attendees:
Chris Baker (standing in for Paul Harris)
- head of intelligence ecosystem rothamptstead, oldest agriculture reserach institute in the world
- Paul: runs platform at northwick farm in devon. collecting farm inputs on experimental farms (8 million data points). Interested in developing process based models, data twin
Harry
Edwin Brown
Ruchi Choudhary
Markus Hauru - Turing, worked on previous iteration of CROP, developing DB base for DT package
Edwin Brown - Turing & Sheffield, RSE,
Giorgio Cerro - PhD high energy physics, Enrichment student collaborating with Markus
Daniel Delbarre - Data Wrangler at Turing, working on E&S GC
Harry Meacher - program manager for E&S GC, facilitating WS2.2

Previously have heard from work at Rothamstead, Cambridge on CROP
goal for this meeting: what can be achieved within the next funding envelop (up until March 2024)
- We need to flesh out all the RSE front components.
- what is necessary to host Chris's data & facilities?
- extending team, which is great to see
- Ruchi suggests to build out the group website, map out tangible outputs etc
- ==Ruchi asking me if we can help with creating website, Harry saying this should be through E&S==
- Chris also involved in 2.3 (scivision), eg detecting and identifying beneficial insects vs pests, phenotyping on wheat cultivars
- Markus: rothamstead working on API, have data dumps they can provide. Markus Edwin and Paul to meet but they have very clear steps in mind for implementing the weather forecasting. Planning to start with the first data dump from rothamstead website, only then move onto more real-time data processing. Only then to discuss models and variety of models
- Giorgio happy to start working on data.
- Markus:lots of transfer from CROP to DT base needed, eg implementing user management system
- CROP was specifically build for underground farm. But can it be repurposed at all? Seems like DT base is build from scratch. > depends user management system should be straight forward to pull over, though we might do some modernisation. Some front-end aspects like the visualisations can copy over. Other aspects require more work, e.g. database, very particular data table (co2 readings etc), now that we want this to be a general DT package we need this to be more general-purpose
- Ruchi: we should create a general data structure that works across Rothamstead, willow farms? in surrey that rebecca will work with. So there should be data dumps from three different environments
- Harvest London confirmed they are on board (Pat will take lead?)
- Willowland farm? willow babytrees grown in controlled environment. Chris also has some experience with willow-specific farming?
- Ruchi: wants to build ecosystem of partnerships
- Harry: yes ecosystem should be build across E&S GC, new RAM will come on to map out stakeholders etc
- Chris: we have clear deliverables we'd like to achieve: we want to be integrated & good citizens, but very focused on our own deliverables.
- In the short term these are ("quick wins")
- we want API up and running
- we want data integrated into DT base
- we want running ?? models
- demonstrate that we can run crop yield models and improve accuracy and reducing the degree of uncertainty.
- visual components
- we have tools to facilitate rich query expressions for generating simulations (semantic querying approach)
- "Simplace" - open source process based models (coming out of germany, Bonn)
- AIUK for first target, what are we aiming to demonstrate for AIUK?
- all data infrastructure for harvest farm and rothamstead (proper data curation)
- focus on showing interoperability of DT base and livestream
- Chris: challenge here is. if we are too focused on the technology we could be any sector, but this is Ag & farm, so the impact needs to be on the farm scale. We cannot just showcase that we have a platform but also what it actually does
- Ruchi: yes, but also connectivity between farm and digital twin is important
- Benchmarking can be post march, we learned that the models worked great until we used live data. Because normally when we model data we curate data quite a lot, whereas with live data models need to update themselves, and that works differently.
- Harry: no demo actually this year at AIUK, but instead we will launch E&S and as part of this crop will be shown. But also a proposal to TPSAC will be send showing achievements and plans for future to gain more funding (4mil?)
- Markus: yes we want to do the live data, but what are the resources for this atm? (It seems some hiring is going on)
- Chris: we are hiring for the following post-docs:
- https://jobs.rothamsted.ac.uk/vacancies/684/digital-twin-scientists.html
https://jobs.rothamsted.ac.uk/vacancies/686/digital-twin-specialist-sensors.html
- they should start early January
- Ruchi's team will be Pat (PhD student) will work on models, Rebecca (front-ending the Surrey collaboration, she has greenhouse simulation model). Rebecca retiring in March. Then will advertising a position by february/march next year.
- Dan: experience with standardisation of data, very interested in helping out on that part. Will these data be shared?
- Daniel: he & another data wrangler will be allocated to E&S, so probably half a person could be committed to Crop
- when working with live data, different sensor types cause distinct problems. This is where we could use data wranglers to understand how to best curate the data
- Chris: data that goes into farm platform atm goes through a lot of preprocessing to avoid data issues.We can demonstrate how uncertainty goes down completely in silico. That's also why we are hiring a sensor specialist as one of the positions
- Markus: relational database, any experience? Daniel yes and happy to help, much experience on this around
- Markus: suggesting meeting with Daniel on interoperability, Xiaoxue might be interested in connecting
-