[WIP] Logging updates for GCHP #846
Closed
Conversation
This file contains 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
…ISTORY This commit fulfills feature request geoschem/geos-chem #571. GeosUtil/grid_registry_mod.F90: - Allocate LONBND array and register as GRID_LONBND - Allocate LATBND array and register as GRID_LATBND History/histcontainer_mod.F90 - Add NB (# of bounds = 2) and bDimId (netCDF dimension Id for bounds) History/histitem_mod.F90 - Add NcBDimId (netCDF dimension for bounds) - Set NcChunkSizes properly for LONBND (dims: bx) or LATBND (dims: by) History/history_mod.F90: - Assing Collection%NB=2 for all collections History/history_netcdf_mod.F90: - Pass Container%bDimId to Nc_Create - Pass Container%bDimId and VarBounds to Get_Var_DimIds - Modify Get_Var_DimIds for LONBND (dims: bx) and LATBND (dims: by) - Pass Current%Item%NcBDimId and VarBounds to Nc_Var_Def - Update CASE statement and Get_Var_DimIds to write out LONBND as "lon_bnds" and LATBND as lat_bnds" - Give "lon" and "lat" the bounds attribute - Update routine IndexVarListCreate to add lon & lat bounds to the list of netCDF index variables NcdfUtil/ncdf_mod.F90: - Nc_Create and Nc_Var_Def now accept nBounds, boundsId arguments Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
In history_netcdf_mod.F90, the lon:axis ="X" and lat:axis = "Y" were inadvertently removed. These have been restored. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
This commit fixes an issue reported by Viral Shah and Chris Holmes in Github issue geoschem/geos-chem #743. The terms 2*ff**2 and 2*kk**2 should have been 2*ff and 2*kk. This corresponds to Eq 3 of Holmes et al, GRL, 2019. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
This commit fixes the issue reported by Eloise Marais in Github bug report geoschem/geos-chem #747. Because ObsPack pressure (the State_Diag%ObsPack_P field) is taken from the GEOS-Chem State_Met%PMID field, the units should be hPa. This is now fixed. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
…onment module on Pleiades
Following the Pull request: #511 Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
We have updated the species_database.yml file with the metadata for the new trace metal simulation species. Most of this metadata can be taken from the DST1, DST2, DST3, and DST4 properties. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
(1) Added template files in run/GCClassic
- HEMCO_Config.rc.metals
- HISTORY.rc.metals
- input.geos.metals
(2) Updated run/GCClassic/createRunDir.sh to allow trace metal sims.
(3) Fixed a few issues in the species_database.yml file.
(4) Added Input_Opt%ITS_A_TRACEMETAL_SIM in Headers/input_opt_mod.F90.
(5) Added "metals" as a new simulation type in input_mod.F90.
(6) Allow tracemetal simulation in hco_interface_gc_mod.F90.
(7) Trimmed trailing whitespace.
Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
(1) Removed an extraneous "-" from the HEMCO_Config.rc.metals file.
(2) Updated ./createRunDir.sh to look for the metals restart file
in GEOSCHEM_RESTARTS/v2021-06.
Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
In test/intTestCreate.sh, add the proper keystroke sequences to set up the following integration tests: - gc_2x25_metals_merra2 - gc_2x25_metals_geosfp - gc_4x5_metals_merra2 - gc_4x5_metals_geosfp Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
Replace "_" with "." in the restart fle name for the metals simulation. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
The default global anthropogenic emissions have been updated from CEDS GBD-MAPS (1970-2017) to CEDS v2 (1750-2019). CEDS GBD-MAPS, which also allow for separation by fuel type are still included as an option. This commit addresses feature request #757. Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
The CEDS entries in HEMCO_Config.rc.tagCO were still using the sectors from the CEDS GBD-MAPS (NRTR, ROAD, RCOR, RCOO) but should now be using those from the CEDS v2 (TRA, RCO). Also fix the logicals for the CHINA_MASK in that file so that it is read if CEDSv2, CEDS_GBDMAPS, or CEDS_GBDMAPS_byFuelType is set to true. Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
Reading in the restart file cannot handle species with a "_" in their name, so we have renamed all trace metal species in the input.geos, HEMOC_Config.rc, and species_database.yml templates. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
The HEMCO_Config.rc.metals template file specified the meteorology section but the meteorology entries are now inserted automatically by the createRunDir.sh script. This was causing the HEMCO code We have removed the manual entries and have left only the template. This fixes the issue. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
The CEDS entries in ExtData.rc were still using the sectors from the CEDS GBD-MAPS (NRTR, ROAD, RCOR, RCOO) but should now be using those from the CEDS v2 (TRA, RCO). Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
We have removed "_" from species names (i.e. Al_F1 is now AlF1, etc), so we had to make the corresponding change in the HEMCO_Diagn.rc.metals template file. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
…ullname for new trace metals Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
Jenny Fisher provided scripts for updating to the latest MODIS LAI product from BNU (v6). This data has been further processed to separate the MODIS LAI field into fields for each land type (aka Yuan-processed MODIS XLAI). The only update required for these new files is to change the file path in HEMCO_Config.rc and ExtData.rc from Yuan_XLAI/v2019-03 to Yuan_XLAI/v2021-06. This commit addresses feature request #717. Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
When using the CEDS GBD-MAPS emission, HEMCO_Config.rc files with the emissions entries are read in via >>>include statements. There was a typo in the file name for these HEMCO_Config.rc files -- "CEDS_GBD-MAPS" should be "CEDS_GBDMAPS". Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
…improve run time
As part of the GEOS-Chem release proposal the GCST sent to the GEOS-Chem
Steering Committee, it was proposed that several diagnostics should be
turned off to increase benchmark run time. From that document:
Diagnostics currently take up about 25% of run-time for full chemistry
benchmarks. This can be reduced by limiting diagnostic output to only
data needed for benchmarking. We propose changes to the following
diagnostic collections in benchmark runs:
* Budget
- Disable for both 1-year and 1-month benchmarks
* DryDep, ProdLoss, WetLossConv, and WetLossLS
- Disable for 1-month benchmarks
- Enable for 1-year benchmarks, but only output Ox species
* CloudConvFlux, ConcAfterChem, KppDiags, LevelEdgeDiags, StateChm
- Disable for both 1-year and 1-month benchmarks
* StateMet
- Limit to only output fields needed for post-processing
Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
In GeosCore/ucx_mod.F90:
(1) In routine CALC_STRAT_AER, make sure that all PRIVATE variables
are initialized or zeroed in the !$OMP PARALLEL loop. This will
help to prevent numerical instability issues.
(2) In routine CALC_SLA_GAMMA, set RXNGAMMA(2), RXNGAMMA(5),
RXNGAMMA(9), and RXNGAMMA(11) to 0.0_f8 instead of TINY(1e0_f8).
The TINY function returns the smallest positive number representable
by REAL*8 (which is ~1e-308), which is very close to the range
of denormal numbers. Using 0.0_f8 will set these elements to
exactly zero, which should eliminate any numerical noise or
numerical instability.
Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
…he AEIC emissions were still emitted in units of kgC instead of kg species. This bug fix requires the definition of two new scale factors, CtoMACR and CtoRCHO.
sulfate_mod.F90 - Add _fp or _f8 to numerical constants where missing wetscav_mod.F90 - Add INV_T298 parameter - Make pH an argument of Compute_L2G for both Luo_Wetdep and default wetdep, passing in the proper pH value in the subroutine call - Remove string comparison tests in IF blocks where special handling has to be done for SO2 and NH3. Replace with tests against the species index. - Streamline Henry's law computations for SO2 and NH3, to avoid recomputing terms that only need to be computed once. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
run/shared/species_database.yml: - The MW_g of EOH (methanol) should be 46.07, which matches https://pubchem.ncbi.nlm.nih.gov/compound/Ethanol run/GCClassic/HEMCO_Config.rc.templates/HEMCO_Config.rc.fullchem run/GCHP/HEMCO_Config.rc.templates/HEMCO_Config.rc.fullchem - Scale factor CtoC3H8 should use a mol wt of 44.11 - Scale factor CtoEOH should use a mol wt of 46.07 Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
The new Luo et al 2020 wet deposition scheme requires the value of State_Chm%phCloud even for non-fullchem or aerosol routines. We have now moved this to the #ifdef LUO_WETDEP block where the other Luo et al 2020-specific fields are initialized. Also, based on a communication from Gan Luo, we make sure to set State_Chm%pHcloud to an initial value of 5.6. This value is appropriate for cloud water (it accounts for dissolved CO2 in cloud water droplets). For the fullchem and aerosol simulations, State_Chm%pHcloud will be computed in the sulfate chemistry routines. But for other simulations, the default value of State_Chm%pHcloud = 5.6 will be used. These updates are necessary to avoid an out-of-bounds error caused by State_Chm%pHcloud not being allocated in simulations other than fullchem or aerosol-only. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
Prior to this commit, a pH value of 5.0 was passed to Compute_Ki in routine Compute_F. We now pass it the value of State_Chm%pHcloud, which by default is set to 5.6. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
The single quotes were removed from NO in commit 1abb448. This, however, causes an issue in GCPy so that the properties for NO are no longer found. To resolve this, we will restore the single quotes. Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
State_Chm%pHcloud was formerly initialized & registered within an IF block that only executed if wetdep & convection was on, and within an #ifdef LUO_WETDEP block too. We have now moved the init/register for State_Chm%pHcloud out of these blocks. This should now allocate State_Chm%phCloud for all simulation types. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
This brings the Luo et al 2020 wet deposition algorithm into the
the dev branch, which will be released as GEOS-Chem Classic 13.2.0.
Resolved conflicts in:
(1) run/shared/species_database.yml
- Merge LuoWd parameters into the existing species_database.yml
containing the trace metals species
(2) test/GCClassic/intTestCreate.sh
- Add LuoWd integration tests
Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
It was brought to our attention that the Luo et al 2020 wet deposition updates had modified some default values in the species_database.yml file. But these should only be applied when using the -DLUO_WETDEP=y option at configure time. We have added modifications to the species_database.yml file, and to the species_database_mod.F90 so that the Luo et al 2020 modifications will not overwrite default values. We have added extra species database tags: 1. DD_DvzAerSnow_Luo 2. DD_DvzMinVal_Luo 3. Henry_CR_Luo 4. Henry_K0_Luo 5. WD_ConvFacI2G_Luo 6. WD_LiqAndGas_Luo 7. WD_RetFactor_Luo that hold values specific to the Luo et al 2020 wetdep scheme. We have also modified the species_database_mod.F90 to apply these values accordingly. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
The DMS entry in the species database had a bunch of extraneous entries following the MW_g tag. These were left over from a Git merge. DMS was incorrectly tagged with Is_Aerosol=true, but did not have a KcScale tag defined. Thus, KcScale was set to [-999, -999, -999], which in turn caused the exponential expression in the GET_F function to blow up. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
Br2 was removed from online seasalt in GEOS-Chem 12.9. The diagnostics is therefore all zeros even when online seasalt is enabled. Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
run/GCClassic/HEMCO_Config.rc.templates/HEMCO_Config.rc.metals - Added a missing "_" in the container name NEI11_INLN_PTEGU_Si_F1, which caused a configuration-time error. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
Headers/species_database_mod.F90 - In the #ifdef LUO_WETDEP block, local variable wd_convfaci2g_luo was being assigned to ThisSpc%WD_RetFactor. It should have been assigned to ThisSpc%WD_ConvFacI2g. This error caused fullchem runs with Luo Wetdep turned to halt when doing wetdep of SO2 because the ConvFacI2G field for SO2 was set to -999. This has now been fixed. - Also moved the #ifdef LUO_WETDEP block to after the SANITY CHECKS section to make sure we don't clobber any other species database values that should be assigned when Luo Wetdep is turned on. Signed-off-by: Bob Yantosca <yantosca@seas.harvard.edu>
… in HEMCO_Config.rc Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
The HEMCO configuration file for GCHP's CO2 simulation was actually a modified configuration file from the TransportTracer simulation and included entries for emissions not needed for the CO2 simulation. The file has now been updated so it contains the same emissions options as the GEOS-Chem Classic CO2 simulation. The CO2 CMS Flux emissions currently used in GCHP's CO2 simulation are still included in this file and are on by default, while the other CO2 emission options used in GEOS-Chem Classic remain off. The Yuan-processed MODIS LAI entries have also been updated in HEMCO_Config.rc for the CO2 simulation, but these entries aren't used in the GCHP simulations since those fields are read directly by ExtData. Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
… Yuan-processed MODIS LAI by default Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
These scripts are used by GCST at Harvard for testing. They are provided as examples for streamlining submission of multiple consecutive GCHP runs. Changes include: - Number of runs is now an argument passed to gchp.multirun.sh - Monthly diag code in gchp.multirun.run is commented out; it is kept as an example of how to auto-update freq/dur in HISTORY.rc for irregular run durations, and for use of instantaneous diagnostics in monthly runs. Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
…cers The suggest memory may be confusing given possible run scenarios. Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
Previously benchmarking used 48 cores over 2 nodes. This update speeds up the GCHP C48 benchmark runs to be closer in timing to GC-Classic 4x5 runs. Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
GCHP integration tests now include only one run per simulation. Across the three type of simulations are runs with different meteorology source and grid resolution. Tests do not yet include RRTMG, but are limited to Transport Tracers, Benchmark, and Standard simulations. This update also includes a bug fix where sed commands in the GCHP run script used in integration test run directories hard initial resource allocation in runConfig.sh hard-coded. This made the integration tests break if the defaults in runConfig.sh change, which they did in a recent commit. The initial values are now wildcards to prevent this in the future. Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
Previously this was omitted from GCHP because the field was not used in GEOS-Chem. It is not used for sea salt aerosol source from blowing snow in the sea salt HEMCO extension. Signed-off-by: Lizzie Lundgren <elundgren@seas.harvard.edu>
Following the benchmarking of GEOS-Chem 13.2.0 with the Luo et al. (2020) wet deposition scheme, Gan Luo noticed a bug in the code. He wrote: Simulated sulfate mass concentration looks lower than what I expected. I checked https://github.com/geoschem/geos-chem/blob/13.2.0-beta.0/GeosCore/sulfate_mod.F90 It seems that we missed some switches for Non-volatile aerosol concentration calculation: (1) At line 5142 D = (2.e+0_fp*SO4nss) - (TNA+2.e+0_fp*TDCA), it should be: #ifdef LUO_WETDEP D = (1.5e+0_fp*SO4nss) - (TNA+2.e+0_fp*TDCA) #else D = (2.e+0_fp*SO4nss) - (TNA+2.e+0_fp*TDCA) #endif (2) At line 5156 D = (2.e+0_fp * SO4nss) - TNA), it should be: #ifdef LUO_WETDEP D = (1.5e+0_fp * SO4nss) - TNA #else D = (2.e+0_fp * SO4nss) - TNA #endif Switching from 2.e+0_fp * SO4nss to 1.5e+0_fp * SO4nss is following the reasons introduced in Luo et al. (2020). Signed-off-by: Melissa Sulprizio <mpayer@seas.harvard.edu>
|
I'll update the base when I have a commit on top of |
|
When I was looking into this I came across GEOS-ESM/MAPL#1002. After that fix the errors are quite a bit better. Closing this for now. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This is a work in progress (WIP).
Objectives
1. Improve error reporting for common pitfalls
The reporting of errors related to reading config files could be improved. The config files are parsed in
Chem_GridCompModso that's the module I'm focusing on. Two updates:__RC__calls withrc=STATUSfollowed by_ASSERT_RC()followed by_VERIFY().Details on 1.1.
__RC__is a PP macro defined byin MAPL/include/MAPL_Exceptions.h. In places like loading config files (e.g., here) these __RC__ calls mean that errors aren't reported very well. For example, if parsing GCHP.rc fails the following non-fatal error is reported
and there is no information about the offending file. Note that the staus is actually helpful,
status=8refers toMAPL_STRING_TOO_SHORTdefined here. We just need it to be printed.The call we want to make is
where
condition=(STATUS ==MAPL_SUCCESS),return_code=STATUS, andrc=MAPL_ERROR_PARSING_CONFIG_FILE.Soon there will be the
_ASSERT_RC()PP macro defined bywhich is close to what we want. It would be better to use
__FILE__though so the full path is printed, and_ASSERT_RCisn't actually available yet.The call we want to make is
_ASSERT_MSG_AND_LOC_AND_RC(STATUS==MAPL_SUCCESS, 'Error parsing GCHP.rc', __FILE__, __LINE__, MAPL_ERROR_PARSING_CONFIG_FILE).