More than half of the world’s population lives in urban areas, a proportion that is set to increase¹ to 68% by 2050. These urban residents depend on supply chains to produce, procure, prepare and deliver food, and they are exposed to potential supply-chain disruptions and food shortages from changes in human activity and natural processes. There is growing recognition that food-system resilience needs to be improved, but how best to buffer against urban food shortages remains an open question for both research and policy. Writing in Nature, Gomez et al.² assess how the flow of agricultural products to a city depends on the diversity of the city’s trading partners. The authors apply ideas from engineering — such as those used when ensuring infrastructure is protected from flooding — to inform the design of food systems that can buffer cities against food shortfalls.

For decades, scientists and industry have been warning governments and consumers about the risks of food shortages. Such shortages have a range of possible causes, including droughts and heatwaves, pest and disease outbreaks, financial downturns and trade policies³. More recently, the COVID-19 pandemic has led academics, and society more generally, to revisit the question of how fragile urban food supplies really are (Fig. 1).

There have been many proposed solutions to deal with the dangers of food shortages, from climate-resilient agricultural management practices to promoting local food systems and self-sufficiency⁴. One solution that is gaining attention is to increase the number and variety of agricultural products, farms and companies procuring and delivering food. Diverse food supply chains might buffer cities against food shortages — in the same way that, in finance, a varied portfolio of stocks limits investment risk and, in ecology, a diverse mixture of species maintains ecosystem functions.

Gomez et al. used data on the origin and destination of different agricultural commodities for 284 cities and 45 non-city geographical areas in the United States. They identified domestic food systems for each city — that is, all of the geographical areas that supply crops, meat, live animals or animal feed to that city. The authors then determined how many cities faced different thresholds of abrupt food-supply disruptions, known as food shocks, using the percentage difference between the minimum and mean of supply amounts for each food sector over four years for each city. More specifically, they counted the number of cities in which the minimum was more than a particular percentage (ranging from 3% to 15%) smaller than the mean in any one of those four years.

Next, Gomez and colleagues combined those data with simple indicators of geographical similarity — such as the physical distance and difference in climate between each city and the geographical areas in that city’s supply network. With this information in hand, the authors tested the idea that groups of cities with more-diverse supply chains are better able to buffer against food shocks than are groups whose supply chains are less diverse. Indeed, they found that cities importing food from suppliers that are more dissimilar from themselves are less likely to face shocks than are cities whose supply-chain partners are less diverse. Such supply-chain benefits would not be reaped from having solely local food systems.

Gomez et al. then considered design concepts from engineering, where infrastructure systems should be planned to withstand shocks — such as extreme flooding — of a given frequency and magnitude. The authors undertook some bold extrapolations, in which they estimated the size of food shocks that would be faced by different US cities given their current supply-chain diversity. They found that a rare shock, such as one occurring once in 100 years, would cause a food-supply loss of about 22–32% across different cities.

The other implicit finding from Gomez and colleagues’ model is that even moderate supply-chain diversity is effective at reducing the probability of extremely large shocks. The authors also applied their analysis to shocks happening in multiple food sectors simultaneously. They obtained similar results to those for single-sector shocks — with supply-chain diversity also providing a buffering effect for these even rarer occurrences.

Gomez and colleagues’ work has major implications for the way in which resilient food systems should be built, but it also has a few caveats. First, the authors used only four years of data for each city, posing problems for characterizing the distribution of shocks at each city. This limited time series makes it difficult to define the baseline variation in food supply — that is, what is considered normal — for consumers and retailers alike. It also makes it hard to see to what extent diversified supply chains buffer against food shortages under normal conditions compared with years marked by extreme events, and whether the net benefits are large enough to trigger a change in food-procurement policies.

Second, the food-flow data used by Gomez et al. do not represent actual flows for each year, but instead are simply annual production quantities proportionally distributed according to observed flows⁵ in 2012. Therefore, the authors’ analysis does not capture, or allow for, rerouting or other social responses at the onset of extreme events. Such social responses within and after shock years would result in changing food flows across the supply network.

Third, Gomez and colleagues did not validate the predictive worth of their model beyond the four years considered, or outside the United States. This lack of verification is perhaps most limiting for applying the findings in practice — partly because the stability of food supply is itself dynamic, and will change with increasing volumes and types of food consumed, as well as with production technology. Although the observed phenomenon and general patterns might hold in other years and geographical regions, no data or analyses exist to validate whether the authors’ design suggestions will protect against future shocks to the degree claimed.

Designing urban food systems to specification is not as easy as engineering a bridge or dam that won’t fail in 100 years. The major global concern with respect to urban food shortages and food security is for populations of middle- to low-income countries, particularly those that are dependent on imports⁶. Theoretically, supply-chain diversity will also have a buffering effect for these populations when the number of urban dwellers starts to drastically increase in the coming years, especially in Africa. However, such nations are probably not accurately described by the model presented. Moreover, they have different policy options and capacities for producing diverse supply chains compared with those possible in the United States. Nevertheless, Gomez and colleagues’ work provides a timely and refreshing reminder that building diverse supply chains offers a crucial mechanism for protecting urban dwellers from food shortages.