Linking Data Infrastructures from Farm to Fork

Linking Data Infrastructures from Farm to Fork A DRI 2025 Workshop led by DARE UK Rob Baxter, National Digital Infrastructure Technical Lead, DARE UK 17/11/2025 Linking Data Infrastructures from Farm to Fork | 2 2 Contents 1. Background & Workshop Goals ............................................................................................................................. 3 2. Use-case Review & Additions ................................................................................................................................ 3 3. Data Infrastructure Systems View ......................................................................................................................... 4 4. Understanding the Links ......................................................................................................................................... 5 5. Conclusions ............................................................................................................................................................... 6 Licence This work © 2025 by HDR UK is licensed under CC BY 4.0 Linking Data Infrastructures from Farm to Fork | 3 3 Linking Data Infrastructures from Farm to Fork A DRI 2025 Workshop 1. Background & Workshop Goals A workshop convened by the DARE UK programme in early 2024 1 was challenged to put aside any constraints and imagine what research could be done if data from multiple sources were easy to link and work with. That workshop came up with 52 use-cases across ten broad data sectors. This hands-on workshop at DRI 2025 began with these use-case, and others brought by attendees and, taking a high-level systems approach, tried to map them out, from data collection through linkage to research publication, by as it were, adding the constraints back in. Along the way our goal was to tease out what kinds of system we already have in the DRI landscape, what (or who) is missing, and what kinds of organisational models we might need to make these use-cases reality. On the day, five groups worked using Google slides or good old-fashioned pen and paper to map and connect the pieces of data infrastructure needed for their chosen use-cases. To keep it as straightforward as possible groups were asked to explore constraints in only three dimensions: data size, quality and sensitivity. The session was structured into three parts: 1. Review & discuss the original DARE UK use-cases, and pick one or two to work on, or add new ones. 2. Map out, at a high level, what data sources might be needed for a given use-case, what kinds of process they might need to go through. 3. Sketch the potential organisational boundaries for the different data sources and processes, and tease out the kinds of “contracts” that the use-case needs in practice. In the original 2024 workshop, the top-voted use-case on the day was “Transforming the food economy”; the 2025 DRI workshop was named in its honour. 2. Use-case Review & Additions The groups were asked to review the 2024 use-cases and discuss their ins and outs, and either agree on one or two to work with or come up with their own. Such was the expertise in the room, the groups came up with six new use-cases: W1 Planning for Housing - National Database. W2 Understanding whole chain carbon emissions associated with food products. 1 See https://zenodo.org/records/14025303 Linking Data Infrastructures from Farm to Fork | 4 4 W3 Enabling regulatory efficiency for food recall events. W4 [skipped] W5 Understanding the factors that most influence cardo-vascular disease and developing targeted interventions for high-risk groups. W6 Towards joined-up healthcare—developing complete medical records for patients covering NHS hospital and GP records. W7 COVID waste water analysis pipeline. The key question for each use-case was: what kinds of data will be needed to answer the research question? The results are captured in the table below. Use-case Health Social Economic Lifestyle Consumer Geographic Climate and environment Care and social system Justice Commercial Agricultural W1 W2 W3 W5 W6 W7 Note that a key feature is that any given use-case needs data from multiple sources, and that these are likely to need to be “linked” in some way. This echoed the findings of the 2024 workshop: research power depends increasingly on the linkage of data from disparate domains. 3. Data Infrastructure Systems View The goal of this part of the session was to encourage participants to take a system-level view of the data sources identified for their use-cases and to think about they would need to be prepared, linked and moved to enable the researchers in question to access them. Data can be classified in all sorts of ways, but for this exercise we focused on size (big v. small) and sensitivity (open v. not-at-all-open). Linking Data Infrastructures from Farm to Fork | 5 5 Constraint-wise, the use of sensitive data typically relies on approvals—legal, ethical or other—and the use of big data relies on significant infrastructure support, in terms of network, storage, tooling and so on. Another potential factor groups were asked to consider was data quality, assessed using a “medallion” model of bronze (raw data), silver (cleaned, tidied, “research-ready”) and gold (augmented, enriched). Below is an example from Group 4 who looked at use-case W7, COVID-19 wastewater analysis, based on actual experience from some of the group members who were involved in the actual activity during the pandemic. 4. Understanding the Links This final part of the session invited the groups to review the system-level diagrams they’d drawn and start to sketch organisational boundaries around groups of elements. From this, pictures began to emerge of the number, and nature, of contracts or other agreements needed between different organisations to realise any given use-case. Over the page is an example from Group 3 who covered use-case W1: what would we need to do to realise a national housing planning database? Linking Data Infrastructures from Farm to Fork | 6 6 The dotted purple lines indicate where the group reckoned the organisational boundaries would be. This recasts a moderately complex dataset integration problem into a major contractual challenge! 5. Conclusions At least anecdotally, session participants found it great fun! There were probably two clear, serious takeaways, though, both related to what we could term the economy of the emerging data linkage ecosystem. 1. Do not neglect data licensing! At least three of the groups identified the need for datasets from commercial providers within their use-cases. A repeat of this exercise would be wise to include cost as a key dimension alongside size and sensitivity. 2. Think of research as a business and research data as business objects. As data objects move through networks like those teased out through the workshop they are transferred between organisations. Thinking of each of those transfers as a business interaction may allow us to bring a different—and useful—set of tools to bear on the challenges involved. And finally, the session organisers would like to thank all the participants for their engagement and insights! Linking Data Infrastructures from Farm to Fork | 7 7