EUVIMEDEUVIMED entdecken
Crossref

Forecasting of GNSS Displacements Based on the Trained Long Short-Term Memory (LSTM) Network

Jakub Rados, Anna Klos

Military Technology and Science · 30.06.2026 · 75 (2) · 9-18

DOI: 10.5604/01.3001.0055.8279 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Military Technology and Science, erschienen am 30.06.2026. Verfasst von Jakub Rados und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Precise monitoring of Earth’s crustal deformation relies heavily on the analysis of Global Navigation Satellite System (GNSS) displacement time series recorded by a network of ground-based antennas. This monitoring is possible during the period in which the

GNSS station is operational. We undertake one of the first attempts to forecast daily GNSS displacements for thirteen selected stations in Europe, for which the displacements were recorded from 1996 to 2023. We use vertical displacements provided by the Nevada Geodetic Laboratory (NGL) and pre-process them thoroughly. We then apply the Long Short-Term Memory (LSTM) network, one of the deep learning approaches, and evaluate its efficiency for long-term forecasting of GNSS displacements over a two-year horizon. This network is compared with a statistical prediction method, the AutoRegressive Integrated Moving Average (ARIMA) model. We also check the impact of gap-filling techniques on forecast accuracy. The tested methods include linear interpolation, the Piecewise Cubic Hermite Interpolating Polynomial (PCHIP), and Least Squares (LS) reconstruction. We evaluate the forecast results based on the Root Mean Square Error (RMSE), the Mean Absolute Error (MAE), and the coefficient of determination (R2). We assess statistical significance using the Friedman test and the Nemenyi post-hoc test. The results indicate that the LSTM network performs better in long-term forecasting than the ARIMA model, and the Friedman and Nemenyi post-hoc tests show that these differences are statistically significant. Forecasting time series with gaps filled using the LS method provides the highest accuracy. Furthermore, the analysis of the degradation of prediction accuracy over time reveals that the LSTM model is more stable over the explored horizon. The results show the significant potential of deep learning methods in forecasting geodetic time series, which encourages their further exploration.

Zitieren

APA

Rados, J., & Klos, A. (2026). Forecasting of GNSS Displacements Based on the Trained Long Short-Term Memory (LSTM) Network. Military Technology and Science, 75(2), 9-18. https://doi.org/10.5604/01.3001.0055.8279

Vancouver

Rados J, Klos A. Forecasting of GNSS Displacements Based on the Trained Long Short-Term Memory (LSTM) Network. Military Technology and Science. 2026;75(2):9-18. doi:10.5604/01.3001.0055.8279

BibTeX
@article{Rados2026,
  author = {Rados, Jakub and Klos, Anna},
  title = {Forecasting of GNSS Displacements Based on the Trained Long Short-Term Memory (LSTM) Network},
  journal = {Military Technology and Science},
  year = {2026},
  volume = {75},
  number = {2},
  pages = {9--18},
  doi = {10.5604/01.3001.0055.8279}
}
Kostenlos starten →