This study extends the semi-automatic Isolation Forest based anomaly detection in GNSS time series towards identification of the driving factors of the detected anomalies. The method combines harmonic trend-seasonal modelling with one-step ARIMA innovations, an Isolation Forest detector with causal features of station motion and the hydrological regime, a catalogue of candidate driving events, and permutation-based statistical verification of the coincidences. The method is applied to six stations of the GeoTerrace and System.NET networks around the Dnister Hydropower Complex in 2020-2024. Non-tidal atmospheric loading is the dominant external driver of the vertical station motion, with a correlation of 0.44-0.50 in the 3–30-day band at all six stations and a 10-14% WRMS reduction after correction with the measured admittance. Significant loading events are traced directly in the daily solutions: the direction of the station displacement matches the modelled deformation in 79% of episodes. With hydrological data and loading features included in the detector feature space, the anomalies concentrate near seismicity: 21 of 22 anomalous days within the seismic catalogue coverage fall within three days of seismic events against a 47% chance level. The results form the basis for factor-based classification of GNSS anomalies in monitoring.
Brusak, I., Haidus, O., & Kuplovskyi, B. (2026). Application of machine learning methods to study the relationships between seismicity and geodynamic features of the Dnister Hydropower Complex. Geofizychnyi Zhurnal, 48(2), 60-76. https://doi.org/10.24028/gj.v48i2.350985
Brusak, I., Maciuk, K., & Haidus, O. (2025). Detection of geodynamic anomalies in GNSS time series using machine learning methods. Geodynamics, 1(38), 37-48. https://doi.org/10.23939/jgd2025.01.037
Dill, R., & Dobslaw, H. (2013). Numerical simulations of global-scale high-resolution hydrological crustal deformations. Journal of Geophysical Research: Solid Earth, 118(9), 5008-5017. https://doi.org/10.1002/jgrb.50353
Dobrovolsky, I.P., Zubkov, S.I., & Miachkin, V.I. (1979). Estimation of the size of earthquake preparation zones. Pure and Applied Geophysics, 117(5), 1025-1044. https://doi.org/10.1007/BF00876083
Haidus, O., Maciuk, K., & Brusak, I. (2024). Semi-automatic anomaly detection in GNSS time series using machine learning algorithms. In International Conference of Young Professionals «GeoTerrace-2024». EAGE. https://doi.org/10.3997/2214-4609.2024510048
Klos, A., Dobslaw, H., Dill, R., & Bogusz, J. (2021). Identifying the sensitivity of GPS to non-tidal loadings at various time resolutions: examining vertical displacements from continental Eurasia. GPS Solutions, 25, 89. https://doi.org/10.1007/s10291-021-01135-w
Liu, F.T., Ting, K.M., & Zhou, Z.-H. (2008). Isolation Forest. In 2008 Eighth IEEE International Conference on Data Mining (pp. 413-422). IEEE. https://doi.org/10.1109/ICDM.2008.17
Mémin, A., Boy, J.-P., & Santamaría-Gómez, A. (2020). Correcting GPS measurements for non-tidal loading. GPS Solutions, 24, 45. https://doi.org/10.1007/s10291-020-0959-3
Nistor, S., Suba, N. S., Maciuk, K., Kudrys, J., Nastase, E. I., & Muntean, A. (2021). Analysis of noise and velocity in GNSS EPN-repro 2 time series. Remote Sensing, 13(14), 2783. DOI: 10.3390/rs13142783
Pedregosa, F., Varoquaux, G., Gramfort, A., et al. (2011). Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 12, 2825-2830. DOI: 10.5555/1953048.2078195.
Savchyn, I., & Pronyshyn, R. (2020) Differentiation of recent local geodynamic and seismic processes of technogenic-loaded territories based on the example of Dnister Hydro Power Complex. Geodesy and Geodynamics, 11(5), 391—400. https://doi.org/10.1016/j.geog.2020.06.001.
Zyhar, A., Yushchenko, Yu., & Savchyn, I. (2023). A study of the influence of water level fluctuations on the geodynamic situation in the natural and technical geosystem of the Dnister HPP and PSPP cascade. Geodesy, cartography and aerial photography, (97), 24—31. DOI: 10.23939/istcgcap2023.97.024.