Full text loading...
Seismic techniques have proven to be highly effective for landslide monitoring. Here, we present a machine learning-based inversion approach for surface wave data to estimate shear-wave velocity (VS) models, which are critical for identifying weak layers and potential failure zones. By using dispersion curves represented as a function of wavelength, the model achieves consistent accuracy across depth. Trained on a large synthetic dataset of 100,000 VS models, the method also incorporates Monte Carlo dropout to quantify epistemic uncertainty, offering insights into the confidence of each prediction. The results highlight the potential of this approach for near-real-time landslide monitoring and geohazard assessment.