Insight Into The Characteristics Of GPS Time Series In North China

As long as more and more Permanent GPS stations are installed in North China, GPS datahas been a huge accumulation for monitoring crustal deformation. We have achieved GPSobservation data for44stations in North China and nearby from late in2008to early in2013.Accompanying with other10IGS permanent stations in China mainland, Russia, we obtaincoordinate daily solutions using Bernese GPS software. To make sure the consistence of frame, wechoose8permanent stations which had not been contaminated by the Great Tohoku earthquake in2011.3.11, in Japan. With such a frame, we achieve all the daily solutions and form54groups ofcoordinate time series. We have also analysis the coseismic deformation of the Great TohokuEarthquake at stations located in North China. The coseismic results show that there are1~17mmdisplacements in the east direction and very little southward displacements.Then, we choose37time series with long time span for time series analysis in the next step.We select a fit model including trend, annual signal, semi-annual signal and offsets, together witha stochastic model constructed by white noise, to fit totally111coordinate time series in threedirections: North, East and Up. We conduct the fitting process iteratively, in order to screen out alloutliers and unreasonable offsets with several specified thresholds. Finally, we achieve estimatedparameters and residuals time series. The estimated trend parameters show that the horitzonvelocity field is uniform and the vertical velocity is diversity with15~22mm/a in the southernpart of the capital area due to the underground water leakage in North China. The estimatedseasonal signals show that the annual signals are notable but with little amplitude at most stations,on the level of0.5~2mm in the north, maximizing in Feburary to April,1~3mm in the east,maximizing in May to July,2~7mm in the vertical direction, maximizing in June to August. Thesemi-annual signal is notable too, with lower than1mm in the horitzontal and0.5~2mm verticalcomponents.According to the residuals time series resulted from the previous step, we first usecross-correlation coefficients between residuals time series to analysis spatial correlation betweenstations qualitatively, proving strong spatial correlation in the regional GPS network. Then we usestacking method or principle components analysis for spatial de-correlation with screening outcommon mode errors from the origin coordinate time series. The root mean square of extracted common mode errors by stacking are0.93mm,0.79mm and3.19mm in the separated north, eastand up directions The contribution percentage of the first principle mode are about60%,65%and55%for three directions in the principle components analysis and the corresponding spatialresponse are uniform in the network which are consistence with the common mode errors. Theroot mean square of extracted common mode errors by PCA are0.84mm,0.95~0.97mm and3.10~3.22mm in the separated north, east and up directions On the whole, the common modeerrors extracted by either stacking or principle components analysis are coherence. After screeningout the common mode errors, the spatial correlation decrease notable to almost zero and theestimated velocity uncertainties and the root mean square of the fit residuals decrease significantwith about40%, while the velocity field seldom changes.In the noise analysis chapter, we first use auto-correlation coefficients in individualcoordinate time series to analysis temporal correlation qualitatively, proving the existence of colornoise in the fit residuals and then use power spectral estimation to roughly estimate the powerindex of color noise, which show “Fractral Guassian” characteristics with power index in theinterval of-1to0. At the same time, we also estimate the power index of the common mode errors,which are approximate to-1, more colored than residuals. At last, we use maximum likelihoodestimation to estimate the characteristics of noise quantitively. The noise analysis results show thatthe fittest noise model before the common mode errors removed is flicker noise plus white noise.And the main components of the common mode errors are flicker noise. The total noise amplitudedecrease about60%and the amplitude of flick noise decrease about70%, the white noise decreaseabout20%when removing the common mode errors and by the way the random walk noise partappears. The best fit noise model after the common mode errors removed is either random walknoise plus white noise or flick noise plus random walk noise plus white noise. The results of theparameters estimation by maximum likelihood estimation show that regardless of the color noisein coordinate time series, the velocity uncertainties are overestimated. While concerning of theinfluence of color noise, the estimated velocity change little, but the velocity uncertainties arescaled2~8times to about0.3mm/a in the horitzontal and1mm/a in the vertical components.According to the horitzontal velocity field fitted from about600permanent and campaignGPS stations, we choose the Gaussian smooth radius as150km to calculate five groups of strainrates coefficients—the principle strain rate, the maximum shear strain rate, the first shear strain rate, the second strain rate and the dilatation, using the distance weighting Gaussian smoothmethod. Combining with tectonical knowledge, the main crustal deformation is left-lateral shear inthe capital area, together with expansion in Tangshan-Qinghuangdao area. The global deformationin Shanxi shear zone is right-lateral shear, but with several extrusion in the south and expansion inthe north. The deformation around Tanlu fracture zone is little, just with slight shear in the northand south. The Erdos block is very stable in the center, but with expansion and left-lateral shear inthe north, left-lateral shear and extrusion from west to east in the west. The southern part of Erdosare very complicated, with left-lateral in the north of Qingling and right-lateral in the south,expansion in the east and extrusion in the west.