zoomy_amrex.malpasset

zoomy_amrex.malpasset#

Project the Malpasset triangular benchmark onto a structured AMReX grid.

The benchmark ships as an unstructured triangular mesh (data/malpasset/geo_malpasset-small.msh) carrying, as gmsh $NodeData, the survey bathymetry B and the reservoir-full initial condition (H depth, U/V velocity — both zero, water at rest). AMReX is a structured (rectangular) framework, so we:

  1. take the axis-aligned bounding box of the triangulation as the rectangular (“quadrilateral”) AMReX domain — the simplest way to embed the irregular catchment without embedded boundaries (EB);

  2. interpolate B and H from the triangle vertices onto the structured cell centres with a linear triangle interpolant (matplotlib.tri.LinearTriInterpolator), which masks points that fall outside the actual (non-convex) mesh — so the real, irregular boundary is respected, not just the convex hull;

  3. turn the masked exterior into a closed wall by raising the bed there above the reservoir surface and leaving it dry.

The result is two row-major float64 .raw rasters (bed → state comp 0, depth → state comp 1) laid out exactly as init_solution.cpp expects (idx = i + nx*j), plus the geometry the driver needs.

Wet/dry on the structured grid is handled by the driver’s h_min flux floor

  • negative-depth clamp (see Source/constants.H / make_rhs.H); here we only set the static bathymetry and the still-water IC.

Module Contents#

Functions#

prepare_rasters

Write bed.raw + depth.raw and return the AMReX geometry dict.

API#

zoomy_amrex.malpasset.prepare_rasters(msh_path, out_dir, ncell_x=180, ncell_y=96, wall_above=5.0, pad_frac=0.0, subsample=4, bed_agg='min')#

Write bed.raw + depth.raw and return the AMReX geometry dict.

Depth is max(0, surface - bed); the still-water surface (reservoir eta=100, sea eta=0) is preserved.

Args

  • ncell_x, ncell_y: Structured resolution.

  • wall_above: Exterior wall bed = max reservoir surface + this (metres).

  • pad_frac: Optional bounding-box padding as a fraction of each span.

  • subsample: Sub-grid factor per cell for the bathymetry aggregation.

  • bed_agg: "min" keeps the lowest sub-grid point (channel-preserving, keeps the narrow gorge open at coarse resolution); "mean" is the plain cell average.

Contents