Quantifying embodied greenhouse gas emissions in food consumption: a multi-spatial scale analysis using Taiwan as a case study Free

According to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2022), densely populated urban areas account for around 70% of global GHG emissions from human activities. Cities are pivotal both in exacerbating climate change and as crucial hubs for mitigation (Mi et al., 2019). The food system is recognized as a significant source of GHG emissions (Crippa et al., 2021; Rosenzweig et al., 2020). Achieving the ambitious net-zero emissions targets of the Paris Agreement is unlikely without considerable reforms within the food sector (Zurek et al., 2022; Tubiello et al., 2021; Mbow et al., 2019; Gil et al., 2019; Niles et al., 2018). Identifying key emission sources is essential for designing targeted interventions to efficiently and effectively mitigate emissions within the food system (Rosenzweig et al., 2021). To this end, precise and comprehensive GHG emission inventories are crucial for setting baselines, prioritizing actions, and monitoring progress toward these critical climate goals.

Managing food-related emissions poses significant challenges, particularly in the context of rapid urbanization, which has resulted in an uneven spatial distribution of food consumption and production. Rural areas typically exhibit higher in-boundary emissions due to their role in food production, whereas densely populated urban areas contribute substantial out-of-boundary embodied emissions from consumption. Over the past decades, trade liberalization and advancements in shipping technologies have further broadened the spatial spectrum of the food supply chain (Chung and Lee, 2020). This has intensified the spatial disconnect between the locations of consumption and the environmental and social impacts of production, resulting in increasingly complex footprints that are difficult to trace and manage (Steen-Olsen et al., 2012; Wiedmann and Lenzen, 2018; Wood et al., 2018). Moreover, comparative advantages in food production and top-down national trade policies play a crucial role in shaping food supply patterns and the associated embodied GHG emissions.

Given that much of the food consumed in urban areas is sourced from other regions, opportunities for mitigating urban emissions extend well beyond the operational boundaries of cities. This reality underscores the need to manage GHG emissions related to food production and consumption across multiple spatial levels, both domestically and internationally. Although existing studies have addressed spatially disaggregated GHG emission inventories (Gately and Hutyra, 2017; Gurney et al., 2019; Han et al., 2020), the interconnections between emissions across different spatial scales remain insufficiently examined. To address this gap, our study develops a methodological framework for quantifying embodied emissions within the food system. This framework enables a multi-spatial comparison, through which we demonstrate that higher-emission areas emerge from distinct underlying drivers—such as population density, dietary structure, income and import dependence—thereby offering policy-relevant implications.

Specifically, the quantitative methodology developed in this study is designed to rigorously analyze food-related emissions across three dimensions: types of food, emission phases, and spatial scales regarding production and consumption. First, food type exerts a decisive influence on emissions, with rice and beef identified as the largest contributors among plant- and animal-based commodities (Xu et al., 2021). Dietary choices and consumption patterns are critical levers for mitigation, primarily through their impacts on supply-side activities (Dalin and Outhwaite, 2019; Hayek et al., 2021; Willett et al., 2019). Second, the phases of emissions critically shape the overall carbon footprint. Food-related emissions are generally classified as direct or indirect (Ritchie et al., 2018). Within this framework, production accounts for the dominant share of life-cycle GHG emissions (approximately 83%), whereas transportation and retail together contribute about 15% (Weber and Matthews, 2008). These findings highlight the necessity of adopting a life-cycle perspective to capture the full spectrum of food-related impacts. Finally, spatial scale is a crucial yet often underexplored dimension of emissions analysis. It reveals how aggregate supply-demand dynamics interact with localized emission patterns, thereby offering a nuanced perspective on mitigation. Although prior studies have examined emissions across different stages of food consumption (e.g. Khan et al., 2024; Li et al., 2024), they have largely focused on domestic supply chains. Yet, in the era of global food trade, spatially explicit data are indispensable for accurate estimates and effective mitigation strategies (Charkovska et al., 2019). Integrating such data into policy design allows decision-makers to identify high-impact intervention points, thereby enhancing both the precision and the effectiveness of GHG reduction efforts in food systems.

In developing our quantitative methodology, Taiwan is selected as the case study due to its high urbanization rate, distinct rice-based dietary preferences, and significant reliance on international imports, which exemplify multi-spatial patterns of food consumption and their broader environmental impacts. The insights derived from this study could inform strategies for effectively managing food-related GHG emissions. The novelty of our quantitative approach manifests in two distinct aspects. First, it involves the comprehensive integration of diverse datasets, including food balance sheets, international trade statistics that detail origin-destination patterns, local agricultural production data, and microeconomic census data capturing consumption differences at finer spatial scales (e.g. cities or counties). Second, our approach employs a life-cycle perspective to estimate the emissions from production and international transportation by specifically calculating these emissions based on the emission factors and shipping distances for different types of food consumed. This approach enhances our understanding of the distribution of GHG emissions from different types of food, emission phases, and spatial scales.

The remainder of the study is organized as follows. Section 2 introduces our data and methodology. Section 3 presents our quantitative results. Finally, the conclusions are spelled out at the end of the study.

This study adopts the data of Taiwan Food Balance Sheet (TFBS) in 2020 ^[1], which aggregates the national food system into 11 major categories: cereals, tubers, sugar and honey, nuts and oilseeds, vegetables, fruits, meats, eggs, aquatic products, dairy products, and fats ^[2]. To maintain consistency with national statistics and ensure comparability across cities/counties, these 11 categories serve as the basis of our analysis. Food consumption across counties and cities is estimated using expenditure shares derived from the Household Income and Expenditure Survey. To align with these data, the food categories in TFBS are consolidated into three groups: staple foods, side dishes, and fruits. As shown in Table 1, the staple food group consists of rice, wheat, corn, sorghum, and other grains. Fats, treated as by-products of animal products, are excluded from this study due to the lack of sufficiently detailed consumption statistics. The remaining food items, with the exception of fruits, were classified as side dishes ^[3].

Figure 1 illustrates the integrated framework for estimating the embodied GHG emissions associated with food consumption at various spatial scales. The emissions are calculated based on the origins of food, distinguishing between domestically produced and imported items. As illustrated in the left flow of the figure, the embodied emissions from consumption of domestically produced food are estimated in three steps. First, the domestic net supply volume of each food item is calculated by subtracting exports from domestic production. Next, the total embodied emissions of each food item are calculated by multiplying the associated net supply volumes and GHG emission factors listed in Table 1. Finally, the total embodied emissions for each food item are downscaled to lower spatial scales (i.e. counties or cities). Specifically, the embodied GHG emissions of consuming domestic-produced food $j$ in city/county $i$⁠, $ETLi,jD$⁠, are calculated as follows:

Let $Pi,jD$ and $Xi.jD$ denote domestic production level and exports of food $j$ in city/county $i$⁠, repsectivley; $ej$ is the GHG emission factor of food $j$⁠; $Ci,j$ is the consumption share of food $j$ in city/county $i$ relative to national consumption. The production data $Pi,jD$ and export data $Xi.jD$ are collected from agricultural production and trade statistics of the Ministry of Agriculture in Taiwan. The consumption shares $Ci,j$ are derived using data from Household Income and Expenditure Survey. This study encompasses all counties and cities in Taiwan, including Northern Taiwan (Taipei City, New Taipei City, Keelung City, Taoyuan City, Hsinchu County, Hsinchu City, and Yilan County), Central Taiwan (Taichung City, Miaoli County, Changhua County, Nantou County, and Yunlin County), Southern Taiwan (Tainan City, Kaohsiung City, Chiayi County, Chiayi City, and Pingtung County), and Eastern Taiwan (Hualien County and Taitung County).

The emission factors $ej$ adopted in this study (Table 1) for most food items are obtained from Poore and Nemecek (2018), which provides a comprehensive global dataset on food-related environmental impacts. However, since emission factors for honey, soybeans, and other grains are not reported in this reference, alternative values are adopted from Kendall et al. (2010), Arrieta et al. (2018), and Porter et al. (2016) ^[4]. As shown in Table 1, there is a significant variation in emission factors across different food items. Specifically, rice exhibits relatively high emission factors compared to other staple foods such as wheat, corn, sorghum, and other grains. This disparity is primarily due to methane generation during rice cultivation, which has a global warming potential 25 times higher than the same amount of CO₂. Furthermore, elevated emission factors for livestock, including beef and lamb, can be attributed to direct methane emissions during the husbandry process and emissions from the production of animal feed throughout the life cycle.

The right-hand side of Figure 1 illustrates the estimation procedure for the embodied emissions of imported food. These emissions are attributed to two main sources: foreign production and international transportation. The calculation involves multiplying the statistics of food import volumes by the emission factors provided in Table 1. To downscale the total emissions for each food item from different import origins to lower spatial scales (i.e. counties or cities), we rely on data from the Household Income and Expenditure Survey. Specifically, the embodied GHG emissions of consuming imported food $j$ from country $k$ in city/county $i$⁠, $ETLi,j,kF$⁠, are calculated as follows:

where $Mj,kD$ represents the import volume of food $j$ from country $k$⁠.

To calculate emissions from international transportation, we gather data on shipping distances between the imported origins and Taiwan. Food imports are further categorized based on shipping modes, which include bulk cargo and containerized cargo. The associated emissions from international transportation are then calculated as the product of food import volumes, shipping distance, and emissions factors of international transportation. Specifically, the embodied international transportation GHG emissions of food j from country k to city/county $i$ is calculated as follows:

where $Dk$ is the shipping distance between country $k$ and Taiwan; $Fj$ is the emission factors of transportation for food $j$⁠. The emission factors $Fj$ are adopted from Lee et al. (2016). For bulk cargo, which applies to rice, wheat, maize, sorghum, other grains, sugar, honey, soybeans, and other oilseeds, the emission factor is 4.9 g CO2​/ton-km. For general cargo, which applies to leafy vegetables, fruiting vegetables and mushrooms, root and tuber crops, fruits, pork, beef, lamb, poultry, eggs, aquatic products, and fresh milk, the emission factor is 18.3 g CO2/ton-km. Based on the statistics of food trade in Taiwan, the origins of food imports are diversified, covering a total of 104 countries. Table 1 provides a summary of the top two import origins for each food item. To capture the emissions associated with expenditure across different cities/counties, we calculate emission intensities, defined as kgCO2e/NTD$1,000.

This section presents our estimated embodied emissions from food consumption in Taiwan, both at the aggregate level and broken down by cities/counties. At the aggregate level, our estimated emissions exhibit significant variation across different sources and food types (Table 2). In total, food consumption contributes 33.45 million tonnes of CO2e (MtCO2e), with domestic production, foreign production, and international transportation accounting for 15.31 (45.77%), 17.14 (51.22%), and 1.01 (3.02%) MtCO2e (% in total), respectively. Notably, embodied emissions from the consumption of foreign food surpass half of the total emissions, primarily due to Taiwan’s relatively low food self-sufficiency. Concerning embodied emissions from various food types, the top five contributors are as follows: (1) domestically produced rice (6.03 MtCO2e from domestic production), (2) imported beef (5.22 MtCO2e from foreign production and an additional 0.03 MtCO2e from international transportation), (3) imported soybeans (4.20 MtCO2e from foreign production and 0.19 MtCO2e from international transportation), (4) domestically produced pork (3.68 MtCO2e), and (5) imported corn (2.40 MtCO2e from foreign production and 0.39 MtCO2e from international transportation).

Although international transport contributes only a small share of total emissions, several foods show outsized transport-related impacts (Figure 2) because of high shipment mass and long distances. Heavy staples and perishables—sugar, root and tuber vegetables, and fruits—raise weight-driven emissions, while bulk grains—wheat, corn, other grains, and soybeans—are largely sourced from distant, non-Asian suppliers, amplifying distance-driven emissions.

The results of county- and city-level food consumption emissions are presented in Figure 3. Among all jurisdictions, Taipei City records the highest embodied emissions, followed by New Taipei City, Kaohsiung City, Taoyuan City, Taichung City, and Tainan City. This pattern suggests a positive association between emission levels and population size. However, income levels and consumption preferences also exert a substantial influence. For instance, although New Taipei City has the largest population and number of households, Taipei City surpasses it in food-related emissions. This outcome can be attributed to Taipei City’s higher average household disposable income and a greater proportion of expenditure devoted to side dishes and fruits, which together elevate its overall emission profile.

To gain deeper insights into emissions from three food categories and their origins, the following analysis concentrates on the six major cities in Taiwan. As shown in Figure 4, foreign side dishes are identified as the predominant source of emissions in these cities, accounting for over one-third of the total emissions. The second and third largest contributors are domestic staple foods and domestic side dishes, respectively. It is noteworthy that international transportation accounts for only a small portion—approximately 3% to 4%—of the total emissions for each city.

Among the six cities, New Taipei City, boasting the largest population, stands out for having the highest embodied emissions from staple food consumption. It leads in emissions from both domestic and imported staple foods. Taichung ranks second for domestic staple food emissions, whereas Taipei City holds the second position for imported staple food emissions. This is likely due to the more Westernized dietary habits prevalent in Taipei City, which result in higher emissions from the consumption of imported staple foods compared to Taichung City. Regarding embodied emissions from the consumption of side dishes, Taipei City claims the top spot, followed by Kaohsiung City, New Taipei City, Taoyuan City, Tainan City, and Taichung City. For embodied emissions from fruit consumption, Taipei City again ranks highest, succeeded by New Taipei City, Kaohsiung City, Taichung City, Taoyuan City, and Tainan City. The elevated emissions in Taipei City can be attributed to its high disposable income and distinctive consumption preferences.

The above findings indicate that the six major cities in Taiwan exhibit higher emissions compared to other cities and counties, primarily due to their larger populations. This raises the question of whether urban food consumption patterns inherently cause greater environmental harm. To investigate this, emission intensities and emission per capita across all cities and counties (Table 3) are compared to uncover spatial variations and key drivers. Among the three food categories, staple food shows the highest average emission intensity (88.58 kgCO2e/NTD$1,000), followed by side dishes (45.19 kgCO2e/NTD$1,000) and fruit (8.7 kgCO2e/NTD$1,000). Despite the six major cities having higher total embodied emissions from food consumption, their emission intensities remain moderate when compared to other regions in Taiwan. Notably, both Taipei City and Kaohsiung City display emission intensities lower than the average of Taiwan. Furthermore, Taipei City’s emission intensity is lower than that of Kaohsiung City, particularly due to its lower emission intensity of staple food (68.76 kg CO2e/NTD$1,000). In contrast, rural counties such as Changhua, Yunlin, Chiayi, Pingtung, Taitung, and Hualien exhibit emission intensities that exceed the average of Taiwan, primarily driven by high emission intensities from staple food consumption. These rural counties have a pronounced “rice-based dietary culture”, with rice occupying a significant and indispensable role as the primary staple in daily diets.

Per-capita emissions exhibit substantial spatial heterogeneity, spanning 961–1,696 kg CO₂e (with mean of 1,351 kg CO₂e). The upper tail is led by Chiayi County (1,696 kg CO₂e), Taipei City (1,663 kg CO₂e), and Yilan County (1,639 kg CO₂e), whereas Taitung (961 kg CO₂e) and Yunlin (983 kg CO₂e) anchor the lower bound. Across the six special municipalities, Taipei is conspicuously high; Taoyuan is mid-to-high (1,524 kg CO₂e); Kaohsiung (1,365 kg CO₂e) and Tainan (1,340 kg CO₂e) cluster near the mean; and both Taichung (1,159 kg CO₂e) and New Taipei (1,121 kg CO₂e) fall well below it. The six-metro average (≈1,362 kg CO₂e) only marginally exceeds the national mean, and several non-metropolitan counties surpass most cities. Overall, the evidence does not support a systematic “urban premium” in per-capita emissions; rather, the pattern likely reflects cross-jurisdictional differences in diet composition (e.g. animal-sourced shares) and disposable income.

Utilizing the food consumption patterns of Taiwan as a test case, this study takes a pioneering step in linking embodied GHG emissions from food consumption across multiple spatial scales. We develop an original and systematic approach to quantify the associated emissions by connecting top-down food production and trade statistics with bottom-up data on household income and expenditure. This scientific effort contributes to advancing the methodological approach in multi-spatial GHG management and sustainability.

Our multi-scale accounting shows that the embodied GHG footprint of Taiwan’s food consumption is driven largely by production, while international transport contributes only a minor share. When analyzing emissions from different food categories, foreign side dishes emerge as the leading contributor, accounting for more than one-third of the total emissions. Domestic staple foods and domestic side dishes represent the second and third largest sources of emissions, respectively. Urban areas exhibit higher total emissions compared to rural areas, primarily due to their dense populations. However, a closer comparison of the six major cities reveals that income levels and consumption preferences also play a significant role in determining their respective emission rankings. For example, despite having fewer households than New Taipei City, Taipei City records the highest embodied emissions from food consumption. While urban areas exhibit relatively high total emissions, their emission intensities remain moderate. Both Taipei City and Kaohsiung City demonstrate lower emission intensities than the average of Taiwan. Interestingly, certain rural counties exhibit higher emission intensities, largely due to their “rice-based dietary preferences”, where rice holds a prominent and indispensable position as the primary staple in daily diets.

The findings highlight several key policy implications that shift emphasis from distance to production, from territorial boundaries to embodied responsibility, and from single-point fixes to integrated portfolios. Firstly, local food cannot be unequivocally equated with sustainable food, aligning with the conclusions drawn by De Cara et al. (2017). Consuming local food alone may not be inherently effective in reducing associated emissions. Policy and managerial attention should therefore pivot from “food miles” to environmentally friendly on-farm practices, product mix and supply-chain performance, together with responsible consumption, such as minimizing food waste (Stein and Santini, 2022). Moreover, non-CO₂ emissions, especially methane, play a significant role in agricultural production. When these emissions are factored in, many food products demonstrate exceptionally high GHG emission intensities associated with their on-farm production. In such cases, the impact of GHG mitigation when shifting from imports to local food depends on the emission intensities specific to the investigated scenario. This coincides with the findings of Avetisyan et al. (2014).

Secondly, emerging international practice underscores the growing salience of agricultural mitigation. New Zealand’s recent initiative on GHG regulation in the livestock sector has attracted widespread attention, signifying a departure from the predominant focus on emissions control in the energy-intensive manufacturing sector seen in most countries. New Zealand’s livestock GHG regulation involves the imposition of taxation and related measures within the livestock sector (Mann and Wang, 2024; Sullivan, 2021), aiming to mitigate the environmental impact of methane and nitrous oxide emissions from animals, particularly cows and sheep. This initiative aligns with New Zealand’s commitment to reducing GHG emissions in the agricultural sector—a crucial component of its national economy— and reflects the broader global effort to combat climate change. As global climate risks intensify, it is anticipated that GHG regulation will increasingly extend to the agricultural sector (Ammal, 2025). However, GHG pricing in agricultural sector could raise food prices, imposing an additional burden on basic livelihood consumption. Food affordability is a cornerstone in food security and affects the level of consumption, balanced nutrition, quality of life, and life expectancy. While debates on climate justice often center on energy poverty, food poverty and affordability safeguards must likewise be addressed. Equity should be embedded in livestock-related GHG policy by pairing pricing or regulatory measures with protections that secure access to affordable, lower-emission diets for vulnerable groups. Concretely, this includes targeted rebates or e-vouchers, nutrition-secure school and social-care meals, and fiscal incentives for affordable plant-based staples (e.g. legumes, whole grains). These complementary measures align emissions mitigation with food security and distributive justice, ensuring that climate policy in the food sector is both effective and fair.

Thirdly, substantial cross-category differences in embodied emissions mean that dietary choices materially shape urban food footprints (Avetisyan et al., 2014). Urban net-zero strategies should therefore align with SDG 12 by prioritizing demand-side management through a bundle of information-based, market-based, regulatory, and nudging instruments (Ammann et al., 2023). Beyond consumption patterns, production practices in rural areas also influence embodied footprints: methane from paddy fields and farm-level manure/nutrient management raise the upstream intensity of rice and animal products. Accordingly, support for locally produced foods should be paired with agronomic measures that lower upstream intensities (e.g. residue and soil-carbon management, nitrogen-efficiency improvements) and with governance instruments—notably green public procurement tracked with Scope-3-like indicators and tied to verified intensity reduction. These interventions reduce production-stage emissions while maintaining yields and can be embedded in city–region food-system programs and climate plans. In parallel, consumer-level food-waste reduction and targeted “green nudge” designs can steer everyday choices towards more climate-responsible behavior, complementing the decarbonized production and supply chain.

Finally, cities are pivotal to climate mitigation given their large and rising contribution to global emissions. The emissions from the top hundred urban regions worldwide constitute 18% of the total global carbon footprint. Our quantitative results show that GHG emissions from food consumption in the six major cities of Taiwan exceed those in rural areas in total terms. Because emissions are generated predominantly upstream during production, and urban consumption relies heavily on external supplies, much of this footprint falls outside territorial city inventories. This creates a governance gap that undermines accurate assessment and effective mitigation. Accordingly, city GHG reporting should adopt dual inventories—territorial and consumption-based—with a dedicated food module detailing totals and intensities per capita and by category mix. Such inclusive statistics provide a clearer picture of environmental impacts of urban activities and support coordinated GHG management strategies across spatial scales, empowering urban areas to take a leading role in reducing overall emissions. In sum, by integrating cleaner production, smarter procurement and logistics, behavior-aware demand policies, and consumption-based urban governance, cities and nations can act on the true drivers of the food footprint and accelerate credible, equitable pathways to net zero.