How much multilateralism do we need? Effectiveness of unilateral agricultural mitigation efforts in the global context

The CAPRI modelling system is a comparative-static partial equilibrium model for the agricultural sector developed for policy and market impact assessments from global to regional and farm type scale (Britz and Witzke 2012). The core of CAPRI is based on the linkage of a European-focused supply module and a global market module. The regional supply module consists of independent aggregate non-linear programming models combining a Leontief-technology for variable costs of the different production activities with a non-linear cost function which captures the effects of labour and capital on farmers' decisions. Each programming module optimizes income under constraints related to land availability, nutrient balances for cropping and animal activities, and policy restrictions. Prices are exogenous to the supply module and provided by the market module. The global market module is a spatial, non-stochastic global multi-commodity model for about 60 primary and processed agricultural products, covering about 80 countries in 40 trading blocks. It is defined by a system of behavioural equations representing agricultural supply, human and feed consumption, multilateral trade relations, feed energy and land as inputs, and the processing industry; all differentiated by commodity and geographical units. Agricultural land types of temporary and permanent cropland as well as fodder areas (i.e. fodder on arable land, as well as grass and grazing) have different land rental prices and land owners are responding to these rental prices according to a multinomial logit land supply system in the model version used in this study. Land types are allocated to single activities using yield elasticities to disaggregate the total supply response into contributions from yields and from areas. The shares of broad land types in the country area respond to rents in a multinomial logit system. Bilateral trade and attached prices are modelled based on the Armington approach. CAPRI endogenously calculates EU agricultural emissions for nitrous oxide and methane based on the inputs and outputs of production activities, taking specific technical GHG mitigation options into account. GHG emissions for the ROW are estimated on a commodity basis in the CAPRI market model (Pérez Domínguez et al 2016, 2020, Fellmann et al 2018).

The GLOBIOM (Havlík et al 2014) is a partial equilibrium model that covers the global agricultural and forestry sectors, including the bioenergy sector. Commodity markets and international trade are represented at the level of 58 economic regions (including EU28 individual member states) in the model version applied in this study (Frank et al 2015). Prices are endogenously determined at the regional level to establish market equilibrium to reconcile demand, domestic supply and international trade. For crops, livestock, and forest products, spatially explicit Leontief production functions covering alternative production systems are parameterized using biophysical models like Environmental Policy Integrated Model (Williams 1995), Global Forest Model (Kindermann et al 2008, Gusti 2010), or the RUMINANT model (Herrero et al 2013). For the EU, the model represents eight different livestock production systems, three tillage systems as well as alternative crop rotations for the EU (Frank et al 2015). For the present study, the supply side spatial resolution was aggregated to 2° (about 200 × 200 km at the equator) and Nomenclature of Territorial Units for Statistics (NUTS2) level for the EU. Land and other resources are allocated to the different production and processing activities to maximize a social welfare function which consists of the sum of producer and consumer surplus. The model includes six land cover types: cropland, grassland, short rotation plantations, managed forests, unmanaged forests, and other natural vegetation land. Depending on the relative profitability of primary, by-, and final products, the model allocates production activities and can switch from one land cover type to another. Spatially explicit land conversion over the simulation period is endogenously determined within the available land resources and conversion costs that is considered in the producer optimization behaviour. Land conversion possibilities are further restricted through biophysical land suitability and production potentials, and through a matrix of potential land cover transitions. GLOBIOM covers major GHG emissions (CO₂, N₂O, and CH₄) from agricultural production, forestry, and other land use and different mitigation options for the land use sector. Detailed information on the parameterization of the different mitigation options for the agricultural sector is provided in Frank et al (2018). More information on the GLOBIOM can be accessed at https://iiasa.github.io/GLOBIOM/.

The MAGNET model is a multi-regional, multi-sectoral, applied general equilibrium model based on neo-classical microeconomic theory (Nowicki et al 2009, Woltjer and Kuiper 2014). It is an extended version of the standard Global Trade Analysis Project (GTAP) model (Hertel 1997). The core of MAGNET is an input–output model, which links industries in value added chains from primary goods, over continuously higher stages of intermediate processing, to the final assembly of goods and services for consumption. Primary production factors are employed within each economic region, and hence returns to land and capital are endogenously determined at equilibrium, i.e. the aggregate supply of each factor equals its demand. On the consumption side, the regional household is assumed to distribute income across savings and (government and private) consumption expenditures according to fixed budget shares. Private consumption expenditures are allocated across commodities according to a non-homothetic constant difference of elasticities (CDE) demand system and the government consumption according to Cobb-Douglas expenditure function.

The MAGNET model, in comparison to GTAP, uses a more general multilevel sector specific nested constant elasticity of substitution production function, allowing for substitution between primary production factors (land, labour, capital and natural resources) and intermediate production factors and for substitution between different intermediate input components (e.g. energy sources and animal feed components). MAGNET includes an improved treatment of agricultural sectors trying to better approximate physical constraints in the CGE context (like various imperfectly substitutable types of land, the land use allocation structure, a land supply function, substitution between various animal feed components (van Meijl et al 2006), agricultural policy (such as production quotas) and different land related payments (Nowicki et al 2009), and biofuel policy (capital-energy substitution, fossil fuels-biofuels substitution (Banse et al 2011). On the consumption side, a dynamic CDE expenditure function is implemented which allows for changes in income elasticities when purchasing power parity-corrected real gross domestic product (GDP) per capita changes. Segmentation and imperfect mobility between agriculture and non-agriculture labour and capital are introduced in the modelling of factors markets.

The model also incorporates emissions from the latest GTAP non-CO₂ database (Irfanoglu and van der Mensbrugghe 2015), including methane (CH₄) and nitrous oxide (N₂O). This is complemented by CO₂ emissions from the GTAP Energy-Environmental database (GTAP-E). Livestock non-CO₂ emissions and Rice CH₄ emissions are tied to the output variables of these respective sectors within the MAGNET model. Whereas N₂O emissions from crop fertiliser use are tied to the fertiliser input variable in these sectors. In addition, data on the marginal abatement costs associated with practices and technologies that can be used to reduce GHG emissions are also incorporated (Henderson et al 2019) based on the US EPA (2013). They cover measures for lowering the main non-CO₂ emission sources including methane from enteric fermentation by ruminants (i.e. cattle, sheep and goats), nitrous oxide and methane from livestock manure, methane emissions from paddy rice and nitrous oxide emissions from soil associated with fertiliser use by crops.

In the baseline scenario, EU agricultural non-CO₂ emissions (CH₄ from enteric fermentation, manure management and rice cultivation, N₂O from manure management, manure applied and dropped on soils, synthetic fertilizers and crop residues) are projected to remain rather stable until 2050 in CAPRI and GLOBIOM, whereas in MAGNET an increase past 2030 is observed related to methane emissions associated to ruminant beef production driven by increasing exports. By 2050, agricultural non-CO₂ emissions are projected to range from around 400–420 MtCO₂eq yr⁻¹ (GLOBIOM, CAPRI) up to 460 MtCO₂eq yr⁻¹ (MAGNET) with the livestock sector accounting for around 80% (75% of which CH₄ and 25% N₂O) and crops for around 20% (95% of which N₂O and 5% CH₄) of the total emissions (figure 2) which is consistent with current shares reported by Food and Agriculture Organization Corporate Statistical Database (FAOSTAT) and UNFCCC. CAPRI and GLOBIOM are well in line with FAOSTAT and GAINS projections that both anticipate a decline of EU agricultural non-CO₂ emissions to around 400 MtCO₂eq yr⁻¹ by 2050 (EC 2018).

Zoom In Zoom Out Reset image size

Emissions developments are related to underlying projections in activity data and changes in emission intensity. For ruminant meat, a decline of around 7% in EU production is projected in CAPRI and GLOBIOM by 2050 following the continued decrease in domestic demand, whereas MAGNET projects +31% production growth by 2050 as compared to 2010. MAGNET results are mainly driven by exports to the ROW (mostly to African countries) caused by huge increase in demand outside the EU driven by population and substantial income growth in an SSP2 world without tighter environmental policies in the future. Production of milk (+9% to +28%), pig and poultry meat (+7% to +20%) is expected to grow across all models driven by increasing domestic and international demand. Since these changes in production levels are accompanied by increasing GHG efficiencies, net impacts on emissions are compensated and sometimes even offset. GHG efficiency increases for dairy in CAPRI (+18% milk, −7% beef) and GLOBIOM (+11% milk, +19% beef) result in an overall stabilization of livestock related non-CO₂ emissions in these models despite the decrease in beef efficiency in CAPRI whereas livestock GHG efficiencies in MAGNET remain rather stable (−1% milk, +1% beef) as no carbon prices or other environmental constraints are implemented in the SSP2 scenario over time thus resulting in an increase in livestock emissions by 19% (figure 2(c)). GHG efficiencies related to crop production increase in CAPRI and GLOBIOM by around 11%–15% up to 32% in MAGNET causing on average across models a 11% decline of crop related N₂O emissions by 2050 despite increases in EU production levels.

To assess the impacts of a unilateral EU mitigation policy (00% buy-in scenario) where only the EU pursues ambitious agricultural mitigation efforts (carbon price of 245 USD tCO₂eq⁻¹ by 2050) compatible with the 1.5 °C target, we compare scenario results to the baseline scenario without mitigation efforts. This allows us to quantify the impacts in- and outside the EU and potential emission leakage effects to other world regions. The 00% buy-in scenario results on average across models in domestic EU agricultural non-CO₂ emission reduction of 155 MtCO₂eq yr⁻¹ (−36%) compared to the baseline scenario in 2050. Across models, GLOBIOM shows the lowest EU non-CO₂ emission savings (130 MtCO₂eq yr⁻¹, −32%), while CAPRI (165 MtCO₂eq yr⁻¹, −39%) and MAGNET (170 MtCO₂eq yr⁻¹, −37%) anticipate higher absolute agricultural GHG abatement (figure 3). The presented mitigation potentials are slightly more optimistic than results from the GAINS model as presented in the EU's Long Term Strategy (EC 2018), which show a mitigation potential of around 130 MtCO₂eq yr⁻¹ (−33%) from the adoption of technological options. This difference can be explained by the consideration of price feedbacks (of carbon pricing) on activity levels in the agricultural sector models as well as the different parameterization of technical mitigation options and in case of MAGNET with different baseline assumptions. All models applied in this study anticipate the highest abatement potential being realized in the livestock sector, which contributes on average 140 MtCO₂eq yr⁻¹ (90% of total GHG abatement) in 2050, mainly through ruminants (beef and milk contribute 75% of the total) while crops are estimated to have more limited emission reduction potentials.

Zoom In Zoom Out Reset image size

Across gases, the models project the highest mitigation coming from methane which provides around 65% of the total GHG abatement. This is more optimistic as compared to GAINS projections (EC 2018) which expects only around 40% of the total non-CO₂ mitigation potential reduction to come from CH₄. With respect to the mitigation option portfolio, technological options such as anaerobic digesters, feed supplements or precision farming, are identified across models as a key technology for non-CO₂ savings, followed by changes in livestock production levels while structural options such as intensification in production systems, seem to be of more limited relevance for the EU.

The unilateral EU mitigation efforts adopted in the 00% buy-in scenario yield smaller emission savings at the global scale as compared to the EU internal mitigation effect. The reason is that domestic EU emission savings are partly compensated by additional emissions outside Europe as part of the EU production is reallocated to regions outside Europe. At the global scale emission leakage decreases domestic EU emissions savings on average by 40% across models. Hence, the 00% buy-in scenario achieves only average emissions savings of around 90 MtCO₂eq yr⁻¹ at global scale (75–100 MtCO₂eq yr⁻¹ across models).

Regional production reallocation and leakage effects in response to the unilateral mitigation policy vary across models (figure 4). Emission leakage is highest in MAGNET (56%) while it is more moderate in CAPRI (39%) and GLOBIOM (27%). Across models, emission leakage mostly occurs for beef while EU dairy related emission reductions are not compensated by increasing emissions in the ROW (except for some increase in Africa according to GLOBIOM). Differences in leakage rates can be mainly explained by different baseline trends. Since the EU is a large net beef exporter in MAGNET in the baseline scenario as compared to more modest net beef exports in CAPRI and GLOBIOM, this allows for higher substitution of EU beef exports and reallocation of domestic production to other world regions under a unilateral EU mitigation policy resulting in higher emission leakage.

Zoom In Zoom Out Reset image size

For beef, the models anticipate a stronger increase in emissions in Latin America induced by increasing exports to the EU and Africa induced by decreasing imports from the EU and limited emission leakage to regions like North America and Other Asia including Oceania with GHG efficient production systems. Results are in line with current trade patterns where Latin America is the main exporter of beef to the EU while around 25% of EU beef exports are currently exported to African countries (according to FAOSTAT). Across models, a fragmentation of international trade (doubling of tariffs) is found to decrease emission leakage and sustain higher emission savings of 85–110 MtCO₂eq yr⁻¹ at global scale.

As production costs of GHG intensive products increase for EU farmers and consequently beef production decreases on average by −20% in response to the unilateral EU mitigation policy (−18% to −24% across models), farmers in the ROW benefit. They become relatively more competitive compared to the EU since they are not targeted by the mitigation policy and increase their production to compensate for the drop in the EU. Consequently, EU is becoming a net beef importer, 40% of it coming from Latin America.

Having quantified potential leakage effects in response to a unilateral EU agricultural mitigation policy, we now assess the impact of ROW mitigation efforts on EU farmers while the EU continues to pursue ambitious mitigation efforts. Figure 5 presents the economic impact (supply, prices, and production value) on EU producers with increasing ROW mitigation efforts (figures S2 and S3 available online at stacks.iop.org/ERL/16/104038/mmedia in the supplementary material provide additional details). ROW mitigation efforts are found to ease the impacts of a unilateral EU mitigation policy on EU farmers. Without any mitigation efforts in ROW in the 00% buy-in scenario, the models anticipate on average a strong decline for beef production (−20%), whereas impacts on oilseeds (−14%), coarse grains (−13%), wheat (−12%), milk (−9%) and non-ruminant production (−6%) are less pronounced. Agricultural prices increase in response to the carbon price, but so do production costs. Furthermore, the reduced production quantities are reflected in a lower production value across commodities as compared to the baseline scenario without mitigation efforts in 2050. This indicates that farmers will experience economic losses in response to the unilateral EU mitigation policy.

Zoom In Zoom Out Reset image size

However, once the ROW starts to pursue agricultural mitigation efforts next to the EU, European farmers are less affected as they can produce more GHG efficient food compared to other ROW regions (figure 6). As production costs for farmers outside the EU start to increase with increasing ROW mitigation efforts, in particular for those with high GHG intensity, EU farmers are able to regain market shares. Consequently, negative EU impacts across all product groups get buffered as compared to the unilateral EU mitigation policy. For beef and non-ruminants, the production value even exceeds the baseline levels if the ROW at least pursues 25% – 50% mitigation efforts (figure 5). Pursuing 1.5 °C compatible mitigation efforts globally (full buy-in scenario) would further decrease the economic impacts on EU producers. However, the less pronounced EU production decreases with increasing ROW mitigation efforts are also mirrored in the EU non-CO₂ emission savings which decreases to 115 MtCO₂eq yr⁻¹ (−27%) on average across models in the full buy-in scenario.

Zoom In Zoom Out Reset image size

Extending the carbon price on agricultural emissions to the ROW drives GHG efficiency improvements also outside the EU. Especially regions with projected high GHG intensity in 2050 like India, Africa, Latin America, or Other Asia can already with moderate efforts substantially decrease their agricultural GHG intensity (figure 6) and consequently reduce non-CO₂ emissions. For example, the 10% buy-in scenario (carbon price of 25 USD tCO₂eq⁻¹) allows to reduce the global GHG intensity for crops and livestock between 19% and 25% across models. At the same this allows to deliver already around 50% (1.6 GtCO₂eq yr⁻¹) of the total non-CO₂ emission savings from the full buy-in scenario (3.3 GtCO₂eq yr⁻¹), in the 50% buy-in scenario even 85% (2.8 GtCO₂eq yr⁻¹). This shows that a substantial amount of the agricultural non-CO₂ mitigation potential can already be realized at low economic costs. With increasing effort level, GHG efficiency increases and thus the marginal rate of additional emission savings decline (figure 6). At global scale, the ruminant sector is again one of the most important sources for GHG mitigation accounting for more than half (around 55%) of the total global GHG abatement, followed by the crop sector (around 25%) and dairy producers (around 15%).

Figure 7 displays in relative terms the change in production and emissions levels for regions when moving from low to higher ROW mitigation efforts and illustrates the sensitivity of the supply and mitigation response in each model with respect to increasing mitigation efforts. Results show that farmers in developed regions like EU and North America are less impacted or may even start to benefit from a more ambitious mitigation policy as compared to regions of the Global South with less GHG efficient production systems such as India, Latin America, and African countries (figure 6). In these regions, agricultural prices tend to increase more drastically in response to the global carbon price when moving towards high mitigation efforts (full buy-in scenario). Thus, extending the mitigation policy to the ROW buffers not only impacts on EU producers but balances the impacts across world regions. Across models, CAPRI is shown to have the most inelastic supply side parameterization outside the EU as increasing ROW mitigation efforts results only in small agricultural production losses in these regions of up to −4% (−4% for crops and −10% for livestock products). In contrast, GLOBIOM (up to −24% agricultural production decrease) and MAGNET (up to −12% agricultural production decrease) show much more elastic supply side response to the agricultural mitigation policy.

Zoom In Zoom Out Reset image size