CO₂ fertilization of crops offsets yield losses due to future surface ozone damage and climate change

Tropospheric O₃ concentrations are controlled by emissions and climate (Jacob and Winner 2009). First, global warming and climate change in response to increased greenhouse gas concentrations can enhance stratosphere-troposphere exchange of O₃ (Zeng et al 2008, Jacob and Winner 2009, Fiore et al 2012). Second, increased temperature also favors O₃ chemical production and peroxyacyl nitrate (PAN) decomposition, which can lead to higher tropospheric O₃ in certain regions (e.g. Zeng et al 2008, Jacob and Winner 2009). PAN is a secondary pollutant found in photochemical smog. It dissociates slowly in the atmosphere into NO₂ and peroxyacetyl radicals. If PAN is formed in or convectively lifted to the troposphere, it remains stable for long enough to be transported far away from urban sources. This process is vital for tropospheric O₃ production as it can indirectly transport NO_x to regions where it can more efficiently produce O₃. Third, enhanced convection from increased temperature transports O₃ precursors and lead to an increase in O₃ production. Lastly, lightning frequency increases with increased temperature and results in a 22% increase of lightning-produced NO_x (Zeng et al 2008). However, a warming climate could lead to higher water vapour in the atmosphere and a decreased in background O₃ concentration. In conclusion, the background O₃ and climate penalty-driven pollution have opposite sensitivities to climate change (Jacob and Winner 2009, Schauberger et al 2019).

In the United States alone, crop losses due to tropospheric O₃ cost more than 5 billion USD annually (Ainsworth et al 2012, Betzelberger et al 2012). It is estimated that the cost in developing countries is much higher, where agricultural techniques are less advanced, and the productivity of staple crops are susceptible to the effects of climate change via more frequent occurrences of droughts, floods, pests and disease outbreaks (The Royal Society 2008). A recent study by Li et al (2022a) showed that O₃ damage causes 7%–19% relative yield losses in China from 2010 to 2017 for rice, wheat and soybean (Li et al 2022a). The estimated global average yield loss due to O₃ damage in 2000 is 5.4%–15.6% for soybean, 7.3%–12.3% for wheat, 2.8%–3.7% for rice, 2.4%–4.1% for maize (Van Dingenen et al 2009).

Several studies have investigated the global impact of O₃ on agricultural crop yields with future air quality scenarios (Van Dingenen et al 2009, Avnery et al 2011, Tai and Val Martin 2017) to the year 2030 or 2050. Van Dingenen et al (2009) provided the first estimate of global crop yield losses due to O₃ in the future, using the optimistic 'current legislation scenario' that assumes current air quality legislation is being fully implemented in 2030. Avnery et al (2011) commented that such scenario is overly optimistic as legislation enforcement often lags behind the announcement. Avnery et al (2011) examined the impact of O₃ exposure on future crop yields using the Intergovernmental Panel on Climate Change Special Report on Emissions Scenarios (IPCC SRES) year-2030 scenarios and found that O₃ causes 12.1% and 16.4% soybean yield loss in 2030 for the Representative Concentration Pathway (RCP)2.6 and RCP8.5 scenarios, respectively, which the RCP2.6 and RCP8.5 scenarios cover the minima and maxima of the O₃ projections However, neither Avnery et al (2011) nor Van Dingenen et al (2009) account for the effects of rising CO₂ concentration and climate change in future scenarios, which could not fully project future crop yield changes. A recent study by Tai et al (2021) summarized three approaches of modeling O₃ impact on crop yields: concentration-based metrics, flux-based metrics, and mechanistic modeling. They found that only mechanistic modeling (i.e. JULES-crop) could address the co-effects of CO₂ fertilization. JULES-crop model is thus selected to further perform factorial simulations with the combination of RCP2.6 climate and RCP8.5 CO₂, climate and O₃ scenarios to investigate the sensitivity of crops to these driving factors.

This study investigates future crop yield response climate change, O₃ and CO₂ and their interactions using JULES-crop over the period 2005–2100. Many existing studies (Fiore et al 2012, Schauberger et al 2019, Hayes et al 2020, Sampedro et al 2020) simulate crop response to climate projections spanning only a few years each due to computational limitations, while this study simulate long transient timeseries. This approach could help distinguish an actual anthropogenic-forced climate signal from internally generated climate variability (Nolte et al 2018). With rising CO₂ expected in the RCP8.5 scenario, the exposure-yield relationship derived from concentration-based metrics would be less applicable due to the effect of CO₂ fertilization (Tai et al 2021). JULES-crop, in this case, could provide insights on how CO₂ interacts with O₃ damage on yield.

To evaluate the performance of JULES-crop simulation, JULES-Crop is applied over the historical period from 1961 to 2005 with climatology from CruNCEP, modelled ozone fields and observed CO₂ data to simulate crop yields for four major global crops: soybean, wheat, maize, and rice (see supplementary material). The order of O₃ damage for the four crops simulated by JULES for year 2000 is consistent with the estimation by Van Dingenen et al (2009), whereby soybean has the highest sensitivity and maize has the least (supplementary material).

For future simulations, JULES was forced over the 2005–2100 period with changing CO₂, O₃ and climate according to the RCP2.6 and RCP8.5 scenarios.

The simulations are initialized from the historical simulations from Osborne et al (2015). However, when a driving variable is fixed (i.e. O₃ or CO₂), it is fixed at the year-2005 value for the transient future period. The model is applied for the 2005–2100 transient simulation using a factorial design varying or fixing CO₂, O₃ and varying climate according to the RCP scenarios (table 2). Given that the historical simulations were driven with merged Climatic Research Unit and National Centre for Environmental Prediction data (CRUNCEP) and the future simulations with HadGEM2-ES climatology, and the fact that the model requires 1–2 years of spin-up for the fast fluxes (soil moisture, soil temperature and NPP), we spun up the model for 5 years using 2005 climatology before the actual simulation.

The future harvested areas for the four crops (i.e. wheat, soybean, maize, and rice) were kept constant at the year-2000 distribution as presented by Monfreda et al (2008) and Ramanutty et al (2008) since the yield changes are independent of changes in the cultivated areas. There were 13 global simulations, including eight factorial future transient simulations with combinations of the two RCP scenarios driving data prescribed with CO₂ and O₃ (table 2). RCP2.6 and RCP8.5 were selected as they represent the upper and lower boundary of future RCP and climate change scenarios. These runs were used to investigate the relative impacts of climate change, CO₂ and O₃ damage on crop yields. The relative contribution of a driving variable is calculated by subtracting the yield of the baseline RCP-2005 run from RCP (to estimate the climate change contribution), RCP + CO₂ (to estimate the CO₂ contribution) and subtracting the baseline yield from RCP + O₃ (to estimate the O₃ contribution).

As shown in figure 1, JULES-crop simulates a higher yield for all crops in RCP8.5 than in RCP2.6 if the CO₂ fertilization effect is included. However, without CO₂ fertilization, RCP8.5 results in a lower yield than RCP2.6 due to O₃ damage and climate change alone (blue line in figure 1). The effect of O₃ damage is greatest for soybean and wheat; as C3 plants they are more sensitive to O₃ damage than C4 maize (Long et al 2005, Williams et al 2017).

Zoom In Zoom Out Reset image size

In JULES-crop, the effect of CO₂ fertilization is higher than O₃ damage; the CO₂ increase in RCP8.5 more than compensates the O₃ damage, and in RCP2.6, the compensation cancels out the O₃ damage effect for soybean and wheat. CO₂ can also reduce the leaf stomatal conductance of vegetation and therefore reduce plant O₃ uptake and damage. Climate change alone in RCP8.5 has a significant negative effect on yields, causing around one-third of the yield loss from 2010 to 2100 for all four crops (see supplementary materials for more information). Global soybean experiences the highest yield loss from O₃ damage among the crops, losing around 1 tonne per hectare. The climate change impact on crop yields in RCP2.6 is significantly less than the impact in RCP8.5; the yield stays mainly constant from 2005 to 2100.

The O₃ impact is substantially higher in RCP8.5 than in RCP2.6 (figure 1). The impact of O₃ on crop yield is higher when CO₂ increases are not included in the simulation. This effect is the result of a higher CO₂ concentration causing stomatal closure and reducing the uptake of O₃ into the stomata, which leads to a lower plant O₃ impact. This means that CO₂ fertilization offsets yield loss due to O₃ damage. However, this mechanism does not apply to maize and rice in the RCP8.5 scenario (figure 1(a): maize and 1d: rice), whereby including O₃ does not further reduce the yield compared to the standard scenario (without CO₂−O₃ interactions). The result suggests that climate change dominates the changes in maize and rice yields, which can be explained by the regional analysis in the next session.

Figure 2 shows that soybean is the most sensitive to O₃ damage, followed by wheat. Rice and maize have similar O₃ damage sensitivity. All three C3 plants (soybean, rice and wheat) show higher sensitivity to CO₂ rise than the only C4 plant (maize) among the staple crops. Under the combined effect of CO₂ fertilization and O₃ damage, soybean shows a slightly higher compensation effect than the other C3 crops when compared with simulations without enhanced CO₂ fertilization. All C3 crops show similar percentage yield change when RCP8.5 + CO₂ + O₃ are compared with the climate change-only scenario.

Zoom In Zoom Out Reset image size

The general yield loss of the four crops due to O₃ damage simulated by JULES-crop are consistent with other studies. Together with results from previous studies, we show that the sensitivity of the four crops to O₃ are ranked in the order from high to low as soybean, wheat, rice, and maize, with soybean being the most sensitive and maize being the least (Van Dingenen et al 2009, Feng et al 2022). Maize is the least sensitive because as a C4 plant, maize has an efficient CO₂-concentrating mechanism, which essentially decouples stomatal conductance and the ensuing ozone damage from photosynthesis.

Climate change-only simulations show that temperature and precipitation change in RCP8.5 scenarios have a considerable negative impact on crop yield. Figures S4 and S5 shows that all crops are negatively impacted by climate change, with rice most impacted. This is because most rice-producing countries are concentrated in the tropical regions, where climate change would cause temperature rises above the optimal temperature threshold of rice (Erda et al 2005, Parry et al 2005, Auffhammer et al 2012).

In the RCP8.5 scenario, the simulated crop yield for all four crops decreased due to climate change alone (see supplementary materials for more detail). Yields for all four crops fall by more than 7%, according to table S4. These simulated reductions in yields resulting from climate change are mainly due to increasing temperature and frequency and intensity of extreme weather events such as storms (Villarini and Vecchi 2012) as reported by many other studies (Kang et al 2009, Asseng et al 2014, Porter et al 2014, Levis et al 2016, Mills et al 2018, Schauberger et al 2019). The reductions will be explained in the next section.

According to figures 1, S4 and table S4, on a global scale, RCP8.5 climate with RCP8.5 +CO₂ + O₃ scenarios simulate increases in yields for all crops. It increases the most for soybean (>60%), and then wheat and rice (20%–30%), and it increases the least for maize (<6%). The fact that maize has the smallest increase in yield of the four crops is due to the phenology of C4 plants. In RCP2.6 scenario, the CO₂ concentration increases to 440 ppm in 2050 and decreases back to 400 ppm in around 2075. Therefore, the overall CO₂ fertilization effect in RCP2.6 is small compared to the RCP8.5 scenario. When the CO₂ fertilization effect is not considered, O₃ damage is very noticeable; yields decrease for all crops in 2050 with the largest decline in soybean (22%) and wheat (16%); yield reductions in rice and maize are smaller in comparison (<8%). O₃ damage on soybean is higher than on other crops in general. According to UNEP (2018), the average O₃ induced soybean yield loss is around 23%, while other crops are less than 10%. Mills et al (2018), using multiple observations, also concluded that soybean is more sensitive to O₃ than other crops. Osborne et al (2016) showed that the O₃ sensitivity of soybean has increased overtime by one third between 1960–2000. It is possible that selective breeding strategies that target high yield and high stomatal conductance has inadvertently selected for soybean cultivars that have higher O₃ sensitivity overtime (Osborne et al 2016).

Figure 3 summarizes the factorial runs by plotting the global yield change from 2010 to 2100 according to the annual CO₂ and O₃ concentration combinations. The colour of the data points represent the yield change. The yield for all crop types decreases when CO₂ is below 430 ppm. When CO₂ is above 500 ppm, even with a high O₃ concentration, the yield still increases. For RCP8.5 + O₃ scenario, with increasing O₃ concentration alone (the horizontal straight line), yields decrease for soybean and wheat because they are both O₃ sensitive crops. Since O₃ damage is most significant during the crop growing period, the regional seasonal O₃ concentrations were plotted in figure S1.

Zoom In Zoom Out Reset image size

At the global scale, O₃ damage impact on rice yield is not as large as expected (Pang et al 2009, Tang et al 2011) according to figures 1 and 3. Therefore, we look at the regional yield change to find which regions contribute to the reduction of yield in RCP8.5 when CO₂ and O₃ are not included in the model in the next section.

Usually, crop growing seasons coincide with high-O₃ seasons. Figure 4 shows that higher O₃ concentrations during the growing season negatively correlate with yield for all regions. According to figure 4, using the RCP8.5 scenario, the higher the O₃ concentration during the growing season, the more crop yield will be affected by higher temperatures. This effect is also reported by Van Dingenen et al (2009) and Avnery et al (2011), who used a concentration-based approach to estimate yield loss.

Zoom In Zoom Out Reset image size

In general, in RCP2.6, decreases in O₃ concentrations are not mainly due to an assumed clean air policy but rather due to a climate change mitigation policy, which reduces emissions of the greenhouse gases such as CH₄ and postulates a transition to more renewable energy production (Chalmers et al 2012, Wild et al 2012, Jones and Warner 2016). The reduction of O₃ is a co-benefit of the decline in O₃ precursors (CH₄, NOx, CO) from the transition to renewable energy. Therefore, this suggests that climate change mitigation is as important as clean air policy in improving food security. However, regarding the yield improvement alone on a plant physiology perspective, a clean air policy will be more beneficial as CO₂ fertilization affords a substantial yield increase, and it can compensate for the O₃ damage on crop yield.

According to figures S6–S8 in RCP8.5, soybean, wheat, and rice experience a significant yield increase in all regions, although O₃ concentration increases because of the strong CO₂ fertilization that overshadows the damaging effect of O₃ (figure 5). Since maize is a C4 crop less sensitive to CO₂ and O₃ increases, it shows a slight increase in yield. RCP2.6 in figure S3 shows that O₃ concentration decreases in 2030, resulting in increased yield in all regions. The yield increase in RCP2.6 is around 10%–15%, while RCP8.5 could be up to 80%.

Zoom In Zoom Out Reset image size

Figure 6 implies that majority of the rice yield reduction is because of climate change. The climate change in Indonesia contributes to the largest decline of rice yield from 6.5 to 4.0 tons ha^–1, mainly due to increases in ambient air temperature with the annual average temperature over 30 °C from 2070 (figure 4).

Zoom In Zoom Out Reset image size

In the RCP8.5 scenario, Asian countries experience a high frequency of extreme weather events with heatwaves of temperatures greater than 30 °C projected to increase by 131% (Lee et al 2014) compared to the current climate (figure 6). The frequency of heavy rainfall is expected to increase by 24% in RCP8.5 for some regions, as shown in figure S5. These extreme weather events would result in frequent floods and droughts (Lee et al 2014). In the RCP8.5 scenario, an intensified monsoon season over East Asia is projected. Since all the key rice-producing countries are in East Asia, this significantly impacts the yield. Advancements in agronomical technology would not be quick enough to adapt to the changing climate (Jagadish et al 2012, Scheben et al 2016).

Since heatwaves in East Asia usually coincide with the rice growing season, it has an enormous impact on yield. Nevertheless, high CO₂ concentrations in the RCP8.5 scenario offset the yield loss from climate change.

In table 3, JULES-crop yield is compared against the FAO Working Paper 2050 projections. According to this data, the FAO projection is much more optimistic than JULES-crop using the RCP8.5 and 2.6 projections. All four crops show a higher yield growth in FAO data than JULES-crop RCP8.5 and 2.6 scenarios simulations. The difference is that because FAO projections are based on socioeconomic factors, it is fundamentally different from the biophysical factors JULES-crop uses. Among the four crops, soybean yield growth for RCP8.5 is the closest to FAO projection, suggesting that soybean calibration improves the yield estimation (table 3).

JULES-crop future simulations allow a comparative study of human and climate influence on crop yields. In general, climate change, O₃ damage and CO₂ fertilization also result from human activities. We can divide the factors into active and passive human influences. Active human influences includes land use change, fertilizer application, irrigation, and other management methods, which follow primarily socioeconomic and technological development. Passive influences are the biophysical effects that JULES-crop currently represents, i.e. impacts of climate change, O₃ damage and CO₂ fertilization on yields. FAO statistics consider the active but not the passive influences. The crop production projection in the FAO report (Alexandratos et al 2015) was based on demand and trade projections. Demand and trade were based on demographic and socioeconomic projections for each country. Alexandratos et al (2015) argues that future yield growth is attained mainly by closing the yield gap in developing countries, whereas yield ceilings have already been reached in some developed countries. Mills et al (2018) also found that O₃ damage is one of the most important factors that contribute to yield gap, especially O₃ sensitive crops, soybean and wheat (Mills et al 2018). The FAO report does not account for extreme weather events or pest infestation outbreaks, and it also neglects CO₂ fertilization, leading to a considerable underestimation of productivity. Therefore, it tends to be overly optimistic (table 3). While JULES-crop does not have nitrogen limitation, for the RCP scenarios with CO₂ and O_3, it is thus also over-optimistic, assuming that the CO₂ effect would be fully translated into yield growth (Long et al 2004).

Besides carbohydrates, a healthy diet should include nutrients such as protein, phosphate, and minerals. (Tulchinsky 2010). A phenomenon has been observed in non-legume C3 crops during high CO₂ conditions. The crops cannot assimilate sufficient nitrogen from soils to maintain the usual C:N ratios in tissues (Bloom et al 2012, Myers et al 2014). This phenomenon has been known as 'carbohydrate dilution', in which CO₂ stimulated carbohydrate production by plants dilutes the rest of the grain components. Of all the elements, changes in nitrogen content at elevated CO₂ concentrations have been the most studied. This carbohydrate dilution effect has been observed consistently in most crops but less significant on leguminous crops. Research suggested high CO₂ concentration could stimulate greater nitrogen fixation to maintain tissue C:N ratios (Erda et al 2005, Myers et al 2014). As projections showed that atmospheric CO₂ would likely continue to increase in the future, programmes such as biofortification, supplementation and breeding crops cultivar for decreased sensitivity to CO₂ would help address the nutrient deficiency caused by CO₂ fertilization.