Testing environment

[pesei]

It was generally agreed that standardized and automated test cases and model/measurement comparison metrics are urgently needed to test new model versions, validate model improvements and inter-compare FLEXPART with other models. Performance benchmarks are also desirable, not only to evaluate hardware but also to quantify the impact of code modifications on performance.
Consistency checks between forward and backward simulations would be useful as well; to reach full consistency, sources need to be defined as grid cell sources on the output grid.

[anphi]

igpis and anphi are working on a draft for requirements regarding the test environment and test cases

[hasod]

I am providing a first test case with a corresponding script that can be a starting point for such a development environment. You can download the data from my ftp drive at ​ftp://iacftp.ethz.ch/pub_read/sharald/flexpart. There are three files, one for the global ECMWF data (ec_lowres_tar.gz), one for a high-resolution nest (ec_highres.tar.gz), and a third file which contains the options and src for the FLEXPART run, and the compilation/execution/testing script (test_case_tnf.tar.gz). The tests so far are on successful compilation, successful starting FLEXPART, successful completion of the run, successful creation of output files. This could be easily automated and the output e.g. be posted on a website. The case is a three-day simulation of a tracer release over the Mediterranean during October 2012.

Someone please take over from here to implement an automatic build test environment.

[dearn]

A basis for a python-based testing environment has been developed in the CTBTO project framework (​https://www.flexpart.eu/wiki/FpCtbto). The testing environment description is available in ​https://www.flexpart.eu/wiki/TestEnv and distributed publicly with the source code in ​http://borealscicomp.com/CTBTO_FLEXPART/snapshot_flexpart_distros/stable-wo4.tgz