Subscribe
Just a fortnight before the deadly landslide in Wayanad this July, the central government activated a new landslide early warning system (LEWS) for three regions in the country. The system, developed by the Geological Survey of India (GSI), was designed to predict the likelihood of landslides based on rainfall data.
The early warning system (EWS) is currently operational in landslide-prone areas like Kalimpong and Darjeeling in West Bengal and Nilgiris district in Tamil Nadu; the government aims to expand it nationwide by 2030. The system is crucial, as India experiences a high number of rainfall-induced landslides each year.
An EWS generates and shares timely information to mitigate the impacts of disasters like landslides, involving risk knowledge, monitoring and warning, dissemination and communication and response capability. For example, EWS can forecast or predict landslides regionally or on specific slopes.
GSI focuses on regional systems and is working to operationalise a rainfall-induced LEWS prototype, which was developed between 2016 and 2021 by nine partners from India, the United Kingdom and Italy as part of the LANDSLIP project.
To make a forecast, scientists must first create susceptibility maps, which estimate the likelihood of landslides in a given area after considering the local terrain, soil type and environmental conditions. These maps only show where landslides are likely to occur but do not provide an estimate of the magnitude or expected duration.
“The susceptibility tells us if landslides are more or less likely to occur in the future. A landslide susceptibility map identifies areas subject to landslides and measures them from low to high. The most basic element of landslide susceptibility is the steepness of a slope,” Bruce Malamud, Professor at Durham University, England, who co-led the LANDSLIP project with the British Geological Survey while at King’s College London, told Down To Earth (DTE).
Many other factors can affect landslide risk. “Examples include an area’s geology, soil types, vegetation types and human infrastructure like roads and buildings. Prior knowledge of where landslides have occurred is key to better understanding landslide susceptibility,” Malamud added.
In 2014, GSI launched the National Landslide Susceptibility Mapping, which covers 0.42 million square kilometres of landslide-prone areas in the country on a scale of 1:50,000. A unit of distance on the map equals 50,000 units of distance on the ground. It considers slopes, vegetation, slope-forming material and geology. LANDSLIP will use this database as a base map and update it as it receives new information.
The second step involves forecasting. The project created a short-term (24-48 hours) rainfall forecast and a medium-range (next 10 days) forecast based on weather patterns in the Nilgiris and Darjeeling.
To make short-term forecasts, researchers use models to calculate rainfall thresholds, determining how much rain is needed to trigger a landslide. Originally developed in Italy, these models were adapted for use in India. For a region with specific geology and geography that has experienced a landslide, the model considers the intensity and duration of rainfall.
“The model looks at the intensity of the rain over a certain period and whether it has caused landslides in the past. It also looks at the amount of rain over a period in this region. Ten millimetres of rain in one hour is more likely to cause landslides than 10 millimetres over five days. This is because water enters soil pores, which are taken by the soil and roots,” he explained.
If researchers lack data for a specific place, they use information from similar regions. If both areas experience similar rainfall, the model can estimate landslide risks.
For medium-term forecasts, the LANDSLIP team identified 30 weather patterns linked to landslides. These patterns, covering the main monsoonal and non-monsoonal types, were tested for their ability to explain rainfall differences, according to a 2019 study published in the International Journal of Climatology.
Malamud said these weather patterns are a real innovation of LANDSLIP. This, combined with susceptibility maps, determines a landslide warning level from ‘low’ to ‘very high’.
SaveTheHills, a Kalimpong-based non-governmental organisation, collaborated with LANDSLIP to provide landslide details in the Darjeeling-Sikkim Himalayas. “GSI shared forecast bulletins with district magistrates, who sometimes shared them with grassroots organisations in an experimental mode during the monsoon since 2020,” Praful Rao, president of SaveTheHills, told DTE.
The LANDSLIP project also roped in social scientists to refine the prototype. Anshu Orga, an assistant professor at the Indian Institute of Technology, Delhi’s School of Public Policy, visited the pilot sites several times while working as a postdoctoral researcher at King’s College London. She was involved with the project from 2018 to 2021 and was tasked with determining how to effectively govern and communicate early warnings.
“We actively conversed with the district offices in both study sites, receiving their feedback for the various forecast bulletins at the prototype stage,” she told DTE. As India does not have a precedence of landslide early warnings, it was important to understand governance and translation of the bulletin in conversation with the users, she added.
Since GSI took over the prototype to operationalise it, they have made some minor adjustments in the rainfall thresholds and will continue to do so as they get newer data along with feedback. “The system will be fine-tuned accordingly as and when required following the similar best practices followed elsewhere in other countries,” said Saibal Ghosh, deputy director general, GSI.
Operationalising a landslide early forecasting system is time-consuming and could take 8 to 20 years to map the environment, collect enough data on rainfall and past landslides and understand forecasting uncertainties, according to Malamud. The problem, he said, is not because we do not lack the knowledge to forecast if an area faces a higher or lower susceptibility to landslides. The real issue, according to him, is human power, money and expertise.
Ghosh also stressed that this is a time-consuming process by drawing parallels between cyclone and landslide prediction. “In 1999, the Great Odisha Cyclone resulted in 14,000 fatalities. In 2024, we have smart cyclone warning systems and see hardly any loss of lives. But this journey took 25 long years,” he explained.
Landslides, he explained, are a much more difficult and complex system and the country has only recently entered this arena. Ghosh added that perfecting the system for improved accuracy will take time.
Experts have identified one major issue: Data on rainfall and past landslides. Ghosh pointed to the same challenges, including fewer rainfall stations, automatic weather stations and automatic rain gauge stations in the hills. Another challenge in making accurate forecasts, according to Ghosh, is erratic patterns and increased frequency of high-intensity rainfall within a smaller area or catchment.
Past landslide occurrences must also be studied in order to improve forecast accuracy. Scientists need to have answers to four main questions: Where and when landslides occur, their size and their type, said Malamud. A landslide early forecasting system, he said, is only going to be as good as the data.
“The GSI, as the nodal agency for landslides in India, has extensive data. They can not be everywhere, so they need help collecting data on small landslides or those that occur in sparsely populated areas,” said Malamud. “One can also use remote sensing images from satellites to collect landslide occurrence information, something the India National Remote Sensing Centre is doing. But there are often clouds or the data is very coarse resolution, making it hard to identify landslides systematically.”
But early warning are not enough, pointed out Rao, asking what will follow these predictions. “If the forecast sounds warning in a region like Darjeeling, how will we carry out evacuations in the whole area? Where will people go?” he asked.
This is the second part of a series on early warning systems for landslides. Read the first part here and the final part here.