Coupled Climate-Carbon Cycle Experiments in CSM1
Introduction:A suite of coupled climate-carbon
cycle experiments have been performed in CSM1 under the auspices of
Biogeochemisty Working Group. The control experiment is documented in
Doney et al. (submitted version) (reference), and the transient experiments are documented
in Fung et al. (submitted version)
(reference). The source code used to perform these
experiments is available below, as is brief documentation on how to run
the experiments. The output from these experiments will be made available
soon. When the output is available, this page will be modified to reflect
The results were also described in a poster presented at the 7th
International Carbon Dioxide Conference. The poster is available for
download through this link.
Support Policy:Though this code, the supporting scripts and
datasets, and the model output is being made available, this is not an
officially released product of CCSM. This is not a supported product of CCSM.
CCSM does not have the resources to provide support for this product.
Experiments:The experiments that have been performed are
depicted in the following figure, along with arrows depicting the various
initial condition sets that are provided below. The first column indicates
what value of CO2 is used in the radiation code. The options are
to use the column average of the transported CO2 tracer, termed
TRACER, or to use a specified constant. How to choose between these is
described below. The second column indicates what value of CO2 is
used in the photosynthesis calculations of the land model. The options are to
use the spatially varying CO2 passed from the atmosphere to the
land, termed BGC, or to use 280 ppmv. Using BGC essentially allows unlimited
CO2 fertilization on the land. How to choose between these is
Downloads:The code used to perform these experiments, as well
as the scripts, forcing datasets, and initial condition (IC) files are
available through the following links:
Running the model:This model is essentially an extension of
CSM1.4, so the documentation at http://www.cesm.ucar.edu/models/ccsm1.4/,
particularly the User's Guide, is relevant for the script and code structure.
Because of this, only differences from the CSM1.4 distribution and options
for configuring carbon cycle parameterizations are described here.
Differences from the CSM1.4 distribution:
- The downloads above should be used instead of the downloads from the
CSM1.4 download page.
- The scripts have been modified to enable the model to run on an IBM SP
cluster running AIX.
- The CSM1 + carbon cycle distribution is only set up for fully coupled
runs with atmospheric resolution of T31 and ocean resolution of x3 (with
enhanced meridional resolution in the tropics). No runs have been performed
with higher resolution and initial condition files are not available.
- Only the source code for the Eulerian dynamical core is provided. The
carbon cycle parameterization has not been tested with other dynamical
- The units of non-water atmospheric constituents are now (kg)/(kg dry
air), as opposed to (kg)/(kg moist air). Physics and dynamics source code
has been considerably modified in order to make the mixing and transport of
non-water constituents consistent with these units.
- The source code tar file contains two source trees, models and
models_base. The models tree contains the source code with the carbon cycle
parameterization, while the models_base tree contains the source code
without the carbon cycle parameterization. Both trees are included to allow
users to better see what changes were made to the model to introduce the
- A new environment variable, CSMROOT, has been introduced that should be
set in the local copy of the scripts to indicate where the source code,
input, and initial condition tar files have been unpackaged.
- To use one of the above initial condition sets, set IC_ORIGIN in
the run script to predefined. Then set IC_PREDEFINED_OPT to one of the
options listed in the run script, which correspond directly to the above
initial condition sets. Note that selecting an initial condition set for
1820 or 2000 does not automatically turn on fossil fuel emissions. See
point 3. from Configuring the Carbon Cycle
on how to do this.
Configuring the Carbon Cycle:
- To select TRACER for radiative CO2, set radco2opt in the
atmospheric model's namelist ccmexp to 'TRACER'. If radco2opt is not
'TRACER', then co2vmr is used for radiative CO2. The default
value of radco2opt is 'TRACER'.
- To select BGC for LAND CO2, set LPCO2 in the land model's
namelist casaexp to 1. If LPCO2 is 0, then LAND CO2 will be 280
- To include fossil fuel emissions,
set ff_emis_on to .true. in the atmospheric setup script. The variables
necessary to select an emissions dataset are described in the script. Note
that the emissions datasets provided do not include emissions that are
due to land use change.
Running at NCAR:The source code, input and initial condition tar
files have been unpackaged on NCAR SCD systems in the directory
"/ccsm/bio/csm1_bgc". So to run at NCAR, simply download the scripts tar file
to a convenient location, such as your home directory, and proceed as you
would above. The scripts in the tar file set CSMROOT to "/ccsm/bio/csm1_bgc"
Model Output Notes:
- TR02 is the net atmospheric CO2.
- TR03 is atmospheric CO2 that responds only to ATM-OCN
- TR04 is atmospheric CO2 that responds only to ATM-LND
- TR05 is atmospheric O2 that responds only to ATM-OCN
- TR06, for runs with fossil fuel emissions, is atmospheric
CO2 that responds only to fossil fuel emissions.
- Atmospheric CO2 is in units of (kg C)/(kg dry air). To
convert these units to ppmv, multiply by 1e6 * 29 / 12, where 29 and 12 are
the model molecular weights of dry air and C respectively.
Code Portability:The code made available below has only been
run on two platforms, an SGI Origin 3000 running IRIX, and an IBM SP cluster
running AIX. The majority of the experiments that were done with this code
were performed on an SGI Origin 3000. Subsequently, the code was modified to
enable it to run on an IBM SP cluster. There are two primary factors that
would inhibit the porting of the code to different platforms:
- The individual components of the coupled model, (atmosphere, land,
sea-ice, ocean, and flux coupler), have only been run as shared memory
executables. This implies that each component needs to reside on an
individual shared memory node of the computer being used. If the nodes have
few CPUs per node, then it will not be possible to achieve significant
throughput. As a point of reference, experiments that have been performed
so far on the IBM SP cluster running AIX at NCAR have used a single 32 CPU
node. The atmospheric component does have the capability to run in a
distributed memory environment, but this option has not been tested with
the modifications that were made with the introduction of the carbon cycle.
These modifications included changes to the constituent transport that
could have implications in a shared memory environment.
- It is likely that the code has non-portable constructs in it.
References: Doney, S.C., K. Lindsay,
I. Fung, and J. John, 2005: Natural Variability in a Stable, 1000 Year
Global Coupled Climate-Carbon Cycle Simulation, J. Climate, submitted.
 Fung, I., S.C. Doney, K. Lindsay, and
J. John, 2005: Evolution of Carbon Sinks in a Changing Climate,
Proc. Nat. Acad. Sci. (USA), submitted.