Test data: 240km (v3)

240km test case for MPAS-A CI workflows.

v3 changes vs v2

Removed q2 (2-m water vapor mixing ratio diagnostic) from both
stream_list.atmosphere.output and stream_list.atmosphere.diagnostics.

The v2 240km init.nc / sfc_update.nc puts MPAS-A into a state where the
q2 diagnostic produces NaN at one or more cells in a 1-hour run with
default physics. NaN appears in q2 regardless of compiler (gcc, NVHPC),
precision (single), I/O backend (SMIOL, PIO), MPI rank count (1, 4), or
GPU offload (OpenACC on/off), so it is not a CI-side problem; it is a
real numerical defect in MPAS-A's q2 diagnostic on this particular
initial condition.

Because q2 is a diagnostic only (not used in the time integration),
removing it from the history stream is a safe workaround that lets BFB
tests run cleanly. The underlying defect has been reported upstream.

All other files (init.nc, sfc_update.nc, namelist.atmosphere, physics
tables, METIS partitions, streams.atmosphere, the other stream lists,
and the graph.info) are byte-identical to v2.

Inherited from v2

x1.10242.init.nc and x1.10242.sfc_update.nc were regenerated upstream
to address GPU ECT failures (config_start_time = 2014-09-10).

Test case: 240km v2

v2 changes (vs testdata-240km-v1)

Regenerated test case from upstream MPAS GPU correctness work. Built by overlaying these files onto the v1 tarball:

File	Change
x1.10242.init.nc	regenerated (size 83 MB → 89 MB)
x1.10242.sfc_update.nc	regenerated

Everything else (physics tables, METIS partitions, namelist.atmosphere, streams definitions, stream lists) is byte-identical to v1. The 240km namelist.atmosphere did not change between v1 and v2 (start time was already 2014-09-10).

Note: the dev's working directory included a regenerated 4-rank METIS partition (x1.10242.graph.info.part.4) that differed from v1, but the underlying mesh adjacency (x1.10242.graph.info) was byte-identical. The v1 partition is preserved in v2 to keep mesh artifacts stable across versions.

240km Quasi-Uniform Mesh Test Case

Used for BFB and general CI validation workflows.

Mesh

• Grid: x1.10242 (240km quasi-uniform)
• Decomposition prefix: x1.10242.graph.info.part.

Dynamics

Parameter	Value
config_dt	1200 s
config_run_duration	0_06:00:00 (6 hours)
config_start_time	2014-09-10_00:00:00
config_time_integration_order	2
config_number_of_sub_steps	2
config_dynamics_split_steps	3
config_epssm	0.1
config_smdiv	0.1
config_horiz_mixing	2d_smagorinsky
config_visc4_2dsmag	0.05
config_monotonic	true

Physics

Parameter	Value
config_physics_suite	mesoscale_reference
config_radtlw_interval	01:00:00
config_radtsw_interval	01:00:00
config_sst_update	true

Streams

Stream	Interval
history output	6:00:00
restart output	none
diagnostics	none

Source

NCAR/mpas-ci-data, regenerated for MPAS GPU correctness fix (April 2026).

Test case: 120km v2

v2 changes (vs testdata-120km-v1)

Regenerated test case from upstream MPAS GPU correctness work. Built by overlaying these files onto the v1 tarball:

File	Change
x1.40962.init.nc	regenerated (ERA5 source, init_case=8 real-data; v1 was CFSR, init_case=7)
x1.40962.sfc_update.nc	regenerated to match new start time
x1.40962.static.nc	regenerated (now includes 30 GOCART aerosol levels; size 53 MB → 71 MB)
namelist.atmosphere	config_start_time 2018-04-14_00:00:00 → 2014-09-10_00:00:00
namelist.init_atmosphere	provenance: documents how new init.nc was built (CI does not run init_atmosphere)

Everything else (physics tables, METIS partitions, streams definitions, stream lists) is byte-identical to v1.

120km Quasi-Uniform Mesh Test Case

Used for Ensemble Consistency Test (ECT) workflows.

Mesh

• Grid: x1.40962 (120km quasi-uniform)
• Decomposition prefix: x1.40962.graph.info.part.

Dynamics

Parameter	Value
config_dt	720 s
config_run_duration	0_02:36:00 (2h 36min, 13 steps)
config_start_time	2014-09-10_00:00:00
config_time_integration_order	2
config_number_of_sub_steps	2
config_dynamics_split_steps	3
config_epssm	0.1
config_smdiv	0.1
config_horiz_mixing	2d_smagorinsky
config_len_disp	120000.0
config_visc4_2dsmag	0.05
config_monotonic	true

Physics

Parameter	Value
config_physics_suite	mesoscale_reference
config_radtlw_interval	00:30:00
config_radtsw_interval	00:30:00
config_sst_update	false

Streams

Stream	Interval
history output	none (ECT workflows override)
restart output	1_00:00:00 (used by spin-up)
diagnostics	none
surface input	none

Notes

• History output interval is set to none by default; ECT workflows override via run-perturb-mpas
• Restart interval is used during the 24-hour spin-up in ect-ensemble-gen

Source

NCAR/mpas-ci-data, regenerated for MPAS GPU correctness fix (April 2026).

ECT data (MPAS v8.4.0)

ECT ensemble summary and spin-up restart for MPAS v8.4.0. Updated automatically by the ect-ensemble-gen workflow.

Test case: 240km v1

240km Quasi-Uniform Mesh Test Case

Used for BFB and general CI validation workflows.

Mesh

• Grid: x1.10242 (240km quasi-uniform)
• Decomposition prefix: x1.10242.graph.info.part.

Dynamics

Parameter	Value
config_dt	1200 s
config_run_duration	0_06:00:00 (6 hours)
config_start_time	2014-09-10_00:00:00
config_time_integration_order	2
config_number_of_sub_steps	2
config_dynamics_split_steps	3
config_epssm	0.1
config_smdiv	0.1
config_horiz_mixing	2d_smagorinsky
config_visc4_2dsmag	0.05
config_monotonic	true

Physics

Parameter	Value
config_physics_suite	mesoscale_reference
config_radtlw_interval	01:00:00
config_radtsw_interval	01:00:00
config_sst_update	true

Streams

Stream	Interval
history output	6:00:00
restart output	none
diagnostics	none

Source

NCAR/mpas-ci-data, namelists updated for CI defaults.

Test case: 120km v1

120km Quasi-Uniform Mesh Test Case

Used for Ensemble Consistency Test (ECT) workflows.

Mesh

• Grid: x1.40962 (120km quasi-uniform)
• Decomposition prefix: x1.40962.graph.info.part.

Dynamics

Parameter	Value
config_dt	720 s
config_run_duration	0_02:36:00 (2h 36min, 13 steps)
config_start_time	2018-04-14_00:00:00
config_time_integration_order	2
config_number_of_sub_steps	2
config_dynamics_split_steps	3
config_epssm	0.1
config_smdiv	0.1
config_horiz_mixing	2d_smagorinsky
config_len_disp	120000.0
config_visc4_2dsmag	0.05
config_monotonic	true

Physics

Parameter	Value
config_physics_suite	mesoscale_reference
config_radtlw_interval	00:30:00
config_radtsw_interval	00:30:00
config_sst_update	false

Streams

Stream	Interval
history output	none (ECT workflows override)
restart output	1_00:00:00 (used by spin-up)
diagnostics	none
surface input	none

Notes

• History output interval is set to none by default; ECT workflows override via run-perturb-mpas
• Restart interval is used during the 24-hour spin-up in ect-ensemble-gen

Source

NCAR/mpas-ci-data, namelists updated for ECT defaults.

ECT data (MPAS v8.3.1)

ECT Data for MPAS v8.3.1

Ensemble summary and spin-up restart for the Ensemble Consistency Test.
Updated automatically by the ect-ensemble-gen workflow.

Assets

File	Description
mpas_ect_summary_120km.nc	Ensemble summary (200 members, 120km mesh)
120km-spinup-restart.nc	24-hour spin-up restart for ECT member runs

ECT Parameters

Parameter	Value
Resolution	120km (x1.40962)
Ensemble size	200
Perturbation variable	theta
Perturbation magnitude	1e-14
Time slice	0
Run duration per member	0_02:36:00
PyCECT version	3.3.1

Regeneration

Run the ect-ensemble-gen workflow to regenerate both assets.
When MPAS-Dev releases a new version with science changes, create a new
ect-vX.Y.Z release and update RELEASE_ECT in ci-config.env.