Reducing global food loss and waste could improve air quality and lower the risk of premature mortality

While the global food system substantially contributes to environmental degradation and climate change, significant amounts of lost or wasted foods along the food supply chain actively contribute to global air pollution and related health risks. In this study, we use an environmentally-extended input–output model to quantify air pollution embedded in global food loss and waste (FLW) and investigate how FLW reduction policies can mitigate air pollution linked to food consumption, decreasing associated premature mortality risks across global regions. While estimating a positive impact of FLW reduction policies on decreasing air pollution levels (from −1.5% of SO2 emissions to −10.2% of NH3 emissions) and mortality reductions (over 67 000 lives worldwide) our findings highlight that rebound effects, wherein a reallocation of consumption from food to non-food commodities, decrease health and environmental benefits by over three quarters (compared to the case with no rebound). Such rebound effects can be substantially mitigated when final consumption shifts towards less pollution-intensive products, such as service activities, rather than conforming to the current composition of non-food consumption. Our results suggest that FLW-related policies would benefit from complementary measures that incentivize sustainable non-food consumption to effectively foster the transition towards a healthier and more sustainable planet.


Introduction
The global food system significantly contributes to environmental degradation and climate change (IPCC 2019, FAO et al 2021), accounting for about 30% of global greenhouse gas (GHG) emissions (Poore and Nemecek 2018), as well as a substantial share of air pollution-between 10% for the case of sulfur dioxide (SO 2 ) and up to 90% of ammonia (NH 3 ) emissions (Crippa et al 2022).While food is essential for a steadily growing world population, air pollution is responsible for 4.2 million premature deaths per year worldwide (WHO 2022) and increasing air pollution embedded in food consumption represents a major environmental mortality risk factor (GBD 2020, Murray et al 2020).
Approximately one-third of food is lost or wasted along the food supply chain (FSC) (FAO 2019), accounting for 6%-10% of global GHG emissions (Poore and Nemecek 2018, UNEP 2021).While indirect emissions result from the production, processing, and transportation of food that is ultimately lost or discarded, direct emissions are driven by the disposal of food loss and waste (FLW).Landfilled food waste is a major contributor to global warming (IPCC 2013), accounting for 16% of global methane emissions (Shindell et al 2020).Air pollution has a direct impact on the productivity of agricultural systems with the potential to reduce crop yields and impair the nutritional quality of food (Lelieveld et al 2015, Domingo et al 2021), inducing farmers to discard produce (Lipinski et al 2013, Giannadaki et al 2018) or sell their crops at lower prices.Additionally, it can impact the storage and transportation of food, as contaminated air can infiltrate storage facilities, transportation vehicles, and packaging materials.At the consumer level, air pollutants affect the shelf life of food, accelerating spoilage and intensifying waste generation (UNEP 2021).Adverse implications of rising air pollution on ecosystems and biodiversity have also been widely recognized in the literature (Lovett et al 2009, Paoletti et al 2010).Short-and long-term exposures to landfilled FLW pollution have also been linked with premature mortality and reduced life expectancy (Kampa andCastanas 2008, Siddiqua et al 2022).
The contribution of FLW to global GHG emissions has been assessed in earlier studies (Katajajuuri et al 2014, Porter et al 2016, Xue et al 2017, Poore and Nemecek 2018) but limited information is available on air pollutants embedded in FLW along global FSC and their impact on premature mortality agricultural production and post-harvest handling and storage represent a global hotspot of food loss generation especially in low-income regions (Kaza et al 2018, FAO 2019, UNEP 2021).As crop and livestock production globally contribute around 75% of the nitrous oxide (N 2 O) (FAO 2020, Tubiello et al 2021), and substantial volumes of particulate matter (PM) (Madden et al 2008), decreasing farmlevel losses could reduce the anthropogenic emissions from agriculture easing the burden of pollutioninduced diseases (GBD 2020, Murray et al 2020).It could further reduce critical levels of ammonia (NH 3 ) emissions (Lassaletta et al 2016, Giannadaki et al 2018) often linked to chronic respiratory illnesses and premature mortality (Wyer et al 2022).Tackling FLW during transportation, processing, and retailing, could reduce carbon monoxide (CO) emissions linked to these stages of the FSC (Tubiello et al 2021).A reduction in food waste, particularly in highincome regions, could decrease emissions of ammonia, sulphides, and carbon monoxide (Kaza et al 2018, UNEP 2021) alleviating impacts on climate change and health-related issues (Shindell et al 2020).
Implementing FLW reduction policies may simultaneously have a positive impact on food availability (UNEP 2021), decreasing average food demand and lowering average food prices (Rutten 2013).While this is crucial for a global food security, it has the potential to decrease consumer expenditures on food simultaneously boosting the consumption of various We build on this FLW-extended MRIO to quantify air pollutants embedded in FLW along global FSC, successively testing how potential FLW reduction policies may mitigate air pollution and related health risks.First, we provide an overview of evolving trends of air pollutants embodied in FLW across a 2004-2014 timeframe outlining drivers and hotspots across countries and stages of the FSC.Following, we explore the impact of the UN-SDG12.3target of 50% cut in FLW, investigating the variation of air pollutants embedded in the final consumption of food under three alternative final demand patterns.The resulting changes in air pollution across scenarios are further linked to the global atmospheric source-receptor model allowing to assess health-related co-benefits across global regions.
We find FLW reduction policies decrease air pollution, but the overall magnitude of the achieved health and environmental benefits substantially varies across assumed changes in consumption patterns.As rebounds are lower when consumption shifts towards less pollution-intensive products, our findings highlight the benefits of FLW-related policies for fostering a healthier and more sustainable global food system but stress on the need of complementary measures to direct consumer choices towards more sustainable non-food products.

Methods
Estimates of FLW across global supply chains are obtained from Gatto and Chepeliev (2023) and are defined as FLW_SHR r,g c,d (1) where r represents the country where losses or waste are generated, g represents a stage of the FSC at which the losses/waste are occurring, c represents a primary food commodity being lost or wasted and d represents a dummy variable that has a value of '1' when a commodity enters the manufacturing stage (i.e. is finally consumed as a processed food product) and a value of '0' when it does not (i.e. is finally consumed as a fresh food product).
To quantify physical (Tones) food supply we retrieve information from the GTAP-FBS database (Chepeliev 2022)3 defining a food supply as where c represents a primary food commodity flowing into primary food, processed food or food services which provides information on the primary composition of non-primary foods, f represents the final food product (primary, processed or from food services) consumed by households, r and s represent regional source (r) and destination (s) of the food supply (if r = s, food is produced and consumed domestically within a country).The computation of coefficient ( 2) is available in (Chepeliev 2022) and is briefly illustrated in supplementary information.Coefficient (2) represents the food matrix derived from our MRIO framework to which FLW shares are applied.
We multiply the FLW coefficient (1) by the physical food flows represented by coefficient (2), tracing FLW along global food supply chains as following FLW_FSC We link FLW estimates to air pollutants derived from the EDGAR Version 5.0 database (Crippa et al 2020) following Chepeliev (2021).First, we map this data to food flows from farm to fork devising air pollutants embedded in food and non-food production and consumption across global supply chains.From this, we quantify air pollution embedded into FLW along five stages of the FSC (namely Agricultural Production, Post-Harvest Handling & Storage, Manufacturing, Distribution & Retail, and Consumption) for nine air pollutants (Chepeliev 2021).The full database providing country-and commodity-specific air pollutants embedded in FLW along the FSC is available in the supplementary information.
We use the FLW-pollution-extended MRIO, to assess the impact of a 50% reduction in global FLW based on the SDG12.3target (U.N. 2019).We decrease the FLW shares by 50% to simulate a FLW-reduction policy, quantifying the associated decrease in food supply assuming a decrease in FLW shares entails a more efficient food production, supply, storage, and consumption.As net food consumption does not change, we assume total gross food demand proportionally decreases, reducing food-related expenditures.We find that 50% reduction in FLW decreases average global food demand by 13.0%.We implement uniform FLW reduction policies across scenarios, assuming expenditures on food decrease by the same amount in each scenario.However, different trends of demand reallocation toward non-food items are assumed across scenarios (table 1).In the 'No Rebound Effect' (NO_REBOUND) scenario, the decrease in food expenditures is assumed to generate no increase in expenditures on non-food products, assuming nonfood demand remains constant, hence no rebound is generated.This scenario corresponds to the potential FLW reduction policy implementation in the partial equilibrium-type models, where interactions and expenditures outside the agri-food sectors are not explicitly represented.In contrast, in the 'Status Quo' (STATUS_QUO) scenario the reduction in food expenditures leads to an increase in the demand for non-food products following the current patterns of consumption outside the food sector.Finally, in the 'Environmental Awareness' (ENV_AWARE) scenario the decrease in food expenditures is reallocated toward demand for products that have the lowest pollution intensity, assuming that consumers decide to shift their preferences towards 'pollution-friendly' goods and services.In addition to lower emission intensity, increasing expenditures on such service activities, including education, can benefit human capital development and benefit economic growth in the long-run (Gyimah-Brempong et al 2006, Haldar andMallik 2010).
We rely on the TM5-FASST model (Van Dingenen et al 2018) to address the impact of evolving air pollutants on human health and premature mortality risks under the FLW reduction policies.We estimate pollutant-related premature mortality rates across global regions, investigating PM2.5-related diseases and respiratory O 3 exposure mortality.Based on the .Mortality associated with PM2.5 is calculated, using the integrated exposureresponse model adopted in the Global Burden of Disease assessment (Stanaway et al 2018), as the number of annual premature mortalities from six causes of death: chronic obstructive pulmonary disease (COPD), lung cancer (LC), lower respiratory airway infections, type 2 diabetes mellitus, ischemic heart disease (IHD), and stroke.To monetize the health co-benefits of improved air quality we rely on the value of social life approach outlined in Markandya et al (2018).A more detailed description of our methodology and estimates of total air pollutants generated across scenarios are available in the supplementary information.

Global hotspots and trends of air pollutants embedded in food loss and waste
Air pollutants embedded in the global FLW range from 2.2% for the case of SO 2 emissions to 18.9% of NH 3 pollution worldwide.The largest volumes of the FLW-embedded pollutants across global supply chains correspond to carbon monoxide (CO-19 802 Gigagrams (Gg) or 5.0% of total emissionspanel A of figure 1), and ammonia (NH 3 -8708 Gg or 18.9% of total emissions-panel B of figure 1), followed by organic carbon (OC-9.3%),particulate matter (PM10-5.4% and PM2.5-5.2%-panelC of figure 1), nitrogen oxides (NO X -4.2%-panel D of figure 1), and sulfur dioxide (SO 2 -2.2%).
Agricultural Production and Consumption represent global hotspots of air pollutants, accounting for 53.8% (16 970.4 Gigagrams (Gg)) of air pollutants associated with discarded or wasted food, and for 1.1% (SO 2 ) to 11.2% (NH 3 ) of the total global pollutants emitted.Key pollutants at these stages are CO and NH 3 , originating from farm operations, organic matter decomposition, and landfill activities.
Across lower income regions such as SEA and Sub-Saharan Africa (SSA), the majority of air pollutants embedded in FLW is generated at the farm-level stages of the supply chain, accounting for an average of 4.3%-8.9% of total air polluting emissions.Differently, in high income countries largest FLW-embedded pollution levels are found at consumer stage and represent an average 2.5%-5.2% of total air pollution generated in these regions.
Between 2004 and 2014, air pollutants embedded in FLW increased primarily in mid-and lowincome regions while in general decreased in highincome countries (figure 2).In SSA, main drivers are found in increasing per-capita gross domestic product (GDP) and population.While these factors have similar effects in SEA, an exception is represented by China and high-income Asian (HI-ASIA) countries where a stronger decrease in generated air pollutants per unit of FLW (in light blue in figure 2) and in food consumption per unit of GDP (in gray in figure 2), resulted in a contained increase or in decreasing amounts of FLW-embedded pollutants.Across global regions, air pollution from services decreased (−2.7%) while increasing from manufacturing (3.6%).As GDP growth was primarily driven by the expansion in service sectors with relatively lower pollution intensity, this structural shift channel had an important contribution to the reduction in pollution-intensity of the final consumption.Parallelly, the changing composition of consumed foods across high-income regions, such as Europe, USA and North America and Oceania (NAMO), has additionally resulted in lower air pollutants embedded in FLW.(1997).A list of countries composing each macro-region reported in the figure is available in the supplementary information.

The impact of FLW reduction policies on air pollution
Reducing FLW by 50% can exert significant benefits on global air pollution but such benefits substantially depend on the changing consumer demand4 .In our 'No Rebound Effect' (NO_REBOUND) scenario, the observed decrease in air pollution is highest compared to other scenarios in which non-food demand is assumed to increase.In NO_REBOUND, a 50% reduction in global FLW decreases CO (2.8% or 11 216.5 Gigagrams (Gg)), NO X (1.7% or 1985.7 Gg) and PM2.5 (3.1% or 870.9 Gg) pollution, globally.When FLW reduction policies are applied worldwide, the most substantial reductions in pollutants are observed in China, Brazil, or SEA (Panel A-figure 3).Production-related pollution decreases mainly in large net food exporting regions such as Brazil, Latin America & Caribbean (LAC), NAMO and Indonesia.In these regions, the largest reduction in pollution is achieved when FLW is reduced globally (average −4.4% or 656.6 Gg in CO emissions; average −4.8% or 46.1 Gg in PM2.5 emissions) but also when policies are applied only domestically (Panel A-figure 3).
In our 'Status Quo' (STATUS_QUO) scenario, the FLW reduction policy proportionally increases demand for non-food products, maintaining current shares of non-food products in total non-food consumption across regions.As non-food products may have higher pollution intensities compared to food products, a global FLW reduction decreases CO (0.3% or 1256 Gg) and PM2.5 (1.4% or 403.3 Gg) pollution but increases NO X (0.2% or 262.9 Gg) pollution, generating heterogeneous effects across regions.A global 50% reduction in FLW increases CO and NO X pollution in China (an average of 0.4%), Nigeria (average 9.3%), HI-ASIA (average 0.8%), and SSA (average 1.0%), as shifting consumer demand specifically towards transportation and construction sectors increases production, hence pollution, in these regions.
Similar effects are observed when policies are applied to a single country/region (Panel B of figure 3).Region-specific policies primarily increase NO X pollution, impacting large pollution-intensive non-food producers such as China or Nigeria.An exception is represented by Brazil and Indonesia, where the reduction in pollution associated with a lower food demand dominates the increase in nonfood pollution notwithstanding whether policies are applied only domestically or worldwide.In relative terms, decreasing FLW in low-income regions such as Nigeria and SSA, exerts a strong increase in domestic (average 7.2% or 1274.7 Gg of CO; average 5.5% or 27.7 Gg of NO x ) and global pollution (average 6.8% or 1262.1 Gg of CO; average 4.4% or 30.8Gg of NO x ).As changes with respect to our baseline are reported in percentage (%), absolute variations (gigagrams) for the illustrated pollutants as well as for other pollutants across the investigated scenarios are reported in the supplementary information.
A Similar pattern is also observed when the FLW reduction policy is applied only in China (0.3% or 280.9 Gg of CO; 0.4% or 103.6 Gg of NO x ; 0.1% or 9.7 Gg of PM2.5) as non-food emission (principally from manufacturing and construction sectors) overrun the decrease in pollution from the food system.All three regions (Nigeria, SSA and China) have relatively high emission-intensity of energy supply as the power generation mix is dominated by fossil fuels (EIA 2023).
The 'Environmental Awareness' (ENV_AWARE) scenario presents the lowest pollution rates.Here, the decrease in air pollution (−0.7% or 29 997.5 Gg of CO; −2.6% or 724.3 Gg of PM2.5; NO X−1 .1% or 1183.1 Gg of NO X ) is almost twice the size of the decrease under STATUS_QUO scenario (Panels B and C of figure 3) showing the benefit of controlling for rebound effect when devising FLW reduction policies.
Major reductions in pollution are observed for large food exporting regions such as Brazil, Indonesia, and LAC (average −2.8% or 656.1 Gg of CO; average −1.2% or 59.3 Gg of NO X ; average −3.4% or 48.8 Gg of PM2.5 principally linked to rice and horticulture production) while CO and NO X pollution grow in Nigeria (17.5% or 3275 Gg of CO; 8.3% or 41.4 Gg of NO X principally linked to transportation and construction sectors and other services).In relation to the STATUS_QUO scenario, region-specific FLW reduction policies generate reverse effects on global air pollution.In China a FLW reduction of 50% in the ENV_AWARE scenario decreases global CO pollution by 1000.7 Gg compared to STATUS_QUO scenario.Differently, CO pollution is higher under the ENV_AWARE scenario for Nigeria and SSA, as CO emission intensity of service sectors, which experience increasing consumption, is relatively high.

Air pollution-related health and mortality impacts from FLW reductions
Evolving trends of air pollution exert a direct impact on pollution-related diseases influencing premature mortality risks.In the NO_REBOUND scenario, the in air pollution reduces premature mortality risks across countries, saving a total of 67 325 lives (Panel A-figure 4).Mortality risks mainly decrease from lower respiratory infections (9625 saved lives), primarily in low-income regions, and from COPD (11 324 saved lives), primarily in highincome regions.Major benefits in terms of saved lives are observed in China (14 510 saved lives) and India (12 225 saved lives).Relatively high benefits are additionally observed in the European Union (EU27) and SEA where reductions in air pollution save a total of 9035 and 6973 lives, respectively.The number of lives saved in the NO_REBOUND scenario has direct global economic benefit ranging from 55.3 to 138.3 billion USD (0.07%-0.17% of global GDP), with highest benefits in terms of savings observed in China (from 7.1 to 17.7 billion USD per year equivalent to 0.06%-0.17% of GDP), EU27 (from 15.0 to 37.5 billion USD or 0.09%-0.24% of GDP), and the United States (USA) (from 12.4 to 31.0 billion USD or 0.07%-0.17% of GDP) (Panel A-figure 5).
In the STATUS_QUO scenario the decrease in premature mortality risks is around 55.3% lower than in the NO_REBOUND scenario, with a total of 30 032 lives saved (Panel B-figure 4).While major benefits remain observed in China (5505 saved lives) and EU27 (6521 saved lives), changing consumption trends in low-income regions such as Nigeria and SSA result in an increase in pollution-related mortality risk with additional 1455 and 1438 lives being exposed to premature mortality, respectively.In these regions growing pollution linked to non-food consumption primarily increases the risk of lower respiratory infection diseases, exposing a total of 1903 people to premature mortality risk per year.This directly translates to higher health-related costs (Panel B-figure 5) which compared to NO REBOUND, increase from 20 to 50 billion USD per year (0.02%-0.06% of GDP).Globally, while across diseases mortality risks decline, an exception is represented by respiratory O 3 exposure mortality risk which increases principally in mid-low-income regions, exposing an additional 496 people to premature mortality per year.
Finally, changing consumer choices towards less polluting products in the ENV_AWARE scenario results in an increase of 62.9% of saved lives compared to the STATUS_QUO scenario, with a total of 48 931 lives saved globally (Panel C-figure 4).In absolute terms, exposure to premature mortality from COPDs and LC decreases mainly in China (4322 saved lives) and EU27 (2582 saved lives), while the risk of lower respiratory infection diseases decreases primarily in India (1797 saved lives) and SEA (1502 saved lives).However, mortality risks increase in Nigeria and SSA by a 42.2% compared to STATUS_QUO, exposing an additional 4116 lives.In comparison to the STATUS_QUO scenario, health-related costs linked to the rise in premature mortality risk in these regions increase by a 42.2%-45.0%,and amount to an additional cost of 11.3-28.2 billion USD per year (average 0.01%-0.07% of GDP) (Panel C-figure 5).On the other hand, the high absolute reduction in pollution results in an increase in health-related co-benefits in China (from 5.4 to 13.6 billion USD or 0.05%-0.13% of GDP), in EU27 (from 13.4 to 33.6 billion USD or 0.08%-0.21% of GDP) and the USA (from 11.1 to 27.8 billion USD or 0.06%-0.16% of GDP).

Discussion
In this study, we explore the relationship between FLW and air pollution, providing a quantification of pollutants embedded in FLW and assessing the potential impact of FLW reduction policies on air quality and associated premature mortality risks.We enhance a FLW-extended MRIO framework by estimating air pollutants embedded in FLW, identifying global hotspots, trends, and driving factors across countries and stages of the global FSC.Using this framework we analyze three scenarios with different final demand patterns.
While available waste-extended input-output models (Kagawa et al 2004, Lenzen and Reynolds 2014, Read et al 2020, Towa et al 2020) primarily concentrate on waste treatment strategies, our study expands current input-output analyses, offering a comprehensive assessment of the impact of FLW changes on air pollution.As studies on FLWrelated emissions limit their scope to GHGs (Porter et al 2016, Poore and Nemecek 2018), we provide an innovative quantification of air pollutants embedded in FLW to assist future policy designs.
Global FLW-embedded air pollutants account for around 5.2% of total air pollutants.In the absence of prior studies specifically addressing the quantification of air pollutants associated with global FLW, we compare food system emissions from the literature with our food-related emissions.In our study air pollution from the food system represents 22.5% of total pollution, in line with the 10%-35% range provided by previous studies (Shukla 2019, Crippa et al 2021, 2022, Rosenzweig et al 2021, Tubiello et al 2021).For each pollutant, we meet the shares of food-related emissions in total emissions reported in the literature (Balasubramanian et al 2021, Chepeliev 2021, Crippa et al 2022).In particular, we find that PM2.5 accounts for a 25.2% (20%-35%)5 of total emissions, NH 3 for a 76.6% (72%-87%), SO 2 for a 10.4% (9%-12%), NO X for a 12.8% (13%-20%), NMVOC for a 19.0%(16%-19%), and BC for a 22.5% (20%-28%).
Global hotspots of FLW-embedded air pollutants are found in Agricultural Production and Consumption stage.Largest levels of FLW-embedded air pollution are found East Asia, notably in China, India, and Indonesia.However, as shares of FLWembedded pollutants in total pollution are relatively low in these regions due to a higher reliance on pollution-intensive energy supply options, FLW reduction policies may provide lower benefits.Differently, in regions like Brazil, Lao, Vietnam, and Bangladesh, shares of FLW-embedded air pollution in total pollution are relatively high, and FLW policies may provide higher benefits on reducing overall pollution levels.
FLW reductions can decrease air pollution but changes in consumer demand play a key role.In the absence of a rebound effect (NO_REBOUND scenario), a 50% decrease in FLW reduces global air pollution by a 3% (25 050 Gg).This is still far from air pollution targets (WHO 2021) but illustrates the interlinked benefit of decreasing FLW on air quality.Moreover, positive effects are found on premature mortality risks which decrease (67 325 lives saved globally), especially in China and India where pollution-related mortality rates are highest (WHO 2021).However, the presence of potential rebound effects associated with FLW policies has been widely discussed in the earlier literature (Druckman et al 2011, Martinez-Sanchez et al 2016, Salemdeeb et al 2017, Reynolds et al 2019, Read et al 2020) and changes in demand must be considered once policies targeting FLW are introduced.
We find rebound effects decrease health and environmental benefits linked to FLW reduction policies.Regions such as China, Nigeria, HI-ASIA, and SSA experience increased pollution levels due to the shift in consumer demand towards nonfood products (STATUS_QUO scenario) such as transportation, construction, and manufacturing, which have higher pollution rates compared to food products.This leads to an increase in pollutionrelated mortality risk and health-related costs, primarily elevating the risk of lower respiratory infection diseases.Differently, in Indonesia or Brazil, a decrease in food demand and production results in a lower level of total pollution as primary food production sectors represent the main drivers of air pollution.Shifting consumption towards 'cleaner' sectors (ENV_AWARE scenario) decreases pollution compared to the STATUS_QUO scenario.Major reductions are observed in large food exporting regions such as Brazil, Indonesia, and LAC.However, Nigeria and SSA exhibit higher levels of CO pollution due to the relatively higher emission intensity associated with the domestic production of services, which are considered less polluting at a global scale.
The findings of our study have important social implications.In low-income regions, where FLW reduction is necessary to increase food security, policies may lead to increased pollution-related mortality risks, showing the need for parallel interventions to direct consumer choices or reduce production-side emissions.This is additionally highlighted by the increase in health-related costs associated with the rise in premature mortality risk and underscores the importance of considering the economic feedbacks of FLW policies on final consumers.The provided estimation of the economic benefits linked to decreased pollution-related mortality risks can contribute to the reduction in FLW policy costs, a channel often overlooked in earlier studies but fundamental to incentivize regional interventions.As economic benefits are evaluated on an annual basis and are enduring, it is essential to acknowledge that investments in FLW reduction policies can yield returns not only during the policy implementation period but also in the long-run.Nonetheless, it is crucial to parallelly promote sustainable production and consumption practices, encourage the use of clean technologies, and invest in pollution control measures in regions where the shift towards non-food products contributes to increased pollution.
It is essential to acknowledge the broader context of causal impact pathways that are not fully accounted for in our analysis.While we highlight the positive spillover effects of decreasing FLW on reducing air pollution levels and premature mortality risks, it is important to recognize that other pathways may play a more significant role in shaping health outcomes.Here, further research that explores the relative importance of these dynamics and their interconnectedness, may provide a more comprehensive understanding of the diverse set of factors influencing health and mortality risks, as well as broader environmental dimensions.While our study contributes valuable insights into the FLW-air pollution-health nexus, it is only a part of a larger puzzle that warrants further exploration and investigation.
As always, our study is subject to limitations.The use of an input-output framework does not allow to fully grasp the impact of policies on the income side of final consumers, omitting changing preferences linked to income shifts.As impacts on wages could affect purchasing power of consumers across regions, demand trends could differ across sectors, influencing total pollution levels.Additionally, as our analysis revolves around 2014, further research could expand our analysis with the use of a dynamic computable general equilibrium or integrated assessment models which allow to investigate impacts of FLW reduction policies on incomes while casting future trends of air pollution.With regards to the applied uniform 50% FLW reduction across all supply chain stages, it is crucial to acknowledge the inherent limitations of such an approach.While the uniform reduction serves as a simplifying assumption for the purpose of our analysis, we recognize that the real-world FLW reduction strategies would likely involve targeted interventions at specific stages of the supply chain.In high-income countries, most FLW occurs at household and retail levels, and interventions should primarily focus on reduction across these stages.Similarly, in low-income countries the majority of FLW occurs at the farm-level and FLW policies should primarily target production stages, as further decreasing household food waste in food insecure regions may not be realistic and/or economically feasible.Nonetheless, in the context of this study, the supply chain location in which the FLW reduction occurs is not a major determining factor for grasping the synergies between FLW policies and air pollution impacts.For this, our decision to adopt a uniform reduction in FLW across stages is motivated by the need for an illustrative reduction scenario to assess the overall impact on air pollution.However, we caution that this approach may not fully capture the nuanced and context-dependent nature of FLW reduction strategies.Future research would benefit from an assessment of a more refined, stage-specific interventions providing a more realistic representation of how FLW reduction policies might unfold across countries and food commodities.
Within the interplay between FLW reduction policies, air pollution, and associated mortality risks, it is imperative to highlight the limitations of the behavioral assumptions underpinning our scenario analysis.The assumption that efforts to reduce food waste have no consequential impact on food demand simplifies our assessment relative to the complexity of real-world behavioral responses.Oversimplification of the demand response to FLW policies could lead to results that may diverge from real-world impacts.As we focus only on secondary rebound effects (i.e. increase in consumption of other products nonrelated with the products directly affected by a policy) we do not investigate the impact of higher food availability and thus lower food prices on food demand.This relates to the negative income elasticity of demand for food observed across global regions (Cirera and Masset 2010, Clements and Si 2018), which reduces the possibilities of the existence of a direct primary rebound effect.At the same time, we also do not assume any explicit costs associated with an implementation of the FLW reduction measures, which could further increase food prices and act as a channel that reduces food demand (compared to the case of cost-free FLW policies implementation).
These aspects underscore the need for additional research to refine the representation of the connection between FLW policies and shifts in food demand, ensuring a more accurate reflection of the dynamics at play.Nonetheless, it is important to acknowledge that this study's scope is confined to exploring rebound effects primarily within the non-food sector.Future research could expand this scope by considering the potential existence of direct rebound effects within the food sector.By employing a dynamic modeling framework, it could be possible to showcase how Engel's law might simultaneously influence pollution levels and health outcomes across both food and nonfood sectors, providing a more comprehensive understanding of the interconnected dynamics and outcomes in relation to rebound effects.
Finally, in the case of Nigeria and SSA, the higher levels of CO emission intensities outside the food sector are associated with the use of biomass and other fuels in residential and other sectors6 .In the air pollution data sourced from Crippa et al (2020), corresponding emission flows are not explicitly distributed across activities and the process for such distribution implemented in the current study follows Chepeliev (2021).In this regard, a more refined identification of the air pollution flows across specific activities might provide additional valuable insights into the implications of the FLW reduction policies on changing emissions throughout the GVC.
In conclusion, our study emphasizes the complex relationship between FLW reduction policies, air pollution and related mortality risks.We provide an innovative quantification of air pollutants embedded in the global FLW, expanding the multidisciplinary relevance of FLW-related polices.We illustrate that decreasing FLW can generate positive spillover effects on reducing air pollution levels, but complementary policies are required to incentivize sustainable non-food consumption, improving air quality while decreasing associated mortality risks.Policymakers need to account for region-specific factors, consumer behavior, and the potential social implications, to effectively reduce FLW creating a healthier and more sustainable planet.
non-food items (Salemdeeb et al 2017, Read et al 2020) with higher pollution intensity.The existence of such rebound effects in the context of FLW reduction policies has been discussed in the earlier literature (Martinez-Sanchez et al 2016, Salemdeeb et al 2017, Reynolds et al 2019) and must be considered when policies are devised.While widely used global FLW databases (FAO 2011, 2019, Xue et al 2017, Kaza et al 2018, UNEP 2021) omit a comprehensive representation of global food trade and FLW embedded in non-primary foods (Gatto and Chepeliev 2023), a multi-regional inputoutput (MRIO) framework inclusive of FLW allows to trace lost and discarded foods and embodied air pollutants along FSC.MRIO models are economic and environmental frameworks apt to quantify the interdependencies between different sectors of the global economy, allowing the examination of crossborder relationships and global supply chains.FLWextended MRIO include the flow of lost and discarded foods across different countries, providing ad-hoc assessments of the environmental and economic impacts of FLW and related activities.The majority of FLW-extended MRIOs (Reutter et al 2017, Usubiaga et al 2018, Read et al 2020) rely on data from Gustavsson et al (FAO 2011), offering a potentially outdated (Sheahan and Barrett 2017, Xue et al 2017) representation of flows of lost and discarded foods.Gatto and Chepeliev (2023) develop a consistent FLW-extended MRIO coupled with a global nutritional database (Chepeliev 2022), relying on up-todate estimates computed consistently with the methodology and definitions adopted by UN-SDG12.3(United Nations 2019).

Figure 1 .
Figure 1.Shares (%) of FLW-embedded air pollutants in total air pollution for four pollutants generated across global countries in 2014.Panel A illustrates shares of FLW-embedded carbon oxide (CO) in total CO emissions by country in 2014.Panel B illustrates shares of FLW-embedded nitrogen oxides (NOX) emissions in total NOX emissions by country in 2014.Panel C illustrates shares of FLW-embedded ammonia (NH3) emissions in total NH3 emissions by country in 2014.Finally, Panel D illustrates shares of FLW-embedded particulate matter 2.5 (PM2.5)emissions in total PM2.5 emissions by country in 2014.Additional data and figures on other air pollutants embedded in FLW are available in the supplementary information.

Figure 2 .
Figure 2. Aggregate change in air pollutants embedded in food loss and waste from 2004 to 2014 based on KAYA identity.Estimates refer to percentage changes in the period between 2004 and 2014.The illustrated estimates are based on changes in the main drivers of air pollutants embedded in food loss and waste and are computed following the approach of Kaya and Yokoburi(1997).A list of countries composing each macro-region reported in the figure is available in the supplementary information.

Figure 3 .
Figure 3. Change (%) in CO, PM2.5, and NOX pollution in production and consumption regions generated by a 50% reduction in FLW by country and globally in 2014, by scenario.Estimates refer to change (%) in total polluting emissions (from food and non-food products) compared to our baseline where no FLW reduction policy is applied.Regions displayed on the vertical side of each square indicate where final consumption occurs and where the FLW reduction policy is applied, while regions displayed on the horizontal side of each square indicate where changes in emission linked to the production of food and non-food products occur.In the displayed squares, each row indicates that the FLW policy is uniquely applied in the corresponding region.The 'GLOBAL' region (row) represents the case in which the FLW reduction policy is uniformly applied to all global countries.The 'GLOBAL' production region (column) represents the impact of a FLW reduction policy (by country or global) on global food and non-food production.Panel A illustrates changes in CO, PM2.5, and NOX pollution in production and consumption regions in the 'No Rebound Effect' (NO_REBOUND) scenario.Panel B illustrates changes in CO, PM2.5, and NOX pollution in production and consumption regions in the 'Status Quo' (STATUS_QUO) scenario.Finally, Panel C illustrates changes in CO, PM2.5, and NOX pollution in production and consumption regions in the 'Environmental Awareness' (ENV_AWARE) scenario.As changes with respect to our baseline are reported in percentage (%), absolute variations (gigagrams) for the illustrated pollutants as well as for other pollutants across the investigated scenarios are reported in the supplementary information.

Figure 4 .
Figure 4. Decreased mortality risk by disease (saved lives (thousand people) in 2014 and per million of population) by a 50% reduction in global FLW, by scenario.Estimates refer to thousand people saved by the reduction in mortality risk linked to air pollution-induced diseases.Positive numbers illustrate a decrease in mortality risks while negative numbers illustrate an increase in mortality risks.Estimates have been calculated based on regional population in 2014 derived from the GTAP 10 Data Base (Aguiar et al 2019).Panel A illustrates changes in premature mortality risk (1000 saved lives and lives save per million of population) linked to air pollution-induced diseases when a 50% reduction in global FLW is applied in the 'No Rebound Effect' (NO_REBOUND) scenario in 2014.Panel B illustrates changes in premature mortality risk (1000 saved lives and lives save per million of population) linked to air pollution-induced diseases when a 50% reduction in global FLW is applied in the 'Status Quo' (STATUS_QUO) scenario in 2014.Finally, Panel C illustrates changes in premature mortality risk (1000 saved lives and lives save per million of population) linked to air pollution-induced diseases when a 50% reduction in global FLW is applied in the 'Environmental Awareness' (ENV_AWARE) scenario in 2014.Stroke and IHD-related mortalities are not reported as the variations in pollutants across scenarios are found not to affect mortality rates linked to such diseases.

Figure 5 .
Figure 5. Health-related co-benefits (billion USD per year) of a reduction in premature mortality risk related to air pollution across global regions in 2014, by scenario.Estimates refer the amount of saved billion USD dollars from the decrease in mortality risks associated to lower polluting emission generated by FLW reduction policies across our scenarios.Negative estimates refer to an increase in health-related expenses due to rising mortality risks linked to increasing air pollution trends.Illustrated estimates have been calculated on regional GDP in 2014 derived from the GTAP 10 Data Base (Aguiar et al 2019) and assuming a Value of Statistical Life of 1.8 mn USD for the lower bound and of 4.5 mn USD for the upper bound.Panel A illustrates health-related co-benefits (billion USD per year) of a reduction in premature mortality risk related to air pollution across global regions in 2014, in the 'NO Rebound Effect' (NO_REBOUND) scenario.Panel B health-related co-benefits (billion USD per year) of a reduction in premature mortality risk related to air pollution across global regions in 2014, in the 'Status Quo' (STATU_QUO) scenario.Finally, Panel C illustrates health-related co-benefits (billion USD per year) of a reduction in premature mortality risk related to air pollution across global regions in 2014, in the 'Environmental Awareness' (ENV_AWARE) scenario.

Table 1 .
Definition of investigated scenarios and quantification of changes in final demand.
a The definition of 'low pollution-intensive products' is based on the volume of air pollutants embedded in $1 of final consumption for each country and sector represented in the GTAP database(Aguiar et al 2019, Chepeliev 2021).We then rank all non-food sectors represented in the GTAP Data Base and determine four sectors with the lowest global average emission intensity of final consumption.Corresponding sectors include the following activities: information & communication (cmn), other financial intermediation (ofi), insurance (ins), and education (edu).TM5-FASST methodology (seeVan Dingenen et al  2018, Belis et al 2022, Sampedro et al 2022), healthrelevant exposure metrics considered in the present study are population weighted annual mean PM2.5 at 35% relative humidity and seasonal daily maximum 8 h average O 3 concentration metric (SDMA8h).The mortality associated with the exposure metrics used to compute health impacts in line with epidemiological studies(Pope III et al 2002, Jerrett et al  2009, Krewski et al 2009) Dietary changes, technological advancements, and broader environmental policies can influence health and mortality risks through different and more direct channels (Willett et al 2019).Reductions in FLW could have more direct impacts on health through an increased food availability (HLPE 2014, Kuiper and Cui 2020) or improved nutritional quality (Kummu et al 2012, Buzby et al 2014, Delgado et al 2021).
World Health Organization (WHO) 2022 Ambient (outdoor) air pollution (available at: www.who.int/news-room/factsheets/detail/ambient-(outdoor)-air-quality-and-health)Wyer K E, Kelleghan D B, Blanes-Vidal V, Schauberger G and Curran T P 2022 Ammonia emissions from agriculture and their contribution to fine particulate matter: a review of implications for human health J. Environ.Manage.323 116285 Xue L, Liu G, Parfitt J, Liu X, Van Herpen E, Stenmarck Å, O'Connor C, Östergren K and Cheng S 2017 Missing food, missing data?A critical review of global food losses and food waste data Environ.Sci.Technol.51 6618-33 Zhu J, Luo Z, Sun T, Li W, Zhou W, Wang X, Fei X, Tong H and Yin K 2023 Cradle-to-grave emissions from food loss and waste represent half of total greenhouse gas emissions from food systems Nat.Food 4 247-56