International demand for food and services drives environmental footprints of pesticide use

Modern agriculture has achieved high yields over the past five decades through irrigation and intensive use of synthetic fertilisers and pesticides, but this strategy degrades ecosystems, depletes water resources, and contributes to climate change. Existing quantifications of global production and consumption footprints cover indicators such as greenhouse gas emissions, water scarcity, biodiversity, nitrogen pollution, acidification, and land use, but have largely overlooked environmental pressures from pesticide use. Pesticides can cause biodiversity loss and disrupt key ecosystem services including pollination, pest control, soil respiration, and nutrient cycling. Global policy initiatives (e.g., the EU Farm to Fork Strategy) seek to reduce pesticide risks, but effective, holistic strategies require quantifying pesticide footprints along entire supply chains and understanding how international trade drives pesticide use and potential footprint leakage. This study addresses that gap by defining and quantifying pesticide footprints as hazard loads attributable to consumption, and tracing them across international supply chains to link producers and consumers.

Environmental impact assessments often use pressure-oriented indicators (elementary flows such as emissions) and impact-oriented indicators (mid- and end-point impacts on health and ecosystems) derived from life cycle assessment (LCA). Pesticides have mainly been included within chemical footprints using bottom-up LCAs and models like USEtox, which provide specificity at product/process level but typically miss consumption-driven hot-spots and the role of globalised trade. Bottom-up LCAs also require system boundary choices that limit quantification of indirect supply-chain impacts. Multi-region input–output (MRIO) analysis offers a top-down approach capable of capturing international supply-chain linkages and spillovers. With platforms like the Global Industrial Ecology Virtual Laboratory (Global IELab), customised MRIO databases now enable assessments targeting specific products and regions. Building on these advances, the study proposes a pressure-oriented pesticide footprint metric (hazard load), complementing existing environmental footprint indicators to assess sustainability and equity from both production and consumption perspectives.

The study defines pesticide footprints as hazard loads (HL) of pesticides used along supply chains to satisfy consumption of goods and services. HL quantifies the body weight (bw) of non-target organisms required to absorb pesticide residues without adverse effects: HL = Mi / (NOAELi × 365), where Mi is the total mass of active ingredient i accumulated in the environment and NOAELi is the no-observed adverse effect level for mammals and birds. This pressure-oriented indicator accounts for pesticide accumulation and toxicity but excludes human health effects and acute toxicities immediately after application. Data and modelling: Global, crop-specific pesticide application rates for 2015 were sourced from PEST-CHEMGRIDS v1.0, which provides georeferenced application rates for 95 active ingredients across ten crop groups (covering 175 crops). This study modelled 80 active ingredients used on cropland and excluded non-cropland uses (pastures/hay, rangelands, livestock skin treatments, aquaculture, urban). Application rates were derived from USGS data, constrained by FAOSTAT country totals where available, and informed by soil, hydroclimatic, agronomic, and socioeconomic covariates, as well as national GM crop approvals and pesticide bans. Environmental fate: A process-based, spatially explicit environmental model (BRTSim) simulated pesticide transport and degradation at 0.5° × 0.5° global resolution, considering 1D variably saturated flow (Richards equation), advection-diffusion, volatilisation (Henry’s law), sorption (Koc and soil organic carbon), and first-order degradation modulated by temperature, moisture, pH, and biological activity. Simulations used 2015 application rates with climate forcing spanning 1970–2017 to reach near steady-state, estimating the fraction and mass of residues remaining in soil and air. Modelled residues were benchmarked against EU soil residue measurements, showing broadly consistent ranges and detectable ingredient counts. MRIO footprinting: Residue masses were converted to hazard loads using conservative (minimum) NOAEL values for mammals and birds, then aggregated by crop and country to build a satellite account aligned to a customised MRIO table constructed on the Global IELab platform. The MRIO covered 221 countries/territories (aggregated to 82 countries/territories plus 8 regions for analysis) and 6,357 sectors (aggregated to 83 sectors). Footprints were computed from both perspectives: (i) primary producer to final consumer to identify where hazard loads occur to satisfy consumption, and (ii) final point of sale to final consumer to attribute footprints to consumed commodities and services. Animal-based products were treated as secondary products whose pesticide footprints stem from feed production. Uncertainty: Robustness checks assessed sensitivity to climate variability and soil properties; Monte Carlo simulations perturbed pesticide applications (±50%), NOAEL values, and MRIO intermediate and final demands to propagate uncertainty, yielding standard deviations for total pesticide footprints across countries/regions.

• Scope: The study accounts for 3.24 Mt of pesticides (about 79% of FAOSTAT’s 4.09 Mt in 2015). Modelled residues indicate about 9.3% (0.302 Mt) of applied pesticides accumulate in the environment, corresponding to a global hazard load of 1.99 Gt-bw y−1.
• Trade: Approximately 32% of global pesticide footprints are embodied in international trade. For developed countries, 49% of their consumption-based footprints occur abroad (0.33 Gt-bw), versus 23% for developing countries (0.30 Gt-bw).
• Composition and sectors: Insecticides contribute >80% of global pesticide footprints; herbicides about 10%. Plant-based foods account for 59% of global footprints, with orchards fruits and grapes contributing 17% (0.34 Gt-bw). Animal-based foods account for ~11%. Servicing sectors (e.g., hotels, restaurants, food services) embed about 8% and other industrial sectors about 5% of global footprints. Empty-calorie foods contribute 17% of developed-country footprints but only 9% in developing countries.
• Per-capita: Global average per-capita pesticide footprint is 0.27 t-bw capita−1 y−1 (range ~0.01–1.6). High-income countries dominate the top-10 per-capita list; Spain is highest and exceeds Portugal and France by ~11% and ~105%, respectively.
• Net trade positions: China, Germany, UK, and Japan are leading net importers of hazard loads; India also ranks highly due to imports of cotton, nuts, and soybeans. The USA is the largest net exporter for insecticides and herbicides (major destinations: China 34%, Japan 7.1%, Mexico 6.9%), followed by Brazil (major destinations: USA 13.5%, China 12%, Germany 6.9%). Spain is also among top net exporters.
• Banned substances: EU27 import about 0.06 Gt-bw of hazard loads from active substances banned domestically (~34% of their imported pesticide footprints). In Sweden, Denmark, Germany, Finland, Lithuania, and Latvia, banned substances account for >90% of imported footprints.
• Major flows: The largest bilateral flow is from the USA to China (0.029 Gt-bw), mostly soybeans (73.4%). Within the EU, large flows include Spain to Germany (0.0084 Gt-bw) and Spain to France (0.008 Gt-bw), dominated by vegetables/fruits, orchard fruits and grapes, and nuts. Total intra-EU27 traded hazard loads sum to ~0.091 Gt-bw.
• Product comparisons: Orchard fruits and grapes have the highest footprints per mass and per calorie. Among grains, wheat has the lowest per-calorie footprint; rice is ~1.3× and maize ~3.5× higher than wheat. Per unit protein, soybeans are lowest among protein-rich crops; nuts are highest. Raw meat has slightly higher per-protein footprint (1.35 kg-bw kg−1 protein) than soybeans (1.24 kg-bw kg−1 protein), while eggs are lowest.
• Validation/uncertainty: Modelled soil residues and detectable ingredient counts align with EU measurements within ranges, with some median differences. Monte Carlo analysis indicates standard deviations for national/regional total footprints between ~2.4% and 16% (average ~4.5%).

The results demonstrate that international trade significantly reallocates pesticide-related environmental pressures from consuming to producing countries, revealing substantial footprint leakage. Developed countries, despite lower population shares, drive a disproportionate share of hazard loads abroad through imports, including substances banned domestically within the EU. Identifying net importers and exporters highlights how resource endowments, trade policies, and regulatory stringency shape the global distribution of pesticide pressures. Sectoral and commodity insights show plant-based foods—especially orchard fruits and grapes and nuts—dominate pesticide footprints, and services embed a notable share through food provision. Product-level normalisations reveal that some plant-based foods can have higher pesticide footprints than animal-based foods per unit mass, calorie, or protein, informing dietary and procurement strategies. These findings support policy interventions that avoid offshoring impacts, align import regulations with domestic bans, and encourage supply-chain transparency and investment in sustainable pest management in exporting countries.

International consumption drives 1.99 Gt-bw y−1 of pesticide hazard loads, with about one-third embodied in trade. Developed countries contribute substantially to pesticide pressures occurring abroad, and a significant fraction of EU imports are associated with substances banned domestically. To reduce global environmental impacts from food systems, policy should: (i) align import regulations with domestic pesticide restrictions to prevent footprint leakage; (ii) support technology transfer and remediation in exporting countries; (iii) promote integrated pest management and reduction in pesticide use; and (iv) pair shifts toward plant-based diets with measures that reduce food waste, overconsumption, and demand for empty-calorie products. Future research should advance open, high-resolution pesticide-use datasets across all settings (including livestock, aquaculture, and urban), and use decomposition analyses to quantify drivers and trends in pesticide footprints over time.

• Coverage of active ingredients: PEST-CHEMGRIDS includes the top 20 active ingredients per crop group, potentially omitting low-mass but highly toxic substances.
• Transferability of statistical relationships: Application rate models leverage relationships inferred from U.S. data; while constrained by FAOSTAT national totals, they may miss country-specific factors not present in the U.S. FAOSTAT data carry uncertainties, especially in Africa and Oceania.
• Spatial inputs: Crop distribution maps are circa 2000; expansion of cropland by ~4% to 2015 could cause slight underestimation of footprints and affect intensity estimates.
• System boundary: Non-cropland uses (pastures, rangelands, livestock skin treatments, aquaculture, urban areas) are excluded; managed pastures generally have low annual inputs, but omissions may still understate totals.
• Export vs domestic practices: Assumes similar application rates for export and domestic crops; adherence to importing countries’ maximum residue limits could bias traded footprint estimates.
• Toxicity characterization: Hazard loads use NOAEL values for mammals and birds only, excluding potential higher toxicity to other organisms (e.g., fish, earthworms) typically reported as LC50, which are not convertible to per-bw units for HL.
• MRIO assumptions: Proportionality between monetary flows and impacts, fixed input structures, and 2015 economic structure may not capture dynamic changes; results reflect 2015 patterns.
• Uncertainty: Monte Carlo analyses indicate national/regional footprint standard deviations between ~2.4% and 16% (average ~4.5%).