Wayanad, Kerala landslides: Can AI predict landslides and what scientists say?

More than 200 people have lost their lives in a series of landslides due to heavy rains in Kerala’s Wayanad district on Tuesday (July 30), and the toll is expected to rise as rescuers search through the debris.

Over 300 houses were completely destroyed in the landslides that hit the Mundakkai and Chooralmala areas of the district.

The Indian Army is coordinating closely with state authorities to ensure swift and effective relief efforts. Four columns from the Defence Security Corps (DSC) Centre, Kannur, and 122 TA Battalion are conducting combined rescue operations along with the National Disaster Response Force (NDRF) and state rescue teams.

Early warning issued by Centre

Union Home Minister Amit Shah on Wednesday said the Kerala government was warned as early as July 23 regarding a possible natural calamity in Wayanad due to heavy rains and nine NDRF teams were rushed to the state the same day. However, the Kerala government did not heed the early warning and also did not get alerted even by the arrival of the National Disaster Response Force (NDRF) battalions, Shah said in the Rajya Sabha. Later, Kerala Chief Minister Pinarayi Vijayan refuted Centre's charges.

Shah said an early warning was sent to the state seven days ahead of the July 30 landslide. Another warning was given on July 24 also.

“Nine NDRF battalions were dispatched on July 23 itself and three more were sent on July 30,” Shah said.

Can AI predict landslides?

As Shah put the blame on the Kerala state government for not acting to the warning issued by the authorities, there are debates on how landslides can be predicted better with the use of artificial intelligence (AI) and various state governments can act accordingly to save loss of lives and property.

India’s neighbour Nepal has used AI to predict landslides, and also in the US, researchers are using AI for the same.

Geologists at the University of California, Los Angeles (UCLA) in the US have developed a new technique that uses AI to better predict where and why landslides may occur.

The new method, described in a paper published in the journal Communications Earth & Environment, improves the accuracy and interpretability of AI-based machine-learning techniques, requires far less computing power and is more broadly applicable than traditional predictive models, according to a press release issued by UCLA.

“Many factors influence where a landslide will occur, including the shape of the terrain, its slope and drainage areas, the material properties of soil and bedrock, and environmental conditions like climate, rainfall, hydrology and ground motion resulting from earthquakes. With so many variables, predicting when and where a chunk of earth is likely to lose its grip is as much an art as a science,” the university said.

“Geologists have traditionally estimated an area’s landslide risk by incorporating these factors into physical and statistical models. With enough data, such models can achieve reasonably accurate predictions, but physical models are time- and resource-intensive and can’t be applied over broad areas, while statistical models give little insight into how they assess various risk factors to arrive at their predictions,” it added.

In recent years, researchers have trained AI machine-learning models known as deep neural networks, or DNNs, to predict landslides. When fed reams of landslide-related variables and historical landslide information, these large, interconnected networks of algorithms can very quickly process and “learn” from this data to make highly accurate predictions, UCLA said in a statement.

Using Superposable Neural Network (SNN)

Yet despite their advantages in processing time and learning power, as with statistical models, DNNs do not “show their work,” making it difficult for researchers to interpret their predictions and to know which causative factors to target in attempting to prevent possible landslides in the future.

The UCLA researchers developed an approach that could decouple the analytic power of DNNs from their complex adaptive nature in order to deliver more actionable results.

Their method uses a type of AI called a superposable neural network, or SNN, in which the different layers of the network run alongside each other — retaining the ability to assess the complex relationships between data inputs and output results — but only converging at the very end to yield the prediction.

The researchers fed the SNN data about 15 geospatial and climatic variables relevant to the eastern Himalaya mountains. The region was selected because the majority of human losses due to landslides occur in Asia, with a substantial portion in the Himalayas.

The SNN model was able to predict landslide susceptibility for Himalayan areas with an accuracy rivalling that of DNNs, but most importantly, the researchers could tease apart the variables to see which ones played bigger roles in producing the results.

The researchers’ new AI programme also requires far fewer computer resources than DNNs and can run effectively with relatively little computing power.

How Nepal is using AI

In Nepal, researchers from Australia’s University of Melbourne are helping villagers predict when treacherous landslides are about to overwhelm their homes.

Scientists from the University of Melbourne, Tribhuvan University in Nepal and the University of Florence have teamed up with the Government of Nepal and Australia’s Department of Foreign Affairs and Trade to deliver a new state-of-the-art artificial intelligence system that can analyse the significant amount of data needed to identify when rain-soaked ground is about to give way.

The forecasting system, called SAFE-RISCCS, continuously analyses satellite images of Earth taken by NASA, the European Space Agency and the Japan Aerospace Exploration Agency. Feeding SAFE-RISCCS forecasts’ into Landslide Early Warning Systems will protect people living in the shadow of landslides, ensuring alerts will be far more accurate, either days or even weeks in advance.

University of Melbourne scientist Professor Antoinette Tordesillas, who is leading the project, said about 59 per cent of Nepal was prone to landslides and had one of the world’s highest deaths per capita due to landslides.

The SAFE-RISCCS platform uses images from space, combined with a new open-access artificial intelligence tool invented at the University of Melbourne, to combine rain measurements and ever-changing ground motion data to continuously monitor and forecast the risk of a landslide at any one time in any one place.

IIT’s low-cost landslide warning system

In a research done by scientists at the Indian Institute of Technology Mandi, Himachal Pradesh, they said they have developed a low-cost landslide monitoring and early warning system. This was in 2019.

“In order to provide landslide monitoring systems of single slopes in affordable price, the Indian Institute of Technology Mandi developed a low-cost landslide monitoring and early warning system. These systems are deployed in Mandi district of Himachal Pradesh, India, and monitoring fifteen plus landslide locations. A recent case study is also discussed in this paper where these systems helped in alerting people and traffic from an impending landslide,” the authors said in their research paper.

According to a report in BBC, they told it will cost an estimated Rs 20,000 rupees to manufacture – a fraction of the cost of existing technologies.

The device was trialled in more than 20 locations in Himachal Pradesh, where landslides kill dozens of people every year. Scientists say they are hopeful that it will help drastically reduce the deaths and damage caused by these natural disasters, the report said.

An accelerometer is a type of motion sensor which measures changes in velocity. In smartphones, this is what allows people to use compass and map applications and even flip their screens horizontally or vertically. The researchers have found that with some modifications, it can be used as a low-cost early warning system for landslides, it added.