The main data we use in this study is the annual agri-food flow at the Freight Analysis Framework scale from 2017 to 2022. The Freight Analysis Framework (FAF) data creates a comprehensive dataset of freight movement among states and major metropolitan areas through a partnership between the US Bureau of Transportation Statistics and the Federal Highway Administration (FAF, 2020).

FAF provides commodity-specific information classified according to the Standard Classification of Transported Goods (SCTG) (FAF, 2020). In this analysis, we focus on SCTG groups 1–7, collectively defined as agri-food commodities (Table 1). To capture the broad understanding of US food supply chains, we consider three types of agri-food flows: (i) domestically produced and consumed, (ii) imported, and (iii) exported commodities. For domestic flows, the origins and destinations of the food flow are both within the United States. For exports, we focus on the domestic origins of shipments for international markets, while for imports, we focus on the domestic destinations receiving foreign-produced products. We examine both the temporal and spatial changes in individual commodity networks and in aggregated food shipments from 2018 to 2022. While FAF datasets are available for the period 2017–2022, we focus our primary analysis on the years 2018–2022, which encompass the major compound shocks of interest, including the US–China trade war, COVID-19 pandemic, and extreme weather events. We retain 2017 as a reference year for selected baseline comparisons where appropriate.

Food supply chain complexity and diversity play a critical role in determining a region's resilience to shocks (Gomez et al., 2021; Doǧan et al., 2023). We thus evaluate the resilience of each FAF region within the agri-food supply chain network by analyzing its network structure complexity. Specifically, we calculate the weighted node degree of each FAF zone, following the approach proposed by Opsahl et al. (2010). Weighted node degree balances the importance of both the number of connections (degree) and the total amount of flow (node strength) for each FAF zone.

In a directed network G = (V, E), let V be the set of nodes and E be the set of edges. Each node i∈V represents a FAF region and each edge w_ij connecting node i to node j also captures the magnitude of directed agri-food flow between two FAF regions i and j. The degree of node i, denoted as k_i, is defined as the number of nodes to which i is directly connected:

where N is the total number of nodes, and x_ij = 1 if node i is connected to node j, and 0 otherwise. In our study, nodes i and j are determined as connected when there is a non-zero agri-food flow between these two nodes.

To account for the magnitude of each food flow, we incorporate edge weights by defining the node strength s_i as:

where w_ij>0 is the weight of the edge between nodes i and j, representing the quantity of food flow. And N is the total number of nodes. Although node strength is often the preferred method in weighted networks due to its focus on flow volume (Barrat et al., 2004; Opsahl et al., 2008), it does not capture the number of links a node connects. Conversely, degree alone does not reflect the volume of flow passing through a node. Therefore, following Opsahl et al. (2010), we combine these two measures into a single metric:

where α is a tuning parameter that determines the relative importance of the degree and strength.

The value of α can range between 0 and 1, providing a continuum between pure degree (when α = 0) and pure node strength (when α = 1). A lower α places greater emphasis on the number of connections, while a higher α places more emphasis on the weight of those connections. In this study, we choose α = 0.5 with an equal balance between the degree and the strength. By doing so, regions that connect to many other regions with smaller flows and those that connect to fewer nodes with very large flows are treated more comparably. The weighted node degree CDwα offers a comprehensive understanding on each region's role within the domestic agri-food network. The change in this metric can signal shifts in a region's resilience under certain disruptions. Higher CDwα values suggest that the region maintains a diverse set of trading partners and routes or that it maintains a substantial volume of agri-food flows, showing its significance as a key supplier or recipient. By tracking how CDwα changes over time, particularly during and after pandemic disruptions, we can gain a clearer understanding of the role each FAF region plays in addressing supply chain challenges. All data processing, network construction, and metric calculations were conducted in Python 3.10.

National-level changes in food flows from 2017 to 2022 are illustrated in Figure 2, considering three food flow types: domestic, import, and export. As shown in Figure 2a, the total domestic food flow remained relatively stable throughout the period. In 2019 and 2022, however, the total domestic food flow mass reached its lowest levels at approximately 3.093 × 10¹² kg and 3.08 × 10¹² kg, respectively. The 2019 decline is likely attributable to extreme flooding across the Midwest. According to the USDA, as of November 1, 2019, a record 19.4 million acres of farmland remained unplanted due to flood conditions (U.S. Department of Agriculture, Farm Service Agency (FSA), 2019), with additional losses caused by harvest delays and freeze damage (Randall, 2019). This decline was followed by a strong rebound in 2020 and a notable peak in 2021, suggesting a two-year recovery trajectory. 2021 recorded the highest domestic mass flux at about 3.23 × 10¹² kg. This increase may reflect a combination of post-pandemic recovery dynamics, including the return of eating out expenditures, and adaptive logistics strategies within the food supply chain. During this period, distributors, logistics providers, and packaging facilities reconfigured supply channels, repackaged goods for retail markets, and addressed labor constraints, thereby enabling a notable resurgence in domestic food movement [U.S. Department of Agriculture, Economic Research Service (ERS), 2023]. In 2022, the total domestic mass flux decreased, largely due to widespread drought conditions. That year, drought affected about 63% of the contiguous United States and caused extensive agricultural stress and an estimated 22 billion dollars in losses (National Centers for Environmental Information (NCEI), 2023). The most severe impacts occurred in the Central and Western states, which include many of the nation's major agricultural production areas (National Oceanic and Atmospheric Administration, 2023). In contrast, the economic and logistical disruptions associated with COVID-19 had largely subsided as supply chains and markets adapted, indicating that drought effects were the primary driver of the 2021–2022 decline.

Export flow volumes followed a similar temporal pattern as Figure 2b suggested, with the lowest value occurring in 2019 at 2.17 × 10¹¹ kg. Unlike domestic food flow, where cereal grains (SCTG 02) dominated, agricultural products (SCTG 03) accounted for the largest share of exports. Notably, SCTG 3 exports rose in 2018 despite the situation of the US—China trade war. While it was expected that export volumes would decline due to tariffs, actual volumes increased as producers redirected supply to alternative international buyers. However, based on our analysis of FAF volumes and export price trends, the total trade value for SCTG 3 declined in 2018, indicating that US agricultural product, particularly soybeans, were sold at lower prices to alternative buyers.

A distinct shift emerged in 2019 for SCTG 03: domestic flows declined sharply shown in Figure 2a, while exports and imports remained relatively stable (Figures 2b, c). The observed divergence in 2019 likely reflects the combined effects of production losses caused by the Midwest floods and the gradual easing of trade tensions toward the end of that year. As shown in Figure 1, discussion on tariff relaxations and renewed agricultural purchases between the United States and China began in late 2019, preceding the signing of the Phase One agreement in January 2020. Reports from USDA and USITC also indicate that soybean inventories accumulated during the record 2018 harvest were drawn down to meet external demand as global prices adjusted and trade conditions stabilized (Cowley, 2020; U.S. Department of Agriculture, Foreign Agricultural Service (FAS), 2020; U.S. International Trade Commission, 2019). Although other factors such as price movements and logistics recovery may also have played a role, the temporal overlap between policy relaxation and export stability suggests that these trade policy adjustments helped prevent further export declines during this period.

Meanwhile, Figure 2c shows that import food flow volumes declined steadily from 2017 to 2019, reaching a minimum of 6.69 × 10¹⁰ kg, followed by gradual recovery in subsequent years. The US–China trade war led to reciprocal tariffs, reduced bilateral agricultural trade, and widespread market uncertainty (Melton and Cooke, 2018; U.S. International Trade Commission, 2019). In 2019, US agricultural imports from China fell by 28%, and the national agri-food trade balance shifted into deficit for the first time since 2007 (Melton and Cooke, 2018). From 2020 onward, import volumes steadily rebounded, aided by stronger consumer demand and improved global trade logistics. This steady growth also aligns with the implementation of the United States–Mexico–Canada Agreement (USMCA) in July 2020, which maintained tariff-free agricultural trade and simplified border procedures that likely supported the increase of imports in 2021–2022.

Figures 2, 3 present spatial and temporal variations in food flows between 2018 and 2022 for all SCTG combined. Figure 2 shows the maximum annual percentage changes in food flows relative to 2017, with panels depicting (Figure 2a) domestic inflows—food shipped into each FAF zone, Figure 2b domestic outflows—food shipped out from each FAF zone, Figure 2c export flows—food shipped from FAF zones to international markets, and Figure 2d import flows—food received from abroad by FAF zones. Figure 4 complements this by indicating the specific years in which these maximum changes occurred for Figure 4a domestic inflows, Figure 4b domestic outflows, Figure 4c exports, and Figure 4d imports. For detailed results by commodity group, see Supplementary material.

For domestic flows in 2018, most maximum percentage changes were positive, particularly in states such as South Dakota and Illinois. Illinois, as a major soybean producer in the US, notably increased its food flow volume in 2018. That year, Illinois soybean production was the highest of any state, reaching nearly 700 million bushels. Yields also rose significantly, increasing by over 10% from the previous year (Newton, 2019).

In 2019, Iowa experienced one of the largest decreases in domestic outflows. This could possibly due to the widespread spring flooding across the Midwest, driven by heavy rainfall and rapid snowmelt. Floodwaters delayed planting, damaged infrastructure, and disrupted farm-to-market transportation across Iowa and surrounding states. These events contributed to a national record of nearly 20 million unplanted acres [U.S. Department of Agriculture, Farm Service Agency (FSA), 2019; Johnson and Monke, 2019], with domestic food trade performance visibly impacted in affected regions (Figure 3b). Beyond farm-level losses, transportation are also disrupted due to the flood and thus, reducing food supply from Iowa and its neighboring states to other states.

In 2020, states such as Montana and Idaho experienced decreases in domestic food flows. In Montana, the cattle industry was significantly affected as meatpacking plants nationwide temporarily shut down due to COVID-19 outbreaks, reducing slaughter capacity and slowing downstream movement (Hettinger, 2020). Idaho's potato industry also faced major disruption following a sharp drop in demand from the food service sector. Since approximately half of US potato production serves restaurants, Idaho farmers were left with about 1.5 billion pounds of surplus potatoes during spring 2020. With reduced processing capacity, potato usage in the state dropped by 26.5% in April 2020 compared to the previous year (Karan et al., 2023; Hobbs, 2021b). Wyoming's sheep ranching sector also faced disruptions, with labor shortages and logistic delays hampering processing and delivery (Kayne Pyatt, 2020).

In 2021, domestic outflows declined sharply in parts of the Central and Northern Plains. Iowa, Nebraska, and South Dakota all showed significant reductions in outflow volumes, aligning with the 2021 drought. By late summer, over 75% of Iowa and neighboring states were under moderate to severe drought (National Centers for Environmental Information, 2021), which likely contributed to lower crop yields and reduced outbound food movement. These stressors were further compounded by soil moisture deficits and degraded pasture conditions, particularly in Nebraska and South Dakota (USDA National Agricultural Statistics Service, 2021). Figure 3b confirms the spatial correspondence between drought exposure and domestic outflow contraction. Meanwhile, the largest percentage increases in 2021 were observed in parts of the Great Lakes region, including Wisconsin, Michigan, Indiana, and Ohio, suggesting a potential recovery following the pandemic.

In 2022, the central United States, including Nebraska, Kansas, Oklahoma, Colorado, and Texas, saw significant declines in both domestic inflows and outflows. These changes align with widespread drought conditions across the region. In Nebraska, 97% of the land experienced drought at its peak, with 41% of pasture rated poor or very poor (Nebraska Farm Bureau, 2022). Colorado experienced similarly severe drought, with 92% of its land affected and over half of both topsoil and subsoil rated as moisture-deficient. As a result, winter wheat and pasture conditions worsened considerably (U.S. Department of Agriculture, National Agricultural Statistics Service, 2022).

Regarding exports, the Midwest did not show maximum percentage changes in 2018, likely due to shifts in trade partners caused by the US–China trade war. In 2019, Oregon, Utah, and Colorado experienced decreases in exports. In Oregon, international trade tensions, particularly from the US–China trade war, reduced demand for several export commodities. For example, Chinese tariffs significantly impacted Oregon's agricultural exports, such as cherries. In 2018, China increased tariffs on US cherries from 10% to 40%, severely affecting exports from the Northwest region (Wojtanik, 2019). Utah, known for producing alfalfa hay, beef, and dairy, also saw substantial declines. During the peak of the trade war, Utah's agricultural exports to China fell by around 15%. Specifically, alfalfa hay exports dropped by over 20%, and beef and dairy exports declined by about 15%–20% due to tariffs and shifts in China's purchasing to other countries such as Brazil and European nations (Farmonaut, 2020).

Agriculture is vital to Colorado's exports, including beef, dairy, wheat, corn, and processed foods. When the US imposed steel and aluminum tariffs, major trading partners like Canada, Mexico, China, and the EU responded with retaliatory tariffs on US agricultural products. Consequently, Colorado's food exports declined sharply. By mid-2019, Colorado's food exports were down 17% year-over-year, primarily due to uncertainties in trade policy. This uncertainty discouraged overseas buyers from committing to long term contracts. Key markets such as Canada, Mexico, China, and Japan which are all significant importers of Colorado products, were particularly affected (Leeds School of Business, University of Colorado Boulder, 2019). For instance, Mexico, Colorado's export market, placed retaliatory tariffs on US's pork, dairy, apples, and potatoes in mid-2018, directly impacting Colorado hog producers (Chappell, 2018). Canada similarly imposed tariffs of 10%–20% on many US food and beverage products. Colorado's exports of non-alcoholic beverages to Canada declined about 40%, dropping from 58 million to 36 million dollar (Colorado General Assembly, 2019).

In 2020, the Midwest experienced significant decreases in exports, primarily due to pandemic related disruptions. For example, Minnesota's exports fell to approximately 20 billion dollars, marking a 10% reduction from 2019. State officials attributed this decline directly to COVID-19, highlighting widespread delays and market uncertainties (Toth, 2021). Similarly, most states across the US experienced reduced exports in 2020, with Midwestern states notably affected due to the global economic slowdown (Numbers, 2021).

For imports, in 2022, the United States experienced a significant increase in agricultural imports, particularly across large areas in the central and southeastern regions, as well as parts of the Northeast and Southwest. US agricultural imports reached approximately 200 billion dollar, an increase of 14.7% compared to 2021. This growth was mainly driven by horticultural products such as fruits, vegetables, wines, and nuts. The rise in imports reflects higher consumer demand for fresh and varied foods available throughout the year, along with favorable exchange rates that made imported goods more competitive (Nigh, 2023; Kaufman, 2025). Specifically, imports of fresh vegetables and fruits saw notable growth, with vegetable imports alone increasing by 7%, reaching around 10.7 billion dollar (Specialty Crop Grower, 2023).

The southwestern United States, particularly states such as Texas, Arizona, New Mexico, and parts of southern California that border Mexico, is a key region for agricultural imports, especially from Mexico. Due to their geographic proximity, these states receive significant amounts of Mexican agricultural products through land transportation. In 2022, US imports of fresh and processed fruits, vegetables, and nuts from Mexico were estimated at approximately 18.7 billion dollars (Paul Schattenberg, 2023). Additionally, policies such as the North American Free Trade Agreement (NAFTA) and its successor, the United States-Mexico-Canada Agreement (USMCA), have facilitated tariff free agricultural trade between the US and Mexico, significantly boosting these imports into the southwestern states (Fresh Fruit Portal, 2025).

Other than higher consumer demand and favorable trading conditions and policies, a reduction in local agricultural supply due to extreme weather events further increases imports in 2022. For example, Florida's Valencia orange production fell to a historic low of 434,000 tons due to damage caused by Hurricane Ian. Because 96% of Florida's Valencia oranges are processed into juice, domestic orange juice production dropped to its lowest level in at least 50 years. Consequently, US orange juice imports increased by 30% in the 2022–23 season, reaching 574 million gallons (Citrus Industry, 2023).

To quantify the resilience of the current US food supply system under various shocks, we utilize weighted node degree to capture both reductions in connectivity and mass flux, as well as their rebound. By comparing temporal changes from 2018 to 2022 (Figures 5a–e), we trace how regional connectivity responded over time to different shocks. As shown in Figure 5, the x-axis represents changes in the weighted inflow node degree for each Freight Analysis Framework (FAF) region, while the y-axis represents changes in the weighted outflow node degree. Each quadrant corresponds to a distinct pattern: the first quadrant indicates increases in both inflows and outflows, the second quadrant shows increased outflows but decreased inflows, the third quadrant represents decreases in both, and the fourth quadrant shows increased inflows but decreased outflows. Figures 5a–e illustrate between-year changes from 2017 to 2022.

The COVID-19 pandemic impacted different regions and commodities in the US supply chain variably, in both the magnitude of disruption and the speed of recovery, as shown in Figures 5d, e. Between 2020 and 2021, some regions like Iowa, Rest of KS, Rest of IL, shifted from the third quadrant (reduced inflow and outflow) to the first or second quadrant, reflecting early stages of structural recovery in food logistics. This rebound suggests improvements in both movement volume and network connectedness. In contrast, several urban nodes such as Portland or Rest of WA remained near the origin or shifted minimally, indicating a slower recovery in more urbanized or import dependent systems. These differences could be linked to regional production structures, commodity specific characteristics, and disruption propagation pathways.

For example, Iowa, a major cereal grain producer (SCTG 02), was notably impacted due to reduced biofuel demand when travel dropped during lockdowns. However, demand rebounded relatively quickly once travel resumed (Hart et al., 2020), restoring flows to prepandemic levels in the next year. Similarly, Rest of CO and Rest of NE also moved upward and rightward, indicating restored export movement and increased inflows. These cases reflect how resilience varied not only by region but also by network function and commodity type.

Extreme weather events such as droughts and floods had strong but different impacts on agricultural production and transportation. These disruptions were often more severe but did not last as long as those caused by trade conflicts or the COVID-19 pandemic. For instance, the severe drought conditions across the United States in 2022 substantially disrupted both agricultural output and transportation logistics, notably pushing more nodes into the third quadrant in Figure 5e, signifying simultaneous declines in inflows and outflows. Similarly, the extensive flooding along critical waterways like the Mississippi River in 2019 severely disrupted grain shipping routes, affecting inflows and outflows at essential nodes, including Illinois and Ohio. Nevertheless, these disruptions showed quicker recovery trajectories, as evidenced by the notable rebound of Midwestern agricultural regions in 2020, even amidst the ongoing pandemic.

In contrast to rural agricultural regions, urban logistics hubs experienced slow recovery following disruptions. Cities such as New York, Boston, and Los Angeles faced prolonged impacts from pandemic-induced labor shortages and shifts in processed food demand. For example, meat, poultry, fish, and seafood commodities (SCTG 05) continued to experience substantial negative impacts in New York through 2022, as shown in Supplementary Figure 16. Similarly, Los Angeles saw persistent declines in milled grain products (SCTG 06) lasting until 2021, with recovery only becoming evident by 2022. Additionally, other prepared foodstuffs, fats, and oils (SCTG 07) exhibited persistent disruptions in major urban centers, in contrast to the relatively stable or positive trends in agricultural production states, like Iowa, Wisconsin, Texas, and Arkansas.

These observations reflect broader resilience patterns across the network. Agricultural production states such as Iowa or Wisconsin shows relatively rapid and stable recovery, likely supported by the natural resumption of seasonal harvest cycles and the gradual reduction of environmental shocks such as drought or flooding. However, urban logistics hubs and coastal port cities like Los Angeles or New York faced more prolonged and uncertain disruptions. In these areas, recovery was shaped not only by the demand but also by complex interactions among freight infrastructure constraints, institutional policy responses, and supply chain bottlenecks. Furthermore, shocks occurring in upstream supplier regions would also propagate through the logistics network, adding more uncertainty and risk to the downstream hubs. These differences suggest that regional resilience is shaped not only by local exposure, but also by the functional role of each node within the national food system, and by the primary mechanism, which could be environmental, infrastructural or institutional, that governs each region's vulnerability and recovery pattern.