The land ice verification and validation toolkit
LIVVkit is a python-based toolkit for verification and validation of ice sheet models. It aims to provide the following capabilities:
- Numerical verification -- "Are we solving the equations correctly?"
- Physical validation -- "Are we using the right physics?"
- Code verification -- "did we build what we intended?"
- Performance validation -- "did we build what the users wanted?"
Within LIVVkit, these capabilities are broken into four components:
Currently, LIVVkit is being used and developed in conjunction with E3SM (Energy Exascale Earth System Model) and CISM (Community Ice Sheet Model), but is designed to be extensible to other models. For further documentation view the full documentation.
Users and contributors are welcome! We’ll help you out – open an issue on github to contact us for any reason.
The latest LIVVkit release can be installed via pip:
pip install livvkit
Additionally, LIVVkit is released on github, and you can clone the source code:
git clone https://github.com/LIVVkit/LIVVkit.git
If you are having any troubles with installation or dependencies, open an issue on the issue tracker or contact us!
LIVVkit is primarily controlled via options specified at the command line. To see the full list of options, run:
In verification mode, LIVVkit analyzes and compares a regression testing dataset to a reference dataset. For example, LIVVkit may analyze the dataset produced from a proposed CISM 2.0.6 release (~400MB; download here) and compare it to the dataset produced from the CISM 2.0.0 release (~400MB; download here). To run this example, first download the two aforementioned datasets to a directory, open a terminal, and navigate to your download directory. Then, un-tar the datasets:
tar -zxvf cism-2.0.0-tests.20160728.tgz tar -zxvf cism-2.0.6-tests.20160728.tgz
For ease, export the path to the two dataset directories:
export REF=$PWD/cism-2.0.0-tests/titan-gnu/CISM_glissade export TEST=$PWD/cism-2.0.6-tests/titan-gnu/CISM_glissade
To run the suite, use:
livv -v $TEST $REF -o cism206v200 -s
LIVVkit will run the verification suite, report a summary of the results
on the command line, produce an output website in the created
directory specified by the
-o/--out-dir option, and launch an http server
-s/--serve option) to easily view the output in your favorite web
browser. LIVVkit will tell you the address to view the website at on the
command line, which will typically look like
LIVVkit is extensible to more in-depth or larger validation analyses.
However, because these validation analyses are particularly data intensive,
many of the observational and example model output files are much too
large to distribute in the LIVVkit package. Therefore, we've developed a
LIVVkit Extensions repository (LEX) which uses
git-lfs (Git Large File Support) in order to
distribute the required data.
git-lfs can be installed either before or
after cloning this repository, but it will be needed before downloading
the required data. You can determine if you have
git-lfs installed on
your system by running this command:
command -v git-lfs
git-lfs is not installed, you can install it by following the instructions here:
git-lfs is installed, clone and enter this repository:
git lfs clone https://code.ornl.gov/LIVVkit/lex.git cd lex
Each extension will have an associated JSON configuration file which will describe the extension's analysis code, data locations, and options. To see a list of available extensions, you can run this command:
find . -iname "*.json"
To execute any of these extensions, point
to any of these extensions config file via the
-e/--extension option (or the
-V/--validate option). For example, to run the minimal example extension,
place the output website in the
val_test directory, and serve the output website
you'd run this command:
livv -e example/example.json -o vv_test -s
Note: All the extension configurations files assume you are working from the
lex directory. You can run any of these extensions from any
directory, but you will need to edit the paths in the JSON configuration files so
livv can find the required files.
Likewise, you can also apply these analyses to any new model run by adjusting the paths to point to your model run.
For more information about using LIVVkit see the documentation.
If you would like to suggest features, request tests, discuss contributions, report bugs, ask questions, or contact us for any reason, use the Issue Tracker.
Want to send us a private message?
Joseph H. Kennedy
- github: @jhkennedy
- email: kennedyjh [at] ornl.gov
Katherine J. Evans
- github: @kevans32
- email: evanskj [at] ornl.gov
If you're emailing us, we recommend CC-ing all of us.