A splinter session yesterday drew a larger-than-expected crowd to talk about the use of free and open source software (FOSS) in the geosciences. Those in attendance spanned the range from developers to end-users and the main outcome is that there will probably be a dedicated FOSS in science session at EGU2013.
A list of FOSS for geoscientists
A lot of the discussion yesterday was about open source software used in computer models (e.g. Glimmer-CISM, a community-written model of ice sheet dynamics). Examples of more general use of FOSS in geosciences are using GRASS and QGIS instead of ArcMap, Python instead of Matlab, Inkscape instead of Illustrator. For a longer list and a discussion of the advantages of FOSS, see the post: All the software a geoscientist needs. For free!
Showcase of FOSS in research presented at EGU2012
Have you used FOSS in research that you are presenting at EGU2012? If so, add an advert for your work in the comments (even if it has already been presented). The list will demonstrate the wide variety of applications for FOSS in the geosciences. Abstracts for the presentations can be looked up on the conference planner website.
Topics discussed at the meeting
Much of the discussion was led by scientists who are also developers of the software that they used. Some of the main themes were:
- The need for reproducibility in science. If people cannot reproduce your results, how can they test if your hypothesis is correct? This issue was highlighted in a recent editorial in Nature, and it is likely that it will be increasingly necessary to publish code alongside results.
- Logistical issues with packaging code. Many open source software projects are built different packages from many sources, each with their own schedule of updates. Sometimes, using a different version of one of these can cause a model to give a different result. The need was discussed to ensure that older versions of packages are always available so that results can be exactly reproduced.
- Getting credit for your work. If an open source model is used by the scientific community, it has, in effect, been peer reviewed. Can a system be developed so that the software itself becomes a citable bit of science, e.g. with a Digital Object Identifier (DOI)? How would this system cope with different versions of the same program?
- Ensuring good documentation. Scientists often write code for their own use and give little thought to documenting it for others. If an open-access journal could be created where documentation for a hydrological or climate model could be published, along with an example use case, this would make it easier for others to use the models as well as giving another way of citing the author(s).