Indian researchers have created an algorithm that provides an 80-94% accuracy in identifying landslide movement, as reported by Mongabay-Inda (MI). Researchers from the Rochester Institute of Technology, the University of Padova’s Machine Intelligence and Slope Stability Laboratory and a team of other researchers introduce a new method of examining slides, flows, and fails, which are the different manners in which landslides occur, as well as finding distinct patterns of landslides that were previously unidentifiable. Their algorithm does not predict landslides, but it gives necessary information, such as causes and mechanisms, to those who are in the business of predicting landslides. Current predictive models categorise different subtypes of landslides into a single group. This new method creates a comprehensive algorithm that tailors a mitigation approach based on the subtype to reduce uncertainty and bias.
As the MI report points out, in 2013, torrential rainfall in Uttarakhand caused devastating landslides and flash floods, leaving an unclear death toll in its wake. Last year, heavy rain and landslides killed at least 72 people in a single week in August in Himachal Pradesh. Deaths and monetary damages from landslides continue to occur around the world and in India as well. One statistic estimates that between 2010 and 2021, at least 3710 people in India lost their lives to landslides, while tens of thousands of people were impacted in other ways.
The Abstract of the study states the death toll and monetary damages from landslides continue to rise despite advancements in predictive modelling, as these models often miss crucial information, e.g., underlying movement types. This study introduces a method of discerning landslide movements by analyzing landslides’ 3D shapes. By examining landslide topological properties, it discovers distinct patterns in their movements. The method achieves an 80-94% accuracy by applying topological properties in identifying landslide movements across diverse geographical and climatic regions, including Italy, the US Pacific Northwest, Denmark, Turkey, and Wenchuan in China. It introduces a paradigm for studying shapes to understand underlying movements through landslide topology, which could aid landslide predictive models and risk evaluations.
In the Introduction, the researchers say that landslides cause economic damages worth 20 billion US dollars every year, and between 2004 and 2019 non-seismic landslides alone caused about 70,000 fatalities worldwide. Within the first two months of 2024, there have been reports of devastating landslides in Colombia, Southern Philippines, and Yunnan, China. Adding to this, recent studies count over one million landslide occurrences, with annual volumes estimated at 56 billion cubic meters globally, presenting a risk to 65 million people. With the increase in urbanization, global climate change, and environmental change trends, the frequency of landslides and the associated risks will keep increasing globally over time. In line with this, landslides are anticipated to evolve and remobilize with increased frequency under changing climatic conditions on a decadal scale. The algorithm’s ability to identify hazards from emerging landslides and dynamically assess impact areas is essential in averting risk to rapidly urbanizing communities and adapting to changing environmental conditions.
While India was notably missing in their sample of countries, according to the MI report, the team said that Indian landslide data is sometimes inaccessible and incomplete, which can skew a predictive model. They hoped that this paper would allow them to collaborate with agencies like GSI and the National Remote Sensing Centre (NRSC) so that their algorithm and predictive models can reach their full potential.
The MI report adds that previous attempts have been made in India at implementing different landslide subtypes in predictive models through knowledge-driven and data-driven approaches. Data-driven approaches do exist, but they have looked at landslides through a limited two-dimensional lens, while landslides are three-dimensional phenomena. By using aerial view and the elevation data of landslide sites, combined with machine learning, the team was able to create three-dimensional landslide data, giving predictive modelers a stronger foundation to build their model on. However, a lack of crucial data is not just a problem in India. Many countries do not have freely accessible, well-documented landslide datasets. Although there is still work to do in making landslide data more accessible and detailed, this algorithm could be a step in the advancement of landslide predictive modelling.