version 1.0 [19 June 2006]
Timeline, Science
Objectives, Metrics
for Scientific Quality of Simulation, Physics, Dynamics, Software
Engineering,
Getting a Contribution into CAM/CSM
CAM4 time schedules are controlled in part by the overall development of
CCSM4 (which is still under discussion). Here is our best guess at the CAM4
timeline. For convenience, the annual CCSM meeting is identified as occurring
in June of each year, and the annual CAM meeting as occurring in February.
back to contents
Below are suggested science objectives for CAM4.
back to contents
A key goal of the development effort is to quantify simulation biases
objectively by constructing measures (metrics) to compare simulations directly
against observational estimates of the same field. A series of metrics (one or
two numbers) is to be developed that can be run on any simulation to provide a
measure of agreement between the model and real world of a particular field
(e.g. precipitation) or phenomena (e.g. ENSO). All CCSM users are invited to
submit candidate metrics (AMWG
Diagnostics) for discussion by the AMWG and possible inclusion into the
standard package.
Please note that the Metrics page above is used to evaluate the
comprehensive model simulation. Individual component processes (e.g.
convection, or dynamics) can and will use specific procedures to evaluate that
component individually. See for example the tests of dynamical cores discussed
in Dynamics
back to contents
To attain the science goals articulated above we believe that certain issues
merit particular attention. These issues can be catagorized in a number of
ways, but we have decided to do so in terms of Physics, Dynamics and Software
Engineering
back to contents
The term "Physics" loosely refers to all processes in the model
that are not "Dynamics" (see Dynamics). These
include all diabatic terms in the evolution equations, source/sink terms in
equations describing the evolution of atmospheric constituents, and
parameterizations of subgridscale (unresolvable) processes.
Our overall goal for the physical parameterizations employed in CAM is a
better intellectual and theoretical basis for the representation of processes
in the model. Our current focus is on the following list of physical
parameterizations: Boundary Layer,Convection,Cloud Fraction,Sub-Columns,Radiation,Microphysics,Gravity Waves,Aerosols and Chemistry
We have currently identified:
We welcome the active involvement of others. Anyone in the AMWG can be added to the list and included in
development wiki access
Contacts:
U Washington Boundary Layer scheme is a candidate parameterization. Boundary
layer might pace progress on other components. Because of the dependancy of
other components on the boundary layer, this may be our highest priority.
A discussion of the scientific requirements for any boundary layer
parameterization and basics of development progress will be contained in a
separate PBL requirements document
back to contents
Contacts:
A separate convection
development document addresses how candidate convection schemes will be
established. This process will be more involved because of the large number of
people and candidate schemes involved. All who are interested are urged to
contact the internal contacts listed.
back to contents
Contact:
Several options are on the table for adjustments to the cloud fraction
scheme. These include a Modified version of the current Slingo Scheme for
consistency (no empty cloud), and a simplified Tompkins PDF moment scheme for
vapor. Contact Gettelman for more details.
back to contents
Contacts:
A proposal exists to use independent sub-columns (Monte Carlo ICA-McICA) for
moist physics. At least: radiation, microphysics (Gettelman/Neale), probably
(eventually) corelated-k radiation. It may build off of total water pdfs in
tompkins cloud fraction scheme (simplifies closure for microphysics) Consider
acceleration options for ICA if we go this route.
Options include schemse by Pincus, Collins (deterministic), or Raisenen
formulations. A small group is working on this effort to get the infrastructure
into the moist physics. Please contact the internal contacts for more
information.
back to contents
Contact:
Ideally, we will utilize a consistent treatment of radiation and physical
processes in CAM, for example, it would desirable to have the microphysical
formulation for stratiform and convective clouds, as well as the scavenging
formulations, and photolysis rates for any photochemistry use the same cloud
overlap, and scattering assumptions
Alternative RT formulations that have been put forward are:
'Beyond correlated-k'. Link to Requirements web page
The AMWG will run or orchestrate an evaluation of these codes against
benchmarks, and interested parties should contact Collins for more details.
back to contents
Contact:
Currently we are developing a size resolved (2 moment bulk) microphysics
code for implementation into CAM. This will also include a radiation interface
(Collins/Gettelman/Rasch/Mitchell) and an Aerosol nucleation interface
(Gettelman/Ghan). The microphysics development has been proceding with a
coordinated group which has its own web site. The basic approach is outlined in
a Roadmap
and progress can be found on the Microphysics
Swiki
back to contents
Contact: Sassi/Richter
Mountain drag formulations may be developed to modify the gravity wave
scheme currently in the model.
back to contents
Contacts
A first aerosol framework has been developed from the MATCH and CAM2
formulation developed by Phil Rasch, Mary Barth, Bill Collins, and Natalie
Mahowald. This aerosol framework has been undergoing continuous development
over the last 5 years. It is a bulk formulation that predicts only size
segregated mass concentrations for externally mixed aerosols.
There has also been a parallel development track occuring in the MOZART
offline transport model by X-X Tie, J.-F. Lamarque, Peter Hess and others over
the last 5 years. It uses a similar framework, although in differs in several
subtle and important ways.
Each of the two frameworks have advantages and disadvantages compared to the
other. These is an active effort taking place to merge the two efforts, and
connect the resulting formulation to the cloud drop activation parameterization
of Ghan and colleagues, to provide a near term parameterization with the best
attributes of both the MATCH and MOZART frameworks as well as the ability
tocrudely represent the first and second indirect effect. We will refer to this
parameterization as the INTERIM aerosol
formulation.
There are several options ranging from a version of the Ghan droplet scheme
that connects with the current aerosol treatment to detailed 7 mode schemes.
The latter may be coupled to MOZART chemistry and aerosols. The goal is to
characterize direct and indirect effects, or at least create the infrastructure
in the code to do so. We hope to do this in each version of CAM as it evolves.
As we add each piece of new functionality (eg new turbulence parameterization,
new dynamical cores, new subgridscale vertical velocity, new cloud overlap, new
PDF clouds, new McICA) we can then test the direct and indirect effects
The ultimate goal is to link the aerosol framework (existing, interim and
proposed options) to the new cloud Microphysics and
to ensure this treatment is as consistent as possible.
DIAGNOSTICS: We are importing AEROCOM framework
back to contents
Chemistry includes two pieces: WACCM4 and Tropospheric Chemistry
WACCM4
Contact: Gettelman/Sassi/(Garcia)
How high do we go? What do we incorporate for an upper atmosphere in the
operational version of CCSM4? Is it WACCM, or MACCM, or a reduced-lid version
of MACCM? This has huge implications for the computational cost. [Who wants to
test this? There is a paper here]
Tropopsheric Chemistry
Contact: (Lamarque) Carbon/Nitrogen
cycles [this might be a CCSM issue]
back to contents
Contacts
The dynamical core refers to the collection of numerical methods that are
used to represent the solution to the equations of motion (Navier Stokes
equations) and the evolution equations for tracer transport. There are
currently three dynamical cores that can be used in CAM and CCSM: a solution
method based upon a combination of spectral, finite difference, and
semi-lagrangian methods framed using the evolution equations in an Eulerian
framework, called the "Eulerian" core; a solution method using the
same combination of techniques but expressed in a Lagrangian framework, called
the "semi-Lagrangian" method, and a control volume formulation in
which the equations are expressed in flux form called the "finite
volume" methods.
Each of these techniques currently use a rectilinear (approximately uniform
in latitude and longitude) grid. We are currently modifying the CAM to allow
its use with alternative grid distributions (for example the "cubed
sphere" or "icosehedral" grids) that resolve the sphere more
uniformly.
Before doing traditional simulations with a full model it is necessary to
test the underlying dynamics of the model (the dynamical core). A test
procedure has been developed to make sure a core is doing what we think it is
doing. A proposed procedure is described in Dynamics
Evaluations
back to contents
There are a variety of issues that need to be considered in any candidates
for the next generation of CAM.
We have developed a number of documents to assist in understanding the steps
used in integrating any new formulations into the model. These documents
describe:
For information on these issues please see Software
Engineering Issues
back to contents
This section defines the process that is required to allow a contribution to
be considered for the next generation of CAM/CSM. They will be made more
specific after initial discussions with the community.
back to contents