Detecting Precursor Onset of Acceleration Based on Temporal Saliency and Non-Parametric Spatial Clustering of Ipta Displacements

Geotechnical land-deformation events, such as slope failures associated with landslides at mining sites, often occur without much apparent warning, thus causing loss of lives and property. Satellite-based multi-temporal interferometric synthetic aperture radar (MT-InSAR) approaches, such as interferometric point target analysis (IPTA) are commonly used to study the spatio-temporal evolution of displacement velocities in landslide-prone regions. These can be utilized in displacement-based time of failure (ToF) forecasting methods such as ‘inverse velocity’ [1], [2] to provide early alerts. However, forecast accuracy is dependent on the correct identification of the precursor date corresponding to the onset of acceleration (OoA). This is further hindered by the noise in the MT-InSAR time series, that leads to a high false-alarm rate. To provide failure alerts for mining operations, we investigated a methodology for detecting OoA precursor robust to temporal and spatial noise. OoA saliency detection, an antecedent to the inverse velocity, performs fast Fourier transform (FFT) based spectral residual (SR) to identify anomalous parts of the original sequence. This framework shall improve the application of ToF prediction in the operational monitoring of mine slopes. We evaluated its performance in the Brumhadino case while showing the effect of the existing smoothing-based velocity calculations on OoA and ToF predictions.