Figure 2 shows the cumulative contribution rate of principal component (PC) eigenvalues, where the horizontal axis denotes the order of PCs and the vertical axis represents the percentage of the cumulative contribution rate of principal component eigenvalues. If we arrange the eigenvectors so that the eigenvalues are in descending order, the first few PCs represent the biggest contributors to the variance of the network residual time series, usually related to the common source time function (Dong et al., 2006). As shown in Fig. 2, the cumulative contribution rate of the first three principal components obtained by the PCA method is 67.11%, and the higher-order PCs are much less than the first few PCs, usually related to local or individual site effects.
The filtered results are obtained by subtracting the corresponding common mode error from the coordinate time series and the residual time series. To quantify the influence of the common mode error derived from PCA on the time series of the station coordinates, Table 1 shows the variations in standard deviations (stds) of residuals before and after filtering. After subtracting the common mode error, the stds of the residuals of all stations are reduced significantly, which indicates the good performance of the PCA.
Station name Before filtering/mm PCA filtering/mm Improvement
rate of std/%
ALRT 6.94 4.18 39.75 QAQ1 5.81 3.63 37.43 SCOR 6.13 4.11 32.98 THU3 6.70 3.67 45.16 KELY 6.11 3.97 35.07 asky 5.8 3.37 41.87 blas 6.70 3.54 47.22 dane 5.89 3.52 40.25 dgjg 5.21 3.11 40.40 dksg 5.90 3.30 44.05 gmma 6.48 3.63 43.92 grok 6.67 3.72 44.24 hel2 6.40 4.27 33.33 hjor 5.85 2.76 52.86 hmbg 4.81 2.79 41.87 hrdg 6.60 3.35 49.30 jgbl 6.72 4.02 40.17 jwlf 7.12 4.01 43.74 kaga 8.14 5.86 27.98 kagz 5.87 2.99 49.09 kbug 7.65 5.58 27.13 kmjp 6.91 4.12 40.42 kmor 6.90 3.84 44.31 ksnb 6.13 3.61 41.00 kuaq 5.98 3.56 40.48 kull 6.15 3.98 35.16 kulu 6.04 3.56 41.00 lbib 5.35 2.49 53.41 lefn 6.45 3.80 41.01 lyns 6.84 3.76 45.06 marg 6.32 3.63 42.58 mik2 5.94 3.55 40.23 msvg 5.93 3.43 42.08 nnvn 5.21 3.94 24.35 nrsk 6.61 4.29 35.07 plpk 6.54 3.95 39.58 qaar 6.25 5.11 18.26 rink 6.37 4.41 30.77 scby 6.76 4.87 28.04 senu 6.30 4.27 32.19 srmp 5.79 3.32 42.70 timm 7.23 4.41 38.99 treo 7.01 4.90 30.09 utmg 5.44 3.69 32.14 vfdg 5.59 3.48 37.80 wthg 5.64 2.71 52.03 ymer 6.16 3.35 45.63
Table 1. Comparison of residual stds before and after filtering at each station
Figure 3 shows the vertical velocity field of stations estimated using CATS software. As shown in Fig. 3, all stations in the coastal area of Greenland rise. The stations near the outlet glacier have higher speeds, such as HEL2 near Helheim (HH), KAGA near Jakobshavn Isbrae (JI), and KUAQ near Kangerdlugssuaq (KL); the station speeds reach 16.3 mm/a, 19.9 mm/a, and 18.7 mm/a, respectively. These fast-moving glaciers and the basin formed by Petermann Glacier (PG) account for approximately one-fifth of the Greenland ice sheet. For Jakobshavn Isbrae, the ice mass loss rate is (22±2) km3/a, as observed by laser altimetry data in 2006–2009 (Khan et al., 2010), and this area’s loss rate is still accelerating. For the Helheim and Kangerdluqssuaq Glaciers, the mass loss caused by dynamic changes in glaciers increased significantly during 2005, and warm ocean and air temperatures further contributed to the thinning of glaciers (Khan et al., 2014). The velocities of stations in the northwest and northeast are relatively small, with an average speed of approximately 7.0 mm/a. The minimum vertical velocity occurs in the eastern region with an average of 3.4 mm/a, in which the speeds at the LBIB and HMBG are 3.2 mm/a and 2.8 mm/a, respectively. Overall, the velocity field of the GPS stations obtained in this paper is consistent with the results of Bevis et al. (2012) and Wake et al. (2016), although there are still differences for some stations. This difference is mainly because the data processing strategy is not exactly the same, and longer time series are used in this paper which can better cover the time span of tide gauge observations.
Accurate estimation of the vertical velocity caused by the current mass change in the Greenland ice sheet is significant for predicting global sea level change. To obtain the vertical deformation rate caused by the current ice and snow mass change, the impact of GIA should first be corrected. The separation of the elastic deformation rate and GIA rate can be achieved by using GIA models and differencing between GPS and gravity data (Nielsen et al., 2014; van Dam et al., 2007). Currently, the commonly used GIA global models mainly include ICE-3G (Tushingham and Peltier, 1991), ICE-4G (VM2) (Peltier, 1994), ICE-5G (VM2) (Peltier, 2004), ICE-6G_C (VM5a) (Peltier et al., 2015), ICE-4G+RF3L20 (β=0.4) (Wang et al., 2009, 2010), Paulson07 (Paulson et al., 2007), and the Geruo13 model (Geruo et al., 2012). The GIA models that show good performance in Greenland mainly include the GREEN1 model (Fleming and Lambeck, 2004), the Huy2 model (Simpson et al., 2009), and the Huy3 model (Lecavalier et al., 2014). Since the GRACE data in this paper used the Geruo13 model for GIA correction, the uplift rate of GPS stations is also corrected using this model.
The Geruo13 model is a global grid model with a resolution of 1°×1°. The cubic spline interpolation method is used to obtain the velocity field caused by GIA at the GPS station, as shown in Fig. 4. The overall GIA speed in Greenland is low, with an average of only 1.6 mm/a. However, the largest contribution to observed mass trends from GIA is in the north-east of Greenland with a maximum of 7.4 mm/a at LEFN station. This is also a region where there is a significant disagreement between different GIA models (Wake et al., 2016). Using the above velocity field to correct the GIA of the velocity field obtained by GPS observation, the vertical displacement velocity field caused by the current ice and snow mass change can be obtained. Due to the same GIA model used in GPS and GRACE data processing, the comparison between GPS and GRACE derived speed is more reasonable and convincing.
Figures 5 and 6 present the velocity fields of stations obtained by GPS, GIA and GRACE. The GIA-corrected vertical velocity of the GPS is in good agreement with the vertical velocity obtained by GRACE, and the correlation coefficient is 0.7. However, there are also regional differences between the two kinds of velocities. For stations in the northeast (LEFN, YMER, JGBL, BLAS, and NRSK), the GIA-corrected speed of GPS and the speed of GRACE are in good agreement, so these stations are considered to be less affected by the current ice and snow mass changes. The stations in the east (DGJG, DANE, and WTHG) are less affected by GIA (less than 1.0 mm/a), and the GPS and GRACE speeds are also low. For stations in the southeast and west (KUAQ, MIK2, HEL2, KAGA, and SRMP), because their locations are close to outlet glaciers, the large mass loss of these outlet glaciers in recent years resulted in the higher speed, while the impact of GIA is small. The GPS results of these stations are quite different from the GRACE results because GRACE indicates large-scale mass movement, while GPS stations are sensitive to regional mass changes. In addition, Khan et al. (2016) pointed out that the impact of GIA in southeastern Greenland is underestimated. Stations in the south (QAQ1, SENU, TIMM, etc.) are also less affected by GIA (1.0–2.5 mm/a). The average vertical lifting speed caused by the ice and snow mass changes is 6.0–7.0 mm/a.
Figure 5. GIA-corrected GPS speed field and GRACE speed field (red arrows represent GPS results, and blue arrows are GRACE results).
Figure 7 shows the monthly absolute sea level at NUUK, QAQO, SCOR, and THUL derived from the tide gauge stations and altimetric products of AVISO, where GPS-based vertical velocities at tide gauge stations were used to convert relative sea level to absolute sea level. Except for NUUK, the other 3 tide gauge stations can provide observations for over 10 years, which is highly valuable for research on sea level change. Although the tide gauges show slightly higher sea level variability than the altimetry data, they have a good correlation with each other, except for THUL. For most areas of Arctic region, due to the presence of sea ice, it is difficult to derive the sea level from satellite altimetry in winter, especially in northwest Greenland, which means continuously operating tide gauge has unique advantages in monitoring sea level changes around Greenland.
Figure 7. Comparisons of monthly absolute sea level changes derived from tide gauge and satellite altimetry.
Figure 8. Absolute sea level trends at four tide gauge stations in Greenland (red arrows represent absolute sea level trend derived from tide gauge, and blue arrows are sea level trend derived from satellite altimetry).
Figures 8 and 9 show the absolute sea level trend at four tide gauge stations in Greenland. Due to vertical crustal movements in Greenland, it is necessary to convert relative sea level to absolute sea level by introducing GPS vertical velocity during the time span of tide gauge. In addition, it must be pointed out that the GPS vertical speed at NUUK is interpolated by inverse distance weighted due to the lack of NUUK GPS station in this paper. The monthly gridded products of absolute sea levels of AVISO were interpolated to the locations of tide gauges to better compare with tide gauge observations.
The GrIS holds 7.2 m of sea level equivalent and rising temperatures have led to accelerated mass loss in recent decades (Aschwanden et al., 2019), and an accumulated global sea level rise of 12 mm for GrIS from 1992 to 2016 was presented (Forsberg et al., 2017). However, regional eustatic movements along the coasts of Greenland are quite complex as shown in Figs 7–9. Notably, there is a dramatic decline in absolute sea level at THUL, which is located in northwest Greenland. Rose et al. (2019) also found a strong negative trend in northern Baffin Bay using radar altimetry. For SCOR and QAQO, the absolute sea level trend derived from tide gauge is also in good agreement with that from altimetry. Comparatively speaking, the sea level difference at NUUK derived from tide gauge and altimetry is non-ignorable. Spada et al. (2014) presented an inferred rate of sea level change of 1.9 mm/a at NUUK using tide gauge records from 1958 to 2002, however, it should be noted that the time period is different in this paper. Given the fact that only 4 years of NUUK tide gauge observations are available and the good fit between tide gauge and satellite altimetry during the 4 years, the sea level fall at NUUK derived from satellite altimetry in 2007–2018 can be taken as a more reasonable result. Spada et al. (2012) found a pattern of regional sea level variations along the coasts of Greenland with measurements from ice cloud and land elevation satellite (ICESat) for the time period between 2003 and 2008, and argued that the regional sea level fall along the coasts of Greenland was strongly anticorrelated with vertical movements. This kind of anticorrelation can also be roughly found in Figs 5 and 8 at THUL, SCOR and QAQO. The fingerprints of sea-level rise is believed as the major driver of regional variations in sea levels (Hsu and Velicogna, 2017; Mitrovica et al., 2001). Sea level fingerprints (SLF) is characterized as changes of near-field sea level fall and far-field sea level rise in areas of intense ice mass loss such as Patagonia, coastal Alaska, the Amundsen Sea sector of West Antarctica, and the GrIS (Adhikari et al., 2019). By utilizing sea level observations from tide gauge stations and satellite altimetry, this research better illustrates the sea level changes in different coastal areas of Greenland, and can be further applied to the analysis of ice sheets and glaciers changes.
Evaluation of vertical crustal movements and sea level changes around Greenland from GPS and tide gauge observations
- Received Date: 2020-08-27
- Accepted Date: 2020-09-29
- Publish Date: 2021-01-25
Abstract: To better monitor the vertical crustal movements and sea level changes around Greenland, multiple data sources were used in this paper, including global positioning system (GPS), tide gauge, satellite gravimetry, satellite altimetry, glacial isostatic adjustment (GIA). First, the observations of more than 50 GPS stations from the international GNSS service (IGS) and Greenland network (GNET) in 2007–2018 were processed and the common mode error (CME) was eliminated with using the principal component analysis (PCA). The results show that all GPS stations show an uplift trend and the stations in southern Greenland have a higher vertical speed. Second, by deducting the influence of GIA, the impact of current GrIS mass changes on GPS stations was analysed, and the GIA-corrected vertical velocity of the GPS is in good agreement with the vertical velocity obtained by gravity recovery and climate experiment (GRACE). Third, the absolute sea level change around Greenland at 4 gauge stations was obtained by combining relative sea level derived from tide gauge observations and crustal uplift rates derived from GPS observations, and was validated by sea level products of satellite altimetry. The results show that although the mass loss of GrIS can cause considerable global sea level rise, eustatic movements along the coasts of Greenland are quite complex under different mechanisms of sea level changes.
|Citation:||Jiachun An, Baojun Zhang, Songtao Ai, Zemin Wang, Yu Feng. Evaluation of vertical crustal movements and sea level changes around Greenland from GPS and tide gauge observations[J]. Acta Oceanologica Sinica, 2021, 40(1): 4-12. doi: 10.1007/s13131-021-1719-0|