Enhancing Climate Change Mitigation through Agriculture

4. Global potential of supply-side and demand-side mitigation options

Abstract

This chapter analyses how agriculture could moderate changes to the climate by simulating supply- and demand-side mitigation strategies. Based on the Aglink-Cosimo model as used for the OECD-FAO Agricultural Outlook 2018-2027 baseline, only direct emissions that result from agricultural crop and livestock production activities are taken into consideration.

Introduction

FAOSTAT reports that 11% of global GHG emissions in 2010 came directly from agricultural production. The subcategories of agricultural emissions distinguished in FAOSTAT are:

• Enteric fermentation

• Manure management

• Rice cultivation

• Synthetic fertilizers

• Manure applied to soils

• Manure left on pasture

• Crop residues

• Cultivation of organic soils

• Burning – crop residues

• Burning – savannah

This percentage does not include emissions that result from converting land, e.g. forest areas to grass or cropland that are categorised in the Land Use, Land-Use Change and Forestry (LULUCF) sector, which accounts for an additional 11% of global GHG-Emissions.

At present, the Aglink-Cosimo model can only capture the first category of emissions.1 About 90% of those direct emissions are captured, as not all production activities are represented and emissions from burning crop residues and savannah, as well as organic soils and crop residues are not accounted for.

Of the emissions studied in this chapter, 73% can be attributed to the ruminant sector, 6% to non-ruminant meat production, 14% to rice cultivation, and 7% to other crops. Geographically, the highest absolute emissions are located in the People’s Republic of China (hereafter “China”) (15%), India (13%), Brazil (9%), the European Union (8%), and the United States (7%). These four countries and the European Union account for over 50% of agricultural GHG-emissions. OECD countries are responsible for about 30% of global emissions.

Figure 4.1 reveals the importance of the ruminant sector in global GHG emissions, implying that there is also significant potential to reduce emissions in this sector. Nevertheless, the 2018 edition of the OECD-FAO Agricultural Outlook (OECD/FAO, 2018[2]) projects that demand for products produced by ruminants will continue to increase up to 2027 for most countries. Global meat production is projected to be 15% higher in 2027 relative to the base period, and the projected output growth is expected to occur predominantly in developing countries. The most rapid expansion is expected to occur in the poultry sector. Consumers in developing countries are expected to increase and diversify their consumption towards more expensive meats, including beef and sheep meat, which have higher emissions per unit of output.

World milk production is projected to increase by 22% over the projection period, with over half the increase originating in Pakistan and India, two countries where emissions per litre of milk are relatively high. Both countries are expected to jointly account for 32% of global milk production by 2027.

Figure 4.1. Composition of GHG emissions, 2010

Note: Energy” includes energy, manufacturing and construction industries and fugitive emissions, “Other” includes residential, commercial and institutional emissions, and emissions from industrial processes and product use, waste international bunkers, and other non-specified sources.

Source: FAOSTAT and Aglink-Cosimo database.

These projections lead to an increase in total GHG Emissions throughout the baseline projection period as shown in Figure 4.2. Total emissions from agriculture are projected to increase by 540 MtCO₂ equivalents (CO₂eq) between the outlook base period (average 2015-2017) and 2030, with about 80% of this increase stemming from the ruminant sector (440 MtCO₂eq), and within that sector methane (CH4) emissions from enteric fermentation will account for 300 MtCO₂eq (70%).

The distribution of total emissions from agriculture varies widely across countries (Figure 4.3). China, India and the rest of South-East Asia account for over 40% of global agricultural GHG emissions at present and are projected to account for 70% of the increase in global GHG emissions. Large increases in emissions are also expected from Sub-Saharan Africa and Brazil.

This baseline represents a business as usual scenario where no additional actions towards emission savings are undertaken, although existing mitigation efforts that are visible in past trends of emissions per production unit are taken into account. This report does not deal with the impact of climate change on the agricultural sector but the overall impact over the analysed time horizon is limited. Hasegawa et al. (2018[3]) showed that even by 2050, the average impact of climate change is expected to be low compared to the potential mitigation impact of agriculture. It should be noted, however, that the impact of potential increases in extreme events has not taken into account.

Figure 4.2. Total GHG emissions from agriculture (MtCO₂eq)

Total emissions do not include emissions from products not covered in Aglink-Cosimo

Source: Calculations based on OECD/FAO (2018), “OECD-FAO Agricultural Outlook”, OECD Agriculture statistics (database), http://dx.doi.org/10.1787/agr-outl-data-en.

Figure 4.3. Regional differences in total GHG emissions from agriculture (MtCO₂eq)

Total emissions do not include emissions from products not covered in Aglink-Cosimo

Source: Calculations based on OECD/FAO (2018), “OECD-FAO Agricultural Outlook”, OECD Agriculture statistics (database), http://dx.doi.org/10.1787/agr-outl-data-en.

Box 4.1. Aglink-Cosimo’s contribution to climate change analysis

The main strength of the Aglink-Cosimo model is its level of detail on agricultural commodity markets and ability to capture interactions with market policies. It can outline the implications of changes in exogenous drivers and policy assumptions for market outcomes. This is exploited here from a new perspective. The Aglink-Cosimo baseline projections are carefully linked to historical developments, so that future projections start from where the world is now (a feature absent in many models). This helps provide a clearer benchmark to assess future contributions to climate change mitigation.

In order to obtain estimates of the emissions that are directly produced by the agricultural sector from each model run, emission factors per commodity produced were inherited from IIASA’s Globiom model which are based on the IPCC guidelines at the tier 1 level and broadly consistent with the FAOSTAT database. The resulting coefficients are not static, but include the most recent trends of emission intensities at the country/region level.

Technological mitigation options are incorporated into the analysis (Scenario 5) using regional marginal abatement cost curves (MACC) for the different direct agricultural emissions. These MACCs were obtained from the technological adjustment behaviour of the Globiom model.

The combination of those two recent developments allow reporting on direct agricultural emissions, as well as analysis of supply-side mitigation scenarios. This analysis will, however, be partial as it does not include the effects on emissions created during the processing of food products, nor those arising from producing inputs for agriculture. Furthermore, changes to input quantities are only implicitly accounted for.

The Joint Research Centre of the European Commission recently published a report on the economic impacts of a low carbon economy on global agriculture using a similar approach to the Aglink-Cosimo model (Jensen et al., 2019_[4]). These two methodologies will be merged in future studies.

The baseline scenario is then compared to alternative scenarios, which address three possible ways to mitigate emissions: reduce the share of food consumed from ruminants; reduce food waste; and impose carbon taxes and improve productivity on the production side.

This analysis, however, also addresses potential trade-offs between food security and emission reduction. Insofar as measures to reduce GHG emissions from agricultural production lead to lower food output or increases in food prices, there may be trade-offs between the dual goals of guaranteeing food security and reducing GHG-emissions.

The FAO definition of food security – encompassing the four dimensions of availability, accessibility, utilization, and stability – is used here.2 Aglink-Cosimo partially addresses the availability and access dimensions via projections for national availability and food prices. Three indicators can be calculated based on the scenario outputs.

• Calorie Availability Index (Availability): The average amount of calories available per capita in each country for the subset of the food basket represented in the model.3

• Consumer Food Price Index (Accessibility): Calculated as a fixed weight index of national consumer prices in real terms. Food consumption quantities in 2015 are used as weights. Higher consumer prices are assumed to lead to lower access to food for parts of the population.

• Agricultural Gross Income Index (Accessibility): Calculated as a fixed weight index of a combination of producer prices and subsidies normalised by an input price index. Agricultural production quantities in 2015 are used as weights. This indicator is most relevant in countries where the agricultural sector is a large contributor to the national GDP.4

Figure 4.4 shows how under the baseline scenario the three food security indicators and emissions from the agricultural sector are projected to evolve relative to their 2015 values. Two of the three indicators show a positive development over the projection period. Prices for consumers, measured by the consumer price index, are projected to decrease in real terms indicating improvements in accessibility. Calorie availability is also projected to improve over the next decades. However, the agricultural income index (in real terms) is expected to decline strongly due to the projected increase in input prices over the outlook period, while real output prices decrease in general. This corresponds to the historical tendency for real agricultural prices to decline over time, exerting income pressure on farmers who are not participating in the productivity gains that drive lower prices. Emissions from agriculture are projected to be 11% higher in 2030, as compared to 2015.

Figure 4.4. Baseline development of global food security indicators and emissions

Source: Calculations based on OECD/FAO (2018[2])

Scenarios to reduce GHG emission

Reducing the consumption share of food produced from ruminants

Ruminant products covered in this analysis contain beef, sheep meat, as well as butter, cheese, fresh dairy milk products, skimmed milk and whole milk powders. Ruminant products account for 70% of the emissions analysed in this report. Strategies to reduce emissions from these products are more promising than for many other food commodities. This section examines the food security and emission impacts of two scenarios that reduce the level of final food consumption of these products.

Scenario definition

Scenario 1 assumes that the average per capita consumption of each of the ruminant products is gradually reduced to reach levels in 2030 that are 10% below the values of 2017. This scenario also assumes that consumption of non-ruminant products increases by the same proportion across all commodities, such that over that same period the same average per capita consumption of total calories is maintained as under the baseline scenario.

In Scenario 2 a consumer demand tax of USD 605 (in year 2000 real USD) per ton of CO₂eq emitted by each product is applied globally, taking country specific emission factors for primary agricultural products into account.6 Given their higher emissions intensity, this measure will result in much higher consumer price increases on ruminant products than for the other commodities. However, Scenario 2 does not address emissions that occur between the farm gate and the final consumer,7 and contrary to Scenario 1, its implementation does not target only the ruminant sector. In addition, the analysis does not include the potential effects arising from redistribution of the collected tax money.8 Both scenarios are applied to all countries with the exception of the Least Developed Countries (LDCs).

Results

Figure 4.5 compares the food security indicators with agricultural emissions saving between Scenarios 1 and 2. The GHG emission savings by 2030 are projected to be much higher under Scenario 1 (870 MtCO₂eq) than under Scenario 2 (160 MtCO₂eq). This can be explained by a lower reduction in calorie consumption of ruminant products. In Scenario 1, the average global per capita consumption of ruminant products in 2030 is reduced from 265 to 197 kcal/cap/day, a reduction of roughly 70 kcal/cap/day. In Scenario 2, this is reduced to 12 kcal/cap/day.

Figure 4.5. Emission savings versus food security in relation to the baseline scenario

Source: Aglink-Cosimo simulation results.

Scenario 1 also performs better in terms of the consumer price index. This index is projected to increase strongly in Scenario 2 because taxing food globally makes food more expensive, putting accessibility to food at risk. Under Scenario 1, the index is expected to decrease because lower demand leads to lower consumer prices, not only for ruminant products but also for food commodities used to feed ruminants. Agricultural income is more affected in Scenario 1 because reduced demand for ruminant products is more pronounced here than under Scenario 2. This decrease not only reduces the producer price of ruminant products, but feed demand as well. Global cereal demand under both scenarios is, for example, lower than under the baseline, implying that in Scenario 1 the effect of reducing feed demand through lower ruminant product consumption dominates the increase in food consumptions of cereals.

Figure 4.6 illustrates the projected relative changes in the consumer price index, the calorie availability index, and agricultural emissions under both scenarios for selected countries. These countries were chosen based on their importance in global agricultural GHG emissions. The two LDC aggregates are included even though these countries were exempt from any mitigation obligations under the two scenarios; nevertheless, they will be affected via global market impacts. Note that calorie availability is not subject to significant changes in either scenario.9

Figure 4.6. Regional emission savings versus food security in relation to the baseline scenario, 2030

Source: Aglink-Cosimo simulation results.

A consumption tax (Scenario 2) is projected to reduce emissions to a lesser extent than Scenario 1 in all selected countries, except for Brazil. The main reasons for the relatively stronger emission-reducing effect of Scenario 2 in Brazil are: it has relatively high emission coefficients for beef production, which translate into higher taxes; beef prices are relatively low so that the new tax accounts for more than 50% of the consumer price (the share would be only 10% in the United States); and, demand in emerging economies such as Brazil is more elastic to price changes than in developed countries.

Emission savings in Scenario 2 are associated with a large cost burden for consumers in terms of consumer prices. Even though small, the spillover effects to least developed countries are positive in terms of cheaper food and reduced emissions. The latter effect is due to lower international producer prices for the ruminant products, which lead to a decrease in production of those products in LDC countries and an increase in imports.10 It is clear that a shift in preferences as simulated in Scenario 1 would lead to lower concerns regarding food security and to higher emission savings than would the introduction of a consumption tax (Scenario 2).

The impact of food waste on GHG emissions

Throughout the supply chain, food is lost or wasted. In general, a distinction is made between the two. Food loss occurring on the supply side is defined by FAO (2011[5]) as food that gets spilled or spoilt before it reaches its final product or retail stage. Food waste occurs at the retail and household levels.

This chapter focuses on the food waste issue, which is considered a part of food demand. Reducing food waste has many benefits but also carries costs. For example, a restaurant owner knows how much food is thrown away every day and knows there are ways to reduce the amount of food wasted, but the costs of applying those measures might be higher than the economic benefits. Similarly, a retail store could reduce the amount of fruits and vegetables thrown away at the end of a day by investing in better cooling systems, but might consider it more profitable not to do so. Finally, a family could reduce food that is thrown away by buying less food during each grocery trip at the cost of shopping more frequently.

Only a few studies examine the impact of reducing food waste and loss at the global level. In Okawa (2015[6]), the medium-term market impacts of reducing food waste and food loss are examined based on the OECD-FAO Agricultural Outlook 2014-2023 projections for world and national agricultural markets. The study applies FAO’s region-specific estimates of producer loss and consumer waste, which are reduced by 20% over ten years on the assumption that these reductions can be achieved without cost. The study finds a greater impact on international markets due to contractions in demand via reduced waste than from the stimulus to supply from lower losses. Savings to consumers total more than USD 2.5 trillion over ten years and reduced crop losses in developing countries lead to higher crop supplies in these countries, with reduced prices from efficiency gains benefiting both developing and developed countries. However, the analysis in this study does not consider the potential environmental impact of reducing food loss and waste.

In order to asses this maximal GHG emissions abatement potential of reducing food waste, the analysis in Okawa (2015[6]) was repeated, focussing on the food waste aspect on the demand side and using the 2018 version of the Aglink-Cosimo model. Okawa (2015[6]) uses the food loss and food waste estimates published in FAO (2011[5]), which are the only estimates currently available on a global scale. The estimates in the FAO study show the shares that are lost or wasted for seven agricultural product groups (cereals, roots and tubers, oilseeds and pulses, fruits and vegetables, meat, fish and seafood, and milk) at seven regional aggregates (Europe including the Russian Federation, North America, and Oceania, Industrialised Asia, West and Central Asia, South and Southeast Asia, Sub-Saharan Africa, North Africa, and Latin America). Those shares are defined for five levels of the supply chain (agricultural production, post-harvest handling and storage, processing and packaging, distribution, and consumption at household level). Figure 4.7 shows the shares of food waste in the distribution and consumption at household levels from that study.

Figure 4.7. Food waste shares

Note: Only waste shares = Sum of distribution and household in FAO (2011[5]). Fish, seafood, fruits and vegetables are not included since they cannot be addressed in this chapter.

Source: FAO (2011[5]).

As indicated above, this section considers the food waste issue as defined by the FAO. As such, only two supply chain levels matter: distribution and consumption at the household level. The food use variable in Aglink-Cosimo implicitly includes waste at these two levels and can therefore be interpreted as food availability. It is assumed that food availability and final food demand are the same, as waste no longer occurs. But since actual waste levels are unknowns in the Aglink-Cosimo database, two assumptions are necessary:

• the waste rates from FAO (2011[5]) are applied to all countries within the seven regional aggregates

• the waste rates from FAO (2011[5]) are applied to all commodities within the five product groups

The LDCs are exempted from all mitigation efforts.

Scenario definition

Scenario 3 assumes that the wasted quantities as defined above disappear linearly between 2018 and 2030 by shifting the product-specific food demand equations to the left (as in the left graph of Figure 4.9). This scenario makes no assumptions about the costs that may be associated with reducing waste. Even though the assumption of zero costs for reducing food loss and waste is unrealistic, such a scenario can be seen as the theoretical upper bound of the potential impacts.

Scenario 4 has a set-up similar to Scenario 3, but incorporates the cost aspects of reducing waste. Under this scenario, it is assumed that the costs of waste reduction increase exponentially, i.e. it costs relatively less to reduce the first rather than the final units of waste, and the cost of eliminating the final unit of waste is set equal to the amount the consumer pays for one unit of the respective commodity in the baseline. It is further assumed that the costs to reduce waste will be manifested in higher prices for consumers. This assumption can be justified as follows: a restaurant will transfer its waste reduction costs to its clients, or improved (and more expensive) packaging at the retail level that improves the storage lifetime of products would have to be paid by consumers. Reducing waste at the household level will not lead to higher observed consumer prices. However, it can be assumed that the perceived consumer prices are indeed higher as the household has to pay an implicit mark-up on each consumption unit to reduce waste. For example, if a household reduces waste by increasing shopping frequency to avoid the deterioration of fresh food, this imposes costs (e.g. fuel costs to drive to the supermarket, opportunity costs of the additional time spent in the supermarket), and for simplification these costs are added to consumer food prices. This assumption illustrates that incorporating the costs of reducing waste significantly changes the cost-benefit calculation of reducing emissions by reducing waste.11

Results

Under Scenario 3, eliminating food waste would reduce the agricultural part of GHG-emissions by 8% or 440 MtCO₂eq by 2030. The reduction would be even higher in Scenario 4: 14% or 800 MtCO₂eq. At a first glance, it appears surprising that reflecting costs to reduce waste increases the emission reduction potential. However, as those costs increase consumer prices, the demand-reducing effect of higher expenditures for food reduces production beyond the levels of Scenario 3.

Figure 4.8. Emission savings versus food security in relation to the baseline scenario

Source: Aglink-Cosimo simulation results.

Figure 4.8 shows how the three food security indexes and the emission savings are projected to evolve under Scenarios 3 and 4 relative to the baseline scenario. Under Scenario 3, consumer prices in 2030 are projected to be 10% lower than under the baseline indicating a positive impact on food accessibility. However, a strong negative effect on agricultural incomes can be observed as producer prices decrease strongly because of lower demand. The calorie availability index indicates that slightly more calories are available for final consumption as compared to the baseline. However, Scenario 3 assumes that the consumer does not have to pay for waste reduction. In Scenario 4, where this assumption is abandoned, the results are very different: the consumer price index increases strongly over time,12 while the agricultural income index and calorie availability decrease.

Clearly the different outcomes under Scenarios 3 and 4 are a direct result of the assumptions regarding the cost of food waste reduction. Even though Scenario 3 is less realistic, it does illustrate that reducing waste without keeping the cost aspect in mind underestimates the emission-saving potential as well as the negative impact on food security. It further underlines that a better understanding of the costs of waste reduction is needed in order to assess the trade-offs between emission-reduction targets and food security issues.

The significant difference between the two scenarios is also apparent in comparisons across major countries. Calorie availability is projected to be much lower under Scenario 4 and accessibility is at risk with much higher consumer prices. (Figure 4.9). The spill-over effect to LDCs is positive in terms of reduced consumer prices and reduced emissions. However, it should be noted that Scenario 4 would perform better in terms of the food security indicators if it were assumed that waste is not reduced to the final unit. The first units of waste reduction are assumed relatively “cheap” and therefore will not affect demand and prices as much as shown in this chapter.

Figure 4.9. Regional emission savings versus food security in relation to the baseline scenario, 2030

Source: Aglink-Cosimo simulation results.

Production-side mitigation

Blanford et al. (2018[1]) identify supply-side mitigation options in agriculture, excluding output reduction, in the form of changes in farming practices and land management that target emissions per unit of input (land or animal), as well as changes in technical and managerial efficiency and technology that lower emissions per unit of output (productivity improvements). The potential for improvement with respect to the practices and management aspects in particular can be found in the heterogeneity of livestock systems across farms (Box 4.2). The major assumption here is that if carbon emissions were priced, this would act as an incentive for farmers to transfer their systems to use less GHG-emitting practices.

Based on the work described in Box 4.2, the Aglink-Cosimo model was adjusted in several. The revised model not only captures the consumption change component of mitigation, but also incorporates technological options and structural changes by introducing dynamic emission coefficients depending on the applied carbon tax level.

Scenario definition

Scenario 5: A production tax of USD 60 (real USD, 2000) per tonne of CO₂eq emitted is imposed on the agricultural production activities by shifting the supply curves for each commodity upwards by the amount of the tax. The individual tax rates per tonne of product thereby differ because beef production has higher emission coefficients compared to wheat for example. The applied emission coefficients in this scenario differ from those in the baseline due to the assumption that a carbon tax will lead to technological and structural mitigation (Figure 4.11) that cannot be explicitly modelled with Aglink-Cosimo, thus reducing the emissions per production unit. The choice of the carbon tax level is the same as that applied in Scenario 2.

Scenario 6: A productivity shift of 10% by 2030 is implemented for all products, linearly increasing from 2018. This means that yields for crops are increasing as well as the output of meat and dairy products per animal. This increase is assumed to be achieved at no cost.

Box 4.2. Heterogeneity of the production system as a source of climate change mitigation in agriculture

Climate change mitigation in agriculture can be modelled through changes on the consumer or producer side. If the agricultural sector were represented in a simplified way whereby each production activity is associated with the same greenhouse gas (GHG) emission coefficient, then reducing the consumption of GHG intensive products would be the only mitigation option. In order to include mitigation options related to the production side, technological options have been incorporated. Examples of technological options are propionate precursors and antimethanogen vaccinations to reduce CH4 emissions from enteric fermentation, or anaerobic digesters to reduce CH4 emissions from manure management. These options are sometimes referred to as add-on technologies since their mitigation potential is typically calculated in addition to current production activities. Another option is to change production to more efficient systems. This structural change option was overlooked for a long time in large-scale assessments, but is valuable as it incorporates the heterogeneity in production systems in terms of their GHG efficiency.

This box presents an overview of recent studies on the structural change option. Most of these studies focus on the livestock sector as it is responsible for 65% of agricultural non-CO₂ emissions. In 2013, Herrero et al. published a detailed dataset describing livestock production systems worldwide in terms of their productivity, feed rations, and GHG emissions. The dataset quantifies the differences in GHG efficiency across individual products, from the relatively high efficiency of poultry production to the relatively low efficiency of beef production. In addition, the study shows that large differences exist for the same product across alternative production systems within the same region and across regions. For example, in Europe 17 kg protein of beef per tonne CO₂eq can be produced in the temperate agro-ecological zone with sufficient concentrate feed supplementation, while only 8 kg protein of beef per tonne CO₂eq can be produced in the grazing systems in the same agro-ecological zone. Similarly, in a comparable system with concentrate supplementation in Australia, typically only 10 kg protein of beef per tonne CO₂eq is produced, 7 kg less than in Europe (Figure 4.10).

Figure 4.10. GHG efficiency for bovine meat from non-dairy cattle

Notes: Non-dairy cattle include here all cattle heads other than dairy cows and replacement heifers. Production systems: ANY – average across all production systems; LGA – grazing systems arid; LGH – grazing systems humid; LGT – grazing systems temperate/tropical highlands; MRA – mixed crop-livestock systems arid; MRH – mixed crop-livestock systems humid; MRT – mixed crop-livestock systems temperate/tropical highlands; URBAN – urban systems; Other – other systems.

Source: Herrero et al (2013[8]).

A key driver of the differences in GHG efficiencies is feed quality. Herrero et al. (2013_[8]) found that increasing the metabolisable energy content in feed, for example from 9 to 10 MJ per kg dry matter feed, reduced the emissions related to beef production from about 0.250 to 0.10 tonne CO₂eq per kg protein. This could be achieved through feeding practices that include less grazing and better quality feeds. These transitions are generally induced by changes in relative factor prices: because of the increased population density, land values are growing faster than the economic opportunity cost of labour.

Havlík et al. (2014_[9]) implemented two scenarios in the Globiom model using the Herrero et al. (2013_[8]) dataset to analyse the future dynamics of livestock production systems and their contribution to reducing GHG emissions. The first scenario is a dynamic one representing business as usual adaptation of the structure of livestock production systems to the future economic conditions. The second scenario is counterfactual whereby the structure of livestock production systems is fixed around the year 2000. In the dynamic scenario, 64% of all ruminants would be reared in mixed systems with feed supplementation in 2030 compared to 56% of ruminants in 2000 (the counterfactual scenario), representing an intensification in feeding strategies.

Under the dynamic scenario, total Agriculture, Forestry, and Other Land Uses (AFOLU) emissions over the period 2010-2030 would be 9% lower than in the counterfactual scenario, indicating that individual adjustments in the production system structure would lead to an average annual saving of 736 MtCO₂eq. In the dynamic scenario, the majority of GHG reductions came from changes in land use (-23%), while agricultural non-CO₂ emissions were reduced by less than 5%. Increased feed use efficiency in mixed systems with concentrate supplementation saved 176 million ha from pasture expansion, and limited cropland expansion to only 14 million ha in the dynamic scenario as compared to the counterfactual scenario.

The role of structural change as a mitigation option under a carbon price policy was also analysed by Frank et al. (2018_[7]) in an integrated framework which considered the technological options mentioned above and the consumption side response to the increased production cost. This study found that at a carbon price of USD 100 per tCO₂eq, the agricultural sector could decrease by 2.6 billion tCO₂eq annually non-CO₂ emissions originating from this sector.

At this level, GHG reduction through structural change, including the transition of livestock production systems, and the technological options would contribute 38% each to the total reduction, while a decrease of consumption in response to increased producer prices would provide the remaining 24% (Figure 4.11). This study also finds indirect benefits of GHG reduction in the agricultural sector from land use change (-0.7 billion tCO₂eq). These examples show that it is important to take into account production system heterogeneity for the baseline emission profile development and the agriculture sector GHG mitigation potential assessment.

Figure 4.11. Economic mitigation potential of non-CO₂ GHG emissions in agriculture

Source Frank et al (2018[7]).

Results

Applying a carbon tax to production reduces emissions from agriculture by about 850 MtCO₂eq in 2030. Globally, more than half of this reduction comes from the assumed improvements in technology adaptation and sub-national reallocation of production (Figure 4.12), while the rest results from changes in production and consumption levels. This reduction is considerably lower than what is presented in Box 4.2 for several reasons. The mitigation potential in Frank et al. (2018[7]) was reported for 2050, hence the baseline level of emissions to mitigate from was higher, and the lead time of the carbon tax was longer which allowed for more pronounced structural change and larger diffusion of mitigation technologies. Finally, the applied carbon tax was higher (USD 100/tCO₂eq) and the LDCs were not excluded from mitigation.

Figure 4.12. Global emission savings in Scenario 5

Source: Aglink-Cosimo simulation results.

The global emission reduction in Scenario 6 amounts to 340 MtCO₂eq (-6%). As expected, the effect on emission savings is lower than the introduced supply shock, as markets adjust to equilibria with lower commodity prices and higher consumed quantities compared to the baseline.

In Figure 4.13, the trade-off between food security indicators and emissions is shown for Scenarios 5 and 6. In Scenario 5 there are more dynamic effects in place than in the scenarios analysed so far because of the lagged response of production to price changes. This is the case for equilibrium prices and the agricultural income index which decreases significantly in the first part of the projection period, increases in the middle part, and levels out towards the end. In the first years, production decisions are already locked in (herd sizes, land allocation) based on expectations that did not include the policy change. Therefore prices do not change much; however, the additional costs in terms of the applied tax strongly reduce income during those years. In the following years, farmers react to the higher costs by reducing production, which then drives prices up and in turn leads to a recovery in production in the next period. By 2030, the increase in prices matches on average the tax applied to each product. Again, it is the consumer who pays most for the GHG emissions reduction. This is clear in the increased consumer price index, which reaches a similar level to that in the consumption tax scenario (Scenario 2).

It seems that once again average calorie availability is not affected strongly; however, the consumer price index increases significantly, putting accessibility at risk especially for the poor. This is different to Scenario 6, where emission reductions are relatively modest but calorie availability shows the strongest increase across all six scenarios. Consumer food prices decrease strongly, putting consumers in a more secure position. The downside of lower prices is visible in the agricultural income index that also decreases significantly, as the effect of additional production quantities through the productivity boost is overcompensated by reduced market prices.

The country comparison (Figure 4.14) shows that the gains in terms of emission abatement in Scenario 5 are higher in Brazil and China compared to the European Union and the United States. This is due not only to larger emission saving potential in the first two countries through technological and structural adjustments, but also because the initial emissions per tonne of product are much higher in these two countries. For example, emissions per produced tonne of beef in 2030 in the baseline amount to over 30 kgCO₂eq/kg in Brazil and China, while the European Union and the United States only emit about 10 kgCO₂eq/kg. The emission reduction in India is projected to be relatively low because the emission reduction potential through technological and structural change in the milk-producing sector, which dominates emissions in India, is limited within the simulated horizon.

Figure 4.13 shows the breakdown of emission savings for several major countries/regions in absolute terms. It illustrates that mitigation in the United States and the European Union is almost entirely based on technical and structural adjustments while production levels hardly change compared to the baseline in 2030. In contrast, technical and structural mitigation options in Brazil, China and India appear to be more costly than in the United States or the European Union, and large parts of the emissions saved come through decreases in production.

Contrary to the other scenarios, the spill-over effect to the groups of LDCs tends to be negative as agricultural prices, especially for ruminant products, increase significantly making food on world markets more expensive. At the same time it stimulates local production which leads to increases in GHG emissions in those regions.

The cross-country comparison for Scenario 6 illustrates how emission savings and calorie availability are interlinked. In Brazil and China, calorie availability is increasing stronger than in India, the European Union, and the United states. Emission savings are lower in the former two countries, however. Obviously, the demand response is more elastic in those countries so that lower prices lead to stronger consumption increases. As a consequence, the emission savings are smaller as higher demand is associated with higher emissions. The spill-over effect is again positive for consumers in LDCs, but negative for net surplus agricultural producers.

Figure 4.13. Emission savings versus food security in relation to the baseline scenario

Source: Aglink-Cosimo simulation results.

Figure 4.14. Regional emission savings versus food security in relation to the baseline scenario, 2030

Source: Aglink-Cosimo simulation results.

Comparison across scenarios

Each of the analysed scenarios reveals potential to reduce GHG-emissions by 2030 as summarised in Figure 4.15. Ranked by emission reduction, the highest potential is found in Scenarios 1 (preference shift in food demand) and 5 (carbon tax on production) where emissions are reduced by about 850 MtCO₂eq (-15%). A reduction by 13% is obtained in Scenario 4 (food waste reduction with costs) followed by Scenario 3 (food waste reduction) (-8%) and the productivity increase Scenario 6 (-6%). The lowest emission reduction is projected under the consumption tax Scenario 2 (-5%). Interestingly, applying a demand tax at the consumer price level appears to be much less efficient than applying the same tax at producer level. This is in line with the general observation that, especially in high income regions, consumers are not so responsive to the final consumer price and so it is not easy to change consumption patterns by implementing taxes on food products. Furthermore, the applied tax is decoupled from the actual carbon produced and thus the only opportunity to reduce emissions is provided through reductions in outputs.

Figure 4.15. Emission savings versus food security across scenarios in relation to baseline scenario, 2030

Source: Aglink-Cosimo simulation results

When including the food security indicators, Scenarios 1, 3 and 6 appear to consistently improve both emissions and food security. However, Scenarios 1 and 3 are also the most difficult to achieve. Reducing food waste without incurring any costs is unrealistic, while influencing consumer preferences such that they consume less ruminant products would most likely prove to be very challenging. Increasing productivity is generally promising, especially given the strong evidence of high returns to research and development, although the cost of implementing such policies could to some extent dampen the price benefits as pointed out by Alston (2010[10]) and Hurley et al. (2016[11]). Scenario 4 clearly illustrates that under the assumption of increasing marginal costs of food waste reduction, the food security goals might be at risk if the objective is to eliminate the final unit of waste. Food security in Scenario 5 might be threatened by higher consumer prices, but still appears to be a promising option for reducing GHG-emissions.

The feasibility of a global carbon tax is questionable. However, production-side mitigation can also be achieved by subsidising emission-reducing technologies, which could reduce emissions without major influences on supply and demand quantities.

It should be noted that comparability across scenarios is limited as the applied measures and scenario assumptions are not always consistent. Figure 4.16 relates the three food security indices to one percentage point of emissions saved. In this figure the signs are chosen such that a positive number is associated with an increase of food security. For example, in Scenario 1 each percent point of reduced emissions has a positive impact on consumer prices of 0.25 percentage points while it impacts negatively on agricultural income (-0.4 percentage points) and slightly reduces calorie availability. It becomes apparent that in most cases, emission savings and food security measures impact in opposite directions. Scenario 3 and 6 are the only cases where two of the indicators – consumer prices and calorie availability – show positive developments, while Scenarios 2 and 4 impact negatively on all food security indices. This perspective generally favours Scenarios 1, 3 and 6 if the agricultural income effect is set aside.

Figure 4.16. Food security indicators per one percentage point of emissions saved

Source: Aglink-Cosimo simulation results.

Figure 4.17 illustrates how emission reductions in the six scenarios benchmarked against the necessary reductions to comply with the 2-degree target specified in the COP 21 agreement. Rogelj et al. (2016[12]) argue that “limiting warming to any level requires net CO₂ emissions to become zero at some point in time and, given the small remaining carbon budget, this moment is estimated to be before the end of this century for a 2 °C limit”. The path to get to that level is, however, not specified. Assuming that all sectors contribute at the same proportionate rate to that target, Figure 4.13 shows the savings that are necessary to reach zero net emissions by 2100 under the assumption that the annual reduction and the emission trajectories of the six scenarios are constant.13

Figure 4.17. Emission pathways under scenarios compared to a linear path to net zero emissions, 2100

Source: Aglink-Cosimo simulation results.

It is clear that none of the six scenarios alone could reduce emissions enough to turn onto a path that would lead to zero net emissions of the sector by 2100. The baseline scenario, under which no additional action towards climate change mitigation is assumed, would increase climate change risks considerably. If that series were to continue towards 2100, emissions from agriculture would be 60% higher than in 2020.14 If all sectors were to perform like this, global warming would be about 4 degrees over that of the current century (Rogelj et al., 2016[12]). To avoid this, climate change mitigation action is needed. Alhough none of the analysed scenarios alone can reduce emissions enough to get on the 2-degree trend, a combination is perhaps sufficient. Given the general positive assessments of Scenarios 1 and 5, a combination of these two scenarios has been carried out. This shows that the emission-savings effect is slightly lower than the sum of those effects in the two single scenarios, leading to an emission level below 4 GtCO₂eq, a stronger mitigation effort than required by the 2-degree target pathway and achieved without increasing significant food security concerns.

Implicit trade of emissions in food products is another layer of the mitigation task. Without global agreement and implementation, countries that implement GHG-abatement strategies run the risk of reducing their emissions, e.g. by a decrease in production, but then importing products responsible for even more emissions to replace the shortfall in supply (carbon leakage). However, trade can also allow production to shift to locations with lower emissions intensity (carbon reallocation). This analysis cannot investigate this aspect in detail because it would require that imports be distinguished by origin, which cannot be done in Aglink-Cosimo. Nevertheless, the analysis does illustrate the global changes in carbon reallocation via trade in the analysed scenarios. The trade quantities recorded in the OECD-FAO Agricultural Outlook 2018-2027 project that traded emissions will increase from 460 in 2017 to 500 MtCO₂eq in 2030. Eighty per cent of these carbon exports can be attributed to trade in beef (55%), sheep meat (10%), and rice (15%). As illustrated in the left graph of Figure 4.18, trade in emissions decreases in all but Scenario 2. In general, rising exports of emissions are not a problem if overall emissions are decreasing, therefore the sign of those exports is not very important. The interesting aspect is a decomposition of the change in emissions trade into changes in trade quantities and emission intensities (Figure 4.18, right column), which reveals that the increase in Scenario 2 is predominantly a source of increase in the emission intensity of trade. This means that the composition of global exports is moving towards countries with higher emission intensities which happen to also be lower-cost producers. For beef, these shifts are dominated by increases of beef exports from Brazil, where beef production emits about 30 kgCO₂eq/kg and decreases in exports from the United States and Canada, where about 10 kgCO₂eq are emitted per kg of beef produced. For sheep meat, global traded quantities decrease, while emissions trade increases. This is mainly caused by export shares that move from the LDCs in Africa, where emissions of producing sheepmeat are estimated at 22 kgCO₂eq/kg in 2020 to Australia (25 kgCO₂eq/kg).

A second observation is that the trade quantity effect dominates in Scenarios 1 and 6. This is intuitive because the scenario setup changes demand (Scenario 5) and supply (Scenario 6) structures in countries through similar shifts, leading to a more equal distribution of the scenario shock across countries so that the relative trade pattern is little affected.

For beef in the food waste scenarios (3 and 4), quantity and intensity changes lead to reductions of traded emissions, with a reduced intensity dimension accounting for the larger shares. Here, the emission intensity reduction stems from export shares moving from Brazil to the European Union. This happens due to higher initial levels of food waste in the meat sectors in the European Union than in Latin America (Figure 4.18), which lead to a stronger shift of beef production to export markets in the European Union. For sheep meat, total emissions traded decrease although emission intensity increases for the same reasons as in Scenario 2, i.e. a shift of export shares towards regions with higher emissions.

The opposite can be observed in Scenario 5 where emissions traded in sheep meat increase and emission intensity decreases. The decrease in emission intensity stems mainly from the reduction of national emissions through technological adjustments and structural change.

At the national level, each of the above mitigation instruments can lead to carbon reallocation through trade. For example, if a country’s emissions reductions were to be achieved primarily through supply side mitigation, with consumption relatively unchanged (as would happen under Scenario 4 border measure to contain imports or raise domestic prices), then final emissions would increase if the emissions intensity of those imports were higher than the emissions intensity of domestic production. Conversely, emissions would fall if the emissions intensity of imports were lower than the emissions intensity of domestic production. A global approach to mitigation is needed to ensure that national mitigation efforts are complementary and lead to positive carbon reallocations rather than carbon leakage.

Figure 4.18. Changes in global emission trade (left graph) and changes in emission intensity of trade (right graphs) in relation to baseline scenario, 2030

Source: Aglink-Cosimo simulation results

Conclusions

Agriculture contributes a significant share of greenhouse gas (GHG) emissions at the global level (about 11% excluding the impact of land use changes), but also holds substantial potential to limit the increase in global warming over the next decades. In a 2018 report prepared for OECD, Blandford and Hassapoyannes (2018_[1]) observed that the agricultural sector could limit this increase by reducing direct emissions in crop and livestock production systems and in indirect emissions that are associated with changes in land use, as well as by increasing carbon sequestration. They found that technological advancements on the supply side and changes in consumer preferences on the demand side that result in land-sparing are promising options, particularly in view of global food security concerns.

This report has analysed agriculture’s potential contribution to climate change mitigation by simulating supply and demand side mitigation strategies. It uses the Aglink-Cosimo model based on the baseline of the OECD-FAO Agricultural Outlook 2018-2027 and takes into consideration only direct emissions that result from agricultural crop and livestock production activities. It does not analyse the most efficient way agriculture can contribute to overall mitigation as this also depends on changes in land use. However, all scenarios presented here should reduce the pressure on land and thus the results underestimate the full mitigation potential of the sector

Six scenarios have been analysed: four relate to demand-side mitigation measures (a preference shift towards less ruminant products, taxing food according to related emissions, and two variants of food waste reduction) and two address supply-side mitigation (carbon tax and productivity shift). The effects on emissions have been compared with the impact on food security as captured by indicators of food availability, food prices and on farm incomes. The analysis finds the following.

• Influencing consumer preferences so that more calories are obtained from non-ruminant animal sources has the highest benefits among the analysed scenarios. However, the mechanism by which such a change could be achieved is not specified.

• Consumption taxes are the least effective measure to reduce greenhouse gases, especially when these are decoupled from the actual carbon produced, owing to the inelasticity of demand for broad food groups, and would raise food prices, potentially leading to food security risks for low income consumers.

• Reducing food waste can be a strategy to mitigate climate change, but it is important to take into consideration that the potentially high costs to reduce waste could raise food prices, and potentially lead to food security concerns.

• Supply side mitigation via carbon taxes has a high potential to reduce emissions from agriculture with limited risks in terms of food security.

• Increasing productivity in agricultural production systems could potentially reduce emissions and increase food availability, in addition to improving access via lower prices.

• A global approach to mitigation is needed to ensure that national mitigation efforts are complementary and lead to positive carbon reallocations rather than carbon leakage.

References

[1] Blandford, D. and K. Hassapoyannes (2018), “The role of agriculture in global GHG mitigation”, OECD Food, Agriculture and Fisheries Papers, No. 112, OECD Publishing, Paris, http://dx.doi.org/10.1787/da017ae2-en.

[5] FAO (2011), Global food losses and food waste - Extent, causes and prevention., http://dx.doi.org/10.1098/rstb.2010.0126.

[7] Frank, S. et al. (2018), “Structural change as a key component for agricultural non-CO2 mitigation efforts”, Nature Communications, http://dx.doi.org/10.1038/s41467-018-03489-1.

[3] Hasegawa, T. et al. (2018), Risk of increased food insecurity under stringent global climate change mitigation policy, http://dx.doi.org/10.1038/s41558-018-0230-x.

[9] Havlík, P. et al. (2014), “Climate change mitigation through livestock system transitions.”, Proceedings of the National Academy of Sciences of the United States of America, Vol. 111/10, pp. 3709-14, http://dx.doi.org/10.1073/pnas.1308044111.

[8] Herrero, M. et al. (2013), “Biomass use, production, feed efficiencies, and greenhouse gas emissions from global livestock systems.”, Proceedings of the National Academy of Sciences of the United States of America, Vol. 110/52, pp. 20888-93, http://dx.doi.org/10.1073/pnas.1308149110.

[11] Hurley, T. et al. (2016), Returns to Food and Agricultural R&D Investments Worldwide, 1958-2015, InSTePP Brief, International Science & Technology Practice & Policy Center, St Paul, Minnesota, https://ageconsearch.umn.edu/record/249356/files/InSTePPBriefAug2016.pdf (accessed on 25 September 2018).

[4] Jensen, H. et al. (2019), “Economic Impacts of a Low Carbon Economy on Global Agriculture: The Bumpy Road to Paris”, Sustainability, Vol. 11/8, p. 2349, http://dx.doi.org/10.3390/su11082349.

[10] M. Alston, J. (2010), “The Benefits from Agricultural Research and Development, Innovation, and Productivity Growth”, OECD Food, Agriculture and Fisheries Papers, No. 31, OECD Publishing, Paris, http://dx.doi.org/10.1787/5km91nfsnkwg-en.

[2] OECD/FAO (2018), OECD-FAO Agricultural Outlook 2018-2027, OECD Publishing, Paris/FAO, Rome, http://dx.doi.org/10.1787/agr_outlook-2018-en.

[6] Okawa, K. (2015), “Market and Trade Impacts of Food Loss and Waste Reduction”, OECD Food, Agriculture and Fisheries Papers, No. 75, OECD Publishing, Paris, http://dx.doi.org/10.1787/5js4w29h0wr2-en.

[12] Rogelj, J. et al. (2016), Paris Agreement climate proposals need a boost to keep warming well below 2 °c, http://dx.doi.org/10.1038/nature18307.

[14] Rogelj, J. et al. (2015), Energy system transformations for limiting end-of-century warming to below 1.5 °C, http://dx.doi.org/10.1038/nclimate2572.

[13] Wollenberg, E. et al. (2016), “Reducing emissions from agriculture to meet the 2 °C target”, Global Change Biology, Vol. 22/12, pp. 3859-3864, http://dx.doi.org/10.1111/gcb.13340.

Annex 4.A. Methodology

Extensions to Aglink-Cosimo

In the course of this project, the Aglink-Cosimo model was extended to capture the necessary aspects to analyse emission-related scenarios.

According to FAOSTAT, the agricultural sector accounts for about 10% of Global Greenhouse Gas (GHG) emissions. Although the database reports this share to be decreasing over time, absolute emissions have been increasing over the past decades. It can be observed, however, that in some countries emissions continue to increase with rising production levels, and in other countries they remain stable or decrease despite increasing production, thus implying decreasing emissions per unit produced over time. The development of such emission factors over time are the result of more efficient input use as well as technological changes and conversion to different production systems.

Earlier work that aimed to include emission factors into the Aglink-Cosimo model did not reflect these dynamics over time, which are essential for comprehensive reporting. The model reflects production technologies and agricultural inputs not explicitly, wherefore a direct link of emissions to their sources is not possible.

During the first half of 2017, collaboration between the OECD and the International Institute of Applied System Analysis (IIASA) was established in order to obtain – as a first step in the direction of analysing climate change related questions with Aglink-Cosimo – dynamic emission coefficients for a baseline scenario. IIASA was chosen because its Globiom model is among the leading models that can capture some of those aspects driving emission factors over time. The idea of this project was therefore to align the production quantities assumed in the OECD-FAO Agricultural Outlook with those of the Globiom baseline and incorporate the resulting emission factors into the Aglink-Cosimo model to calculate total emissions from agriculture.

Since the two models differ in regional and commodity scope, a mapping between the two model codes was developed, including possible conversion/aggregation factors that reflect possible differences at the processing stage of the two models. The OECD then provided the GLOBOIM team with detailed time series and assumptions needed to align the Globiom baseline with the Outlook, and a time series of emission factors from 2000 to 2030 was calculated. These factors were then added to the post-model calculation of Aglink-Cosimo. They are available disaggregated according to the categories available in FAOSTAT:

• CropSoil_N2O N2O emissions from applying mineral fertilizer to crop soil

• Rice_CH4 CH4 emissions from cultivating Rice

• ManmgtTot_N2O N2O emissions from Manure management

• ManaplTot_N2O N2O emissions from applying manure on the field

• ManprpTot_N2O N2O emissions from manure left on pasture

• ManmgtTot_CH4 CH4 emissions from manure management

• Entferm_CH4 CH4 emissions from enteric fermentation

These positions sum up to the

• Total_CH4N2O Total direct non-CO₂ from agriculture

All coefficients have been converted to CO₂ equivalents using the conversion coefficients reported in the IPCC Fourth Assessment Report and are available for the following countries/country aggregates:

• USA United States

• EUN European Union

• BRA Brazil

• CAN Canada

• CHN China

• JPN Japan

• MEX Mexico

• KOR Korea

• TUR Turkey

• ZAF South Africa

• IND India

• RLAM Rest of Latin America

• AUNZ Australia and New Zealand

• MNAF Middle East and North Africa

• OCEL Other Oceania

• REUW Rest of Western Europe

• SSAF Sub Saharan Africa

• SSEA South-East-Asia

• ECSI Eastern Europe (including the Russian Federation)

As Aglink-Cosimo does not cover all food commodities, it currently captures the emissions of the agricultural sector partially. Emission coefficients exist for the following commodities:

• BV Beef and veal

• CT Cotton

• EG Eggs

• MA Maize

• MK Milk

• OCG Other coarse grains

• OOS Other oilseeds

• PL Palm oil

• PK Pork

• PT Poultry

• RI Rice

• RT Roots and tubers

• SB Soybeans

• SCA Sugar cane

• SH Sheep and goat meat

• WT Wheat

This subset of products covers about 80% of global emissions caused by the agricultural sector. Although it would be clearer to link the emission coefficients to the activity levels (e.g. land use and herd sizes), due to a simplified presentation of the animal sector in Aglink-Cosimo in particular, this is not done and emissions are calculated as a function of production quantities:

${\mathrm{E}\mathrm{m}\mathrm{i}\mathrm{s}}_{\mathrm{c},\mathrm{p},\mathrm{e},\mathrm{t}}=\frac{{\mathrm{\beta }}_{\mathrm{c},\mathrm{p},\mathrm{e},\mathrm{t}}{\mathrm{Q}\mathrm{P}}_{\mathrm{c},\mathrm{p},\mathrm{e},\mathrm{t}}}{1000}$

The Emission Emis of type e in region c for product p in year t measured in MtCO₂eq per year are calculated by multiplying the production QP (measured in 1000 t) of a product by the emission coefficient $\mathrm{\beta }$ measured in kgCO₂eq by kg of product.

Total emissions are then the sum over the single emission types e.

This enhancement makes it possible to assess the emissions path which is inherent to the OECD-FAO agricultural outlooks, but also in most scenarios that can be analysed with Aglink-Cosimo. It does not allow to address supply side mitigation policies, as the average emission coefficient of a country would adjust endogenously as soon as those technologies change. The Globiom Team at IIASA was asked to address this problem.

Globiom has been used in the past to develop the MACCs (Marginal Abatement Cost Curves) for the agricultural sector e.g. for the UK DECC model GLOCAF, and for integrated assessment models such as MESSAGE (IIASA), POLES (JRC), or WITCH (FEEM). A detailed agricultural non-CO₂ MACC analysis is published in Frank et al (2018[7]) However, MAC curves cannot be used directly in Aglink-Cosimo, and an alternative approach was developed.

Globiom explicitly covers the following non-CO₂ emission sources: N2O from application of synthetic fertilizer, CH4 from rice cultivation, N2O from manure dropped on pastures, N2O from manure application, CH4 and N2O from manure management, and CH4 from enteric fermentation. The global amount of emissions in a mitigation scenario will be the result of three endogenous mechanisms.

Regional GHG intensity change

• Management / production system change

• Spatial relocation within a region/country (crops)

• Technological mitigation options (e.g. biodigesters)

Global GHG intensity change

• Average regional GHG intensities for individual products differ substantially across regions, hence worldwide relocation of the production through international trade to more or less GHG intensive regions will change the global average GHG intensity

Global production volume change

• Result of a change in food and feed consumption potentially related to increased market prices, themselves being result of the additional production cost related to mitigation policies

Aglink-Cosimo has the internal capacity to deal with Scenarios 2 and 3 – reduction of GHG emissions via international trade and consumption side adjustments – based on the GHG emissions coefficients derived. For example, for a simple climate policy implemented as a carbon tax, the supply curve of a given product in a given region would be shifted upwards by the product of the emission factor and the carbon price and the model would endogenously adapt demand and international trade.

The only missing mitigation mechanism that needed to be added to the Aglink-Cosimo based on additional Globiom input were the “Regional GHG intensity” improvements as a response to climate change policy. Similarly, as for the above mentioned MACCs, the parameters from Globiom to Aglink-Cosimo would be derived by running a series of scenarios covering the whole relevant range of carbon prices representing the cost efficient contribution of the agricultural sector to a wide range of overall mitigation effort levels. In order to avoid double counting, since Aglink-Cosimo represents endogenously consumption side and international trade adjustments, these mechanisms were fixed in Globiom to the baseline levels. Globiom being a standard economic model, endogenous adjustments in management systems, intra-regional spatial allocation of activities, and adoption of mitigation technologies take place as long as the marginal cost of emissions reduction are less or equal to the carbon tax. This approach allowed to derive a new set of emission coefficients corresponding to each carbon price level. The new coefficients can now be implemented in Aglink-Cosimo depending on the carbon price / mitigation effort scenario considered allowing the model to capture the production side adjustment to mitigation incentives.

Given that in the Globiom scenarios the marginal cost of reduction (abatement) of the emissions coefficient is always equal to the carbon price, the Aglink-Cosimo supply curves would need to be shifted by the “baseline” emission coefficient multiplied by the carbon price. Potential alternative effects of the management change on the necessary supply curve shift could be included in Aglink-Cosimo for specific additional climate policies. For example, the farm-level mitigation measures were implemented in the form of a subsidy rather than a tax. In this case, if the region-level management adjustment led to a halving of the emission intensity of a given product, and the adoption of such management were supported by a program fully covering the cost difference between the “baseline” and “mitigation” management system, the supply curve would be shifted only by half compared to the scenario without a subsidy.

The missing elements to get substantially closer to the complete AFOLU GHG emissions accounting are emissions from other land uses and carbon sequestration. Since there are potential interactions between non-CO₂ emissions and CO₂ emissions, it would be useful to account for both to avoid unintentional negative effects, such as reduction of nitrogen fertilizer use leading to reduced N₂O emissions but also larger CO₂ emissions from additional land cover change due to lower yields. This aspect will be added in the future.

Technical scenario implementation

Scenario 1 reflects a preference shift in demand away from ruminant production such that demand for ruminant products decreases 10% below the base year values. This scenario was implemented by shifting the demand curves for different food items. Figure 4.A.1 illustrates how Scenario 1 was implemented using a simplified example whereby only three food commodities exist in the consumption basket: beef, wheat and pork. The graph gives the initial demand curves for the three products in calorie equivalents (Do) with the corresponding price quantity combinations (Po and Qo) in the baseline. The demand function of beef is moved to the left (D1) so that the new quantity Q1 would be demanded at the baseline price Po. Q1 would be derived from the assumption described above (10% below current per capita values). The difference between Q1 and Qo in the left graph is then distributed between the other two commodities with wheat taking larger shares, because the initial demand quantities were higher for wheat than for pork. Consequently, the new demand curves for wheat and pork are located to the right of the original ones.

Naturally, as demand remains an endogenous scenario output, the simulated demand quantities will not be exactly equal to the initial assumptions, as prices will move away from the baseline equilibrium as supply adjusts. Such a shift in demand behaviour could be achieved by influencing consumer preferences. This shift in preferences was applied globally, except in the LDCs.

Annex Figure 4.A.1. Shifting demand preferences in Scenario 1

Scenario 2: In this scenario, all products are taxed at the consumption level based on their primary emissions. The consumption tax is set at USD 60 (in year 2000 real USD) per ton of CO₂eq emitted by each product. It is applied globally, again except in the LDCs. Technically this tax is added to consumer prices that enter the food demand equations.

Scenario 3: In this scenario, wasted food at consumption level is reduced to zero without any costs. The food demand variable in Aglink-Cosimo includes wasted quantities implicitly. Those implicit values were estimated based on the FAO (2011[5]) study and then gradually deducted from the food demand variable by shifting the demand curves to the left. Technically this is done by adjusting the R-factors of the food demand equation:

$\log \left({FO}_{r,c,t}\right)=\alpha +\sum _{c1\left(food\right)}{\beta }_{c1}*\log \left(\frac{{CP}_{r,c1,t}}{{CPI}_{r,t}}\right) +{\beta }_1*\log \left(\frac{{GDPI}_{r,t}}{{POP}_{r,t}/{POP}_{r,2005}}\right)+\log \left({POP}_{r,t}\right)+{\beta }_2*TRD+\mathrm{l}\mathrm{o}\mathrm{g}\left(R\right)$

Where:

FO = Food use

CP = Consumer price

CPI = Consumer price index

GDPI = Nominal GDP index

c1(food) = Commodities with food use

β_c1 = Cross- and own-price elasticities

POP = Population

TRD = Trend

R = Residual calibration factor

r = Regions

c = Commodities

t = Years

An additive logged variable in this double log representation corresponds to a multiplicative variable in the un-logged version. Therefore, the R-factor is a multiplicative scaler to the food use variable. Since the idea in this scenario is that if no waste exists demand goes down by the wasted amounts, the R-factors can be used to define those. For example, FAO data says that in the European Union, 25% of wheat is wasted at consumption (and retail) level. If those quantities vanish, 25% less of the product would be needed to achieve the same final consumption level. Therefore, the original R-factor of the food use equation of wheat in the European Union was multiplied by 0.75 in the final simulation year (2030). Between the first and the final simulation year this factor has been gradually decreased from 1 to 0.75.

Scenario 4: This scenario is implemented in the same way as Scenario 3 but it increases consumer prices in relation to waste abatement levels. A crucial assumption is the level of consumer price increase for the final unit of waste abatement. Since data on this does not exist, assumptions were made to illustrate the importance of reflecting waste abatement costs. The assumption that the final unit of waste abatement costs as much as the consumer pays for the respective product in the baseline was taken. This means the consumer must effectively pay in the final simulation year twice as much for one unit of product as in the baseline. Between the first and the final simulation year, these costs are not assumed to increase linearly, but exponentially. This reflects the assumption that the first units of waste abatement are relatively cheaper than the following ones.

Scenario 5: A production tax of USD 60 (real USD, 2000) per tonne CO₂eq emitted was imposed on agricultural production activities by shifting the supply curves for each commodity upwards by the amount of the tax. In other words, marginal production costs increase by the amount of tax applied. Based on the collaboration with IIASAA described above, the emission coefficients that were derived from Globiom simulations with the same tax level were also applied.

Scenario 6: A productivity shift in the crop sector was simulated by shifting the yield equations using the R-factor in a similar way as described for Scenario 3. For animal products, where yields are not explicit variables but inherent from the production – and herd size equations, production equations were shifted by the applied 10% and it was ensured that the resulting higher production quantities were achieved with the same herd sizes c.p. Note that by doing this, the yield increase came at no additional cost.

Notes

← 1.  Ongoing investments in Aglink-Cosimo should allow emissions from LULUCF to be included in future analysis. The exclusion of the land use sector from this partial analysis should not, however, be viewed as a significant shortcoming. Instead, it allows to focus on the analysis on direct emissions. The six scenarios developed for this chapter should reduce pressure on land, whereas including the land use sector would not lead to a compensation of the analysed emission reduction potential. Relative potentials might change slightly depending on differences in the impact of total cropland.

← 2.  According to the FAO definition, “food security exists when all people at all times have physical and economic access to sufficient, safe and nutritious food to meet their dietary needs and food preferences for an active and healthy life”.

← 3.  Hasegawa et al (2018[3]) illustrate that food availability is also correlated with prevalence of undernourishment, so an increase of that availability should also improve the undernourishment situation. This study does not, however, go deeper into that issue.

← 4.  The Aglink-Cosimo model does not include explicit costs, and this index only covers implicit cost increases via input price changes, but not via input quantity adjustments.

← 5.  The choice of the carbon tax level was based on the fact that this price corresponds to the value that some modelling studies suggest will be required to limit temperature increases to 1.5°C (Rogelj et al., 2015[14])

← 6.  This assumption is simplified as the emissions that are implicitly contained in final consumption depend on the origin of each product. A large share of imported commodities could change the emission coefficient significantly. This effect could, however, only be captured by a model capturing bilateral trade.

← 7.  This tax is applied at the primary consumer level as emissions arising from the processing industry are not accounted for.

← 8.  Indeed, if the effect of the collection of taxes was included, the various possibilities of use could impact income or final prices, and the conclusions drawn from this scenario might change. This is also true for Scenario 5. However, one principle of this report is not to have too many overlaying parameters changed in one scenario in order to assess the pure effect of the instrument in place.

← 9.  This is by construction in Scenario 1, while it is an endogenous outcome in Scenario 2.

← 10.  Trade in emissions is not covered in this analysis, but it is fair to assume that imports which are consumed instead of goods domestically produced, the former would be produced with lower emission intensity in the exporting countries as emission intensities are higher in the LDCs.

← 11.  The scenario assumptions are a strongly simplified representation of the mechanisms that would be in place when applying different waste reduction technologies. This scenario is primarily meant to illustrate that neglecting waste reduction costs, as in Scenario 3, is dangerous. The EU Agricultural Outlook 2018-2030 includes a box on the effects of reducing food waste in European households. Using a CGE model, the authors attribute waste avoidance costs mainly to improved packaging, and they estimate those costs to be between 1% and 5% of sales.

← 12.  The evolution of the consumer price index in Scenario 4 cannot be interpreted directly as increasing consumer prices. It does indicate that other (opportunity) costs of acquiring food by households increases significantly when waste is eliminated.

← 13.  The two-degree compatible emission reduction from agriculture by 2030 using this method amounts to about 650 MtCO₂eq. This is below the range of estimates reported in Wollenberg et al (2016[13]). They estimate that on average it would require an annualized reduction of 1 GtCO₂eq in 2030. It should also be noted that such an individual trajectory, for a given sector, has only very limited scope and relevance. On the one hand, the various sectors do not have the same mitigation potential, and on the other hand, these potentials differ according to the boundaries of the various sectors. For example, the relative reduction potential for agriculture would not be the same if the effect of the substitution of fossil energy by biomass was attributed to it, or if it was accounted for in the industry or transport sector, as is currently the case. It is likely that agriculture will continue to have net positive emissions, but other sectors, especially LULUCF, could compensate for this.

← 14.  A linear extrapolation is most likely no proper approximation of a long-term business as usual scenario. Slowing population growth and increasing saturation of demand will lead to lower growth in emissions. Havlik et al (2014[9]) estimate that agricultural emissions will be only 30% higher in 2100 compared to 2020.