LANDSLIDE MONITOR: a real-time landslide monitoring system
Rawat, Pradeep Singh
DOI: https://doi.org/10.1007/s12665-024-11526-0
IF: 2.8
2024-04-03
Environmental Earth Sciences
Abstract:In the current era of computing, communication, and technology, hydrological, metrological, and geographical parameters supported by sensor-based systems are available to detect, monitor, and analyze natural disasters like landslides. The landslide-related information from the study area is collected in offline mode through site visits. This process of collecting data in offline mode may cause delays in prediction and proactive decision-making in real-time mode. Although landslides cannot be prevented, their impact on human life and the environment can be reduced through real-time monitoring and prediction using IoT, Cloud, and Machine Learning Technologies. This manuscript aims to present a robust, real-time monitoring system that can minimize losses and save lives. The proposed model utilizes Internet of Things (IoT) technology integrated with cloud services to monitor and analyze landslides in a specific study area. The real-time monitoring system relies on three types of parameters: hydrological, meteorological, and geographical. These parameters are used to collect and store real-time information in an IoT cloud platform. The IoT cloud information is fetched on the LANDSLIDE MONITOR application for proactive decisions. To predict landslide events in areas prone to disasters, supervised learning classifiers were employed. The prediction analysis takes into account meteorological, hydrological, and geographical factors. The effectiveness of the proposed real-time landslide monitoring system was tested and evaluated in the Varunavat hills of the Uttarkashi District in Uttarakhand, India. The performance of the system was assessed by analyzing the accuracy of the model at different levels. The major focus of the developed system includes real-time data storage of landslide-prone areas in the IoT cloud, predictive modeling, and lastly the real-time landslide responses on LANDSLIDE MONITOR. The present landslide monitoring system uses long short-term memory networks (LSTM) and gated recurrent units (GRU) for predictive modeling of the landslide events in the study area. Hence the unique contribution of the work includes the technologies integration and real-time data collection from the study area and stored in the IoT cloud. The novel contribution of the work also includes the predictive modeling of landslide events using LSTM and GRU, and study area people awareness using LANDSLIDE MONITOR for level of risk from landslide events. The accuracy rates for the alert classes, 'no threat', 'mild threat', and 'high threat' events are 96.83%, 97.07%, and 98.56%, respectively. With a mean F 1 score of 0.96 across the three classes of landslide occurrences, the proposed system demonstrates a high level of accuracy.
environmental sciences,water resources,geosciences, multidisciplinary