In the past 20 years, many scholars have used GNSS coordinate time series to carry out different aspects of geoscience research work, mainly including periodic signal analysis in GNSS coordinate time series (Blewitt et al., 2001; Ray et al., 2008; Klos et al. al., 2018a, 2018b), stochastic model research (Williams et al., 2004; Langbein, 2008), environmental loading model (van Dam et al., 1994, 2001, 2010; Jiang et al., 2013a, 2014; Gu et al., 2017; Chanard et al., 2018; Yuan et al., 2018; Klos et al., 2018c) and other aspects. GNSS coordinate time series contains a lot of information about the elastic deformation of the earth caused by environmental factors such as the atmosphere, seabed pressure, snow, ice, surface water, and groundwater (Farrell, 1972; Li et al., 2020). Existing studies have shown that environmental loading factors including ATML, NTOL and HYDL are the main cause of nonlinear changes in GNSS time series (Jiang et al., 2010; Sun et al., 1995). By establishing the surface loading model and calculating the elastic deformation displacement sequence of GNSS stations, the geophysical influence mechanism at the station position can be effectively studied. At the same time, it compares and analyzes the characteristic changes of nonlinear deformation in the GNSS coordinate time series before and after the surface loading correction, and detects the effects of various surface loadings on the GNSS coordinate time series. This is essential to correctly understand the geophysical sources of nonlinear changes in GNSS coordinate time series.
For ATML model, van Dam et al. (1994) analyzed the elevation time series of 19 global positioning system (GPS) stations in the northern hemisphere. They concluded that when the atmospheric loading correction was applied, the variance of the GPS elevation time series could be reduced by 24%, and the atmospheric loading displacement series fluctuates greatly in high latitude regions. van Dam et al. (2010) also found that when the influence of terrain factors on ATML was taken into account, the nonlinear motion of GPS elevation time series could be better explained. Li et al. (2020) calculated the surface displacement caused by five atmospheric loading products and compared them with the coordinate time series of 596 global GNSS stations. The results showed that the ERA-Interim model performed best in reducing the dispersion of GNSS coordinate time series, and for all five atmospheric models, the correction effect of inland stations was better than that of island stations.
For HYDL model, van Dam et al. (2001) analyzed the elevation time series of 147 GPS stations around the world. They found that HYDL can cause the RMS value of GPS elevation time series increase by 8 mm. After hydrological loading correction, the dispersion of the GPS height time series at 92 stations decreased. Dill and Dobslaw (2013) used a global high-resolution hydrological model. The study concluded that after corrections of ATML and NTOL, HYDL could explain up to about 54% of the vertical nonlinear deformation of GPS stations. Among these loadings, HYDL is an important factor that causes periodic vertical deformation of GPS stations (van Dam et al., 1998; Liao et al., 2010; Jiang et al., 2014). The increase of HYDL on land will cause the ground to sink, and the vertical position of GPS stations will move downward. The weakening of HYDL on land will cause surface to rebound, which will cause the vertical position of GPS stations to move upward.
For NTOL model, Williams and Penna (2011) studied the NTOL of 17 GPS stations at coastal areas of Europe. The study showed that the NTOL of these stations were of similar magnitude to ATML, and the sum of the two types of environmental loadings could explain about 20–30% of the RMS value of GPS elevation time series. On a global scale, the maximum NTOL calculated using the latest oceanic bottom pressure model could reduce the RMS value of GPS elevation time series by 3.7 mm. RMS values of about 70% global stations were reduced (van Dam et al., 2012).
For sum environmental loading (SUML) model, Dong et al. (2002) quantitatively calculated the impact of ATML, NTOL, HYDL and the sum of the three on 128 GPS stations around the world. The study indicated that SUML can explain about 40% of the seasonal signals in GPS elevation time series. Jiang et al. (2013b) studied the causes of nonlinear changes in 11 IGS stations in China, and believed that ELCs could effectively weaken the annual amplitude of GPS elevation time series, but the correction effect on horizontal component of GPS stations was not ideal. The displacement of the GNSS stations caused by the environmental loadings can reach several centimeters and produce a seasonal signal with an amplitude of millimeters, which not only has a significant impact on the estimated speed of these stations (Johnson et al., 2017), but also deviates the establishment of the earth reference frame (Freymueller, 2009; Zou et al., 2014). By correcting environmental loadings, the nonlinear deformation in GPS coordinate time series can be reduced to a certain extent. Taking into account the differences of the surface loading data sources used by different agencies, the current surface loading models can explain up to 50% of the nonlinear deformation of GNSS stations in the ideal case, and less than 20% in the horizontal direction (Yan et al., 2009; Xu et al., 2017). Therefore, the ELCs cannot completely eliminate the nonlinear deformation in GNSS coordinate time series.
At present, many research institutions around the world provide environmental loading products, including GFZ (http://rz-vm115.gfz-potsdam.de:8080/repository) (Dill and Dobslaw., 2013), EOST (http://loading.u-strasbg.fr/) in Strasbourg, France (Mémin et al., 2020) and IMLS (http://massloading.net/#Download). In addition, softwares such as QOCA (Dong et al., 2002) and OMD (Jiang et al., 2013a) can also calculate the ground displacement caused by environmental loadings. Due to different measured data (such as global atmospheric pressure, ocean bottom pressure, land water reserves), different geophysical models and methods, the environmental loading displacement at the same stations obtained by the above institutions and softwares may be quite different. For example, Andrei et al. (2018) calculated the 15-year coordinate time series of the Finnish Antarctic research station, namely ABOA, and compared the environmental loading products provided by GFZ, EOST and IMLS. The study found that the non-tidal oceanic loadings given by GFZ and EOST were significantly different.
Therefore, this study selects 21-year coordinate time series of 631 International GNSS Service (IGS) reference stations around the world, and mainly compares 6 kinds of ATML products, 7 kinds of HYDL products and 5 kinds of NTOL products provided by EOST, GFZ and IMSL. We analyze the effects of various models on GNSS coordinate time series, then select the optimal three environmental loading models, and further explore their effects on periodic signals in GNSS coordinate time series.