Six decades of warming and drought in the world’s top wheat-producing countries offset the benefits of rising CO2 to yield

Wheat production and yield vis-à-vis climate trends

Wheat is currently grown in all six continents except Antarctica. The leading producers include China, the Russian Federation, Ukraine, Kazakhstan (RUK), India, USA, France, Canada, Pakistan, Germany, Argentina, Turkey, Australia, and United Kingdom (Fig. 1 and Supplementary Table 1). The total grain production of these twelve countries is estimated at 600 megatons (2019 data), which accounts for over 78% of the global wheat production. The top three producers are China with 133.6 megatons per year (Mt y⁻¹), RUK with 114.1 Mt y⁻¹, and India with 103.6 Mt y⁻¹. RUK contains the largest harvested area of 45.8 million hectares, followed by India with 29.3 million hectares and China with 23.7 million hectares (Fig. 1A). Despite a relatively small harvested area of 10.1 million hectares (only 22% of RUK’s harvested area), the United Kingdom, France, and Germany account for the world’s highest yields per hectare, with 8.93 tons ha⁻¹, 7.74 tons ha⁻¹, and 7.40 tons ha⁻¹, respectively (compared with the world’s average yield of only 3.2 tons ha⁻¹), accounting for a total yearly production of 79.9 Mt y⁻¹.

While all these twelve major wheat producers saw an increase in yield during the last six decades (Fig. 1B), China displayed the most noteworthy increase with a nearly sevenfold higher yield in 2019 than in 1961 and a mean total increase of 5.19 tons ha⁻¹ for the period of 1961–2019. Germany, the UK, and France reported comparable yield increases of 5.20 tons ha⁻¹, 5.19 tons ha⁻¹, and 4.81 tons ha⁻¹, respectively, during this period, suggesting an approximately 1.6-fold improvement since 1961 (Fig. 1B). Australia, RUK, and Turkey reported the lowest gains with only 0.87 tons ha⁻¹, 1.26 tons ha⁻¹, and 1.71 tons ha⁻¹, respectively, representing improvements of 67%, 150%, and 175% in yield per hectare since 1961.

Yield increase occurred despite the steep rise in temperature (nearly 1.2 °C) in the twelve countries during the last six decades (Fig. 1C). Water deficit—calculated as the difference between potential evaporative demand and precipitation (PET—P; mm H₂O y⁻¹)—also increased by an average of (sim) 29 mm of H₂O for the same period. Increases in yield since the early 1960s were likely due to breeding and agrotechnological advances, improved management, and a steep rise in atmospheric [CO₂] of (sim) 98 μmol mol⁻¹, from 315.9 μmol mol⁻¹ in 1961 to 413.4 μmol mol⁻¹ in 2019 (insert in Fig. 1C).

Unraveling the impacts of climate and [CO₂] on yield

Based on previous studies^30,31, we used a log-linear model to quantify the impact of [CO₂] and daily minimum (T_min), maximum (T_max), and mean (T_mean) temperatures, as well as seasonal water deficit (PET-P), and rainfall distribution on wheat yield. Climate variables were obtained from the TerraClimate data set³², while monthly records of [CO₂] from the Mauna Loa station were used to model the effects of CO₂ (see “Methods”). To quantify wheat yield as a function of climate variables and [CO₂], we included all 12 countries in the regression analysis. Supplementary Table 2 presents summary statistics of all variables, while Supplementary Fig. 1 depicts trends in T_mean and PET-P per country.

Since climate variables tend to be correlated over time (Supplementary Table 3), controlling for all of these variables in the model facilitates the estimation of their distinct effect on yield. We used country-specific trends to distinguish changes in wheat yield related to climate and [CO₂] from those attributed to agrotechnological advancements, changes in country-specific policies, and other local-changing factors (e.g., economic and population growth; more information on how this was done can be found in “Methods”). We also included country-specific effects across all models to account for unobserved time-invariant heterogeneity at the country level, such as geographical properties, edaphic characteristics, and other local-specific features (see “Methods”).

Table 1 reports the estimated regression coefficients of four models, (1) using only temperature variables (T), (2) temperature and water-related (i.e., seasonal rainfall distribution and water deficit as PET-P) variables (T + W), (3) including [CO₂] (T + W + C), and (4) the interaction between [CO₂] and climate variables (T + W + C + interactions).

Among the temperature measures, only T_mean had a consistently significant effect on yield (p < 0.01, two-sided t-test), with increases in T_mean during the growing season leading to reduced wheat yields (negative effect) in all four models. This negative effect weakened (p < 0.1, two-sided t-test) when the water deficit variable was included in the model (T + W) but strengthened again (p < 0.01, two-sided t-test) when [CO₂] was further incorporated (T + W + C and T + W + C + interactions models). The negative effect of water deficit suggests a reduction in annual yield when the evaporation demand is greater than precipitation because of reduced rainfall or increased evaporative demand. Rainfall distribution is an important variable in predicting wheat yields^30,33. In our models, rainfall distribution had a negative effect on yield (Table 1), implying reduced annual yield when rainfall is less uniformly distributed throughout the season, consistent with other studies³⁰. This effect is likely because of decreased water retention in the soil when rainfall is more intense and less frequent during the season. Notably, the negative effect of rainfall distribution in all models was not statistically significant (p > 0.1, two-sided t-test), indicating that rainfall distribution might be an important predictor of yield in some but not all places.

Among all variables, atmospheric [CO₂] had the most substantial effect on yield, with a net beneficial effect in both T + W + C and T + W + C + interaction models. This effect was two and four orders of magnitude higher than T and W, respectively (Table 1). This much stronger effect of [CO₂] indicates that the [CO₂] (times) T or W interaction is mainly through T and W moderating or enhancing the effect of [CO₂] and not the other way around. The negative sign of the T_mean (times) [CO₂] interaction (Table 1) indicates that the adverse (negative) impact of warming on yield weakens under elevated [CO₂] (i.e., when atmospheric [CO₂] rises), consistent with previous experimental studies³⁴. Interestingly, the water deficit (times) [CO₂] interaction was positive suggesting that a rise in [CO₂] increases the negative effect of water deficit on yield, contrary to prior studies^12,13,21,35.

Since the irrigated area of wheat in India increased dramatically from a low 10% in the early 1970s to almost 100% by the late 2000s with a substantial impact on wheat yield³⁶, we further repeated the analysis by excluding India from our models. As shown in Supplementary Table 4, the estimated regression coefficients of the four models that excluded India from the analysis are similar to those presented in Table 1 when India was included in the model. The only exception was the insignificant negative T_mean (times) [CO₂] interaction (p > 0.1) when India was excluded from the analysis (Supplementary Table 4), which implies that a significant proportion of the beneficial effect of [CO₂] that neutralized the negative impact of warming was due to the increased irrigation in India³⁶.

Yield changes due to climate and [CO₂]

We further analyzed yield changes due to climatic variations and [CO₂] rise by applying regression analysis under different scenarios adjusting climate variables and [CO₂] to the levels reported in 1961–1965. We then established the T + W + C model (Table 1) with each variable fixed at its 1961–1965 average, each variable at a time. We translated these effects into relative changes in yield due to climate and [CO₂] by calculating the difference between predicted yields using our baseline regression and predicted yields using climate variables or [CO₂] levels recorded in 1961–1965. We also assessed current changes for each country separately by estimating the contribution of [CO₂] and each of the climate variables to the yield in 2018–2019 compared with the baseline yield in 1961–1965.

In general, the benefits of [CO₂] rise outweighed the adverse impact of warming and water deficit over time (black line in Fig. 2A). The yield gained due to an almost 100 μmol mol⁻¹ increase in [CO₂] was 6.8% on average, with a linear increase from 1961 to date (yellow line in Fig. 2A). Such an increase in wheat yield per (Delta)[CO₂] of (sim) 100 μmol mol⁻¹ is within the range of 5–10% reported from FACE experiments for a similar increase in [CO₂]^12,13. Warming of (sim) 1.2 °C from 1961 to 2019 (Fig. 1C) reduced the annual yield by nearly 2.8% (red line in Fig. 2A), and an increased water deficit of 28.8 mm H₂O m^–2 for the same period reduced the yield by almost 1% (blue line in Fig. 2A). On average, the gain in wheat yield from the recent rise in atmospheric [CO₂] was enough to overcome the loss due to warming and water shortage (red and blue dashed lines in Fig. 2A), consistent with other studies³⁷. The finding that warming of (sim) 1.2 °C suppressed the yield by (sim) 3% is slightly lower but generally in line with previous grid-based and point-based simulations^8,9. However, despite the previously reported benefit of elevated [CO₂] on yield under drought conditions^20,38, our findings of a positive water deficit (times) [CO₂] interaction (Table 1 and Fig. 2A) and reduced yields in the presence of increased [CO₂] combined with intense drought (blue dashed line in Fig. 2A and Supplementary Fig. 1) question this conclusion.

Figure 2B presents the relative change in yield across the 12 major wheat-producing countries in 2018–19. Yield gains due to the beneficial effect of [CO₂] (yellow bars in Fig. 2B) outweighed losses due to warming and intensified drought in most countries (red and blue bars in Fig. 2B). Germany and France—two of the three countries with the highest yield per hectare in the world (Fig. 1B)—were the only countries that saw loss (− 10.1% due to warming and drought compared to + 7% due to [CO₂] rise) and no gain (− 7% due to warming and drought, and + 7% due to [CO₂] rise) in yield, respectively, at the end of the sampling period relative to the 1961–1965 baseline. These two countries experienced the worst temperature rise and intense drought conditions in their wheat-growing areas at the end of the 2010s (Supplementary Fig. 1), enough to obliterate or reverse the gains due to the rise in atmospheric [CO₂] into significant yield losses. The largest wheat producers—China and RUK—also experienced a substantial decrease in yield, with losses of 5.6% and 5.3%, respectively, mainly due to similar warming of + 0.31 °C decade^–1 (p < 0.01) in both regions, and a water deficit of 21.5 mm decade^–1 (p < 0.05) in China and 6.1 mm decade^–1 (p < 0.1) in RUK (Supplementary Fig. 1). Nevertheless, such losses were insufficient to reverse gains in these two countries due to the rising [CO₂] in the last six decades (Fig. 2B).

Source: Ecology - nature.com