CFMIP Standard Output

Prepared by

Mark Webb and Karl E. Taylor

Based, originally, on AMIP Standard Output Tables developed by

PCMDI and the AMIP panel

A pdf version of this page can be downloaded from here.

26 September 2005

Change history:

Added request for 3D fields to be on model levels rather than pressure levels and a more explicit distinction between stratiform, convective and total 3D cloud fields. Also added some extra diagnostics requested daily. KDW 26/09/05

Changed headings on monthly table to state preference for monthly means over climatological fields, and to allow for annual mean climatological CO2 forcing fields. MJW 29/11/04

· Background and summary of format requirements

· Tables of output fields

Monthly data and time-independent fields

Table CF1a: 2-d atmosphere or surface data (longitude, latitude, time)

Table CF1b: ISCCP simulator data (longitude, latitude, pressure2, tau, time)

Table CF1c: 3-d atmosphere data on pressure levels (longitude, latitude, pressure, time)

Table CF1d: 3-d data on model levels (longitude, latitude, model_level, time)

Table CF1e: time-independent 2-d surface data (longitude, latitude)

Annual mean data

Table CF2a: 2-d atmosphere or surface data (longitude, latitude, time)

Table CF2b: ISCCP simulator data (longitude, latitude, pressure2, tau, time)

Daily mean data

Table CF3a: 2-d atmosphere data (longitude, latitude, time)

Table CF3b: ISCCP simulator data (longitude, latitude, pressure2, tau, time)

Table CF3c: 3-d atmosphere data on pressure levels (longitude, latitude, pressure, time)

Table CF3d: 3-d atmosphere data on model levels (longitude, latitude, pressure, time)

Doubled CO2 radiative forcing

Table CF4: Monthly-mean 2-d radiative forcing data for doubled co2 (longitude, latitude, time).

Background and overview of format requirements

In order to facilitate participation in various community modeling experiments (e.g., IPCC, CMIP, AMIP, CFMIP), the Program for Climate Model Diagnosis and Intercomparison (with encouragement from the WGCM and WGNE and in collaboration with project leaders) is attempting to establish a more uniform set of model output requirements. CFMIP was initiated before this effort had matured sufficiently, so initial CFMIP output has been collected in various structures and formats, but to facilitate analysis of future model output, any new CFMIP data should conform to the new requirements.

Modeling groups contributing output to the CFMIP database are henceforth asked to ensure that it meets rather strict format and metadata requirements. These requirements yield files produced in network Common Data Form (netCDF; see http://www.unidata.ucar.edu/packages/netcdf), which has become the most popular form for exchanging ocean-atmosphere model output. The files are "self-describing" and the metadata contained in the files conform with the NetCDF Climate and Forecast (CF) Metadata Conventions (see http://www.cgd.ucar.edu/cms/eaton/cf-metadata). The CF conventions for netCDF data generalize and extend the Cooperative Ocean / Atmosphere Research Data Service (COARDS) conventions developed in the 1990s. Note that the CF convention establishes standard names for climate and weather variables, which identify the physical quantity. These standard names are given in the tables below. Note that more than one field can be associated with the same standard name because different fields sampled in different ways (e.g., surface air temperature vs. upper air temperature) refer to the same physical quantity. Nevertheless, one can uniquely identify each stored field by considering additional metadata stored in the file (e.g., dimension information). Extended definitions of CF standard names, which basically answer the question "What do you mean precisely by this quantity?", may be found on the Web at http://www.cgd.ucar.edu/cms/eaton/cf-metadata/standard_name.html.

The standards for CFMIP contributions closely follow those established for IPCC/CMIP, which are given in the document http://www-pcmdi.llnl.gov/ipcc/IPCC_output_requirements.htm. This document should be read carefully before preparing contributions. Perhaps the easiest way to meet these requirements is to rewrite your model output through CMOR, a software library available from PCMDI and further described in the next paragraph. Put briefly, the requirements specify required metadata and file organization, along with acceptable coordinate systems.. Units and sign conventions of the data must conform to the tables below. Latitude-longitude grids must be rectilinear, i.e., have a unique set of longitudes that applies to all latitudes. Data on non-rectilinear grids must be interpolated to rectilinear grids before transmission to the PCMDI. With the exception of certain cloud-related fields (which as noted below should be provided on model levels), three-dimensional atmosphere variables must be interpolated to standard pressure levels given below. Each output file should contain only a single output variable, though there may be many time points per file. This file organization contrasts with the typical model output history files, which contain all variables for a single time step.

To facilitate adherence to these standards, a PCMDI team has written (in FORTRAN 90) a standard output code called CMOR (pronounced "see more"), which is now available as open source. This code structures the data uniformly and writes netCDF files in full compliance with IPCC and CFMIP requirements. Use of CMOR is being encouraged (and in some cases required) by various other ongoing model intercomparison projects. A cmor_readme file (see http://www-pcmdi.llnl.gov/software/cmor/cmor_readme.html) and code documentation in pdf format (see http://www-pcmdi.llnl.gov/software/cmor/cmor_users_guide.pdf) are available. For further information, contact taylor13@llnl.gov.

The IPCC/CMIP requirements as defined by http://www-pcmdi.llnl.gov/ipcc/IPCC_output_requirements.htm should be satisfied by new submissions of CFMIP data except that in that document:

All references to the IPCC Standard Output should refer instead to the CFMIP Standard Output (i.e., this document).
The required global attribute "project_id" should be set to "CFMIP"
As appropriate, the "experiment_id" should be set to one of the following:

'slab ocean control experiment'

'2xCO2 equilibrium experiment'

'+2K perpetual July experiment'

'0K perpetual July experiment'

'-2K perpetual July experiment'

4. The CFMIP files should be stored in the following directory structure: A separate directory should be created containing all the variables in each model/experiment/standard_output_table category (e.g., CSIRO/Slabcntl/A1). This will ensure that filenames within each directory are unique. For uniformity, it is suggested that the trailing lower case letter, found on the tables below, should be dropped (e.g., 'CF1', not CF1a) and it is suggested that the experiments be abbreviated as follows:

Slabcntl (i.e., the slab ocean control experiment)

2xCO2 (i.e., the 2xCO2 equilibrium experiment)

Plus2KJul (i.e., the +2K perpetual July experiment)

CntlJul (i.e., the 0K perpetual July experiment)

Minus2KJul (i.e., the -2K perpetual July experiment)

5. For perpetual July experiments the calendar should be set to 'none', no matter what calendar is assumed by your model. The basetime should be set to "days since 1-7-15". This is the CF-convention method of indicating perpetual July. The time coordinate values will normally record the number of days since the beginning of the experiment. For the time mean, the bounds specify the beginning and end of the interval over which the time mean is formed.

6. For slab ocean monthly mean data, climatological values should be reported (averages of several January's, several February's, etc.). The CF-convention specifies that time bounds information should be stored in an array identified by the "climatology" attribute (http://www.cgd.ucar.edu/cms/eaton/cf-metadata/CF-1.0.html#climatology), and the "cell_methods" attribute (http://www.cgd.ucar.edu/cms/eaton/cf-metadata/CF-1.0.html#cell-methods) provides further information concerning how the climatological values are computed. For CFMIP, the CF conventions should be followed, but the complete set of climatological metadata accommodated by the conventions is not yet produced by CMOR. It is therefore currently acceptable to omit the "climatology" attribute and simply record the cell bounds as if they were monthly means (and CMOR will indicate in the cell_methods attribute that a time-mean is stored, but not that it is a climatological mean). In the future CMOR will be modified to accommodate climatological data.

The notes that appear in the tables below are meant to provide precise definitions of the requested fields. Sometimes it may be impossible to satisfy the requests; in these cases, any deviations from the specifications below should be described in the "history" and/or "comment" attributes associated with the variable.

Model output summary

Various fields, sampled at various frequencies, are needed to carry out various proposed CFMIP analyses (see http://cfmip.metoffice.com/DR.html). The tables below derive from tables (again see http://cfmip.metoffice.com/DR.html) that were originally modified from the AMIP standard output tables (see http://www-pcmdi.llnl.gov/projects/amip/OUTPUT/AMIP2/index.html) to meet the needs of CFMIP. These "original" CFMIP tables have been restructured to be more consistent with IPCC standard output and to be consistent with the CMOR input tables that have been created for CFMIP. The "analysis packages" indicated by letters in the fifth and sixth columns of each table refer to various proposed CFMIP analyses. A few fields are not called for by any of the currently proposed projects, but have been requested as part of the IPCC standard model output, so they are included here because they may eventually be of some interest.

Tables CF1, CF2, and CF3 contain fields reported at monthly, annual, and daily frequencies, respectively. Table CF4 contains doubled CO2 forcing fields. Tables CF1, CF2, and CF3 are further separated below into component tables, depending largely on whether the fields are spatially two or three dimensional and whether the three dimensional fields are stored on pressure levels or model levels. Finally, a separate component table contains ISCCP simulator output.

CFMIP output is requested for different time periods, depending on the reporting frequency and the experiment. In the case of "monthly mean" data:

For slab ocean experiments 12 x 20 months of monthly mean data should be reported after equilibrium is reached.
For perpetual-July experiments 24 months of monthly mean data (or longer if there is high variability) should be reported after equilibrium is reached.
For slab ocean runs, annual mean data should be reported for each year simulated (including spin-up). No annual means are requested from prescribed SST experiments.
For slab ocean experiments, report daily values only for the final 5 years at equilibrium.
For perpetual-July experiments, report daily values for the final 2 years at equilibrium.

As noted below some of the tables parallel the IPCC standard output tables, but have been assigned a different table number. Also a few minor differences between these tables and the original CFMIP tables exist. The current lists are preferable, however submission of data based on old CFMIP diagnostic lists are acceptable.

Tables of output fields