How Artificial Intelligence Can Help Nepal Build Resiliency Against Future Earthquakes?

• Sumesh Shiwakoty

With all that is going on politically and economically, it seems we have started to forget that Nepal sits on a high seismic zone. As per geologists, Nepal lies atop a major fault line between two tectonic plates. Since the unbroken upper part of the fault is continuously building more pressure over time, researchers at Oxford University in recent years have predicted that another ‘major tremor could hit Nepal’s Gorkha district within years or decades rather than the centuries that typically elapse between quakes.’ If a major quake were to hit Nepal today, we are not in any better position to deal with its impact than we were on that darkest of days in April, 2015.

In recent years, many countries that sit in high seismic zones, like Nepal, are turning to artificial intelligence (AI) to build resiliency against future earthquakes. There is a broader consensus in the scientific community that AI can drastically reduce the loss of life and property during future quakes. This is because AI can analyse massive amounts of seismic data that can help better understand earthquakes and thus can provide faster and reliable early warnings. For instance, scientists have concluded that AI is more accurate in processing seismic data than the most useful existing model, referred to as the Coulomb stress transfer prediction model. As per their assessment of accuracy, on a scale of 0 to 1, where 1 is perfectly accurate, and 0.5 is as accurate as flipping a coin, scientists found that the Coulomb model scored 0.583, while the AI prediction model tested scored 0.849.

The future holds promise

The type of AI that researchers are using for processing seismic data is referred to as deep learning, a more advanced kind of machine learning that utilises a neural network. A neural network is a complex mathematical system, modeled like a web of neurons in the human brain, that mimic the thinking processes of the brain and learn new tasks on its own, like neurons in a human brain would. Scientists maintain that this particular type of artificial intelligence is better suited to process complex seismic signals. With so many complex variables to consider, from the position of the tectonic plates to the type of ground involved, scientists believe that neural networks can promptly analyse a massive amount of data, pulling out only relevant patterns and signals that could be relevant for earthquake forecasting.

Further, researchers have also concluded that AI can use previously underestimated signals, such as a particular sound made by the fault which can tell when an earthquake would arrive. This signal was previously assumed to be meaningless. In recent days, researchers at the Los Alamos National Laboratory in New Mexico have stated that AI can learn to discern a very specific pattern in the sound emitted by the fault before it ruptures. This pattern, researchers say, can tell us how much stress the fault is undergoing, and with the help of AI, can make an accurate prediction of the time remaining before tremors begin. Thus, with these discoveries, it is certain that AI will revolutionise early warning systems in the days to come.

Lagging behind

Although Nepal installed its first batch of earthquake early warning sensors in June 2015, there are neither robust nor adequate. The Nepal Academy of Science and Technology has estimated that Nepal will need 320 sensors to cover the entire country while, with the Chinese assistance, Nepal only installed its first 80 sensors in 2015. It has also been estimated that we would require approximately $20 million for setting up supercomputers and sophisticated broadband seismometers for this purpose. Since the use of AI is expected to dramatically raise the accuracy of the early warning system, Nepal should consider investing in it. Of course, it would be an expensive project for Nepal, but the benefits would far outweigh the costs. For example, Japan’s early warning system prevented the derailment of high-speed trains during a 9.1 magnitude earthquake in 2011.

Nepal should also keep an eye on how other countries that are located in high seismic zones are utilising newer technologies for early warning purposes. The government should partner with the private sector and non-governmental organisations to adopt innovative solutions that are adaptable to our needs. Earlier this year, the Los Angeles city council, in partnership with telecom company AT&T, launched a mobile phone app called ShakeAlertLA, which is designed to give users location-based early warning alerts. Closer to home, Build Change, a US based entity, is working on creating a smartphone-based artificial intelligence that aims to inform rural Nepalis on whether their houses can be seismically retrofitted. The main ambition of this app is to provide free engineering advice to poor homeowners whose houses are vulnerable to seismic events. Recently, this project was selected as one of the top three finalists for the global Call for Code Developer Challenge, hosted by IBM.

Newer technologies, like artificial intelligence, are expected to minimise the loss of life and property during future earthquakes. Many countries which are located in seismic zones are already investing in these technologies. It is high time for Nepal to seriously consider investing in these newer technologies for building resilience against future earthquakes. While this will require significant investment, when the lives and property of our people and the future of our cultural heritage are at stake, the benefits trump any costs of implementation.

Shiwakoty is a 2015 Andrew W Mellon Foundation environmental analysis research fellow.

-This article was originally published by The Kathmandu Post. Here is the link to the article: https://kathmandupost.ekantipur.com/news/2019-04-19/tools-for-better-seismic-detection.html