Enhancing Agri-Food Transparent Sustainability - EATS
Lead Research Organisation:
University of Aberdeen
Department Name: Computing Science
Abstract
The UK has a legally binding target of 'net zero' greenhouse gas (GHG) emissions for 2050 (Scotland, 2045) and the Food and Drink sector has a vitally important role to play in helping to achieve this. This must be done while also improving nutrition, protection of ecosystems, reduced risks to soil, water and air quality. Delivery against these ambitious targets will require a range of measures to be adopted across the agri-food supply chain - not just primary producers but also processors, retailers and ultimately consumers. Over the last few decades rapid advances in processes to collect, monitor, disclose, and disseminate information (broadly classified under the concept of 'transparency') have contributed towards the development of entirely new modes of environmental monitoring and governance for supply chains. Unfortunately, existing approaches often suffer from limitations in terms of collection and dissemination of data; over-simplification of supply chains; power dynamics influencing information inclusion/exclusion decisions; and potentially perverse outcomes regarding how the information is used, by whom and to what effect.
Given these issues, we need to consider how best to capture information about supply chains in order to document existing sustainability practices in sufficient detail; this is necessary to not only support monitoring and reporting needs of all stakeholders, but also to promote additional pro-environmental behaviours and even re-configuration of the supply chain. Our vision is built around an actionable information ecosystem whose purpose is to deliver transparent sustainability - realised via three pillars that we refer to as: SEE-SHARE-ACT. The first of these encompasses the role of sensors and carbon reporting tools in capturing data about agri-food processes (SEE); the second is a trusted digital platform able to manage sustainability data and report it across supply chain actors(SHARE); the third is the use of data-analytics and machine learning to support decision-making and action (ACT).
But what would a trusted infrastructure for transparent sustainability look like, and how would it be framed by (and operate within) its wider environmental, social and economic context? Also - how would such a framework go beyond simply documenting the elements of a supply chain (actors, processes, inputs, outputs) to enable a holistic approach to monitoring, pro-environmental decision-making and action? We have assembled an interdisciplinary team of academics and user organisations spanning the livestock, soft-fruit and brewing sectors to investigate transparent sustainability. Together we will explore the following questions:
What datasets, indicators and decision-making processes are relevant to the different actors participating in supply chains to realize sustainable food futures (in the DE)?
How do we formulate appropriate vocabularies with which to characterise sustainability practices, their context and rationale, and facilitate data capture and integration?
Can we realize a provenance-based sustainability solution for supply chains, operating across a range of technologies and organisational boundaries, that is trusted and able to facilitate pro-environmental decision-making and action?
How do we exploit sustainability data assets and ML/AI technologies to inform decision making towards net-zero, resulting in demonstrable changes to practice and behaviour?
Answers to these (and the many other questions that will certainly emerge) will lead us to develop prototype solutions that will be evaluated with project partners. Our ambition is to create a means by which farmers and other food and drink supply chain stakeholders can create a more sustainable economy built upon trusted data regarding the lifecycle history of products for enhanced environmental and product safety in (therefore more resilient) food supply chains.
Given these issues, we need to consider how best to capture information about supply chains in order to document existing sustainability practices in sufficient detail; this is necessary to not only support monitoring and reporting needs of all stakeholders, but also to promote additional pro-environmental behaviours and even re-configuration of the supply chain. Our vision is built around an actionable information ecosystem whose purpose is to deliver transparent sustainability - realised via three pillars that we refer to as: SEE-SHARE-ACT. The first of these encompasses the role of sensors and carbon reporting tools in capturing data about agri-food processes (SEE); the second is a trusted digital platform able to manage sustainability data and report it across supply chain actors(SHARE); the third is the use of data-analytics and machine learning to support decision-making and action (ACT).
But what would a trusted infrastructure for transparent sustainability look like, and how would it be framed by (and operate within) its wider environmental, social and economic context? Also - how would such a framework go beyond simply documenting the elements of a supply chain (actors, processes, inputs, outputs) to enable a holistic approach to monitoring, pro-environmental decision-making and action? We have assembled an interdisciplinary team of academics and user organisations spanning the livestock, soft-fruit and brewing sectors to investigate transparent sustainability. Together we will explore the following questions:
What datasets, indicators and decision-making processes are relevant to the different actors participating in supply chains to realize sustainable food futures (in the DE)?
How do we formulate appropriate vocabularies with which to characterise sustainability practices, their context and rationale, and facilitate data capture and integration?
Can we realize a provenance-based sustainability solution for supply chains, operating across a range of technologies and organisational boundaries, that is trusted and able to facilitate pro-environmental decision-making and action?
How do we exploit sustainability data assets and ML/AI technologies to inform decision making towards net-zero, resulting in demonstrable changes to practice and behaviour?
Answers to these (and the many other questions that will certainly emerge) will lead us to develop prototype solutions that will be evaluated with project partners. Our ambition is to create a means by which farmers and other food and drink supply chain stakeholders can create a more sustainable economy built upon trusted data regarding the lifecycle history of products for enhanced environmental and product safety in (therefore more resilient) food supply chains.
Organisations
- University of Aberdeen (Lead Research Organisation)
- UNIVERSITY OF OXFORD (Collaboration)
- The Open University (Collaboration)
- Technical University of Madrid (Collaboration)
- Anheuser-Busch InBev nv/sa (Project Partner)
- Food Standards Scotland (Project Partner)
- Angus Growers Ltd (Project Partner)
- Scottish Pig Producers Ltd (Project Partner)
- Food and Drink Forum (Project Partner)
- 2 Sisters Food Group (United Kingdom) (Project Partner)
- Opportunity North East (Project Partner)
- Muntons plc (Project Partner)
Description | Hosting a Distinguished Visiting Fellow (Academic or Industrial) |
Amount | £1,800 (GBP) |
Organisation | SICSA Scottish Informatics and Computer Science Alliance |
Sector | Academic/University |
Country | United Kingdom |
Start | 06/2022 |
End | 06/2022 |
Title | Neural network pruning for federated learning |
Description | Federated Learning (FL) can be leveraged to train a model in a decentralised manner for the agri-food sector while ensuring data safety and privacy of individual farms or data silos. However, the conventional FL approach has two major limitations. First, data on individual silos may experience heterogeneity, resulting in an aggregated global model that can perform well for some clients but not all, as the update direction on some clients may hinder others. Second, significant communication costs may arise due to the constant updates between the clients and the central server. To address these issues, we proposed a technical solution using network pruning. The method was applied to a soybean yield prediction dataset using an algorithm that identifies lottery tickets on individual clients before passing the pruned models to the server for aggregation. The results demonstrated that this approach can reduce the amount of data that needs to be sent from each client to the server compared to the conventional FL method. Furthermore, rather than having a universal global model, client models can be personalised to enhance local yield predictions, hence reducing the overall overheads that contribute to resource usage. |
Type Of Material | Computer model/algorithm |
Year Produced | 2023 |
Provided To Others? | No |
Impact | The method has just been developed and evaluated on an openly available dataset (soybean forecasting). The method and model will further be tested on datasets that will be provided to us by our project partners in due course. |
Description | Transparent Emission Calculations (TEC) WG |
Organisation | Open University |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | A working group lead by University of Aberdeen that aims to develop open source toolkit for enhancing transparency of carbon emission calculations trough the use of semantic web technologies. |
Collaborator Contribution | One researcher from each partner institution attends weekly meetings and contributes towards development of open source software elements of the TEC toolkit. In additions, WG group is also working collaboratively on preparing academic papers. |
Impact | Two provenance based ontological models for describing carbon footprint calculations https://w3id.org/ecfo and https://w3id.org/peco |
Start Year | 2022 |
Description | Transparent Emission Calculations (TEC) WG |
Organisation | Technical University of Madrid |
Country | Spain |
Sector | Academic/University |
PI Contribution | A working group lead by University of Aberdeen that aims to develop open source toolkit for enhancing transparency of carbon emission calculations trough the use of semantic web technologies. |
Collaborator Contribution | One researcher from each partner institution attends weekly meetings and contributes towards development of open source software elements of the TEC toolkit. In additions, WG group is also working collaboratively on preparing academic papers. |
Impact | Two provenance based ontological models for describing carbon footprint calculations https://w3id.org/ecfo and https://w3id.org/peco |
Start Year | 2022 |
Description | Transparent Emission Calculations (TEC) WG |
Organisation | University of Oxford |
Department | Oxford University Innovation |
Country | United Kingdom |
Sector | Private |
PI Contribution | A working group lead by University of Aberdeen that aims to develop open source toolkit for enhancing transparency of carbon emission calculations trough the use of semantic web technologies. |
Collaborator Contribution | One researcher from each partner institution attends weekly meetings and contributes towards development of open source software elements of the TEC toolkit. In additions, WG group is also working collaboratively on preparing academic papers. |
Impact | Two provenance based ontological models for describing carbon footprint calculations https://w3id.org/ecfo and https://w3id.org/peco |
Start Year | 2022 |
Title | Emission Conversion Factors Ontology (ECFO) |
Description | An ontology for modelling emission conversion factors |
Type Of Technology | Software |
Year Produced | 2023 |
Open Source License? | Yes |
Impact | none to date |
URL | https://w3id.org/ecfo |
Title | Provenance of Emission Calculations Ontology (PECO) |
Description | Ontology for describing the provenance traces of carbon emission calculations |
Type Of Technology | Software |
Year Produced | 2023 |
Open Source License? | Yes |
Impact | none to date |
URL | https://w3id.org/peco |
Title | RML Mappings and Datasets for generating Emission Conversion Factor Knowledge Graphs |
Description | Modified public datasets and custom RML mappings for creating novel Emission Conversion Factor Knowledge Graphs described using the ECFO ontology |
Type Of Technology | Software |
Year Produced | 2023 |
Impact | none to date |
URL | https://github.com/TEC-Toolkit/cfkg |
Title | Semantic Carbon Emission Calculator |
Description | A software prototype for calculating carbon emissions for training of Machine Learning models. The software adapts existing open source project and enhances it with semantic technologies that make use of other project outputs including PECO and ECFO ontologies as well as Emission Conversion Factors knowledge graphs. |
Type Of Technology | Software |
Year Produced | 2023 |
Open Source License? | Yes |
Impact | none to date |
Description | EATS Workshop - Brewery |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | Regional |
Primary Audience | Industry/Business |
Results and Impact | A collaborative workshop with brewers of aimed at eliciting challenges related to carbon emission tracking in their sector |
Year(s) Of Engagement Activity | 2022 |
Description | EATS workshop - Soft Fruits |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | Regional |
Primary Audience | Industry/Business |
Results and Impact | A collaborative workshop with soft fruit growers aimed at eliciting challenges related to carbon emission tracking in their sector |
Year(s) Of Engagement Activity | 2022 |