Open Source ensures visibility
Dr Wolf-Peter Schill talks about his experience with Open Science
Three key learnings:
- Open models are a necessary condition for transparency and peer review.
- Open Source is often an accelerator for publications.
- Open Science increases visibility in the international community.
Which Open Science practices are of particular benefit in your area of research?
WPS: My research group studies the changeover to renewable energies in power supply and sector coupling. We use electricity sector modelling for quantitative studies of possible future scenarios so we can make predictions about the role of individual technologies. That was my introduction to the topics of Open Science and Open Data, because we need input data for our models, of course, and the modelling results essentially depend on which kind of data have been inputted. Of course you can describe the model equations in a paper, but ideally you show the complete model, so that the computations are reproducible and transparent. That is why we decided to make the codes and input data of our electricity sector models freely available a few years ago.
Has this been a positive experience for you?
WPS: I actually had a key moment: a few years ago I submitted a paper to a rather highly ranked journal. During the first review round, the comments ranged from positive to mixed and there were a few queries. At that stage I decided to release the code. I simply uploaded the model to Zenodo and it was evident that this was the moment when I completely convinced three reviewers in one go. That’s probably also why the paper was published in that journal in the end. This experience clearly proved to me that openness has enormous benefits.
Do you think this would still happen today?
WPS: A lot has surely changed since then. Speaking for myself, I see the reproducibility of such model analyses as a basic requirement for being able to do a serious review. A peer review based merely on results put into writing must necessarily remain superficial. Therefore it is good and correct that using and sharing model codes or open data are becoming standard operating procedure. However, the fact that model analyses are Open Source doesn’t mean they can be reproduced simply and in a user-friendly way. In my opinion, Open Source is a necessary condition for reproducibility, but not sufficient for actual implementation. We in our group still have a lot of work to do before our models are easily understood and usable. We’re looking especially at the area of documentation, but you have to keep in mind that this is very much a question of time.
Do you believe that Open Source is going mainstream?
WPS: At universities, definitely yes. At other research institutions it’s not quite so clear-cut. Big energy system models have certain path dependencies, they have been developed with much effort and sometimes use proprietary data. This increases the institutional difficulties of providing models as Open Source. For us at DIW Berlin it’s easier: we have our own electricity sector model, fondly named DIETER, which we developed ourselves and where Open Source was part of the design from the first moment.
Has this benefited your collaboration with other institutions?
WPS: Definitely. I see two aspects here. In the first place, Open Source has brought a certain visibility. Several times I have been contacted by colleagues who have used our model for totally different contexts. A group in Singapore has worked with a derivative of DIETER, and a government institution in Australia has used our model for an official report. And the other day I found out on Twitter that a group in France has applied one of our models. That’s brilliant – and new contacts are made that wouldn’t have happened without Open Source.
The second aspects concerns how model codes are shared. As I mentioned before, the minimum is the deposition of the code somewhere, for instance on Zenodo. But more and more modellers now use model-based systems such as GitHub and GitLab. These do not only offer the possibility of sharing the code, but of showing its development. Most important of all, they enable the participation of outsiders. Other people can suggest improvements or provide their own contributions, and you can check immediately if they should be added to the master branch or not. This enables a true outside collaboration beyond mere sharing.
There is another reason why we structure our work with Git: although we are a comparatively small research group we ended up with too many versions of the model, it became unwieldy. And if you do an internal version control to unclutter anyway, you can just as well provide a public version. I am pretty much a Git newbie myself, and I profit a lot from my younger colleagues for whom working with Git is completely normal.
Regarding collaboration, I absolutely want to mention the Open Energy Modelling Initiative. It definitely helped us with opening up our research and it has brought us into contact with many interesting people. The website in general is a brilliant resource for anyone who wants to enter the field of open energy modelling. There’s also a wiki that gives a good overview, for instance about different kinds of open licences.
Does this openness have its challenges, too?WPS: Sure. For instance, we were involved in the development of the platform Open Power System Data. The goal is to gather data for energy market and energy system modelling from various free sources and to provide them in a central location as quality-controlled and processed versions. A lot of questions came up there, especially about licences. It was a big learning process for all involved, but a lot is going on now, too.
The interview was conducted on December 7, 2020.
About Dr Wolf-Peter Schill
Wolf-Peter Schill is deputy head of the department Energy, Transportion and Environment at the German Institute for Economic Research (DIW Berlin) and heads the research area “Transformation of the Energy Economy”. His research interests are renewable energy integration, energy storage, electric mobility, and other sector coupling strategies. His methodological focus is Open Source power sector modelling. At DIW Berlin, he works in various research projects funded by German federal ministries and the European Commission. He is an active member of the Leibniz Research Alliance Energy Transition and the National Platform Future of Mobility.