Adjusting agricultural emissions for trade matters for climate change mitigation

Food systems emit 21–37% of global anthropogenic GHGs, with the largest share generally from agricultural production (farm-gate). National emission inventories typically follow production-based accounting within borders, rooted in the Kyoto Protocol and IPCC frameworks, but growing international food trade means consumption often occurs far from production. Consequently, trade needs to be incorporated into agricultural emissions accounting, including the roles of producer, consumer, and intermediary trading countries. Prior research has contrasted production- versus consumption-based accounting and shown disparities, but relatively few studies focus specifically on agricultural emissions with trade adjustments across food items. A comprehensive national-level overview that accounts for agricultural emissions embodied in trade remains a gap. This study explores global, regional, and national trends (1986–2017) in trade-adjusted agricultural emissions (TAE) for all food items, how TAE differs from production-based emissions (PBE), and emissions embodied in trade between producer and consumer regions. The analysis focuses on farm-gate emissions, uses FAOSTAT data on production, trade, and emission intensities, and applies a bilateral trade input–output (BTIO) approach to define TAE as PBE plus imports minus exports. Three sensitivity analyses test alternative specifications regarding emission intensities, technology-adjusted exports, and re-exports. An additional analysis explores trade-adjusted agricultural land-use emissions (FOLU).

Studies have documented differences between production-based and consumption-based accounting and highlighted the implications for climate policy. A limited number of works have applied trade-adjusted approaches to agricultural emissions, often emphasizing producer or consumer perspectives while overlooking intermediary trading countries. There is a noted gap in national-level GHG inventories for food-related agricultural emissions that fully incorporate trade flows. Prior work also discusses multi-regional input–output (MRIO) versus bilateral approaches and the need to consider emission intensities alongside trade volumes to avoid misestimating embodied emissions. This study extends these discussions by focusing on agricultural emissions and explicitly incorporating intermediary roles using a BTIO approach and FAOSTAT data.

Data: FAOSTAT (accessed 02.12.2019) for 211 countries grouped into eight UN M49 regions. Trade data include primary and processed items from two datasets: Trade – Crops and Livestock (TCL) for totals and Detailed Trade Matrix (DTM) for bilateral flows (1986–2017). The scope is restricted to 307 food-related items. DTM reporter-country data were primary; gaps were filled using partner reports. DTM totals were scaled to match TCL to ensure consistency. Conversion of processed to primary equivalents: Mass of processed items was converted to primary equivalent using caloric content ratios (me=mp*(Cp/Ce)), with caloric data from FAO and USDA. Emission intensities: FAOSTAT provides farm-gate emission intensities for 14 food groups (using IPCC SAR GWPs), covering 82.3–92.5% of agricultural emissions (1986–2017). An additional 'others' group (e.g., vegetables, fruits, oilcrops) was created by subtracting emissions of the 14 groups from total agricultural emissions and dividing by production volumes to derive country-, regional-, and global-level intensities. Agricultural land-use (FOLU) intensities: Separate land-use emissions (IPCC AR5 GWPs) covering net forest conversion, degradation of organic soils, and biomass burning in humid tropical forests/peat soils were divided by total production volumes to estimate a uniform intensity per item due to lack of item-specific linkage. BTIO approach and assumptions: A bilateral trade input–output (BTIO) method was used, without distinguishing final versus intermediate trade, to align with FAOSTAT bilateral data. Assumptions: (1) Exports originate from domestic production; (2) if an exporter is a non-producer, the exported item is assumed imported from the same region; if the region is also a non-producer, global sourcing is assumed. These assumptions determine which emission intensities apply in embodied trade calculations. Embodied emissions calculation: Emissions E_cyi = m_ecy,i × EI_cy,i for production, exports, and imports. TAE_c,y = PBE_c,y + ImE_c,y – ExE_c,y. Values are typically presented as three-year averages. Sensitivity analyses: (1) Global intensities for non-producer countries (imports/exports) instead of regional; (2) technology-adjusted approach (Kander et al.): use global average intensity for all exports to reflect substitution effects; (3) re-exporter approach: compute new intensities based on combined domestic production and imports, assuming proportional shares in exports, capturing first-order trade effects. Differences relative to the main approach were quantified as weighted differences. Land-use emissions test: For 2015, agricultural land-use emissions were adjusted by trade using analogous procedures to assess how trade reallocates FOLU-related emissions.

- Global TAE rose from 3.86 Gt CO2e/yr (1987) to 5.02 Gt CO2e/yr (2015), while per capita TAE fell from 0.77 to 0.68 t CO2e/cap/yr, reflecting efficiency gains and declining emission intensities. - Regionally, most areas saw rising total TAE except Europe, Oceania, and the Former Soviet Union (FSU), due to increased productivity/efficiency (Europe, Oceania) and structural changes post-1991 in FSU. - Country-level highlights: Mainland China’s TAE increased from 452.2 Mt CO2e/yr (1987) to 705.2 Mt CO2e/yr (2015, largest globally). Large increases occurred in Pakistan (+60.0 Mt) and Nigeria (+41.6 Mt) between 1987 and 2015. Per capita TAE was highest in parts of Oceania and LAC (e.g., Australia 2.41; Uruguay 3.77 t CO2e/cap/yr in 2015). Some low-income/emerging economies (e.g., Mongolia; southern/central Africa) exceeded 3.0 t CO2e/cap/yr. Many Asian countries had low per capita TAE (<0.5 t CO2e/cap/yr) despite large totals. - Differences between TAE and PBE: Import-dependent economies showed TAE ≫ PBE, e.g., Hong Kong and Singapore (>50× PBE in 2015); Bahrain, Kuwait, UAE (≥4× PBE). China’s TAE was 7.9% higher than PBE in 2015 (difference 51.5 Mt CO2e/yr), small per capita difference (<0.05 t/cap). Major exporters had TAE < PBE: Australia and New Zealand averaged 59.3% lower TAE than PBE (2015). Brazil’s PBE–TAE gap widened by 45.6 Mt CO2e/yr from 1987 to 2015, driven by export growth; ruminant meat export emissions rose from 10.007 to 34.453 Mt CO2e/yr; the 'others' group (dominated by soybeans) contributed 1.229 to 12.727 Mt CO2e/yr. - Emissions embodied in trade: Export-embodied emissions increased from 306.5 Mt CO2e/yr (1987) to 695.7 Mt (2015); import-embodied from 317.0 to 724.5 Mt CO2e/yr. Regional patterns mirror trade volumes: Europe had the largest embodied emissions in both exports and imports (including intra-regional). LAC’s export and Asia’s import emissions grew, with Asia’s import emissions surpassing Europe’s in the last two decades. - Intra/inter-regional flows (2015): Europe’s embodied emissions were mostly intra-regional (exports 80.3%, imports 76.9%). Africa, WANA, FSU, and Asia had higher inter-regional import shares (66.7–85.8%). Despite low export volumes, Oceania contributed >27% of imported emissions to Asia and North America on average (2015), due to high-intensity ruminant meat exports and higher 'others' intensity versus other exporters to those regions. For Africa’s imports, rice dominated embodied emissions; Africa’s paddy rice intensity (1.14 kg CO2/kg) exceeded Asia’s (0.90 kg CO2/kg), meaning volume shares alone understate embodied emissions. - Food groups: Ruminant meat and milk products had the largest embodied emission shares in most regions (Europe, Oceania, North America). In Asia, paddy rice was a dominant contributor. The 'others' group’s share has grown in recent years, linked to rising crop-based feed demand (e.g., soybeans’ global exports increased fivefold from 1986 to 2017). - Sensitivity analyses: Results were robust overall. Using global intensities for non-producers yielded mean changes of +1.5% (increases) and −0.47% (decreases); outliers in small import-dependent states (e.g., Kiribati up to ~19× in 2000), but >98% within ±10%. Technology-adjusted approach: mean +5.5%/−8.5%, with UAE up to ~60× in 1997; >81% within ±10%. Re-exporter approach: mean +1.3%/−5.4%, UAE up to ~62× (1997); specific food-group outliers (e.g., paddy rice in 2015: Netherlands ~5×, Slovenia ~6×) due to high re-export shares; >91% within ±10%. - Land-use emissions (2015): Major exporters had high agricultural land-use emissions and large decreases when trade-adjusted: Indonesia −471.9 Mt CO2e/yr (>40% of original), Brazil −120.1 Mt CO2e/yr (>20%). Import-dependent countries saw large increases when trade-adjusted: China and India increased by >50× and >10× respectively relative to their low domestic FOLU emissions; several European countries (e.g., Italy, Spain >60×) and Singapore (>100×). These patterns reflect displacement of land-use emissions via trade.

Incorporating trade into agricultural emissions accounting substantially alters national inventories, especially for major importers and exporters, and highlights the role of intermediary trading countries. Using a BTIO framework tied to bilateral data, the study shows that embodied emissions depend not only on trade volumes but critically on relative emission intensities of trading partners and the composition of traded food groups (notably ruminant meat, milk, and rice). The limited deviations across sensitivity analyses (including re-export considerations) and similarities with other studies suggest robustness of the main approach; observed discrepancies are often attributable to differing emission factors or food groupings. Policy implications are significant: current NDCs are largely production-based and risk displacing emissions abroad. Trade-adjusted accounting can better inform targets, border adjustment measures, and trade policies by recognizing consumption and intermediary roles, encouraging consideration of emission intensities in sourcing. The dominance of animal-source foods in embodied emissions supports dietary shifts toward lower-emission options. The BTIO approach, while less granular for final consumption than MRIO, is transparent and well-suited to assessing bilateral agreements, border tax adjustments, and policy effects on embodied emissions.

This study constructs a comprehensive, trade-adjusted accounting of agricultural emissions across countries and regions from 1986–2017, demonstrating that national agricultural emissions depend strongly on trade flows, relative emission intensities, and food group composition. It highlights large divergences between production-based and trade-adjusted inventories for import- and export-intensive economies and underscores the policy relevance of including intermediary trading countries. Results are robust under alternative specifications and show substantial displacement of agricultural land-use emissions via trade. Future research directions include: integrating nutrient-based embodied accounting, examining the impacts of trade agreements and tariffs on embodied emissions, extending analyses to transport emissions and other supply chain stages, and further reconciling BTIO and MRIO perspectives to inform effective, trade-aware climate policies for the agricultural sector.

Key limitations arise from FAOSTAT data coverage and aggregation: limited item-specific emission intensities required grouping many items as 'others' with a single derived intensity; nevertheless, the 14 FAOSTAT groups capture most emissions and country-level results depend on both intensities and volumes. The main TAE estimates focus on farm-gate emissions and exclude other pre/post-production stages and transport (the latter being a small share of food system emissions). Linking FOLU emissions to specific commodities is challenging; the land-use analysis assumes proportional allocation by production volumes. Methodological assumptions include exporters sourcing from domestic production and regional sourcing for non-producers, which were tested via sensitivity analyses; MRIO could capture final consumption patterns differently, but BTIO better aligns with bilateral policy analysis. Data inconsistencies in bilateral trade reporting required scaling between DTM and TCL datasets.