Physical and virtual nutrient flows in global telecoupled agricultural trade networks

The study addresses how international agricultural trade redistributes nitrogen and phosphorus through both physical (nutrients contained in traded products) and virtual (nutrient inputs required for production) flows under a telecoupling framework. With globalization, agricultural trade volumes and their environmental implications have expanded, interlinking distant socio-environmental systems. Prior work has examined either N or P flows or only physical versus virtual flows separately, leaving unclear the comparative magnitude and impacts of both flow types across products and countries. The authors aim to quantify global physical and virtual N and P flows across 320 agricultural products and 221 countries from 1997–2016, assess sending-receiving and spillover effects, and evaluate implications for resource depletion, environmental risks, and policy to enhance sustainable nutrient management.

Existing research shows dramatic increases in nutrient flows via trade: nitrogen embodied in food and feed trade increased eightfold during 1961–2010 and accounted for about one-third of global N production, while phosphorus flows increased by roughly 750% during 1961–2011, reaching 17% of global P-fertilizer input. Studies have quantified either virtual N or P flows, examined specific sectors (e.g., animal feed), or assessed limited country sets. Work has also decomposed drivers of reactive nitrogen emissions and explored interactions among multiple virtual flows (water, energy, land). However, comprehensive simultaneous measurement of both physical and virtual N and P across a broad product spectrum and global country coverage has been lacking, making it difficult to compare effects, identify inefficient flows, and appraise telecoupling effects at scale.

Scope: 320 agricultural products (including major grains, livestock products, and processed items; fish excluded) covering about 95% of global caloric consumption; 221 countries or regions; period 1997–2016. Trade volumes sourced from FAOSTAT Detailed Trade Matrix. Telecoupling framework used to conceptualize sending (exporting), receiving (importing), and spillover (re-export and indirectly affected) systems. Quantification of flows: For each trade route A→B and product i, nutrient flows equal traded mass times nutrient content. Physical flows: use FAO trade data and product-specific physical N and P contents to compute nutrients contained in traded products. Virtual flows: source-sink analysis sums all nutrient inputs required for production (inorganic and organic fertilizers, seeds, irrigation water, atmospheric deposition, biological N fixation). Virtual nutrients include both absorbed and unabsorbed fractions, thereby capturing losses to environment and soil accumulation; they represent production-side inputs effectively transferred via trade. To assign virtual flows to true producers and consumers, re-exports were excluded when tracing production origins (trade traceability per established methods). Sending-receiving effects: For physical nutrients, effects are the net physical nutrients transferred from sending to receiving systems, accounting for re-exports. For virtual nutrients, effects indicate nutrient savings or waste relative to producing in the importing country, calculated from differences in country-specific virtual nutrient intensities and adjusted for re-exports. Positive values denote efficient flows (saving nutrients), negative values denote inefficient flows (wasting nutrients). Spillover effects: Quantify nutrient impacts due to re-export processes when trade importers serve as transit hubs. Computed from differences in nutrient contents (physical or virtual) between original producers and ultimate destinations, weighted by re-export volumes. Telecoupling effects: For each route and globally, sum sending-receiving and spillover effects for both physical and virtual nutrients to obtain total telecoupling effects. Data and computation: Product-level physical nutrient contents and country- and product-specific virtual nutrient intensities compiled from literature and supplementary datasets; trade matrices cleaned for re-export traceability. Analyses and visualizations performed in R 4.0.2 and Python 3.8.

• Growth in flows (1997–2016): Physical nutrient flows increased from 10.3 to 27.1 Tg N and from 1.4 to 3.5 Tg P; virtual nutrient flows increased from 13.4 to 36.6 Tg N and from 8.9 to 24.5 Tg P—both about 2.8-fold increases.
• Trade structure: 48 countries experienced more than tenfold increases in both physical N and P exports over the period. China’s physical imports rose 6.98-fold (N) and 6.27-fold (P); it became the largest nutrient importer from 2003 to 2016.
• Product composition: In 2016, crop and processed crop trade accounted for about 90% of physical flows. Virtual flows were roughly split between crops and animal products. Soybeans and derivatives contributed 29.8% of physical N in 2016. Beef and cattle products made up 12.4% (N) and 16.8% (P) of total virtual flows but only 5.0% (N) and 4.3% (P) of physical flows.
• Spatial patterns: Net receiving systems concentrated in Asia; net sending systems concentrated in North and South America. Net-sending systems declined from 99 to 62 (1997→2016), while net-receiving increased from 122 to 159. In 2016, top 10 net senders contributed >70% and top 20 contributed >90% of exported physical nutrients. The top 5 net receivers imported 33% of total physical flows.
• Country highlights (2016): China net imported 6.06 Tg N and 0.62 Tg P (20.0% and 16.0% of global physical N and P flows). Japan (0.95 Tg N, 0.13 Tg P) and Mexico (0.78 Tg N, 0.23 Tg P) followed. The United States (5.20 Tg N, 0.55 Tg P) and Brazil (4.21 Tg N, 0.39 Tg P) were leading net exporters.
• Trade routes and adjacency: Trade routes increased from 7,170 (1997) to 27,819 (2016). Forty-one routes exceeded 0.1 Tg physical N and 59 exceeded 0.01 Tg physical P. Average flow volume between adjacent systems was over three times that between non-adjacent systems; for non-adjacent systems, average flow increased with distance. Many virtual flows were orders of magnitude larger than physical flows (e.g., 256 virtual N routes >1000× corresponding physical N; 409 virtual P routes >1000× physical P).
• Dominant bilateral flows: Soybeans drove the largest US→China physical N flow; 40.1% of China’s imported physical N came from the United States, representing 20.2% of US physical N exports.
• Telecoupling contributions: In 2016, physical N+P flows equaled about 27% of global nutrient content in consumed agricultural products; virtual N+P flows equaled about 33.7% of total soil nutrient inputs to global agriculture. Virtual sending-receiving effects reached up to 62.3 Tg N and 73.9 Tg P, indicating substantial nutrient savings via trade from higher- to lower-efficiency producers. Some commodities (lentils, hazelnuts, coconuts, sunflower seeds, seed cotton) showed virtual saving effects over 100 times their physical effects.
• Spillover effects: Largest N spillovers observed on the Germany–Netherlands (physical, 1.37 Tg) and China–US (virtual, 7.56 Tg) routes. In 2016, total N spillover effects were 3.81 Tg and P spillovers 1.23 Tg. Wheat, corn, rapeseed, and soybeans had the largest spillover magnitudes.

The global agricultural trade network shows a highly uneven nutrient export structure, with about 10% of countries accounting for roughly 90% of exports in 2016. Positive feedbacks from growing demand in receiving systems have incentivized production expansion in sending systems, sometimes driving deforestation and ecosystem conversion, notably via soybean expansion in Brazil influenced by shifting US–China trade dynamics. Net sending systems face risks of soil nutrient depletion and fertility decline, particularly for phosphorus due to the difficulty and cost of recycling. Case evidence includes Argentina’s Pampas facing P deficits amid high physical P outflows and low recycling rates. Virtual nutrient flows shift environmental burdens from receiving to sending systems, externalizing pollution associated with production (e.g., reactive N losses) to exporters. Conversely, countries with nutrient-poor soils or limited fertilizer input (e.g., parts of Africa, Jordan) benefit from inflows of physical nutrients and conserve domestic resources via imports of virtual nutrients; yet increased consumption, particularly of meat, can elevate local nutrient loads and eutrophication risks in receiving systems with limited absorptive capacity. Land-use changes induced by imports can also exacerbate local nutrient pollution. Despite overall global nutrient savings due to trade (positive sending-receiving effects), inefficient flows persist, where products move from lower- to higher-efficiency producers, wasting nutrients. Addressing these inefficiencies through production and trade restructuring, demand-side shifts (e.g., reduced beef and dairy consumption), and technology upgrades in low-efficiency regions can enhance conservation outcomes. Incorporating agents and governance elements into telecoupling analyses and aligning trade policies with green procurement can further reduce inefficient flows and strengthen sustainability.

This study simultaneously quantified global physical and virtual nitrogen and phosphorus flows embedded in agricultural trade across 320 products and 221 countries (1997–2016) within a telecoupling framework. The analysis reveals rapid growth in both flow types, substantial contributions to global nutrient redistribution, and net positive global nutrient savings via sending-receiving effects. However, benefits are uneven, with environmental burdens and resource depletion risks concentrated in net sending systems, and pollution risks rising in some net receiving systems. Policy actions should target elimination of inefficient flows, improvement of production efficiencies, promotion of nutrient recycling (e.g., organic manure, biosolids, bio-fertilizers), and integration of environmental risks and soil nutrient balances into agricultural planning and trade policies. Future research should incorporate agent dynamics and detailed upstream/downstream supply-chain representations, improve data granularity on product nutrient contents, and refine assessments of spillover effects to guide sustainable nutrient management and trade.

• The analysis did not differentiate nutrient contents between exported and non-exported products due to data constraints, potentially affecting precision.
• The study boundary focused on bilateral agricultural trade flows and excluded detailed upstream fertilizer production and downstream refined product chains, as well as explicit modeling of agents and causes within the telecoupling framework.
• Fisheries products were excluded; while capture fisheries involve relatively low virtual nutrient inputs, their omission limits completeness for some regions and diets.
• Data limitations in re-export tracing and virtual nutrient intensity estimates may introduce uncertainties; further data integration at higher resolution is needed.