Correlation of 4 Greenhouse Gas Emissions datasets using the Family Food Module of the UK’s Living Costs and Food Survey for food eaten at home. LEAP 2025 conference, Oxford

Aim and Objectives

This study aims to compare the results of four distinct Greenhouse Gas Emissions (GHGE, kg CO2e) datasets by applying them to UK Family Food Survey purchase data to estimate the kg CO2e per kg for 19 broad categories of food per person per week.

By analysing the alignment and discrepancies between these datasets, we seek to understand the accuracy and reliability of GHGE estimates in the context of household food purchases. We selected four datasets to compare: : (A) Poore and Nemecek (2018) at a global resolution; and at UK level resolution (B) Ali et al (2022), (C) Wrieden et al. (2019), and (D) Reynolds et al. (2019a) . We mapped each dataset to the food and drink codes in the Family Food Module of the UK’s Living Costs and Food Survey. Our comparison provides insights into the strengths and limitations of each dataset, contributing to more informed policy-making by highlighting which regularly purchased food products consistently are associated with elevated GHGE and which have lower environmental impact. These findings can help policy-makers identify products for future policy intervention to improve environmental sustainability and contribute to national net zero goals, while also allowing them to evaluate unintended consequences of public health food policy on the environment.

Methods

Four distinct greenhouse gas emissions (GHGE, kg CO2e) datasets were identified, each matched to the Family Food Module of the Living Costs and Food Survey (UKGov 2024; Office for National Statistics, 2024).

• Poore and Nemeck (2018) consolidated data on the environmental impacts of approximately 38,000 farms producing 46 different agricultural goods. We selected the mean greenhouse gas emissions data (kg CO2eq/FU, IPCC 2013 including CC feedbacks) for these goods, representing the global average impact. These 46 agricultural goods were matched to the FoodEx2 classification system (see Reynolds et al. 2019b, 2021, 2022), which were then matched to the categories of the Family Food Module of the Living Costs and Food Survey using recipes from Reynolds et al. (2019a).
• Ali et al. (2022) provided emission factors for foods purchased in the UK, based on Scarborough et al. (2014), which in turn were derived from Audsley et al. (2010). The system boundary for these emissions was from production to the retail distribution centre.
• Wrieden et al. (2019) mapped total greenhouse gas emissions for 129 commodities from Audsley et al. (2010) to 526 food and drink categories of the Living Costs and Food Survey. The system boundaries included primary production, processing, and transport of raw materials and final products. We excluded cooking impacts for consistency with other databases.
• Reynolds et al. (2019a) provided concordance mapping for greenhouse gas emissions of 101 food items to 337 categories (eaten at home) and 316 categories (eaten out) of the Living Costs and Food Survey. The emissions data was based on Audsley et al. (2010), with the system boundary extending to the regional distribution centre. Reynolds et al. (2019a) also disaggregated complex categories and composite meals in the LCFS into individual components.

The four databases of greenhouse gas emissions (GHGE, kg CO2e/kg) were matched to the weekly average purchase data (g per person per household) for 2022-2023, sourced from the Family Food Module of the Living Costs and Food Survey (UKGov 2024; Office for National Statistics, 2024). Only the 337 categories (eaten at home) were used for this analysis. Results are presented on this poster in 19 'parent’ categories. The 316 categories (eaten out) were excluded from this poster.

The Living Costs and Food Survey includes purchase data from 3,993 households across the UK for 2022-2023. Households recorded all purchases of food and drink over two weeks, including those eaten at home and out of home. The LCFS collected data on the weights of all foods purchased and the amount spent (£) on each food and drink item per person per week, reported at the individual level. Prior to calculation, the number of eggs purchased per week per person was adjusted from number to weight, assuming 60g per egg. Figures were created with Datawrapper graphics software.

The greenhouse gas emissions metrics of comparison on this poster are 1) kg CO2e per kg of food products comparing databases directly, 2) the kg CO2e per person per week calculated using LCFS 2022-2023 purchase weights.

Results

Complexity: Figure 1 shows the number of multiplications between each greenhouse gas emissions dataset and food weights within the Living Costs and Food Survey. These differ as each dataset is mapped to different parts of the hierarchy. Figure 2 shows the example of the greenhouse gas emissions of bread being mapped to different parts of the Living Costs and Food Survey hierarchy. The greater the numbers of calculations performed the more complexity and possible variation between results.

Correlation between datasets: The correlations between raw datasets (figure 3a) provided in Figure 3 suggest that there are moderate (0.6) to strong (0.79) positive relationships between the datasets. The strongest correlations are observed between Ali et al (2022) and both Wrieden et al (2019) (0.79) and Reynolds et al (2019) (0.76), indicating that these datasets are more closely related in terms of greenhouse gas emissions data. When correlations are compared at the aggregated 'parent’ 19-food category level (figure 3b) all correlations are above (0.7) highlighting strong er correlations in aggregate.

In total the per person per week impacts ranged from 51.4kg of CO2e (Poore and Nemecek (2019)), to 16.8 kg of CO2e (Wrieden et al (2019)), with 38.7 kg of CO2e (Ali et al (2022)) and 18.4kg of CO2e Reynolds et al (2019). Figure 4 and Figure 5 show the embodied greenhouse gas emissions (kg CO2e) per person per week for the 19 'parent’ categories according to the 4 datasets, while Figure 6 shows the percentage contribution of the same data. These figures highlights significant variations in greenhouse gas emissions across different food categories and datasets. Poore and Nemecek (2019) generally report higher emissions for most categories, while other datasets like Wrieden et al (2019) and Reynolds et al (2019) tend to report lower values. These differences may be due to variations in attribution methodologies, system boundaries, and data sources used by each study.

There are considerable variations in emissions across different food categories and datasets. Poore and Nemecek (2019) generally report higher greenhouse gas emissions for most food categories, with notable exceptions such as cheese and non-carcass meat where Ali et al (2022) report significantly higher values. Emissions for milk and milk products, carcass meat, and fresh vegetables are particularly high in Poore and Nemecek (2019), while Wrieden et al (2019) often report the lowest emissions. Across all datasets, emissions for fish, eggs, and beverages are relatively low, with consistent values for eggs. Reynolds et al (2019) report the lowest emissions for fats and fresh vegetables, but higher values for cakes and fresh fruit.

Discussion & Conclusion

This analysis highlights significant variations in greenhouse gas emissions (GHGE) across different food categories and datasets. Poore and Nemecek (2019) generally report higher emissions for most food categories, particularly for milk and milk products, carcass meat, and fresh vegetables. This highlights the nuances of using a dataset with global average impacts compared to a country specific dataset. In contrast, Wrieden et al (2019) often report the lowest emissions. Ali et al (2022) report notably higher emissions for cheese and non-carcass meat, while Reynolds et al (2019) provide lower emissions for fats and fresh vegetables but higher values for cakes and fresh fruit. The consistency in emissions for fish, eggs, and beverages across all datasets suggests a more uniform impact in these categories. Considering three of the datasets are all based on interpretations of data from Audsley et al (2010), the simultaneous level of agreement in some categories and diversion in others is a surprising result. This underscores the importance of considering multiple sources when evaluating the environmental impact of food consumption.

The main implication for further research arising from our work is that researcher need to carefully select and interpret GHGE data when performing sustainable dietary analysis and policy modelling. We recommend that future studies use multiple emissions datasets when presenting their results to highlight possible variation and bias from LCA data selection.

Funding Declaration

Research on this poster was funded by the NIHR funded programme grant Health Economic Analysis incorporating effects on Labour outcomes, Households, Environment and Inequalities (HEALTHEI) for food taxes, NIHR133927). CR is also funded by UKRI (through the Building a green future and Building a secure and Resilient world cross UKRI themes), Defra and NERC and administered by NERC on behalf of the partners by the Joined up Landscapes (Project Reference: TBA); UKRI and NIHR as part of the Building A Green Future strategy the THRIVING Food Futures research hub(MR/Z506485/1) , and UKRI through the Healthy soil, Healthy food, Healthy people (H3) project (Project Reference: BB/V004719/1).

References

Ali, M. et al. (2022). Assessing pre-pandemic carbon footprint of diet transitions in UK nations and regions. International Journal of Production Research, 61(18), 6115–6130. https://doi.org/10.1080/00207543.2022.2104182

Audsley, E. et al. (2010). How low can we go? An assessment of greenhouse gas emissions from the UK food system and the scope reduction by 2050. Report for the WWF and Food Climate Research Network. http://dspace.lib.cranfield.ac.uk/handle/1826/6503

Office for National Statistics. (2024). Living Costs and Food Survey. UK Data Service. SN: 2000028. http://doi.org/10.5255/UKDA-Series-2000028

Poore, J., & Nemecek, T. (2018). Reducing food’s environmental impacts through producers and consumers. Science, 360, 987-992. https://doi.org/10.1126/science.aaq0216

Reynolds, C. et al. (2021). Comparing the environmental impacts of recipes from four different recipe databases using Natural Language Processing. Livestock, Environment and People (LEAP) Conference 2021 https://openaccess.city.ac.uk/id/eprint/27235/

Reynolds, C. et al. (2022). Comparison of Greenhouse Gas databases using FoodEx2 codes. 7th Annual Agriculture, Nutrition and Health (ANH) Academy https://openaccess.city.ac.uk/id/eprint/28348

Reynolds, C.J. et al. (2019a). Healthy and Sustainable Diets That Meet Greenhouse Gas Emission Reduction Targets and Are Affordable for Different Income Groups in the UK. Public Health Nutrition, 22(8), 1503–17. https://doi.org/10.1017/S1368980018003774

Reynolds, C.J. et al. (2019b). A Pilot Method Linking Greenhouse Gas Emission Databases to the Foodex2 Classification, Livestock, Environment and People (LEAP) Conference 2019. Saïd Business School. http://dx.doi.org/10.13140/RG.2.2.15990.34889

Scarborough, P. et al. (2014). Dietary Greenhouse Gas Emissions of Meat-Eaters, Fish-Eaters, Vegetarians and Vegans in the UK. Climatic Change, 125(2), 179–192. https://doi.org/10.1007/s10584-014-1169-1

UKGov (2024). Family Food FYE 2023 - GOV.UK. https://www.gov.uk/government/statistics/family-food-fye-2023

Wrieden, W. et al. (2019). Sustainable diets in the UK—developing a systematic framework to assess the environmental impact, cost and nutritional quality of household food purchases. Sustainability, 11(18), p.4974. https://doi.org/10.3390/su11184974