-
Notifications
You must be signed in to change notification settings - Fork 696
WRF V4.2 Known Problems
Problem: For at least the last several major release versions there has been a problem with restart runs if they are not restarted at a boundary time. The simulation completes without errors, but if one were to compare output files for the original simulation and the restart run at identical times, there would be minor differences in the file.
There is no solution at this time. It is recommended to try to restart at boundary times.
Problem: When using lightning_option = 1, 2, or 11, and running a restart, simulation results will not be bit-for-bit identical when comparing output for the same time for the original simulation and the restart simulation unless cudt is set to 0.
Solution: For now, it is recommended to set cudt = 0 for all domains when using the lightning option and restarts.
Problem: If using lightning_option = 3, the run will seg-fault soon after the simulation begins. This behavior has been seen with both GNU and Intel compilers (but is likely a problem with others too).
There is currently no solution. Users will not be able to use lightning_option = 3 at this time. This has been a problem for several versions now.
Restarts not bit-for-bit identical when using any of the following options: adaptive time-stepping, Eta/Ferrier microphysics, SBM (full) microphysics, GFDL radiation, and SSIB LSM
Problem: When using any of the options use_adaptive_time_step=.true. (for multiple domains), mp_physics=5, mp_physics=32, ra_lw_physics=99 & ra_sw_physics=99, or sf_surface_physics=8 and running a restart, simulation results will not be bit-for-bit identical when comparing output for the same time for the original simulation and the restart simulation.
There is currently no solution to this problem. The restart simulation will complete and results should likely be reasonable, regardless of the minor differences.