Part of open science is that researchers make their data FAIR: Findable, Accessible, Interoperable and Reusable. But how to do this? In this series, researchers tell us about their data management choices.

In this edition, we highlight the article ‘Putting FAIR into practice for ecologists: How to make ecological data more reusable’ published in the Elsevier open access journal Ecological Informatics.

We asked co-author Cherine Jantzen, PhD student at the Faculty of Science and Engineering, a few questions about the publication.

"We hope to lower the threshold to make data FAIR and also increase the motivation to do so."

Your recent article addresses a challenge many researchers face: understanding how to translate the FAIR principles into concrete actions. What motivated you and your co-author to develop this workflow, guide, and FAIR+SET tool?

When  I first started making ecological datasets FAIR, I quickly became overwhelmed by the number of resources online, leaving me with no good idea of where to start. Some of these resources were very technical, making it difficult for me as an ecologist to understand how to implement this for the data I have and decide what is relevant for me. 

By collating information from many different sources and making several ecological datasets FAIR, my co-author and colleague Stefan Vriend and I realised many commonalities in the steps we took for each dataset and the discussions we had over different ways to implement these steps. We were further discouraged by many of the FAIR assessment tools, as many of them require your data to be available online, which was not always true for the data we used. These tools also required technical background knowledge on FAIR to assess the data yourself, but did not provide you with any help on how to improve your data. 

Talking to colleagues further made us realise that many ecologists do not make their data FAIR, simply because they do not know where to start or how to implement the rather abstract FAIR principles. This motivated us to use the experiences we gained in our own work to develop something that provides users with a simple way of getting started on FAIR, bundles existing resources and information in an accessible way, and makes FAIR usable in practice, resulting in the FAIR+SET tool, the hands-on guide for ecological data and our recent publication.

"Many ecologists do not make their data FAIR, simply because they do not know where to start or how to implement the rather abstract FAIR principles."

The FAIR+SET tool focuses on helping researchers identify concrete improvements rather than assigning a score, as many existing FAIR assessment tools do. What are the advantages of this approach?

Making data FAIR may not be the favourite activity of many researchers, as they often enjoy data collection much more than data management. The motivation for making your data FAIR may therefore not be very high. If you then were to assess your dataset with a FAIR assessment tool and you get a very low score, this can often feel quite judgemental, a bit like “your data is not good enough”, which will likely not help the motivation to continue the process either. Most of these tools also leave you with this negative assessment and do not provide any clear instructions on how you can improve your data. 

With the FAIR+SET tool, we tried to change that by giving an indication of which parts of the data can be improved, directly connected to the respective sections in our FAIR guide that describe why this is important, and how you can achieve improvement. We also highlight in the guide which aspects are most important to work on first, which together with the tailored suggestions can make it less overwhelming to start making your data FAIR, especially if you are time constrained. The guide also provides practical examples from ecological use cases and some coding examples in R. Compared to other tools, we thereby hope to lower the threshold to make data FAIR and also increase the motivation to do so.

You emphasize that FAIR does not necessarily mean open. Could you elaborate on this distinction, and why it is important for researchers to understand it?

Findability and accessibility are often misinterpreted to mean the data have to be openly available, but you can have a perfectly FAIR dataset with restricted access, as long as the metadata (the description of the data) is openly available and describes where to find the data and how to access it. A dataset is findable when it has a persistent identifier (e.g., a DOI) which ensures that the dataset can always be found and referenced, and a detailed metadata description, which includes this identifier. While it does not matter whether the data are open, it is however, important that the metadata are open and indexed in a searchable resource. This enables both humans and machines to know about the existence of the data and to locate them. 

Accessibility does not mean that each user has immediate access to the data but rather refers to a clear description of how the data can be accessed (e.g., whether authorization is needed). Researchers have many reasons why they would be hesitant to make their data openly available, and assuming FAIR means open then keeps them from making their data FAIR, which is why emphasizing this distinction is very important. Even though I would also encourage researchers to make their data open and support the open science movement, this can be completely independent from FAIR. Whether or not the data are open, FAIR data can enhance understandability and reuse of the data, while also securing the data’s persistence and enabling integration with other datasets.

"Whether or not the data are open, FAIR data can enhance understandability and reuse of the data, while also securing the data’s persistence and enabling integration with other datasets."

Although your workflow was developed for ecology, many other disciplines face similar challenges with research data. Which elements do you think could be transferred to other scientific fields, and which parts are uniquely tailored to ecology?

The workflow we developed is not specific to ecology, only the implementations of each step are (partly) ecology-specific. We believe that the general steps of making data FAIR are the same independent of the domain, the data structure, or the data contents. As the FAIR principles always require detailed metadata, a persistent identifier, or the use of FAIR vocabularies, the general steps of implementing this to data will always remain the same, and the general workflow can therefore easily be transferred to other fields. 

For the implementation of the workflow steps, this can be different, as those depend more strongly on the data format and contents. For example, using a standardized vocabulary to describe the data structure of an ecological dataset will require different terms than one describing a medical dataset. When choosing how to implement steps, such as choosing data standards or a repository to archive the data, it can further be helpful to choose options that are widely used within your research community, because that will facilitate and enhance the (re)use of your data by colleagues. The rationale behind the implementation suggestions described in our guide can therefore still be helpful for other scientific fields, but the practical steps of improving the data can differ.

"We believe that the general steps of making data FAIR are the same independent of the domain, the data structure, or the data contents."

You argue that making data FAIR can feel overwhelming for researchers who are not trained in data management. What practical first steps would you recommend to researchers who want to apply the FAIR principles in their research?

The most important  – and likely even easiest – step is to describe your data. Creating detailed metadata that contains information on the contents of the data, but also on the data collection, administrative information, and the data structure, is a big step towards FAIR and can help both yourself and others tremendously to understand and reuse the data. 

Secondly, it is important to ensure the long-term persistence of the data and make it findable. This can be achieved by archiving them in a good repository, which is one that (amongst other things) assigns a persistent identifier, allows for the desired access option (e.g., restricted access), indexes the data and ensures that your metadata stays available, even if the data does not.

Choosing a good repository can require some time, but there are good tools to help with that (e.g., re3data.org), and it is a crucial step to prevent valuable data from getting lost. Even if the data are not yet interoperable, which definitely is the most technical and time-consuming step of FAIR, archiving your data together with detailed metadata in a repository is an easy start towards FAIR and will bring you already half the way.

Besides that, my own experience taught me that making data FAIR becomes a bit easier and quicker with every new dataset you work on. Lastly, we hope that the guide we developed can further help to make FAIR less overwhelming and find the start point that is needed for the dataset at hand. 

"Making data FAIR becomes a bit easier and quicker with every new dataset you work on."