Solvi Thrastarson , Martin van Driel, Michael Afanasiev, Andreas Fichtner

In recent years, seismic station coverage in polar regions has been increasing steadily, providing new insight into the deep structure and dynamics of these remote parts of the globe. Numerical seismic wave propagation through polar regions is complicated by the presence of ice sheets. At periods relevant for regional-scale waveform tomography, the ice acts as a thin layer that, for an exact solution, needs to be meshed by elements that are small compared to the minimum wavelength. As a consequence, computational costs may become prohibitively high.

In this study, we investigate the effects of various simplifications of the mesh using spectral-element simulations of regional seismic wave propagation, with propagation distances of several hundred kilometres. Our goals are to quantify errors in phase and amplitude, and to identify simplifications that significantly reduce computational costs while producing numerical errors that are tolerable given a concrete application, dataset and frequency band. Using the example of western Greenland, we specifically consider meshes where 1) the ice sheet is replaced by bedrock, 2) where the ice sheet is removed but bedrock topography is honored, and 3) where any topography and the ice sheet are ignored.

For periods of 10 s we find that the group velocity of fundamental-mode surface waves can be off by more than 2 % when the ice sheet topography is ignored (2 and 3). The actual elastic properties of the ice play a comparatively smaller role. Furthermore, the ice sheet-related traveltime errors diminish with increasing period, whereas the topography-related errors are more frequency-independent. This suggests (1) that regional-scale topography with height much less than a wavelength should be modelled also in ice-free regions, and (2) that computational-costs may be reduced substantially by replacing ice with bedrock in the simulations, while fully honoring topography.

Amplitude errors in the fundamental-mode surface waves are equivalent to variations in attenuation that easily exceed 100 %, regardless of the type of simplification. It follows that the ice sheet should be meshed accurately when amplitude information is used to infer regional-scale Earth structure.