Tropospheric chemistry mechanism

From Geos-chem
Revision as of 17:27, 22 February 2019 by Melissa Payer (Talk | contribs) (Previous issues that are now resolved)

Jump to: navigation, search

On this page you can find information on GEOS-Chem's tropospheric chemistry mechanism (formerly referred to as "NOx-Ox-HC-aerosol-Br").

Recent chemistry updates

The following table lists notable updates to the GEOS-Chem tropospheric chemistry ("tropchem") mechanism. For more information about a particular feature, click on the appropriate link below.

Version Features
GEOS-Chem 12.0.0
(aka "v11-02-final")
GEOS-Chem v11-01
GEOS-Chem v10-01
GEOS-Chem v9-02
GEOS-Chem v9-01-03
GEOS-Chem v8-03-01
GEOS-Chem v8-02-04
GEOS-Chem v8-02-03
GEOS-Chem v8-02-01

--Bob Y. (talk) 21:07, 27 May 2015 (UTC)


GEOS-Chem v10-01 and newer versions now use the FAST-JX v7.0 photolysis mechanism. Prior versions used the older FAST-J photolysis mechanism.

--Bob Y. (talk) 21:03, 27 May 2015 (UTC)

CH4 concentrations

Latitude bands

Katherine Travis has added CH4 latitudinal data for years 2007 through 2013 in GEOS-Chem v10-01 (approved 15 Jun 2015). For more information, please see this post on our GEOS-Chem v10-01 wiki page.

Katherine Travis has added CH4 latitudinal data for years 2014 through 2016 in GEOS-Chem v11-02d. For more information, please see this post on our GEOS-Chem v11-02 wiki page.

Chris Buten wrote:

When running the NOxOxHC simulation, I presume that CH4 concentration is specified as an input and not determined by emissions and transport (and therefore not a tracer). If so, is there a way to easily visualize or extract these input concentrations?

Bob Yantosca wrote:

Yes, in the full chemistry mechanism, CH4 is a constant. The values for years and latitude bands are set by routine GeosCore/get_global_ch4.F.

--Bob Y. 09:52, 14 July 2008 (EDT)

Lee Murray wrote:

The methane concentrations used for non-UCX full chemistry runs in GEOS-Chem (fixed throughout the troposphere to annual zonal mean values in four bands) have not have their values updated since 2008, which was the tail end of the hiatus in the growth rate. I can provide updated values through the present. This is something we should flag as necessary for every public model release — it's trivial to do, I'm happy to be point person.

--Melissa Sulprizio (talk) 18:06, 20 May 2015 (UTC)

Monthly mean surface distributions

This update was included in v11-02e (approved 24 Mar 2018).

Lee Murray wrote:

I have prepared spatially kriged (interpolated) monthly mean surface methane distributions from the NOAA GMD flask observations from the early 1980s through the present in a NetCDF file. This allows for seasonality in the methane concentrations and local maxima (e.g., south central US; western Russia) and minima. It is not appropriate to use these surface distributions fixed throughout the entire column (spatial heterogeneity will be reduced in the free troposphere), but I use them in all my current simulations to fix surface concentrations and advect methane at the cost of an additional transported tracer. Since the surface concentration is fixed, tropospheric chemistry equilibrates much faster than if we were to have online methane emissions, on the time-scale of tropospheric mixing. UCX does the same thing for its treatment of methane by default, but with a single mean surface value. Fixing surface concentrations and allowing advection is common treatment in other atmospheric chemistry models.
It is now generally standard practice for global models to prescribe surface methane concentrations and allow it to advect. It doesn’t have large impacts on the short-lived chemistry, but my maps eliminate the artificial step gradients that the zonal bands impose and introduce realistic and dynamic spatial gradients in the troposphere. I think it is worth putting in tropchem as an advected tracer, but at the very least, we should use my mean meridional gradients to impose the fixed concentrations.
It’s no more work to maintain than the current treatment, which is constantly falling behind (and the UCX surface data is even further behind than tropchem).
The NetCDF methane files are ready and contain observations from Feb 1983 through Dec 2016, with extrapolated trends for 1979-2020.

--Melissa Sulprizio (talk) 16:49, 28 November 2017 (UTC)

Bromine chemistry

Please see our Bromine chemistry mechanism wiki page for complete information about the bromine species used in this mechanism.

--Bob Y. (talk) 21:07, 27 May 2015 (UTC)

Removal of NOx and Ox partitioning

This update was tested in the 1-month benchmark simulation v9-02h and approved on 15 May 2013.

Family tracers NOx and Ox have been removed and are now replaced with their individual constituents: NO, NO2, NO3, HNO2, and O3. In order to remove NOx and Ox, the following updates were also made:

  1. NOx emissions are now NO emissions and Ox emissions are now O3 emissions. This is consistent with the previous treatment of NOx and Ox emissions. Previously, in the Tracer Menu of input.geos, NO and O3 were denoted with parentheses, indicating NOx was emitted as NO and Ox was emitted as O3.
  2. Dry deposition of NOx is now dry deposition of NO2.
  3. Routine OHSAVE has been updated to remove FRACO3, FRACNO, FRACNO2, SAVENO, SAVENO2, and SAVENO3. These are no longer needed because O3, NO, NO2, and NO3 are now tracers.
  4. Routine PARTITION has been updated to remove partitioning of NOx and Ox.

--Melissa Payer 15:21, 1 April 2013 (EDT)

Evolution of mean OH

Please see our wiki page which tracks the evolution of the mean OH concentration taken from the GEOS-Chem 1-year benchmark simulations.

--Bob Y. 12:56, 16 February 2010 (EST)

Isoprene papers

Mat Evans wrote:

I thought it would be useful to put together all the isoprene papers that are coming out at the moment.

New papers

Mat Evans wrote:

This page is somewhere to put comments about new papers that have a relevance to the chemistry scheme in GEOS-Chem and assocated parameters.

Acetone treatment in GEOS-Chem

Bryan Duncanwrote:

I've asked you before about how you treat acetone in GEOS-Chem, but I thought to ask you again since Eric Nielson is implementing the GMI Combo chemistry package into the NASA GEOS-5 Chemistry Climate Model (Steven Pawson/Rich Stolarski = PIs). In the Combo CTM, we simply read in static acetone fields, instead of having acetone emissions. We did this a while back following Harvard's lead. So, I'd like to know if there has been any change to the treatment of acetone that we should adopt?

Daniel Jacob replied:

We still read the same static acetone fields (from Jacob et al. 2002) in the standard full-chemistry version of GEOS-Chem and I recommend that you still do that too in GMI. The reason is that there's a 300 ppt acetone background in the atmosphere that we don't understand and can't model. In my 2002 paper I tried to simulate it with an ocean source but subsequent work showed that the ocean is more often a sink (so much for that). At least the 2002 paper has a good simulation of observed concentrations, even if it's for the wrong reason.

--Bob Y. 09:09, 2 December 2008 (EST)

Updates to chemistry mechanism

NOTE: These updated reaction rates were released in GEOS-Chem v8-02-01 in May 2009.

Rob Pinder wrote:

We downloaded the code from the latest version of GEOS-Chem (v8-01-01), but the documentation that we found is from version 5-07-08. Do you know if these rate constants have been overhauled recently, especially to reflect guidance from JPL 2006 ?

Jingqiu Mao replied:

I updated most of the reactions according to JPL06 and IUPAC06. Please find a document with the revised reaction rates.

--Bob Y. 16:52, 16 February 2010 (EST)

HO2 uptake

Original implementation

Helen MacIntyre wrote:

I noticed that in the new version of the model, HO2 uptake has been switched off to 'give better results over the tropics'. I would be very grateful if you could give me some more information regarding this issue, or point me in the direction of some.

Daniel Jacob replied:

Here's the history of HO2 uptake in GC:
  1. Old versions of the model had an HO2 reactive uptake coefficient of 0.1 as per Jacob (AE 2000)
  2. Then results started coming out of Jon Abbatt's lab showing that HO2 uptake was much slower than that. At the same time, Randall Martin and Bastien told me that they were getting better results with HO2 uptake shut off. So that's what we did.
  3. Most recently, Lyatt Jaegle worked with Joel Thornton on inclusion of HO2 uptake in GC by the aqueous-phase recombination mechanism. That paper is now published in JGR (Thornton et al., 2008) and Lyatt was going to give us the code for including it in GC.

Bastien Sauvage replied:

I confirm the history version concerning the reason why we turned off the HO2 uptake, as decribed in Sauvage et al. ACP, 2007. It was based on simulations over the Tropics and comparison of GC ozone simulation versus observations (SHADOZ and MOZAIC).

NOTE: This update from Lyatt Jaegle was released in GEOS-Chem v8-02-01.

--Bob Y. 16:51, 16 February 2010 (EST)

Improved HO2 uptake

This update was tested in the 1-month benchmark simulation v9-02j and approved on 28 May 2013.

Daniel Jacob replied:

I have conferred with Jingqiu and Mat on this latest update and we agreed that using gamma_HO2 = 1 in the standard model is too high since it is given in Mao et al. 2013 as an upper limit (and obviously is). We will use instead gamma_HO2 = 0.2 giving H2O, which can be citable as Jacob [2000] (for the gamma) and Mao et al. [2013] (for the yield).

--Melissa Payer 11:08, 22 May 2013 (EDT)

One thing I would like to add is, the effective gamma_HO2 could be higher than 1, if one considers the uptake of H2O2 induced by HO2 uptake (shown in our paper). This remains largely uncertain.

--Jmao 09:54, 8 November 2013 (EST)

gamma = 0.2 (instead of previous 0.07) is consistent with our independent analysis on Cu and Fe mass fractions in Chinese aerosols (e.g., Cu fraction in China is a factor of 1.6-12 higher than the U.S. mean; shown in our 2012 ACP paper). The number is also in line with the campaign measurements over China (0.13-0.34 over Mt. Tai and 0.09-0.40 over Mt. Mang; shown in Taketani et al. 2012 paper).

--Jintai Lin 0:05, 23 Decemeber 2013 (EST)

Methyl chloroform lifetime

This content has now been moved to the separate Methyl chloroform lifetime wiki page.

--Bob Y. 15:30, 8 June 2011 (EDT)

Previous issues that are now resolved

Most of the issues described below pertained to the SMVGEAR chemical solver (which was replaced by FlexChem in v11-01) and/or the FAST-J photolysis mechanism (which was replaced by FAST-JX in v10-01).

NIT should be converted to molec/cm3 in calcrate.F


SMVGEAR was removed from GEOS-Chem v11-01 and higher versions. The code in calcrate.F will be replaced by the KPP master equation file.

In calcrate.F, we have:

                    ! Nitrate effect; reduce the gamma on nitrate by a
                    ! factor of 10 (lzh, 10/25/2011)
                    IF ( N == 8 ) THEN
                       TMP1 = State_Chm%Tracers(IX,IY,IZ,IDTSO4) +
    &                         State_Chm%Tracers(IX,IY,IZ,IDTNIT)
                       TMP2 = State_Chm%tracers(IX,IY,IZ,IDTNIT)
                       IF ( TMP1 .GT. 0.0 ) THEN
                          XSTKCF = XSTKCF * ( 1.0e+0_fp - 0.9e+0_fp
    &                            *TMP2/TMP1 )

Here NIT is added to SO4 but NIT is in different units than SO4. This unit difference can be traced to the definition of IDTRMB, which is only nonzero for species that are in the SMVGEAR mechanism. Since NIT is not a SMVGEAR species, IDTRMB = 0 for NIT and it is therefore skipped in the unit conversion from kg --> molec/cm3 in partition.F.

This issue was discovered during the implementation of FlexChem. In GEOS-Chem v11-01g and later versions, units of NIT are properly accounted for in routine HETN2O5 (found in gckpp_HetRates.F90).

--Melissa Sulprizio (talk) 20:25, 12 September 2016 (UTC)
--Bob Yantosca (talk) 20:27, 31 January 2017 (UTC)

rate of HNO4


SMVGEAR was removed from GEOS-Chem v11-01 and higher versions. The globchem.dat file is now replaced by the KPP master equation file.

Ellie Browne found a typo in the globchem.dat (GEOS-Chem v8-02-01 and beyond)

A   73 9.52E-05  3.2E+00 -10900 1 P   0.60     0.     0.         
       1.38E+15  1.4E+00 -10900 0     0.00     0.     0.         
      HNO4          +                         M                                
=1.000HO2           +1.000NO2           +                   +

This should be corrected as:

A   73 9.52E-05  3.4E+00 -10900 1 P   0.60     0.     0.         
       1.38E+15  1.1E+00 -10900 0     0.00     0.     0.         
      HNO4          +                         M                                
=1.000HO2           +1.000NO2           +                   + 

The difference is within 2%.

--J Mao. 19:04, 30 Aug 2010 (EDT)
--Bob Yantosca (talk) 20:29, 31 January 2017 (UTC)

near-IR photolysis of HNO4

This update was added to GEOS-Chem v8-02-04.


SMVGEAR was removed from GEOS-Chem v11-01 and higher versions. The globchem.dat file is now replaced by the KPP master equation file. Also, FAST-JX has now replaced FAST-J photolysis.

1. Since FastJX already takes this into account with cross section data at 574nm, we do not need to redo this in calcrate.f. We can therefore comment out this entire IF block:

        ! Prior to 10/27/09:
        ! FastJX has taken near-IR photolysis into account with
        ! cross section at 574nm, so we don't need to add 1e-5 anymore.
        ! According to Jimenez et al., "Quantum yields of OH, HO2 and
        ! NO3 in the UV photolysis of HO2NO2", PCCP, 2005, we also
        ! changed the branch ratio from 0.67(HO2)/0.33(OH) to 0.95/0.05
        ! This will put most weight of near-IR photolysis on HO2 channel.
        ! (jmao, bmy, 10/27/09)
        !! HARDWIRE addition of 1e-5 s-1 photolysis rate to 
        !! HNO4 -> HO2+NO2 to account for HNO4 photolysis in near-IR -- 
        !! see Roehl et al. 'Photodissociation of peroxynitric acid in 
        !! the near-IR', 2002. (amf, bmy, 1/7/02)
        !! Add NCS index to NKHNO4 for SMVGEAR II (gcc, bmy, 4/1/03)
        !IF ( NKHNO4(NCS) > 0 ) THEN
        !   ! Put J(HNO4) in correct spot for SMVGEAR II
        !   NKN     = NKNPHOTRT(PHOTVAL,NCS)
        !   DO KLOOP=1,KTLOOP
        !      RRATE(KLOOP,NKN)=RRATE(KLOOP,NKN) + 1d-5
        !   ENDDO

2. We need to change the branch ratio of HNO4 photolysis in ratj.d. Change these lines from:

13 HNO4       PHOTON     OH         NO3                  0.00E+00  0.00     33.3  HO2NO2 
14 HNO4       PHOTON     HO2        NO2                  0.00E+00  0.00     66.7  HO2NO2 


13 HNO4       PHOTON     OH         NO3                  0.00E+00  0.00      5.0  HO2NO2 
14 HNO4       PHOTON     HO2        NO2                  0.00E+00  0.00     95.0  HO2NO2

This is based on Jimenez et al. (Quantum yields of OH, HO2 and NO3 in the UV photolysis of HO2NO2, PHYSICAL CHEMISTRY CHEMICAL PHYSICS, 2005) shows that HO2 yield should be 0.95 and OH yield should be 0.05 for wavelength above 290nm.

This way all the near-IR photolysis will have most weight on HO2 channel(Stark et al., Overtone dissociation of peroxynitric acid (HO2NO2): Absorption cross sections and photolysis products, JOURNAL OF PHYSICAL CHEMISTRY A, 2008).

This update has now been added to the chemistry mechanism documentation file.

--J Mao. 11:00, 26 Oct 2009 (EDT)
--Bob Y. 16:08, 4 November 2011 (EDT)

yield of isoprene nitrates

This update was added to GEOS-Chem v8-03-02 as a post-release patch, and standardized in GEOS-Chem v9-01-01.


SMVGEAR was removed from GEOS-Chem v11-01 and higher versions. The globchem.dat file is now replaced by the KPP master equation file.

Fabien Paulot found a problem in current chemistry scheme. In GEOS-Chem v8-02-01 and beyond, isoprene nitrates are produced twice: one through channel A and one through 10% loss in channel B. This makes the loss of NOx larger than it should be (18.7% vs. 10%) and also reduces the yield of MVK/MACR/CH2O by about 13%.

A  453 2.70E-12  0.0E+00    350 1 B   0.00     0.     0.         
       5.00E+00  0.0E+00      0 0     0.00     0.     0.         
      RIO2          +     NO                                              
=0.900NO2           +0.900HO2           +0.340IALD          +0.340MVK     
+0.220MACR          +0.560CH2O          +                   +  
A  453 2.70E-12  0.0E+00    350 1 A   0.00     0.     0.         
       5.00E+00  0.0E+00      0 0     0.00     0.     0.         
      RIO2          +     NO                                              
=1.000HNO3          +                   +                   +             

So it should be corrected as (no channel A):

A  453 2.70E-12  0.0E+00    350 0 0   0.00     0.     0.         
      RIO2          +     NO                                              
=0.900NO2           +0.900HO2           +0.340IALD          +0.340MVK     
+0.220MACR          +0.560CH2O          +                   +       

D  453 2.70E-12  0.0E+00    350 1 A   0.00     0.     0.         
       5.00E+00  0.0E+00      0 0     0.00     0.     0.         
      RIO2          +     NO                                              
=1.000HNO3          +                   +                   +  

--J Mao. 18:04, 30 Aug 2010 (EDT)
--Bob Yantosca (talk) 20:31, 31 January 2017 (UTC)

Potential issue with reading restart.cspec file

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


The binary-punch format restart.cspec.YYYYMMDDhh file is slated to be replaced by a netCDF-format restart file, starting in GEOS-Chem v11-01 and higher versions. But during a transition period, you can still request binary-punch format output.

Jingqiu Mao discovered a mis-indexing problem when using the restart.cspec.YYYYMMDDhh file. Please see this wiki post for more information.

--Bob Y. 16:02, 4 November 2011 (EDT)
--Bob Yantosca (talk) 20:33, 31 January 2017 (UTC)

GLCO3, GLPAN bug in standard mechanism

This update was tested in the 1-month benchmark simulation v9-01-03a and approved on 08 Dec 2011.


SMVGEAR was removed from GEOS-Chem v11-01 and higher versions. The globchem.dat file is now replaced by the KPP master equation file.

Fabien Paulot wrote:

I think there is a relatively serious bug in the standard chemistry. GLPAN and GLCO3 are set to inactive but their production and loss reactions are active. As a result they never reach equilibrium and this results in an artificial loss of NOx.
If this is the only cause of the imbalance between sources and sinks of NOx in my simulations, this would account for ~5% of NOy losses. I don't see that problem in a simulation with a different chemistry that among other changes does not feature those reactions. So hopefully that's it.
To fix the error, I made the following modifications in globchem.dat:
  1. I set GLPAN and GLCO3 rxns from active to dead. These rxns were causing an artificial loss of NOx.
  2. I have physically removed GLCO3, GLP, GLPAN, GPAN, ISNO3, MNO3, O2CH2OH, MVN2 and their associated reactions.
  3. I have made GLYX active. I'm not sure why it's not active by default.
and to ratj.d:
  1. I deleted photolysis reactions for MNO3 and GLP, since these species have also now been deleted in globchem.dat

--Bob Y. 14:51, 10 November 2011 (EST)
--Melissa Payer 10:49, 15 December 2011 (EST)
--Bob Yantosca (talk) 20:35, 31 January 2017 (UTC)

Bug in routine ARSL1K

This update was tested in the 1-month benchmark simulation v9-01-03m and approved on 06 Jun 2012.


SMVGEAR was removed from GEOS-Chem v11-01 and higher versions. The ARSL1K routine was replaced by an equivalent function in KPP's rate law library.

A bug in routine ARSL1K became problematic in the implementation of Justin Parrella's tropospheric bromine chemistry mechanism for GEOS-Chem v9-01-03. In the bromine chemistry mechanism, a sticking coefficient of 0.0 is passed to the routine ARSL1K for non-sulfate, non-sea salt aerosol. The IF statement modified in GEOS-Chem v8-02-04 resulted in the reaction rate being set to the default value of 1.0d-3. A 1-month benchmark for July 2005 indicated that the simulated BrO was a little more than twice the expected zonal mean. Modifying the default value from 1.0d-3 to 1.0d-30 resulted in reasonable simulated BrO values.

Mat Evans wrote:

I've re-run two 2 month simulation [using GEOS-Chem v9-01-02]. One with the error handling value of 1e-3 (standard) and one with it being 1e-30. There are 5127 time and space points where the model traps the problem and invokes the 1e-3 or 1e-30 value. There are 30*24*2*37*72*46 (roughly 200 million) time and space points when the error could have occurred so we are looking at a relatively infrequent event.
The simulations show virtually no difference between the two simulations.
mean and stddev ratio of all grid boxes with and without the fix are shown below
    NOx     0.999996  0.000409291
    Ox      1.00000   1.27233e-05
    O3      1.00000   1.52284e-05
    PAN     0.997849  0.0111997
    CO      1.00000   4.21768e-06
    ALK4    0.990514  0.0351941
    ISOP    0.999979  0.0108033
    H2O2    0.992067  0.0264659
    DST1    1.00000   0.00000
    HO2     0.999996  0.00309464
    OH      1.00003   0.00767954
So although there are some differences they are very minor. For completeness we should put this in as a bug fix (make the error value 1d-30 rather than 1d-3). But it is not a major problem.

--Melissa Payer 17:52, 14 May 2012 (EDT)
--Bob Yantosca (talk) 20:35, 31 January 2017 (UTC)

NOTE: Please also see our list of GEOS-Chem bugs and the versions in which they were fixed for the latest information.

Variable Tropopause

Please see the following links for more information about issues with the dynamic tropopause that have since been fixed.

  1. Buggy implementation in v7-04-12
  2. Polar cap @ 200 hPa is now applied

Other errors in SMVGEAR

  1. May Fu and Philippe Le Sager fixed a bug in SMVGEAR that caused concentrations of certain tracers in STT to go to zero. This bug was fixed in GEOS-Chem v7-04-13.
  2. Lok Lamsal reported a bug with NaN's in SMVGEAR. Bob Yantosca recommended a fix for this error. Visit the discussion here.

--Bob Y. 12:52, 13 September 2010 (EDT)