Development of IoT Slope Monitoring System and its Applications for Kratu-Patong Road Landslide in Phuket, Thailand

Abstract

This paper presents an on-going development of Internet-of-Things (IoT) slope monitoring for landslide early warning system in Thailand. The current system employs a variety of sensors, namely MEMs-based tensiometers, piezometers, soil moisture sensor, tiltmeter, in-placed inclinometer and tipping bucket raingauge, all connected to Arduino-based microcontroller which relied on Narrowband, NB-IoT, protocol for data transmission to the cloud server. A specially designed application platform was developed to convert the sensor readings to engineering unit and ultimately geotechnical parameters, such as factor of safety, which enable engineers to readily understand the situation and make an informed-decision based on such parameters. A weighted approach was proposed in calculating the overall landslide hazard level based various kinds of sensor readings. A case history of Kratu-Patong Road Landslide in Phuket, Southern Thailand, taking place in Year 2022 was presented to demonstrate how the developed IoT system was used real-time together with geotechnical analysis to aid in traffic management during the critical time. The warning event primarily stemmed from spikes in slope movement, spurred by heightened traffic intensity. Rapid slope movement during the incident was characterized by a tilting magnitude of -2 to 1.2 degrees and a velocity ranging from -1.7 to 1.8 degrees per hour. Notably, the calculation of the warning index based on tilting magnitude provides a continuous warning message, in contrast to an intermittent message based on tilting velocity. The tensiometer effectively detected the decrease in suction caused by slope movement, while the piezometer only registered changes in pore-water pressure when the groundwater table rose above the measurement point. Finally, an Artificial Neural Network (ANN) model was used to predict the pore-water pressure at different depths based on 5 rainfall parameters, namely, 5-min, 1-hour, 1-day, 3-day and 7-day antecedent rainfalls. The model demonstrated satisfactory predictive accuracy (R² = 0.644, RMSE = 3.637 kPa), offering promising potential for integration with the IoT platform in the future.