Future warming increases probability of globally synchronized maize production shocks

Changes in Mean Yields

Present-day maize yields vary widely across the globe, depending on the regional climate and crop management system. Yields are highest in intensive, temperate-zone production systems, such as the US Corn Belt and western Europe, followed by regions in China, Argentina, and South Africa. Maize research and development organizations, such as the International Maize and Wheat Improvement Center and the Australian Center for International Agricultural Research (ACIAR), have previously used the different growing conditions of maize to define so-called “maize mega-environments” across which maize cultivars perform similarly (20, 21). The response of maize yields to climate variability and climate change will differ between these different environments and will also depend on management characteristics, such as cropping intensity and irrigation. We therefore use climate (growing season mean temperature, growing season total precipitation, and latitude) and cropping (mean yield and level of irrigation) data as input for a K-means cluster analysis in which we divide the globe into seven growing regions that correspond closely to the maize mega-environments (SI Appendix, Figs. S1 and S2). For each of these seven regions, we develop statistical models relating changes in yield to climate variability (Materials and Methods).

Empirical models of climate-crop relationships use a variety of indicators to capture the effects of climate variability on crop yields (10). Growing season mean temperature and precipitation are easily calculated and commonly used metrics. However, these season-mean variables smooth out the opposing contributions of early season warming and extreme summer heat on plant development. Alternative model formulations therefore make use of growing and killing degree days (GDDs and KDDs, respectively) to distinguish between thermal time to development and the harmful effects of high temperatures (11, 12). Using temperatures of individual months rather than temperatures averaged over the growing season is another way to capture the differential effect of temperatures on plant development by phenological phase. For each growing region, we test a number of model formulations and select the one providing the best fit (Materials and Methods). In six of seven clusters, this is a model using a degree day formulation. The estimated model coefficients (SI Appendix, Table S8) are consistent with well-understood biological constraints (SI Appendix).

In all but the most draconian emission reduction scenario, global annual mean temperature rises by about 2 °C by the mid-21st century compared with the 1980–1999 average. We evaluate the effect of this common temperature target on crop yields by normalizing the monthly multimodel mean patterns of end-of-century temperature changes from the Coupled Model Intercomparison Project Phase 5 (CMIP5) Representative Concentration Pathway 8.5 (RCP8.5) simulations by the global annual mean temperature and adding the temperature anomaly pattern to present-day temperature fields (Materials and Methods and SI Appendix, Fig. S4). We thus create a future climate history that features identical variability as in the historical record but acting on top of an annual cycle in temperature associated with global annual mean warming of 2 °C. Additionally, we assess the risk of failing to reduce carbon emissions by quantifying the potential impacts of 4 °C global annual mean warming (SI Appendix, Fig. S4). Under business-as-usual emissions (RCP8.5), the global mean temperature is projected to increase by 2 °C as early as 2042, with a median prediction of 2055, and by 4 °C between 2075 and 2132. Even in an emissions scenario aiming to stabilize greenhouse gas concentrations by the mid-21st century (RCP4.5), global mean temperature could rise by 2 °C as early as 2052 (SI Appendix, Table S1). We do not consider the effect of changes in the annual cycle of rainfall on future crop yields because of the high uncertainties in the magnitude and pattern of future precipitation changes, and the comparatively higher signal-to-noise ratio in temperature changes (13).

With the exception of a few locations in western Europe and China, maize yields decline everywhere in response to 2 °C of warming, with particularly strong declines in the southeastern United States, eastern Europe, and southeastern Africa (Fig. 2). In the midlatitudes, an increase in GDDs contributes positively to crop yields, while more KDDs lead to yield declines. At 4 °C warming, the negative contribution of additional KDDs far outweighs the positive effect of increased GDDs, so that substantial changes in mean maize yields of >40% are predicted in many places, most notably in the United States, Mexico, eastern Europe, and southern Africa. These values are within the range of response to warming found in previous empirical and crop modeling studies (5). The magnitude of projected yield changes is comparable between the three linear regression models that we test (SI Appendix, Fig. S5). When including quadratic terms in the models, the model fit is generally equivalent or slightly better, and the projected yield reductions are even greater than those projected using the linear models, especially for the 4 °C warming scenario (SI Appendix, Fig. S5). The same is true when we use yield ratios as the dependent variable rather than absolute yield anomalies. Our predictions may therefore be on the conservative side.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Warming-induced changes in mean yield. Relative change in average yield (%) following annual mean global warming of 2 °C (Left) and 4 °C (Right).

Global maize production is highly concentrated within a few locations: Just four countries (United States, China, Brazil, and Argentina) produce 68% of the world’s maize, and the top four maize-exporting countries combined (United States, Brazil, Argentina, and Ukraine) account for 87% of global maize exports (22) (SI Appendix, Table S2). In the United States, China, Brazil, and Argentina (the top four producing countries), mean total production is projected to decline by 18% (17.4–18.3), 10% (10.1–10.7), 8% (7.6–8.1), and 12% (11.3–11.9), respectively, under 2 °C of global warming and by 46% (45.4–47.5), 27% (26.7–28.0), 19% (19.0–19.9), and 29% (27.9–29.0) with 4 °C of warming (mean and 90% confidence intervals; SI Appendix, Table S6). Averaged over the 2012–2017 period, global annual maize exports were 125 million tons and global maize production was 986 million tons (22). In response to 2 °C of global warming, total production in the top four maize-exporting countries is projected to decline by 53 million tons (51.9–54.8), equivalent to 43% (41.5–43.8) of global maize export volume. For 4 °C of warming, projected production declines in these top-exporting countries increase to 139 million tons (135.5–142.0), which amounts to 14% (13.7–14.4) of current global production and exceeds present-day exports.

Changes in Yield Variability

Not all variability in yields is a result of variability in weather and climate. Socioeconomic drivers, plant breeding, pests and pathogens, and other agronomic variables also contribute to yield and production variance. In highly managed, high-yield cropping systems, such as those in the United States, Europe, and China, climate variability accounts for a relatively large share of the total yield variance compared with low-yield environments (Fig. 3 A and B and SI Appendix, Table S3). Irrigation generally reduces sensitivity to changes in temperature (23), lowering the climate-driven share of yield variability in intensive, highly managed environments (SI Appendix, Table S3). In this study, we do not consider future changes in precipitation and, as such, do not quantify changes in future yield variance that result from the covariance between temperature and rainfall variability. Depending on the growing region, the covariance between temperature and precipitation currently explains, on average, 2% of the total yield variance and 17% of the climate-driven variance.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Warming-induced changes in yield variability. (Top) Present-day CV due to all sources of variability (Left) and climate variability (Right). (Bottom) Ratio of changes in climate-driven CV following annual mean global warming of 2 °C (Left) and 4 °C (Right).

A commonly used measure of yield variability is the coefficient of variation (CV) (24), which captures both changes in the SD and mean. In our yield projections, climate warming causes the CV to increase in most places, especially in the United States, eastern Europe, and southern Africa (Fig. 3 C and D), to values many fold higher than present-day values (Fig. 3 A and B). A decomposition of CV changes into contributions from changes in the mean and SD (SI Appendix, Fig. S6) shows that both factors contribute to this increase: Not only will mean maize yields decrease with warming (Fig. 2), leading to an increase in CV, but absolute variability is also projected to increase, including in the major maize-producing regions of the United States, Europe, China, and Argentina (SI Appendix, Fig. S6). The increase in CV due to increased yield variability is comparable to or greater than the increase due to decreasing mean yields. In locations where crop failures become the norm under high (4 °C) warming (mostly in tropical, low-yield environments), the SD of yields ultimately decreases.

An increase in yield variability has implications both for farmers, who rely on their crops for income stability, and for global markets, where crop availability influences food prices (2, 16, 25). Six countries, the United States, China, Brazil, Argentina, Ukraine, and France, collectively account for 73% of global production and 93% of total exports, and in all of these countries, mean yields decrease and yield variability increases under higher temperatures compared with present-day values (Fig. 4 and SI Appendix, Figs. S7 and S9 and Table S7). In line with the theoretical prediction that yield losses will decline precipitously above an optimum temperature (Fig. 1), our results show that extreme yield losses become increasingly likely under global warming. For Ukraine, in particular, yield losses of up to 100% become a possibility under 4 °C warming, where the losses are compared with the mean yield baseline for the period 1999–2008 (Materials and Methods).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Warming-induced changes in yield variability in top-producing regions of the six largest maize-producing and -exporting countries (SI Appendix, Table S2): Probability density functions of yield anomalies with respect to present-day mean yield for present-day climate (black), following 2 °C of annual mean global warming (blue), and following 4 °C of annual mean global warming (red). The vertical gray line denotes a relative yield reduction of 20%, and boxed values indicate mean present-day yield in these areas for present-day climate (1999–2008; black) and for 2 °C (blue) and 4 °C (red) warming.

Extreme crop losses in large-producing areas are presently rare because of the highly controlled environments in which maize is grown in these regions. Yields are tightly constrained around their mean (Fig. 4), and climate-induced yield losses of >10% only occur every 15 to 100 y (Table 1). Climate-induced yield losses of >20% are virtually unseen. In a warming climate, however, these extreme crop failures become increasingly likely. With a 2 °C warming, the probability of a >10% yield loss in any given year in the world’s top four producing countries (United States, China, Brazil, and Argentina) rises to 69%, 46%, 39%, and 50%, respectively. Assuming that weather varies independently between our regional clusters, the probability that maize production will fall by more than 10% in the high-productivity areas of all four countries in the same year is 0% today but increases to 6% under 2 °C warming and 87% under 4 °C warming. Given that these four countries alone account for almost 70% of global maize production, such synchronized production shocks are likely to have tremendous impacts on global cereal markets. This pattern is even more pronounced for the top four exporting countries (United States, Brazil, Argentina, and Ukraine). With a 2 °C warming, the probability of a >10% yield loss in any given year for each of these four countries is 69%, 39%, 50%, and 52%, respectively. Collectively, the probability that these large-exporting countries will incur simultaneous production losses greater than 10% in any given year is virtually zero under present-day climate conditions but rises to 7% under 2 °C warming and 86% under 4 °C warming. The projected changes in variability are robust to statistical uncertainty in model coefficients (SI Appendix, Fig. S9 and Table S7).

Implications for Food Security

The projected increase in maize yield variability across major producing and exporting countries has important implications for global food security, as defined by the ability to provide adequate and affordable food supplies to all people at all times and to ensure economic access to a nutritious diet for all. Meeting this food security goal will become increasingly difficult as the world’s population grows from 7.5 billion today to ∼10 billion by 2050, a 30% increase (26). Virtually all of this growth will occur in developing countries, most notably within sub-Saharan Africa, and over half the global population will reside in urban areas, where international trade plays a key role in ensuring affordable food supplies.

Temperature-induced volatility in global maize prices results from production shocks in both exporting and importing countries. When rising temperatures affect yields in large maize-exporting countries, such as the United States, Brazil, Argentina, and Ukraine, global export supply falls. Similarly, when shocks affect large maize-producing countries that also import, such as China and Mexico, global import demand rises. Synchronous production shocks across multiple large trading countries is therefore expected to lead to a higher frequency in international price spikes. Our analysis indicates that the probability of a synchronous decline in yield of >10% for the world’s three largest maize exporters and three largest maize importers is virtually zero today but jumps to 69% under 4 °C warming (Table 1). Simultaneous production shocks among these large trading countries will have a direct impact on urban consumers, as well as on agribusiness, grain producers, and poor households that spend a large share of their income on staple foods. The degree to which these groups in any given country is affected by international market volatility depends importantly on trade policies designed to insulate domestic markets from international price fluctuations. From a global dataset of 82 countries for the period 1985–2010, the price transmission coefficient for maize is estimated to be 0.63 (27), meaning that, on average, domestic markets experience about two-thirds of the price volatility seen in international markets.

Recent experience from the 2006–2008 global food crisis suggests that if food prices spike and urban consumers become discontent, governments become vulnerable to collapse, particularly in developing countries where consumer food safety nets (e.g., the US Supplemental Nutrition Assistance Program) do not exist (2). As a result, many governments restrict cereal trade to stabilize domestic food markets, resulting in even greater international market instability (27⇓–29). During the 2006–2008 crisis, large maize-exporting countries, including Brazil, Argentina, and Ukraine, imposed export bans on maize, and thus further reduced global export supply, while maize-importing countries introduced trade incentives to lower the price of imported grain, and thus raised import demand. Together, these policies accounted for about one-ninth of the 83% increase in the international maize price over the short period (29). Similar trade policies were implemented for rice and wheat, with even greater contributions to international price spikes (27⇓–29).

The combination of global crop yield variability and disruptive government trade intervention in staple grain markets suggests that the pattern of high price volatility experienced over the past decade (SI Appendix, Fig. S8) will likely persist, or even intensify in the future, with rising temperatures. Maize is an important crop to examine in this context, as it accounts for roughly one-third of both global cereal production and trade, and it is closely connected to other cereal and oil crops through its versatile role in food, animal feed, and fuel markets (30). Grain and oil crop prices are highly correlated over time as a result of substitutions in consumption and production (SI Appendix, Table S5). The impact of rising mean temperatures on maize yields and maize yield variability thus has far-reaching consequences for the stability of global food systems overall as well as for global food security.