Global food insecurity and famine from reduced crop, marine fishery and livestock production due to climate disruption from nuclear war soot injection

We use a state-of-the-art global climate model to calculate the climatic and biogeochemical changes caused by a range of stratospheric soot injections, each associated with a nuclear war scenario¹⁸ (Tables 1 and 2). Simulated changes in surface air temperature, precipitation and downward direct and diffuse solar radiation are used to force a state-of-the-art crop model to estimate how the productivity of the major crops (maize, rice, spring wheat and soybean) would be affected globally, and changes in oceanic net primary production and sea surface temperature are used to force a global marine fisheries model. We combine these results with assumptions about how other crop production, livestock production, fish production and food trade could change and calculate the amount of food that would be available for each country in the world after a nuclear war.

Table 2 Changes in food calorie availability (%) in Year 2 after a nuclear war for the nations with nuclear weapons and global average assuming no trade after simulated nuclear wars under the Livestock case, the Partial Livestock case and the No Livestock case with 50% livestock feed to human consumption

The simulated surface climate disruptions due to the nuclear war scenarios are summarized in Fig. 1. Averaged over the current crop regions, surface downwelling solar radiation reduces by 10 W m⁻² (5 Tg soot injection) to 130 W m⁻² (150 Tg soot injection). With less energy received, the maximum average 2 m air temperature reductions range from 1.5 °C (5 Tg soot injection) to 14.8 °C (150 Tg soot injection), peaking within 1–2 years after the war, with temperature reduction lasting for more than 10 years. The cooling also reduces precipitation over summer monsoon regions. Similar but smaller reductions of solar radiation and temperature are projected in marine regions (Fig. 1b,d), with resulting changes in lower trophic-level marine primary productivity. We applied local changes at every grid cell to the crop and fish models.

Climate model

All nuclear war scenarios^9,18 are simulated using the Community Earth System Model (CESM)³⁹. This model includes interactive atmosphere, land, ocean and sea ice. Both atmosphere and land have a horizontal resolution of 1.9° × 2.5°, and the ocean has a horizontal resolution of 1°. The atmospheric model is the Whole Atmosphere Community Climate Model version 4 (ref. ⁴⁰). The land model is the Community Land Model version 4 with the carbon–nitrogen cycle. CESM output at 1- and 3-hour resolution, including 2 m air temperature, precipitation, specific humidity and downward longwave radiation and solar radiation (separated into direct and diffuse radiation), is used to drive the offline crop model simulations. There are three ensemble members of the control simulation, which repeats the climate forcing of 2000 for 15 years, three ensemble members of the 5 Tg case and one simulation for each other nuclear war scenario.

In all the simulations, the soot is arbitrarily injected during the week starting on May 15 of Year 1. Our scenarios assume that all stored food is consumed in Year 1 and we present analysis of the remaining food in Year 2. If the war occurred at the end of a calendar year, there would still be food available in Year 2, so what we label Year 2 should be relabelled Year 3. However, since the severe climate and food impacts last for more than 5 years (Figs. 1 and 2), the same conclusions apply to a world after a nuclear war.

Direct climate model output use

Because climate models have biases, it is typical to bias correct model output before using it as input for crop models. There are various techniques that attempt to use past observational data to address changes in the mean and variance, but none are perfect and all are limited by assumptions that future relations between model output and crop model input can be based on the recent past. A common method¹⁴ is the delta method in which an observational reanalysis weather dataset is used and monthly means of temperature, precipitation and insolation are modified according to the climate model simulations. This comes with the advantage of realistic internal variability important for crop modelling^12,13,14 but does not adjust changes in variance, which might be an unrealistic assumption under higher emission scenarios, such as the 150 Tg case. Here, because we are using a crop model that has already been calibrated with the same climate model that we are using, we use raw climate model output (1.9° × 2.5°) to force the crop model, and this allows variance to change too.

Crop model

Crop simulation uses the Community Land Model version 5 crop (CLM5crop)^41,42,43 in the Community Earth System Model version 2 (CESM2). Dynamic vegetation is not turned on. CLM5crop has six active crops, maize, rice, soybeans, spring wheat, sugar cane and cotton, and also simulates natural vegetation, such as grasses. In this study, we used the output of the cereals (maize, rice, soybeans and spring wheat) and grasses. Although CLM5crop does not simulate winter wheat, we assume winter wheat production is changed by the same amount as spring wheat, which has been found in other studies¹⁴; however, this may underestimate the winter wheat response, because winter wheat would experience colder temperatures during its growing period that would be more likely to cross critical thresholds¹⁴. Surface ozone and downward ultraviolet radiation would also be impacted by nuclear war³⁶, but CLM5crop is not able to consider those impacts, which might exacerbate the losses. In addition, the crop model does not consider the availability of pollinators, killing frost and alternative seeds. The model simulates rainfed crops and irrigated crops separately, and all results presented here refer to the total production of rainfed and irrigated crops. Irrigated crops are simulated under the assumption that freshwater availability is not limiting⁴³. Although evaporation is reduced with cooling, it is possible that our result may underestimate the negative impact from precipitation reduction, especially for the large injection cases.

CLM5crop was evaluated⁴¹ using FAO observations (average of 1991–2010), and it does a reasonable job of reproducing observed spatial pattern of maize, rice, soybean and spring wheat yield. Also, time series of crop yields simulated by CLM5crop compare with FAO data from 2006 to 2018, and CLM5crop reasonably represents global total production and average yields of maize, rice, soybean and spring wheat⁴².

CLM5crop is spun up for 1,060 years by repeating the past 10 years of the CESM control to reach the equilibrium of four soil carbon pools. The crop simulations are at the same resolution as CESM simulations (1.9° × 2.5°). The crop planting date is determined by growing degree days, and the location of cropland is fixed for all crops.

Fishery model

Fish and fisheries responses are simulated with the BiOeconomic mArine Trophic Size-spectrum (BOATS) model^15,44,45. BOATS was used to calculate the size-structured biomass of commercially targeted fish based on gridded (1° horizontal resolution) inputs of sea surface temperature and oceanic net primary production from CESM. The model also interactively simulates fishing effort and fish catch through a bioeconomic component that depends on fish price, cost of fishing, catchability and fisheries regulation¹⁵. Details are found in ref. ¹⁵ and references therein.

Combining crop and marine fish data

Supplementary Table 1 shows the total calorie reductions for each of the nine nuclear states from just the simulated crops and marine fish. Data for countries can be found in Supplementary Table 2.

To calculate nation-level calories available from simulated crops and fish, we weight the production by the calorie content of each type of food. We use data from FAO^23,24,46,47. Nation-level calorie reduction (%) from total production of maize, rice, soybean, wheat and marine fish is thus calculated as:

$$w_\mathrmiy = \fracP_ic_iR_\mathrmiy\mathop \sum\nolimits_i = 1^5 P_ic_iR_\mathrmiy $$

(1)

and

$$R_y = \mathop \sum\limits_i = 1^5 R_iyw_iy ,$$

(2)

where index i is maize, rice, soybean, wheat or marine fish wild catch, w_iy is the calorie weight of each commodity per country each year, _Pi is the national production of item i in FAO-Food Balance Sheet (FBS)^23,24, c_i is calories per 100 g dry mass for each item²³, R_iy is national production reduction (%) of each item in year y after the nuclear wars and R_y is nation-averaged calorie reduction (%) of the five items in year y after the nuclear wars.

Effects on other food types

Other crops

National averaged calorie reduction (%) of the four simulated crops is applied to the total calories of all crops in 2010 to estimate simulated nuclear war impacts on this category.

Livestock and aquaculture

We assume these two types of food share a similar response to simulated nuclear war as they involve feeding animals in a relatively controlled environment. For global calculations for livestock, we assume that 46% are fed by pasture and 54% are fed by crops and processed products⁴⁸ and use national-level data²⁶ to calculate reduction of livestock feed from pasture and crop-based products. We assume that livestock production is linearly correlated with the feed. Annual leaf carbon of grasses (both C3 and C4) is used to estimate pasture changes, and reduction of the four simulated crops is used for crop feed changes. For aquaculture, the feed is only from crops and processed products, and the production is also correlated with the amount of feed fish receive. Direct climate change impacts on livestock and fish are not considered.

Inland fish capture is not considered in this study. Because inland fish contribute only 7% of total fish production⁴⁶, adding inland fisheries would not change the main conclusions of this study.

International trade

All food commodity trade calculations are based on the 2010 FAO Commodity Balance Sheet (FAO-CBS), FAO-FBS and processed data from previous studies^24,27,28,47. This dataset provides the production and usage of each food and non-food agricultural product for each country and imports and exports and thus allows the calculation on a national basis of food usage and calorie availability.

Domestic availability of a food in each country comes from domestic production and reserves, reduced by exports and increased by imports. We calculate no international trade by applying the ratio of domestic production and domestic supply to each food category and the food production in different usages:

$$C_\mathrmfood – notrade = C_\mathrmfood \times \fracP_\mathrmdpP_\mathrmds$$

(3)

where C_food is national-level calorie supply from different food types^26,46, C_food-notrade is national-level calorie supply from different food types with the assumption of no international trade, P_dp is national-level domestic production for each type of food in FAO-CBS and P_ds is national-level domestic supply for each type of food in FAO-CBS. Domestic supply is the available food on the market, including domestic production, export and import.

Food usage of maize, soybean, rice and wheat is calculated from FAO-CBS. In FAO-CBS, maize products are maize and byproduct maize germ oil, soybean products are soybean and byproducts soybean oil and soybean cake, rice products are rice and byproduct rice bran oil and wheat product is wheat. Products for food purposes are the sum of food supply in each category and the processing product minus the total byproducts (the difference includes processing for the purpose of alcohol or sugar).

Calorie calculations

For the Livestock case, national-level available calories are calculated by

$$\beginarray*20l C_L \hfill & = \hfill & C_\mathrmplantbased \times \left( 1 – R_\mathrmcy \right) + C_\mathrmlivestock – ruminant \times \left( 1 – R_\mathrmgrass \right) \hfill \\ \hfill & \hfill & + C_\mathrmlivestock – monogastric \times \left( 1 – R_\mathrmcy \right) + C_\mathrmlivestock – monogastric\\&& \times R_\mathrmgrass \times \left( 1 – R_\mathrmcy \right) \times \fracF_\mathrmruminant – cropfeedF_\mathrmmonogastric – cropfeed \hfill \\ \hfill & \hfill & + C_\mathrmaquaculture \times \left( 1 – R_\mathrmcy \right) + C_\mathrmmarine – catch \times \left( 1 – R_\mathrmmarine – catch – y \right) \hfill \\ \hfill & \hfill & + \left( 1 – R_\mathrmcy \right) \times C_\mathrmplantbased \times \fracf_\mathrmfinal – product – biofuelf_\mathrmfood \hfill \endarray$$

(4)

where C_L is calories available in each nation L (kcal per capita per day) under the Livestock case, C_plantbased, C_{livestock-ruminant} and C_{livestock-monogastric} are calories available from plant-based products, ruminants and monogastrics²⁷ and C_aquaculture and C_marine-catch are calculated by calorie availability from fish²⁷ multiplied by the ratio of aquaculture and catch⁴⁶. R_grass is grass production change, and R_{marine-catch-y} is marine capture change. F_{ruminant-cropfeed} is the fraction of crop feed for ruminant, and F_{monogastric-cropfeed} is the fraction of crop feed for monogastrics²⁶. R_cy is crop production change calculated as:

$$w_\mathrmiy = \fracP_ic_iR_\mathrmiy{{\mathop \sum\nolimits_i = 1^4 P_ic_iR_\mathrmiy }},$$

(5)

and

$$R_\mathrmcy = \mathop \sum\limits_i = 1^4 R_\mathrmiyw_\mathrmiy$$

(6)

where index i is maize, rice, soybean or wheat, w_iy is the calorie weight of each commodity per country each year, _Pi is the national production of item i in FAO-CBS⁴⁷, c_i is calories per 100 g retail weight for each item²³ and R_iy is national production change (%) of each item in year y after the nuclear wars. $f_\mathrmfinal – product – biofuel$ is the fraction of final product of biofuel in plant-based product, and f_food is the fraction of food in plant-based product.

For the No Livestock case, national-level available calories are calculated by

$$\beginarray*20l C_\mathrmNL \hfill & = \hfill & C_\mathrmplantbased \times \left( 1 – R_\mathrmcy \right) + C_\mathrmmarine – catch \times \left( 1 – R_\mathrmmarine – catch – y \right) \hfill \\ \hfill & \hfill & + C_\mathrmplantbased \times f_\mathrmfeed – to – food \times \left( 1 – R_\mathrmcy \right) \times p_\mathrmfeed – for – human \hfill \\ \hfill & \hfill & + \left( 1 – R_\mathrmcy \right) \times C_\mathrmplantbased \times \fracf_\mathrmfinal – product – biofuelf_\mathrmfood \hfill \endarray$$

(7)

C_NL is national-level available calories in the No Livestock case. f_feed-to-food is the fraction of food crops that are used as feed relative to their usage as food, calculated based on their calorie content²⁶. p_{feed-for-human} is the percentage of livestock grain feed used for human consumption. We tested 0%, 20%, 40%, 50%, 60%, 80% and 100% and used 50% for Table 2 and Fig. 4.

For the Partial Livestock case, national-level available calories are calculated by

$$\beginarray*20l C_\mathrmPL \hfill & = \hfill & C_\mathrmplantbased \times \left( 1 – R_\mathrmcy \right) + C_\mathrmmarine – catch \times \left( 1 – R_\mathrmmarine – catch – y \right) \hfill \\ \hfill & \hfill & + C_\mathrmplantbased \times f_\mathrmfeed – to – food \times \left( 1 – R_\mathrmcy \right) \times p_\mathrmfeed – for – human \hfill \\ \hfill & \hfill & + (1 – p_\mathrmfeed – for – human) \times \left( C_\mathrmlivestock – ruminant \times \left( 1 – R_\mathrmgrass \right) \right. \hfill \\ \hfill & \hfill & + C_\mathrmlivestock – monogastric \times \left( 1 – R_\mathrmcy \right) \hfill \\ \hfill & \hfill & {\left. { + C_\mathrmlivestock – monogastric \times R_\mathrmgrass \times \left( 1 – R_\mathrmcy \right) \times \fracF_\mathrmruminant – cropfeedF_\mathrmmonogastric – cropfeed} \right)} \hfill \\ \hfill & \hfill & { + \left( 1 – R_\mathrmcy \right) \times C_\mathrmplantbased \times \fracf_\mathrmfinal – product – biofuelf_\mathrmfood} \hfill \endarray$$

(8)

C_PL is national-level available calorie in Partial Livestock case. On the basis of the assumed percentage of livestock crop feed to convert to human consumption, instead of wasting the remaining portion of livestock crop feed as in No Livestock case, here we use the remaining livestock crop feed to raise livestock.

The percentage of national household waste is calculated by

$$P_\mathrmwaste = 100\% \times \frac{{C_\mathrmavailable – C_{{\mathop\rmintake}}}}C_\mathrmavailable$$

(9)

P_waste is the percentage of national household waste of food calorie availability in 2010, C_available is the food calorie availability per day per person in each country and C_intake is the national calorie intake per day per person²⁷.

Calorie requirements

The population percentage supported by available calories calculated for the Livestock, Partial Livestock and No Livestock responses indicates the macro-level consequences for food security (Fig. 4). The current average human available calorie supply is 2,855 kcal per capita per day, including food intake and food waste (Fig. 3). Calorie requirements vary significantly with age, gender, size, climate, level of activity and underlying medical conditions. Ref. ²⁷ estimated the national-level calorie availability, calorie intake, calorie from plant-based product, livestock and fish and also calculated the calorie intake of an underweight population with current physical activity of an underweight population with sedentary physical activity and calorie intake lower than the basal metabolic rate. We assume that the calorie intake of an underweight population with current physical activity is needed to support life and regular labour activity.

Uncertainties

This work was done with one Earth system model, with only one ensemble member for all the cases with soot injections >5 Tg, only one crop model and only one fishery model. For the 5 Tg case and the control, there are three ensemble members, but only the ensemble averages are used. The three ensemble members for the 5 Tg case are very similar (Supplementary Fig. 8), so climate variability for the larger forcings would be much smaller than the signal.

CESM is a state-of-the-art climate model, and its simulations of the impacts of nuclear war have been almost identical to simulations with other models for the 5 Tg (refs. ^49,50) and 150 Tg (ref. ⁹) cases. However, further developments in climate models, such as including organic carbon in fire emissions, and better simulating aerosol growth and interactions with the surrounding environment, may improve climate prediction after a nuclear war.

CLM5crop and BOATS are also state-of-the-art models, but future simulations with different models would certainly be useful. CLM5crop compares well with other crop models in response to nuclear war forcing¹⁴ (Supplementary Fig. 2). If anything, CLM5crop underestimates the crop response to nuclear war (Fig. 2 and Supplementary Fig. 2). Because most crop models were developed for the current or warmer climates, further research is needed to understand how crops react to a suddenly cold environment. Our study is the first step to reveal national food security after nuclear wars, but crops may not respond uniformly to the same forcing in each nation, given different farming practices. In addition, multi-model assessment will be essential to fully investigate this problem, and crop model developments are important to understand impacts from surface ozone, ultraviolet radiation and freshwater availability. Furthermore, local radioactive contamination and climate change from nuclear war would impact the insect community. The influence on pests, pollinators and other insects is unclear, and hence further studies are needed.

Some assumptions in this study could be examined in future work. For example, to turn off international trade, the ratio of local production to domestic supply is applied on a national level. Also, to calculate national calorie intake after nuclear wars, we assume that food is evenly distributed in each country. Economic models will be necessary to further understand the contributions of trade and local food distribution systems to human calorie intake after nuclear wars.

This study uses calorie intake from ref. ²⁷, and food loss from harvesting is not considered. If human behaviour and the food industry would change substantially, this would affect our conclusions.