MPAS Atmosphere release testing

Items to test for MPAS-Atmosphere v5.0

Scientific validation runs

• 2-d mountain wave idealized case
• baroclinic wave idealized case
• 3-d supercell idealized case
• real-data case using MODIS land-use
• What sorts of tests should be used to validate 2nd vs. 3rd order, splitting options, etc.?

Basic tests for memory errors

• Successful run with DEBUG=true with "mesoscale_reference" suite
• Successful run with DEBUG=true with "convection_scale_aware" suite * Fails with a FPE at module_mp_thompson.F:5036
• Successful run with DEBUG=true and only microphysics * Works for WSM6 and Kessler, but fails for Thompson
• Successful run with DEBUG=true with "none" suite

Restartability tests

• Bit-identical restart with "mesoscale_reference" suite
• Bit-identical restart with "convection_scale_aware" suite
• Bit-identical restart with only microphysics
• Bit-identical restart with "none" suite

Initialization

• Initialize with specific humidity
• Attempt to initialize with specific humidity when only RH is available
• Add native GWDO fields to existing static file
• Process both static and GWDO fields in the same execution of init_atmosphere_model
• Check for proper message when config_nfglevels is set too low
• Check for warning message if config_tc_vertical_grid = true and config_nvertlevels is not 41 or 55
• Verify no sea ice over land
• Can create monthly surface update files * Crucial that reference_time is set in the stream definition
• Can initialize using both USGS and MODIS land-use datasets
• Can initialize with native GFS
• Can initialize with high-res ECMWF model-level data

Framework tests

• Build and run with PIO2

Parallelization tests

• Bit-identical results for different MPI task counts w/o physics
• Bit-identical results for different OpenMP thread counts w/o physics * Code must be built with -fp-model precise

Diagnostics

• Verify that writing no diagnostics means no measurable time spent in diagnostics framework * Overhead is about 7 ms/timestep. For typical runs that require ~1 s/timestep, this is probably acceptable overhead for now.
• Can write soundings?
• Check that convective diagnostics look reasonable

Miscellaneous functionality

• Basic sanity test for IAU capability
• Basic test of 1-d OML
• Confirm reasonable speedup when compiling with -DCONST_INNER_DIMS -DCONST_NVERTLEVELS=?? -DCONST_MAXEDGES=?? -DCONST_MAXEDGES2=?? -DCONST_NUM_SCALARS=??