The system can forecast multiple hazards like landslides, earthquakes, floods, forest fires, droughts, building management system including dams and reservoirs with minor changes.
N.B. Nair
New Delhi (ISJ): A research lab in an Indian university has developed a first-of-its-kind advanced landslide warning system integrating machine learning and artificial intelligence techniques. It has a sensing system, Internet of Things (IoT) and wireless technology to ensure that the entire system works with least human intervention.
It’s sensing system reaches as deep as the bedrock itself and constantly monitors for any change in hydro-geological parameters of the soil. The system also relies on multi-layered wireless communication architecture that ensures fault-tolerant connectivity even during extreme weather.
“We have built advanced algorithms that run on the edge processor and cloud servers, which ensures optimum system performance with least to no maintenance. It is fully automated and doesn’t require any human intervention as many landslide-prone areas are inhospitable. You can change the software or change the frequency of the sensing system remotely. The whole ecosystem is built in such a manner that it requires very minimal human intervention,” Dr Maneesha Sudheer, Provost of Strategic Initiatives, Research and Innovation at Amrita Vishwa Vidyapeetham (Amrita Global University) explained.
Dr Maneesha, who is also Professor & Director of Amrita Centre for Wireless Networks & Applications told Indian Science Journal, that since the launch of the landslide detection system in 2009 in Munnar – a hill station in Kerala, they have effectively issued warnings to the state government in 2009, 2011, 2013, 2018, 2019, 2020 and the latest in 2021. She said, the system consists of more than 100 geological sensors and more than 10 wireless sensor nodes at six different locations. It runs 24×7, collecting and processing real-time data, such as rainfall, soil-moisture, pore-water pressure, vibration and tilt. It can issue landslide warnings 3-24 hours ahead of a disaster.
The system is lightweight and can adapt to network traffic dynamically by every milli-second. It works on even low internet bandwidth, in every weather pattern. The system will automatically work and decide how much data needs to be transmitted, so that it could still be able to give the warning. It also has an advanced data driven decision support system (DSS). DSS is the strongest part of the system, which looks at all the different parameters, its inter-relationship, learnt by itself and accordingly configuring the decision models. The reliability of the DSS increases dynamically as more data arrives from the field.
“It automatically transmits real-time data to the stakeholders in the government, with options to alert them through SMS or e-mail,” she added. However, the Disaster Management Rules do not allow any private entity to send alerts to the public directly.
Majority of existing landslide warning systems use only rainfall threshold as a parameter to provide alarms. This often causes false alarms due to delay in the time it takes for rainfall to infiltrate into the hills. Therefore, rainfall alone is not an accurate indicator. Existing low-cost solutions for landslide monitoring, serve as a landslide detection system for shallow landslides alone, as it fails to provide the opportunity to give enough lead time for the administration to get prepared for imminent landslides and time for evacuation. This won't serve the purpose of saving lives. They also bring in a large number of false alarms, which causes the citizen to lose trust in such systems
Amrita-designed and developed multi-level integrated system analyses other triggering factors, besides rainfall infiltration, such as pore water pressure, vibrations, movements and slope instability. It takes into account meteorological, geographical and hydrological data.
Researchers at Amrita University have collected real-time data for the past 10 + years. During this period, the rain-fall pattern has changed drastically due to climate change. In such drastic changes, the researchers dynamically learnt how the earth’s subsurface characteristics are changing with respect to the weather parameters, and based on that, refined the system to calculate the time of the landslide or when the trigger would happen. “That became automatic machine learning and artificial intelligence-based decision models for effective forecasting of imminent landslides,” said Dr Maneesha.
Dr Maneesha said, the Amrita landslide warning system has tremendously improved the reliability of the landslide warnings and thus reduced false alarms for both regional and site-specific landslide warnings. “For any early warning systems, world around, this is a major problem – false alarms and people don’t believe such alarms.” This is the only comprehensive system capable of monitoring multiple types of landslides, and delivering reliable real-time advance warnings in both regional and site-specific scale.
The system is capable of monitoring, detecting and forecasting multiple hazards – it can be used for landslides, earthquakes, floods, forest fires and droughts with minute changes. It can also be used for building management system including dams/reservoirs; only sensors change, and enhanced decision models will be used to suit the requirements of different applications.
Besides Munnar, Amrita has developed and deployed a similar real-time landslide early warning system at Chandmari, in Gangtok District of Sikkim. Custom-made for Himalayan geology, the Sikkim system comprises more than 200 sensors that measure many geophysical and hydrological parameters, such as rainfall, pore water pressure and seismic activity. The system is currently monitoring a densely populated area spanning 150 acres.
The state of Maharashtra and Uttarakhand have also evinced interest in deploying the system in landslide-prone regions in those states.
India is the second most affected nation due to global warming. An assessment by World Meteorological Organisation (WMO) pegged India’s loss due to this global phenomenon at 238 billion US dollars. According an estimate released by WMO recently, natural disasters, such as cyclones, floods and droughts have cost India around 87 billion US dollars in 2020 alone.
India’s 12.6 percent land area, barring snow covered region, is prone to landslide hazard, including the Himalayan region, Western and Eastern Ghats.
According to data given to Rajya Sabha by the Ministry of Earth Science early December, over 2000 people were killed in cyclonic storms, heavy rains, floods and landslides in 2021.
Image courtesy: Wikimedia Commons
