Skip to main content

Developing eco-labelling for food sold on campus

Date

Can the probabilistic mathematics underpinning the hydrogen bomb aid us in creating accurate environmental impact labels for food consumed on campus?

Project overview

Food production significantly influences global environmental degradation and climate change, contributing to an estimated 35% of greenhouse gas emissions (GHG) in the UK. However, most studies on food's environmental footprints focus on GHGs and home consumption, neglecting other impacts and settings. Recognising and mitigating environmental challenges through its Environmental Policy, Sustainability Strategy and Environmental Management System, the University of Leeds is committed to becoming a net-zero organisation by 2030. This involves addressing the environmental impact of foods, a quarter of which in the UK is purchased and consumed outside the home, in places like the University of Leeds. One solution is developing environmental impact labels for food on campus to help consumers make environmentally sound choices. However, due to the high variability of the data (for example, GHG emissions for beef being drastically higher (216.68 kg per kg of meat) than those for low-impact dairy beef (4.85Kg per kg of meat) (Poore & Nemecheck, 2018), the Monte Carlo simulation (MCS) is implemented to calculate the variability in environmental footprints of campus catering recipes and integrated into an app.

Data and methods

This data science project is underpinned by the MCS developed by Stanislaw Ulam in the 1940s, during the development of nuclear warfare by the Manhattan Project. Although described as embarrassingly parallel computation, this simple simulation method is designed to account for variability in complex systems. While this method was originally used for calculating neutron diffusion in nuclear reactions, in this project, it was adapted to model environmental impacts for eco-labelling.

Figures utilised in the MCS were sourced from Poore and Nemecek's 2018 paper ‘Reducing food’s environmental impacts through producers and consumers’ and its accompanying dataset. The dataset is derived from a comprehensive study of 38,700 farms and 1600 processes from the supply chain in addition to packaging types and retailers. The data was extracted and transformed into a format denoting weights, units, ingredients, and various environmental impacts. Since Poore and Nemecek's procurement data often used broad categories—such as "Citrus Fruit" to encompass items like limes or lemons, or "Nuts" to represent various types of nuts— I created a lookup table to ensure that all related ingredients are counted under their respective broad categories. While this approach simplifies the environmental calculations, it provides solid estimations for understanding the impacts of different recipes. Recipe data were provided by the University of Leeds catering team.

To transform these results into actionable insights, the MCS code was integrated into a web application titled ‘Earth Plate.’ Earth plate automates the MCS for environmental impact assessment from user-inputted recipes. The app also automatically generates eco-labels and menus based on the environmental footprint of the ingredients used. The labels display icons for GHG emissions, land use and water use.

Figure 1. User interface of the EarthPlate app, with options to Calculate environmental impacts and also generate menus and labels.

The application was designed with both hospitality managers and workers, as well as consumers, in mind. For hospitality providers, the app assesses the environmental impact of recipes via the MCS and populates menus and labels with this information. For consumers of campus catering, the labels were designed by following the methods of Margaret Calvert who designed road signs that communicate essential information to motorists quickly and effectively with bold minimalistic symbols for rapid legibility. Like drivers who may be distracted, or in a rush, campus consumers are likely buying food on the go, or grazing a buffet at a conference. As such, the eco-labels were designed to convey key environmental impact information clearly and swiftly. Even when consumers are distracted or pressed for time, they can quickly understand the impacts of their food choices with little effort.

Figure 2. Examples of simplistic icons denoting Fresh water withdrawal, land use and GHG emissions

Figure 3. User interface of EarthPlate allowing users to select recipes and generate labels.

Figure  4.  Label generated for a home-made vegan sausage roll displaying key environmental impact metrics per 100g, including freshwater usage (107.17 L), land use (0.58 m²), and greenhouse gas emissions (0.11 kg CO₂eq). The blank square serves as a placeholder for a QR code, which can provide additional product information. The label is designed for integration with POS systems

Monte Carlo Simulation Implementation

The MCS was employed to model the variability inherent in environmental impact data. Each simulation consisted of 10,000 iterations to achieve statistical significance. Random samples were drawn from provided data ranges for each environmental metric of an ingredient. Custom Python functions utilising libraries like NumPy and pandas were created to perform the simulations efficiently. Ingredient impacts were normalised based on the total recipe weight, allowing for per-100g impact calculations.

Web Application Development

Flask, a Python web framework, was chosen for developing the Earth Plate application due to its simple integration with MCS written in python . The app's architecture consists of a backend that processes data and runs MCS, and a frontend that allows users to input recipes and view results. Users can manually enter recipes or upload files, after which the app parses the inputs and maps ingredients using a lookup table. The user can then choose to generate menus or labels. The results are presented in an accessible format displaying mean values and icons.

Figure 5. User interface of the Menu Generator.

Figure 6. Generated menu using Sausage role with varying percentages of lentils.

Data Management and Storage

Google Cloud Storage was implemented for storing datasets, recipe files, simulation results, and lookup tables. The cloud offers scalability and reliable access. Data retrieval and updates are managed through the app, ensuring users interact with the most recent information. The lookup table is maintained by updating it whenever new ingredients or substitutes are added, with changes saved back to the cloud storage. Security measures, including user authentication and secure communication protocols, were implemented to protect data and comply with regulations.

Case use

The web application enabled an evaluation of the impact of replacing varying percentages of meat with lentils in items such as sausage roles, addressing a key question from the catering department. Recipes for meat sausage rolls with 10% to 50% lentil replacement were inputted into EarthPlate. Linear regression was then calculated from outputted MSC summaries, providing answers to the departments query.

Key findings

  • Lentils replacement significantly reduces GHG emissions, land use, and freshwater withdrawals, with particularly strong effects on GHG emissions and freshwater withdrawals.
  • High R² values indicate that lentil content is an excellent predictor of environmental impact across all three factors.
  • Practical Implications: Shifting towards more lentil-based sausage rolls would likely have a very positive environmental impact, particularly in terms of reducing GHG emissions and water use.

Figure 7 The linear regression in this graph shows the relationship between GHG emissions and lentil percentage in the sausage roll. As the lentil percentage increases, GHG emissions per 100g decrease linearly, as indicated by the regression line.

Figure 8 This graph demonstrates the impact of lentil substitution on land use. Increasing the lentil percentage results in a linear decrease in land use per 100g, as shown by the regression line.

Figure 9 illustrates how freshwater usage declines with an increase in lentil percentage. The regression line shows a steady decrease in water withdrawals as lentil replacement increases

Value of the research

  • Provision of reliable information about the environmental impact of ingredients and recipes to consumers and hospitality professionals at the University of Leeds.
  • For hospitality professionals, the Earth Plate app can assess recipes' environmental footprints, enabling them to make more sustainable menu decisions.
  • Consumers will benefit from easy-to-understand labels that can help them make environmentally conscious choices.
  • This research work will contribute to the broader goal of reducing the university's overall carbon footprint as part of its commitment to becoming a net-zero organisation by 2030.

Potential further benefits lie in the scalability of the app, allowing anyone in the hospitality industry to use the tool to assess their recipes' environmental impacts. This opens the possibility for hospitality businesses to adopt similar sustainable practices with little effort at zero cost. This could support broader efforts in encouraging more sustainable practices within the food service industry, contributing in a practical way to environmental initiatives and sustainability goals.

Quote from project partner

“As a project partner, I’m excited about the impact that this research will have at the University of Leeds. The Earth Plate app will empower our catering professionals to make more sustainable menu decisions by assessing the environmental footprints of their recipes. Our customers will also benefit from clear, easy-to-understand labels on food items that can guide them towards making environmentally conscious choices. This initiative marks a significant step towards the university’s goal of becoming a net-zero organisation by 2030. Catering can play a crucial role in this journey, as food bought and sold on campus directly impacts our carbon footprint. By making informed decisions, we can collectively reduce our environmental impact and contribute to a more sustainable future.” Charlotte Davenport, Sustainability Programme Officer, Sustainability Service, University of Leeds.

Insights

  • The Monte Carlo Simulation, originally developed for calculating neutron diffusion in nuclear reactions, has been adapted to account for the variability in environmental impact data. This ensures that the environmental impact labels are based on probabilistically sound and statistically significant calculations, which helps to reflect real-world uncertainties in food production data.
  • The data analysis reveals that replacing meat with lentils in sausage rolls has a linear, significant positive impact on reducing GHG emissions, land use, and freshwater withdrawals. The linear regression shows high R² values, indicating that lentil content is a strong predictor of reduced environmental impact.
  • The Earth Plate app translates complex environmental impact data into user-friendly labels, enabaling hospitality providers and consumers alike to make more environmentally conscious decisions. The system is designed to scale, offering the potential for broader adoption across the hospitality industry, which can contribute significantly to sustainability goals like the University of Leeds' net-zero objective by 2030.

Research theme

Identify which LIDA research theme(s) this sits under:

  • Environment

Programme theme

  • Statistical Data Science
  • Data Science Infrastructures

People

Dr William James (School of Geography, Teaching fellow)

Dr Susan Lee (CDRC, Research fellow)

Charlotte Devenport (Sustainability Programme Officer)

Partners

Leeds University Sustainability Service and Leeds University Catering and procurement.

Funders

Consumer Data Research Centre (CDRC)