GeoClaw in One Space Dimension

As of Version 5.10.0, the geoclaw repository contains some code for solving problems in one space dimension. This can be used for solving plane wave problems on planar topography (including onshore inundation), as well as radially symmetric problems on the plane or axisymmetric problems on the sphere (see Coordinate systems).

Some general notes:

  • The standard 2d version of GeoClaw can be used to solve 1d problem by simply specifying a domain that is only a few cells wide in the y-direction, and insuring that the topography, initial data, and any dtopo files varies only in x as well. By setting the AMR refinement ratios to be 1 in the y-direction, it is possible to still use adaptive mesh refinement in x. For some 1d problems this may be the best approach.

  • By contrast, the newly introduced 1d code does not support AMR at this time. Instead, a fixed grid is used.

    However, the grid spacing may be variable and some tools are provided to compute a mapped grid that has the property that the Courant number (based on the linearized shallow water wave speed sqrt(g*h))is roughly constant, so that cells in deep water are larger than cells in shallow water (transitioning to a uniform grid in very shallow water and onshore). For some problems a fine 1d grid of this nature can be used to compute a very accurate solution and is sometimes preferable to using AMR.

  • In addition to shallow water equations, the 1d code also supports two different forms of Boussinesq equations, which include dispersive terms and better model waves whose wavelength is short compared to the fluid depth. For more information, see Boussinesq solvers in One Space Dimension. (Two-dimensional Boussinesq solvers have also recently been implemented, with AMR, and will appear in a future release; see Boussinesq solvers in Two Space Dimensions.)

  • Topography data (topo files) and moving topography (dtopo files) can be specified much as in 2d GeoClaw; see topo1d below.

The 1d library routines are found in $CLAW/geoclaw/src/1d_classic/shallow, with some additional routines needed for the Boussinesq solvers in $CLAW/geoclaw/src/1d_classic/bouss.

Some examples illustrating usage can be found in $CLAW/geoclaw/examples/1d, and some plots and animations can be viewed in the GeoClaw Gallery

Coordinate systems

In setrun.py, the parameter rundata.geo_data.coordinate_system can be used to specify the coordinate system to be used.

  • rundata.geo_data.coordinate_system == 1: x is measured in meters. This is the most natural coordinate system for many 1d problems, e.g. modeling waves in a wave tank, or plane waves in the ocean (provided the topography only varies in the direction of propagation).

  • rundata.geo_data.coordinate_system == -1: x >= 0 is measured in meters and represents radial distance. In this case, the solution is assumed to a 1d cross section of a rotationally symmetric 2d solution. The 2d shallow water (or Boussinesq) equations can then be reduced to 1d equations that have a similar form to the plane wave equations, with the addition also of a geometric source term [BergerLeVeque2023]. This source term is built in to the solution procedure in this case.

  • rundata.geo_data.coordinate_system == 2: x is measured in degrees for a problem that is rotationally symmetric on the sphere about some axis of rotation, e.g. waves spreading out from a radially symmetric crater on topography that is also rotationally symmetric about the same axis. In this case -90 <= x <=90 with the endpoints corresponding to the two points where the axis intersects the sphere, so it represents latitude with respect to this axis. (If the axis passes through the poles then x is the ordinary latitude with x = -90 at the south pole and x = +90 at the north pole.)

    As in the case of radial symmetry, the spherical case requires some changes in the equations and the addition of a geometric source term. Near each pole the solution behaves much as in the radial symmetric case, but note that waves from a disturbance at one pole will initially decay as they spread out but after passing the equator they will start to refocus at the other pole.

Uniform and mapped grids

In setrun.py, the parameter rundata.grid_data.grid_type can be used to specify the computational grid to be used.

  • rundata.grid_data.grid_type == 0: A uniform grid is used, with spacing determined by the domain length and the number of grid cells specified.

  • rundata.grid_data.grid_type == 1: A mapped grid is used. In this case a function mapc2p.f90 must be provided to map the computational grid specified in setrun.py to physical cells. See The mapc2p function. In this case 0 <= x <= 1 is used in the examples, but any computational grid interval can be used as long as mapc2p maps equally spaced points on this interval to te desired physical grid.

  • rundata.grid_data.grid_type == 2: A nonuniform grid is used with a user-specified set of grid cell edges. In this case rundata.grid_data.fname_celledges should be set to a string giving the name of the file that contains the cell edges (one per line). Also, the computational grid should be in the domain 0 <= x <= 1, i.e.:

    clawdata.lower[0] = 0.           # xlower
    clawdata.upper[0] = 1.           # xupper
    clawdata.num_cells[0] = mx       # number of grid cells
    

    In this case the number of celledges in the data file should be mx+1. A mapc2p function that maps the unit interval to the physical grid must then be specified in setplot.py, and can be generated using the function clawpack.geoclaw.nonuniform_grid_tools.make_mapc2p().

    The module clawpack.geoclaw.nonuniform_grid_tools.py also includes tools to create a nonuniform grid with the property that a specified uniform grid width is used onshore an in very shallow water, but are much larger in deeper water, with the physical grid widths chosen so that the CFL number is roughly uniform (based on the propagation speed sqrt(g*h)). Most of the examples in $CLAW/geoclaw/examples/1d_classic/ illustrate this.

Note that when using grid_type 1 or 2, any gauges specified in setrun.py must be specified in computational coordinates, not physical coordinates. See, e.g. $CLAW/geoclaw/examples/1d_classic/ocean_shelf_beach/setrun.py for an example.

Topograpy data

Topography data is specified in a file that has two columns, with values x, B specifying the topo value B at spatial locations x. The topography is viewed as being piecewise linear connecting these points. As in 2d GeoClaw, the finite volume cells used for the computation have a single cell-averaged B value that is obtained by cell-averaging this piecewise linear function.

Note that if a mapped grid is used and if the topography values are specified at the cell edges, then the cell-averaged finite volume values are simply the average of the B values from each edge of the cell. In this case, the same file that is used to specify the topography can also be used to specify the grid. (The second column is ignored when it is read in as a grid specification.)

In setrun.py, the parameter rundata.topo_data.topofiles is set to a list of topofiles, each of which is specified by a list containing the topo_type and topofile_path, the path to the file, as in 2d. Currently only one topofile is supported, and so this should have the form:

rundata.topo_data.topofiles = [[topo_type, topofile_path]]

Currently only topo_type == 1 is supported, which has the form described above.

Moving topograpy (dtopo) data

In setrun.py, the parameter rundata.dtopo_data.dtopofiles is set to a list of dtopofiles, each of which is specified by a list containing the dtopo_type and dtopofile_path, the path to the file, as in 2d. Currently only one dtopofile is supported, and so this should have the form:

rundata.dtopo_data.dtopofiles = [[dtopo_type, dtopofile_path]]

Currently only dtopo_type == 1 is supported, and the dtopofile should have a form similar to what was described for topofiles above, except that each line starts with a t value for the time, so each line contains t,x,dz

The x,dz values give the displacement dz at x at time t. It is assumed that the grid is uniform and that the file contains mx*mt lines if mt different times are specified on a grid with mx points.

One way to specify a dtopo file is to use the Okada model (see Earthquake sources: Fault slip and the Okada model) in a situation where the fault is dipping in the x-direction and the fault geometry and slip are assumed to be constant in the y-direction over a long enough distance that a 1d slice in x is a reasonable model. Tools are provided create such a dtopo file, see the example in $CLAW/geoclaw/examples/1d/okada_dtopo.