Telecoupling Perspective on the Evolution and Driving Factors of Virtual Cropland Networks in Global Wheat Trade

Abstract

The international wheat trade serves as a vital pathway for balancing the global food supply and demand while facilitating the cross-regional allocation of cropland resources. Based on the telecoupling framework, this study constructed a global virtual-cropland-flow network using wheat trade data from eight time points between 1995 and 2023. Social network analysis and quadratic assignment procedure regression were applied to examine its structural evolution and driving factors. The findings reveal that (1) while growing in connectivity, the virtual cropland network exhibits structural vulnerability and evolutionary complexity. (2) The network demonstrated a clear telecoupled structure, with the sending system shifting from U.S.–Canada dominance towards multipolarity, and the receiving system centered in Asia, Africa, and Latin America, with China at its core. The United States and France are major spillover systems. (3) Economic development and foreign demand significantly promote the establishment and intensification of trade relationships between countries. Geographical distance has a dual effect: it strongly negatively influences trade initiation but can be overcome by high complementarity between countries during trade deepening. (4) International wheat trade contributes to global cropland savings but also introduces systemic risks and environmental spillovers in some countries. The results provide theoretical support for building sustainable food trade and agricultural resource governance systems and offer important insights for advancing SDG 2 (Zero Hunger), SDG 12 (Responsible Consumption and Production), sustainable land systems, and the optimization of global land governance.

1. Introduction

Cultivated land use, a quintessential example of human–environment interaction, is a core focus of geographical research on human–land relationships [1]. These relationships persist under dynamically evolving productivity levels and social structures, with their scope and complexity expanding as society develops. Since the mid-20th century, deep-seated industrialization, urbanization, information, and globalization have markedly amplified human impacts on the Earth system, propelling the planet into an epoch dominated by human activity: the Anthropocene [2,3].

In the context of accelerating global interconnectivity, the localized paradigm of human–land relations—often captured by the adage “one’s native land sustains its people”—has become increasingly inadequate for explaining the growing complexity of human–environment interactions in the context of globalization. The cognitive framework for understanding human–land relations is progressively shifting from a place-based, territorial perspective to a globally networked one [4]. Within this intellectual shift, the telecoupling framework, proposed by Liu et al. [5] in 2013, offers a valuable perspective for examining socioeconomic and environmental interactions between coupled human–natural systems across distances. This theoretical framework comprises five interconnected components: coupled human–natural systems (including sending, receiving, and spillover systems), flows, agents, causes, and effects. It emphasizes long-distance socioeconomic and environmental linkages across multiple locations and is oriented toward global sustainability [6]. Leveraging its strengths in integrating multiple spatiotemporal scales and facilitating cross-sectoral systems analysis, the telecoupling framework has rapidly emerged as a cutting-edge research frontier in the field of sustainability science. Scholars are not only focusing on theoretical refinements, including literature reviews [7], framework standardization [8], and conceptual discussions [9], but are also actively extending the framework’s application to empirical research domains. Within this empirical expansion, agricultural trade has become a particularly prominent area for applying the telecoupling framework [10,11,12].

As one of the world’s three major staple grains, wheat occupies a central role in the increasingly integrated international agricultural trade [13]. As a land-intensive commodity, wheat exports have long been concentrated in a limited number of land-abundant countries, primarily in Europe and North America. In recent years, Russia and Ukraine have enhanced their positions in the wheat export network by capitalizing on their resource endowments [14]. Constrained by the immobility of croplands and region-specific consumer preferences, international wheat trade serves as a crucial mechanism for addressing geographical imbalances in wheat supply and demand and facilitating the cross-regional allocation of cropland resources [15,16]. Scholarly research on wheat trade has addressed several aspects. Internationally, studies have examined the evolution and resilience of global wheat trade networks (e.g., Fair et al.) [17], assessed the vulnerabilities of wheat producers (Gutierrez-Moya et al.) [18], and explored the effects of heterogeneous extreme-weather stresses on trade flows (Vishwakarma et al.) [19]. Early research in China mainly adopted qualitative, food security-oriented approaches to analyze shifts in wheat trade patterns [20] and the related policy measures [21]. As research deepens, quantitative approaches such as network analysis [14], the gravity model of trade [22], and Quadratic Assignment Procedure (QAP) modeling [23] are increasingly employed to examine the structure and determinants of wheat trade. However, current studies remain predominantly centered on the wheat trade network itself, with insufficient attention paid to the cross-regional flows and reconfiguration patterns of cropland resources embedded in trade and their linkages with land systems.

International wheat trade fundamentally embodies the flow of cropland resources, demonstrating long-distance interactions across transregional human–land systems [24]. To quantify the land resources embedded in agricultural trade, Luo Zhenli advanced the concept of “virtual land”, building upon the earlier notion of “virtual water” [25]. As research progressed, scholars further refined this idea and introduced the term “virtual cropland.” Unlike physical croplands, which can be directly observed, virtual croplands are an analytical abstraction derived from agricultural products, representing cropland resources implicitly traded through commodity flows as a factor of production [26]. Compared to the broader concept of the “land footprint,” virtual cropland specifically highlights arable land as a scarce and productive input. The flows of virtual croplands not only reflect the global matching of cropland supply and demand but are also intrinsically linked to core issues in land system science, such as land cover change, the differentiation of land productivity, and land degradation risks.

In terms of research scale and focus, existing studies have primarily concentrated on the Chinese context, covering the spatial patterns [27] and drivers [28,29] of virtual cropland flows and the relationship between virtual cropland trade and national food security [30]. At the global scale, while studies have examined the characteristics of virtual cropland flows associated with agricultural products such as soybeans [31] and grains [32], as well as the land-saving effects of global grain trade [33], and some scholars have utilized network analysis to reveal the connectivity features of virtual cropland flows within global agricultural trade [34], research specifically targeting wheat—a core staple crop—remains lacking. No study has systematically analyzed the structural characteristics and influencing factors of the virtual cropland resource network embedded within international wheat trade from a global perspective. Moreover, global croplands are projected to face large-scale and uneven loss pressures under current climate mitigation measures [35]. Therefore, key questions deserve further in-depth investigation: What are the flow patterns and network characteristics of the cropland resources implicitly transferred through global wheat trade? What are the main factors influencing the reallocation of cropland resources? How is it telecoupled with land system sustainability?

Building on this foundation, this study applied the telecoupling framework to examine international wheat trade. We employed social network analysis (SNA) to quantitatively characterize the spatiotemporal evolution of virtual cropland flows embedded in the trade network. Furthermore, QAP regression was used to identify the key influencing factors. By employing the telecoupling framework to analyze virtual cropland flows within the global wheat trade, this study empirically demonstrates its efficacy in examining long-distance land-system interactions across “sending-receiving-spillover” systems in the context of globalized staple food trade, thereby extending its application to the study of transnational resource flows. Moreover, the structure and drivers of the virtual cropland network identified here offer scientific insights and policy-relevant guidance for optimizing the global allocation of cropland resources, fostering a sustainable food trading system, and contributing to the attainment of the United Nations Sustainable Development Goals (SDGs).

The remainder of this paper is organized as follows. Section 2 describes the data sources and outlines the research design, which comprises the virtual cropland accounting method, social network analysis, and quadratic assignment procedure regression analysis. Section 3 systematically presents the empirical results, covering the overall and nodal structural evolution of the global virtual cropland flow network, an analysis of its driving factors, and the identified telecoupling effects. Section 4 concludes the paper, synthesizes the key findings, discusses their policy implications, and suggests directions for future research.

2. Data Sources and Methods

Drawing on the telecoupling framework, this study combined virtual cropland accounting, social network analysis, and quadratic assignment procedure (QAP) regression to examine the spatiotemporal evolution and driving factors of the virtual cropland network embodied in the global wheat trade. The analysis was conducted in two stages. First, the structural features of the virtual cropland network were analyzed. Virtual cropland flows were estimated based on wheat trade volumes and national yield data per unit area. To ensure consistency, countries lacking yield per unit area statistics were excluded, retaining only those with complete data sets. Second, QAP regression was performed. Building on the sample from the first stage, countries were screened further according to the model specifications. Where independent variables were missing for individual years, gaps were interpolated using trends from adjacent years, and countries with extended periods of missing data were omitted. The overall methodological workflow is illustrated in Figure 1.

2.1. Data on Wheat Trade and the Calculation of Virtual Cropland

This study used wheat export volumes as a proxy for trade volumes. Data on wheat exports (measured in metric tons) and yields (measured in metric tons per hectare) were obtained from the Food and Agriculture Organization Statistical Database (FAOSTAT). The wheat trade volumes for China’s regions of Hong Kong, Macao, and Taiwan were integrated into the data for China, whereas data for overseas territories lacking independent sovereign governments were combined with their respective administering countries.

To systematically examine the influence of events, technological advances, and policy shifts on wheat trade, this study employs eight time points: 1995, 2001, 2007, 2014, and the consecutive period of 2020–2023. These points were chosen to capture pivotal junctures in the evolution of global wheat trade: 1995 coincides with the establishment of the WTO and the onset of agricultural trade liberalization; 2001 aligns with China’s WTO accession and a phase of production technology upgrades; 2007 reflects structural adjustments triggered by the global food crisis; 2014 captures synergies between technological innovation and expanding regional trade agreements; and the 2020–2023 interval provides a continuous lens to track sustained disruptions and adaptations resulting from the COVID-19 pandemic, the Russia–Ukraine conflict, and evolving national food security policies [36]. Collectively, these time points span the progression of wheat trade from globalization and regional integration to greater diversification.

For each selected year, the analysis included only countries that were actively engaged in wheat trade and for which wheat yield per unit area data were reported in FAOSTAT. Countries lacking reliable yield-per-unit-area statistics owing to reporting gaps or country-specific conditions were excluded. The number of country samples and exclusions at each time point are presented in

Table 1. The number of country samples and exclusions.

Year	The Initial Number of Trading Countries	The Number of Countries Excluded (Due to the Absence of Data on Yield per Unit Area)	Final Number of Trading Countries
1995	166	5	161
2001	167	3	164
2007	168	3	165
2014	180	3	177
2020	182	4	178
2021	181	1	180
2022	178	0	178
2023	179	4	175

. Using the resulting sample, wheat trade volumes were converted into virtual cropland equivalents, forming the basis for constructing annual trade networks.

The concept of virtual cropland is derived from that of “virtual water” and denotes the abstract utilization of cropland resources embodied in specific physical products [37]. Currently, there are two primary methods for calculating the trade volume of virtual croplands. The first method, from the producer’s perspective, defines virtual croplands as the actual amount of cropland resources used to produce a given product in its origin region, thereby reflecting the real land resources embodied in agricultural trade. The second method, from the consumer perspective, defines it as the amount of cropland resources required to produce an equivalent product in the consuming region [28]. The former approach relies on the yield of the exporting country as the accounting benchmark, capturing the real land resource appropriation and spatial patterns associated with trade, thus providing a foundation for evaluating resource flow sustainability. The latter adopts the yield of the importing country as the accounting basis, illustrating the potential reduction in cropland use under an “import substitution” scenario and highlighting the resource conservation potential enabled by trade [38]. For example, consider Ukraine’s export of wheat to Ethiopia in 2023: calculated using Ukraine’s yield (4.6 t/ha), this trade embodies approximately 13,043.5 hectares of virtual cropland, representing the actual resource cost of the export. If assessed using Ethiopia’s domestic yield (3.1 t/ha), producing the same quantity of wheat locally would require approximately 19,354.8 hectares of cropland. A comparison of the two figures reveals that by importing wheat from a region with higher agricultural productivity, Ethiopia theoretically conserved approximately 6311.3 hectares of cropland. This example underscores the role of international trade in promoting a more efficient global allocation of cropland resources.

Based on this, the present study used national wheat yield as the conversion factor to translate international wheat trade volumes into virtual cropland areas. Although this method inherently assumes cropland homogeneity, it remains sufficiently applicable for quantifying the scale of virtual croplands, which is the focus of this study. The associated limitations are discussed in Section 4.1. Building on this foundation, the virtual cropland volume from the exporter’s perspective, virtual cropland volume from the importer’s perspective, and virtual cropland balance were calculated separately using the following formulas:

$${VCT}_{E,t}={\displaystyle \frac{Q_t^{ab}}{Y_{E,t}}}$$

$${VCT}_{I,t}={\displaystyle \frac{Q_t^{ab}}{Y_{I,t}}}$$

$${VCTD}_t={VCT}_{I,t}-{VCT}_{E,t}$$

where ${VCT}_{E,t}$ denotes the virtual cropland volume from the exporter perspective in year t. ${VCT}_{I,t}$ denotes the virtual cropland volume from the importer perspective in year t. $Q_t^{ab}$ represents the volume of wheat trade between trading countries a and b in year t. $Y_{E,t}$ denotes the wheat yield of the exporting country between trading countries a and b in year t. $Y_{I,t}$ denotes the wheat yield of the importing country between trading countries a and b in year t. ${VCTD}_t$ represents the virtual cropland difference between trading countries a and b in year t. A positive value indicates that wheat trade flowed from the country with higher land productivity to the one with lower land productivity, thereby conserving global cropland resources for the same level of output, enhancing land-use efficiency, and supporting sustainable cropland use. Conversely, a negative value suggests that trade results in an inefficient allocation of cropland resources. Detailed data are available in Supplementary Materials.

2.2. Social Network Analysis

Social Network Analysis (SNA) is an analytical method that characterizes complex relational patterns into network configurations. It reveals the influence of network structure on the functions of groups and individuals by examining the interplay between structure and function [39]. This approach is widely applied in trade network research. Telecoupled systems represent a typical form of geographical network that exhibits distinct structural features [7]. The network topology and interconnections among the components can be effectively characterized using SNA.

Based on this, this study employs SNA to construct a telecoupled network model, denoted as G, which represents the global virtual cropland trade embodied in wheat trade. Participating countries (regions) in global wheat trade serve as network nodes, bilateral wheat trade flows constitute the edges, and the volume of virtual cropland trade embodied in wheat trade defines the edge weights. The network is formally defined as follows.

G = (N, E, W, T)

where $N$ represents the set of all nodes (countries or regions). $E$ represents the set of all edges (trade linkages). $W$ represents the set of all edge attributes (trade weights), specifically the virtual cropland trade volume embodied in each bilateral wheat trade flow. $T$ represents the set of years for which international wheat trade embodied the virtual cropland trade network.

Based on this, we constructed an unweighted adjacency matrix indicating the presence or absence of trade linkages between country pairs (matrix element = 1 if a trade linkage exists, 0 otherwise) and a weighted adjacency matrix representing the intensity of trade (matrix element = the total volume of virtual cropland trade between the two countries). Using these matrices, we calculated the key network topology metrics using Ucinet6. This enables the identification of the sending, receiving, and spillover systems of virtual croplands embodied in wheat trade, along with the direction and intensity of the trade flows. Finally, we quantified the evolutionary characteristics of the global virtual cropland resource network embodied in international wheat trade. Based on existing research, in this study, we operationalize the measurement of social network structure using three key metrics at the macro level—network density, average clustering coefficient, and average path length—to quantify global connectivity and cohesion. At the micro level, we employ relative degree centrality, relative closeness centrality, and betweenness centrality to examine the structural prominence and brokerage potential of individual nodes within the network (see

Table 2. Overall network structure metrics.

Indicator Name	Indicator Description	Expression
Network density (D)	The ratio of the actual number of trade connections in the network to the maximum possible number of connections. It measures the overall connectivity of a network. A higher density indicates more connections between nodes, reflecting more frequent trade interactions between countries.	$D={\displaystyle \frac{2m}{n\left(n-1\right)}}$ $m$: number of actual trade links. $n$: total number of nodes.
Average clustering coefficient (C)	The average clustering coefficient of all nodes in the network. The clustering coefficient of a node is defined as the ratio of the actual number of links between its neighbors to the maximum possible number of links. This reflects the extent to which neighboring nodes are clustered together.	$C={\displaystyle \frac{1}{n}}{\sum }_i^n{\displaystyle \frac{2m_i}{k_i\left(k_i-1\right)}}$ $k_i$: number of neighbors of node $i$. $m_i$: number of actual links between the neighbors of node $i$.
Average path length (L)	The average number of edges along the shortest paths for all possible pairs of nodes in the network. It indicates the efficiency of the connectivity between the nodes.	$L={\displaystyle \frac{{\sum }_{ij}{d}_{ij}}{n\left(n-1\right)}}$ $d_{ij}$: distance (shortest path length) between node $i$ and $j$ node.

and

Table 3. Node-level structural metrics.

Indicator Name	Indicator Description	Expression
Relative degree centrality ($C_D\left(i\right)$)	The number of nodes directly connected to a given node in the network. A higher value indicates a stronger ability of the node to form connections and a more central position in the network.	$C_D\left(i\right)={\displaystyle \frac{{\sum }_i{a}_{ij}}{n-1}}$ ${\sum }_i{a}_{ij}$: absolute degree centrality of node $i$.
Relative closeness centrality ($C_C\left(i\right)$)	Reflects how close a node is to all other nodes in the network and its ability to avoid being controlled by others. A higher value indicates greater independence and efficiency in reaching the other nodes.	$C_C\left(i\right)={\displaystyle \frac{\sum _j{d}_{ij}}{n-1}}$ $d_{ij}$: distance (shortest path length) between node $i$ and node $j$.
Betweenness centrality ($C_B\left(i\right)$)	The proportion of all shortest paths between pairs of nodes in the network that pass through a given node. It reflects the role of the node as a bridge or intermediary in the network.	$C_B\left(i\right)={\sum }_j^n{\displaystyle \sum _k^n}{\displaystyle \frac{g_{jk\left(i\right)}}{g_{jk}}}$ $g_{jk}$: total number of shortest paths between node $j$ and node $k$. $g_{jk\left(i\right)}$: number of the shortest paths that pass through node $i$.

). Detailed data are available in Supplementary Materials.

2.3. QAP Regression Analysis

The Quadratic Assignment Procedure (QAP) is a method that takes “relational” data as its research object. Based on the permutation of matrix data, it compares the corresponding element values in two or more square matrices, calculates the correlation coefficient between the matrices, and simultaneously performs non-parametric tests on this coefficient [40]. QAP regression analysis can effectively overcome the issue of multicollinearity among the data [41] and is widely applied in social network research.

2.3.1. Selection of Variables

This study employs the virtual cropland trade flow matrix and dichotomized trade relationship matrix between countries (regions) as the dependent variable matrices. Guided by the telecoupling perspective and integrating theories of spatial interaction, factor endowment, comparative advantage, and new trade theory, this study selects influencing factors for the evolution of the virtual cropland trade network embodied in international wheat from two aspects—complementarity and accessibility between regional systems and across four dimensions: demand, supply, distance and facilitation [11]. Complementarity refers to the supply demand relationship of a specific element between regional systems, which serves as a fundamental condition for coupling interactions between systems [42]. In this study, three indications were selected to represent the demand-side factors of systemic complementarity: economic level, foreign wheat demand, and the consumption structure. Three other indicators were chosen to represent the supply side factors: wheat cultivation area, wheat yield per unit area, and renewable freshwater resources. Accessibility is a quantitative measure of the ease of connection between sending and receiving systems. It is influenced by both distance and facilitating conditions and serves as a critical factor for the coupling between systems [43]. This study selects two indicators—national distance and contiguity—to represent the distance dimension between systems, and two additional indicators—national governance level and WTO co-membership—to represent the facilitating conditions between systems. The selection basis and specific explanations of the variables are presented in

Table 4. Definition and selection basis of variables.

Dimensions	Variables	Symbols	Explanations
Demand	Economic level	G	Measured by the gross domestic product (GDP). Countries at different economic development levels often exhibit differences in their positions within the global value chain and variations in the structure of production and consumption of specific production [44]. Greater disparities in economic development imply larger industrial gaps, which may lead to greater trade opportunities [45].
	Foreign wheat demand	N	It is represented by the wheat self-sufficiency ratio (SSR). A lower self-sufficiency ratio indicates that domestic wheat production is less able to meet national demand, implying a higher reliance on imports. The self-sufficiency ratio was calculated using the following formula: Production/(Production + Imports − Exports).
	Consumption structure	S	It is expressed as total domestic wheat demand/total population. Higher values indicate stronger consumer preference for wheat products. The total domestic wheat demand was estimated using the following formula: production + imports − exports.
Supply	Wheat cultivation area	A	It was measured as the proportion of the actual harvested wheat area to the total cultivated land area. A higher proportion reflects more favorable endowment conditions of cropland resources for wheat production in a country (or region), indicating the greater importance of wheat in the agricultural structure. This may also imply a higher capacity to meet domestic demand and pursue export opportunities than other countries and wheat products.
	Wheat yield per unit area	O	It is measured as wheat output per hectare of land. A higher yield per unit area suggests improved economies of scale and thus greater export advantages in the international market.
	Renewable freshwater resources	F	It is expressed as per capita renewable internal freshwater resources. Countries with scarce freshwater resources often tend to develop water-saving industries, while those with per capita renewable freshwater resources above the world average are typically net exporters of virtual croplands [15].
Distance	National distance	D	Measured as the spherical distance between national capitals. Geographical distance partly reflects transportation costs and competitive dynamics in agricultural trade [16].
	Contiguity	C	This indicates whether two countries (or regions) share a land border. A value of 1 was assigned if they were adjacent, and 0 was assigned otherwise. Contiguous countries often leverage their geographic proximity to facilitate grain trade [46].
Convenience	National governance level	P	A sound political environment is a key component of the international trade landscape and an essential factor in sustaining stable agricultural trade. Greater disparities in governance between countries may reduce the likelihood of trade [47]. Drawing on the methodology of Cheng et al. [23], a national governance index was constructed as the arithmetic mean of six Worldwide Governance Indicators from the World Bank database: Voice and Accountability, Political Stability and Absence of Violence/Terrorism, Government Effectiveness, Regulatory Quality, Rule of Law, and Corruption Control. This index reflects the quality of a country’s political environment.
	WTO co-membership	W	The WTO plays a significant role in establishing trade rules, reducing trade barriers, and promoting international cooperation and policy coordination. Shared WTO membership facilitates regional trade cooperation. A value of 1 is assigned if both countries (or regions) are WTO members and 0 otherwise.

.

2.3.2. Model Specification

Based on the theoretical framework, we constructed a Quadratic Assignment Procedure (QAP) multiple regression model to analyze the determinants of the virtual cropland network embedded in international wheat trade. The general form of the model is specified as

Y = β₀ + β₁G + β₂N + β₃S + β₄A + β₅O + β₆F + β₇D + β₈C + β₉P + β₁₀W + μ

where Y is the dependent variable, represented by both unweighted and weighted adjacency matrices of the virtual cropland network. β₀ is the intercept. β₁ to β₁₀ are the coefficients to be estimated in the model. μ denotes the error term. G denotes the difference matrix in GDP between trading partners. N denotes the difference matrix of wheat self-sufficiency rates. S denotes the difference matrix of consumption structure. A denotes the difference matrix in the share of the wheat harvest area. O denotes the difference matrix of wheat yields per unit area. F denotes the difference matrix of per capita renewable inland freshwater resources. D denotes the geodesic distance matrix of the national capitals. C denotes a binary matrix indicating shared borders. P denotes the difference matrix of national governance performance. W denotes a binary matrix indicating mutual membership in the WTO.

The names, detailed descriptions, construction methods of the matrices, hypothesized effects, and data sources for all independent variables are summarized in

Table 5. Description of independent variables.

Criterion Layer	Element Layer	Variables Name	Variables Description	Matrix Processing	Expected Effect	Data Source
Complementarity	Demand factors	Economic level (G)	Gross domestic product	Difference matrix	+	World Bank database
		Foreign wheat demand (N)	Wheat self-sufficiency rate	Difference matrix	+	FAO database
		Consumption structure (S)	Total domestic wheat demand/Total domestic population	Difference matrix	+	World Bank database
	Supply factors	Wheat planting area (A)	Percentage of wheat-harvested area to total arable land	Difference matrix	+	FAO database
		Wheat yield (O)	Difference in yield per unit area	Difference matrix	+	FAO database
		Renewable freshwater (F)	Per capita available productive inland freshwater resources	Difference matrix	+	World Bank database
Accessibility	Distance factors	National distance (D)	Spherical distance between national capitals	Multi-value matrix	−	CEPII database
		Contiguity (C)	Whether territories are adjacent	Binary matrix	+	CEPII database
	Convenience factors	Governance level (P)	Worldwide governance indicators	Difference matrix	−	World Bank database
		WTO membership (W)	Whether both are WTO members	Binary matrix	+	WTO official website

Note: “+” and “−” represent the expected positive and negative directions of the variable effects, respectively.

. Regarding data processing, missing values in single-year cross-sections were addressed by imputation using data from adjacent years. This method was applied to supplement renewable freshwater resource data for 2023 and national governance indicators for 1995, 2001, and 2023. To meet the model requirements and maintain a consistent number of country nodes across all variables within each cross-section, observations with extensive missing data across multiple years and variables were excluded from the analysis. Finally, to eliminate scale differences, the independent variable matrices, including both difference and multi-value matrices, were standardized using the normalization function in UCINET 6.

3. Results

3.1. Overall Network Evolution Characteristics

Between 1995 and 2023, the global wheat trade volume surged from 8.6298 million tons to 16.6535 million tons, whereas the virtual cropland trade expanded from 2.9097 million hectares to 4.8889 million hectares. The consistent growth in both trade volume and virtual cropland underscores not only the practical necessity for various countries and regions to secure food supplies but also the pivotal role international trade plays in the global distribution of cropland resources.

Network connectivity indicators reveal that the virtual cropland network within the global wheat trade has continued to evolve and deepen (Figure 2). In 1995, the network displayed a decentralized small-world cluster pattern, marked by a “low density, low average path length, and high average clustering coefficient,” indicating generally loose global connections. Regional blocs like the European Union and the North American Free Trade Agreement formed small local clusters, yet comprehensive global trade linkages were lacking. By 2007, the network’s average clustering coefficient had dropped to its lowest point, whereas the network density increased. The world food crisis and extreme weather events led to the dissolution of the original local clusters, resulting in more diverse wheat trading partnerships. As international wheat trade interactions evolved, network density steadily rose, peaking between 1995 and 2020, while the average clustering coefficient rebounded. This suggests that global interconnections in virtual cropland trade have intensified, with emerging export clusters and a restructured regional trade order following the food crisis.

Between 2020 and 2023, the global wheat trade virtual cropland network exhibited a slight decrease in density, a steady rise in the average clustering coefficient, and a notable reduction in average path length. These structural shifts illustrate the network’s adaptation to the compounded external shocks of the COVID-19 pandemic and the Russia–Ukraine conflict. In response to the pandemic outbreak in 2020, numerous countries introduced wheat export restrictions to bolster domestic food security. Coupled with widespread logistical constraints and supply chain disruptions, these policies have weakened the established trade connections. Concurrently, risk-averse behavior has encouraged nations to deepen cooperation with neighboring countries and traditional trade patterns, fostering the emergence of regional trading clusters. The escalation of the Russia–Ukraine conflict in 2022 further reconfigured the network topology. Reduced wheat production in Ukraine, combined with export bans adopted by several nations, disrupted prevailing trade arrangements and contributed to the fragmentation of the wheat trade network [48]. For example, Ukrainian wheat exports declined from 19.4 million tons in 2021 to 11.22 million tons in 2022 due to the war. Consequently, several Middle Eastern and North African countries, historically dependent on wheat imports from Russia and Ukraine, were compelled to diversify their sources, turning to alternative supplies such as Australia.

In conclusion, influenced by a confluence of factors, including economic development, geopolitical influence, and trade agreements, the global wheat trade virtual cropland network has evolved to exhibit greater structural complexity. Among these drivers, the Russia–Ukraine conflict acted as a critical catalyst, significantly altering global wheat supply patterns and trade routes, thereby steering phased structural adjustments in the network during this period.

3.2. Evolutionary Characteristics of Individual Structure

The individual structural characteristics of the virtual cropland trade network embodied in wheat trade were measured using Ucinet 6 software, yielding degree centrality, closeness centrality, and betweenness centrality for each node across six time points spanning 1995 to 2023. The specific analysis is as follows.

(1)

Degree centrality.

To enhance comparability across countries with disparate trade volumes and explicitly characterize the flow trajectories of virtual croplands, this study implements standardization on degree centrality, yielding relative out-degree centrality and relative in-degree centrality. Overlaying these standardized metrics with the virtual cropland flow maps for each time cross-section from 1995 to 2023 (Figure 3) enables the direct identification of core network nodes, which offers robust visual evidence for interpreting the spatiotemporal evolution of the global virtual cropland trade regime.

From the perspective of out-degree relative centrality, countries such as the United States, Canada, Australia, Russia, and Kazakhstan demonstrate high out-degree relative centrality, reflecting their strong capacity to export virtual cropland through wheat trade. These nations serve as the primary sources of virtual croplands, having developed extensive export connections with other countries and holding a central position in the global network. From 1995 to 2023, the structure of virtual cropland outflows underwent significant evolution, shifting from a duopoly dominated by the United States and Canada to a competitive landscape involving four major exporters: Australia, Canada, Kazakhstan, and the United States (Figure 4). During this period, Australia surpassed the United States to become the largest exporter, and Kazakhstan emerged as the third largest. International wheat trade exhibits distinct seller market characteristics, where exporting countries play a decisive role in shaping trade structures and resource flows. Moreover, their roles in the global allocation of cropland resources differ. As traditional leading exporters, the United States and Canada, with average wheat yields of 2.89 t/ha and 2.61 t/ha, respectively, act as net conservers of virtual croplands through their exports. Conversely, emerging exporters such as Australia and Kazakhstan, with lower average yields of 1.76 t/ha and 1.04 t/ha, respectively, generally represent net consumption of virtual croplands. In certain trade relationships with importing countries, their export activities have led to inefficient substitution of cropland resources in the exporting countries. Additionally, by 2023, Brazil in South America showed a significant rise in ranking, whereas Argentina’s ranking fluctuated, dropping from third in 1995 to eighth in 2023. The position of Black Sea region countries, such as Ukraine, in virtual cropland exports has fluctuated but has shown an upward trend.

From the perspective of in-degree relative centrality, China, Egypt, Japan, Brazil, Indonesia, and Italy consistently maintained relatively high in-degree centrality from 1995 to 2023, identifying them as the main recipients of virtual croplands in the wheat trade. Among these, China ranked first in in-degree relative centrality in 1995, 2020, and 2023, underscoring its core position in the virtual cropland-inflow network. Notably, the primary source of China’s virtual cropland inflows has shifted from the United States to Australia. This transition is closely linked to China’s vast wheat consumption market, limited domestic cropland resources, and evolving international relations. In 2023, Uzbekistan overtook Brazil to become the second-largest receiving system, with all its virtual cropland inflows originating in Kazakhstan. Egypt, in North Africa, has consistently ranked highly, reflecting its large population, wheat-dependent dietary traditions, and relatively scarce cropland. Concurrently, the global significance of Central Asian countries has been steadily increasing in both the sending and receiving of virtual croplands embedded in wheat trade.

(2)

Closeness centrality.

Closeness centrality measures the relative distance of a country’s trade relationships with others within the network. In the virtual cropland flow network embedded in international wheat trade, out-closeness centrality and in-closeness centrality reflect the accessibility and dependency of a node (country or region) from the perspectives of virtual cropland outflow and inflow, respectively.

Specifically, from 1995 to 2023, countries ranking high in out-closeness centrality were primarily developed nations in Europe and America, such as the United States, Germany, Canada, and France. Among them, the United States held a dominant position from 1995 to 2001. The primary reason is that countries like the United States, Canada, and France dominate the global wheat market through their exports. These nations have established direct export relationships with numerous partners, granting them a strong capacity to export virtual cropland via trade. Additionally, the out-closeness centrality values for countries in the Black Sea region, such as Russia and Ukraine, have shown an upward trend, indicating their increasing influence in the outflow of virtual cropland through wheat trade. Meanwhile, the overall trend of in-closeness centrality over time is characterized by “a decline from 1995 to 2001, an increase from 2001 to 2014, and a decrease again from 2014 to 2023.” As illustrated in Figure 5, the in-closeness centrality values are generally low, with no single node holding a monopolistic position. This is likely due to the high concentration of global wheat exporters, which limits the options for major suppliers available to importing countries. It is noteworthy that if countries with high out-closeness centrality implement policy changes, such as export restrictions, the direct impact would affect a broader range of countries, potentially triggering cascading effects throughout the network.

(3)

Betweenness centrality.

Betweenness centrality reflects the bridging role of individuals within a social network and serves as an effective method for identifying telecoupling spillover systems [11]. As shown in

Table 6. Top 10 countries by betweenness centrality (1995–2023).

Year	1	2	3	4	5	6	7	8	9	10
1995	U.S. 7.930	France 4.559	Germany 4.271	Netherlands 2.409	U.K. 2.025	Italy 1.688	Spain 1.657	Belgium 1.083	Denmark 0.877	Canada 0.874
2001	U.S. 8.392	France 6.162	Germany 4.740	Canada 4.185	Argentina 3.940	Australia 3.565	U.K. 3.063	Russia 2.711	Japan 2.612	Turkey 2.604
2007	U.S. 11.072	France 7.573	Germany 4.396	Italy 4.235	Russia 4.119	Ukraine 3.120	U.K. 2.863	Canada 2.822	China 1.902	Australia 1.876
2014	U.S. 8.117	France 4.484	Germany 4.400	Canada 4.369	U.K. 3.244	Italy 2.925	India 2.301	Russia 2.087	South Africa 1.495	China 1.256
2020	U.S. 8.376	France 6.552	Germany 4.768	U.K. 3.896	Kenya 3.416	South Africa 3.201	Canada 3.016	Russia 2.950	Uganda 2.456	Italy 2.139
2023	France 5.511	Kenya 3.869	U.S. 3.811	France 3.484	U.K. 3.191	Canada 2.026	South Africa 1.914	Brazil 1.891	Australia 1.719	Tanzania 1.567

, from 1995 to 2020, the betweenness centrality of the United States far exceeded that of other countries, consistently ranking first, while France and Germany stably held the second and third positions, respectively. In 2023, France surpassed the United States to take the top spot. This pattern underscores the longstanding and significant intermediary roles played by the U.S., France, and Germany within the virtual cropland network embedded in the international wheat trade. This structure is closely tied to the underlying global trade architecture shaped by multinational agribusiness firms, notably ADM, Bunge, Cargill, and Louis Dreyfus (LDC) [49]. Leveraging worldwide subsidiary networks, dedicated logistical infrastructure, and long-term contractual arrangements, these corporations exercise intermediary influence and resource-allocating power that extends beyond their home countries’ borders [50].

More specifically, the persistently high betweenness centrality of the United States rests on a globally integrated production model orchestrated by multinational grain traders, such as ADM, Bunge, and Cargill. By controlling the entire supply chain—from production sites and storage through logistics to final markets—these entities have established the U.S. as a critical junction within multiple key trade corridors [51]. In contrast, France’s intermediary strength is grounded in the global trading networks of firms, such as Louis Dreyfus, which are enhanced by institutional synergies within the European Union. This has given rise to an institution-embedded intermediary model that amplifies the country’s network coordination capacity. Although Germany is not a major wheat producer, its highly efficient ports, including Hamburg and Bremen, have rendered it an essential physical transit and dispatch node within the global logistics network dominated by the ABCD companies (ADM, Bunge, Cargill, and Louis Dreyfus).

France’s ascent to the highest betweenness centrality in 2023 highlights the divergent resilience characteristics of distinct intermediary models in the face of global shocks. During systemic disruptions, such as the pandemic, the highly centralized, corporate-led direct-linkage model characterizing the U.S. approach revealed certain vulnerabilities. Conversely, the French model—embedded in the more institutionally coordinated and regionally resilient fabric of the European Union—demonstrated stronger buffering and adaptive capacities. This shift suggests that the robustness of the global wheat trade network may be transitioning from a primarily “efficiency-first” orientation to a paradigm that balances efficiency and resilience.

Notably, the intermediary function of the global virtual cropland network embedded in wheat trade is not monopolized by the developed nations. Taking Kenya as a case in point, its betweenness centrality exhibited a consistent upward trend during 2020–2023, indicating a strengthened role in transit and intermediation. During the 2020 pandemic, disruptions in shipping operations at several major global ports coincided with the United Nations World Food Programme (WFP) designating Kenya as a distribution hub for food aid in East Africa. As the region’s largest deep-water port, the Port of Mombasa’s hub role became more prominent, prompting some virtual cropland flow paths to shift from a “global direct supply” model to an “exporter–Kenya–importer” pattern. Additionally, following the outbreak of the Russia–Ukraine conflict in 2022, several East African countries redirected their wheat imports from Ukraine to alternative sources like Australia. This further solidified Kenya’s position as a crucial transit node for virtual cropland flowing from exporting countries into landlocked East African nations.

Overall, from 1995 to 2023, the United States, France, and Germany have sustained their dominance as pivotal intermediaries in the cropland network of international wheat trade, a position fundamentally supported by the deep-seated trade and logistical frameworks established by multinational agribusiness corporations in these nations. France’s recent ascent to the top position underscores the varying resilience of different trade governance models. Concurrently, the emergence of new intermediary hubs, exemplified by Kenya, signals a broader structural shift within the global virtual cropland network, gravitating toward a multipolar architecture.

3.3. Drivers of Evolution in the Virtual Cropland Trade Network

The QAP regression results (

Table 7. Regression results of the unweighted trade network.

	Variable	1995 Year	2001 Year	2007 Year	2014 Year	2020 Year	2023 Year
Demand factors	G	0.2664 *** (0.001)	0.2324 *** (0.001)	0.2132 *** (0.001)	0.1936 *** (0.001)	0.1726 *** (0.001)	0.1749 *** (0.001)
	N	0.1582 *** (0.001)	0.2663 *** (0.001)	0.2054 *** (0.001)	0.2471 *** (0.001)	0.2234 *** (0.001)	0.1760 *** (0.001)
	S	0.0901 *** (0.009)	0.0251 (0.199)	0.0947 *** (0.007)	0.0246 (0.208)	0.0591 ** (0.048)	0.0909 *** (0.007)
Supply factors	A	−0.0905 *** (0.001)	−0.0580 *** (0.005)	−0.0332 * (0.074)	−0.0845 *** (0.001)	−0.0873 *** (0.001)	−0.0644 ** (0.011)
	O	0.0724 ** (0.012)	0.0764 *** (0.002)	0.0491 ** (0.027)	0.0664 ** (0.022)	0.0164 (0.266)	0.0150 (0.3133)
	F	0.0746 ** (0.033)	0.0358 (0.147)	0.0224 (0.214)	0.0653 * (0.051)	0.0971 ** (0.012)	0.0777 ** (0.035)
Distance factors	D	−0.2105 *** (0.001)	−0.2096 *** (0.001)	−0.2876 *** (0.001)	−0.266 *** (0.001)	−0.2533 *** (0.001)	−0.2422 *** (0.001)
	C	0.1685 *** (0.001)	0.1891 *** (0.001)	0.1306 *** (0.001)	0.1560 *** (0.001)	0.1626 *** (0.001)	0.1408 *** (0.001)
Convenience factors	P	−0.0606 *** (0.005)	−0.0833 *** (0.001)	−0.0844 *** (0.001)	−0.0905 *** (0.001)	−0.0912 *** (0.001)	−0.0706 *** (0.002)
	W	0.1076 *** (0.002)	0.0425 * (0.077)	0.0109 (0.393)	0.0799 ** (0.013)	0.0914 *** (0.006)	0.1138 *** (0.001)
	R2	0.1859	0.2105	0.2158	0.2040	0.1948	0.1631
	Adj-R2	0.1853	0.2099	0.2152	0.2034	0.1942	0.1625

Note: Values reported are standardized regression coefficients, with significance levels in parentheses; *, **, *** indicate statistical significance at the 0.1, 0.05, and 0.01 levels, respectively; the constant term is 0 for all models.

and

Table 8. Regression results of the weighted trade network.

	Variable	1995 Year	2001 Year	2007 Year	2014 Year	2020 Year	2023 Year
Demand factors	G	0.1579 *** (0.002)	0.1294 *** (0.003)	0.1796 *** (0.001)	0.1354 *** (0.001)	0.0855 ** (0.012)	0.0849 ** (0.012)
	N	0.0644 ** (0.017)	0.1992 *** (0.001)	0.1406 *** (0.001)	0.1390 *** (0.001)	0.0762 *** (0.008)	0.0600 ** (0.018)
	S	0.0131 (0.205)	−0.0010 (0.520)	0.0158 (0.160)	−0.0013 (0.513)	0.0323 * (0.053)	0.0300 * (0.065)
Supply factors	A	−0.0219 (0.103)	−0.0052 (0.360)	0.0239 * (0.071)	0.0036 (0.396)	−0.0187 (0.144)	−0.0040 (0.413)
	O	−0.0290 ** (0.039)	−0.0185 (0.104)	0.0062 (0.353)	−0.0257* (0.098)	−0.0164 (0.169)	−0.0013 (0.474)
	F	0.0528 ** (0.025)	0.0234 * (0.083)	0.0362 ** (0.042)	0.0539 ** (0.022)	0.0575 ** (0.025)	0.0366 * (0.057)
Distance factors	D	−0.0317 ** (0.029)	−0.0331 ** (0.012)	−0.0753 *** (0.001)	−0.0688 *** (0.001)	−0.0459 *** (0.007)	−0.0256 * (0.073)
	C	0.0643 *** (0.001)	0.1084 *** (0.001)	0.0829 *** (0.001)	0.1067 *** (0.001)	0.0906 *** (0.001)	0.1211 *** (0.001)
Convenience factors	P	0.0047 (0.380)	−0.0044 (0.401)	−0.0405 *** (0.004)	−0.0116 (0.255)	−0.0316 ** (0.024)	−0.0010 (0.278)
	W	0.0193 (0.141)	−0.0000 (0.499)	−0.0326 ** (0.034)	0.0032 (0.460)	0.0200 (0.177)	0.0082 (0.367)
	R2	0.0373	0.0747	0.0751	0.0573	0.0282	0.0278
	Adj-R2	0.0366	0.0739	0.0744	0.0566	0.0274	0.0271

Note: Values reported are standardized regression coefficients, with significance levels in parentheses; *, **, *** indicate statistical significance at the 0.1, 0.05, and 0.01 levels, respectively; the constant term is 0 for all models.

) indicate that, at the global scale, demand and distance factors serve as the primary drivers influencing the evolution of the virtual cropland network in international wheat trade. Supply and facilitation factors also affect the trade network, although their impacts are relatively weak. A comparison between Table 7 and Table 8 reveals that the correlation coefficients and significance levels of all explanatory variables are stronger in the unweighted trade network than in the weighted network. A detailed analysis is presented in the following sections.

(1)

Demand factors constitute a key driver of virtual cropland flows.

The economic level variable shows a statistically significant positive correlation with the volume of virtual cropland flows, suggesting that countries with larger economic disparities are more likely to engage in wheat trade, thereby facilitating virtual cropland transfers. This finding is consistent with the previously identified structural characteristic, where developed economies like the United States, Canada, and Australia act as primary senders, while developing regions in Asia, Africa, and Latin America serve as main receivers. Notably, although the regression coefficients for the economic level variable remain relatively high in both unweighted and weighted network models, they exhibit a general downward trend over time, indicating a gradual weakening of its influence in driving virtual cropland resource flows.

The foreign wheat demand variable is significantly positive in both network types, indicating that countries with higher wheat self-sufficiency rates typically have greater exportable surpluses. Differences in national self-sufficiency capacities thus underpin differentiated demand in the international market. The greater the disparity in wheat self-sufficiency between two countries, the more likely they are to establish trade linkages, facilitating cross-regional coordination in allocating virtual cropland resources.

The influence of the consumption structure variable is more complex. In the unweighted network, although the coefficient is positive, its magnitude is small, indicating that when trade volume is not considered, countries with greater differences in consumption structure are more likely to establish wheat trade ties. In the weighted network, however, the direction of this variable’s coefficient is unstable and its statistical significance is weak. This suggests that when trade volume is considered, the effect of consumption structure on virtual cropland flows becomes less clear, and its mechanism may be influenced by additional contextual factors.

(2)

Supply factors exhibit differentiated effects on the virtual cropland trade network.

The variable for the wheat-harvested area exhibited an overall negative influence, which diverged from the theoretical expectations. In the unweighted network, its regression coefficient remains small and exhibits a fluctuating decline, indicating that when trade volume is excluded from consideration, countries with smaller differences in the proportion of their wheat-harvested areas are more inclined to establish virtual cropland trade linkages. However, in the weighted network, this variable was not found to be statistically significant. This indicates that once trade volume is incorporated, the influence of area-based differences is diluted by more dominant factors such as trade scale and market demand. The structural pattern of global wheat trade, characterized by highly concentrated exports in a few nations (e.g., the United States, Canada, and Australia) and widely dispersed importers, fundamentally limits the explanatory power of the harvested area variable. For example, in 2020, despite significant disparities in the wheat-harvested area between Russia and countries such as Turkey and Egypt, their virtual cropland trade volume remained substantial, driven by robust import demand and Russia’s strong export capacity. Conversely, trade flows between China and the United States and Australia remained considerable, even with minimal differences in the harvested area, owing to stable trade relations and consistent market demand (Figure 6).

The wheat yield variable demonstrates a positive effect in the unweighted network, aligning with expectations: the greater the disparity in yield levels, the more pronounced the comparative advantage in wheat production between countries, thereby facilitating trade connections. However, its coefficient did not achieve statistical significance in 2020 and 2023, suggesting that comparative advantage is no longer the sole consideration for wheat trade amid external shocks such as the pandemic and increased geopolitical uncertainty. In the weighted network, the yield variable shows a negative and insignificant effect, contrary to expectations, further illustrating that virtual cropland flows are shaped by the complex interplay of multiple factors. Additionally, the renewable freshwater resource variable exhibits a significant positive effect in both network types, with stronger significance in the weighted network. This indicates that differences in freshwater resource endowments between countries not only promote the formation of trade linkages but also help expand actual trade volume, thereby reinforcing connections within the virtual cropland trade network.

(3)

Distance factors exhibit dual nature within the virtual cropland network;

The geographic distance between countries has a notably negative impact on virtual cropland flows. In the unweighted network, its regression coefficient consistently remains below −0.2, with its absolute value increasing over time. Although the absolute value of the coefficient decreases in the weighted network, it still remains significantly negative. In contrast, the contiguity variable exhibits a statistically significant positive coefficient at the 1% level in both network types, suggesting that geographical proximity actively fosters the formation of trade relationships. From 1995 to 2023, the average distance of virtual cropland flows in global wheat trade decreased from 4801 km to 4664 km (Figure 7). On one hand, trade volumes over medium distances (5000–10,000 km) increased substantially, with neighboring countries like Uzbekistan and Kazakhstan capitalizing on their proximity to forge closer trade ties. On the other hand, ultra-long-distance trade links exceeding 10,000 km—such as those between China and Canada, Canada and Japan, and China and the United States—consistently facilitated large-scale virtual cropland transfers. This illustrates that while distance serves as a fundamental barrier to trade connections, it can be surmounted by high complementarity between countries when driven by large-scale supply and demand dynamics.

Virtual cropland flows can originate not only from nearby receiving systems but also directly “jump” to distant ones. When demand in a receiving system is sufficiently concentrated and substantial, these virtual cropland flows can overcome the friction of distance, linking remote sending systems and creating leapfrogging telecoupling connections. This results in a complex spatial pattern marked by the coexistence of localized clusters and globalized connections.

(4)

Facilitating factors exhibit differentiated effects on virtual cropland flows.

The governance variable shows a significantly negative regression coefficient at the 1% level in the unweighted network, with the absolute value of the coefficient fluctuating upward over time. This indicates that smaller differences in governance quality between countries facilitate the establishment of wheat trade linkages and promote virtual cropland flows, and the marginal impact of this factor on trade formation continues to strengthen. This finding aligns with the conclusions of Cheng et al. [23]. In the weighted network, the significance of this variable weakens, suggesting that governance quality has a relatively limited influence on trade volume.

The WTO co-membership variable generally has a positive, albeit modest, impact on virtual cropland flows. As a multilateral trade coordination mechanism, the WTO fosters the establishment of wheat trade connections between countries by enhancing trade facilitation. However, with the rise of regional trade agreements and the restructuring of global value chains, its economic facilitation influence in international wheat trade may increasingly be diverted by regional governance mechanisms, leading to a weakening effect on the virtual cropland trade network.

The flow of virtual cropland embedded in international wheat trade essentially represents a cross-regional reconfiguration of global cropland resources, facilitated by the physical trade of wheat. It reflects increasingly complex and networked human–land interactions. Within this telecoupled system, spatial disparities in economic development levels and the cross-regional transmission of wheat consumption demand drive sending systems (exporting countries) to leverage their natural endowment advantages, transforming local cropland resources into tradable wheat production capacity and virtual cropland output. Concurrently, governance coordination, trade facilitation, logistical efficiency, and transnational agribusinesses provide the operational channels for virtual cropland to flow from sending systems to receiving systems (importing countries). This flow, shaped by supply-demand matching, forms a bidirectional feedback loop. It transcends the rigid constraint of the immobile geographical location of cropland resources, promotes the reallocation of these resources within the global social-ecological system, and establishes a cross-territorial system for the synergistic utilization of cropland resources.

3.4. Telecoupling Implications of Virtual Cropland Trade Embodied in Global Wheat Trade

The virtual cropland difference is a crucial indicator for assessing the efficiency of cropland resource allocation in international trade. From 2019 to 2023, global wheat trade generally optimized cropland use, leading to a cumulative saving of 47.5812 million hectares of cropland. However, this optimization effect is waning: global cropland savings have declined each year since 2020, dropping from 12.9118 million hectares to 10.0939 million hectares by 2023. This downward trend is closely tied to a series of shocks affecting the global wheat trade network, including the COVID-19 pandemic, the Russia–Ukraine conflict, and wheat export bans by several countries. These disruptions have reduced the overall efficiency of virtual cropland trade allocation and have disproportionately impacted low- and middle-income countries [47].

Using Egypt as a case study, we found that its dietary structure heavily relies on imported wheat. Typically, virtual cropland inflows help mitigate the strict limitations of domestic resources. However, this reliance makes the country vulnerable to external risks. From 2020 to 2023, there was a significant restructuring of its import sources. The share of virtual cropland from Russia and Ukraine dropped dramatically from 86.19% to 36.04%, while the contribution from the United States and Canada increased from less than 1% to about 9%. This shift extended the geographical distance of trade links, resulting in higher transport-related carbon emissions and increased shipping costs, which in turn raised the landed price of wheat. Additionally, fluctuations in the global trade network imposed fiscal pressure and posed risks to social stability within Egypt. In summary, while global wheat trade redefines local resource boundaries and creates cropland savings, its network volatility also redistributes global risks, increasingly linking the food security of low- and middle-income countries to political and environmental changes in distant regions.

The rise in virtual cropland trade volume within the global wheat market highlights a deeper cross-regional allocation while also revealing its concealed eco-environmental costs (Figure 8). In Australia, a burgeoning wheat exporter, the area of permanent grassland and pasture shrank by nearly 1.22 million km² between 1995 and 2020, with a significant portion converted into wheat-growing regions. Meanwhile, wheat expansion in South American countries like Brazil and Argentina has been accompanied by extensive deforestation. According to World Bank data, the cumulative forest loss in these two countries during the same period reached approximately 800,000 km², severely impairing the carbon sequestration and biodiversity functions of their ecosystems. In terms of chemical inputs, India and Brazil increased their fertilizer application rates per unit of cropland from 85 kg/ha and 98 kg/ha to 210 kg/ha and 353 kg/ha, respectively, illustrating a growth pattern driven by intensified chemical use. In contrast, while the volume of virtual cropland flows from traditional major exporters such as the United States and Canada has not expanded significantly, their cropland fertilizer application rates remained high, reaching 131 kg/ha in 2020, exerting cumulative pressure on local soil and water environments.

While international wheat trade has optimized the allocation of global cropland resources on a broad scale, it has increasingly strayed from its optimal state due to the impacts of pandemic shocks, regional conflicts, and rising international uncertainties. This has created a transmission chain of “shock impact → network evolution → ecological cost → economic effect,” highlighting the deeper challenges of global food security and environmental governance. Future research should further quantify the loss of cropland quality and ecological footprints associated with virtual cropland flows under different trade pathways, thereby providing a more precise scientific underpinning for the sustainable management of land systems.

4. Discussion and Conclusions

4.1. Discussion

The vulnerability of the virtual arable land trade network in the international wheat market has become increasingly evident. The trade pattern, characterized by exports dominated by Europe, the United States, Canada, and Australia, and imports dependent on Asia, Africa, and Latin America, has yet to undergo a fundamental transformation. Meanwhile, the seller’s market characteristics further intensify the risk of imbalance in the trade system [52]. In this context, a trade model focused solely on efficiency is no longer sufficient to address the myriad challenges, such as economic disruptions and hidden transfers of ecological costs. Therefore, strengthening the governance of the global wheat trade system is urgent. Both importing and exporting countries must embrace the vision of a community with a shared future and enhance coordinated governance of the receiving, sending, and spillover systems within the global wheat trade. This approach should ensure food security while balancing ecological sustainability and equity.

From the perspective of virtual cropland-receiving systems, importing countries should implement integrated strategies that combine risk management with environmental standards. This approach involves utilizing geographical proximity and existing cooperation mechanisms to prioritize the creation of regional wheat reserve alliances with neighboring nations. For example, China could leverage agricultural cooperation under the Belt and Road Initiative to develop emerging suppliers with seasonal complementarity, such as Kazakhstan, while partnering with multinational agribusinesses, such as COFCO, to establish sustainable production bases. Through these collaborations, green production standards and traceability systems can be disseminated along the supply chain, with tariff preferences offered for wheat that meets these standards, thereby encouraging ecological transformation at the source. Meanwhile, core importing countries such as China could incorporate virtual cropland imports into a dynamic balance framework for their cultivated land preservation systems. By establishing a synergistic mechanism that links virtual cropland inflows with domestic cropland conservation and integrating this approach with high-standard farmland construction to enhance the quality of local cropland, it is possible to mitigate the risks of domestic cropland abandonment and quality degradation associated with over-reliance on imports. Differentiated technological empowerment is essential to mitigate the risks stemming from economic disparities. Less-developed importing countries should focus on enhancing innovation to increase yields and reduce post-harvest losses, thereby replacing high-ecological-cost imports with environmentally friendly domestic production. More developed importers can assist by investing in green technology research and development in exporting developing countries. Additionally, developing importers should expedite alignment with international standards to bridge the governance gap with developed exporters and foster standardized and unified inspection protocols at trade ports.

At the level of virtual cropland sending systems, major exporting nations must take greater responsibility for minimizing the environmental impacts of production. Countries that rely heavily on fertilizers, for instance, should expedite the development and implementation of green production technologies to reduce their environmental costs. Sustainable practices, such as conservation tillage techniques from countries like Canada, can be adapted and transferred to emerging suppliers in regions like South America and Africa, including Argentina, to help build resilient and sustainable production systems that strengthen the global wheat supply network. Drawing on the green direct payment scheme within the EU’s Common Agricultural Policy (CAP), major wheat-exporting countries could provide subsidies to wheat farmers who adopt conservation cultivation and crop rotation practices.

Overall, creating a trade system that balances efficiency, security, and sustainability requires a cooperative mechanism focused on the synergy of receiving, sending, and spillover systems. Building upon the Land Degradation Neutrality (LDN) target under the United Nations Convention to Combat Desertification (UNCCD), the ecological costs associated with virtual cropland flows can be incorporated into environmental impact assessments for international trade. Furthermore, drawing on the rationale of mechanisms like the Carbon Border Adjustment Mechanism (CBAM), an ecological compensation tariff could be imposed on wheat exports from countries with high fertilizer input or significant forest-to-cropland conversion. The revenue generated should be earmarked specifically for cropland quality restoration in less-developed regions. Importing and exporting countries should establish regular platforms for policy and technical exchange to reduce trade barriers, especially between nations with similar dietary patterns but different governance systems, such as India and Argentina. At the same time, multinational agribusinesses, as key intermediaries influencing trade flows, should be encouraged through incentives or regulations to actively fulfill inclusive governance responsibilities and maintain a fair and open trading environment. Ultimately, the governance of the trade network must be closely aligned with the United Nations Sustainable Development Goals, particularly SDG 2 (Zero Hunger) and SDG 12 (Responsible Consumption and Production), to achieve a synergistic win-win outcome for global food security and ecological sustainability.

While this study broadens the research perspective on resource flows in international agricultural trade through the lens of telecoupling, it has certain limitations. First, at the methodological level, the quantification of virtual croplands relies solely on the single indicator of “area,” without incorporating dimensions such as carbon storage, biogeochemical characteristics, and human well-being disparities across different croplands. This makes it challenging to accurately reflect the environmental costs associated with different trade routes. In the QAP regression, variables such as “wheat production area”—measured as the proportion of harvested area—and “WTO proximity”—which does not differentiate governance capacities between new and existing members—may warrant reconsideration in their construction. Furthermore, the model includes relatively few environmental variables. Additionally, telecoupling in international agricultural trade significantly influences both intracoupling and pericoupling dynamics [10]. This study does not address the interactive mechanisms between telecoupling, pericoupling, and intracoupling, such as examining the mutual influence between intra-European wheat trade and transatlantic trade between the United States and Europe.

Future research could advance in two main directions. On the one hand, a comprehensive virtual cropland assessment framework integrating both area and environmental quality could be developed. Introducing environmental correction factors would enable a more precise quantification of environmental costs and effect disparities across trade pathways. On the other hand, systematic analysis is needed to elucidate the processes, mechanisms, and effects of telecoupling, pericoupling, and intracoupling within international wheat trade systems. The reshaping of these coupled relationships by accelerating regional integration and profound geopolitical shifts also calls for systematic exploration in future research.

4.2. Conclusions

Based on the telecoupling framework, this study integrates social network analysis and QAP regression to trace the evolution and identify the drivers of the global virtual cropland trade network from 1995 to 2023. The main conclusions are as follows.

(1)

A distinct telecoupled structure characterizes this network. Evolving under tightening connectivity and external shocks reveals the dual nature of vulnerability and adaptive restructuring. The sending system has transitioned from a US–Canada duopoly to a multipolar regime led by Australia, Canada, Kazakhstan, and the United States. Receiving systems are concentrated in developing nations across Asia, Africa, and Latin America, with China being the predominant inflow hub. Within spillover systems, the institutionally embedded model exemplified by France has surpassed the U.S.-led global direct-linkage model in terms of coordinative capacity.

(2)

The drivers of network evolution vary significantly. Demand and distance emerged as core forces, whereas the influence of supply was unstable, and the facilitating factors showed a limited overall effect. Economic development and foreign demand significantly strengthen trade ties. Distance acts as a fundamental barrier but can be offset by high bilateral complementarity under certain conditions.

(3)

External shocks recalibrate the logic of the network between efficiency and security. The COVID-19 pandemic, Russia–Ukraine conflict, and subsequent national policy shifts have collectively steered the network from an efficiency-oriented to a security-prioritized architecture. For instance, Egypt’s shift from Russian and Ukrainian to U.S. and Canadian suppliers—trading efficiency for stability—has reduced the system-wide efficiency of resource allocation.

(4)

A new food trade architecture that balances efficiency and security is urgently required. This study concludes that isolated national or sectoral strategies cannot achieve the sustainable optimization of global cropland resources. Future governance must foster a multilevel, regionally coordinated, resilient system. Institutional innovation should balance network resilience with developmental efficiency to prevent fragmentation and systemic efficiency losses.