Global catastrophic risks and food security

Abrupt global catastrophic risks (GCRs) include nuclear war, extreme pandemics, technological catastrophes, massive volcanic eruptions, asteroid/comet impacts, or solar storms, among others [1]. Collectively these risks are not improbable [2,3] and some risks may be increasing (e.g., those associated with artificial intelligence, bioengineered pandemics and nuclear war). A GCR could cause immense disruption to global trade [4–6]. Such major trade disruption could lead to shortages of critical commodities, such as liquid fuels, necessary for industrial food production and distribution. Degradation of food production or distribution systems could lead to famine, whether impacting food production [7], or supply chains [8], and famine is a key historical driver of societal turmoil and civilization collapse [9,10].

Even if measures have been taken to secure liquid fuel supply, for example upscaling biofuel feedstock production and refining capacity [11], then shortages of industrial agricultural inputs might still threaten famine [7], and the two-thirds of countries that are net food importers [12], would be at risk, as would remote isolated nations with potentially fragile linkages to global shipping routes [4].

One possible way to mitigate the impact of a severe disruption to global trade, and thereby prevent famine and/or societal collapse, is to pivot food production to more fuel-efficient, low input methods, closer to where food is consumed. Urban agriculture (UA) is the closest and most fuel-efficient source of food for city dwellers and may be part of one such solution, with some amount of near-urban land needing to be farmed as well.

Urban agriculture

UA can be defined as, “an industry located within a town, a city or metropolis, which grows or raises, processes and distributes a diversity of food and non-food products (re-)using largely human and material resources, products and services found in and around that urban area, and in turn supplying human and material resources, products and services largely to that urban area” [13]. UA may include such entities as home gardens, community gardens, allotments, school gardens, balcony and rooftop gardens, or larger farms at urban/rural interface [14], as well as indoor farms, vertical farms, and building integrated agriculture [15]. These can be conceived as ‘micro-UA’ (e.g., home gardens), ‘meso-UA’ (e.g., community gardens), or near-urban farms on the urban perimeter [16]. In this paper we use the term ‘near-urban’ agriculture to include industrial agriculture that is close to the edge of the built urban environment (i.e., where transport fuel needs are very low).

UA offers many potential benefits, including the provision of nutritious food, alleviation of poverty/food insecurity, strengthened communities, inclusion of marginalized groups, food security in the face of climate change, reduced ‘food miles’ (i.e., greenhouse gas emissions due to transport of food) [13], heat island reduction, employment, biodiversity, urban water retention, ecosystem services, circularity (reuse of waste and regeneration of inputs), dietary variety and progress towards the Sustainable Development Goals [15]. UA also contributes to education, ‘urban green infrastructure’ and the functioning of natural cycles disrupted by urban sprawl [14].

On the other hand, barriers to UA include potential contamination of urban soils [13,17], shortages of water supply, sunlight shading due to the built urban environment, or the economics of food production versus higher rent functions [14].

Previous research on urban agriculture

Previous studies have shown that UA makes a significant contribution to global food supply [18], particularly vegetable supply [19,20], with production area of home gardens potentially threefold that of community gardens [21]. Table 1 describes a selection of studies that report the potential of UA in terms of human diet. Previous findings indicated that vegetable requirements of some urban populations could be met through UA, but UA falls short of meeting total food demands. Quantitative estimates of UA production are still relatively rare, and there is considerable potential to increase urban food production [20].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Examples of studies that estimate the potential contribution of urban agriculture.

https://doi.org/10.1371/journal.pone.0321203.t001

Most UA activity and research has focused on community gardens and green open spaces (rather than private gardens), or on residential lots alone, and also on issues of sustainability, poverty, and climate change [13], or regional disasters such as earthquakes [22]. Some have argued that UA focus should be on nutrients and nutrient security alone and that ‘only partial self-sufficiency is required and desired’ [23], rather than focus on potential abrupt global catastrophes, international supply chain risks and general resilience.

Some UA research exaggerates the impact of UA by applying the advantages of some growing practices to all potential UA [23]. Additionally, when residential lots have been considered, some research did not investigate residential properties as a stand-alone resource, and ‘yards’ were simplistically considered to be the total space of the residential block minus the building footprint [24], which likely includes many paved areas.

When considering resilience to GCRs, limitations of many previous studies include that they did not examine all three components of UA (micro-, meso-, and near-urban), or examined only existing UA land area rather than potential UA land area, or studied only actually grown foods rather than optimized combinations of foods, or did not report the offset in industrial agriculture (rural) that UA might entail. Much previous research on UA also focuses on large urban areas, rather than more representative median cities [25]. The median city population globally is 85,000 [26]. The estimated area required to satisfy global cereal (661.6 Mha) and vegetable (47.2 Mha) demand exceeds global urban land area (64.3 Mha) [25], but this does not consider the differing density and context of cities, such as small semi-rural cities.

Food system resilience

Overall, what is needed is food system resilience and UA may be a key component in the face of potential GCRs. Food system resilience can be defined as the “capacity over time of a food system and its units at multiple levels, to provide sufficient, appropriate, and accessible food, to all, in the face of various and even unforeseen circumstances” [15]. A primary consideration is whether there is sufficient land to produce food. In a GCR that has severely degraded trade, fuel supply and transportation, productive land must be near to consumers.

Given this background, we aimed to determine the potential contribution that a radical pivot to UA might have in building resilience to GCRs, by calculating the potential land area available for UA in a globally median-sized city for population (the New Zealand city of Palmerston North). Our approach is unique in considering all the following factors in combination: micro-, meso-, and near-urban UA, a median-sized city, optimized rather than typical crops, food energy/protein needs – not just vegetable intake, maximum catastrophe-context land use, analysis of shortfall to be made up by near-urban agriculture, consideration of altered climate in a nuclear winter scenario, and the volume of liquid fuel conserved by offsetting off-road (on-farm) diesel use. We used a simple method that can be replicated by other small cities. Additional considerations beyond land area and optimal crops, such as water supply [16], fertilizer and organics [24], agrichemicals, and expertise [27] are addressed in the Discussion.

Study aims

Overall, this study aimed to address the following interrelated research questions:

1. What are the optimal crops to grow in the UA and near-urban spaces to maximize the number of people fed?
2. How much land is available for UA in this median-sized case-study city?
3. How many people could this combination of UA space and crop selection feed?
4. How much additional near-urban industrial cropping is needed to make up the shortfall of micro- and meso-UA and feed the whole city?

Study site

For this initial case study, we chose a globally median sized city by population in a temperate climate. Our method and study are demonstrative, and the approach could be replicated for cities of different sizes in different locations. The study site was the city of Palmerston North, New Zealand. Palmerston North lies in a temperate location (Cfb/oceanic climate, mean yearly temperature ~ 12 C, with ~ 1000mm annual rainfall), has a metropolitan administrative area of 395 square km (official city limits), and a contiguous built urban environment of approximately 35 km². The city had an urban population of 91,800 in 2023 [31], mostly living in medium density suburban-style standalone housing.

To investigate the potential role of UA in mitigating the impact of a trade disabling global catastrophe, we sought to answer the following questions about our case-study site.

What are the optimal crops to grow in the UA and near-urban spaces?

Previous research has shown a large discrepancy in total land area required to satisfy population protein and food energy needs by type of crop grown. For example, industrial agriculture producing only milk would require 642 thousand hectares (ha) to feed the total New Zealand population, whereas wheat requires 117 thousand ha and potatoes just 84 thousand ha [11]. The same issue arises with UA. Therefore, in our analysis, unlike many previous UA studies, we optimized for potential protein and food energy yield.

For the near-urban land area, we used the same general optimization approach as in our previous published work on optimizing frost-resistant crops for industrial agriculture in post-catastrophe New Zealand [32]. In this previous work, we considered published crop yield (kg/tonne of harvested crop per land area) and actual dietary energy (kJ) and dietary protein consumption of average New Zealanders, based on published nutrition survey research (specifically 8686 kJ of dietary energy and 81 g of dietary protein per person per day for a year) [33]. The optimization involved linear programming conducted with Excel Solver using the “Solver LP” method. Linear programming was selected as an optimization method due to its computational efficiency and ability to effectively handle multiple constraints while maximizing a specific objective function, making it well-suited for agricultural resource allocation problems where we need to simultaneously optimize for both protein and energy requirements under land constraints. For the nuclear winter scenarios, we considered frost-resistant crops that are commercially grown in New Zealand [34]. A simplifying assumption was one crop cycle per year.

For the micro- and meso- levels of UA, we analyzed as above but used crop yield data for UA from a recent meta-analysis by Payen et al. [35]. We considered all crops where the mean yield was over 1 kg/m² (gross weight), where at least two study results were reported in the meta-analysis and used data for open-air soil-based agriculture (i.e., excluding: polytunnels, green houses, aquaculture, and indoor settings with hydroponics and artificial lighting). Herbs, garlic and cereal crops were not included for the micro-/meso- UA analyses. The rationale for this was that herbs and garlic are typically used as condiments and unlikely to be a palatable diet on their own; and because cereal crops typically have low yields (<1 kg/m²) in UA settings according to Payen et al. and are more complex to process than vegetables and fruits. We calculated arithmetic means, as per the Payen et al. meta-analysis. We present specific tabulated data on yields and nutrient levels for these crops in the Supporting Information. These protein and energy levels were obtained from the food composition database of the New Zealand Institute for Plant and Food Research [36]. We additionally present yields reported by experienced permaculture growers for comparison [27], but excluded this from our main analysis as ordinary growers may lack the relevant skill set and results were extracted from a single detailed report.

How much land is available for UA in this median-sized case-study city?

We defined the boundary of the city by constructing a Google Earth polygon around the built urban environment, where it typically abuts farmland (see Fig 1). Within this boundary we estimated land area potentially available for UA in the form of micro-UA (residential lots), and meso-UA (city green spaces). This was estimated through manual analysis of samples, using orthorectified Google Earth images.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Calculating land area for urban food production.

Publicly available aerial imagery (CC BY 4.0) from Palmerston North City Council (2024), similar though not identical to the views obtained for the study using Google Earth. (A) Detail of the built urban boundary showing clear demarcation between urban environment and surrounding near-urban land (cemetery to east of urban area was excluded, as although part of ‘built’ space it is unlikely to be used for agriculture; also excluded was new development at the eastern boundary, as houses were not yet built). (B) Detail of coding residential lots for potential urban agriculture (UA)-space. Note coding of front yard and back yard where there is clear open space (front yard sighted on Google Street View and polygon drawn around clear yard space). Note paved driveway excluded, as well as trees, hedges, and areas with uncertain UA potential (not shown, meso-UA area, where the same process used for yards was applied to city green spaces).

https://doi.org/10.1371/journal.pone.0321203.g001

Micro-UA.

We used a random geolocation generator to identify specific residential lots within the built urban boundary (http://www.geomidpoint.com/random/), piloting 20 lots to establish inter-rater reliability (both authors) and then completing a sample of the entire urban area (one rater). Residential site area was divided into roof area, paved area, non-paved area, and tree area [16]. We considered only non-paved area as suitable for UA, identifying these areas using Google Earth imagery (data attribution 20 February 2024), a local real estate database (www.homes.co.nz) providing aerial imagery and property photographs, supplemented with Google Street View to confirm non-paved green space. We defined unpaved areas with UA potential using the polygon tool in Google Earth to calculate land area. We chose Google Earth because the spatial resolution of the aerial images is high and analysis required minimal training [21]. This analysis could easily be replicated by city officials or community groups in diverse locations. We sampled iteratively until we had estimated the land area potentially suitable for UA in a number of residential lots equal to the desired sample size (see ‘Statistical Analysis’ below). We then extrapolated the mean potential UA area of each lot to the wider city by simple multiplication of the mean potential UA area and the number of residential lots in the city (see Table 2). The method was piloted and refined in another New Zealand town, where physical site visits allowed comparison between aerial imagery and on the ground reality. This revealed that adding the use of Google Street View and the photographs on the real estate website (www.homes.co.nz), greatly enhanced the accuracy of identifying potential UA space. We recorded whole property land area and the 2021 capital valuation of each residential lot (www.homes.co.nz). Fig 1 illustrates coding of an example residential lot.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Key variables considered in the analysis.

https://doi.org/10.1371/journal.pone.0321203.t002

How many people could this combination of UA space and crop selection feed?

We multiplied the total land area potentially available for UA by the protein and food energy (kJ) yield per area of the optimal crop(s) and determined what fraction of the total Palmerston North urban population’s kJ and protein needs could be met if optimal food crop(s) were grown on the total potential UA land area with yields equal to those in the meta-analysis. We also considered scenarios where the yield of UA was reduced due to the sunlight blocking effects of a nuclear winter. Assuming both 150 Tg (equivalent to 150 megatonnes) or 5 Tg of soot has reached the stratosphere [37], with this previous study by Xia et al. specifically considering impacts on crop yields in New Zealand (with yield coefficients of 0.921 and 0.391 respectively, as in Table 2). Additional factors could also impact food production such as water supply, and inputs of fertilizer/organic material, agrichemicals and expertise. We address these additional factors in the Discussion.

How much additional near-urban industrial cropping is needed to make up the shortfall of micro- and meso-UA?

Outside of the built urban area, near-urban cropping can take advantage of industrial methods to achieve high yields, while minimizing transport and fuel requirements to processing and consumption in the city. That said, different crops in different scenarios will require different quantities of liquid fuel input to provide the expected yield. Additionally, in a GCR with trade collapse, usual supplies of imported diesel may not be available and local production of biofuel may be required [11]. Previous research has estimated the land area and diesel fuel requirements to feed the entire New Zealand population assuming any one of several different crops and documented industrial yield [11]. We used the results of these calculations for supplying minimum required food energy and protein, along with the diesel requirements per ha to calculate how much near-urban land would be needed to overcome shortfall in food energy and protein from micro- and meso-UA. We also calculated the land area needed for a biofuel feedstock such as canola (rapeseed) in case trade in liquid fuel is halted and local liquid fuel production is needed for agricultural machinery. We assumed zero transport fuel requirements, given that near-urban agriculture could be within walking distance of the city.

Statistical considerations

With 26,304 standalone residential lots in Palmerston North, a sample of 96 lots was needed to achieve 95% confidence of a 10% margin of error. To ensure reliability, two researchers (both authors) iteratively discussed and trialed the method, and then independently calculated the potential UA area in a pilot of 20 randomly selected residential lots. Interrater agreement was high, with Pearson’s r of 0.99, ICC 0.979, and a 2.4% difference in means (rater 1 mean 157.05 m², rater 2 mean 160.85 m²). Further analysis by a single rater proceeded on this basis. A similar process was followed for meso-UA with two raters independently estimating the potential UA area in a randomly selected km² of the built urban area. There was a 6.0% difference in rater totals, and this was deemed acceptable. On the basis of the number of individual spaces coded in one km² (mean 22) it was estimated there would be 720 potential UA spaces across the city, with sample size of 85 needed to achieve 95% confidence with a 10% margin of error. One rater proceeded to estimate the potential UA area in random km², as above, until this number was reached. All calculations were performed using Microsoft Excel for Mac v16.9.

Research Q1: Optimized cropping

For micro- and meso-UA areas under normal climate conditions, the optimized crop based on UA yields from the meta-analysis used was peas (at 292 m² of land to satisfy one person’s dietary energy and protein needs per year) (Fig 2). If peas were excluded from the analysis, the next optimized crops were a mix of sugar beet and spinach. In the nuclear winter scenario, the optimized micro- and meso- area frost-resistant crops were a mix of sugar beet (326 m²) plus spinach (18 m²). If those two crops were excluded from the analysis, the next optimized frost-resistant crop was turnips (417 m²) plus chicory (316 m²).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Crop optimization results: Crops requiring least land area to feed one person - different agriculture types and post-catastrophe climate scenarios.

The figure shows the minimum land area required to feed one person in both the normal climate scenario (top panel) and nuclear winter scenario (bottom panel – only frost-resistant crops considered), whether deploying urban agriculture methods (UA) or near-urban industrial agricultural methods. Only the crops/crop combinations found to require the least and next least land area are displayed. Shorter bars in the graph indicate less land is required to provide protein and food energy needs (i.e., most optimal crops). Yields in the nuclear winter scenario are typical yields of frost-resistant crops and not adjusted here for reduced sunlight (see Table 3 for such an adjustment). GCR – global catastrophic risk; UA – urban agriculture.

https://doi.org/10.1371/journal.pone.0321203.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3.

Potential for UA to meet combined protein and food energy needs of the palmerston North population.

https://doi.org/10.1371/journal.pone.0321203.t003

For near-urban areas under normal climate conditions, the optimized crop based on commercial agriculture yields was potatoes (at 156 m², Fig 2). The next optimized crops were a mix of wheat plus carrots. In the nuclear winter scenario, the optimized near-urban frost-resistant crops were a mix of wheat (219 m²) plus carrots (7 m²) (as found for previously published work [32]). The next optimized frost-resistant crop was sugar beet (276 m²).

Research Q2: Potential micro- and meso-UA land area

The land area of the Palmerston North built urban environment was estimated to be 34.2 km² (34,181,997 m²). The 96 sampled residential lots had a mean capital valuation (2021) of NZ$782,263 (US$480,539; 12 June 2024) (sd NZ$277,440). The mean lot area was 739 m² (sd 299.2). The mean area per lot potentially available for UA was 149.8 m², (sd 115.7; 95% CI: 126.7–172.9) which was 20.3% of overall lot area. With 26,304 ‘separate house’ dwellings in Palmerston North, the result was 3.9 km² (3,940,339 m²) of potential micro-UA space in this city.

Coding of open spaces resulted in an estimate of 9.83% of total urban area having meso-UA potential (5,111,000 m² sampled from 34,181,997 m² built urban area), resulting in an estimated 3.4 km² (3,360,090 m²) of potential meso-UA space. The resulting sum of micro- and meso- potential UA space for the whole city was 7.3 km² (7,300,429 m²).

Research Q3: UA food production potential

Results indicate that UA using 75% of the available urban land could supply up to 20.4% of population food needs (Table 3). Proportionately less of the population could be fed by UA in lesser land use scenarios (e.g., 50%, 25% of the micro- and meso-UA areas), or lower yield scenarios (nuclear winter, necessitating frost-resistant crops),

Research Q4: Near-urban land requirement

This scenario where UA supplies food to feed 20.4% of the population, would necessitate at least 1140 ha of near-urban cultivation (industrial cropping adjacent to the city) to make up the shortfall (Table 4), equivalent to 33% the size of the built urban area.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Additional near-urban land area (ha) required for industrial cropping to feed remainder of population (i.e., in addition to UA detailed in Table 3).

https://doi.org/10.1371/journal.pone.0321203.t004

This near-urban industrial cropping would require at least 134,468 L of liquid fuel for farm machinery per annum. If external supply of liquid fuel was not possible, then liquid fuel could be supplied by cultivating 110 ha of canola seed as a feedstock to produce biodiesel (Table 5). Proportionately more near-urban land and biofuel feedstock would be required in the scenarios where proportionately less UA land area was cultivated.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Diesel fuel (L) needed for near-urban industrial agricultural machinery (ha of canola to supply it as biodiesel)^*.

https://doi.org/10.1371/journal.pone.0321203.t005

Fig 3 illustrates schematically the minimum land area requirements for feeding the city. The total additional near-urban area required (combining both food crop and biofuel feedstock area) is equivalent to 39.5% of the total land area of the built urban environment.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Land required for food security for a median-sized city.

The figure visually depicts the food security land area footprint of the city. Circles illustrate relative scale of land available for micro-, and meso-urban agriculture (UA) and the population this land area could feed. Also represented is the additional land area of near-urban industrial cropping (including the required biofuel feedstock production), that could ensure food security (dietary energy and protein) for the population of a median-sized city (in this case Palmerston North, New Zealand).

https://doi.org/10.1371/journal.pone.0321203.g003

Exploratory results

If looking at data from experienced permaculture growers [27], rather than the meta-analysis of Payen et al., then Jerusalem artichokes (sunchokes) and peas are high yield in a no fertiliser/no pesticide scenario, especially if growing artichoke as perennial and harvesting tubers, requiring just 113 m² of area to fed a person per year (compared to the above 292 m² for UA peas and 156 m² for near-urban industrial potatoes). We provide direct comparison between the meta-analysis and results from these permaculture growers for the other less efficient crops in Supporting Information Table S3 in S1 Text. However, most city residents will not be experienced permaculture growers, so we consider this finding based on a single detailed report to be a sensitivity analysis, not the main result.

Urban and near-urban agriculture potential

With the potential to feed up to 20.4% of the population, we found there is insufficient micro- and meso- UA potential within the bounds of a suburban-style globally median-sized city to feed the city after an abrupt global catastrophe, even when public spaces and residential yards are optimally prepared and planted with the most productive UA crops. A modest amount of near-urban land is additionally required for industrial cropping to produce food along with biofuel to run the machinery for industrial agriculture.

Situating our results in the literature, we note that most previous studies we identified (some of which are listed in Table 1), examined only fruit and vegetable intake needs, and not food energy or protein. McDougall et al.’s 2020 study of urban agriculture potential in Sydney, Australia, was an exception. These authors reported that UA could supply calories feeding 33.8% of the population (compared to our finding of 20.4%), however, as we have noted McDougall et al. probably overestimate supply due to their method of calculating potential area for UA (which includes paved yard area by default). Munya’s study of Auckland, New Zealand, also considered overall diet, finding that 25% of yard area in low-density suburbs could contribute 9.4–12.3% of diet. This implies that 75% of yard use could supply 28–37%, which is higher than our finding, but in a lower density residential area, emphasizing the value of each city conducting a local analysis, potentially using the simple method we have described.

If the most efficient crops are chosen (peas in the UA space and potatoes on near-urban land) then an additional area of 1250 ha near-urban land must be cropped in a normal climate scenario, even if trade in global fuels ceases (1250 ha is the sum of food crop area in Table 4 and biofuel feedstock area in Table 5). This production would ensure sufficient food energy and protein for the city population as well as biofuel to run the agricultural machinery. It is important to note that less than 9% of this land (110 ha) is required for the biofuel feedstock production. When considering only a minimal amount of biofuel to run agricultural machinery, the argument that biofuel feedstock would displace food for humans does not apply. In fact, sowing this small percentage of land with a crop like canola maximizes food production (i.e., by allowing industrial machinery to operate) in no-fuel-trade scenarios. In normal times the canola can be used to produce food oil. We have further addressed the criticisms of biofuel production and various trade-offs in previous work [11].

The total area of 1250 ha is equal to 39.5% of the built urban environment. More land would be required for less efficient crops, or in the nuclear winter scenarios where frost resistant crops such as wheat, carrots, sugar beet and spinach would be optimal. This near-urban cropping could be configured to take advantage of the most fertile soils, transport routes such as electric railways or rivers, or other features of local geography. Larger denser cities which have high-rise apartments would require even more near-urban land. All this cultivation could be encouraged through judicious land zoning policy in advance of a catastrophe.

To meet the liquid fuel needs, in the base scenario, refining capacity for at least 134,000 L of biodiesel per annum would be required. This volume is equal to only 0.05% of New Zealand’s business as usual national consumption of 295 million L of diesel for off-road agricultural purposes and is less than 1% of the annual output capacity of one former New Zealand canola seed biodiesel refinery [11]. In the 5 Tg nuclear winter scenario 81–92,000 L is needed. The requirement for biodiesel is less in the winter scenario because the frost resistant crops wheat and carrot require less fuel to produce than potatoes, but require more growing land. Some cities will require more fuel for transport if they are not surrounded by arable land, meaning that ‘near-urban’ farming must be some distance away. Electric drying and milling capacity for wheat, if grown, would be needed and the wheat would need to be stored appropriately to ensure year-round supply. Anticipation of these needs is required, given the high versatility and storage potential of wheat.

Although other studies of UA have noted high potential for food production [35], in our analysis of optimized crops the near-urban yields were higher than micro-/meso-UA yields. This finding may reflect the temperate climate in this case study country setting and the high efficiency of its industrial cropping sector relative to UA. Furthermore, near-urban cropping involves vastly fewer hours of human labor to feed the population than scaling up UA.

That said, prudent GCR planning should assume some risk of break-down in industrial cropping and should anticipate the need for scaling up micro-/meso-UA. Key benefits of including micro-/meso-UA in the mix include reduced transport needs and conservation of liquid fuel, the ability to grow crops with high yields by mass and with high water content, where freshness is needed, requiring minimal capital-intensive processing, as well as diversifying the range of foods available, to meet nutrient needs, and psychological reassurance for urbanites concerned with major levels of social collapse. Therefore, UA cultivation is useful in addition to producing optimized crops in the near-urban space for survival.

UA does not feature in many cities’ urban plans, yet an integrated food and energy strategy, based on simple calculations of minimum requirements, could help ensure food supply in an extended global catastrophe. The focus of any strategy should be on facilitating and optimizing industrial cultivation of near-urban agricultural land, supplemented by the scale-up of UA (including assistive tools such as electric rototillers) and local biofuel and grain processing facilities. We acknowledge the competing demands on near-urban land, including development. This is more reason to carefully plan activities at the city fringe, including a GCR perspective, rather than allowing mere urban sprawl.

Our method is straightforward and could be undertaken for other communities, having contextual understanding of what is possible in their areas. There could be projects to prepare land to be fertile for cropping, or for citizens to learn how to do so if ever required.

Previous studies have quantified UA potential of residential lots. However, many of these overestimate residential potential and few previous UA studies attempted to quantify self-sufficiency by optimizing crop selection or considering the most efficient way to overcome food energy and protein shortfall (see example studies in Table 1). We have considered both factors.

Our results are based on open-air and ground grown vegetables. We excluded hydroponic/LED/artificial light/indoor/greenhouse/polytunnel production data, thereby avoiding the need for large quantities of materials, which may not be available in a post-catastrophe scenario. Wheat has the advantages that less volume per day needs to be eaten (compared to large volumes of potatoes to meet protein needs), it is frost-resistant, has long shelf-life, it is easier to store the seed than for potatoes, it provides a multipurpose food when milled, and surpluses can be used to feed to poultry for egg production.

Potential challenges

Policy recommendations

Although we did not set out to develop a set of policy recommendations, some broad policy approaches are immediately apparent. Food production is often divorced from urban planning [41], and there is often discrepancy between urban land with UA potential and land zoned for UA. For UA to provide GCR resilience it must be embraced pre-catastrophe at the policy level.

• Central and local government organizations responsible for long-term resilience planning should include food resilience workstreams, particularly with a view to food system and trade disrupting global catastrophes.
• Municipal authorities should consider zoning plans that encourage cultivation of near-urban land, especially existing cropping land, and manage urban sprawl. Cities are often built where they are because the nearby land is highly productive.
• Near-urban production zones might most optimally be configured along transport routes such as railway lines or navigable waterways thereby linking the most productive near-urban land with processing facilities (e.g., for wheat or biodiesel feedstock).
• Authorities could prioritize robust infrastructure in the near-urban area (renewable electricity grid; water supply for irrigation; rail transport links). Development of specific amenities and utilities could consider UA, for example, potential UA water use needs to be supported by water planning and infrastructure [42].
• Additionally, UA involving industrial systems could be highly productive, e.g., potentially growing fully mature lettuces and chicories in a month, and all year round [35]. Such systems could be supported where it is cost-effective, thereby providing year-round supply.
• Governments could incentivize pilot programs to enhance UA (especially near-urban agriculture) in different sized towns and cities. This could include central government mandates for use of local near-urban food for government funded food provision in institutions (e.g., meals in hospitals, and school lunch programs).
• Current local government regulations could be amended to facilitate UA and to facilitate trade in local agricultural produce, e.g., local government support for farmers’ markets.
• Local governments could consider reducing rates/land taxes for land used for UA and near-urban cropping.

Future research directions

The results above imply the following further research directions:

• Central and local government agencies could replicate this type of study for different city types and geographies and expand its dimensions (e.g., cost-effectiveness analyses relating to different crops) to quantify population survival needs in a global catastrophe.
• Near-urban farming is much more plausible than micro-/meso-UA to supply food needs in a global catastrophe and a geographic analysis and map of needed and available land for each city would be useful.
• Cost-effectiveness analysis should include small local biodiesel refineries and wheat mills, with the analyses accounting for the expected harm of low probability, high impact global catastrophes.
• Future studies could use citizen science approaches to estimate UA yields under various conditions (inputs, expertise, growing methods; school gardens, prison gardens) to help inform the above planning. Additional studies of intensive permaculture setups tended by experienced growers would be useful.
• Research needed includes justifying the best practices at each scale of UA, judicious use of production factors, sources of water, fertilizer use etc. Digital technology could be investigated to help implement/coordinate UA and food distributions across many stakeholders.

Study limitations

In this study we provided a detailed analysis of UA potential and then estimated the additional near-urban land required to feed the population. We have not analyzed the geography of the near-urban land to determine the ‘where’ of near-urban cropping. That would be the next step. The UA and near-urban yields described would depend on sustaining a degree of societal order that prevented theft of crops and maintained in-country trade in gardening/cropping inputs. Another limitation was that the main meta-analysis of UA yields did not include any data from New Zealand, albeit most of the studies were in high-income countries and temperate zone settings. There was often quite large variation in crop yields for industrial agriculture [34], and also for UA, based on region/country, crop variety, and how early/late in the season that planting occurred [35].

Nevertheless, our method excluded the UA potential from polytunnels, green houses, gardening on rooftops and balconies, existing fruit trees, trees with edible leaves, mushrooms in sheds/basements, aquaculture, indoor food production systems (hydroponics/artificial lighting), and from small livestock in sheds fed on kitchen waste etc (poultry, rabbits, etc). Similarly, we also did not consider potential mechanisms for increasing available land for UA such as by removing paving stones or pavement (e.g., ripping up car parks and tennis courts), growing crops that spread over paved areas (e.g., pumpkins), and planting on street berms etc. Ornamental trees could be trimmed or cut to expand access to sunlight for urban crops.

We also assumed only one harvest per crop per year and yet for some crops in certain settings, two harvests might be possible in normal climate conditions, this could greatly increase the contribution of UA or decrease the near-urban land needed. Furthermore, we did not account for higher yields from ongoing partial harvests (e.g., the outer leaves of lettuce, kale, spinach and beet greens). For some root crops (e.g., sugar beet, beetroot) we only considered the nutrition from the roots and not the edibility of the leaves for either human or livestock feed. Similarly, for some crops we only considered the above ground harvest and not the less typically used roots which are also edible (e.g., for chicory). Given all these factors, our analyses probably provide a likely underestimate of the total theoretical potential UA production of a post-catastrophe city.

Conversely, our modeled yields could be poorer due to deficiencies of horticultural inputs, or poor initial soil quality. Other than for the nuclear winter scenario (with reduced sunlight and precipitation), we made the simplifying assumption of current day yields (i.e., which are typically supported using hybrid seed varieties, fertilizer and pesticide inputs). Furthermore, in catastrophe situations some potential UA and near-urban land might also be diverted for non-food uses such as: medicinal plants, recreational drugs including crops for beer making, and firewood crops etc.

From a nutritional perspective, our optimization was simplistically based on just dietary energy and protein (and not other macro- and micro-nutrients). Also, as per previous work [32], we reviewed but did not include food wastage calculations, given that these could potentially be markedly reduced in post-catastrophe situations with the threat of food shortages.