the logistics of disaster relief.
AI’s ability to monitor for natural disaster risks is a major benefit for early warning systems, but actually getting the warning out to populations, with clear instructions about what to do when they do occur, is also vital. AI works to fill the gaps in early warning systems by collecting live data gathered by sensors and satellites, making predictions based on these data, and transmitting the message to vulnerable populations.
Recent examples of AI being used in this way include Chinese company Xiaomi’s earthquake warning function, which is integrated into its MIUI operating system. The system alerts Xiaomi users, sending quake warnings “seconds to tens of seconds” before the first tremors are felt. Another recently proposed system aims to fill the critical gap of reaching those who don’t have access to phone or radio networks because they are located inside of buildings. The system uses AI to analyze CCTV footage in real time in order to alert residents or workers of emergencies using the alarms already fitted within buildings.
Some AI-powered systems are also in use for disaster relief and response. These systems work to help emergency crews, technical teams, and first responders to pinpoint the areas where their attention should be directed first. For example, One Concern uses AI to direct first responders to where they are most needed after a natural disaster, and IBM’s Outage Prediction predicts power outages as much as 72 hours before storms are expected, giving utility companies a grace period to plan and execute their response. Other groups use social media posts following a natural disaster to analyze where infrastructure damage is most severe, and where provision of aid should be prioritized.
Every Tool has its Limitations
As climate change drives natural disasters to become more prevalent and severe across the globe, AI will prove to be a powerful tool for saving lives through the prediction and management of catastrophic situations. While these capabilities are invaluable, it’s important to tread cautiously — especially when human lives are at stake.
The current landscape of AI-powered tools being used to curb the worst impacts of natural disasters have very real limitations to overcome. First and foremost, efforts to operationalize AI for these types of projects are limited in scope. The solutions that do exist are highly localized and focused on specific use cases. The private sector actors creating these initiatives usually have one or a few NGO or government partners, and their efforts are not well-integrated into the larger disaster relief community. This can result in a fragmented approach, where some communities benefit while others are completely left out.
Another set of challenges arise from the data used to train these potentially life-saving algorithms. Though bountiful data exists that can be used for disaster prediction, preparedness, warning, and relief, it may not always be easy or ethical to access. At the same time, distinct datasets are rarely combined in ways that could yield important insights. Poorly integrated and captured ground-view insights from experienced groups working or studying “on-the-ground” could be combined with findings from large datasets to paint a far better picture of the whole than big datasets alone can provide.
Yet, while current efforts may be patchy and fragmented and many meaningful data samples remain unintegrated, there is hope that AI-powered tools can help us to combat the worst effects of the climate change-driven natural disasters to come. It’s important to remember that we are only at the beginning of AI’s utility in society. This sentiment is both a warning as well as an opportunity. In a time when fancy new AI-powered tools seem to be under development left, right, and center, effective design and deployment will require a holistic and equitable approach.
