[Go-essp-tech] new expt. to add to CMIP5 and the DRS document
Karl Taylor
taylor13 at llnl.gov
Sat Nov 13 12:33:07 MST 2010
*** nb. Those of you developing software in support of CMIP5 may need
to modify things to include one more CMIP5 experiment. See below:
Dear all,
At the WGCM meeting in Exeter there was agreement that the "historical"
runs should be extended from their official ending date at the end of
2006 to about year 2012. The concentrations and land-use changes
specified in these runs will not be the same for all groups and may not
even be based on true observations (but instead be estimated in various
ways). Since these runs should not be confused with the "historical"
runs where everyone uses the same forcing (based on observations), the
feeling is these short runs (2006-2012) should be labeled as a new
experiment. I've sent a message to the modeling groups describing what
they should do. You might want to read items 6e and 6f in that message
(copied below). To accommodate these new runs, we need to add a line to
the second table of Appendix 1.1 of the DRS document with the following
entries:
7.4 historicalExt historical extension extension of
the historical simulation (experiment 3.2) through year 2012.
I understand this will require additions to much of the software,
possibly including CMOR2 tables, ESG publisher?, ESG node software?,
METAFOR questionnaire, ESG search?, etc.
I apologize for these late changes, but the WGCM thought it important to
include these extensions of the historical runs so that detection
attribution studies could be done.
Please pass on this message to anyone who might need to know this
(beyond those already on the GO-ESSP-tech mail list).
Best regards,
Karl
---------------------------
Dear all,
Here are some items that should be of interest to those of you
participating in CMIP5. Please pass on this information, especially to
those who are responsible for preparing the model output for the CMIP5
archive.
1. For those of you doing the emissions-driven simulations: Unless it's
too late, it is recommended that the recently produced gridded fossil
fuel emissions data from Andres (hosted by IPSL) and the the land use
data from Houghton (hosted at MPI) should be used for the historical
simulations. More information should appear soon on the CMIP5 website.
These fossil fuel emissions data can be retrieved from:
http://dods.ipsl.jussieu.fr/cpipsl/ANDRES/
File name is :
CMIP5_gridcar_CO2_emissions_fossil_fuel_Andres_1751-2007_monthly_SC_mask11.nc
It is a monthly dataset, units are gC/m2/s
Also provided is an ascii file that contains the globally integrated
emissions for every month.
CMIP5_gridcar_CO2_emissions_fossil_fuel_Andres_1751-2007_monthly_SC_grid1x1.txt
The land use file can be found at MPI:
http://www.mpimet.mpg.de/en/wissenschaft/land-im-erdsystem/wechselwirkung-klima-biogeosphaere/landcover-change-emission-data.html
File name is: carbon_emissions_landuse_20person.nc
It is an annual dataset, units are also gC/m2/s
2. Again for those of you with coupled carbon climate models (ESM's):
The CMIP5 expts. 5.4 and 5.5 are designed to isolate the climate change
effects on carbon uptake from the uptake due to CO2 concentration
increases (in the absence of climate change). Originally there were two
options proposed for these experiments: analyze 1%/yr CO2 increase runs
or analyze historical+RCP4.5 runs. At the WGCM meeting last month and in
subsequent discussion, it was decided that for groups who have not yet
performed these experiments, it would be better if they would base these
runs on the idealized 1%/yr CO2 increase (rather than the
historical+RCP4.5 simulations). There will, of course, also be interest
in the historical+RCP4.5 runs, so groups who have already done these
runs, should contribute them to the archive.
3. We recently posted a revised document describing the model output
requirements for CMIP5 (see
http://cmip-pcmdi.llnl.gov/cmip5/output_req.html?submenuheader=2#metadata ).
(Those of you who have chosen *not* to use CMOR2 to rewrite your model
output should study this document carefully; those of you using CMOR may
refer to the CMOR documentation for most of the information you need to
know. TheCMOR documentation has also been recently revised with expanded
descriptions of what you must supply.) At the beginning of the output
requirements document there is a link directing the user to the bottom
of the document where a list of the changes that were made can be
found. Nearly all the changes were simply to improve clarity. Please
note, however, that a new global attribute is now required for most
simulations (parent_experiment_rip), which identifies which ensemble
member the child experiments was spawned from. This information will be
essential for many CMIP5 studies.
4. The so-called "data reference syntax" document has also been
revised. Again, many of the changes should simply make it easier to
understand. Data providers will be especially interested in the
official "short names" of the CMIP5 experiments, since these names are
used in constructing filenames. This document can be obtained through
the following link:
http://cmip-pcmdi.llnl.gov/cmip5/output_req.html?submenuheader=2#req_format
5. The latest version of CMOR2 was just released on 12 November 2010.
The changes made were summarized on an announcement made to those of you
on the cmor email list. It also includes the latest "requested
variables" CMOR tables (also posted at
http://cmip-pcmdi.llnl.gov/cmip5/output_req.html?submenuheader=2#req_list ),
which includes a number of formerly missing "standard names", and also
some corrections. Please download the latest version of CMOR, the CMOR
tables and the documentation before continuing to process your data.
6. At the WGCM meeting there was agreement that it would be useful for
model evaluation and detection/attribution studies to extend the CMIP5
historical runs to near-present (as we have for AMIP), rather than
ending them in 2005. In fact since the CMIP5 project is ongoing, it
would be useful to have simulations extended to at least the end of 2012
using some estimate of recent and future forcing. There is, however, no
community-wide accepted observationally-based concentration/emissions
past 2005. Groups are therefore free to use whatever concentrations,
solar forcing, SO2 emissions etc. they want to use in extending these
runs. It is also o.k. for detection/attribution studies to simply
splice one of the RCP runs to the end of the historical simulations. No
matter what forcing is chosen it is important to consider the following:
a) The groups should take care that there are no substantial
discontinuities in the forcing in passing from the "past" to the
"future", defined to be the end of 2005.
b) It is recommended that if an ensemble of "all-forcings" historical
simulations have been run, then *each* member of the ensemble should be
carried to the end of 2012. Thus, a full ensemble of runs (through
year 2012) would be available for analysis.
c) It is recommended that all historical runs with only a subset of
forcing (e.g., GHG only, natural forcing only) should also be extended
through the year 2012.
d) If one of the RCP forcings is used to extend the historical run, it
may not matter too much which RCP is chosen, and CMIP5 makes no strong
recommendation. If a modeling group has no preference, they might choose
the RCP8.5 expt., as at least one group (the Hadley Centre) has made
this choice already.
e) For all-forcing (anthro + natural) historical runs, the extended
portions of these runs should be treated as a new runs spawned from the
parent historical runs at the end of year 2005. If this run is forced
by an RCP that extends at least to the end of the 21st century, then
nothing special needs to be done. If, however, some other forcing is
used or if the run is an RCP run that is truncated after a few years
(say ending in 2012), then the run should be considered a "historical
extension" experiment with its output placed in a directory named
historicalExt. The "forcing" attribute (a netCDF global attribute)
should describe what forcing is used to extend the run, and this
information will also need to be recorded when entering information
about the run in the METAFOR questionnaire. Placing the extended
portion of the historical runs in a separate place will help guard
against users assuming that these runs are necessarily based on
historically-measured concentrations, land-use changes, solar forcing,
etc. For these historicalExt experiments, the ensemble member
(designated by the "rip" value appearing in the filename and recorded as
netCDF global attributes) will be identical to the historical run it
extends. Also for the runs, the identifying netCDF global attributes
should be defined as follows: experiment="historical extension" and
experiment_id="historicalExt". The "Data Reference Syntax" document and
the CMOR tables will be revised shortly to include these "new" experiments.
f) If one chooses to do historical runs with only a subset of forcing
(e.g., GHG only, natural forcing only, single-forcing experiments,
etc.), then all the data for the complete historical period and in the
extended portion (from 2006-2012) would be kept together, no matter what
forcing was used (in the historicalNat, historicalGHG, or historicalMisc
directories). (In these clearly "unrealistic" cases, naive users will
be less likely to access the output and possibly misuse it.)
7. For your information, We've posted (see
http://cmip-pcmdi.llnl.gov/cmip5/terms.html?submenuheader=3 ) the latest
(and final) version of the the two different terms of use governing the
CMIP5 model output. It looks like about half the groups plan to release
their data for "unrestricted" use and half for "non-commercial
educational and research purposes" only.
Carry on!
Best regards,
your faithful CMIP5 comrades (Karl's lame attempt to try to develop a
little esprit de corps here),
Karl and Ron
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/go-essp-tech/attachments/20101113/fadf1062/attachment.html
More information about the GO-ESSP-TECH
mailing list