|
Testing Information for Contributors
Code Testing by Prospective Contributors
Developers seeking to add code to WRF should consult
the document Information for WRF Code Contributors. This describes the
WRF coding standards, code testing, and code submission procedures.
The required types of testing for new WRF code prior to
submission are: (a) software testing and (b) case testing.
The materials for the software and case testing can be obtained
through this page.
The purpose of the software testing is to check different
compilers and model configurations for bit-for-bit
reproducibility. The software testing is done
through the "WRF Testing Framework" utility (see below).
The purpose of case testing is to confirm code robustness and reasonableness
through the running of short simulations. Robustness is addressed by
running cases for summer and winter.
Reasonableness of results is checked by comparing the new output
of the modified WRF with output in existing plots
of selected simulations (provided below), or by otherwise reviewing
the new output to make sure it is reasonable.
Prospective contributors should do reasonableness testing both for one of
the summer cases and for one of the winter cases.
Software Testing
Modified WRF code must be run through the
WRF Testing Framework (WTF). The WTF is designed to build and test WRF compilation and running, and to analyze test results for versions of
the model with new modifications.
The WTF is described here:
WRF Software
Testing.The WTF materials can be obtained from the WRF users download page at: WRF Testing Framework.
Case Testing
Testing for at least one summer case and one winter case,
in the form of at least a 24-hr forecast for each, is
required. The developer may choose the specific cases
from among the ones listed below. Input files
for these are provided below.
Summer Cases
Listed below are the Metgrid files for
six test cases from Summer 2010. The domain is over the US,
with a grid size of 20 km covering 290x190 points.
The namelist.input template has 40 eta levels.
For the testing, developers may generate 24-hr forecasts using any of the
initialization times. To assist the reasonableness review,
provided below are
verification plots to compare against one of the summer cases run.
Domain map
A map of the domain for the test cases is here: domain map.
*This case (2010061100) is also provided as a case to check if the results are reasonable. Plots for various fields from a reference simulation of this case are provided below for comparison purposes.
**To create your own plots to compare with the verification plots provided above, please use plot.ncl and make changes as necessary.
Winter Cases
Listed below are the Metgrid files for
six test cases from Winter 2012. The domain is over the US,
with a grid size of 20 km covering 290x190 points.
The namelist.input template has 40 eta levels.
For the testing, developers may generate 24-hr forecasts using any of the
initialization times. To assist the reasonableness review,
provided below are verification plots to compare against one of
the winter cases run.
Domain map
A map of the domain for the test cases
is here: domain map.
* This case (2012020300) is also provided as a case to check if the results are reasonable. Plots for various fields from a reference simulation of this case are also provided, for comparison purposes.
** To create your own plots to compare with the
ones provided here, please use plot.ncl, and make changes as necessary.
Updated Code
If you are considering contributing code to WRF,
please use the most recent WRF release or obtain a current version of
the WRF repository code from a member of the Developers' Committee.
The most recent release may be obtained from the WRF download site, shown below.
WRF Downloads
|
