Skip to content

Releases: GEOS-ESM/GEOSradiation_GridComp

Updates for ifx

27 Nov 18:59
cf12672
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp has update to the Solar Gridded Component to allow it to work with ifx.

All testing shows it to be zero-diff.


What's Changed

  • v11: Convert Solar GC to use gFTL StringVector by @mathomp4 in #50
  • GitFlow: Merge develop into main with ifx fix by @mathomp4 in #51

Full Changelog: v1.9.0...v1.10.0

Solar Bugfix, CMake update, NRL Preprocessing

21 Aug 19:01
733c6c7
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp has the following updates:

  1. A bugfix for some code in Solar (#45)
  2. Update the ESMF CMake target to ESMF::ESMF
  3. Add the NRL SSI2 preprocessing scripts

All tests show these are zero-diff to v1.8.0


What's Changed

Full Changelog: v1.8.0...v1.9.0

Update to RRTMGP v1.7

01 Mar 17:13
00bb036
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp moves it to use RRTMGP v1.7. Note this release will not compile unless also used with RRTMGP v1.7 as there are file changes in CMake. Moreover, to run RRTMGP you also need a corresponding change in AGCM.rc, see GEOS-ESM/GEOSgcm_App#585

This release is zero-diff to GEOSradiation_GridComp v1.7 when not running RRTMGP, but is non-zero-diff for RRTMGP.

Note also that RRTMGP v1.7 can be run in Single Precision mode with the CMake option:

-DRRTMGP_SINGLE_PRECISION=ON

Tests show that single-precision RRTMGP is about 10-20% faster than DP. But, for now we are keeping the default build to be double precision by default until further testing. Per @dr0cloud:

I have noticed around a -0.15W/m2 difference in global OSR in SP mode, but individual clouds can be -2 W/m2 or more.

Updates for RRTMGP Validation and OceanBioGeoChem

22 Feb 16:27
46ee430
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp has two updates.

The first is a small change needed for ongoing OceanBioGeoChem work (#28). To use this completely needs updates in GEOSgcm_GridComp and GOCART. Contact @mfmehari for more information.

The second is an update for on-going validation of RRTMGP (#33). This adds a new CMake option, ENABLE_SOLAR_RADVAL which is default OFF. Enabling it will enable many new exports for validation purposes.

This update is non-zero-diff for RRTMGP but that is not run by default currently. For RRTMG they are either zero-diff or 'truncation zero-diff', but these changes should only affect diagnostic outputs. That said, the Solar Internal Restart will have some changes (see below) but the overall state of the model is zero-diff with RRTMG:

  • Metadata changes (long_name)
    • CLDHISW: "high-level_cloud_area_fraction_rrtmg_sw_REFRESH" → "high-level_cloud_area_fraction_RRTMG_P_SW_REFRESH"
    • CLDLOSW: "low-level_cloud_area_fraction_rrtmg_sw_REFRESH" → "low-level_cloud_area_fraction_RRTMG_P_SW_REFRESH"
    • CLDMDSW: "mid-level_cloud_area_fraction_rrtmg_sw_REFRESH" → "mid-level_cloud_area_fraction_RRTMG_P_SW_REFRESH"
    • CLDTTSW: "total_cloud_area_fraction_rrtmg_sw_REFRESH" → "total_cloud_area_fraction_RRTMG_P_SW_REFRESH"
  • Removed fields
    • TAUHIPAR
    • TAULOPAR
    • TAUMDPAR
    • TAUTTPAR
  • Added fields
    • COTDENHIPAR
    • COTDENLOPAR
    • COTDENMDPAR
    • COTDENTTPAR
    • COTHIPAR
    • COTLOPAR
    • COTMDPAR
    • COTTTPAR
    • COTNUMHIPAR
    • COTNUMLOPAR
    • COTNUMMDPAR
    • COTNUMTTPAR

What's Changed

New Contributors

Full Changelog: v1.6.0...v1.7.0

Load Balance Updates and Profiler Fixes

15 Dec 14:48
9cc656f
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp has a few updates:

  1. The Load Balancing code in Solar was reworked to not call MAPL_LoadBalance code when not load balancing. Previously, it used a naive approach which still did MPI calls for no reason. Now it avoids that.
  2. Adds the ability to change the MaxPasses parameter when loadbalancing via the SOLAR_LB_MAX_PASSES: AGCM.rc resource. By default it uses 100 which is the default in MAPL
  3. Fixes for the profilers. Testing by @wmputman and @atrayano at high-res found that some columns over an integration period might not have, say, any clouds. This would then avoid some code in RRTMG which meant some timers were not turned on-off on some processes. This lead to deadlocks in Finalize when the timers were outputted. Likewise, when the loadbalaning code was updated, it found that processes without lit points also avoided some profiler calls.

All testing shows this is zero-diff to GEOSradiaiton_GridComp v1.5.1


What's Changed

  • GitFlow: Merge hotfix from main to develop by @mathomp4 in #25
  • Bypass MAPL_LoadBalance when not running load balancing, add maxpasses as resource by @mathomp4 in #27
  • GitFlow: Merge develop into main for release by @mathomp4 in #29

Full Changelog: v1.5.1...v1.6.0

Precision-fix for selectable T_MIN_OK

22 Aug 17:47
72d8b05
Compare
Choose a tag to compare

This release fixes a bug in the user-selectable T_MIN_OK code added in v1.5.0


What's Changed

Full Changelog: v1.5.0...v1.5.1

User-selectable T_MIN_OK for RRTMGP

03 Aug 12:14
0d8e5e7
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp allows the user selectable leniency of minimum temperature for LUT violation with RRTMGP. It is controlled by the following settings in AGCM.rc:

RRTMGP_LW_TMIN_INC_OK_K: 
RRTMGP_SW_TMIN_INC_OK_K: 

The default for both is 10.0 Kelvin (which was the hard coded value before).

This change is non-zero-diff, but only if you are running RRTMGP, so it should not change results as RRTMG is still the default radiation scheme.


What's Changed

Full Changelog: v1.4.0...v1.5.0

Updates for RRTMGP v1.6

24 Apr 15:12
588f7d7
Compare
Choose a tag to compare

Various changes to merge in v1.6 RRTMGP and upgrade the structure of the radiation codes:

  • Update @GEOSradiation_GC to run with RRTMGP v1.6.
  • RRTMGP now has a generalized maximum-random overlap option (the default) with cloud condensate inhomogeneity making it consistent with RRTMG (i.e., Oreopoulous et al., 2012)
  • Make type of sub-gridscale condensate inhomogeneity selectable at runtime.
  • Update condensate inhomogeneity system to be more lightweight.
  • Add internal RRTMG timing for cost analysis
  • various and significant structural (readability, speed, current practise, etc.), commenting, and cosmetic improvements.
  • In particular, a number of REFRESH-timed EXPORTS are now handled implicitly via the FRIENDLYTO mechanism for INTERNALs. The INTERNAL state of SOLAR now has COSZSW, CLDxxSW, and TAUxxPAR added to it.
  • effective radii used in RRTMGP are clipped to limits for RRTMGP lookup tables.

Changes are zero-diff for RRTMG runs except for nine extra 2D fields added to solar internal state.

Note: To run with RRTMGP v1.6 requires:

v1.3.0

24 Mar 13:34
dec6974
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp has updates from @wmputman from his refactored physics. This is non-zero-diff to v1.2.0


What's Changed

  • pmn: added vrtqdr code explanations, 0-diff by @dr0cloud in #14
  • Merging refactored physics code into develop by @sdrabenh in #15
  • GitFlow: Merge Develop into Main for release by @mathomp4 in #16

New Contributors

Full Changelog: v1.2.0...v1.3.0

v1.2.0

02 Dec 17:36
be45f26
Compare
Choose a tag to compare

This release of GEOSradiation_GridComp extends the work from v1.1.0 that enabled production of OBIO surface solar downwelling direct and diffuse fluxes on OBIO bands. The previous release worked for RRTMG only. This release expands the facility to all three radiation codes.

It is activated by putting SOLAR_TO_OBIO: .TRUE. in AGCM.rc and then associating the solar export fields DROBIO and DFOBIO either by exporting them in history or adding future connectivity to the OradBioGC. It is off by default and therefore zero-diff. Even if it is on, it is effectively zero-diff ... it only potentially produces two currently unused new exports, now for RRTMGP and Chou-Suarez as well as for RRTMG (see #3).


What's Changed

New Contributors

Full Changelog: v1.1.0...v1.2.0