EUVIMEDEUVIMED entdecken
Crossref

Assessing the potential of generalized Gauss–Markov plus white noise as a parameter-unified noise model for GNSS coordinate time series

Lei Peng, Yueyang Huan, Huachao Yang, Guobin Chang, Nijia Qian, Hefang Bian, Lubin Chang, Chao Chen

Measurement Science and Technology · 28.08.2026 · 37 (34) · 346304

DOI: 10.1088/1361-6501/ae96cb ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 24.08.2026. Verfasst von Lei Peng und 7 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Gefördert von: Basic Research Program of Jiangsu, National Natural Science Foundation of China, Fundamental Research Funds for the Central Universities.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Accurate stochastic noise modelling is essential for reliable velocity uncertainty estimation from GNSS coordinate time series. This study evaluates the generalized Gauss–Markov plus white noise model (GGM + WN) as a parameter-unified representation. We analysed 753 daily east, north, and up coordinate time series from 251 IGS stations using maximum-likelihood estimation. GGM + WN was compared with eight candidate models using logarithmic power spectral density (PSD) residuals, Akaike information criterion (AIC) selection, Akaike weights, velocity-uncertainty comparisons, time-span grouping, and Bayesian information criterion (BIC) sensitivity analysis. GGM + WN produced the largest proportion of best PSD fits for the east, north, and up components (44.2%, 48.2%, and 37.8%, respectively). Under strict AIC selection, GGM + WN was selected for 57.1% of all series, with stronger support for the horizontal components (64.1% east; 69.3% north) than for the vertical component (37.8%). GGM-selected cases with numerically identical matched GGM + WN fits were classified as de facto GGM + WN-optimal (GGM + WN-eff); together with strict GGM + WN selections, they accounted for 79.8% of the dataset. This diagnostic proportion does not replace strict information-criterion selection. AIC support for GGM + WN was higher in longer observation-length groups, particularly for horizontal components. Under BIC, the strict and combined proportions were 31.7% and 45.9%, respectively. For 152 series in which GGM + WN was neither strictly AIC-optimal nor GGM + WN-eff, velocity-uncertainty estimates remained close to those of the AIC-optimal model, with a median relative deviation of −0.027 and a 95th-percentile absolute deviation of 0.183. These results support GGM + WN as a practical parameter-unified representation for many long GNSS coordinate time series, while showing that model choice depends on the information criterion, coordinate component, and observation length.

Förderung

  • Basic Research Program of Jiangsu
  • National Natural Science Foundation of China
  • Fundamental Research Funds for the Central Universities
  • China Railway Design Corporation Limited

Zitieren

APA

Peng, L., Huan, Y., Yang, H., Chang, G., Qian, N., Bian, H., Chang, L., & Chen, C. (2026). Assessing the potential of generalized Gauss–Markov plus white noise as a parameter-unified noise model for GNSS coordinate time series. Measurement Science and Technology, 37(34), 346304. https://doi.org/10.1088/1361-6501/ae96cb

Vancouver

Peng L, Huan Y, Yang H, Chang G, Qian N, Bian H, et al. Assessing the potential of generalized Gauss–Markov plus white noise as a parameter-unified noise model for GNSS coordinate time series. Measurement Science and Technology. 2026;37(34):346304. doi:10.1088/1361-6501/ae96cb

BibTeX
@article{Peng2026,
  author = {Peng, Lei and Huan, Yueyang and Yang, Huachao and Chang, Guobin and Qian, Nijia and Bian, Hefang and Chang, Lubin and Chen, Chao},
  title = {Assessing the potential of generalized Gauss–Markov plus white noise as a parameter-unified noise model for GNSS coordinate time series},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {34},
  pages = {346304},
  doi = {10.1088/1361-6501/ae96cb}
}
Kostenlos starten →