Evaluation of filtering methods for use on high-frequency measurements of landslide displacements

Displacement monitoring is a critical control for risks associated with potentially sudden slope failures. Instrument measurements are, however, obscured by the presence of scatter. Data filtering methods aim to reduce the scatter and therefore enhance the performance of early warning systems (EWSs). The effectiveness of EWSs depends on the lag time between the onset of acceleration and its detection by the monitoring system such that a timely warning is issued for the implementation of consequence mitigation strategies. This paper evaluates the performance of three filtering methods (simple moving average, Gaussian-weighted moving average, and Savitzky-Golay) and considers their comparative advantages and disadvantages. The evaluation utilized six levels of randomly generated scatter on synthetic data, as well as high-frequency global navigation satellite system (GNSS) displacement measurements at the Ten-mile landslide in British Columbia, Canada. The simple moving average method exhibited significant disadvantages compared to the Gaussian-weighted moving average and Savitzky-Golay approaches. This paper presents a framework to evaluate the adequacy of different algorithms for minimizing monitoring data scatter.

Automated summary

Displacement monitoring is crucial for managing risks associated with sudden slope failures. This paper evaluates the performance of three filtering methods in reducing scatter in measurement data and considers their advantages and disadvantages.