Impact of cutting meat intake on hidden greenhouse gas emissions in an import-reliant city

Greenhouse gas emissions embodied in trade is a growing concern for the international community. Multiple studies have highlighted drawbacks in the territorial and production-based accounting of greenhouse gas emissions because it neglects emissions from the consumption of goods in trade. This creates weak carbon leakage and complicates international agreements on emissions regulations. Therefore, we estimated consumption-based emissions using input-output analysis and life cycle assessment to calculate the greenhouse gas emissions hidden in meat and dairy products in Hong Kong, a city predominately reliant on imports. We found that emissions solely from meat and dairy consumption were higher than the city’s total greenhouse gas emissions using conventional production-based calculation. This implies that government reports underestimate more than half of the emissions, as 62% of emissions are embodied in international trade. The discrepancy emphasizes the need of transitioning climate targets and policy to consumption-based accounting. Furthermore, we have shown that dietary change from a meat-heavy diet to a diet in accordance with governmental nutrition guidelines could achieve a 67% reduction in livestock-related emissions, allowing Hong Kong to achieve the Paris Agreement targets for 2030. Consequently, we concluded that consumption-based accounting for greenhouse gas emissions is crucial to target the areas where emissions reduction is realistically achievable, especially for import-reliant cities like Hong Kong.


Introduction
The standardized greenhouses gas (GHG) calculation framework provided by the Intergovernmental Panel on Climate Change (IPCC) captures the emissions occurring within country's geographical boarder and usually adopts a producer-oriented approach (e.g. IPCC 2006, Caro et al 2015, Druckman et al 2008, Bastianoni et al 2014. The national GHG estimates are the foundation in binding international agreements such as the Kyoto Protocol and the Paris Agreement from the 21st Conference of Parties (COP21; Rogelj et al 2016, UNFCC 2015, Kanemoto et al 2014. However, the production, or territorialbased accounting, ignores emissions embodied in international trade (e.g. Peters 2008, Lenzen et al 2004. In a globalized and highly-connected economic system, country's emissions are intimately linked with each other because international trade divides production and consumption in different geographical regions (Schmitz et al 2012, Su and Ang 2010, Wiebe et al 2012, Lenzen et al 2004. Therefore, the economic activities and consumption of one country have impacts beyond its geographical boundary. There is an increasing trend of emissions transfer via international trade and thus emissions are shifted from one country to another (Peters et al 2011). Neglecting international trade in the global GHG estimation lead to an incomplete understanding of emissions nationally and undermine the emissions reduction targets Figure 1. Comparison between production-based (black boxes) and consumption-based methods (red boxes) of emissions accounting for Hong Kong, which relies on import and a producer country (Steininger et al 2014).
(e.g. Davis and Caldeira 2010, Peters et al 2012, Weinzettel et al 2013, Peters and Hertwich 2008, Scott and Barrett 2015, Athanassiadis et al 2016. A number of studies have examined another type of GHG accounting, consumption-based accounting (figure 1), which allocates emissions to the final consumers (e.g. Barrett et al 2013, Peters et al 2011, Su and Ang 2010, Chen and Chen 2011, Caro et al 2014a, Su and Ang 2011, Gavrilova and Vilu 2012, Marin et al 2012, Caro et al 2015. It considers emissions embodied in local production and global trade, which are thought to better reflect the actual geographical distribution of emissions characteristic of international trade (e.g. Caro et al 2015). For instance, consumption-based accounting revealed that more than 30% of emissions from the UK and France were unaccounted for under the production-based accounting (Davis and Caldeira 2010). As such, it has been suggested that using production-based accounting to estimate emission from a city is 'insufficient and even misguiding' (Athanassiadis et al 2016).
The objective of this study is to evaluate the importance of using consumption-based accounting in cities with limited local production like Hong Kong. Food consumption, especially livestock products, have been identified as a significant contributor to anthropogenic GHG emissions, accounting for 7%-20% of the total emissions depending on the accounting method and system boundary (Steinfeld et al 2006, Rojas-Downing et al 2017. GHG emissions associated with animal production originate mainly from land use change (carbon dioxide), enteric fermentation (methane) and manure management (nitrous oxide) (FAO 2011). Caro et al (2014b) estimated that 32.8 Megatonnes CO 2equivalent (Mt CO 2 -e) were embodied in international traded livestock between 1990 and 2010. Here, we use a life cycle assessment (LCA) approach to highlight the contribution of livestock emissions on the city's consumption-based emissions and evaluate the mitigation potential of a low-meat diet. To assess the consumption-based GHG emissions in meat and dairy, a hybrid of input-output analysis and life cycle assessment was used, the former one focusing on the economic and the latter on the emissions in production process (Andrade et al 2018, Jones et al 2016, Peters 2008, Davis and Caldeira 2010, Su and Ang 2010, Wiebe et al 2012, Athanassiadis et al 2016. Hong Kong is an excellent case study to explore the complexity of livestock emissions and consumption emissions. Hong Kong is a net importer of goods and GHG emissions, ranked second worldwide with respect to the net import of GHG emissions per capita with 9.2 tonnes of CO 2 -equivalent per capita per year for a total of 64 Mt CO2-e imported in 2004 (Davis and Caldeira 2010). In contrast, Hong Kong's government reported total GHG emissions of 41.6 Mt CO 2 -e (5.7 tonnes CO2-e per capita) in 2015 using a production-based accounting method (Environment Bureau 2017). Under production-based accounting, the two largest contributors of emissions are electricity generation (67%) and transportation (18%). This suggests that the reported emissions from Hong Kong's government are considerably lower than those included from international trade.
Meat and dairy consumption in Hong Kong is the highest per capita in the world (500 g of meat and 281 g of dairy products per day). The city relies heavily on imports, with less than 1% of consumption sourced from local livestock (Agriculture Fisheries and Conservation 2017, FAO 2017). Furthermore, both the environmental and health impacts of excessive meat and dairy consumption has raised increasing concerns in recent studies (Friel et al 2009, Godfray et al 2010, Bodirsky et al 2014. Using a consumptionbased accounting method, it was suggested that the European Union could reduce agricultural-related GHG emissions by 25%-40% from halving its meat, dairy and egg consumption (Westhoek et al 2014). Moreover, reducing the consumption of beef and mutton, ruminant meat, was found to be necessary for the European Union to achieve climate targets (Bryngelsson et al 2016). We hypothesize that a similar situation is true for Hong Kong.
Here, we highlight the importance of consumption emissions in an import-driven city by quantifying meat and dairy consumption emissions in Hong Kong using an input-output analysis and LCA. We recalculated GHG emissions and targets to achieve COP21 goals in Hong Kong using consumption-based emissions accounting. Dietary change scenarios were also modelled to estimate the emissions reduction potential of changes in meat consumption under a consumptionbased accounting system. This research can help to inform policy objectives to achieve COP21 goals.

Materials and methods
Meat-and dairy-based consumption emissions in Hong Kong were calculated using a combination of multiple data sets and analyses; (1) meat and dairy consumption was estimated from the Hong Kong government's trade data (Hong Kong Census and Statistics Department 2017), (2) embodied GHG emissions were calculated using detailed LCA analysis from the Food and Agriculture Organization of the United Nations (FAO), (3) input-output analysis was used to trace the emissions in international trade and (4) projected GHG emissions for 2016-2030 were obtained from the Hong Kong Environmental Protection Department (HKEPD).

Meat and dairy consumption data
Consumption data were extracted from the Hong Kong Merchandize Trade Statistic published by the Hong Kong government, which compiles trade data in Hong Kong and identifies the import, domestic export and re-export of goods (Hong Kong Census and Statistics Department 2017). Food intake and wastage are considered as consumption here. The monthly report is divided based on the type, origin, value and quantity of commodity, which follows Standard International Trade Classification Revision 4 (SITC Rev. 4). Meat products include fresh, chilled, processed meat, offal and related dairy products (Hong Kong Census and Statistics Department 2017). In total, 22 meat-related commodities were analyzed in this study. The commodities were aggregated into 8 major groups (beef, pork and poultry, sheep, egg, milk, cheese and yogurt). Domestic livestock supply was extracted from the Agriculture, Fisheries and Conservation Department (Agriculture Fisheries and Conservation 2017). The trade dataset identifies the location of production between Hong Kong and other countries, which was used to build an input-output analysis. The meat consumption was calculated by summing annual amount of imports and local livestock supply, and subtracting the domestic export and re-export for each type of meat. All meat consumption is expressed in kilograms (kg). The conversion of heads of live animal to kg were done according to the average weight of livestock (pork: 80 kg; cattle: 450 kg; poultry: 0.185 kg; USDA Foreign Agricultural Service 2015).

Greenhouse gas emissions using life cycle assessment
The Global Livestock Environmental Assessment Model (GLEAM) developed by the FAO was used to quantify GHG emissions. Emissions in the supply chains from land use change, livestock production, and processing and transportation to final retail are embodied in the exported product (Hristov et al 2013). GLEAM reports carbon emissions for different production systems and regions with respect to types of livestock . It is a holistic model assessing the bio-physical process on livestock production using LCA . The model captures the carbon dioxide (CO 2 ), methane (CH 4 ) and nitrous oxide (N 2 O) from all major emissions activities in each stage of the livestock supply chain, such as feed and animal production, processing and transport (figure 2). The dataset includes a range of livestock such as cattle, chicken, pigs, buffalos, goats and sheep. This model was previously used in the FAO report on ruminant, non-ruminant and dairy-sector emissions , Gerber et al 2010. GHG emissions are quantified according to the IPCC guidelines as CO2-e using 100 years global warming potential, where carbon dioxide represents one global warming potential, nitrous oxide 298 CO2-e and methane 34 CO2-e (IPCC 2006). Additionally, GLEAM estimates the emissions intensity per protein of meat products in different regions, which is expressed as kg of CO2-e per kg of protein (kg CO 2 -e ⋅ kg protein −1 ). Hong Kong's livestock consumption inventory was divided into the origin countries to consider the regional differences in production methods and energy used. In order to compare the GHG emissions between various meat and dairy products, edible protein was used as the standard unit for carbon intensity (Flachowsky and Kamphues 2012). The meat consumption was expressed as protein per kilogram of edible product, assuming 0.19 kg protein/kg edible product (de Vries and de Boer 2010).
Combining the trade balance and LCA, we calculated emissions in the bilateral trade for livestock products in Hong Kong. Therefore, the GHG emissions from meat and dairy consumption were estimated by summing the product of consumption, emissions intensity derived from GLEAM and protein mass: where E represents GHG emissions (kg CO 2 -e), M is the consumption of meat and dairy products in weight (kg), P is the protein factor (kg protein ⋅ kg −1 ) and I is the emissions intensity (kg CO2-e ⋅ kg protein −1 ), summed for each product i. Meat and dairy product To compare the relatively high meat consumption and the associated emissions in Hong Kong, we also present an input-output analysis and LCA for the United Kingdom (UK), which we consider as a typical western diet. Like those in Hong Kong, western dietary habits are also often characterized by an overconsumption of animal-based products (Röös et al 2016). Garnett (2007) quantified the livestock sectors' contribution on UK's consumption GHG emissions with a similar method, which allowed direct comparison with Hong Kong. It should be noted that no uncertainty on the input-output data is available and thus it is not possible to assess the accuracy of the calculated emission values. However, this is usually the case for such calculation as we followed a well-established, state-of-the-art, methodology (e.g. Wiedmann et al 2011, Caro et al 2015 and thus our results can be compared to similar study.

Projection of Hong Kong's emissions and new targets
Hong Kong's consumption-based emissions projection for 2016-2030 were calculated by summing the emissions from local production and trade (figure 1). The future trend was estimated using the population projection from 2016-2077 and constant per capita GHG emissions (Census and Statistics Department 2015, Davis and Caldeira 2010). By keeping the reduction target percentage for 2020 and 2030 (20%, 26% to 36% respectively), new targets were set in accordance with the Paris Agreement (Environmental Bureau 2017).

Dietary change scenario
Dietary change scenarios were designed to investigate the mitigation potential of dietary change on GHG emissions. Here, we focused on meat consumption (beef, pig, poultry and sheep). Four scenarios were assessed a; 1) 25%, 2) 50%, and 3) 80% reduction in the consumption of meat relative to 2016 consumption in addition to four) a healthy diet in compliance with the nutritional guidelines of the Hong Kong Department of Health. These guidelines recommend an average daily intake of 180 g for meat, fish, egg and alternatives, in addition to two portions of milk and alternatives for a normal adult (equivalent to 480 ml of milk). Following the current diet composition, we defined the healthy diet scenario as 72 g of beef, 72 g of pork, 36 g of poultry, 133 g of cheese, 187 ml of milk, 3 g of yogurt per day (Department of Health 2015). It is assumed that the reduction of livestock consumption will proportionally reduce the livestock production in the exporting country. These scenarios were only tested considering a consumptionbased accounting method as there are virtually no

Livestock consumption and embodied emissions
Total meat and dairy consumption in Hong Kong was approximately 2.3 billion kilograms in 2016 (992 g day −1 /capita), with pork and beef consumed the most (figure 3). Compared to the UK, Hong Kong consumed over nine times more beef and pork per person each day (Garnett 2007).
Using consumption-based accounting, we estimated that 57.5 Mt CO2-e per year (7.8 tonnes CO2-e per capita) can be attributed to Hong Kong's meat and dairy consumption. Beef (47 Mt CO 2 -e) produced the highest total GHG emissions followed by pig (5.8 Mt CO 2 -e) and poultry (2.5 Mt CO 2 -e) (figure 3). Since the carbon intensity per protein of beef is the highest (276 kg CO2-e per kg of protein supplementary material, figure S1 available at stacks.iop.org/ERL/13/064005/mmedia), it translates to much higher meat and dairy consumption-derived GHG emissions per year in Hong Kong compared to the UK.
Hong Kong's is an import-oriented city where the local production is minimal (Harris et al 2012). As mentioned earlier, less than 1% of the meat and dairy products were produced in Hong Kong (USDA Foreign Agricultural Service 2016, FAO 2017, Agriculture Fisheries and Conservation (AFCD) 2017). In total, Hong Kong imported livestock products from 97 countries with the top three importers being Brazil, the US and China (supplementary material, figure S2). By calculating the embodied emissions of such imports, 57.5 Mt CO2-e were outsourced to other countries, and are therefore missing from the Hong Kong's emissions inventory. From our estimation above, meat and dairy consumption (57.5 Mt CO2-e) alone exceeds production emissions   (table 1).

Projection of Hong Kong emissions for 2016-2030
Here It is important to note that there is a difference between the absolute quantity of emissions if we consider only the production or if both local production and consumption are taken into account ( figure 4). Therefore, we set new consumptionbased reduction targets for 2020 and 2030 as 82.4 Mt CO2-e and 65.9 Mt CO2-e and 20.6 Mt CO2-e respectively using 2005 as base year, which are two times higher than the production-based targets of 35.8 Mt CO2-e by 2020 and 28.6 Mt CO2-e by 2030 (table 1). Therefore, different emissions accounting may lead to different emission targets and policy. The new targets were applied to examine the mitigation potential of dietary change.

Emission reduction scenario based on dietary change
The mitigation potential of dietary change scenarios was examined based on current diet and reduction scenarios. Diets with 25%, 50% and 80% reduction of meat consumption could reduce 38% to 76% of livestock-related GHG emissions in Hong Kong using consumption-based accounting ( figure  5). In addition, our results show that a healthy diet saves approximately 40 Mt CO 2 -e, which represents a reduction of around 67% compared to the current diet. The healthy diet scenario would require meat reduction per capita from 664 g d −1 to 180 g d −1 (supplementary material, figure S4). Therefore, dietary changes can manifest as considerable reductions in Hong Kong's GHG emissions. When combined with the projected consumption emissions that include population growth and a potential progressive shift towards healthy diet, we project a 43% decrease in total consumption-based emissions in Hong Kong from 2016-2030 (from 109 Mt CO2-e to 62 Mt CO 2 -e; figure  6). If Hong Kong shifts to the healthy diet scenario by 2030, we forecast that consumption emissions in Hong Kong will meet the 2030 target ( figure 6).

A meat-heavy diet and its impact on consumption-based emissions in Hong Kong
We estimated that an average Hong Kong inhabitant consumed about 664 g and 258 g of meat and dairy products daily in 2016 (figure 3), which is consistent with previous reports (FAO 2017). Hong Kong's relatively high meat consumption can partially explain its large consumption-based emissions. The estimated emissions from livestock consumption (57 Mt CO 2 -e) are approximately 50% of total consumption emissions in Hong Kong (103 Mt CO 2 -e) extracted from Davis and Caldeira (2010). While this number may seem unusually large, it is important to note that the GLEAM model used in this study includes a detailed life cycle assessment and the emissions associated with land use change (figure 2), as opposed to Davis and Caldeira (2010). This means that numerous sources of emissions listed under 'meat and dairy' consumption in our study were listed under other rubrics (e.g. transport, land use, production of goods, etc.) in Davis and Caldeira (2010). Therefore, we believe that the total amount should still be similar, as our estimate simply enlarges the boundary of what is considered 'meat and dairy' consumption. Moreover, we account for both food intake and wastage as consumption, which may result in a higher consumption than other self-reported meal intake studies. Our results highlight the importance of consumption-based emissions accounting in Hong Kong, as the population of Hong Kong consumes an inordinate amount of mostly-imported meat.

Potential carbon leakage due to meat consumption
Total consumption-based GHG emissions in Hong Kong were estimated to be 109 Mt CO2-e in 2016, which is more than twice the emissions reported by the government (41.6 Mt CO2-e) using productionbased accounting ( figure 6). This demonstrates that the government underestimates GHG emissions by 62% because it fails to recognize the emissions associated in trade. This large difference is consistent with Energy and Climate Change Committee (2012), in which it was estimated that Hong Kong's actual GHG emissions should be approximatively 80% greater than production-based GHG emissions. Similar results were also reported in Hertwich and Peters (2009), whereby estimates of Hong Kong's consumption-based emissions in 2001 were 4 times higher than the government-reported emissions. This large discrepancy shows that the emissions inventory derived from production-based accounting does not provide a complete analysis in cities with considerable emissions associated with trade. The large gap between consumption-based and production-based emissions is due to the high carbon intensity of GDP and trade imbalances (Davis and Caldeira 2010). Even within China, cities with higher GDP per capita, such as Shanghai, generally have higher consumption-based emissions than production-based emissions compared to less developed cities, such as Shijiazhuang (Mi et al 2016). Similar patterns were also observed in Southeast Asia, where both Hong Kong and Singapore behave as net importers with limited local resources and high GDPs. Therefore, most of the carbon emissions in Hong Kong and Singapore are captured in international trade (Davis and Caldeira 2010). By tracing the emissions from meat consumption, we demonstrate that Hong Kong outsources considerable emissions to producer countries, yet these emissions are not allocated to Hong Kong. Therefore, a weak carbon leakage may occur because of the increasing demand driven by population growth in Hong Kong. Kanemoto et al (2014) observed that many developed countries such as Japan and US meet the Kyoto Protocol targets by outsourcing their emissions. The missing emissions embodied in trade will potentially undermine the climate change targets and policy, highlighting the need for more effective emissions accounting.
Currently, countries follow the IPCC guidelines to report their emissions inventory (Dahal andNiemelä 2017, European Environment Agency 2013). Hong Kong also follows the IPCC guidelines as its emissions are included within China's inventory (Harris et al 2012). Territorial-based inventories neither allocate GHG emissions embodied in import beyond city borders to the consumer countries nor subtract the GHG emissions from export in the producer country (Dahal and Niemelä 2017). In order to navigate the existing caveats of the territorial-based inventory and estimate the emissions embodied in trade, Hong Kong should use a consumption-based accounting approach to calculate its GHG inventory (e.g. Lenzen et al 2012, Steckel et al 2010, Caro et al 2014a, Gavrilova and Vilu 2012, Marin et al 2012, Davis and Caldeira 2010.

The mitigation potential of dietary change on consumption emissions
In the scope of the COP21 agreements, Hong Kong committed to China's target of reducing 20% on its absolute carbon emissions by 2020 and 26%-36% by 2030, relative to 2005 levels (Environmental Bureau 2017). In 2017, the government released Hong Kong's Climate Action Plan 2030+, which proposed policies in reducing coal-fire electricity and increasing energy efficiency. The report focused on the emissions targets and estimated a decline in Hong Kong's GHG emissions (figure 6) (Environmental Bureau 2017). However, when the emissions in trade are taken into account, our results demonstrate that Hong Kong no longer achieves the target despite government's mitigation efforts ( figure 6). Therefore, the existing emissions inventory could mislead policy-makers.
Therefore, we examined the potential of using consumption-based targets and dietary change to meet Hong Kong's COP21 objective. By addressing the consumption-based targets, we formulated new targets for 2020 and 2030 to 82.4 Mt CO 2 -e and 65.9 Mt CO 2 -e. At the same time, we applied dietary change scenarios to estimate the mitigation potential of a reduction in meat consumption. In line with other studies, we found that decreasing meat consumption can manifest in extensive GHG emissions reductions (De Boer et al 2014, Hallström et al 2014, Tilman and Clark 2014, Springmann et al 2016. Our results demonstrated that following the protein intake recommended by the government (healthy diet) can reduce livestock-related emissions by 67% (figure 6). By comparing different scenarios, adopting government-recommended nutritional guidelines is likely the best option to optimize both environmental and human health. Hence, increasing awareness of consumer choices and dietary habits on the environment is very important as such personal decisions may affect the producer country and even global climate change. In addition, such dietary changes will also yield health benefits for the consumer (Friel et al 2009). High meat consumption can notably increases the risk of obesity and cardiovascular disease (Micha et al 2010, Koeth et al 2013, WHO 2003, and in the case of red meat, the risk of colorectal cancer (WCRF/AICR 2007, O'Keefe 2016).

Conclusions
Using Hong Kong as a case study, this research demonstrates the differences between consumption and production-based emissions accounting. While the existing production-based accounting framework provides technical and comparable results, it does not capture the emissions embodied in trade, especially in a city with a high dependency on import. Given the large gap between conventional productionbased emissions accounting and consumption-based emissions accounting, future work should focus on building a comprehensive consumption-based inventory for Hong Kong and other import-dependent cities such as Singapore. As the Paris agreement leaves flexibility for domestic-driven policies to achieve global climate goals, it is worth discussing the possibility to complement the existing GHG reporting methodology to better reflect the contribution of GHG emissions in import-reliant large cities. Moreover, our analysis indicates that Hong Kong's trade-related emissions are partially driven by the demand of consumers for a heavy-meat diet. Therefore, the role of consumer behavior in reducing global GHG emissions is important to consider. By adopting new, climate-friendly dietary habits, Hong Kong consumers can play a nontrivial role in emissions reductions. Incentivizing the public to participate in the global effort to reduce GHG emissions may prove to be key in addressing climate change and, in the case of Hong Kong, improving public health by reducing meat consumption.