Impact of Healthy Diet Shifts on Soil Greenhouse Gas Emissions Across Europe

Authors: Vasilis Michailidis, Emanuele Lugato, Panos Panagos, Florian Freund, Diego Abalos

Abstract

Shifting towards healthy, plant‐based diets is widely recognized as a strategy to reduce greenhouse gas emissions (GHG) from food systems, primarily through reduced methane emissions from livestock. However, the implications of this transition for soil‐based GHG emissions, a major contributor to climate change, remain uncertain. We used the MAGNET economic model and the DayCent biogeochemical model to assess the impacts of dietary shifts aligned with the EAT‐Lancet guidelines on soil organic carbon (SOC), nitrous oxide (N2O) emissions, and the soil GHG balance across the European Union and the United Kingdom. Adopting the EAT‐Lancet diet reduced livestock production, organic carbon (C) and organic nitrogen (N) inputs from manure, and permanent grassland areas for agricultural use. This results in potential SOC losses of an EU average of 14 Mg CO2e ha^−1^ and reaching up to 50 Mg CO2e ha^−1^ in livestock‐intensive regions by 2100. However, afforestation of land released from production could offset approximately half of the diet‐induced soil C losses by 2100. When above‐ground biomass from afforestation is factored in, this could yield an additional 65 Mg C ha^−1^ in afforested areas, resulting in net CO2 removal at the European scale. N2O emissions exhibited more moderate and heterogeneous changes by 2100, ranging from 10 to −13 Mg CO2e ha^−1^ across the continent, and dependent on land use change (LUC) and increased synthetic N inputs. The changes in SOC were driven by LUC, lower organic inputs, soil types and, to a lesser degree, climatic zones. This study's findings underscore the importance of dietary changes in tackling climate change. However, practitioners and policymakers should carefully consider potential soil‐related trade‐offs by supporting and implementing appropriate soil conservation practices, such as no‐tilling or afforestation, to realize the full co‐benefits of more sustainable diets.

Introduction

Diets are crucial for human health and the environment. Accordingly, the European Union (EU) has adopted the farm‐to‐fork strategy to mitigate climate change by envisioning healthier and more sustainable food consumption patterns (Schebesta and Candel 2020). Modern societies are characterized by a high intake of animal‐sourced and highly processed food, and a low intake of plant‐based products. These diets are responsible for various non‐communicable diseases (Tilman and Clark 2014). In this context, the EAT‐Lancet Commission on Food, Planet and Health proposed a planetary healthy diet for adults, a diet rich in plant‐based foods and low in red meat, emphasizing the need to reverse current dietary patterns (Willett et al. 2019).

Food systems currently account for approximately one‐third of the total anthropogenic greenhouse gas (GHG) emissions (Crippa et al. 2021). These emissions are predominantly due to the livestock sector, a major contributor to environmental impact primarily through methane (CH4) emissions from enteric fermentation, global warming by feed imports and transport, and emissions related to land use change (LUC) (Leip et al. 2015; Tullo et al. 2019; Espinosa‐Marrón et al. 2022). However, diet shifts can have implications beyond livestock‐related emissions, particularly on soil GHG emissions which are currently a missing element in diet‐related impact assessments (Ran et al. 2024). For instance, livestock production—especially ruminants—requires large grassland and grass lays areas for feed production (Weindl et al. 2017). Reducing animal‐based food products would decrease the fraction of these systems within the agricultural land cover (Aleksandrowicz et al. 2016; Weindl et al. 2017; Poore and Nemecek 2018), and possibly increase the share dedicated to annual crops for human food to meet the nutritional requirements (White and Hall 2017).

Since grasslands and grass lays have a higher soil carbon (C) storage capacity compared to annual crops due to higher root C inputs and lower soil disturbance by tillage (Bai and Cotrufo 2022), such dietary shifts may reduce soil organic carbon (SOC) stock. This is concerning because maintaining and increasing SOC in soils is crucial to mitigating climate change and to enhancing the soil's capacity to provide essential ecosystem services. The extent to which these potential reductions in soil C will offset the GHG benefits of reducing animal‐related emissions remains unknown (Sun et al. 2022).

Changes in agricultural land use due to diet shifts can also have strong impacts on the emissions of nitrous oxide (N2O) (Wiesmeier et al. 2019; Hong et al. 2021). Both grasslands and arable lands are significant sources of N2O, a potent GHG primarily emitted from agricultural soils (Butterbach‐Bahl et al. 2013). Applying manure to soil can increase N2O emissions and enhance SOC storage, depending on soil and climatic conditions (Gross and Glaser 2021; Petersen et al. 2023). Therefore, shifting towards lower animal product consumption would reduce manure inputs, with uncertain effects on the overall soil GHG balance. Additionally, crop selection and management practices specific to grasslands and croplands play a substantial role in the overall soil GHG balance, with large variations across regions and timescales (Van Groenigen et al. 2017; Godde et al. 2020). Practices such as no‐tillage, crop rotations, plant residue incorporation, and cover crops can enhance SOC stocks, while conventional intensive farming may lead to SOC losses and higher N2O emissions (Lugato et al. 2018; Grados et al. 2022).

Assessing the biogeochemical flows of C and nitrogen (N) at the continental scale requires precision in capturing the spatiotemporal variability of soil and climatic conditions. Across the EU, there is a large variability in soil types, ranging from sandy soils with low organic C content to clay‐rich soils with high water retention, as shown in recent European Commission (EC) projects monitoring EU soils (Tóth et al. 2013; Orgiazzi et al. 2018). These variations profoundly affect soil GHG fluxes, especially N2O production and SOC turnover (Oertel et al. 2016). Climate conditions are also crucial in determining soil C and N fluxes (Koven et al. 2017); SOC stocks tend to be higher under cool, humid conditions such as the northern regions of the EU. In contrast, in warmer and drier climates, such as the Mediterranean region, soil organic matter (SOM) decomposes faster due to increased microbial activity (Conant et al. 2011). Conversely, soil N2O emissions are often higher in alkaline, clay‐rich humid soils with high precipitation rates (Li et al. 2022). Short‐lived climatic events lasting a few hours or days, such as rainfall after fertilization, can trigger N2O emissions, accounting for 39%–76% of the total annual emissions (Abalos et al. 2016). Therefore, understanding how diet changes affect the soil GHG balance can only be achieved with a comprehensive framework that fully captures the continental variation in soil and climate conditions at high spatiotemporal resolution, as well as the land use and management practices that these diet changes entail.

Various methods have been used to estimate the consequences of diet shifts on the soil GHG balance. Studies focusing on SOC impacts often employ basic soil C models or data‐driven models that oversimplify the soil processes regulating the C cycle (Nijdam et al. 2012; Goglio et al. 2015; Moberg et al. 2019; Bessou et al. 2020; Hammar et al. 2022; Sun et al. 2022). Other models operate at low spatial resolution and do not account for management practices affecting SOC levels (Weindl et al. 2017; Knudsen et al. 2019; Saarinen et al. 2023). To evaluate N2O emissions, assessments frequently rely on fixed emission factors (EF) that only consider N inputs to the soil (Nijdam et al. 2012; Theurl et al. 2020). While these studies consider soil emissions under a diet shift, they may yield inaccurate estimates, particularly in large‐scale assessments, as they overlook crucial biogeochemical interactions that drive C and N flows in soils in interaction with climate and management components (Li et al. 2005; Butterbach‐Bahl et al. 2013; Lal 2018).

Our study evaluates the impact of the EAT‐Lancet diet recommendations on the EU's soil GHG balance. Given that food systems are linked across global trade networks, shifts in demand and supply can have far‐ranging implications. Adopting the EAT‐Lancet reference diet in the EU, for instance, will affect agricultural commodities worldwide, influencing everything from trade to land use (Rieger et al. 2023; Navarre et al. 2023; Rulli et al. 2024). We considered these demand‐driven changes in food consumption and production under a diet shift scenario in the EU using the economy‐wide global economic simulation model MAGNET. We then used the EU‐wide DayCent modeling framework, as developed by Lugato et al. (2018), to assess the implications for the GHG balance of agricultural soils for the EU and its member states (MS). DayCent is a process‐based full ecosystem model simulating the C, N, and water cycles at a daily time step (Parton et al. 1998). This temporal resolution and its structure allow the model to estimate long and short‐term C and N feedback under heterogeneous land uses, soil‐climate conditions and agricultural management practices across the continent. The model has been extensively used to evaluate soil nutrient fluxes and yield, as well as the impact of agricultural management at a continental scale (Lugato et al. 2018; Quemada et al. 2020; Muntwyler et al. 2023; Pacifico et al. 2024). The modeling framework integrates the most advanced datasets, including the Land Use/Cover Area Frame Statistical Survey (LUCAS), an EU top‐soil survey on soil properties, land use, and land cover, which collects data following harmonized protocols that are ingested into advanced digital soil mapping methods to infer soil properties at 1 km^2^ resolution (Ballabio et al. 2016, 2019). By combining the insights from MAGNET regarding dietary shifts with soil biogeochemical simulations with the DayCent modeling framework, we provide a comprehensive evaluation of how implementing dietary shifts affects the net GHG balance of agricultural soils at a continental level, covering the period from 2030 to the end of the century. We mainly focus on the full adoption of the EAT‐Lancet reference diet across the EU, but we also provide sensitivity scenarios with lower adoption rates in Supporting Information. This study addresses the following research How does adopting the EAT‐Lancet reference diet affect agricultural production, LUC, and agricultural management practices (i.e., organic inputs and fertilization rates) at a continental scale?How is the soil GHG balance affected by the diet shift, and what are the main drivers of the observed changes?To what extent can afforestation offset the adverse effects of these dietary changes?

Materials and Methods

An overview of the framework linking MAGNET and DayCent, and their spatial resolution, is shown in Figure 1.

MAGNET Model

The Modular Applied General Equilibrium Tool (MAGNET) is a commonly used model for analyzing the effects of agricultural, trade, land, and bioenergy policies on the global economy, land use, agricultural prices, and nutrition (https://www.magnet‐model.eu/model/). MAGNET is a computable general equilibrium (CGE) model that determines relative prices that ensure all markets clear simultaneously. While it covers the entire economy, its emphasis lies on agricultural and land markets.

MAGNET encompasses primary agricultural commodities, such as raw milk and paddy rice, as well as value‐added products like dairy goods and processed rice. It features a dedicated module for first‐ and second‐generation biofuels (Banse et al. 2008; Philippidis et al. 2019) and incorporates segmented factor markets to capture differences in factor prices between agricultural and non‐agricultural sectors (Keeney and Hertel 2005). Additionally, its flexible production structure enables detailed modeling of feed supply systems including direct substitution between produced feed and land‐based inputs like grass (Woltjer and Kuiper 2014). The model also integrates agricultural subsidies aligned with the OECD's producer support estimates, allowing adjustments to reflect changes in subsidies and structure (e.g., coupling versus decoupling) over the projection period (Boulanger et al. 2015; OECD 2022; Springmann and Freund 2022). Further information on MAGNET's sectoral and regional aggregation in Tables S7 and S8.

In this study, we employed the MAGNET model to simulate changes in the production and consumption of food commodities, as well as land use changes at a continental scale, assuming three levels of adoption of the EAT‐Lancet guidelines across the EU Scenario 10%, Scenario 30%, and full‐adoption Scenario 100%. The focus is on the full adoption scenario, with the partial scenarios providing a sensitivity analysis.

DayCent Model

We employed DayCent, a state‐of‐the‐art biogeochemical process‐based model, known for its advanced capacity for simulating the dynamics of N, C, and GHGs in various ecosystems. DayCent is the daily timestep version of the CENTURY model. The model can simulate a range of N and C cycling processes by employing a variety of sub‐models including soil profile water movement and temperature distribution, plant growth and net primary production (NPP), decomposition of crop residues and soil organic matter, nutrient mineralization, N gas emissions from nitrification and denitrification, and CH4 oxidation in non‐saturated soils (Parton et al. 1994).

Specifically, for GHGs, the decomposition of C inputs leads to a shift in the available N and C either into mineral and soil pools, uptake from plants (C moves to other pools or as dissolved organic C), loss in gaseous form, and/or leach into the groundwater. Mineral N can also be incorporated into the organic soil pools for maintaining their C:N ratio. The rate and flows of C and nutrients, nitrification and denitrification, and consequently the emission rates, are influenced by soil physicochemical properties, climatic factors, and the availability of C and mineral N. Additionally, agronomic management practices can affect the biogeochemical composition of the soil. These factors make the model a powerful tool for predicting how changes in different soil and crop types impact the soil GHG balance while considering varying management and climatic conditions. By integrating robust regional data within the modeling framework, the EU‐wide DayCent modeling framework used here stands out as the most effective tool for large‐scale assessments (Lugato et al. 2018; Quemada et al. 2020; Muntwyler et al. 2023; Pacifico et al. 2024). In this study, we applied an established biogeochemical modeling framework that leverages a range of consistent EU datasets. The model operates at a continental scale and produces outputs at a high spatial resolution of 1 km^2^, as detailed below.

Data Inputs and Model Implementation

Soil physicochemical attributes were sourced from the LUCAS topsoil survey (Orgiazzi et al. 2018) and interpolated to a 500 m resolution (Ballabio et al. 2016, 2019). For meteorological data, we used the E‐OBS gridded dataset (http://www.ecad.eu) for the current climate until 2020. The climate datasets included minimum and maximum temperature, as well as precipitation data, provided on a 0.1° resolution grid. Future climate projections were derived from the CNRM‐CM5.4.1 general circulation model, a widely used and well‐documented tool in European climate assessments. These projections followed the RCP4.5 scenario (http://www.ecad.eu), which represents a moderate stabilization pathway through the year 2100. The land cover map for arable land and grasslands was obtained from the EU harmonized land cover and land use dataset Corine Land Cover (CLC), which offers a comprehensive pan‐European inventory of land cover and land use, created using a standardized methodology. The EU and its member states extensively utilize it for reporting, developing indicators, environmental modeling, and supporting a wide range of studies at the continental scale (Gallardo and Cocero 2023). Irrigated areas were extracted from the FAO‐AQUASTAT database by Siebert et al. (2005). All data inputs were interpolated at 1 km^2^ for this study.

Planting and harvesting dates for each crop were determined using the crop calendar map from the SAGE Center (Sacks et al. 2010). Within the identified arable land of the CLC, we established a most representative 4‐year crop rotation based on the average arable crop area at country level and as reported by the Member States to the EU Statistical office (named hereafter EUROSTAT). Additionally, to ensure a diverse representation of all crops within the rotation in a given year, we applied a random shift in the crop sequence across individual grid cells within each country. For future land‐use projections under different dietary scenarios, land cover changes were adapted based on LUC projections from MAGNET. Since the LUCAS survey lacks detailed information on certain management practices, we assumed that conventional agricultural methods involved primary (mouldboard) and secondary tillage, as well as mineral N applications, split into two events based on crop type (Lugato et al. 2018). Additionally, policy‐driven best management practices such as cover crops, and reduced and minimum tillage were obtained from the Farm Structural Survey of 2010 and 2016, reporting area coverage at regional (NUTS 2) level (Panagos et al. 2020); these management practices were fully implemented in the modeling exercise.

Inorganic N and P fertilization rates by crop and country were sourced from the latest International Fertilizer Association (IFA) survey (Ludemann et al. 2022). Organic N inputs are based on the “Livestock Gridded of the World” and were modified to reflect new livestock densities under different scenarios as simulated by MAGNET (Robinson et al. 2014).

For a more comprehensive description of the model development, input data and configuration, uncertainty ranges, and model validation within the EU modeling framework, we direct the reader to previous studies (Lugato et al. 2018; Muntwyler et al. 2023; Pacifico et al. 2024; Michailidis et al. 2025).

Diet Scenarios and Crop Simulations

Three diet shift scenarios were developed based on MAGNET outputs, representing different levels of adoption of the EAT‐Lancet diet within the EU, implemented in 2030 and evaluated through 2100. These scenarios reflect adoption rates by the EU's total population, set at 10%, 30%, and 100%. The 100% scenario aligns with the EAT‐Lancet diet recommendations while the 10% and 30% were chosen as more realistic future developments for the diet shift. The MAGNET baseline aligns with the Shared Socioeconomic Pathway 2 (SSP2) projections through 2030, which assumes moderate changes in population growth and medium economic development (Geibel and Freund 2023). Established agricultural and trade policies are gradually implemented over this period, while baseline diets are similar to those outlined in Geibel and Freund (2023) but with different regional aggregation (here all EU countries are singled out). From 2030 onward, MAGNET outputs were kept constant, providing static land use and dietary conditions. The baseline scenario reflects country‐ and commodity‐specific diets aligned with projected developments up to 2030. These diets are typically characterized by a high consumption of animal‐based products, processed foods, and sugars, alongside a low intake of fruits, vegetables, and nuts. In contrast, the diet shift scenarios are derived by comparing the baseline with the EAT‐Lancet diet recommendations for a daily intake of 2500 kcal per capita (Willett et al. 2019). These recommendations emphasize increased consumption of plant‐based foods, such as fruits, vegetables and nuts, while reducing the intake of red meat and processed foods.

MAGNET simulates a range of commodities; however, we focused on the number of live cattle (beef and dairy), sheep and goats, poultry, and swine as these directly impact organic input from livestock densities (Figure 2a). Additionally, we used the LUC outputs to quantify changes in the EU's land cover. The model estimates LUC for various crop types including grains, fruits and vegetables, sugar crops, oilseed, and fodder crops (Figure 2b). We aggregated the commodities to arable land, permanent grasslands, fodder crops, horticulture, and permanent crops to illustrate the LUC and the corresponding freed‐up land for each scenario (see Table S2).

The impact of dietary shifts on the EU's and UK's agricultural soil GHG balance was evaluated against a baseline scenario reflecting business‐as‐usual diets. Reducing livestock production following the EAT‐Lancet diet recommendations is associated with a decrease in the number of livestock animals. The outputs from the MAGNET model are presented as percentage differences relative to the baseline. Based on these outputs, we adjusted the livestock model input density by categories reported by EUROSTAT to reflect the projected changes. This adjustment corresponds to the reduced demand for the reference diet, resulting in fewer animals being raised. MAGNET projections, which account for reduced meat production and increased fruit and vegetable consumption, were used to inform changes in land allocation for animal feed and crops for human consumption. While MAGNET includes most of the commodities included in DayCent simulations, it does not specifically address the allocation of temporary grasses and permanent grasslands. To address this, we developed a framework to allocate permanent grasslands and fodder crops based on livestock type and region (see Table S3).

MAGNET provides projections on LUC and production changes for the previously mentioned commodities up to 2030 for the baseline, while the projected diet scenarios show relative differences (percentage) to the 2030 baseline thereafter. After 2030, we assume that these differences remain constant relative to the 2030 baseline. We used the land cover of the CLC dataset and the organic input data for each MS based on EUROSTAT statistics until 2022 and, from 2022 to 2030, we aligned the land cover/use of the Daycent modeling framework to the MAGNET projection (Figure 1). After 2030, we assumed that land‐use patterns and production levels remain at their 2030 values until 2100. We adapted accordingly for each crop category, but we focused on crops that DayCent simulates (orchards and vegetables were excluded; the model is not calibrated for these crops, but we simulated the LUC as a proxy). During the conversion of LUC from baseline to the scenarios, we prioritized the conversion of fodder crops to arable land when needed, and lastly, the transformation of permanent grasslands into arable land. To account for the lower organic N in the scenarios, we increased the mineral N inputs proportionally to the reduction in organic N at the pixel level, assuming a 70% efficiency of the organic fertilization, and limiting at a maximum of 100 kg of N replacement. We ran DayCent from 1980 to 2100 and assessed the soil changes with different metrics (see below) for the period 2030 to 2100 of the scenarios in comparison to the baseline.

Environmental Indicators and Afforestation Scenario

We calculated the net GHG flux and compared it with the baseline scenario. This included changes in soil C fluxes, and direct N2O emissions from soil. Indirect N2O emissions (via leaching) were excluded from the calculations to avoid relying on emission factors and because they are an order of magnitude lower than direct emissions but can be found in Table S1. We calculated the soil GHG balance (Mg CO2e ha^−1^ year^−1^) according to Grados et al. (2024):(1)GHGbalance=directN2O×273×4428−4412×ΔSOCwhere direct N2O emissions describe the soil N2O emissions (kg N2O–N ha^−1^ year^−1^), and ΔSOC represents annual changes in SOC within the top 30 cm of the soil. The global warming potential of N2O over a 100‐year horizon was set at 273 (IPCC 2023).

We also assessed the impact of dietary changes on CH4 emissions from enteric fermentation and manure management due to reduced livestock production and consumption, to approximate potential CH4 reduction under the diet shifts. These estimates were not part of the DayCent soil emission simulations and are included in Supporting Information. Total CH4 emissions from enteric fermentation and manure management were calculated using the IPCC Tier 1 approach, which assigns an EF to each livestock category (IPCC 2023). This EF was then multiplied by the number of livestock heads for each species (IPCC 2023). Emissions from manure management based on the Tier 1 approach are classified by average annual temperature. The total emissions are the sum of each category and subcategory (Gg CH4 year^−1^). To convert CH4 emissions to CO2e, we applied a global warming potential of 28 over a 100‐year timeframe.

We estimated how much of the soil GHG emission changes due to diet shifts could be affected if the spare agricultural land, as projected by MAGNET (6 M ha of permanent grasslands in the total EU), is used for afforestation (i.e., Afforestation Scenario). To establish this scenario, we substituted the spare agricultural land of the diet shift scenarios with afforested area across the EU. LUC emissions from the conversion of cropland to forest were calculated based on a time‐dependent exponential decay function (Equation 3).(2)ΔSOCt=SOCinitial+SOCequilibrium−SOCinitial×1−e−ktwhere ΔSOC is the SOC at time t (year^−1^), SOCinitial is the SOC of the previous land use at the time of afforestation (2030) (Mg C ha^−1^ year^−1^), SOCequilibrium is the mean SOC of forests for each MS and is equivalent to SOC at equilibrium (Mg C ha^−1^ year^−1^) (Lugato et al. 2021), and k is the growth rate constant here defined as in Equation (3).(3)k=Rate ofSOCchangeSOCequilibrium−SOCinitial

The rate of SOC change (Mg C ha^−1^ year^−1^) after afforestation is derived from the literature (Shi et al. 2013; Poeplau and Don 2013; Bárcena et al. 2014; Hou et al. 2020) and set at 0.2 (standard deviation = 0.05) Mg C ha^−1^ year^−1^. Emissions of N2O from land under afforestation were set at 1 kg‐N2O ha^−1^ year^−1^ as frequently reported (Yu et al. 2022; Cen et al. 2024).

The same approach was used to estimate the above‐ground biomass (Mg C ha^−1^ year^−1^) of the afforested areas. The forest biome at saturation as an approximation across the EU was taken from Miettinen et al. (2024), while the rate of above‐ground biomass was set at ~1.2 (standard deviation = 0.4) Mg C ha^−1^ year^−1^ (Houghton 2005; Pretzsch et al. 2023).

Results

Production and Land Use Changes

Figure 2 illustrates the changes in the livestock sector for the scenario with 100% diet shift adoption across the EU. The changes are generally proportional to the degree of diet adoption for each scenario (Figures S1 and S5). Cattle for beef production, along with sheep, goats, and swine, had the largest reductions, followed by poultry and dairy. For the full scenario, the countries with the largest cattle production sector, Italy and France, experienced the highest reduction in cattle production, reaching up to 74%. Poland and Germany follow closely with reductions reaching up to 69%, while Spain, the Netherlands, and Belgium see reductions of 60%. Ireland and the UK experienced smaller reductions of 45% and 40%, respectively.

Italy exhibits the biggest reduction in the sheep and goats sector with 71%, followed by France with 58%. Other countries with a prominent livestock sector, such as the UK, Spain, Romania, Greece, and Ireland, have reductions of 27%, 49%, 48%, 35%, and 45%, respectively. The countries with the strongest swine production sector include Spain, Germany, France, Denmark, the Netherlands, Poland, and Italy. Their reductions vary, with Denmark seeing the smallest decrease (16%) and other countries ranging between 36% and 59%. Among the largest poultry‐producing MS, reductions ranged from 31% to 47%. In the dairy sector, the reductions are reaching up to 40% for France, the Netherlands, and Italy, while Germany, Ireland, the UK, and Poland show smaller decreases of 24%, and 16%, respectively. In total the EU would face a reduction of approximately 70 million swine (−36%), 2 billion poultry (−32%), 50 million cattle (−47%) for meat production, 7 million dairy cattle (−24%), and 36 million sheep and goats (−42%).

The changes in LUC are shown on Figure 2b. Similar trends are observed across the three scenarios in all MS (Figures S2 and S6). Following a reduction in the livestock sector, the shares of permanent grasslands and fodder crops decline in each MS, while arable land and crop areas for fruits and vegetables generally increase. In the 100% scenario, 4% (5.5 Mha) of the total land simulated is converted from permanent grasslands to arable crops to meet the demand for human consumption, and approximately 1% (1 Mha) of arable crops are converted to permanent grasslands. Just over 4% (6 Mha) are grasslands that can be used for non‐agricultural purposes. Germany (1.3 Mha), Italy (1 Mha), and Poland (1.1 Mha) had the highest freed‐up land, while 1.2% (1.7 Mha) is needed as additional agricultural land for Spain and Greece. The land sparing associated with adherence to the EAT‐Lancet diet could, in principle, be even greater. However, it is limited here due to the focus on the EU alone. Because of this, there is considerable scope for expanding exports of fodder crops from the EU to regions that do not adopt diet changes. For instance, wheat exports increase by 24% in the 100% EAT‐Lancet scenario.

Italy, France, and Sweden had the largest reduction in fodder crops by 575, 408, and 305 thousand hectares, respectively. Conversely, arable land expanded in nearly half of MS. France, the UK, Romania, and Finland increased their arable land by 1631, 697, 287, and 160 thousand hectares, while Italy, Poland, Spain, and the Czech Republic removed about 571, 299, 140, and 103 thousand hectares, respectively. The shares of horticulture increased in all of MS. The countries with the highest increase in their share of horticulture are Ireland, Finland, Denmark, Austria, and France with more than 50%. Netherlands, Belgium, Slovenia, Spain, Germany, Italy, Hungary, Poland, Croatia, and Portugal show a moderate expansion of the agricultural land ranging from 30% to 39%.

Impact of Diet Shifts on GHG Emissions at National Level and Main Drivers

In all scenarios, the simulations showed a decline in SOC for the majority of MS and a mixed effect on N2O emissions relative to the baseline scenario (Figure 3, Figures S3 and S7). The magnitude of these impacts corresponded directly to the level of diet adoption within each scenario. Dietary changes were implemented starting in 2030 and compared to the business‐as‐usual diet. By 2060, the highest SOC losses (Mg CO2e ha^−1^) were observed in the Netherlands (30.9), Ireland (36.4), Belgium (15.6), Portugal (11.4), Austria (11.4), Germany (10), France (23.1), and the UK (20) and Slovenia (20). The remaining MS had losses ranging from 0.5 to 8 Mg CO2e ha^−1^. The average EU‐27 + UK experienced losses of approximately 10 Mg CO2e ha^−1^ by 2060. Nitrous oxide emissions followed a similar pattern, with the Netherlands having the highest increase in emissions by 2060 (6.2 Mg CO2e ha^−1^), followed by Ireland, Belgium, and Croatia (4.7 Mg CO2e ha^−1^). Substantial reductions in N2O emissions would occur in Italy (−4.8 Mg CO2e ha^−1^). The increases in N2O emissions for the rest of MS vary from 0.3 to 2.5 Mg CO2e ha^−1^. By 2100, the changes in the GHG balance resulting from dietary shifts compared to current diets followed a consistent pattern across countries. We identified three distinct an upper class of MS ranging from 15 to 50 Mg CO2e ha^−1^ (Germany, Austria, Portugal, Belgium, Slovenia, the UK, Ireland, Netherlands), a middle class (Greece, Spain, Denmark, Slovakia, Romania, Finland, Lithuania, Cyprus, and Croatia) from 4 to 8 Mg CO2e ha^−1^, and a lower‐to‐negative class (Hungary, Bulgaria, Italy, Czech Republic, Poland, Malta, Latvia, Estonia, Sweden, Luxemburg) from 1 to −6 Mg CO2~e ha^−1^.

There is substantial variation in SOC changes in comparison with the baseline (ΔSOC) across EU‐27 + UK under the 100% EAT‐Lancet diet adoption scenario (Figure 4a). Areas with intensive livestock production, such as western and central Europe, exhibited the largest SOC losses by 2100, while regions with lower livestock density showed more moderate changes. The magnitude of SOC losses linearly correlated with the reductions in organic C inputs (R ^2^ = 0.49) (Figure 4b, right and left panels) due to decreased manure application, and land use change from permanent grasslands to arable crops (R ^2^ = 0.30), Figure 4c illustrates the effects of soil type and temperature on ΔSOC. Elevated SOC losses are evident across all soil types, with sandy soils showing the highest SOC stock reductions. The results indicate a complex relationship between ΔSOC and climatic variables (Figure S9, Table S4). Precipitation had a stronger effect on ΔSOC than temperature. However, this effect was weaker than that induced by changes in organic inputs.

The spatial distribution of N2O emission differences (ΔN2O) under the 100% EAT‐Lancet diet adoption scenario shows varying patterns across EU member states (Figure 5a). The reductions in N2O emissions are more pronounced in Southern and Eastern Europe and in the Scandinavian countries. Conversely, some regions in France and the UK exhibited minimal changes, likely reflecting smaller increases in mineral N inputs.

The relationship between ΔN2O and mineral N input changes (Figure 5b, left panel) showed a positive correlation (R ^2^ = 0.41), indicating that regions with a higher replacement of organic N by mineral N inputs experienced higher N2O emissions. In contrast, the relationship with organic C input changes (Figure 5b, right) was weak (R ^2^ = 0.06). Land use change correlated positively with the change in N2O (R ^2^ = 0.31).

Figure 5c shows the effects of soil type and precipitation on ΔN2O. Increases in N2O emissions were evident across all soil types, with clay soils emitting more compared to sandy soils. The effect was also significant among all soil types. The effect of climate on ΔN2O was less pronounced (Figure S10, Table S5)—we observed a weak negative and a positive and significant correlation between temperature and precipitation and ΔN2O, respectively (Table S5).

Afforestation Scenario

Several trends can be observed implementing the afforestation scenario (Figure 6). SOC stocks in the baseline increased until 2050 and then reached equilibrium and started stabilizing, while afterwards, SOC decreased by 2100. The scenario with a diet shift without afforestation (with the abandoned grasslands) showed gradual SOC losses, reaching a reduction of 3.75% compared to the baseline by 2100. However, the inclusion of afforestation in the diet scenario compensated for most of these SOC losses, maintaining SOC stocks closer to the baseline. In terms of cumulative N2O emissions across scenarios by 2100, the diet‐only scenario slightly decreased N2O emissions compared to the baseline. However, the afforestation scenario reduces cumulative N2O emissions to 175 kg N2O. Figure 6c illustrates the C sequestration potential of the planted forests through above‐ground biomass. We estimated that by 2100, these 6 Mha of forests planted could store approximately 65 Mg C ha^−1^ in their aboveground biomass, indicating a large C sink potential.

Impact of Diet Shifts at High Spatial Resolution and Regional Variability

The spatial distribution of the net soil GHG balance after an EU‐wide diet adoption (Figure 7) revealed a distinct heterogeneity across the MS. High‐emitting regions, represented by yellow areas (> 0.60 Mg CO2e ha^−1^ year^−1^), are concentrated in parts of western and central Europe including regions of France, the UK and Ireland, the Netherlands, and western Germany. Conversely, areas with minimal changes in emissions after diet adoption dominated in eastern Europe and the Mediterranean. Moderate changes in emission levels after diet adoption are widespread across the European continent, particularly in parts of eastern Spain, northern Italy, and central Poland.

Discussion

The livestock sector is responsible for more than two‐thirds of food‐related GHG emissions (Crippa et al. 2021), primarily through CH4 from enteric fermentation and manure management. Based on an IPCC Tier 1 simple approach, our estimates confirm that the GHG mitigation potential of diet shifts towards lower animal‐based food products is extremely high compared to our current diets (Figure S11). However, unlike previous research, our study goes beyond these well‐known reductions by shedding light on the previously overlooked impacts on long‐term GHG projections, revealing an environmental trade‐off associated with such transition. Our results depict profound transformations in food production, land use and the soil GHG balance, presenting challenges and opportunities for mitigating GHG emissions from the agricultural sector during the transition towards healthier diets.

Production, Land‐Use Change, and International Trade After a Dietary Shift in the EU

The reductions in livestock production (Figure 2a) across the EU align with the EAT‐Lancet reference diet recommendations. While all MS are expected to experience high reductions in livestock numbers, the actual numbers vary substantially. Countries with a strong livestock sector, such as Spain, the UK, France, Denmark, Poland, the Netherlands, and Italy, are projected to face the greatest challenges. These reductions could pose economic and social challenges, particularly in rural areas highly dependent on livestock farming, where local economies may lose income (Rieger et al. 2023; Geibel and Freund 2023).

Land‐use changes associated with the 100% dietary shift are projected to reduce the area of permanent grasslands across the EU by approximately 10 Mha (−22% compared to the current baseline), largely due to the reduced demand for livestock products (Figure 2b). This reduction aligns with previous assessments of dietary shifts (Humpenöder et al. 2024). This goes along with an expansion of arable land (arable crops, orchards, and vegetables) required to support increased production for direct human consumption. As a result, in several countries, permanent grasslands are converted to cropland to meet this shift in demand. However, at the individual country level, in some MS such as Italy and Poland, arable land requirements decrease, which allows the transformation of cropland to permanent grasslands or afforested areas. Taking into consideration the increased demand for human consumption, the total EU freed land is estimated at approximately 6 Mha. This provides an opportunity to use this land for policy interventions targeting reductions in GHG emissions and protecting biodiversity.

According to MAGNET results, the international trade of food commodities is likely to change dramatically after an EU‐wide adoption of the EAT‐Lancet diet guidelines (Table S6). Compared to the baseline scenario, the imports of beef commodities are projected to decrease by approximately 76% in the EU, similarly to the imports of swine, sheep, and goats. Poultry imports are slightly less affected by diet shifts, with a reduction of 49%. In contrast, imports of fruits and vegetables are projected to increase greatly across the EU by 150%. Comparing the exports in the diet scenario with the baseline, beef meat and horticultural commodities would decrease whereas grain products, oil seed commodities, sheep and goat and poultry meat, would increase. Given the changes in trade flows, it is likely that emissions will also fall in the rest of the world, which is supported by the MAGNET results aggregated at the global level, showing a net saving in GHG emissions (Table S6).

Soil GHG Balance: Spatiotemporal Dynamics

Cumulative SOC losses after a healthy diet adoption were substantial across the EU (Figure 3), and they were largely driven by the reduced manure input from livestock systems and LUC (Figure 4b). The reduced SOC stocks were more pronounced in regions with high livestock densities and animal manure production (Köninger et al. 2021). Cattle manure, one of the main sources of organic inputs to soils in the EU, is closely linked to beef production. Our findings show that countries with the highest emissions (e.g., Ireland, Netherlands, France) are also among the largest producers of cattle manure, explaining their evident SOC losses under the diet shift since manure application significantly increases SOC levels (Maillard and Angers 2014). Still, the extent of this increase is influenced by various factors, including soil properties and, to a lesser degree, climatic conditions. In our study, SOC losses reached an average of 3.38 Mg C ha^−1^ in sandy soils, compared to 3.20 Mg C ha^−1^ in clay soils by 2100. The modeling outcomes are supported by the evidence that sandy soils have reduced aggregate formation and less physical protection (Poeplau et al. 2015), due to a coarse mineral size texture. The low mineral surface leads to fewer opportunities for organic matter to bind to mineral surfaces, leading to less stabilization of SOC (Basile‐Doelsch et al. 2020). Additionally, the low water retention and enhanced aeration of sandy soils increase decomposition rates (Osman 2018), as also accounted by DayCent. The higher prevalence of sandy soils in Northern Europe helps explain the increased SOC losses in these regions after the diet shift adoption.

In terms of climatic conditions, we did not observe any clear patterns. Most studies suggest that warmer climates may increase the decomposition rate of SOC due to enhanced metabolic rates of microbial communities (Hartley et al. 2021; Gutierrez et al. 2023; Georgiou et al. 2024). On the contrary, Koven et al. (2017) found that SOC turnover rates are more sensitive to colder temperatures. We also noticed a negative relationship between temperature and ΔSOC (Conant et al. 2011; Georgiou et al. 2024) although, in our study, the relationship was not significant. Soil organic carbon turnover can also be strongly related to soil moisture (Kerr and Ochsner 2020). In our study, we observed a negative relationship between SOC changes and precipitation rates. Higher precipitation rates tend to accelerate SOC turnover by enhancing microbial activity (Carvalhais et al. 2014), but in waterlogged soils, decomposition can slow down due to oxygen limitations. Increased water availability can also stimulate plant growth, which, in turn, increases C inputs and SOC. The lack of consistent patterns between SOC and temperature and precipitation suggests that other factors, such as soil texture and land management practices, were more important drivers of SOC changes after the diet adoption than climate.

We found that SOC losses occurred rapidly after diet implementation, with significant reductions observed within 35 years, highlighting the temporal dependence of soil C changes. The temporal dynamics result from C stabilization processes, where SOC losses gradually reach an equilibrium because of physical, chemical, and biological mechanisms (Singh et al. 2018; Kan et al. 2022), as predicted by the model. These findings suggest that, as manure availability declines, alternative C recovery strategies, such as the application of digestate or biochar, crop diversity, inclusion of cover crops, conservation tillage, and crop residue retention, should be considered to mitigate SOC losses (Lehtinen et al. 2014; Frank et al. 2015; Poeplau and Don 2015; Abdalla et al. 2019; Greenberg et al. 2019; Beillouin et al. 2023; Rosinger et al. 2023; Gurmessa et al. 2024).

In addition to the impact of manure availability on SOC, the LUC induced by the diet shifts was also responsible for the projected C losses. The contribution of grasslands to the global SOC stocks (Taube et al. 2014), and the threat posed by LUCs to annual crops is widely established (Ledo et al. 2020; De Rosa et al. 2023). Grassland plant species typically develop deep root networks and contribute to SOC through root turnover, exudation, and depositing of organic materials deep in the soil profile (Bai and Cotrufo 2022). In DayCent, the allocation to belowground biomass is indeed higher for grassland plants compared to annual crops. Additionally, these systems are less disturbed and maintain year‐round soil cover. In our scenario, the conversion of permanent grasslands and lays contributing to lower SOC stocks was apparent in countries like France (−8.5%), Ireland (−5.3%), the UK (−5.2%), Netherlands (−6.4%), Slovenia (−20%), Austria (−5.2%), Portugal (−10%) and Spain (−6.5%), where grasslands are highly prevalent (Peeters 2009). Conversely, in cases like Italy and Poland, which typically have high livestock productivity, they required less arable land under a diet shift and therefore arable land was converted to permanent grasslands, resulting in lower overall SOC losses (−3.3% and −2.6%, respectively).

Nitrous oxide emissions were less affected by diet shifts compared to SOC losses. Some regions**—that is, France, Ireland, the Netherlands, the UK, Portugal, and Slovenia—**experienced increases in N2O due to the conversion of perennials to annual crops. This shift is associated with increased tillage frequency and faster turnover of organic matter (Abalos et al. 2016; Helfrich et al. 2020) which, in DayCent, results in enhanced mineralization of N and a larger substrate availability for nitrification and denitrification. Conversely, in countries such as Italy, Poland, Estonia, and Sweden, in which arable land was converted back to undisturbed land, the emissions declined. Additionally, we observed significant variations in N2O emissions across soil types (Figure 5c), with clay soils having 194% higher N2O emissions than sandy soils. This can be explained by the greater water retention capacity of clay soils, which creates an anoxic environment after rainfall and irrigation events that favors denitrification (Butterbach‐Bahl et al. 2013; Oertel et al. 2016), as accounted for by DayCent. Conversely, in sandy soils, nitrate accumulation from nitrification is more prone to leaching rather than gaseous losses from denitrification. There was a weak but significant negative relationship between increases in N2O and temperature, which is counterintuitive given that higher temperatures typically stimulate microbial activities and N2O emissions through positive feedback mechanisms (Tu et al. 2024). As with the changes in SOC, this may indicate that management and land use factors were stronger predictors than climatic factors of the changes in N2O after the diet adoption.

Reducing mineral N applications while maintaining crop yields for a growing population requires significant improvements in N use efficiency and nutrient management. Previous studies estimated a reduction in N2O emissions under reduced meat intake on a global scale (Popp et al. 2010; Reay et al. 2012; Geibel and Freund 2023) primarily due to reductions in N2O emissions from manure management, which can account for more than 11% of the N losses (Köninger et al. 2021). Our study did not account for emissions from manure management, focusing solely on soil emissions. Policies promoting precision fertilization, enhanced‐efficiency fertilizers (e.g., nitrification inhibitors), and diversified cropping systems could help mitigate the increased N2O emissions associated with the diet shifts (Abalos et al. 2022).

Afforestation Opportunities

Afforestation in the land that became available after the adoption of EAT‐Lancet guidelines showed potential for mitigating soil C losses and reducing net GHG emissions, mostly by above‐ground biomass C sequestration. We estimated a SOC accumulation of an average 0.4% year^−1^ in the EU, which aligns with the studies from Poeplau et al. (2011) and Bárcena et al. (2014). However, these SOC increases were region‐specific, as converting permanent grasslands to forest can also decrease SOC in mineral soils by altering soil biological composition and the biochemical quality of organic inputs (Joly et al. 2025). By 2100, above‐ground biomass of the afforested areas sequestered on average 65.3 Mg C ha^−1^, contributing to 390 Mt. of C in total by 2100, and representing a substantial contribution to the EU's carbon neutrality goals. This means that, by increasing SOC storage and providing an additional C sink through above‐ground biomass sequestration, afforestation almost fully offsets the SOC losses induced by diet adoption, sequestering approximately 75% more C (SOC and aboveground biomass compared to the baseline by 2100), resulting in a significant net C gain. Our estimates show a time lag of around 75 years between planting trees and achieving significant C sequestration, suggesting that EU policies would require additional, shorter‐term measures to protect SOC losses if initiatives to promote healthy diet shifts are implemented.

Limitations and Policy Implications

We projected a 37% increase in fruits and 48% in vegetable production on average across the EU, in comparison with the baseline. However, in our modeling study we excluded the simulation of fruit and vegetable production with DayCent. Instead, when land‐use conversion from grassland to fruits and vegetables was required, we approximated this transition by simulating the conversion of grassland to arable land, serving as a proxy to capture the general effects of converting perennial systems to more intensive land uses. This limitation likely affects our SOC and N2O estimates, although the extent remains uncertain. It is likely that the N2O emissions from these crop types would not differ largely from those of arable crops (Rezaei Rashti et al. 2015). In contrast, SOC under perennial fruit and nut trees can be higher in most cases (Ledo et al. 2020). Future studies should improve the SOC estimates with these crops, which are currently limited due to the low availability of benchmark field datasets with fruit and nut trees to validate and improve model simulations.

While a comprehensive uncertainty analysis is beyond the scope of this study, sources of variability should be acknowledged when interpreting the results. This study combines output from the MAGNET economic model with DayCent, both of which carry inherent uncertainties. In MAGNET, the main uncertainties arise from assumptions related to market responses regulated by price and trade elasticities that can influence projected land‐use patterns and production levels. Additionally, these projections remain static throughout our analysis. In DayCent, uncertainties originate from generic model parameterization, structure, and input data including interpolation of LUCAS points, weather data aggregation, and homogeneous regional crop management. Rigorous uncertainty assessments and using other models (e.g., in an ensemble) would be necessary to quantify the variability of the reported results.

Our findings demonstrate that a diet shift towards more plant‐based food products could lead to substantial reductions in SOC across European soils (an average of 3.75% across the EU‐27 + UK without afforestation). Accordingly, existing EU policies including the European Green Deal and Farm to Fork Strategy that promote sustainable food systems and healthier diets, must be refined to address these unintended diet‐induced trade‐offs and ensure the sustainability of soil systems. Our model estimates, considering interactions between agricultural land use, management practices, climate conditions and soil properties, provide a spatially explicit map at high resolution of the regions with higher increases in soil GHG emissions after adoption of the EAT‐Lancet diet. These areas represent key targets for implementing GHG mitigation strategies. A range of regenerative practices, including diversified cropping systems (Yan et al. 2022), cover cropping (Poeplau and Don 2015), crop residue retention (Lehtinen et al. 2014), conservation tillage (Frank et al. 2015), and biochar and digestate application (Gurmessa et al. 2024), may enhance SOC stocks and simultaneously improve soil fertility. Additionally, conservation agriculture can be beneficial for both SOC and N2O as a mitigation strategy (Li et al. 2023). In regions with high risk for N2O emissions, targeted interventions such as optimized fertilization rates, nitrification inhibitors, and liming can significantly reduce emissions (Zhang et al. 2022; Grados et al. 2022). As our results indicate, these interventions are particularly important for sandy soils and wetter regions, where SOC losses and N2O are most pronounced.

Our results illustrate a massive increase in fruit and vegetable imports, meaning that GHG emissions associated with the production of these food commodities may increase overseas. However, this is offset by a decline in meat and dairy imports, which typically have a higher C footprint. To minimize emissions displacement and ensure the success of EU‐level dietary shifts, policy mechanisms must be implemented to address trade‐related risks. These could include Carbon Border Adjustment Mechanisms (CBAM), which implement C tariffs on imported plant‐based products. CBAM would encourage global producers to adopt more sustainable practices while preventing carbon leakage, although the effectiveness of such a measure in agriculture is uncertain and requires careful consideration (Arvanitopoulos et al. 2021; Fournier Gabela et al. 2024). Supporting a global dietary shift beyond the EU is required to ensure global sustainability. Similarly, our LUC was adopted according to simulations of the economic model MAGNET, which are based on agricultural and land markets. Policy decisions that prioritize environmental and biodiversity targets over purely economic considerations will be needed to minimize the negative side effects on the soil GHG balance of healthy diet shifts.

We focused on sustainable diets in the EU. Although regional dietary changes are a crucial step in fighting sustainability and health problems, diet shifts on a global scale are needed to stay within planetary boundaries and save up to 11.6 million premature deaths per year (Springmann et al. 2018; Willett et al. 2019). If other countries' diets were to align with the EAT‐Lancet diet as well, this would have implications for the production and land use in the EU and elsewhere. The impact of such a scenario on soil‐related GHGs is unknown, and future research should be directed towards global‐scale scenarios, as soon as necessary data are available.

Conclusion

Shifting towards healthy, plant‐based diets is widely recognized as a strategy to reduce GHG emissions from food systems, primarily through reduction in CH4 emissions. However, this study revealed a critical trade‐off. While reducing livestock production in Europe lowers direct emissions, it also results in substantial SOC losses in agricultural soils, particularly in livestock‐intensive regions such as Ireland and the Netherlands. Across the continent, the rise in soil emissions resulting from a dietary shift would offset approximately 17% of the GHG reductions achieved through decreased enteric CH4 emissions associated with lower livestock production. The SOC losses primarily result from the reduction in manure due to a decline in livestock numbers across the EU, and from the effects of LUC from permanent grasslands to arable land, which typically involve tillage and are associated with higher SOC turnover. Increased reliance on mineral N fertilizers in intensified cropping systems, combined with the mineralization of available C from the LUC, led to moderate increases in N2O emissions, emphasizing the need for careful N management. These changes are subject to strong spatiotemporal dynamics, mostly driven by livestock intensity and soil types, and to a lesser degree by climatic factors. From a policy perspective, these results highlight the need for well‐timed targeted and region‐specific strategies such as reduced tillage, cover cropping, and the retention of crop residues. These strategies should prioritize preventing soil C losses in regions with sandier soils and N2O emissions in clay soils, as we have shown that they are more sensitive to the impacts of diet shifts. If well applied, adaptation of conservation management practices can mitigate and potentially fully offset the negative consequences associated with dietary shifts.

Beyond on‐farm mitigation measures, adopting the EAT‐Lancet diet recommendations provides a unique land‐sparing opportunity that can be strategically repurposed for climate and nature‐based solutions. We showed that the afforestation of 6 Mha of agricultural land that becomes available for non‐agricultural purposes can increase C sequestration (aboveground biomass and SOC) by 75% by 2100 and in comparison with the baseline. While overall the EAT‐Lancet diet offers clear climate benefits (net mitigation of 131 Mt. CO2e year^−1^ at the EU level), its success will depend on implementing additional soil and land‐use strategies that ensure the environmental sustainability of food systems.

Author Contributions

Vasilis Michailidis: conceptualization, data curation, formal analysis, methodology, methodology, software, software, writing – original draft, writing – original draft, writing – review and editing, writing – review and editing. Emanuele Lugato: formal analysis, methodology, software, supervision, writing – review and editing. Panos Panagos: supervision, writing – review and editing. Florian Freund: methodology, software, writing – review and editing. Diego Abalos: conceptualization, formal analysis, funding acquisition, methodology, supervision, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information