calving_dt_convergence

The landice/calving_dt_convergence test group supports tests for assessing the timestep convergence of calving physics in MALI. The tests all use pre-generated meshes.

The test group includes a single test case with many variants for using different meshes, calving laws, and velocity solver settings.

config options

There currently are no config options.

dt_convergence_test

landice/calving_dt_convergence/dt_convergence_test runs short integrations repeatedly with different values for the fraction of the calving CFL limit applied in the adpative timestepper. Time series of the total calving flux and the calving CFL to actual timestep ratio are then plotted, as well as summary of the number of calving warnings for each choice of calving CFL fraction (see below). The individual tests are named for the mesh, the calving law, and the velocity solver setting, separated by periods.

Example results of calving dt test for the humboldt.specified_calving_velocity.none test. The top plot shows the total calving flux over time for different choices of the calving CFL fraction. Results should be similar for small fraction values. As the fraction is increased, significant differences can be seen, indicating values greater than about 1.0 are less accurate. The second panel shows the cumulative total calving flux. The third panel shows the actual dt used by the adaptive timestepper divided by the timestep limit due to the calving CFL. This allows assessment of the current implementation limitation that the calving CFL limit is lagged by one time step; if this value is significantly less than the fraction specified, the method of lagging the calving CFL is not accurate. The final panel shows the number of calving warnings (left axis) and the fraction of total timesteps with warnings (right axis) for each value of the calving CFL fraction that was run. Warnings are triggered when the calving routine is unable to calve more than 1% of the calving volume it was asked to calve.