Difference between revisions of "GEOS-Chem v9-01-02 benchmark history"

From Geos-chem
Jump to: navigation, search
Line 59: Line 59:
 
|-
 
|-
 
|At 500 hPa, list all species that changed by 10% or more:
 
|At 500 hPa, list all species that changed by 10% or more:
|ALK4, ISOP, MEK, ALD2, RCHO, PPN, R4N2, PRPE, C3H8, C2H6. NH3 and NIT exhibit what seem to be randomly noisy changes at 500 hPa. Is this a plotting bug?
+
|ALK4, ISOP, MEK, ALD2, RCHO, PPN, R4N2, PRPE, C3H8, C2H6. NH3 and NIT exhibit what seem to be randomly noisy changes at 500 hPa.
 
|-
 
|-
 
|Comments on 500 hPa differences:
 
|Comments on 500 hPa differences:
|The changes over and downwind of southern Africa propagate to the free troposphere for some of the species. Changes are weaker than at the surface and exhibit the expected patterns (with the exception of NH3 and NIT).  
+
|The changes over and downwind of southern Africa propagate to the free troposphere for some of the species. Changes are weaker than at the surface and exhibit the expected patterns. For NH3 and NIT the 500 hPa absolute changes are near zero so this is apparently just noise in the ratio b/c of near-zero concentrations.
 
|-
 
|-
 
|In the ZONAL MEAN differences, list all species that changed by 10% or more:
 
|In the ZONAL MEAN differences, list all species that changed by 10% or more:

Revision as of 15:34, 12 August 2011

On this page we have posted complete information about all benchmark simulations (both 1-month and 1-year) for GEOS-Chem v9-01-02.

v9-01-02h

Here is the assessment form for 1-month benchmark simulation v9-01-02h.

Description
New features added into GEOS-Chem:
Developer name(s) and institution(s):
  • Dylan Millet (U. Minnesota)
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Thu Aug 11 11:30:21 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-05-c.as.harvard.edu
  • Wall time: 00:58
  • Scalability: 7.6116
Compared to previous benchmark: v9-01-02g
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Diagnostics, Other (please specify):
Plots may be viewed at: Primary site (available Fri 8/12): ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02h/

Quick-look site (available now): ftp://ftp.as.harvard.edu/pub/exchange/bmy/1mo_quick_look/v9-01-02h/

Metrics
Global mean OH (from log file): 13.1718562359138 x 105 molec cm-3
Methyl chloroform lifetime: 4.7556 years
Did either of these change by more than 5%? No
At the SURFACE, list all species that changed by 10% or more: Large changes (> 50%) for: ALK4, PAN, ACET, MEK, ALD2, RCHO, PPN, R4N2, PRPE, C3H8, C2H6, N2O5, MP. Smaller changes (~10-25%) for NOx, ISOP, HNO3, H2O2, MVK, MACR, PMN, CH2O, HNO4, DMS, SO4, NH3, NH4, NIT
Comments on SURFACE differences: Changes are all over and immediately downwind of southern Africa, as expected. Large changes to the anthropogenic hydrocarbons because their emissions were strongly decreased over Botswana to correct scaling bug. Large changes to some of their immediate oxidation products (ACET, RCHO, PPN, ALD2, PAN, etc.). Some smaller changes to other species due to secondary chemical effects (OH seems to have increased somewhat in the outflow from southern Africa because we no longer have the artificially elevated VOC concentrations).
At 500 hPa, list all species that changed by 10% or more: ALK4, ISOP, MEK, ALD2, RCHO, PPN, R4N2, PRPE, C3H8, C2H6. NH3 and NIT exhibit what seem to be randomly noisy changes at 500 hPa.
Comments on 500 hPa differences: The changes over and downwind of southern Africa propagate to the free troposphere for some of the species. Changes are weaker than at the surface and exhibit the expected patterns. For NH3 and NIT the 500 hPa absolute changes are near zero so this is apparently just noise in the ratio b/c of near-zero concentrations.
In the ZONAL MEAN differences, list all species that changed by 10% or more:
Comments on ZONAL MEAN differences:
In the EMISSION RATIO maps, list all species that changed by 10% or more:
Comments on EMISSION RATIO differences:
Additional or summary comments:
Approval
Requires further investigation:
Approved:
Date of approval:

--Bob Y. 14:09, 10 August 2011 (EDT)

v9-01-02g

Here is the assessment form for 1-month benchmark simulation v9-01-02g.

Description
New features added into GEOS-Chem:
Developer name(s) and institution(s):
  • Sungshik Patrick Kim (Harvard)
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Wed Aug 3 15:00:30 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-02-c.as.harvard.edu
  • Wall time: 01:32
  • Scalability: 7.6115
Compared to previous benchmark: v9-01-02f
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Diagnostics, Other (please specify):
Plots may be viewed at: ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02g/
Metrics
Global mean OH (from log file): 13.1512226781986 x 105 molec cm-3
Methyl chloroform lifetime: 4.7648 years
Did either of these change by more than 5%? No
At the SURFACE, list all species that changed by 10% or more: None
Comments on SURFACE differences:
At 500 hPa, list all species that changed by 10% or more: NH3, NIT
Comments on 500 hPa differences: These are are likely due to small number differences, probably caused by ISORROPIA.
In the ZONAL MEAN differences, list all species that changed by 10% or more: NIT
Comments on ZONAL MEAN differences: Caused by small number differences near the limit of single precision.
In the EMISSION RATIO maps, list all species that changed by 10% or more: None
Comments on EMISSION RATIO differences:
Additional or summary comments: Aerosol optical depths (AOD) are recomputed after chemistry but before diagnostics are archived. This will make sure that any removal of aerosol by chemistry or deposition will be reflected in the archived AOD diagnostics.

There is a known issue in ISORROPIA that can cause random differences, especially in species like HNO3, NH3, and NIT. This issue has not been resolved in v9-01-02g.

In order to remove the effects of the ISORROPIA II issue, we have created this frequency distribution histogram. The histogram reflects results from GEOS-Chem 1-week simulations of v9-01-02g vs v9-01-02f with ISORROPIA turned off. As you can see, 100% of all of the transported tracers fall into the central bin in the histogram. This lets us say with certainty that the small differences that we see in the output plots for v9-01-02g are not caused by the AOD diagnostic updates, but are caused by the bug in ISORROPIA II.

Approval
Requires further investigation: No
Approved: Yes
Date of approval: 10 Aug 2011

--Bob Y. 10:32, 10 August 2011 (EDT)

v9-01-02f

Here is the assessment form for 1-month benchmark simulation v9-01-02f. This is a re-submission of v9-01-02b.

Description
New features added into GEOS-Chem:
Developer name(s) and institution(s):
  • Helen Amos (Harvard)
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Mon Aug 1 13:36:33 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-03-c.as.harvard.edu;
  • Wall time: 01:28
  • Scalability: 7.6258
Compared to previous benchmark: v9-01-02e
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Diagnostics, Other (please specify):
Plots may be viewed at: ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02f/
Metrics
Global mean OH (from log file): 13.1512320861717 x 105 molec cm-3
Methyl chloroform lifetime: 4.7648 years
Did either of these change by more than 5%? No.
At the SURFACE, list all species that changed by 10% or more: ISOP, HNO3, H2O2, MVK, MACR, PMN, PRPE, CH2O, MP, DMS, SO2, SO4, SO4s, NH3, NH4, NIT
Comments on SURFACE differences: Largest changes: H2O2, NIT, NH3
At 500 hPa, list all species that changed by 10% or more: ISOP, HNO3, H2O2, MVK, MACR, PMN, PRPE, CH2O, MP, DMS, SO2, SO4, NH3, NH4, NIT
Comments on 500 hPa differences: Largest changes: H2O2, ISOP, NIT, NH3
In the ZONAL MEAN differences, list all species that changed by 10% or more: ISOP, HNO3, H2O2, MVK, MACR, PMN, PRPE, CH2O, N2O5, HNO4, MP, DMS, SO2, SO4, SO4s, NH3, NH4, NIT,
Comments on ZONAL MEAN differences: Species change by 10% north of 30N and south of 30S.
In the EMISSION RATIO maps, list all species that changed by 10% or more: None.
Comments on EMISSION RATIO differences: None.
Additional or summary comments:
Approval
Requires further investigation: No
Approved: YES
Date of approval: 01 Aug 2011

--Bob Y. 12:32, 1 August 2011 (EDT)

v9-01-02e

Here is the assessment form for 1-month benchmark simulation v9-01-02e.

Description
New features added into GEOS-Chem: Various source-code updates and fixes that will have no impact on the standard 1-month full-chemistry benchmark simulation:
  1. Various bug fixes for the offline Hg simulation
  2. Bug fixes for leap years when using GCAP met fields
  3. Minor bug fixes for emissions routines affecting offline simulations
  4. Fixes for minor issues affecting nested-grid simulations
  5. Fixes for better compatibility with the PGI compiler
  6. Add MERRA SEAICExx fields to the planeflight diagnostic
Developer name(s) and institution(s):
  1. Jenny Fisher (Harvard); Helen Amos (Harvard); Chris Holmes (UC Irvine)
  2. GEOS-Chem Support Team
  3. GEOS-Chem Support Team
  4. Luke Schiferl (Colo. State); Lin Zhang (Harvard)
  5. Gan Luo (SUNY Albany)
  6. Jenny Fisher (Harvard)
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Fri Jul 29 16:00:03 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-02-c.as.harvard.edu
  • Wall time: 01:30
  • Scalability: 7.6186
Compared to previous benchmark: v9-01-02d
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Diagnostics, Other (please specify): We expect to see no changes from v9-01-02d
Plots may be viewed at: ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02e/
Metrics
Global mean OH (from log file): 13.1838752732688 x 105 molec cm-3
Methyl chloroform lifetime: 4.7493 years
Did either of these change by more than 5%? No. Both are identical to the previous run v9-01-02d.
At the SURFACE, list all species that changed by 10% or more: None
Comments on SURFACE differences:
At 500 hPa, list all species that changed by 10% or more: None
Comments on 500 hPa differences:
In the ZONAL MEAN differences, list all species that changed by 10% or more: None
Comments on ZONAL MEAN differences:
In the EMISSION RATIO maps, list all species that changed by 10% or more: None
Comments on EMISSION RATIO differences:
Additional or summary comments: All of the changes were to fix mostly minor issues in (1) the Hg simulation, (2) the GCAP simulation, (3) nested-grid simulations, (4) offline simulations, and (5) the planeflight diagnostics. We did not add any scientific updates to the code; therefore, the standard chemistry simulation was not affected. The output file from this benchmark (ctm.bpch.v9-01-02e) is 100% binary identical to the output file from the previous benchmark (ctm.bpch.v9-01-02d).
Approval
Requires further investigation: No
Approved: YES
Date of approval: 01 Aug 2011

--Bob Y. 10:37, 1 August 2011 (EDT)

v9-01-02d

Here is the assessment form for 1-month benchmark simulation v9-01-02d.

Description
New features added into GEOS-Chem:
Developer name(s) and institution(s):
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Tue Jul 26 12:35:13 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-01-c.as.harvard.edu
  • Wall time: 01:07
  • Scalability: 7.4958
Compared to previous benchmark: v9-01-02c
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Diagnostics, Other (please specify): We expect to see no changes from v9-01-02c
Plots may be viewed at: ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02d/
Metrics
Global mean OH (from log file): 13.1838752732688 x 105 molec cm-3
Methyl chloroform lifetime: 4.7493 years
Did either of these change by more than 5%? No.
At the SURFACE, list all species that changed by 10% or more: none
Comments on SURFACE differences:
At 500 hPa, list all species that changed by 10% or more: none
Comments on 500 hPa differences:
In the ZONAL MEAN differences, list all species that changed by 10% or more: none
Comments on ZONAL MEAN differences:
In the EMISSION RATIO maps, list all species that changed by 10% or more: none
Comments on EMISSION RATIO differences:
Additional or summary comments: This benchmark was to test the initial importation of source code for ESMF. We did this in such a way so that the standard chemistry simulation is not affected. The output file from this benchmark (ctm.bpch.v9-01-02d) is 100% binary identical to the output file from the previous benchmark (ctm.bpch.v9-01-02c).
Approval
Requires further investigation: No
Approved: YES
Date of approval: 26 Jul 2011

--Bob Y. 14:52, 26 July 2011 (EDT)

v9-01-02c

Here is the assessment form for 1-month benchmark simulation v9-01-02c.

Description
New features added into GEOS-Chem:
Developer name(s) and institution(s):
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Wed Jul 20 16:47:00 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-06-c.as.harvard.edu
  • Wall time: 00:57
  • Scalability: 7.5008
Compared to previous benchmark: v9-01-02a
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Diagnostics, Other (please specify): Chemical species restart file
Plots may be viewed at: ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02c/
Metrics
Global mean OH (from log file): 13.1838752732688 x 105 molec cm-3
Methyl chloroform lifetime: 4.7493 years
Did either of these change by more than 5%? No. The Mean OH is identical to v9-01-02a and the MCL lifetime has a small difference in the very last digit (see Summary Comments).
At the SURFACE, list all species that changed by 10% or more: none
Comments on SURFACE differences:
At 500 hPa, list all species that changed by 10% or more:
Comments on 500 hPa differences:
In the ZONAL MEAN differences, list all species that changed by 10% or more: none
Comments on ZONAL MEAN differences:
In the EMISSION RATIO maps, list all species that changed by 10% or more: none
Comments on EMISSION RATIO differences:
Additional or summary comments: NOTE: This update corrects an erroneous situation in the chemical species restart file (aka "CSPEC restart file". This fix ONLY gets invoked for long GEOS-Chem simulations that have to be broken into several stages for the purpose of submitting to computational queues with fixed time limits. When each run stage ends, GEOS-Chem will write the chemical species concentrations into a restart file. When the next run stage begins, GEOS-Chem will read the chemical species information back into the CSPEC array (which gets passed to the SMVGEAR and KPP solvers).

For short simulations that can finish within the time limits of the computatational queues (such as this 1-month benchmark simulation), we do not use the CSPEC restart file. Therefore, when comparing v9-01-01c to v9-01-01a, we should expect to see identical results to v9-01-02a (which we do see as evidenced by the plots and Mean OH concentration).

In v9-01-02c, we use the grid box surface areas saved out directly from GEOS-Chem (via the ND69 diagnostic) to compute the MCL lifetime. In v9-01-02a, we computed the grid box surface areas using IDL code. The surface areas from both methods should be similar, but perhaps not strictly numerically equal to all double-precision digits. This causes the very slight difference in the last digit of the MCL lifetime for v9-01-02c.

Approval
Requires further investigation: No
Approved: YES
Date of approval: 21 Jul 2011

--Bob Y. 10:51, 21 July 2011 (EDT)

v9-01-02b

GEOS-Chem 1-month benchmark simulation v9-01-02b was withdrawn and re-submitted as v9-01-02f.

--Bob Y. 12:30, 1 August 2011 (EDT)

v9-01-02a

Here is the assessment form for 1-month benchmark simulation v9-01-02a.

Description
New features added into GEOS-Chem:
Developer name(s) and institution(s):
  • Dynamic tropopause fix: Daven Henze (Colorado University) and Jamin Koo (MIT)
  • Out-of-bounds error: GEOS-Chem Support Team
Version, resolution, met fields used: v9-01-02, GEOS-5 (47L), 4x5
1-month benchmark finished on: Mon Jun 13 18:02:07 EDT 2011
Performance statistics:
  • Ran on 8 CPUs of terra-02-c.as.harvard.edu
  • Wall time: 01:29
  • Scalability: 7.6141
Compared to previous benchmark: v9-01-01
This update will impact:
(select all that apply with boldface)
Advection, BL Mixing, Convection, Met Fields, Dry Dep, Wet Dep, Stratosphere, Anthro Emiss, Biogenic Emiss, Biomass Emiss, Photolysis, Chemistry, Other (please specify): Tropopause
Plots may be viewed at: ftp://ftp.as.harvard.edu/pub/geos-chem/1mo_benchmarks/v9-01-02a/
Metrics
Global mean OH (from log file): 13.1838752732688 x 105 molec/cm3
Methyl chloroform lifetime: 4.7490 years
Did either of these change by more than 5%? No
At the SURFACE, list all species that changed by 10% or more: None
Comments on SURFACE differences: None
At 500 hPa, list all species that changed by 10% or more: NIT, NH3
Comments on 500 hPa differences: Follow up calculations indicate that these changes are owing to an unresolved OpenMP issue, not the v9-01-02a code updates. NIT seems particularly sensitive to this issue, and this probably drives changes in NH3.
In the ZONAL MEAN differences, list all species that changed by 10% or more: NOx, N2O5, NIT, NH3
Comments on ZONAL MEAN differences:
  • Small changes in tropospheric OH on order of +/-0.05%.
  • NOx changes follow from the variable tropopause fix and are small in absolute value.
  • There are also small percent changes for most gas-phase species near the tropopause.
  • The changes to NIT and NH3 have a random structure with large (>1000%) differences in spots throughout the troposphere, likely owing to the issue mentioned above. However, absolute changes are still small.
  • There are small (a few percent or less) changes visible for all chemically active gas and aerosol species. These changes are likely owing to both the code updates and the random OpenMP error, with the former having structure around the tropopause and the latter contributing to random scatter throughout the troposphere.
In the EMISSION RATIO maps, list all species that changed by 10% or more: None
Comments on EMISSION RATIO differences: No changes in emissions.
Additional or summary comments:
  • There is an unidentified OpenMP error which contributes to differences shown here, as confirmed by independent tests of consecutive evaluations of a single executable compiled with and without parallelization. The impacts of this OpenMP error are typically less than a fraction of a percent, and negligible in terms of absolute value, with the exception of NIT and NH3. This is an issue to investigate further.
  • I think this code update obfuscates the need to copy CSPEC to CSPEC_FULL and back when variable tropopause is turned on, as this fix now recalls the proper mapping from the previous time step when performing partitioning. Perhaps a future update could clean this up.
Approval
Requires further investigation: See above
Approved: Daven Henze
Date of approval: 06/28/2011

--Bob Y. 10:30, 21 July 2011 (EDT)