The DYAMOND Model intercomparison(s)#

The initiative DYAMOND (DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains) sets up a framework for the intercomparison of global storm-resolving models, i.e. an emerging class of atmospheric circulation models.

DYAMOND addresses three main questions:

  1. Where do the storm-resolving simulations agree and provide deeper insights into the climate system?

  2. How sensitive are the simulations to a particular implementation?

  3. What are performance and analysis bottle necks associated with global storm-resolving models?

The intercomparison has resulted in several publications. Find more details at the ESiWACE DYAMOND page.

Important

If you work with the data stored at DKRZ, we request that you acknowledge DKRZ and ESiWACE in any publication resulting from this work.

The DYAMOND intercomparison is coordinated by the Max Planck Institute for Meteorology (MPI-M) and the German Climate Computing Center (DKRZ) and supported by the Centre of Excellence in Simulation of Weather and Climate in Europe (ESiWACE).

Contact: In case of questions, please contact the DYAMOND coordination team via: dyamond@esiwace.eu.

Phases#

So far, the following models participated in at least on of the two DYAMOND pahses: ARPEGE NH, GEOS, GRIST, ICON, IFS, MPAS, NICAM, SAM, SCREAM, SHiELD, Unified Model (UM). More information about the models and their providede experiment runs can be found in dedicated DYAMOND phase pages.

Getting the data#

The two data collections, DYAMOND Summer and DYAMOND Winter are archived in DKRZ’s HSM tape archive. These data sets are accessible via the The DYAMOND Data Library, which is situated at DKRZ’s new supercomputer Levante at /fastdata/ka1081/DYAMOND/.

1) Get/check your user account at DKRZ:

To access the DYAMOND data library you will need a user account for the post-processing project “1153”, named “DYAMOND Data Intercomparison” at DKRZ. To do so, please visit https://luv.dkrz.de and use your institution e-mail address. Once you have requested the account, DKRZ will have to process your request. When this is done (usually within a few minutes to hours), you will get a confirmation mail with your new user id (typically something like b123456). After you get a confirming e-mail you are able to login to Levante (e.g. ssh -X <user-id>@levante.dkrz.de). For more information about how to start at DKRZ see: https://docs.dkrz.de/doc/getting_started/index.html.

2) Look at available data files on disk:

While most of the data is archived in the DKRZ tape system, available DYAMOND Summer and DYAMOND Winter files can be found in /fastdata/ka1081/DYAMOND/data/summer_data and /fastdata/ka1081/DYAMOND/data/winter_data, respectively. Currently, most of the DYAMOND Winter collection is available on disk.

3) Request missing files from tape archive:

Since the DYAMOND data sets form hundreds of terabytes of data, only parts are available on disk. For an overview which models and runs are overall availabel, see Participating models and description (Summer) and Participating models and description (Winter).

To access further DYAMOND Summer data files, you will need to make a request within the DYAMOND data library. The ESiWACE team at DKRZ will download the data for you to Levante, thereby making it accessible to all DYAMOND users. A detailed description how to search for and request archived data files is given at The DYAMOND Data Library and the README file at /fastdata/ka1081/DYAMOND/README.

To retrive DYAMOND Winter data files from tape, DKRZ provides the Python tool outtake. It is setup to download the files (with read only permissions) to /fastdata/ka1081/DYAMOND/data/winter_data, thereby making it accessible to all DYAMOND users. Examples how to use it in a Python notebook or with the command-line script find_files can be found at Loading data from the catalog and Getting file names from intake with a command-line python script.

Unused data files will be deleted on a regular basis.

Working with the data#

You are most invited to use our systems for analyzing the data (instead of copying them around). Our post-processing project 1153 has some compute resources that can be used for analysis scripts. Therefore, we recommend to store project data under /work/bb1153/<user-id>/ and use the temporary storage in your scratch folder /scratch/*/<user-id> for processing of large data sets. For more information about the Levantes file systhem its quotas and backups, see https://docs.dkrz.de/doc/levante/file-systems.html#levante-file-systems.

Further information on the runs and models and how to work with it can be found at DYAMOND Summer and DYAMOND Winter.

easy.gems needs your expertise!

We want to make this documentation a community effort. Please fell free to contribute your tips, scripts, information about experiments and other information you consider relevant.

Contributing to easy.gems

Other info#