An Automated Learning System for Monitoring and Forecasting Rainfall Induced Landslides

Abstract

Technological evolution has contributed to new solutions, that help in reducing the impact of landslides by monitoring early warning vulnerable areas. However, issuing reliable early-warnings ahead of time is still a challenge. In this thesis, major challenges in landslide early-warnings such as (i) reliability of data for early warnings during disastrous scenarios, (ii) inadequacy of time for early-warnings (iii) extending the lifetime of the monitoring system when there is less resource availability, (iv) ensuring data from the monitoring system resilient to faults, are addressed using a data-driven newlineapproach. Amita Center for Wireless Networks and Applications-AmritaWNA has designed, developed, deployed an \IoT System for landslide monitoring early warning quot in Munnar, Western Ghats. The IoT system in the deployment area captures the realtime variability in the environment and slope using various sensors deployed at different depths. Among all the sensed information, the information about pore-water pressure at different soil layers is considered vital for the following reasons: (1) Pore-water pressure change happens slowly over time, hence there is enough time to issue an early warning when threshold limits of the piezometer sensors are overcome; (2) the slope accounts for a slide when the soil loses its cohesion and pore-water pressure is indirectly proportional to soil cohesion. This study focuses on understanding the complex behavior of two prominent landslide triggers, rainfall and pore-water pressure leading to slope instabilities using the rainfall and piezometer data from the IoT system. The following models are implemented to achieve reliable, resilient, and real-time landslide early warnings. i) Methodology for extending the life-time of the IoT system through intelligent algorithms in the EDGE layer of IoT system. ii) Implementing methodologies to ensure data resilient to faults using a quotFault Diagnostic newlineSystemquot. iii) Spatio-temporal variation of pore-water pressure and localization of vulnerable..