Can AI Act Against Natural Disasters?

Climate change has led to increased instances of natural disasters such as earthquakes and tsunamis, resulting in massive scale of destruction and loss of lives. Technologies such as AI can come to our rescue here. Research to use AI for predicting such occurrences is underway, thereby ensuring necessary evacuation to mitigate damages. AI is also being developed to help in disaster management and rescue operations following such incidences. However, whether AI will function as a saviour against natural disasters is yet to be seen.

The planet is witnessing increasing instances of natural disasters due to climate change. Disasters such as earthquakes, tsunamis, floods, and volcanoes are highly destructive and fatal. Therefore, there is a need for technologies that can predict these disasters, allowing time to undertake evacuation and saving lives as well as property. Entities ranging from government organizations and companies to universities funding and researching projects focus on using artificial intelligence (AI) as a predictive tool to predict and estimate the effects. The technology also has the potential to effectively support relief measures, thus making them efficient and limiting casualties. However, major initiatives are still at the level of university research and it might take a couple of years for the efforts to bear fruit.

Prediction
Earthquakes – Earthquakes are caused due to seismic movements and AI can be a valuable tool for detecting and analyzing the first signs of these types of events. The appropriate data is required to train the algorithm properly, especially for high magnitude earthquakes. The “micro-earthquakes” occurring near fault lines can be a repository for the needed information.

University research - A group of researchers at Stanford University have developed a new machine learning algorithm named Earthquake Transformer. This algorithm can provide a level of accuracy close to that of human analyses and detecting a large number of earthquakes, particularly those of low intensity, which are usually not identified by traditional detection methods.

Government agency – Researcher Paul Johnson and his team from Los Alamos National Laboratory, in collaboration with various universities, are using AI to create earthquake prediction tools. The team creates lab-simulated earthquakes and collects data before, during, and after these events. AI algorithm sifts through this data to look for patterns that can indicate when an artificial quake will occur. The team has now begun this analysis using raw seismic data from real temblors.

Floods – Variables such as excessive rainfall and large amount of snowfall melting and leading to rising water levels make it difficult to accurately predict floods.

University research – Researchers from the Lassonde School of Engineering use AI-based models and data from Don River in Toronto and Bow River in Calgary to predict the water levels in rivers much in advance of floods. Still under study, this research could help make reliable predictions.

Company research - Google Flood Forecasting Initiative is an offshoot of the popular search engine and uses its infrastructure and AI abilities to forecast accurate real-time flood information. This system is powered by AI and physics-based modelling and creates scalable inundation models in real-world settings. The on-ground data is taken from government agencies, allowing the system to predict not only the place of occurrence but also its severity.

Volcanic eruptions – There are about 1,500 potentially active volcanoes around the world, of which 6% (50–85) erupt each year.

University research - Juliet Biggs is a respected volcanologist from University of Bristol, UK. She and her team implement machine learning algorithms to determine the formation of ground distortions around volcanoes. They use radar observations from two satellites to collect data on the various identified volcanoes every 6, 12, or 24 days. As the satellites repeatedly pass over the same spot, they measure the distance between themselves and the ground and record any difference over time, thereby indicating ground shifts as magma moves beneath a volcano. They use AI-based models for data sifting.

Company research – IBM, in partnership with researchers from California Institute for Technology, University of Texas at Austin, and New York University, has created a simulation of the earth’s tectonic plates to help predict earthquakes and volcanic eruptions using its AI system, Watson.

Tsunami – Tsunamis are among the natural disasters that cause considerable destruction. This destructive event is usually caused by earthquakes on converging tectonic plate boundaries.

University research - The National Science Foundation in USA funds three of Oregano’s largest universities and other colleges in southern Washington to use big data and AI models to predict earthquakes and tsunamis.

Company research - Japan’s Fujitsu Laboratories created an innovative AI model to predict tsunami flooding in coastal regions using the world’s fastest supercomputer, Fugaku. The researched used training data from 20,000 possible tsunami scenarios via this computer and created an AI model based on them. In case of an earthquake, inputting tsunami waveform data observed offshore can predict the flooding that will occur in coastal areas before landfall. The system can help predict infrastructure damage and make evacuation measures more effective.

Disaster Management
In addition to prediction, AI can also be implemented to aid in rescue operations following a disaster. Some of these innovative tools have already been used to save lives after disaster struck.

Robot rescuers – Using robots to search for survivors in case of building collapses can help expedite rescue operations. A rescue robot called Snakebot, enabled by AI and equipped with lights and cameras, was developed by Carnegie Mellon University in Pittsburgh. Prototypes of this robot were used in rescue missions following the Mexico City earthquake in September 2017. With advancement in this technology, developers now plan to add microphones and sensors that can detect hazardous gas.

Targeted help – Disaster strikes lead to immense flow of data in emergency services. It is necessary to effectively understand the data and deploy resources accordingly. AI can analyze through a vast pool of data quickly, identifying patterns that can prove helpful. In 2019, after Cyclone Idai, DigitalGlobe’s Open Data Program was used to obtain high-resolution satellite imagery and applied AI to design disaster response.

First responders – The first responders on any scene of natural disaster are usually the medical staff, firefighters, and police officers who can assess the situation and take quick decisions. Hence, it is necessary that they receive communication as soon as possible. A mobile communication platform, BlueLine Grid, connects users to a set network of first responders, security teams, and law enforcement bodies. This platform is effective as users can quickly find public employees by geography.

Conclusion
AI is slowly gaining traction as a technology that can prove effective in various ways to combat natural disasters. The technology is yet to realize its full potential in being an efficient tool in complete disaster management. AI can combine data from earth observation street imagery, connected devices, and volunteered geographical details. However, technology alone cannot resolve all the related issues. People need to work creatively to tackle challenges in a collaborative manner.