18We present a new 3D unstructured-grid model (SCHISM) which is an upgrade from an existing model (SELFE). 19The new advection scheme for the momentum equation includes an iterative smoother to reduce excess mass 20 produced by higher-order kriging method, and a new viscosity formulation is shown to work robustly for generic 21 unstructured grids and effectively filter out spurious modes without introducing excessive dissipation. A new higher-22 order implicit advection scheme for transport (TVD 2 ) is proposed to effectively handle a wide range of Courant 23 numbers as commonly found in typical cross-scale applications. The addition of quadrangular elements into the 24 model, together with a recently proposed, highly flexible vertical grid system (Zhang et al. 2015), leads to model 25 polymorphism that unifies 1D/2DH/2DV/3D cells in a single model grid. Results from several test cases 26 demonstrate the model's good performance in the eddying regime, which presents greater challenges for 27 unstructured-grid models and represents the last missing link for our cross-scale model. The model can thus be used 28 to simulate cross-scale processes in a seamless fashion (i.e. from deep ocean into shallow depths). 29 30 31 32 54 accuracy, efficiency and robustness. For instance, the eddying regime sets a high standard for numerical dissipation 55 and stability (control of modes), whereas the order of numerical schemes is less important in the estuarine 56 applications, as the strong forcing therein favors stable and often lower-order numerical schemes. For such 57 applications, more emphasis should be placed on faithfully resolving geometric and bathymetric features that act as 58 the 1 st -order forcing for the underlying processes. The rich diversity of the processes as found from shallow to large 59 depths likely precludes a 'one-size-fits-all' approach, and different numerical options may prove to be useful in 60 different regimes. This has been the guiding principle when we built our cross-scale model. 61As far as the model (SELFE) we have been developing for the past 15 years is concerned, we have made steady 62 progress in the baroclinic regime in the shallows (ZB08; Burla 2010). Although all implicit models have inherent 63 numerical diffusion, SELFE seems to have struck a good balance between numerical dissipation (due to implicit 64 time stepping), numerical dispersion (due to Finite Element Method), and stability demanded by such type of 65 applications. However, the following areas need to be improved before it can become a bona fide cross-scale model. 66First, the stratification is often under-estimated. This is related to the transport scheme as well as the vertical grid 67 Zhang et al. Page 3system used (which is a hybrid system with part terrain-following S coordinates and part Z coordinates). The 68 situation improves significantly with the introduction of TVD scheme for transport, and recently a flexible LSC 2 69 vertical grid (Zhang et al. 2015). Second, the model has not been applied in the eddying re...

[1] We present a new modeling system for wave-current interaction based on unstructured grids and thus suitable for very large-scale high-resolution multiscale studies. The coupling between the 3D current model (SELFE) and the 3rd generation spectral wave model (WWM-II) is done at the source code level and the two models share same sub-domains in the parallel MPI implementation in order to ensure parallel efficiency and avoid interpolation. We demonstrate the accuracy, efficiency, stability and robustness of the coupled SELFE-WWM-II model with a suite of progressively challenging benchmarks with analytical solution, laboratory data, and field data. The coupled model is shown to be able to capture important physics of the wave-current interaction under very different scales and environmental conditions with excellent convergence properties even in complicated test cases. The challenges in simulating the 3D wave-induced effects are highlighted as well, where more research is warranted.

No abstract

Characterizations of tsunami hazards along the Cascadia subduction zone hinge on uncertainties in megathrust rupture models used for simulating tsunami inundation. To explore these uncertainties, we constructed 15 megathrust earthquake scenarios using rupture models that supply the initial conditions for tsunami simulations at Bandon, Oregon. Tsunami inundation varies with the amount and distribution of fault slip assigned to rupture models, including models where slip is partitioned to a splay fault in the accretionary wedge and models that vary the updip limit of slip on a buried fault. Constraints on fault slip come from onshore and offshore paleoseismological evidence. We rank each rupture model using a logic tree that evaluates a model's consistency with geological and geophysical data. The scenarios provide inputs to a hydrodynamic model, SELFE, used to simulate tsunami generation, propagation, and inundation on unstructured grids with <5-15 m resolution in coastal areas. Tsunami simulations delineate the likelihood that Cascadia tsunamis will exceed mapped inundation lines. Maximum wave elevations at the shoreline varied from ~4 m to 25 m for earthquakes with 9-44 m slip and M w 8.7-9.2. Simulated tsunami inundation agrees with sparse deposits left by the A.D. 1700 and older tsunamis. Tsunami simulations for large (22-30 m slip) and medium (14-19 m slip) splay fault scenarios encompass 80%-95% of all inundation scenarios and provide reasonable guidelines for landuse planning and coastal development. The maximum tsunami inundation simulated for the greatest splay fault scenario (36-44 m slip) can help to guide development of local tsunami evacuation zones.

No abstract