An Adaptive Thresholding Framework for Detecting the Subsurface Movements in Landslide-prone Sites

Landslides are mainly caused by heavy rainfall, soil erosion, changes in groundwater, earthquakes, and human activities such as deforestation, shifting cultivation, mining, and urbanization. Understanding surface and subsurface ground movements is crucial for developing effective landslide monitoring and early warning systems. One of the major challenges in developing such systems is the identification of the signal patterns that trigger landslides. Due to the weak nature of the landslide signals and the high interference of external noises, accurately detecting these signals can be difficult. Existing systems often rely on decision models with fixed thresholds, for event detection which may not be suitable for different landslide-prone areas. To address this issue, this paper proposes an adaptive thresholding mechanism that uses a Linear Energy-Based Detector and Gaussian Kernel to detect subsurface activities in landslide-prone sites. The onset arrivals of the subsurface activities are calculated using a recursive STA/LTA algorithm. This adaptive strategy can be used in sites with different environmental and climatic conditions and offers greater flexibility than fixed threshold systems. Using geophone sensors, the proposed methodology is tested by conducting experiments that capture subsurface ground movements caused by human footsteps and vehicular vibrations.