Add time-step-exit and aes-graupel savepoint validation - #1363
Closed
jcanton wants to merge 5 commits into
Closed
Conversation
Readers for the three new savepoints generated by the instrumented ICON run (icon-nwp add_exp_ape_aes): - IconTimeStepExitSavepoint: end-of-timestep prognostic state, written after integrate_nh returns (all physics applied, time levels swapped) - IconAesGraupelInit/ExitSavepoint: state + prm_tend accumulators around the mig block (cloud_mig = satad + graupel + satad), plus aggregated surface precip (rsfl, ssfl, pr, ufcs) Replace test_standalone_driver_moist_physics: instead of only checking the fields muphys does not touch against diffusion/nonhydro exits, the full prognostic state (vn, w, rho, exner, theta_v, all six tracers) is now validated against the single time-step-exit savepoint. Provisional tolerances document the known phy2dyn fidelity gaps (linearized exner update, no theta_v update, no negative-tracer clipping, full-column vs jks_cloudy) until they are ported. Add test_aes_graupel_granule: first validation of the muphys granule against Fortran ICON (previous coverage was netCDF references from the C++ reference implementation only). Compares component tendencies and surface precip against the aes-graupel-init/exit pair, restricted to the jks_cloudy..nlev levels ICON computes, and checks the granule produces (near-)zero tendencies above. Tolerances are provisional until measured on the regenerated exclaim_ape_aesPhys v05 archive (overwritten in place, no version bump). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Reader classes, provider methods, test file and function names now use the muphys nomenclature; the serialbox savepoint names (aes-graupel-init/exit) and the Fortran sources keep the aes_graupel naming so it stays clear the data interface is the icon-nwp Fortran mig block, not the icon-mpim Kokkos port. Also drop the qnc-units bug commentary obsoleted by the MuphysConfig fix (qnc now 50.0e6 m^-3, matching Fortran cloud_num) and read qnc from the config again in the datatest. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The muphys port followed the older muphys C++ reference implementation, while icon-nwp carries the newer MPIM rain-microphysics revisions. This adds the icon-nwp formulation (mo_aes_graupel.f90) as a second scheme, selected via MuphysConfig.scheme / a use_icon_nwp flag threaded through graupel_run and muphys_run: - cloud_to_rain: rho-dependent polynomial accretion kernel in log(clamped rho*qr) instead of the constant 5.25 - rain_to_vapor: exp-polynomial evaporation instead of the power law - cloud_to_snow: riming tuning factor 3.0 - snow_number/snow_lambda: operate on the snow mass density rho*qs with a lower clamp instead of the additive (qs + 2e-6) offset, and branch on rho*qs > qmin - fall speeds (vm): rain becomes a log-polynomial (with the constant term not density-corrected, faithfully reproducing the Fortran operator precedence), ice prefactor 0.80, snow prefactor 115.6 with snow_number evaluated at the clamped falling-mass density and the previous level's temperature (now carried in the scan state), and a two-sided clamp of the hydrometeor density for all species - ice_dep is computed from the vapor_x_ice deposition before the nucleation contribution is added, matching the Fortran ordering Defaults: setup_graupel/setup_muphys/MuphysComponent keep CPP_REFERENCE, so the existing netCDF-based tests are unchanged (verified: graupel_only/full_muphys/component datatests pass at rtol 1e-14 after the refactor); MuphysConfig defaults to ICON_NWP, so the standalone driver and the aes-graupel serialbox datatest use the ported scheme matching the Fortran that generates the reference data. Also fix the fused muphys_run to return the temperature updated by the final saturation adjustment (as the Fortran does), consistent with the adjusted qv/qc; the separate-programs path already did this. New stencil tests validate every ported formula against independent numpy references (pass on gtfn_cpu); an execution smoke run confirms the ICON_NWP path compiles, stays finite, and differs from CPP_REFERENCE by physically plausible magnitudes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
yiluchen1066
added a commit
that referenced
this pull request
Jul 8, 2026
Scheme names now say what each formulation is: KOKKOS_MUPHYS is the muphys C++/Kokkos reference implementation, AES_GRAUPEL is the icon-nwp mo_aes_graupel.f90 formulation. All _icon_nwp suffixes (field operators, programs, constants class, stencil tests, the use_icon_nwp flag) become _aes_graupel accordingly. The Python-level scheme dispatch in setup_graupel/setup_muphys uses match/case with an explicit case per scheme. Inside the GT4Py operators the branches stay scalar-bool if/else with a labeled else: the DSL has no match statement, and an if/elif pair on a bool is rejected by its definedness analysis (it cannot prove the two conditions exhaustive, so names assigned in the branches count as possibly undefined). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
Mandatory Tests Before merging, run the When developing, you can test your changes on CSCS CI before merge with the You can pass options to override pipeline variables, for example:
Available options are:
For each option, See The Optional Tests To run benchmarks you can use:
For more detailed information please look at CI in the EXCLAIM universe. |
Contributor
Author
|
already merged into #1301 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Readers for the three new savepoints generated by the instrumented ICON run (icon-nwp add_exp_ape_aes):
Replace test_standalone_driver_moist_physics: instead of only checking the fields muphys does not touch against diffusion/nonhydro exits, the full prognostic state (vn, w, rho, exner, theta_v, all six tracers) is now validated against the single time-step-exit savepoint. Provisional tolerances document the known phy2dyn fidelity gaps (linearized exner update, no theta_v update, no negative-tracer clipping, full-column vs jks_cloudy) until they are ported.
Add test_aes_graupel_granule: first validation of the muphys granule against Fortran ICON (previous coverage was netCDF references from the C++ reference implementation only). Compares component tendencies and surface precip against the aes-graupel-init/exit pair, restricted to the jks_cloudy..nlev levels ICON computes, and checks the granule produces (near-)zero tendencies above.
Tolerances are provisional until measured on the regenerated exclaim_ape_aesPhys v05 archive (overwritten in place, no version bump).
sync with https://gitlab.dkrz.de/icon/icon-nwp/-/merge_requests/2200