The numerical oscillations visible in the Fluidity output become negligible at 1000 m resolution, with little difference between results at resolutions between 1000 m and 125 m Veliparib ic50 ( Fig. 2). The observed numerical oscillations are caused by the sharpness of the leading and trailing edges of the slide, where minimal smoothing of 1000 m was used ( Haugen et al., 2005). Increasing

the smoothness of these edges (by increasing S in (9)) removes the oscillations. Clearly, the mesh resolution must be high enough to capture the smoothing length or the slide will have an effective flat front. To check that this was the cause of the spurious oscillations, the 5000 m resolution case was re-run with a smoothing length of 7500 m. The results show much reduced oscillations, but with the

wave form shifted due to the new location of maximum height ( Fig. 2). This experiment confirms the correct implementation of the boundary condition and shows how the assumed shape of the slide dictates the mesh resolution required in the slide area. A slide with steeper leading and trailing edges requires higher spatial resolution to eliminate numerical oscillations. To extend our validation 17-AAG in vivo of Fluidity’s new slide-tsunami model to three dimensions, we also replicated a simulation of landslide generated waves that are only weakly dispersive (Ma et al., 2013). Recent work by Ma et al. (2013) simulated the wave train produced by a rigid-block model in a three-dimensional domain on a constant slope. We can therefore compare Fluidity to the results shown in Ma et al. (2013). The domain is 8 ×× 8 km, with a constant slope of 4°. We set the minimum depth to be 12 m and the maximum to be 400 m. We used a horizontal model resolution

was 25 m in x   and y   and explored the influence of vertical resolution by performing simulations with 1–4 layers. Ma et al. (2013) use a different slide geometry to that described above, based on the work of Enet and Grilli (2007). The slide geometry is given by: equation(13) hs=hmax1-∊1coshkbx1coshkwy-∊where kb=2C/b,kw=2C/wkb=2C/b,kw=2C/w and C=acosh(1/∊)C=acosh(1/∊). The slide has length b=686b=686 m, width w=343w=343 m and thickness hmax=24hmax=24 ADP ribosylation factor m. The truncation parameter, ∊∊ is 0.717. The slides moves according to: equation(14) s(t)=s0lncoshtt0where s0=ut2/a0,t0=uta0,a0=0.27 m s−2, and ut=21.09ut=21.09 m s−1 as detailed in Ma et al. (2013). We use these definitions of the slide height and speed for comparisons to Ma et al. (2013). The resulting wave is very similar in magnitude and waveform to that shown in Ma et al. (2013), even using only a single layer in the vertical (Fig. 3). Convergence of the Fluidity model results is observed for three or more element layers (c.f. 40 layers used by Ma et al. (2013)), indicating that the wave is only weakly dispersive. In more detail, Fluidity produces slightly lower amplitude waves than those reported by Ma et al. (2013) (Fig.