Initial conditions for GEOS-Chem

From Geos-chem
Jump to: navigation, search

On this page we provide some information about initial conditions and restart files for GEOS-Chem.

Restart files for use with the GEOS-Chem Unit Tester

In GEOS-Chem v11-01 and higher versions, we now keep a folder of GEOS-Chem restart files for use with the GEOS-Chem Unit Tester. These restart files will also be used when you create GEOS-Chem run directories with the Unit Tester.

You can download the files from:

wget ""

--Bob Yantosca (talk) 21:31, 22 December 2016 (UTC)

Restart files from benchmark simulations


NOTE: This information is now out of date. GEOS-Chem v11-01 and higher versions read initial conditions from restart files in netCDF format.

We maintain a repository of restart files from the 1-month and 1-year simulations that are used to benchmark the "full-chemistry" simulation.

We do not maintain restart files for the various offline simulations in GEOS-Chem. To obtain restart files for these simulations you should directly contact the person or persons who are using these simulations. If you do not know whom to contact, then please ask your Working Group chairperson.

NOTE: You should always spin up your simulation for a sufficient period of time (such as 1-year) to remove the signature of the restart file from your output.

1-month benchmarks

For v9-01-01 and later

The restart files generated from the 1-month benchmark simulations are all contained within the vXX-YY-ZZ*.tar.gz files that are available from the following directory:

1-month Benchmark Location of restart files
All versions
cd gcgrid/geos-chem/1mo_benchmarks/vXX-YY-ZZ/

For v8-03-01 and earlier

The restart files generated from the 1-month benchmark simulations are all contained within the GEOS-Chem-vXX-YY-ZZ*.stdrun.tar.gz files that are available from the following directories:

1-month Benchmark Location of restart files
All "Beta" releases
cd gcgrid/geos-chem/beta_releases/
All "Public" releases
cd gcgrid/geos-chem/public_releases/

1-year benchmarks

The restart files generated from the 1-year benchmark simulations may be obtained via FTP download. The table below shows the data download locations of the most recent benchmarks. Please click on the links below for a full description of each 1-year benchmark simulation.

Version Location of restart files

GEOS-Chem v9-02
GEOS-Chem v9-01-03
GEOS-Chem v9-01-02
GEOS-Chem v9-01-01
GEOS-Chem v8-02-04
GEOS-Chem v8-02-01
GEOS-Chem v8-01-04
GEOS-Chem v8-01-01
GEOS-Chem v7-04-13

--Melissa Sulprizio 13:01, 17 January 2014 (EST)

Known issues

Problem caused by zero concentrations

NOTE: The file partition.f was removed from GEOS-Chem v11-01 along with other modules from SMVGEAR. But the advice about starting simulations with a well spun-up restart file remains valid.

Eric Sofen wrote:

I just switched over to v8-01-01 from v7-03-06, and have found that it will run fine using a restart file with reasonable "modern" concentrations (e.g. the restart file that is in the run directory when I download the code), but an error occurs when I try to run using a restart file with all of the concentrations set to zero, which is what we want to use at the start of a spin-up run for a pre-industrial simulation.
The error output to the log file is as follows:
   ---> DATE: 1989/02/01  GMT: 04:00  X-HRS:       4.000
   STOP at partition.f
       - CLEANUP: deallocating arrays now...
   ERROR running GEOS-CHEM!
This error code refers to line 300 in partition.f
I have found that if the restart file has concentrations of exactly zero, the error occurs during the first timestep, and if I use small non-zero concentrations (1e-20, 1e-15), the error still occurs, but after several timesteps.
I am not entirely clear on why this error is occurring and am wondering whether we can change something to make it possible to run with a restart file with all of the concentrations set to zero, as we were able to in version 7-03-06.

Bob Yantosca wrote:

In general you can't start a full-chemistry run from a restart file with zero concentrations. bThe nature of the SMVGEAR solver (and the very non-linear chemistry mechanism) is such that you can get funny problems when concentrations get too low. Also, very small concentrations are not good to have because they can cause logarithms to blow up, which can cause floating-point NaN values, which then can propagate through your code and cause other problems elsewhere. Small numbers in the aerosol thermodynamic routine RPMARES were problematic to us recently and it took us a while to implement a fix for that so that we don't do a division that will result in a NaN or other abnormal value.
Please see the following wiki pages for more information:
What I would do is to create a restart file with "reasonable" values for a preindustrial atmosphere (cf. the literature) rather than one that is zero everywhere. Theoretically, it shouldn't really matter what value you start from as long as you spin up the model for long enough. The spinup will remove the initial value. I think if you feed the model preindustrial emissions and spin it up for a couple of years then you will eventually get to the right preindustrial atmosphere.

--Bob Y. 13:52, 1 July 2008 (EDT)