A GNSS velocity field for estimating tectonic plate motion and testing global glacial isostatic adjustment models

Abstract

The two main causes of the long-term deformation of the Earth on a global scale are tectonic plate motion and Glacial Isostatic Adjustment (GIA). GIA results in vertical as well as lateral movements of the Earth’s surface. It is difficult to distinguish from local and regional effects, such as the deformational response to decadal and longer-term changes in continental water storage and the mass balance of glaciers and ice sheets. On a global scale, GIA is also, to some extent, difficult to distinguish from millennial-term lateral motion due to plate tectonics. The effects of GIA must therefore be modelled. GIA models use an ice sheet history combined with an estimate for Earth rheology to produce predictions of present-day GIA velocities. GIA models are typically tuned to fit evidence for past and present vertical motion, as determined from historical relative sea-level data, and they may additionally be tuned to fit GNSS-derived present-day uplift rates. However, GNSS-derived horizontal rates have not traditionally been used to tune GIA models. Lateral Earth structure can significantly influence horizontal GIA rates, and most GIA models do not account for lateral structure, these are so-called 1D GIA models. Recently, GIA models accounting for lateral Earth structure have been developed, known as 3D GIA models. Vertical GIA velocities are important for studies of surface mass loading, sea-level change, mass balance of glaciers and ice sheets, and vertical reference systems. Horizontal GIA velocities are also important for interpreting surface mass loading, as well as tectonic plate rigidity, with implications for horizontal components of reference systems. Consequently, this project aims to create a bespoke 3D GNSS surface velocity field to test and compare a set of recent 1D and 3D GIA models and investigate tectonic plate motion. In turn, this velocity field has several applications beyond this project. It may be used to investigate present-day surface loading due to ice melting as well as other aspects of the global hydrological cycle, and loading studies in general. The main motivation for creating a bespoke 3D velocity field (as opposed to using, e.g. the most recent International Terrestrial Reference Frame ITRF2014) is to include a larger number of GNSS sites in the GIA-affected areas of investigation, namely North America, Europe, and Antarctica. GIA and plate motion velocities are at the mm level so the choice of a stable and accurate reference frame plays a crucial role. Here I create the GNSS surface velocity field using the IGS repro2 data and other similarly processed GNSS datasets. The networks are deconstrained, combined and aligned to ITRF2014 on a daily level. For this, I use the Newcastle University-developed reference frame combination software Tanya. Within this project, the software has been updated to be compatible with ITRF2014, including the discontinuity information and post-seismic deformation models. This resulted in 57% reduction of the WRMS of the alignment post-fit residuals compared to the alignment to ITRF2008. The time series of daily GNSS solutions were used to create the GNSS velocity field. After additional data screening and quality control, the final GNSS velocity field has horizontal uncertainties mostly within 0.5 mm/yr, and vertical uncertainties mostly within 1 mm/yr, which make it suitable for testing GIA models. I use a suite of GIA models that have been produced by combining three different ice models (ICE-5G, ICE-6G and W12) with a range of 1D and 3D Earth models. By subtracting this ensemble from the velocity field, I identify and compare a range of plate motion models (PMMs), which are then expected to be unaffected by GIA. The impact of GIA on the PMM estimates is investigated and the resulting PMMs are compared with previously published ones. The results show that there can be significant GIA-related horizontal motion which may be modelled into the plate motion if left uncorrected. Using an extensive set of 1D and 3D GIA models allows to include more GNSS sites in the PMM estimate. These sites are in GIA-affected areas which have typically been excluded from PMM estimates. A joint estimation of PMM with GIA is beneficial when investigating GIA with GNSS observations because it reduces dependency on a pre-existing PMM which can be contaminated by GIA. Next, the predicted horizontal and vertical velocities of each GIA model are subtracted from the GNSS surface velocity field after removing the respective PMM. Median Absolute Deviations (MADs) are computed for the suite of residual fields including the null-GIA case, where GIA predictions were not taken into account. For the 3D GIA models, applying GIA corrections reduces the MAD in all regions. For 1D GIA models, applying GIA corrections reduces the MAD in the majority of regions. Exceptions are found for the vertical component of the velocity field in Antarctica, and the horizontal component in the global case. The latter result indicates that it is not possible to replicate the global horizontal GIA velocity field by combining a 1D Earth model with the global ice models being tested here. Based on the results of this project, it is not possible to conclude that 3D GIA models consistently outperform 1D GIA models or vice-versa. However, it is possible to identify common GIA model features that correspond to better MADs. Furthermore, a group of best-performing GIA models is selected for each region of interest based on their MADs, and the range of GIA predictions from this group is assumed to represent the uncertainty of GIA models. For Antarctica, a range of equivalent water height values is computed from the group of best-performing GIA models, which in turn can be used as an uncertainty measure when applying GIA corrections in GRACE studies of ice mass change. The total GIA contribution to annual mass change in Antarctica ranges from 5 Gt/yr to 45 Gt/yr depending on which of the best-performing GIA models is used.