Research on warning threshold of horizontal displacement of landslide monitoring based on clustering idea

Unreasonable threshold setting is an important factor leading to high false alarm rate of landslide monitoring and warning. In order to obtain accurate threshold warning intervals for most landslide warning levels, we taken GNSS horizontal displacement parameters as an example, selected 60 successful landslide warning cases in Ankang City, Shaanxi Province, and extracted the corresponding GNSS horizontal displacement changes. After data filtered, applied parameter standard normalization and K-means clustering algorithm to obtain four centroid positions adaptively, thus divided five landslide monitoring threshold warning level intervals. Used this threshold level standard, we judged the threshold warning level of the remaining 5 successful landslide warning cases and obtained good warning results. The research showed that the K-means clustering method can obtain the monitoring and warning threshold level suitable for most landslide hazard points, which was conducive to reduce the false alarm rate of landslide disaster threshold warning.