GNSS Time Series Non-Structural Motion Analysis

The Earth’s crust is the only directly observable circle on the Earth,and it is the region where the interactions between the Earth’s interior and exterior(including interstellar space)meet.The Earth’s crust moves and deforms under various internal and external factors.According to the influencing factors and characteristics of crustal deformation,it can be divided into tectonic deformation and non-tectonic deformation.As a multi-dimensional and high-precision crustal deformation observation technique within the geophysical field source,Global Navigation Satellite System(GNSS)is not only controlled by the large regional tectonic motion,but also particularly sensitive to the changes of local observation environment,and the coordinate time series contains rich tectonic and non-tectonic deformation information.At Present,China has basically built a continuous and mobile station observation network covering the whole Chinese continent and accumulated abundant observation data,but how to accurately extract the crustal deformation information effectively from the actual observation data containing various trends,noise,errors and periodic non-tectonic motion information still needs continuous exploration of different ideas and methods.In this paper,the empirical mode decomposition(EMD)and its several developed decomposition methods are compared and analyzed for their ability to extract periodic signals through simulated signals,and finally the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(CEEMDAN)method is selected and combined with wavelet transform for practical data processing.By analyzing the N,E and U direction time series of 10 GNSS continuous stations in different regions of mainland China,we found that not only annual and semi-annual period signals existed,but also shorter or longer period signals existed in some stations,and the vertical periodicity was more obvious than the horizontal direction,and the corresponding amplitude was larger.After the periodic term correction,the average root mean square error(RMSE)of the time series in the N,E,and U directions is reduced by 19.8%,19.0%,and 29.9%,respectively,compared with the original N,E,and U directions,indicating the effectiveness of the method for periodic term extraction.In our paper,after determining the non-tectonic deformation information extraction method,we apply it to the data processing and analysis of 24 continuous stations and 2 mobile stations in the Sichuan-Yunnan area.The results show that it is necessary to correct the deformation generated by non-constructive motion,and the CEEMDAN method can extract the periodic components with different frequencies and amplitudes adaptively according to the own characteristics of each station signal,which is also more in line with the actual situation,and the average RMSE is reduced by 21.70%,12.34%and 36.60%compared with the original N,E and U direction time series,respectively,which is a more accurate and effective method than harmonic model correction method.By using the corrected continuous station time series to simulate the mobile observations,we found that relatively reliable station motion velocities could be obtained after 5～6 years/period of observations,and the stability and reliability of the CEEMDAN method were verified by correcting the mobile station period term with the actual continuous stations at a closer distance.This provides a reference and theoretical basis for the implementation of mobile GNSS observations and the use and correction of the data.We analyze,compare and interpret the GNSS time series non-constructive deformations in terms of amplitude,phase and cross-correlation coefficients by integrating GRACE and International Mass Loading Service(IMLS)data.GNSS,GRACE and IMLS show good consistent in the vertical component,and the seasonal variation is mainly related to hydrological and atmospheric loads.The differences in the horizontal component were large,and the GNSS seasonal term amplitude was larger than that of GRACE,indicating that the GRACE resolution in the horizontal component was lower than that in the vertical component,and the whole estimation accuracy was lower than that of GNSS.The IMLS amplitude was larger than GNSS,indicating that some model data might be overestimated.All three components of GNSS have obvious annual and semi-annual period components,and the vertical component mainly shows seasonal deformation of annual period with the strongest seasonality;the horizontal component mainly shows seasonal deformation of semi-annual period,and the N component has stronger seasonality than the E component,and the two are negatively correlated with a complex correlation coefficient of-0.90.In addition,precipitation can cause instantaneous surface deformation of loading,but also due to the water In addition,in addition to the instantaneous surface deformation caused by precipitation loading,the pore pressure of the subsurface medium will also change due to the diffusion of water,thus producing some surface deformation characteristics with a lag relative to instantaneous precipitation loading.