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

Revision as of 20:28, 10 August 2011

Overview

BETA RELEASE -- Summer/Fall 2011

• Previous version: GEOS-Chem v9-01-01
• Next version: GEOS-Chem v9-01-03

What's new in this version

GEOS-Chem v9-01-02 contains the following new features:

--Bob Y. 15:09, 10 August 2011 (EDT)

New data directories

The following new data directories have been added for GEOS-Chem v9-01-02. You will have to download the directories relevant to your simulation.

You can download these directories with anonymous FTP or the Wget utility. For instructions, please see Chapter 2.4, Downloading the GEOS-Chem shared data directories in the GEOS-Chem Online User's Guide.

GEOS_1x1/APM_201011                                             # For APM microphysics only
GEOS_1x1/Streets_200607/NH3-Streets-SeasonalScalar.generic.1x1  # For NH3 seasonality in Streets 2000 

GEOS_2x2.5/CO_201003/fossilfuel_andres/annual_v2010             # CO2 annual emissions from CDIAC
GEOS_2x2.5/CO_201003/fossilfuel_andres/monthly_v2010            # CO2 monthly emissions from CDIAC
GEOS_2x2.5/RETRO_201103                                         # RETRO anthropogenic VOC emissions

GEOS_4x5/CO_201003/fossilfuel_andres/annual_v2010               # CO2 annual emissions from CDIAC
GEOS_4x5/CO_201003/fossilfuel_andres/monthly_v2010              # CO2 monthly emissions from CDIAC
GEOS_4x5/RETRO_201103                                           # RETRO anthropogenic VOC emissions

--Bob Y. 12:50, 6 July 2011 (EDT)

Previous issues now resolved in v9-01-02

Patches from v9-01-01

The following patches that were issued after the initial release of GEOS-Chem v9-01-01 are now standardized in v9-01-02:

1. Double counting of biofuel emissions over Asia
2. Updated annual and monthly CO2 emissions from CDIAC
3. Typos in vdiff_mod.f90
4. Negative tracer caused by DO_MERRA_CONVECTION
5. Omitted code in Hg simulation
6. Bug fixes for drydep (ND44) and wetdep (ND17,ND18,ND37,ND38,ND39) diagnostic archival
7. Bug fix for time-in-tropopause (ND54) diagnostic
8. Only apply wetdep error trap when using MERRA met fields
9. Apply washout to the proper fraction of the grid box
10. Better representation of aqueous SO2 chemistry

--Bob Y. 11:51, 24 June 2011 (EDT)

Mis-indexing error in partition.f

This update was tested in the 1-month benchmark simulation v9-01-02a and approved on 28 Jun 2011.

The implementation of the Dynamic tropopause was causing a mis-indexing error in subroutine partition.f. On every timestep, partition.f uses the concentrations of grid boxes at the previous timestep to partition family tracers (e.g. NOx, Ox) into individual species (e.g. NO2, NO, O3, etc.). The way in which this was done did not take into account the fact that some boxes near the tropopause could be in the troposphere at one timestep and in the stratosphere at the next timestep (or vice-versa). Thus, the concentrations that partition.f was sending to the chemistry solver was incorrect.

This problem affected both the forward and adjoint models.

--Bob Y. 10:14, 6 July 2011 (EDT)

Out-of-bounds error in readchem.f

This update was tested in the 1-month benchmark simulation v9-01-02a and approved on 28 Jun 2011.

It was discovered that this code near the end of subroutine readchem.f:

                 ELSEIF ( J == MAPPL(ILBRO2H,NCS) .or.
    &                     J == MAPPL(ILBRO2N,NCS) .or.
    &                     J == MAPPL(ILTRO2H,NCS) .or.
    &                     J == MAPPL(ILTRO2N,NCS) .or.
    &                     J == MAPPL(ILXRO2H,NCS) .or.
    &                     J == MAPPL(ILXRO2N,NCS) ) THEN
                    ITS_NOT_A_ND65_FAMILY(J) = .FALSE.
                    EXIT

will cause an out-of-bounds error if any of the IL* indices are zero. This error manifested itself in a standard 43-tracer simulation.

We have corrected this error by first making sure that each IL* index is nonzero before doing the corresponding IF ( J = ... ) test:

                 ! Avoid out-of-bounds-errors by making sure that each
                 ! IL* index is nonzero before doing the IF test for J. 
                 ! (bmy, 3/1/11)
                 IF ( ILBRO2H > 0 ) THEN 
                    IF ( J == MAPPL(ILBRO2H,NCS) ) THEN
                       ITS_NOT_A_ND65_FAMILY(J) = .FALSE.
                       EXIT
                    ENDIF
                 ENDIF

                 ... etc ...

--Bob Y. 10:42, 24 June 2011 (EDT)

Mis-indexing when reading restart.cspec file

This update was tested in the 1-month benchmark simulation v9-01-02c and approved on 21 Jul 2011.

Please see this post on the "Restart files" wiki page for more information about an indexing problem when reading the restart.cspec.YYYYMMDDhh file.

--Bob Y. 15:22, 21 July 2011 (EDT)

Various bug fixes for the offline Hg simulation

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

The following issues were discovered in the offline Hg simulation shortly after the release of v9-01-01. These have all been fixed in v9-01-02. Please follow the links below for more information:

1. De-hardwire file path BR_DIR in global_br_mod.f
2. Division by zero error
3. Bug fixes for ND03 mercury diagnostics
4. Negative concentrations of aquatic mercury
5. VEGEMIS bug fix for GCAP simulation

--Bob Y. 10:41, 1 August 2011 (EDT)

Bug fix for leap years with GCAP

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

Recall that the GCAP met fields are climatological GCM output, and do not have any leap years (i.e. there are no Feb 29ths).

In GEOS-Chem v9-01-01 and prior versions, routine SET_CURRENT_TIME in time_mod.f) uses the Astronomical Julian Date algorithms (located in julday_mod.f), which assume that there are leap-years. If you ran GEOS-Chem with the GCAP met fields for more than one year, the current code caused the time & date to be computed incorrectly at the start of the second year. GEOS-Chem would die because the wrong met fields would be read in from disk.

We have added an improved algorithm fix into SET_CURRENT_TIME for GCAP met fields only. Whenever a leap-year day (Feb 29th) is encountered, we shall just skip ahead a day to March 1st and compute the corresponding time & date values. However, we must also keep track of how many leap years we have encountered since the start of the GEOS-Chem simulation. This count is needed so that we can increment the current Astronomical Julian Date so as to "skip over" from Feb 29th to Mar 1st for each new leap year that we encounter.

This fix now skips from Feb 28 to Mar 1st for all leap years. We have tested this for multiple leap years and it works fine.

--Bob Y. 10:42, 1 August 2011 (EDT)

Minor bug fixes for emissions routines affecting offline simulations

We have now corrected some bugs that caused errors when running the offline aerosol-only simulation and tagged CO simulation. (NOTE: These errors do NOT occur when running the full-chemistry simulation, only the offline simulations.)

Fixes in biomass_mod.f

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

In biomass_mod.f, there are some IF blocks such as:

           IF ( ITS_A_FULLCHEM_SIM() ) THEN              
              BIOMASS(:,:,IDBCO) = BIOMASS(:,:,IDBCO) * 1.05d0
           ELSE IF ( ITS_A_TAGCO_SIM() ) THEN
              BIOMASS(:,:,IDBCO) = BIOMASS(:,:,IDBCO) * 1.11d0
           ENDIF

However, if your simulation does not use CO (e.g. offline aerosols), then the ID flags IDTCO and IDBCO will both equal zero. If IDBCO = 0, this will cause an out-of-bounds error.

The solution is to add an outer IF block to prevent the code from being executed if IDBCO = 0.

           IF ( IDBCO > 0 ) THEN
              IF ( ITS_A_FULLCHEM_SIM() ) THEN              
                 BIOMASS(:,:,IDBCO) = BIOMASS(:,:,IDBCO) * 1.05d0
              ELSE IF ( ITS_A_TAGCO_SIM() ) THEN
                 BIOMASS(:,:,IDBCO) = BIOMASS(:,:,IDBCO) * 1.11d0
              ENDIF
           ENDIF

--Bob Y. 10:44, 1 August 2011 (EDT)

Fixes for minor issues affecting nested-grid simulations

The following typographical errors affect nested-grid simulations:

Wrong directory specified in nei2005_anthro_mod.f

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

Luke Schiferl wrote:

I have been working with the nested NA simulation and ran into a bug in the code. In GEOS-Chem v9-01-01, line 779 of nei2005_anthro_mod.f should point to the NEI2005_201007 directory rather than NEI2005_200710. There is a comment in line 709 that confirms this.

--Bob Y. 13:34, 27 July 2011 (EDT)

CAC, BRAVO emissions inadvertently turned off in NA nested grid runs

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

Lin Zhang wrote:

I found a small bug in recent GEOS-Chem v9-01-01 when running nested NA. The error is in routine READ_EMISSIONS_MENU of module input_mod.f, around line 1700. The line:

   #if defined( GRID05x0666 ) || defined( NESTED_CH ) 

should be

   #if defined( GRID05x0666 ) && defined( NESTED_CH ) 

or just simply

   #if defined( NESTED_CH )

Otherwise, it will turn off CAC, BRAVO emissions when running nested NA.

--Bob Y. 10:44, 1 August 2011 (EDT)

Bug fixes for compatibility with the PGI compiler

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

Gan Luo from SUNY/Albany pointed out several minor typos that cause the PGI compiler to choke when building the GEOS-Chem executable. These will all be fixed in v9-01-02.

(1) Line 303 of GeosCore/depo_mercury_mod.f:

Change this:
 #include 'define.h'

to this:
 #include "define.h"

(2) Line 169 of GeosCore/ocean_mercury_mod.f:

Change this:
 REAL*8, PARAMETER     :: Kd_part   = 10**(5.5)

to this:
 REAL*8, PARAMETER     :: Kd_part   = 10d0**(5.5d0)

(3) Line 1281 of GeosCore/ocean_mercury_mod.f:

Change this:
 HgPaq(I,J)   = MAX ( HgPaq(I,J) , 0.0 )

to this:
 HgPaq(I,J)   = MAX ( HgPaq(I,J) , 0d0 )

(4) In subroutine DO_MERRA_CONVECTION of convection_mod.f

Change this:
 REAL*8            :: ALPHA,       ALPHA2,   CLDBASE, CMFMC_BELOW

to this:
 INTEGER           :: CLDBASE
 REAL*8            :: CMFMC_BELOW, ALPHA,    ALPHA2

(5) In module vdiff_mod.f90: Several numerical constants had to be rewritten so that explicitly use "D" exponents. In other words:

Change numbers like these:
 1.
 1.e-12

to these:
 1.d0
 1.d-12

In short, the PGI compiler expects constants to agree strictly with the type of the variable that they are being assigned to (i.e. if A is a REAL*8, then you must use A = 1.0d0 instead of A = 1.0). IFORT will correct for this situation during compilation, but PGI will throw an error.

--Bob Y. 10:44, 1 August 2011 (EDT)

Planeflight diagnostic update

This update was tested in the 1-month benchmark simulation v9-01-02e and approved on 01 Aug 2011.

The MERRA SEAICExx fields are now saved to the plane flight diagnostic, per request of Jenny Fisher.

--Bob Y. 10:45, 1 August 2011 (EDT)

Washout fix for non-aerosol species

This update was tested in the 1-month benchmark simulation v9-01-02f and approved on 02 Aug 2011.

The current Wet deposition algorithm was underestimating washout of non-aerosol species. Please see this wiki post for more information about how we corrected this situation.

--Bob Y. 14:34, 2 August 2011 (EDT)

Overhaul of aerosol optical depth diagnostics

This update was tested in the 1-month benchmark simulation v9-01-02g and approved on 10 Aug 2011.

Patrick Kim discovered some inconsistencies in the way aerosol optical depth diagnostics were being computed in GEOS-Chem v9-01-01 and prior versions. We have now corrected these inconsistencies.

Sungshik Patrick Kim wrote:

Here is an example of a daily (first page) and monthly (second page) average AOD for January 2008 from the public release of v9-01-01 and a newer version that addresses the changes we discussed earlier vs. the corresponding average AERONET data, a difference plot, and a histogram of the differences by grid box. [T]he largest differences appear to be due to the change in met field rather than changing the timing of the AOD calculation and it may be largest in the Arctic due to uncertainties in the previous calculation of RH.

--Bob Y. 15:15, 10 August 2011 (EDT)

Fix for GEIA emissions scaling factor over Botswana

Please see this wiki post about a problem in the scale factors for GEIA emissions over Botswana and southern Africa. We are planning to fix this in GEOS-Chem v9-01-02.

--Bob Y. 14:37, 2 August 2011 (EDT)

Fix for vertical regridding of offline OH for offline simulations

Chris Holmes wrote:

I have a bug fix for offline simulations that use offline OH fields. The way the code previously lumped layers in the stratosphere was incorrect for OH. Other offline fields (e.g. O3, Br, BrO) are OK, as long as they are in mixing ratio units.

Unlumped layers are unaffected. OH number densities in the lumped layers increase up to 7%. I tested for Hg only and found that they Hg changes were correspondingly modest.

Here are my comments from the commit:

Offline OH fields are stored as number concentrations (molec/cm3) on a 3D grid with unreduced vertical coordinate (e.g. in the file OH_3Dglobal.geos5.4x5). The model lumps layers in the stratosphere into a reduced vertical grid within global_oh_mod.f. The regridding routine (subroutine transfer_3d) requires input as a mixing ratio, but this assumption was violated for OH fields. To correct this problem, we convert OH number concentration to a mixing ratio, then regrid, then convert back to number concentration.

OH concentrations in unlumped layers are unaffected, while increases up to 7% occur in lumped layers in the stratosphere.

--Bob Y. 15:15, 10 August 2011 (EDT)

Under further investigation

The following updates (which should ideally not yield any differences) seem to cause slight numerical discrepancies when added to the standard code. We shall postpone including these into the standard code, pending investigation.

Minor bug fixes in ISORROPIA II

Please see this wiki post for a description of two proposed bug fixes that should go into source code file ISOROPIA/isorropiaIIcode.f. This bug should not affect the current GEOS-Chem ISORROPIA configuration, but may affect future development.

Update 7/28/11: Implementing this change causes the 1-month standard benchmark simulation to have divergent results than the previous version. Therefore, we should understand the source of these differences before we add this update into the standard repository.

--Bob Y. 14:40, 28 July 2011 (EDT)

Fixes in emep_mod.f

In routines READ_EMEP_UPDATED and READ_EMEP_UPDATED_05x0666 of emep_mod.f, there were some IF statements that were used to pick the proper filename and tracer number with which to read data from the bpch file format:

        IF ( TRACER .eq. IDTNOx ) THEN
           FILENAME = TRIM( DIR ) // 'SeasonalVariation/' 
    &        // 'NOx-EMEP-SeasonalScalar.geos.1x1'
           RATIOID = 71
        ELSEIF ( TRACER .eq. IDTCO ) THEN
           FILENAME = TRIM( DIR ) // 'SeasonalVariation/'
    &        // 'CO-EMEP-SeasonalScalar.geos.1x1'
           RATIOID = 72
        ELSEIF ( TRACER .eq. IDTSO2 ) THEN
           FILENAME = TRIM( DIR ) // 'SeasonalVariation/'
    &        // 'SOx-EMEP-SeasonalScalar.geos.1x1'
           RATIOID = 73
        ELSEIF ( TRACER .eq. IDTNH3 ) THEN
           FILENAME = TRIM( DIR ) // 'SeasonalVariation/'
    &        // 'NH3-EMEP-SeasonalScalar.geos.1x1'
           RATIOID = 74
        ENDIF

where TRACER is passed to these routines via subroutine calls such as these:

        CALL READ_EMEP_UPDATED(  1, EMEP_YEAR, EMEP_NOx, 0 )
        CALL READ_EMEP_UPDATED(  4, EMEP_YEAR, EMEP_CO,  0 )
        CALL READ_EMEP_UPDATED( 26, EMEP_YEAR, EMEP_SO2, 0 )
        CALL READ_EMEP_UPDATED( 30, EMEP_YEAR, EMEP_NH3, 1 )

The problem is the code is set up to assume that IDTNOX will always be 1, IDTCO will always be 4, IDTSO2 will always be 26, and IDTNH3 will always be 30. This is true only if you are running a full-chemistry simulation. However, for an offline-aerosol simulation NOx and CO are undefined, and SO2 and NH3 will not equal 26 and 30. Therefore, the IF statement will not return the proper file name and tracer ID.

The solution is to remove all references to IDTNOX, IDTCO, IDTSO2, IDTNH3 from that section of code. The new code looks like:

        ! Pick proper filename & tracer #
        SELECT CASE( TRACER )

           ! NOx
           CASE( 1 ) 
              RATIOID  = 71
              FILENAME = TRIM( DIR )                        // 
    &                    'SeasonalVariation/'               //
    &                    'NOx-EMEP-SeasonalScalar.geos.1x1'
           ! CO
           CASE( 4 )
              RATIOID  = 72
              FILENAME = TRIM( DIR )                        // 
    &                    'SeasonalVariation/'               //
    &                    'CO-EMEP-SeasonalScalar.geos.1x1'
           ! SO2
           CASE( 26 )
              RATIOID  = 73
              FILENAME = TRIM( DIR )                        // 
    &                    'SeasonalVariation/'               //
    &                    'SOx-EMEP-SeasonalScalar.geos.1x1'
           ! NH3
           CASE( 30 )
              RATIOID  = 74
              FILENAME = TRIM( DIR )                        // 
    &                    'SeasonalVariation/'               //
    &                    'NH3-EMEP-SeasonalScalar.geos.1x1'
        END SELECT

There are a couple more instances of this error in emep_mod.f, but the fix is the same.

Update 8/1/11: Implementing this change causes the 1-month standard benchmark simulation to have divergent results than the previous version. Therefore, we should understand the source of these differences before we add this update into the standard repository.

--Bob Y. 10:43, 1 August 2011 (EDT)

Fix for offline simulations in nei2005_anthro_mod.f

Routines EMISS_NEI2005_ANTHRO and EMISS_NEI2005_ANTHRO_05x0666 in nei2005_anthro_mod.f contained similar errors to the errors in emep_mod.f described above. Again, the crux of the problem is that the IDTxxxx tracer flags are being used for the indices of tracers in the binary punch file. And as described above, the various IDTxxxx flags will have different values for an aerosol-only simulation as for a full-chemistry simulation.

The fix is to keep a separate array of tracer indices in the code, and use those for aerosol-only simulations:

     ! list of ID of available species
     SPECIES_ID = (/ IDTNOX,  IDTCO,   IDTSO2,  IDTSO4,  IDTNH3,
    &                IDTACET, IDTALK4, IDTC2H6, IDTC3H8, IDTOCPI, 
    &                IDTBCPI, IDTALD2, IDTCH2O, IDTPRPE, IDTMEK   /)

     ! ID #'s for that are not tied to IDTxxxx flags
     OFFLINE_ID = (/ 1,       4,       26,      27,      30,
    &                9,       5,       21,      19,      35,           
    &                34,      11,      20,      18,      10       /) 

     ! Loop over species
     DO KLM = 1, SIZE( SPECIES_ID )

        ! Species ID (IDT* flags)
        SNo         = SPECIES_ID( KLM )

        ! Skip undefined tracers (bmy, 4/11/11)
        IF ( SNo == 0 ) CYCLE
           
        ! Pick the proper tracer # for fullchem or offline simulations
        ! This prevents crashes for e.g. offline aersol sims (bmy, 4/11/11)
        IF ( ITS_A_FULLCHEM_SIM() ) THEN
           BPCH_ID = Sno
        ELSE
           BPCH_ID = OFFLINE_ID(KLM)
        ENDIF

where BPCH_ID is the index for the given species in the binary punch file. If you are running the full-chemistry simulation, then BPCH_ID is defined from the IDTxxxx flags. If you are running an offline aerosol-only simulation, then BPCH_ID is defined from the values in the OFFLINE_ID array.

Update 7/28/11: Implementing this change causes the 1-month standard benchmark simulation to have divergent results than the previous version. Therefore, we should understand the source of these differences before we add this update into the standard repository.

--Bob Y. 15:42, 28 July 2011 (EDT)

Outstanding issues not yet resolved in v9-01-02

Apparent optimization or parallelization error in ISORROPIA II

Please see this description of small random errors that are apparently being generated when ISORROPIA II is run in parallel with compiler optimization turned on. This is also present in v9-01-01. Shannon Capps (Ga. Tech) is currently investigating this.

--Bob Y. 10:52, 1 August 2011 (EDT)

Apparent bug in LINOZ

Prasad Kasibhatla wrote:

There is a bug in LINOZ related to calculation of column ozone associated with very tiny negative O3 mixing ratios in strat, that can cause the model to crash. This is compiler dependent because sometimes the tiny concentrations appear as tiny positive numbers and the model does not crash. Dylan Jones. is looking into this - a quick fix is possible to prevent the model from crashing, but a bigger issue is the non-physical tiny O3 concentrations that occassionally occur in stratosphere - I suspect this has something to do with handling of STT and CSPEC arrays when a box goes [from the troposphere to the stratosphere] in a time-step.

--Bob Y. 10:52, 1 August 2011 (EDT)

Inconsistencies when partitioning tracers into species

Prasad Kasibhatla wrote:

There seem to be some inconsistencies in how [the CSPEC, CSPEC_FULL, and STT] arrays are used in the code. I am working with Daven Henze to sort out, and hope to have this taken care off by end of Aug. I would however recommend that a sub-group take a careful look at how partioning of transported species before chemistry and lumping of chemical species after chemistry is done to ensure that there are no remaining bugs.

--Bob Y. 10:52, 1 August 2011 (EDT)