To avoid Apoptosis Compound Library order sharp edges, which would cause numerical oscillations, a smoothing length, S  , is used at the front and back of the slide and the slide is smoothed along the whole width laterally as described in Harbitz (1992). S   is 1 km in the 2-D validation

study and 7.5 km in the Storegga simulations. The slide movement is then governed by x′x′ and y′y′, which describe the slide motion in the x–y plane and are defined by: equation(10) x′=(x-xs)cosϕ+(y-ys)sinϕ,x′=(x-xs)cosϕ+(y-ys)sinϕ,and equation(11) y′=(x-xs)sinϕ+(y-ys)cosϕ.y′=(x-xs)sinϕ+(y-ys)cosϕ.This gives a total volume of the slide, V: equation(12) V=0.9Bhmax(L+0.9S).V=0.9BhmaxL+0.9S.The motion given by (2) is then weakly imposed in the normal direction on the lower boundary to simulate the rigid block slide. This is a similar method to Ma et al. (2012) and Harbitz (1992), though differs in that Harbitz (1992) alter the h term in the shallow water equations. In practice, all methods should give very similar results. To ensure correct operation of the slide-tsunami model for weakly dispersive or non-dispersive waves

we replicated simulations from independent Natural Product Library screening numerical modelling studies in the literature. The first is a flat two-dimensional model, with dimensions approximately equivalent to the Storegga slide (Haugen et al., 2005), which produces a non-dispersive wave. The second is a smaller-scale, three-dimensional slide on a gentle slope (Ma et al., 2013), which produces a weakly dispersive wave. Comparisons to these previous studies verify correct implementation

of the slide boundary condition. Fluidity’s ability to capture highly dispersive slide-tsunami will be examined in future work. Haugen et al. (2005) simulated wave generation by the Storegga slide using a two-dimensional (x–z) approach, with an idealised rigid-block slide geometry and constant water depth. They showed that the very large length of the Storegga Dimethyl sulfoxide slide compared to the water depth resulted in a very long wave with little-to-no dispersive characteristics. Here we reproduce this simulation using Fluidity with a single element in the vertical. Tests with more vertical layers (not shown) produced almost identical results, confirming that wave dispersion is negligible in this scenario. The test case uses a flat-bottom domain, 1000 m deep, and 2000 km long. The slide has the parameters detailed in Table 1. Fluidity simulated the same scenario at six different horizontal resolutions: 5000 m, 2000 m, 1000 m, 500 m, 250 m, and 125 m. The mesh in this case is formed of 1D elements in the horizontal, which are then extruded downwards to 1000 m. A single layer of triangular elements was used in the vertical and the timestep was fixed at 1 s. The Fluidity results are compared to Haugen et al. (2005) in Fig. 2.