Defining alert thresholds using the acceleration-velocity time series of surficial and subsurface ground movement of deep-seated landslides

Abstract

Landslide forecasting is difficult due to the variability of geomechanical properties of slopes across different sites, and the effect of external factors such as rainfall and earthquake. Active landslides tend to undergo cycles of acceleration and deceleration, which if interpreted incorrectly, may lead to high false alarm rates. We outline a new method of defining alert thresholds using historical critical velocity and acceleration values of surficial and subsurface ground movement of deep seated landslides. Data from the Dynaslope Project of PHIVOLCS is used to test this method. We used a Gaussian weighted average spline interpolation to construct the velocity-acceleration time series from surficial and subsurface displacement data, and computed thresholds from the statistical interval of critical values. We find that in contrast to traditional fixed-velocity thresholds, no alerts are generated during the deceleration of slope movement. The threshold line can also be calibrated using local and site-specific values to better represent the movement behavior of a particular landslide.