WP2: Exploiting the data - Project Management

Lead Partner: UOY
Objectives: This WP aims to address SO3 and SO6. We do this by better institutionalisation of food footprint information systems, achieved by (i) gathering and assessing the technological readiness, FAIRness and compliance to Product Environmental Footprint standards of key food-impact linked (LCA, footprint and other) datasets; (ii) developing recommendations for (and implemented selected) improvements via data-provider engagement; and (iii) increasing dissemination and uptake of knowledge via accessible AI technologies.

Tasks

T2.1: Generate a centralised repository of food-linked data Lead: ETHZ (5PM). Effort by contributors: UOY (5PM), ZHAW (4PM), SEI (2PM), COOL (2PM). Timing: M3-12. ata on the environmental impacts of food is scattered. Here, we will gather and interrogate existing food-related datasets (LCA, IO-based/footprinting, production and processing, nutritional, supply-chain management-linked, geospatial/remote-sensing and trade resources) encompassing information and datasets linked to production, processing, consumption and end-of-life/waste stages of the food system. The purpose of this review is to support standardization efforts and help build a roadmap for how to move from the current state of the art to robust PEF standards. The purpose of the review is not distil knowledge, i.e. it is not to get a robust understanding of food system impacts. The purpose is simply to support standardization efforts. We will centrally and systematically record key components of the data. We will document the scope of information being provided (e.g. supply chain stage covered, indicators specified/available, commodity coverage, specificity/resolution, time-series etc), the management of the data by data-owners (e.g. ownership, periodicity of uptake, project resourcing and marketing etc), examples of use-cases and applications (e.g. impactful case studies, known utilisation in policy or private-sector decision making) and availability to end-users (openness, licensing, availability of meta-data and methodology materials, investment in training materials etc). Contributes to: D2.1 (M21).

T2.2: Targeted investments for FAIRness and PEF compliance Lead: UOY (6PM). Effort by contributors: ULEI (3PM), DDS (3PM), SEI (2PM), COOL (2PM). Timing: M12-21. Data uptake grossly lags behind data availability. We will assess datasets (including those resources developed/owned by the project team, but also external parties) for technological ‘readiness’. Initial triage (based on Task 2.1 datasets) will identify those most important in providing (or having the potential to provide) actionable information to private sector and policy decision maker needs. Key criteria will include the development status of datasets (i.e. maturity as a product and potential longevity etc), uniqueness (i.e. ability to fill gaps in knowledge), levels of impact (i.e. existing uptake in practice, demonstrable use-cases), and scientific quality, including reported uncertainties. Triage will identify datasets which span the breadth of food sustainability issues (environmental domains, supply chain stages, end-users targeted etc) with the most ‘valuable’ use cases. For this subset of ‘most valuable products’ we will undertake assessment of adherence to FAIR principles and Product Environmental Footprint standards; systematically identifying weaknesses. For the latter, we will benchmark data provision and documentation against PEF guidance from the perspective of dataset readiness to provide information to the sixteen core environmental footprints, plus the optional ‘further environmental information’ aspects of PEF implementation. Best practice examples will be collated across the tools and datasets. Contributes to: D2.1 (M21), D2.5 (M42). This task is linked to the following milestones: MS2.1 (M15).

T2.3: Improve the technological readiness of key datasets Lead: UOY (5PM). Effort by contributors: NILU (5PM), CNRS (3PM), DDS (3PM), ZHAW (3PM), ULEI (2PM), SEI (2PM), ETHZ (2PM), COOL (2PM). Timing: M22-34. To enhance uptake, we will develop recommendations for each tool/data provider based on the assessment of weaknesses and best practice examples (T2.2). We will actively disseminate findings to data providers, who will be invited to attend one-to-one discussion meetings with the project team. We will also host two workshops across multiple partners (covering the same topic, but split to allow attendance from more parties and deeper interaction) to encourage further sharing of best practice and ‘challenges’ in responding to recommendations. An academic publication will also be produced to share outcomes openly. For tools provided by project partners we will, via project funding, implement selected recommendations. This task contributes to D2.2 (M24), D2.3 (M32) and D2.4 (M34). Milestone: M2.2 (M28) Stakeholder workshops (M28); M2.3 (M28): Assessment of which internal tool/data provider recommendations will be implemented (M28). Contributes to: D2.2 (M24), D2.3 (M32), D2.4 (M34). This task is linked to the following milestones: MS2.2 (M28), MS2.3 (M28).

T2.4: Building statistical robustness to improve PEF quality Lead: DDS (28PM). Effort by contributors: ZHAW (3PM), ETHZ (2PM), COOL (2PM). Timing: M6-36. So far almost no work exists estimating the reliability or confidence parameters of any food footprint data. The motivation of this task is to better understand statistical robustness within and across LCA databases, and thus to establish error budgets and prioritise improvements. Building on T2.1 and T2.2, and outputs from WP1, we will harmonise project data and other open data sources using a common glossary, and enter the resulting outputs into a high-performance data storage. We will also deduce logical relationships between inputs and outputs using current inventory datasets, supplemented with consortium expertise and AI. This logic will be used to create product system models which can be linked into flexible supply chains, and used to calculate the synthesised best-available knowledge of agricultural and other LCA data. The models and quantitative results produced can support decision making directly, either by providing direct or user-specific LCA results through an open web research portal, or through an interface to the AICA system (T2.5). Contributes to: D2.5 (M36).

T2.5: Launch an AI conversational agent (AICA) for food LCA supply chain and environmental impact info Lead: INTEC (24PM). Effort by contributors: ARBO (22PM), UOY (2PM), SEI (2PM), NILU (1PM), ULEI (1PM), ETHZ (1PM), CNRS (1PM), DDS (1PM). Timing: M26-46. We will develop a prototype AICA that answers user-posed questions in natural language about the impacts of the EU’s food system. It will utilise food linked LCA, footprint and other data developed within and outwith the project (incl. sources developed in WP1 and assessed in WP2), collated into a central structured database. Quantitative data will be complemented by information derived from research outputs (including outputs from WP3 and WP4) and other assessments. Training of the AICI will take place via development of common questions inspired by the EU’s policy context and the guidance received via stakeholders (WP5). We will undergo a thorough validation process, drawing on expertise from across the consortium. Iteration of the model based on feedback from the experts will take place before final production launch towards the end of the project. Contributes to: D2.6 (M42), D2.7 (M42). This task is linked to the following milestones: MS2.4 (M30), MS2.5 (M32), MS2.6 (M34).

Milestones

Table 3.1d: List of Milestones

Number	Name	Work Package No.	Lead Beneficiary	Means of Verification	Due Date (in months)
MS2.1	Triage of most valuable products from initial collated datasets	2	UOY	Initial list of most valuable datasets available to the project consortium	15
MS2.2	Stakeholder workshops	2	UOY	Two workshops with data providers held	28
MS2.3	Data collection protocol developed and circulated to allow data from partners to be provided via common standards (e.g. CSV/JSON/SQL)	2	UOY	Protocol available to all project partners for the purposes of providing data	28
MS2.4	Current project data mapped and entered into synthesis database	2	UOY	Project partners confirm their data is present via online review process	30
MS2.5	Data mapping to determine sources to be used in the AI database	2	UOY	AICA tool developers are provided with a clear list of datasets to be included in the model	32
MS2.6	Customised Generative Language Model (GLM)	2	UOY	GLM developed based on an open-source LLM	34
MS2.7	Validation completed via expert engagement with prototype AICA model	2	UOY	AICI model is ‘fit for purpose’ following validation of the model’s outputs by the research team.	42

Deliverables

Number	Relative Number in WP	Name	Lead Beneficiary	Type	Dissemination Level	Due Date (in months)
D2.1	1	Data assessment report	UOY	R	PU	21
D2.2	2	Recommendations sheets	UOY	OTHER	PU	24
D2.3	3	FAIR data recommendations	UOY	R	PU	32
D2.4	4	Summary of implemented recommendations	UOY	R	PU	34
D1.5	5	PEF-compliant LCI data of food	ZHAW	DATA	PU	42
D2.5	5	LCA results calculated and published using open infrastructure	DDS	DEC	PU	36
D2.6	6	Database	INTEC	DATA	PU	42
D2.7	7	Green Grocer FoodprintGPT	INTEC	OTHER	PU	42