Spatiotemporal noise in GPS position time-series from Crustal Movement Observation Network of China. Issue 3 (29th November 2018)

• Record Type: Journal Article
• Authors: Wang, Wei; Qiao, Xuejun; Wang, Dijin; Chen, Zhengsong; Yu, Pengfei; Lin, Mu; Chen, Wei
• Is Part Of: Geophysical journal international. Volume 216:Issue 3(2019)
• Page Start: 1560
• Page End: 1577
• Publication Date: 2018-11-29
• Subjects: Satellite geodesy -- Space geodetic surveys -- Asia -- Time-series analysis; Geophysics -- Periodicals; 550
• DOI: 10.1093/gji/ggy506
• Languages: English
• ISSNs: 0956-540X

Abstract

SUMMARY: We analyse noise characteristics of spatial and temporal correlation of 260 continuous GPS sites from Crustal Movement Observation Network of China (CMONOC). These data sets were mainly collected between 2010 and 2016, with an average of 6 yr of position time-series. In the functional analysis, a clear regional dependence of seasonal movements has been observed and other significant periodic signals are detected nearby the GPS draconitic period and its harmonics. The distribution of these periodic signals shows a spatial correlation, along with non-negligible local inconsistencies. In the stochastic analysis, impacts of the periodicities on the noise assessment have been investigated and Maximum likelihood estimation is used to study noise properties for deseasonalized residual time-series having the seasonal signals removed and filtered residual time-series having other periodic signals removed further. We demonstrate that for both solutions, the flicker noise is thought to be the dominant time-correlated noise and velocity uncertainties may be underestimated 8–10 times if assuming a pure white noise. Ignoring the periodicities could bias the estimation of noise amplitude, spectral index and velocity uncertainty. After removing the periodic signals, the median flicker noise magnitude shows an average of ∼10 per cent reductions and the noise process shifts closer towards white noise. A correlation between the index variations and RMS variations has been observed, indicating that the index varies more significantly for sites with more periodic signals removed. Besides, the spectral index in the vertical component has a better spatial correlation than that in the horizontal and the spatial distribution of the index of deseasonalized solutions seems to correlate well with the amplitudes of seasonal signals, probably implying common sources of spatial variations of these characteristics. Furthermore, analyses of intersite correlations indicate that the correlation induced from the periodic signals displays a similar pattern to the deseasonalized solutions, confirming that the period pattern is spatially correlated and can induce time-series correlations. Finally, the stochastic processes of the common mode noise are predominantly featured by spatially correlated flicker and white noise over a wide range, consistent with previous results.