Challenges and opportunities from an in-house cross collaboration between three research infrastructure data repositories

The ICOS Carbon Portal is the main data repository for the Integrated Carbon Observation System Research Infrastructure (ICOS RI), covering the domains Atmosphere, Ocean, and Ecosystems. Data from ICOS is available and accessible for humans and machines with a rich set of metadata under a CC BY 4.0 licence. The core services for the data portal (https://data.icos-cp.eu/portal/) are open-source software and are available on GitHub (https://github.com/ICOS-Carbon-Portal). The main goal for the development was to make the European greenhouse gas measurements accessible as FAIR as possible. This led to a mature and stable data portal which was subsequently adapted to be applied by another Research Infrastructure namely SITES, a national Swedish Infrastructure for Ecosystem Science, and the European Horizon 2020 project PAUL, pilot applications in urban landscapes (ICOS Cities). Although all three data portals use the same software core and are hosted at the ICOS Carbon Portal, they are independent from each other and base on slightly different ontologies. Hence, we have a unique opportunity to explore the challenges and opportunities of accessing and combining data from three or more different data sources and compare FAIR aspects of the datasets. How do we deal with attribution of the used data using correct citations? Do we have access to the licence for each data sets, are they different and what are the implications? How do we combine the data for further analysis keeping track of provenance and origin?

Further we will try to step back from the implementation of a service on specific data sets (which is kind of a hands-on bottom-up approach) and look at scalability to include other (environmental/ENVRI) data portals and think more about the top-down approach like the European Open Science Cloud EOSC. Can we offer a generalised service level for automated data processing from machine to machine? What do we need to process cross domain data sets?