Note the maximum mesh element size was very coarse, but this only occurred in very deep, flat areas away from the slide region: all shallow regions or regions where depth varies rapidly had much finer resolution due to the choice of metric. Note also that the horizontal resolution around the coastlines was much less

than that of the bathymetry data (1 km or 500 m mesh resolution vs. 1.8 km bathymetry resolution) and hence all bathymetric features were well resolved in these regions. The palaeobathymetric domain was generated by first adding the isostatic this website adjustment data from Bradley et al. (2011) to the GEBCO bathymetry dataset to generate a palaeobathymetry. Note that the isostatic data only has extent of −20° to 20° west to east and 40° to 70° south to north, and hence we extrapolated the data by setting the extended domain corners to the same values as the corners of the true domain and then using GMT to interpolate the missing data. We extrapolated

data to match our domain (−43° to 24° west to east and 22° to 80° south to north). Results from this simulation are therefore only valid within 20° west to 20° east and 43° north to 70° north. Note that all wave gauges are situated within this region except gauge 1 (Greenland). All comparisons to the multiscale mesh were carried out within this sub-domain. Once the palaeobathymetry was generated, HDAC inhibitors in clinical trials the 0 m contour was used to generate a coastline as GSHHS was no longer valid. Inland seas and lakes were removed. Mesh resolution, including refinement in the vicinity of the slide and around bathymetric features, was identical to the modern multiscale simulation, except all coastlines were generated using 1 km element lengths.

As before any small islands and features were removed if they could not be resolved. The resulting coastline and bathymetry are shown in Fig. 1 which also shows the comparison to the high resolution GSHHS data. There are clear differences in coastline configuration around the eastern coast of the UK, but no significant differences around the central and southern Norwegian coasts. The mesh contains just over 1 million elements, around 300,000 fewer than the modern mesh, which is largely due to the difference in coastline resolution and the reduced ocean area (Table 3). For each simulation we compare the basin-wide Urocanase free-surface (i.e. sea surface) height and the free-surface variation at the 34 virtual wave gauges. We compare against a subset of these locations for each simulation. Fig. 6 shows the large-scale free-surface patterns and the qualitative convergence between 25 and 12.5 km mesh resolution. There are no discernible differences in free surface at 60 min simulated time for resolution of 25 km and below. Minor differences between the 25 km and 12.5 km simulation output at 120 min can be seen, but there is no visible difference between 12.5 km and 6.