Micro-scale Landslide Displacements Detection Using Bayesian Methods Applied to GNSS Data

Abstract

In this chapter, we evaluate the movement of 6 points near a landslide body, which were surveyed with GNSS receivers over time. We apply Bayesian inference to identify the areas on the ground with statistically significant vertical (downwards) shifts. Traditional statistical methods work well only when point displacements between different survey epochs are sufficiently large compared to the standard deviations of related coordinates. In such cases, coordinate differences of some points can be marked as potential displacements. The Bayesian analysis can help to improve discrimination when height differences, computed with respect to the first measurement epoch, are at the same order of magnitude as the uncertainties of the measures. After the application of the classical statistical test, one network point, close to the upper part of the landslide area, seemed to be more unstable than the remainder. In order to remove or validate the hypothesis of instability, the Bayesian statistical inference was applied, and all three of the upper group of points show significant shift, depending on the data prior parameters. This application shows that the Bayesian approach can be considered as an integration to classical statistical significance testing (e.g. z-test), reliably showing significance in vertical directional (i.e., downwards) coordinate shifts, thus supporting detection of movements having lower magnitude.