Each directory here contains a test problem or example. Before trying to run these tests, make sure you can get the unit-tests and parallel-tests to PASS (as discussed here), which will confirm you've got all the dependencies working.
You can run the tests individually using the run_model.sh scripts in each directory. Alternatively you can run all the test problems in one go using the script in ../tests/validation_tests/. The latter takes about an hour on the authors home desktop (with an Intel(R) Xeon(R) CPU E5-1650 v4 @ 3.60GHz, for tests as of May 2022). It is regularly run when the code is updated. In each subdirectory it will compile and run the model, make a series of the plots, and report multiple PASS/FAIL criteria for the problem (the latter is printed by the R script that runs the test).
The problems are:
-
circular_island -- This compares the model with a 2D analytical solution for the linear shallow water equations, for periodic waves around a conical island.
-
dambreak -- This compares the model with analytical solutions to the dam-break problem with various initial depths.
-
generic_example -- This is useful for modelling single-grid global-scale tsunami propagation. It is also set up as a regression test; we simulate a PTHA scenario that is loosely similar to the Tohoku tsunami, compare a range of flow algorithms with observations, and check that the results are the same as when we first set up the problem (to within some tolerance that will likely catch changes in the model behaviour, but should allow for differences between compilers). We also check that adding a uniform rise-time to the earthquake source is equivalent to smoothing time (as is mathematically true for linear equations).
-
isolated_building -- This compares the model with a well-know dam-break type experiment.
-
landslide_tsunami -- This compares the model with a 1D analytical solution to runup over a plane beach, with an initial condition that is heuristically like a landslide.
-
merewether -- This compares the model with a flooding scenario for merewether, similar to the ANUGA test, which is based on field observations and laboratory experiments.
-
nesting_plane_wave -- This is a simple 1D plane wave propagation test problem which includes a nested grid.
-
overbank_flow -- This computes steady, overbank flow in a channel with floodplains, and compares the model with an analytical solution for the cross-channel distribution of velocity, including the effect of eddy-viscosity.
-
parabolic_canal -- This compares the model with a 2D analytical solution for flow in a parabolic canal, which also includes a Coriolis term and periodic wetting and drying.
-
paraboloid_bowl -- This compares the model with a 2D analytical solution for flow in a paraboloid bowl, which is periodic with strong wetting and drying.
-
periodic_convergence -- This checks the order of accuracy of the model via convergence tests, using test problem from the literature.
-
periodic_multidomain -- This is not a test, but illustrates a multidomain model with periodic east-west boundary conditions, and allows a rise-time in the earthquake source.
-
radial_dam_break -- This runs a radial dam break problem.
-
spherical_box -- This compares the results of 3 different flow algorithms in an idealised test problem using spherical coordinates. The problem involves a large area domain with quasi-realistic topography, and we check that all algorithms produce similar results.
-
uniform_channel -- This tests the model with steady-uniform-flow in a uniform channel
-
uniform_slope -- This tests the model with steady-uniform-flow in a uniform slope
-
uniform_slope_shallow -- This tests the model with steady-uniform-flow in a uniform slope, when the flow is very shallow.
See here for additional tests using the NTHMP benchmark problems, which are also run by the aforementioned validation test script.
All tests should PASS for the default setup, which largely focusses on the rk2 solver for the nonlinear shallow water equations and the leapfrog linear solver for the linear shallow water equations. There are variables controlling the default timestepping methods in global_mod.f90 which can be changed. The validation tests do exercise other solver types in any case - but less thoroughly.
If you are experimenting with a non-default flow algorithm, you should not be surprised if some tests FAIL, perhaps even when the results are ok. This can happen because some PASS/FAIL tests require ad-hoc thresholds to define the boundary between PASS and FAIL, and these criteria are debatable. In other cases the problem setup might require some adjustments specific to the non-default solver, or the non-default solver might just be poorly suited to some kinds of problems. In any case examination of the model outputs and plots will give insight into the model performance.