9:15 AM - 9:30 AM
[HDS07-02] Finding tiny landslide movements by clustering seismic noise data
Keywords:K-means, deep-seated landslide, slow-moving landslide, Lantai, landslide early warning, ambient noise
Under the right circumstances a seemingly harmless creeping landslide can turn deadly. Monitoring creeping landslides is essential to prepare for disaster. With seismic instruments, detailed particle motion from seismic noise data can provide critical information such as the initial direction of landslide movement, as well as threshold values of rain and earthquakes that cause movement. Finding the seismic noise of landslide movement is a necessary step to understanding the complex particle motion since other disturbances appear in seismic noise. However, this task can be challenging as there is too much data for any individual to handle and judge without bias. We propose using the cosine K-means clustering method to group seismic noise signals and then isolate groups related to landslide displacement via GPS. Our results from the Lantai landslide region in Taiwan confirm displacement associated with the frequent occurrence of some seismic noise groups during rainfall. From these groups we are able to determine the first-order characteristics of their power spectral density. Our work shows clustering methods can be effective for studying the most significant seismic noise signals and have potential for improving early warning systems.