Increasing CO₂ levels fertilize C₄ grass production

The potential for direct CO₂ effects on plant productivity has long been recognized, motivating a network of free-air CO₂ enrichment (FACE) experiments^7,8. These show widespread but context-dependent direct CO₂ effects, including increased photosynthesis, reduced stomatal conductance and water use, and increased biomass production^9,10. However, in situ CO₂ enrichment experiments have largely omitted major tropical and subtropical ecosystems^7,8. In tropical forests, long-term vegetation monitoring is filling that gap, strongly implicating CO₂ fertilization as a driver of enhanced tree productivity⁴. In tropical and subtropical savannas, long-term observations and modelling have also suggested a role for CO₂ fertilization as a driver of C₃ woody plant encroachment^11,12. However, although appropriate monitoring data are available¹³, equivalent long-term analyses of CO₂ responses in C₄ savanna grasses are lacking. This uncertainty means that net outcomes for tree–grass dynamics remain unclear, including possible effects on fire regimes¹⁴, limiting our ability to model and forecast savanna responses to global change.

In situ experiments in temperate grasslands and chamber experiments with savanna grasses provide a basis for generating reasonable hypotheses for how tropical C₄ grasses will respond to increasing levels of CO₂. Because the C₄ photosynthetic pathway evolved as a CO₂-concentrating mechanism that suppresses photorespiration, C₄ grasses are predicted to show smaller direct photosynthetic responses to rising atmospheric CO₂ levels than C₃ grasses. In C₄ crops, which have undergone intensive selection for high yields, this CO₂-concentrating mechanism eliminates CO₂ fertilization under well-watered conditions¹⁵. However, photosynthesis in wild C₄ grasses is less efficient, and CO₂ fertilization has been observed across a diversity of wild C₄ grass species¹⁶. Although this may be partially due to enhanced activity of the photosynthetic enzyme Rubisco, higher levels of CO₂ also reduce stomatal aperture, alleviating water deficits to promote carbon fixation. FACE experiments with C₄ crops show that this mechanism enables CO₂ fertilization under water scarcity¹⁰. Field-based experiments in mixed C₃–C₄ North American temperate prairies have also shown that grass productivity is enhanced by FACE, especially in seasonally dry environments, in dry years, at drier sites and under high soil nutrient availability^8,9,17 (but see ref. ¹⁸). However, existing syntheses of field experiments do not disentangle C₃ from C₄ plant responses, and existing quantitative synthesis of smaller-scale experiments is out of date¹⁶ and includes only a subset of experiments^8,9. Together, these stop short of evaluating in situ grass production in wild C₄-dominated systems.

Here we provide a comprehensive synthesis of C₄ savanna grass responses to rising levels of CO₂, combining a meta-analysis of experiments, new field observations and a vegetation model case study. First, we used a meta-analysis of CO₂-enrichment experiments to evaluate whether CO₂ fertilization is globally widespread across wild C₄ grasses and interacts with water availability. Second, we analysed a unique 32-year time series of savanna grass aboveground biomass from 533 plots in the Kruger National Park, South Africa (‘Kruger’ henceforth), to evaluate whether experimental findings translate into in situ biomass production. Finally, we used a land surface model (the Community Land Model (CLM)⁶ to explore potential outcomes for grassy ecosystems in the future. Combining independent lines of evidence overcomes limitations inherent in previous studies and enables us to evaluate whether hypothesized CO₂ fertilization of C₄ savanna grasses is increasing their production in the field.

Systematic meta-analysis of 70 experimental studies showed that CO₂ fertilization of aboveground biomass is widespread among wild C₄ grasses under water limitation (estimated marginal mean log response ratio under elevated levels of CO₂ (logRR-CO₂) = 0.34 (95% confidence interval: 0.06–0.63)), but not well-watered conditions (logRR-CO₂ = 0.16 (−0.04 to 0.37); Fig. 1a). Photosynthetic rate also increased under elevated levels of CO₂ in water-limited plants (logRR-CO₂ = 0.21 (0.00–0.42); Fig. 1a). Variable responses of root biomass to elevated levels of CO₂ (logRR-CO₂ = 0.00 (−0.31 to 0.36) for well-watered plants; logRR-CO₂ = 0.00 (−0.49 to 0.50) for water-limited plants) contributed to variation in total (aboveground plus belowground) biomass trends (logRR-CO₂ = 0.10 (−0.39 to 0.60) for well-watered plants; logRR-CO₂ = 0.12 (−0.41 to 0.68) for water-limited plants). Experimental factors (CO₂ treatment size, pot size and experimental duration) had minimal effects on grass trait responses (Extended Data Fig. 1). In contrast with a previous meta-analysis on C₄ grass CO₂ responses¹⁶, supported by 24 years of additional research, we found that CO₂ enrichment effects were larger under water limitation for aboveground biomass. This fertilization effect on shoot biomass was consistently stronger under water-limited conditions, although variable in size, across C₄ subtypes, clades and regions (Extended Data Fig. 2).

As water limitation in experiments (due to a reduction of watering or rainfall) consistently amplified biomass responses to elevated levels of CO₂, we next examined how CO₂ altered C₄ grass water use (Fig. 1b). CO₂ enrichment decreased stomatal conductance in well-watered plants (logRR-CO₂ = −0.40 (−0.60 to −0.26)) but not in water-limited plants (logRR-CO₂ = −0.08 (−0.27 to 0.10); Fig. 1b,c), such that stomatal conductance decreased significantly under water limitation at ambient levels of CO₂ but not at elevated levels of CO₂ (Fig. 1c). Under water limitation, leaf water deficits were alleviated under elevated levels of CO₂ (that is, higher midday leaf water potential; logRR-CO₂ = −0.43 (−0.71 to −0.15); Fig. 1b,c). However, decreased transpiration under elevated levels of CO₂ did not increase soil water content in water-limited treatments (logRR-CO₂ = 0.08 (−0.10 to 0.25)). Modelling of C₄ grass water relations under declining soil water (Fig. 1c and Extended Data Fig. 3) suggests that lower stomatal conductance under elevated levels of CO₂ interacts with the hydraulic system to maintain leaf water potential. Together, the results of this experimental synthesis suggest that increased levels of CO₂ substantially reduced water deficits in wild C₄ grasses, which translated into photosynthetic gains under water limitation. As decreasing soil moisture and increasing drought frequency are likely outcomes of climate change arising from increasing levels of CO₂ (refs. ^19,20), these physiological interactions have large potential effects on the vegetation trajectories of grassy ecosystems in the future.

Second, we examined the in situ responses of the C₄ grass layer in a tropical savanna (Supplementary Fig. 1) to ambient changes in climate and atmospheric CO₂ levels to determine whether experimental results translated into increases in grass biomass production over time as CO₂ levels increased. Grass biomass varied spatially and temporally but increased overall across the 32-year observation period in Kruger (Supplementary Tables 1 and 2). Grass biomass increased for 2 years after wetter years (Fig. 2a,b and Supplementary Tables 1 and 2), reflecting the accumulation of grass biomass across years (Fig. 2e and Supplementary Tables 1 and 2, showing increased grass accumulation with time since fire), and perhaps also legacy effects of rainfall on biomass production from year to year²¹. Fewer rainfall events were also associated with greater grass biomass (Fig. 2c and Supplementary Tables 1 and 2), a relationship that arises because fewer events correspond to more rainfall per event (after controlling for annual rainfall). However, neither annual rainfall, the number of rainfall events, nor the mean rainfall event size changed directionally across the multi-decadal study period to explain the observed increase in grass biomass (Fig. 2f and Extended Data Fig. 4a–c). Conversely, high temperatures tended to reduce grass biomass (Fig. 2d and Supplementary Tables 1 and 2), yet a warming trend over the study period (Extended Data Fig. 4d), associated with more days exceeding the temperature optimum for C₄ photosynthesis (Extended Data Fig. 4e), was insufficient to completely offset the overall biomass increase over the study period.

Changes in herbivory could not explain increases in C₄ grass biomass in the field²². Overall herbivore density has increased in Kruger over the observation period²³ (Supplementary Fig. 2), suggesting that grazing pressure has probably increased, which would, if anything, decrease grass standing biomass²². To further examine this possibility, we ran statistical models on the subset of plots with low relative abundance of herbivory-adapted lawn grasses. Subsetting the data in this way did not eliminate overall increases in grass biomass with time (Extended Data Fig. 5 and Supplementary Table 2), reinforcing the idea that decreased grazing is not the main driver of increases in biomass production through time.

Another potential explanation for increased grass biomass could be atmospheric nitrogen pollution from fossil fuel combustion. Globally, nitrogen deposition has alleviated nitrogen limitation to favour grass biomass production²⁴, and nitrogen deposition from coal-fired power stations in South Africa has the potential to reach Kruger. However, nitrogen derived from coal carries a characteristic stable isotopic signature (δ¹⁵N), and an analysis of δ¹⁵N in herbarium specimens collected from Kruger over the late twentieth and twenty-first centuries shows no evidence of this signal²⁵ (Supplementary Fig. 3). If anything, carbon-to-nitrogen ratios appear to have increased, consistent with progressive nitrogen limitation as grass biomass production has increased. Kruger may simply not experience high nitrogen deposition by global standards²⁴.

After ruling out these alternatives, it appears likely that rising grass production in semi-arid Kruger savannas reflects a response to the 65 ppm increase in atmospheric CO₂ levels during the past three decades²⁶, as predicted under water-limited conditions in our experimental meta-analysis. Across all sites, grass biomass measured in the first year after fire increased by 75.1 g m⁻² across 32 years (95% confidence interval (CI) of the mean 74.5–75.8 g m⁻²), representing a 28% increase in annual grass production. Although the absolute magnitude of plot-scale biomass increases was invariant to rainfall, drier savannas experienced larger relative biomass increases than wetter savannas (logRR-CO₂ of 0.270 (95% CI 0.265–0.274) at 475 mm mean annual rainfall (MAR) versus 0.184 (0.178–0.190) at 625 mm MAR; Fig. 2g and Supplementary Tables 1 and 2), again consistent with predictions based on experiments of CO₂ stimulation of photosynthesis in water-limited conditions.

Changes in grass biomass differed across the two dominant soil types in Kruger, with larger increases on nutrient-rich, clay soils derived from basalt than on nutrient-poor, sandy soils derived from granite (117.0 g m⁻² (95% CI 108.3–125.7 g m⁻²) versus 52.0 g m⁻² (95% CI 45.0–59.1 g m⁻²) in the first year after fire; logRR-CO₂ = 0.33 (95% CI 0.28–0.37) versus 0.20 (95% CI 0.15–0.25); Fig. 2g and Extended Data Fig. 5). One possibility is that the fertility of the basalt-derived clays in Kruger may release grasses from soil nutrient limitation in the face of CO₂ fertilization²⁷. Alternatively, soil texture may change plant available moisture via differences in field capacity and wilting point. In a field experiment in a prairie ecosystem, grass biomass on clay soils, with higher field capacity, showed a larger CO₂ response than on sandy soils²⁸ (but see also ref. ²⁹).

Next, we considered potential contributions of shifts in species abundance within communities to observed biomass increases, as past work has shown that long-term experimental exposure of mixed C₃–C₄ prairies in North America to CO₂ enrichment can cause community turnover²⁹. To account for functional and composition turnover in Kruger, we incorporated data on abundance and productivity-related traits (specific leaf area and plant height) of the five most common grass species in the park. Taller, more productive grasses (for example, Megathyrsus maximus) were increasingly dominant through time, a process which in itself may be partly determined by rising levels of CO₂ (ref. ²⁹). Statistically accounting for turnover in species and functional traits reduced, but did not eliminate, the observed grass biomass increase (to 66.7 g m⁻² (95% CI 66.0–67.4 g m⁻²) controlling for species composition and 63.2 g m⁻² (95% CI 62.2–64.2 g m⁻²) controlling for functional turnover in the first year after fire; Extended Data Fig. 5 and Supplementary Table 3).

Overall, the potential effect size of CO₂ fertilization in the field was smaller than water-limited CO₂ responses in the meta-analysis (Fig. 2g). On the one hand, this is expected as experimental treatments included much larger differences in CO₂ (180–400 ppm over ambient levels compared with an increase of only 65 ppm across 32 years in Kruger); on the other hand, a saturating response to future increases in CO₂ levels is anticipated due to non-linear relationships between plant photosynthesis and atmospheric CO₂ (although the point of saturation may differ depending on, for example, soil nutrient availability²⁹). A more informative direct comparison can be made with a North American experiment that provided a continuous CO₂ gradient of 200–500 ppm to four mixed C₃–C₄ prairie communities on three different soil types, yielding an increase of 38–58 g m⁻² of aboveground annual net primary productivity (ANPP; 95% CI 26–62 g m⁻²)²⁹. For the equivalent CO₂ level increase on sandy versus clay-rich soils, 52 and 117 g m⁻² of grass biomass, respectively, accumulated in the year after fire in Kruger (a reasonable estimate for ANPP), which overlapped but exceeded the range observed across experimental prairie communities on similar variation in soil texture at similar rainfall²⁹. Direct comparisons can also be made with trends inferred from satellite observations for C₄-dominated ecosystems¹, which, scaled to the full 32-year span of our dataset, yielded a closer match to our estimates, at 80 g m⁻² of ANPP (95% CI 3–175 g m⁻²; Methods).

We next used CLM5 to predict future ecosystem trajectories. We evaluated future CO₂ effects against experimental meta-analysis results, factorially switching each driver off to isolate the modelled effects of CO₂ versus climate change, focusing on CO₂-only runs for experimental validation (Extended Data Fig. 7a). Under future elevated levels of CO₂, modelled photosynthesis and biomass production increased (see Fig. 3a for SSP3-7.0; Extended Data Fig. 8a–d). The magnitude of increase was larger at both sites than the experimental results but particularly at the semi-arid site, consistent with our findings that water limitation drives CO₂ fertilization of photosynthesis and aboveground biomass (Fig. 3a; RR_{north/semi-arid} = 0.42 and RR_south/mesic = 0.38 for photosynthesis; RR_{north/semi-arid} = 0.47 and RR_south/mesic = 0.43 for aboveground biomass across approximately 700 ppm of CO₂ under SSP3-7.0 from 2015–2075). The model predicted a strong increase in root biomass in response to increasing levels of CO₂, which contrasts with the highly variable response of C₄ grass root biomass to elevated levels of CO₂ found in our experimental synthesis and in grassland FACE experiments (for example, logRRs ranging from −0.4 to +0.4 in ref. ³⁵; Fig. 3a and Extended Data Fig. 8c), suggesting that the model overestimates future grass root biomass and associated soil carbon storage. Modelled effects of CO₂ on plant water status (Fig. 3b and Extended Data Fig. 9a–c) suggest contrasts with both our synthesis (Fig. 1b,c) and with FACE experiments³⁶, predicting consistently reduced stomatal conductance instead of the observed sustained high stomatal conductance under drought with CO₂ enrichment. The difference implies that refinements to the CLM representation of stomatal and hydraulic regulation under elevated levels of CO₂ and water limitation on productivity may be beneficial, particularly as this is an active area of physiological research^37,38.

However, overall, comparison with both historical observations and experimental meta-analysis suggests that CLM5 is able to phenomenologically capture C₄ grass responses to CO₂, indicating that model projections could be informative of future C₄ grass production. Simulations showed a clear continued increase in C₄ grass production under increased levels of CO₂ but also showed that climate change reduced, but did not eliminate, CO₂ fertilization of C₄ grass biomass into the future (Fig. 3c and Extended Data Fig. 8b). Modelled climate effects—showing lower levels of photosynthesis and productivity with higher air temperature and lower levels of rainfall—are consistent with observations from Kruger field data (Fig. 2d). In the model, these outcomes resulted from lower stomatal conductance and associated higher leaf temperature (Extended Data Fig. 7b), which depresses photosynthetic capacity once it exceeds 36.3 °C (Supplementary Fig. 4). This model response concords with past work showing that photosynthesis in C₄ grasses declines at such high temperatures through a number of mechanisms^39,40,41,42. The combination of CO₂ and climate effects on productivity was larger in the high-warming (SSP3-7.0) than in the low-warming (SSP1-2.6) scenario, suggesting that temperature suppresses productivity but not enough to eliminate CO₂ fertilization (Fig. 3c). Together, the combination of rising CO₂ levels and climate change is projected to result in increases in C₄ grass production of between 0.01 and 0.05 kg C m⁻² by 2100 (representing an increase of 6–21% from 2015 levels; Extended Data Fig. 8b).

In summary, across 70 experiments and 32 years of in situ field observations, we found consistent evidence that higher CO₂ levels increase C₄ grass biomass production under water limitation. In experimental settings, increased productivity was linked to decreased water use and improved plant water status. This suggests that the CO₂ fertilization that we observed in the field is principally driven by reduced plant water deficits, which could extend growth following rainfall events and into the dry season⁴⁴, rather than by a direct stimulation of Rubisco activity. CLM5 simulations suggest that observed productivity increases in situ could continue through the twenty-first century, exceeding suppressive effects of increasing temperatures and aridity.

Overlooked C₄ grass CO₂ fertilization has major implications for savanna ecosystem function. In forests, CO₂ fertilization has been linked to carbon storage⁴. In C₄ grasslands, carbon is stored long term primarily in soil and in coarse woody roots^45,46, as aboveground carbon is vulnerable to herbivory and fire⁴⁷. Most soil organic carbon is derived from grass root litter⁴⁸, which did not increase in experiments, suggesting that responses of soil organic carbon due to CO₂ effects on grasses could be negligible. However, CO₂ fertilization could alter forage quality (higher carbon-to-nitrogen ratio and lower protein levels)⁴⁹, changing animal consumption, litter decomposability and conversion to recalcitrant soil carbon. CO₂ fertilization could also change dynamics of tree–grass interactions, amplifying C₃ woody plant encroachment and losses of C₄ grass cover⁵⁰. Interactions with soil water and fire⁶ make this difficult to predict, but stronger CO₂ fertilization of C₄ grasses in drier savannas could help to explain their slower rates of woody plant encroachment⁵¹. Together, these potential mechanisms result in highly uncertain responses of savanna carbon to increasing CO₂, making them a prime target for future work. Overall, these results combine with past work to demonstrate CO₂ fertilization in both trees and grasses across the tropics⁴ and to highlight the importance of direct CO₂ effects on past and future vegetation trajectories.

Meta-analysis

Search methodology

Suitable studies were identified through a systematic literature search. To create a comprehensive search string, a naive Boolean search string of (‘elevated CO₂’ or ‘high CO₂’) and ‘C₄’ and (‘grass*’ or ‘Poaceae’) was used initially in ISI Web of Science, resulting in 186 results (4 March 2021). The citation record and abstracts of all results were then exported for keyword co-occurrence network analysis using the R package Litsearchr⁵². This generated the final Boolean search string of (‘C₄’ or ‘C-4’) and (‘grass*’) and (‘elevat*’ or ‘increas*’ or ‘enrich*’) and (‘carbon dioxide’ or ‘CO₂’) and (‘biomass’ or ‘growth’ or ‘product*’ or ‘respons*’).

The final search terms were used in ISI Web of Science and SCOPUS, resulting in 768 unique results (March 2021). Google Scholar was not used because it does not recognize wildcard components (*) and a search without these components produced too many results (more than 36,000). Several e-theses repositories (White Rose repository, British Library EThOS and South African ETD Collections) were also queried, resulting in two further suitable studies.

Inclusion and exclusion criteria

The studies were screened by title and abstract and were selected if they investigated the responses of wild, C₄ grass species of grass-dominated ecosystems (grasslands and savannas) to elevated [CO₂]. Therefore, studies on non-grass, C₃ and domesticated species were excluded, as well as those from other ecosystems (for example, agricultural weeds and wetland species such as Spartina patens). We analysed data using a phylogenetically controlled method (see the ‘Modelling’ section below) so responses needed to be identified to the species level. Therefore, studies that did not isolate responses of particular species (for example, grassland community responses) were excluded.

For inclusion, studies had to include both ambient [CO₂] (aCO₂) and elevated [CO₂] (eCO₂) treatments. A range of [CO₂] has been used in both treatments, reflecting increasing atmospheric [CO₂] during the approximately 40-year timeframe of these studies and changing predictions of near-future [CO₂]. Ambient treatments used [CO₂] less than or equal to 450 ppm. [eCO₂] treatments ranged from 460 to 1,005 ppm. To investigate how C₄ grass responses to eCO₂ interacted with plant water status, studies that also included treatments relating to water availability (in addition to [CO₂] treatments) were included.

Studies had to report at least one of the following metrics: aboveground biomass, belowground biomass, total biomass, stomatal conductance measured at the growth [CO₂] (G_s) or light-saturated net CO₂ assimilation rates measured at the growth [CO₂] (A).

Data extraction

Grass responses at aCO₂ or eCO₂ were extracted either from the main body of the text, tables or Supplementary Information, or manually digitized from figures using WebPlotDigitizer⁵³. When data were not available, we contacted authors directly to request data.

For each species within each study, we extracted the following information: the two [CO₂] values, the mean value of the variable of interest (biomass, G_s or A) under aCO₂ and eCO₂, the sample size and the standard error (s.e.; either as presented or calculated from the standard deviation and sample size). Occasionally, when multiple data points were reported for one species in a study (for example, measurements were taken over several years), data were recorded separately but linked by a study identification number.

In studies in which both CO₂ and water availability treatments were applied, the CO₂ response was collected separately by water treatment level. ‘Low’ water availability treatments involved a reduction, relative to ‘high’ water availability treatments, in the amount of water supplied to plants, either due to the restriction or withholding of supplementary watering, an experimental reduction of rainfall (by using rain-out shelters, for example), or the occurrence of a low-rainfall event (for example, a growing season with below-average rainfall). If additional environmental stresses (such as salinity or ozone treatments) were imposed factorially, only the CO₂ response at the ambient, non-stressful level of this other factor was used.

To account for other factors that could influence grass responses to eCO₂, metadata describing the experimental methodology of each study were recorded. These included pot size (categorized as less than 10 l pots, more than 10 l pots or if the plants were in the ground), the length of exposure to CO₂ treatments (from treatment initiation to plant trait measurement; categorized as less than 60 days, 60–120 days or more than 120 days), and the exposure methodology. The latter consisted of four categories: FACE, open-top chamber, indoor-controlled environment growth chamber or outdoor enclosed greenhouse.

The final dataset for analysis consisted of 655 pairs of observations under aCO₂ and eCO₂, of which 550 did not limit water availability and 105 were droughted, on 82 C₄ grass species from 70 studies^{43,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122}. The species represented 42 out of the 314 genera containing C₄ grass species, all 4 grass subfamilies where C₄ photosynthesis has evolved (Aristoideae, Chloridoideae, Micrairoideae and Panicoideae), and the 3 C₄ subtypes (NAD-ME, NADP-ME and PCK). These species are found across broad temperature and rainfall ranges, covering a significant proportion of climatic space from temperate to tropical biomes (Supplementary Fig. 5a). The dataset had global coverage, with data from C₄ grass species from Africa (15 studies), Asia (4 studies), North America (34 studies), Oceania (9 studies) and South America (8 studies). Twelve of these 70 studies were also included in a previous analysis by Wand et al.¹⁶ of C₃ and C₄ grass responses to eCO₂ (refs. ^{70,72,75,88,90,99,106,110,113,117,118,122}). The other 51 studies utilized in this previous meta-analysis¹⁶ were not included here because they investigated C₃ (33 studies) or non-grassland species (wetland or agricultural land species; 7 studies), did not report any variables of interest here (for example, a focus on leaf traits; 6 studies), reported community level responses (2 studies) or were not accessible (3 studies).

Effect size calculations

Log-transformed response ratios (RR) were used to calculate the size of CO₂ effects on grass traits, a common approach in ecological meta-analyses when comparing a control and a treatment group¹²³. RRs were calculated as:

where \({\overline{x}}_{T}\) is the measured value for the treatment group (the eCO₂ group) and \({\overline{x}}_{C}\) is the measured value for the control group (the aCO₂ group). A negative RR (RR < 0) represents a decrease in the measured variable under eCO₂ (relative to aCO₂), and a positive RR (RR > 0) represents an increase in the measured variable under eCO₂.

s.e. values were imputed for the small number of studies that did not report s.e. or standard deviations. Missing values were imputed using Bracken’s approach¹²⁴, which uses the mean variable value and an average s.e.-to-mean ratio for that CO₂ treatment.

When pooling RR from multiple studies, each RR should be weighted by its precision to give less weight to studies with high s.e. and vice versa. Weights were calculated as inversely proportional to the sampling variance of a RR observation¹²³ as

where s.e._C and s.e._T are the s.e. for the control (aCO₂) and treatment (eCO₂) groups, respectively.

Modelling

Linear mixed effects models were fitted using the ‘MCMCglmm’ function (MCMCglmm package)¹²⁵ in the R language and environment¹²⁶. This approach implements Markov chain Monte Carlo (MCMC) routines for fitting generalized linear mixed models, while accounting for non-independence and correlated random effects arising from phylogenetic relationships. To do this, the dataset was linked to a completely sampled and dated Bayesian phylogeny that incorporated 11,297 grass taxa¹²⁷ pruned to only include species in this study.

For each of the five grass metrics, a linear mixed effects model was fitted with RR as the response variable and ‘water stress’ (either none or drought), as the main fixed effect of interest. To account for the effect of differences in experimental methodology that may influence RR, additional fixed effects were fitted to account for ‘CO₂ treatment size’ (the difference in [CO₂] between aCO₂ and eCO₂ treatments), ‘pot size’ (the volume of soil the plant was grown in) and ‘exposure duration’ (to CO₂ treatments). Exposure methodology was not included as an additional fixed effect because it tended to be correlated with exposure duration and pot size. For example, in the aboveground biomass data, all plants in FACE experiments were grown in the ground with a long exposure to eCO₂ (mean duration = 301 days), whereas plants grown in indoor growth chambers were exposed for shorter durations (mean duration = 89 days) and were grown in pots (45% in pots less than 10 l). ‘Species’ and ‘study’ (an identifier for the study) were included as random effects. Observation-level sampling variance was included as mev = s.e.², because MCMCglmm defines mev as the known measurement-error variance for each effect size. This separates within-study sampling error from between-study heterogeneity, so more precise estimates are weighted more strongly, whereas the random effects capture residual among-study variation. Models were run for 1,000,000 iterations with a burn-in of 1,000 iterations, and parameter-expanded priors for the random effects (V = diag(1), nu = 1, alpha.mu = 0 and alpha.V = diag(1) × 1,000). Convergence of models was tested using the Heidelberger convergence diagnostic test (‘heidel.diag’ function in the Coda package¹²⁸). The Emmeans package¹²⁹ was then used to produce estimated marginal mean response ratios and their 95% confidence intervals from MCMCglmm model outputs.

To ensure our approach of accounting for CO₂ treatment size effects on trait responses described above (that is, including CO₂ treatment size as a fixed effect in models) was robust, an alternative, complementary approach was taken. A different metric (‘relativized β-factor’), proposed by Walker et al. (equation (1) in ref. ¹³⁰), that standardizes CO₂ effects on traits by the change in CO₂ concentration was used as an alternative to RRs. For this metric, β = 1 represents direct proportionality between a variable’s CO₂ response and the change in CO₂. MCMCglmm models were run for all five plant traits as described above, except the response variable was the relativized β-factor (rather than RR) and CO₂ treatment size was removed as an explanatory variable. The results were similar to those gained from the RR approach (Supplementary Fig. 6), and so we only describe the RR model outputs in the main text for ease of comparison with the field and modelling components of this paper.

Publication bias

Publication bias was assessed visually through funnel plots of the inverse variance and effect size residuals (Supplementary Fig. 7). There was little potential for bias across the whole dataset, and both positive and negative RRs have been well represented in the data.

CO₂ effects on plant water relations

To further explore CO₂ effects on plant water relations under different water availabilities, values of leaf water potential and soil water content were extracted from eight studies identified in the meta-analysis^{43,57,65,69,75,80,95,97} (study details given in Supplementary Table 6) that measured these traits under different CO₂ and water-availability treatments. The low-water-availability treatments of these studies involved reductions in soil water contents ranging from 23% to 83%, or (when soil water content information was not given) a reduction in water supply of 33–99%, relative to the high-water-availability treatment (see Supplementary Table 6 for details). This dataset contained 41 paired leaf water potential measurements (n = 82 total; 38 control and 44 drought) and 90 paired soil water content measurements (n = 180; 88 control and 92 drought) at aCO₂ and eCO₂ for 11 C₄ grass species. MCMCglmm models were fitted to these data with CO₂ and water treatment (and their interaction if significant) as explanatory variables. All other model specifications were the same as above except ‘exposure duration’ (to CO₂ treatment) was not included as an explanatory variable due to a bias in the proportion of observations (that is, 62% and 78% of soil water content and leaf water potential measurements) arising from experiments with a long exposure (more than 120 days) to CO₂ treatments in this subset of studies. No observation-level sampling variance was included in the soil water content model as the error on these measurements was rarely provided. The extracted model coefficients are plotted in Fig. 1b.

For insights into underlying physiological mechanisms, we explored how interrelated plant water relations parameters changed in relation to CO₂ and water using a hydraulic–stomatal modelling approach. Using stomatal and hydraulic parameters derived for a widespread, C₄ savanna grass, Themeda triandra¹³¹ (Supplementary Table 7), the responses of stomatal conductance (G_s) and leaf water potential (Ψ_leaf) to declining soil water potential (Ψ_soil) and different CO₂ treatments (aCO₂ versus eCO₂) were modelled. Three equations were used to simulate these responses (following ref. ¹³²). Equation (1) determines Ψ_leaf from plant hydraulic conductance (K_plant) and G_s at a given Ψ_soil and vapour pressure deficit (VPD) based on steady-state water transport according to the Ohm’s law analogy¹³³,

Equation (2) models the vulnerability of K_plant to declining Ψ_leaf as the sigmoidal function given by

where K_max is the maximum leaf hydraulic conductivity under well-watered conditions, and P_K50 is Ψ_leaf at 50% loss of hydraulic conductivity in leaves. The constant c defines the shape of the sigmoidal curve. Equation (3) models the decline in G_s with more negative Ψ_leaf as the sigmoidal function

where \({G}_{{s}_{\text{max}}}\) is the maximum light-saturated stomatal conductance under well-watered conditions, P_Gs50 is the Ψ_leaf at 50% stomatal closure, and c is a constant that defines the shape of the sigmoidal curve.

The responses of G_s and Ψ_leaf were simulated for a range of Ψ_soil (from 0 to −5 MPa), at a VPD of 2.5 kPa (chosen to represent a high VPD scenario typical of savanna ecosystems), and under both aCO₂ and eCO₂ by simultaneously solving equations (1)–(3) in base R. The modelled relationships are displayed in Fig. 1c, along with estimated marginal means of leaf water potential and stomatal conductance under different CO₂ and water-availability treatments (derived from studies listed in Supplementary Table 6).

Field observations

Study area

Kruger covers nearly 20,000 km² in low-elevation areas (260–839 m; the ‘lowveld’) of northeastern South Africa, spanning tropical and subtropical latitudes (22° 20′ to 25° 30′ S; 31° 10′ to 32° 00′ E; Supplementary Fig. 1). It is dominated by two underlying parent materials, a granite and a basalt, producing soils that are broadly characterized as sandy and nutrient poor versus clay and nutrient rich, respectively¹³. Mean annual rainfall ranges from 350 mm in the north to 750 mm in the south, although inter-annual variation is significant. Park management continuously maintains 21 daily rainfall measurement stations throughout the park and one daily temperature measurement station is maintained by South African Weather Services. The flora of Kruger includes 400 or more species of tree and shrub and 200 or more species of grass.

Fire is a major ecological feature of the park; the average fire return interval is about 3.5 years, but fire regimes vary locally. Fire frequencies range from one fire per year to one every 34 years. The park is also host to a diverse assemblage of African mammals. Elephants (Loxodonta africana) occur at a density of 0.7 km⁻² and make up the largest propotion of herbivore biomass (1,900 kg km⁻²). Herbivore biomass totals approximately 6,080 kg km⁻² (10 large stock units (LSU) km⁻²). Because spatially resolved herbivore density estimates are not available, we have not included this explicitly in this analysis, but note that herbivore population sizes have increased significantly in the past several decades (Supplementary Fig. 2) and therefore cannot be responsible for any increases in grass biomass (but see ref. ¹³⁴ for short-term grass–grazer interactions). However, rhino poaching has locally resulted in shifts from grazing lawn to bunchgrass, which we examine via changes in grass species composition (see below).

Grass biomass data

Kruger management established 533 veld condition assessment (VCA) sites (343 on granitic and 190 on basaltic soils) to monitor grass biomass to inform fire management, starting in 1989 (Supplementary Fig. 1). Sites were thus located primarily in areas where herbivore impacts are relatively mild (that is, sites were not located in areas dominated by grazing lawns, such as sodic sites and near rivers, where fires are rare). Grass biomass was measured every year in February or March, which corresponds to the end of the growing season and therefore peak standing grass biomass, in plots of 50 m × 60 m with a calibrated disc pasture meter (DPM). Measurements to derive plot-level grass biomass estimates were taken every 2 m along four 50-m transects, running at 0 m, 20 m, 40 m and 60 m along the length of the plot, for a total of 104 DPM estimates per plot. A local DPM calibration is available from Trollope and Potgieter (1986)¹³⁵ (Supplementary Fig. 8), showing reasonable accuracy at the rate of DPM sampling performed here. The most recent grass data available for this analysis were collected in 2021. The subset of plots that were actually sampled for grass biomass and community composition each year varied substantially in number and identity; plot sampling intensity has decreased over time to approximately 150 per year in 2020, reflecting a decrease in investment in monitoring by park managers.

Grass species were identified and recorded. From 1989 through 2006, grass identifications were restricted to 21 indicator species, termed ‘key’ species, with additional categories for ‘other grasses’, ‘forbs’ and ‘bare ground’. Species composition was not recorded in 2007 or 2010–2012. In 2008–2009, grass identification reverted to the original methodology, but with the addition of 9 key species (for a total of 30). Starting in 2016, all grass species were identified, and a category for ‘sedges’ was added. See ref. ¹³⁶ for comprehensive descriptions of ‘key species’ and species-composition methodology. Here, to aid comparison across years, we have degraded more recent community composition sampling to the taxonomic resolution of the original 1989–2006 dataset, focusing our analysis on changes through time in the relative abundances of the five most common key species that were recorded throughout the sampling period: Bothriochloa radicans, Digitaria eriantha, Megathyrsus maximus, Themeda triandra and Urochloa mosambicensis.

Across the VCA plots, 94 C₄ grass species have been identified, representing 43 out of the 314 genera that contain C₄ species, and 3 out of the 4 grass subfamilies where C₄ photosynthesis has evolved (Aristoideae, Chloridoideae, Panicoideae but not Micrairoideae, which is concentrated in Australia and Asia). All three C₄ subtypes are present in approximately equal proportions (in 2021, 32%, 39% and 29% of individual species observations were NAD-ME, NADP-ME and PCK subtypes, respectively). Species are broadly representative of those found in arid-to-mesic tropical and subtropical savannas but not high-rainfall savannas (Supplementary Fig. 5b).

Rainfall and fire occurrence

Since 1988, park management has maintained 21 daily rainfall measurement stations throughout the park, which we used to generate rainfall surfaces via inverse distance weighting for extraction and comparison with grass biomass and fire occurrences at monitoring sites. Of these stations, three have been continuously maintained since 1960, which were used to estimate rainfall for comparison with parkwide burned area. Maximum daily temperature has been measured at one station across both time periods. Averaging across the entire park, there has been no systematic change through time in rainfall, number of days with rainfall or rainfall per rainfall day, but mean annual daily maximum temperature increased between 1989 and 2021 (Extended Data Fig. 4).

Kruger management also maintains records of the spatial distribution of fires throughout the park, whether accidental or set by management; older records were kept by hand and have been digitized, whereas more recent records were based on fire-scar mapping from MODIS satellite-derived data (accurate to 250 m).

Grass functional trait data

This dataset includes maximum culm height (hereafter, ‘height’), specific leaf area (SLA), leaf carbon-to-nitrogen ratio, leaf nitrogen per area, leaf nitrogen per mass and stomatal conductance (G_s). Species-level data on height were extracted from ref. ¹³⁷, and for the other traits from the TRY database^{33,138,139,140,141,142,143,144,145,146,147,148,149,150,151,152,153,154,155,156,157,158,159,160,161,162,163,164,165,166}. For each trait, we calculated a community mean trait value weighted by relative abundance (community weighted mean) for all plots with more than 80% of species relative abundance represented. Trait coverage was sufficient to calculate community weighted means for more than 50% of plots for only height and SLA, so we restricted our analysis to those traits.

Data analysis

Except where noted, statistical analyses were performed on all data except from the year 2000, when record rainfall resulted in widespread flooding (mostly from surface runoff) and thus a vegetation response that is uncoupled from total rainfall amount. Data were analysed via linear mixed effects models using the R ‘lmer’ function in ‘lme4’¹⁶⁷ with plot as a random effect, and effect sizes were estimated in the ‘emmeans’ package¹²⁹.

Model selection was done via minimization of Akaike Information Criterion corrected for small sample sizes (AICc); we selected as ‘best’ the simplest model within two AICc units of the model with the lowest overall AICc. We tested four different models for grass biomass: (1) grass biomass dependence on soil type, annual rainfall, previous year annual rainfall, rainfall events (in number of days), annual mean daily temperature, time since fire, and year, on a data subset including all grass observations; (2) the same set of factors as (1) but on a data subset excluding sites with more than 20% relative abundance of three common lawn grass species (Urochloa mosambicensis, Digitaria eriantha and Megathyrsus coloratum) to exclude the possibility that local loss of large grazers such as rhinos could be driving temporal change; (3) the same set of factors as (1) and also including relative abundances of the five most common species across our dataset (Urochloa mosambicensis, Bothriochloa radicans, Digitaria eriantha, Themeda triandra and Megathyrsus maximus) to disentangle the role of community turnover in producing changes in grass biomass; and finally (4) the same set of factors as (1) and also including community weighted means of specific leaf area and individual height to further disentangle the role of trait turnover in creating changes in grassy biomass production. We considered only additive combinations of factors but tested the significance of interactions between year and rainfall and between year and soil type, to evaluate whether possible increases in grass biomass have varied systematically depending on environmental context. In addition, we performed a sensitivity analysis examining the potential effects of DPM-associated errors in estimated grass biomass by randomly sampling observation errors from the DPM calibration and repeating this process 100 times; all runs yielded the same best statistical model as the non-randomized analysis, showing that statistical results were robust to DPM sampling errors (Supplementary Fig. 8). We also evaluated possible model sensitivity to decreases in sampling intensity through time by randomly sampling the full dataset to standardize effort through time; this process also yielded the same best statistical model as the statistical model on the full dataset, with similar effect sizes.

We examined drivers of variation in fire activity in two ways. First, we modelled fire occurrences (yes or no) at each of the VCA sites over time, which we modelled via binomial mixed effects models with ‘glmer’ (lme4 (ref. ¹⁶⁷)), using site as a random effect, in two ways: (1) fire occurrence depending on rainfall, previous year rainfall, rainfall events, temperature and year; and (2) fire occurrence depending on grass biomass and year. Second, we modelled parkwide proportional burned area through time using a linear model with respect to variation in annual rainfall, previous year rainfall, rainfall events, temperature and year.

Comparison of magnitude of CO₂ effect with published estimates

Estimates of CO₂-driven increases in aboveground grass productivity have been compared with published estimates from refs. ^1,29.

In ref. ²⁹, grasses were subjected to a gradient of CO₂ concentrations from 250 to 500 ppm, allowing us to estimate the difference in productivity between [CO₂] of 350 ppm (equivalent to 1989 levels) and 415 ppm (equivalent to 2021 levels). We digitized the data from figure 2a–c in ref. ²⁹, and fitted new linear models to estimate increases for all three soil types included in that work. Confidence intervals were estimated from modelled increases spanning all three soil types.

In ref. ¹, we used only estimates that were directly based on satellite-informed estimates of gross primary production (GPP), avoiding estimates that made strong theoretical assumptions about physiological optimality, which generally have been developed for C₃ plants. To do this, we digitized supplementary figure 5b, subpanel ‘C₄ biomes’, focusing on satellite-derived estimates corresponding with bar ‘C₁’. Confidence intervals were estimated by digitizing estimates from all remote sensing products. To convert GPP carbon into biomass ANPP, we assumed an average carbon-use efficiency of 0.48 for savannas in the conversion of photosynthate to biomass¹⁶⁸ and a carbon concentration for grass of 0.42 (from ref. ¹⁶⁹).

Land surface and vegetation modelling

To simulate land surface processes in African savannas, we used the CLM5 (ref. ⁶), which is the terrestrial component of the Community Earth System Model version 2 (CESM2; ref. ¹⁷⁰). We used the standard download of the CLM5 model, available publicly on GitHub (https://github.com/escomp/ctsm), and ran the configuration of this model recommended by its developers. CLM5 represents terrestrial carbon and nitrogen biogeochemistry, vertically resolved soil carbon, and nitrification–denitrification to treat the response of vegetation to future climate and fires. The CLM5 land units are vegetated, lake, urban, glacier and crop. Vegetation and crops are represented by plant functional types (PFTs), each with its own set of ecophysiological, morphological, phenological and biogeochemical parameters. The default PFT distribution of natural vegetation and crops was derived from satellite observations (for example, MODIS) and agricultural census data^171,172. There are 16 types of natural vegetation (including C₄ grasses) and 8 active crops⁶.

For C₄ grasses, photosynthesis follows the Collatz et al. (1992) model³⁰, which calculates net photosynthesis based on Rubisco and phosphoenolpyruvate (PEP) carboxylase-limited rates of carboxylation, assuming a sharply saturating response to intercellular CO₂ concentrations and including limitations from light availability and temperature⁶. The temperature optimum for C₄ productivity, set via the maximum carboxylation rate of Rubisco (V_Cmax), is 36.3 °C (ref. ¹⁷³) (Supplementary Fig. 21). Stomatal conductance for CO₂ and water is calculated using the Medlyn et al. (2011) model³², which assumes a linear increase with net leaf photosynthesis and non-linear declines with VPD and aCO₂ concentration. This framework is applied to both C₃ and C₄ plants. For C₄ grasses, CLM5 uses empirical parameter values for the Medlyn slope parameter (g1) derived from the global synthesis of Lin et al.³³, following implementation tests in CABLE by de Kauwe et al.¹⁷⁴. Both photosynthesis and stomatal conductance are calculated for sunlit and shaded leaves to capture their distinct microenvironments.

Carbon uptake by C₄ grasses is driven by prescribed photosynthetic parameters (for example, V_pmax (the maximum catalytic rate of PEP carboxylase, V_Cmax and J_max (the maximum rate of photosynthetic electron transport)). These parameters do not adjust dynamically in response to CO₂ or nitrogen⁶. However, carbon–nitrogen interactions are represented through fixed carbon-to-nitrogen ratios for plant tissues, and nitrogen uptake is regulated by the fixation and uptake of nitrogen (FUN) model¹⁷⁵, which assigns carbon costs to nitrogen acquisition from soil and internal remobilization. Nitrogen limitation can constrain tissue growth (for example, leaf area) and thereby indirectly influence canopy-scale uptake under eCO₂.

Carbon allocation in CLM5 follows a hierarchical priority scheme, in which carbon is first used for maintenance respiration, then stored for future needs, and finally allocated to new growth. For grasses, including C₄ grasses, carbon allocation is primarily determined by fixed allometric parameters that set the proportion of carbon allocated to leaves, stems and roots. Specifically, fine root allocation is set as a fixed ratio relative to leaf biomass, whereas coarse root allocation follows a predefined proportion relative to stem biomass. Although these allocation ratios remain fixed, environmental factors such as CO₂ concentration, temperature, soil moisture and nutrient availability can influence photosynthesis and overall carbon uptake, indirectly affecting carbon accumulation.

The water–carbon coupling in CLM5 is managed by a plant hydraulic stress approach¹⁷⁶, which models water transport through plant tissues and calculates water potentials for roots, xylem and leaves. This balances water supply and demand by modulating maximum stomatal conductance to prevent high xylem tension and low leaf water potential. Soil water limitation effects on photosynthesis are quantified using a diagnostic soil moisture stress factor, which assesses the influence of plant water status on carbon cycling processes. A more detailed explanation with specific formulations and parameterizations is provided in refs. ^6,31.

CLM5 also includes a prognostic treatment of fires based on a modified version of the fire algorithm from refs. ^177,178. The fire algorithm accounts for agricultural, deforestation, peat and landscape fires, and estimates area burned and fire emissions using information about climate and weather conditions, vegetation composition and structure, and human activity. After the burned area is determined, the fire module calculates the impact of fire, including biomass and peat carbon losses, fire-induced vegetation mortality, adjustment of the vegetation carbon-to-nitrogen pools, and total fire carbon and other trace gas emissions.

Historical simulations

To examine the C₄ grass sensitivity in the model to climate conditions, we performed single-point simulations at Kruger. We selected two sites with different rainfall conditions: a drier, semi-arid site in the north (22.83° S, 31.25° E; average 420 mm rainfall per year) and a wetter, more mesic site at the southern end of the park (25.12° S, 31.88° E; average 620 mm rainfall per year). At both sites, we converted all PFTs within the 50 × 50 km CLM5 grid cell to C₄ grass, to focus specifically on C₄ grass responses and avoid effects from other PFTs (for example, via competition and succession processes), and spun up the model for about 600 years so that all the state variables in the model, including total ecosystem soil carbon, reached equilibrium. The present-day spin-up was based on a historical simulation for 1850–2014 using historical nitrogen and aerosol deposition as well as aCO₂ forcing⁶. The meteorological forcings were from the Global Soil Wetness Project (GSWP3 version 1; http://hydro.iis.u-tokyo.ac.jp/GSWP3/, last access: 15 November 2023), with forcing data available from 1901 to 2014 and cycled from 1901 to 1920 for years before 1901.

Future simulations

For our future sensitivity modelling experiment, and following the historical simulations, the Kruger simulations were run from 2015 to 2100 with changes in CO₂ concentration and climate following the SSP1-2.6 and SSP3-7.0, representing low and high future warming scenarios, respectively. Following standard practice for offline CLM5 simulations, we applied an anomaly forcing approach^6,179. Atmospheric forcing variables (for example, temperature, precipitation, wind, pressure, humidity and radiation) from the GSWP3 observations (2001–2014) were used as a baseline, onto which monthly mean anomalies from CESM2 fully coupled simulations performed for the Coupled Model Intercomparison Project phase 6 experiments were superimposed. This approach preserved observed high-frequency variability while imposing long-term climate trends consistent with the Coupled Model Intercomparison Project phase 6 SSP projections.

For both scenarios, simulations were performed by changing only CO₂ and holding all other forcings at 2015 levels (CO₂ only); only changing climate and holding all other forcings at 2015 levels (climate only); and changing all forcings (CO₂ + climate; see Extended Data Fig. 7a for driving variables). For each run, we extracted simulated photosynthesis, aboveground and belowground biomass and total biomass, stomatal conductance, leaf water potential, soil moisture, leaf temperature and total fire carbon emissions from fires at each field site (Extended Data Figs. 8 and 9), which represent specific C₄ grass PFT conditions.

Validation

CLM5 simulations have been extensively evaluated against satellite and tower flux observations, on a global scale, across Africa and including grasslands^180,181. For this study, we evaluated aboveground biomass model predictions against field observations collected at Kruger for 1989–2014 at sites within each modelled grid square for local validation. We plotted all individual site observations as well as annual means against model predictions for the time period 1989–2014 (Extended Data Fig. 6). This was done for simulations both where CO₂ concentrations increased during this time period and where they were held at 1989 levels (Extended Data Fig. 6b,c).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

The data supporting the findings of this study are available from the Dryad Digital Repository (https://doi.org/10.5061/dryad.rbnzs7hq7). The meta-analysis was informed by literature searches conducted using the following databases and repositories: Web of Science (https://clarivate.com/academia-government/scientific-and-academic-research/research-discovery-and-referencing/web-of-science/), Scopus (https://www.elsevier.com/en-gb/products/scopus/search), the White Rose Research Online repository (https://eprints.whiterose.ac.uk/) and the South African National ETD Portal for theses and dissertations (https://spu-za.libguides.com/az/national-etd-portal-south-african-theses-and-dissertations). Trait values for common grass species found in the veld condition assessment plots at Kruger were extracted from the TRY database (https://www.try-db.org/TryWeb/Home.php). Citations for the TRY database and the original publications have been provided.

1. Chen, C., Riley, W. J., Prentice, I. C. & Keenan, T. F. CO₂ fertilization of terrestrial photosynthesis inferred from site to global scales. Proc. Natl Acad. Sci. USA 119, e2115627119 (2022).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

2. Forzieri, G., Dakos, V., McDowell, N. G., Ramdane, A. & Cescatti, A. Emerging signals of declining forest resilience under climate change. Nature 608, 534–539 (2022).

   Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

3. Jia, G. et al. in Climate Change and Land: an IPCC Special Report on Climate Change, Desertification, Land Degradation, Sustainable Land Management, Food Security, and Greenhouse Gas Fluxes in Terrestrial Ecosystems (eds Shukla, P. R. et al.) 131–242 (IPCC, 2019).

4. Hubau, W. et al. Asynchronous carbon sink saturation in African and Amazonian tropical forests. Nature 579, 80–87 (2020).

   Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

5. Grace, J., José, J. S., Meir, P., Miranda, H. S. & Montes, R. A. Productivity and carbon fluxes of tropical savannas. J. Biogeogr. 33, 387–400 (2006).

   Article  Google Scholar 

6. Lawrence, D. M. et al. The Community Land Model version 5: description of new features, benchmarking, and impact of forcing uncertainty. J. Adv. Model. Earth Syst. 11, 4245–4287 (2019).

   Article  ADS  Google Scholar 

7. Leakey, A. D., Bishop, K. A. & Ainsworth, E. A. A multi-biome gap in understanding of crop and ecosystem responses to elevated CO₂. Curr. Opin. Plant Biol. 15, 228–236 (2012).

   Article  CAS  PubMed  Google Scholar 

8. Pan, Y. et al. Contrasting responses of woody and grassland ecosystems to increased CO₂ as water supply varies. Nat. Ecol. Evol. 6, 315–323 (2022).

   Article  PubMed  Google Scholar 

9. Nowak, R. S., Ellsworth, D. S. & Smith, S. D. Functional responses of plants to elevated atmospheric CO₂—do photosynthetic and productivity data from FACE experiments support early predictions? New Phytol. 162, 253–280 (2004).

   Article  Google Scholar 

10. Ainsworth, E. A. & Long, S. P. 30 Years of free-air carbon dioxide enrichment (FACE): what have we learned about future crop productivity and its potential for adaptation? Glob. Chang. Biol. 27, 27–49 (2021).

    Article  ADS  CAS  PubMed  Google Scholar 

11. Bond, W. J. & Midgley, G. F. Carbon dioxide and the uneasy interactions of trees and savannah grasses. Phil. Trans. R. Soc. B 367, 601–612 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

12. Buitenwerf, R., Bond, W. J., Stevens, N. & Trollope, W. S. W. Increased tree densities in South African savannas: >50 years of data suggests CO₂ as a driver. Glob. Chang. Biol. 18, 675–684 (2012).

    Article  ADS  Google Scholar 

13. Staver, A. C., Botha, J. & Hedin, L. Soils and fire jointly determine vegetation structure in an African savanna. New Phytol. 216, 1151–1160 (2017).

    Article  CAS  PubMed  Google Scholar 

14. Alvarado, S. T., Andela, N., Silva, T. S. F. & Archibald, S. Thresholds of fire response to moisture and fuel load differ between tropical savannas and grasslands across continents. Global Ecol. Biogeogr. 29, 331–344 (2020).

    Article  Google Scholar 

15. Leakey, A. D. et al. Photosynthesis, productivity, and yield of maize are not affected by open-air elevation of CO₂ concentration in the absence of drought. Plant Physiol. 140, 779–790 (2006).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

16. Wand, S. J. E., Midgley, G. F., Jones, M. H. & Curtis, P. S. Responses of wild C₄ and C₃ grass (Poaceae) species to elevated atmospheric CO₂ concentration: a meta-analytic test of current theories and perceptions. Glob. Chang. Biol. 5, 723–741 (1999).

    Article  ADS  Google Scholar 

17. Morgan, J. A. et al. C₄ grasses prosper as carbon dioxide eliminates desiccation in warmed semi-arid grassland. Nature 476, 202–205 (2011).

    Article  ADS  CAS  PubMed  Google Scholar 

18. Reich, P. B., Hobbie, S. E., Lee, T. D. & Pastore, M. A. Unexpected reversal of C₃ versus C₄ grass response to elevated CO₂ during a 20-year field experiment. Science 360, 317–320 (2018).

    Article  ADS  CAS  PubMed  Google Scholar 

19. Spinoni, J., Naumann, G., Carrao, H., Barbosa, P. & Vogt, J. World drought frequency, duration, and severity for 1951–2010. Int. J. Climatol. 34, 2792–2804 (2014).

    Article  Google Scholar 

20. Chiang, F., Mazdiyasni, O. & AghaKouchak, A. Evidence of anthropogenic impacts on global drought frequency, duration, and intensity. Nat. Commun. 12, 2754 (2021).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

21. Sala, O. E., Gherardi, L. A., Reichmann, L., Jobbágy, E. & Peters, D. Legacies of precipitation fluctuations on primary production: theory and data synthesis. Phil. Trans. R. Soc. B 367, 3135–3144 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

22. Staver, A. C., Abraham, J. O., Hempson, G. P., Karp, A. T. & Faith, J. T. The past, present, and future of herbivore impacts on savanna vegetation. J. Ecol. 109, 2804–2822 (2021).

    Article  Google Scholar 

23. Smit, I. P., Peel, M. J., Ferreira, S. M., Greaver, C. & Pienaar, D. J. Megaherbivore response to droughts under different management regimes: lessons from a large African savanna. Afr. J. Range Forage Sci. 37, 65–80 (2020).

    Article  Google Scholar 

24. Stevens, C. J. et al. Anthropogenic nitrogen deposition predicts local grassland primary production worldwide. Ecology 96, 1459–1465 (2015).

    Article  Google Scholar 

25. Del Toro, I., Case, M. F., Karp, J., Slingsby, J. & Staver, A. C. Carbon isotope trends across a century of herbarium specimens suggest CO₂ fertilization of C₄ grasses. New Phytol. 243, 560–566 (2024).

    Article  PubMed  Google Scholar 

26. Keeling, R. F., Morgan, E. J. & Keeling, C. D. Atmospheric monthly in situ CO₂ data—Mauna Loa Observatory, Hawaii. In Scripps CO2 Program Data. UC San Diego Library Digital Collections https://doi.org/10.6075/J08W3BHW (2017).

27. Terrer, C. et al. Nitrogen and phosphorus constrain the CO₂ fertilization of global plant biomass. Nat. Clim. Change 9, 684–689 (2019).

    Article  ADS  CAS  Google Scholar 

28. Fay, P. A. et al. Soil-mediated effects of subambient to increased carbon dioxide on grassland productivity. Nat. Clim. Change 2, 742–746 (2012).

    Article  ADS  CAS  Google Scholar 

29. Polley, H. W. et al. CO₂ enrichment and soil type additively regulate grassland productivity. New Phytol. 222, 183–192 (2019).

    Article  CAS  PubMed  Google Scholar 

30. Collatz, G. J., Ribas-Carbo, M. & Berry, J. A. Coupled photosynthesis-stomatal conductance model for leaves of C₄ plants. Funct. Plant Biol. 19, 519–538 (1992).

    Article  Google Scholar 

31. Kennedy, D. et al. Implementing plant hydraulics in the Community Land Model version 5. J. Adv. Model. Earth Syst. 11, 485–513 (2019).

    Article  ADS  Google Scholar 

32. Medlyn, B. E. et al. Reconciling the optimal and empirical approaches to modelling stomatal conductance. Glob. Chang. Biol. 17, 2134–2144 (2011).

    Article  ADS  Google Scholar 

33. Lin, Y. S. et al. Optimal stomatal behaviour around the world. Nat. Clim. Change 5, 459–464 (2015).

    Article  ADS  CAS  Google Scholar 

34. Noy-Meir, I. Desert ecosystems: environment and producers. Annu. Rev. Ecol. Syst. 4, 25–52 (1973).

    Article  Google Scholar 

35. Piñeiro, J. et al. Effects of elevated CO₂ on fine root biomass are reduced by aridity but enhanced by soil nitrogen: a global assessment. Sci. Rep. 7, 15355 (2017).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

36. Purcell, C. et al. Increasing stomatal conductance in response to rising atmospheric CO₂. Ann. Bot. 121, 1137–1149 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

37. Scoffoni, C., Albuquerque, C., Buckley, T. N. & Sack, L. The dynamic multi-functionality of leaf water transport outside the xylem. New Phytol. 239, 2099–2107 (2023).

    Article  CAS  PubMed  Google Scholar 

38. Márquez, D. A., Wong, S. C., Stuart-Williams, H., Cernusak, L. A. & Farquhar, G. D. Mesophyll airspace unsaturation drives C₄ plant success under vapor pressure deficit stress. Proc. Natl Acad. Sci. USA 121, e2402233121 (2024).

    Article  PubMed  PubMed Central  Google Scholar 

39. Crafts-Brandner, S. J. & Salvucci, M. E. Sensitivity of photosynthesis in a C₄ plant, maize, to heat stress. Plant Physiol. 129, 1773–1780 (2002).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

40. Boyd, R. A., Gandin, A. & Cousins, A. B. Temperature responses of C₄ photosynthesis: biochemical analysis of rubisco, phosphoenolpyruvate carboxylase, and carbonic anhydrase in Setaria viridis. Plant Physiol. 169, 1850–1861 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

41. Yin, X., van der Putten, P. E. L., Driever, S. M. & Struik, P. C. Temperature response of bundle-sheath conductance in maize leaves. J. Exp. Bot. 67, 2699–2714 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

42. Sonawane, B. V., Sharwood, R. E., von Caemmerer, S., Whitney, S. M. & Ghannoum, O. Short-term thermal photosynthetic responses of C₄ grasses are independent of the biochemical subtype. J. Exp. Bot. 68, 5583–5597 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

43. Raubenheimer, S. L., Simpson, K., Carkeek, R. & Ripley, B. Could CO₂-induced changes to C₄ grass flammability aggravate savanna woody encroachment? Afr. J. Range Forage Sci. 39, 82–95 (2022).

    Article  Google Scholar 

44. Reyes-Fox, M. et al. Elevated CO₂ further lengthens growing season under warming conditions. Nature 510, 259–262 (2014).

    Article  ADS  CAS  PubMed  Google Scholar 

45. Abreu, R. C. R. et al. The biodiversity cost of carbon sequestration in tropical savanna. Sci. Adv. 3, e1701284 (2017).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

46. Zhou, Y. et al. Limited increases in savanna carbon stocks over decades of fire suppression. Nature 603, 445–449 (2022).

    Article  ADS  CAS  PubMed  Google Scholar 

47. Poulter, B. et al. Contribution of semi-arid ecosystems to interannual variability of the global carbon cycle. Nature 509, 600–603 (2014).

    Article  ADS  CAS  PubMed  Google Scholar 

48. Zhou, Y. et al. Soil carbon in tropical savannas mostly derived from grasses. Nat. Geosci. 16, 710–716 (2023).

    Article  ADS  Google Scholar 

49. Robinson, E. A., Ryan, G. D. & Newman, J. A. A meta-analytical review of the effects of elevated CO₂ on plant–arthropod interactions highlights the importance of interacting environmental and biological variables. New Phytol. 194, 321–336 (2012).

    Article  CAS  PubMed  Google Scholar 

50. Luo, X. et al. Mapping the global distribution of C₄ vegetation using observations and optimality theory. Nat. Commun. 15, 1219 (2024).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

51. Skowno, A. L. et al. Woodland expansion in South African grassy biomes based on satellite observations (1990–2013): general patterns and potential drivers. Glob. Chang. Biol. 23, 2358–2369 (2017).

    Article  ADS  PubMed  Google Scholar 

52. Grames, E. M., Stillman, A. N., Tingley, M. W. & Elphick, C. S. An automated approach to identifying search terms for systematic reviews using keyword co-occurrence networks. Methods Ecol. Evol. 10, 1645–1654 (2019).

    Article  Google Scholar 

53. Rohatgi, A. WebPlotDigitizer. GitHub https://github.com/ankitrohatgi/WebPlotDigitizer (2023).

54. Abdalla, A. L. et al. Nutritive value and enteric methane production of Brachiaria spp. under elevated [CO₂]. Int. J. Plant Prod. 14, 119–126 (2020).

    Article  Google Scholar 

55. Abdalla Filho, A. L. et al. Fiber fractions, multielemental and isotopic composition of a tropical C₄ grass grown under elevated atmospheric carbon dioxide. PeerJ 7, e5932 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

56. Abdalla Filho, A. L. et al. CO₂ fertilization does not affect biomass production and nutritive value of a C₄ tropical grass in short timeframe. Grass Forage Sci. 74, 670–677 (2019).

    Article  CAS  Google Scholar 

57. Adam, N. R., Owensby, C. E. & Ham, J. M. The effect of CO₂ enrichment on leaf photosynthetic rates and instantaneous water use efficiency of Andropogon gerardii in the tallgrass prairie. Photosynth. Res. 65, 121–129 (2000).

    Article  CAS  PubMed  Google Scholar 

58. Anderson, L. J., Maherali, H., Johnson, H. B., Polley, H. W. & Jackson, R. B. Gas exchange and photosynthetic acclimation over subambient to elevated CO₂ in a C₃–C₄ grassland. Glob. Chang. Biol. 7, 693–707 (2001).

    Article  ADS  Google Scholar 

59. Barbehenn, R. V., Chen, Z., Karowe, D. N. & Spickard, A. C₃ grasses have higher nutritional quality than C₄ grasses under ambient and elevated atmospheric CO₂. Glob. Chang. Biol. 10, 1565–1575 (2004).

    Article  ADS  Google Scholar 

60. Baruch, Z. & Jackson, R. B. Responses of tropical native and invader C₄ grasses to water stress, clipping and increased atmospheric CO₂ concentration. Oecologia 145, 522–532 (2005).

    Article  ADS  PubMed  Google Scholar 

61. Bellasio, C., Quirk, J. & Beerling, D. J. Stomatal and non-stomatal limitations in savanna trees and C₄ grasses grown at low, ambient and high atmospheric CO₂. Plant Sci. 274, 181–192 (2018).

    Article  CAS  PubMed  Google Scholar 

62. von Caemmerer, S., Ghannoum, O., Conroy, J. P., Clark, H. & Newton, P. C. D. Photosynthetic responses of temperate species to free air CO₂ enrichment (FACE) in a grazed New Zealand pasture. Funct. Plant Biol. 28, 439–450 (2001).

    Google Scholar 

63. Carter, D. R. & Peterson, K. M. Effects of a CO₂-enriched atmosphere on the growth and competitive interaction of a C₃ and a C₄ grass. Oecologia 58, 188–193 (1983).

    Article  ADS  CAS  PubMed  Google Scholar 

64. Carvalho, J. M. et al. Elevated CO₂ and warming change the nutrient status and use efficiency of Panicum maximum Jacq. PLoS ONE 15, e0223937 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

65. Clark, H., Newton, P. C. D. & Barker, D. J. Physiological and morphological responses to elevated CO₂ and a soil moisture deficit of temperate pasture species growing in an established plant community. J. Exp. Bot. 50, 233–242 (1999).

    Article  CAS  Google Scholar 

66. de Oliveira, A. C. G., Rios, P. M., Pereira, E. G. & Souza, J. P. Growth and competition between a native leguminous forb and an alien grass from the Cerrado under elevated CO₂. Austral Ecol. 46, 750–761 (2021).

    Article  Google Scholar 

67. Dijkstra, F. A., Blumenthal, D., Morgan, J. A., LeCain, D. R. & Follett, R. F. Elevated CO₂ effects on semi-arid grassland plants in relation to water availability and competition. Funct. Ecol. 24, 1152–1161 (2010).

    Article  Google Scholar 

68. de Faria, A. P., Marabesi, M. A., Gaspar, M. & França, M. G. C. The increase of current atmospheric CO₂ and temperature can benefit leaf gas exchanges, carbohydrate content and growth in C₄ grass invaders of the Cerrado biome. Plant Physiol. Biochem. 127, 608–616 (2018).

    Article  PubMed  Google Scholar 

69. Fravolini, A., Williams, D. G. & Thompson, T. L. Carbon isotope discrimination and bundle sheath leakiness in three C₄ subtypes grown under variable nitrogen, water and atmospheric CO₂ supply. J. Exp. Bot. 53, 2261–2269 (2002).

    Article  CAS  PubMed  Google Scholar 

70. Ghannoum, O., Caemmerer, S. V., Barlow, E. W. R. & Conroy, J. P. The effect of CO₂ enrichment and irradiance on the growth, morphology and gas exchange of a C₃ (Panicum laxum) and a C₄ (Panicum antidotale) grass. Funct. Plant Biol. 24, 227–237 (1997).

    CAS  Google Scholar 

71. Ghannoum, O., von Caemmerer, S. & Conroy, J. P. Plant water use efficiency of 17 Australian NAD-ME and NADP-ME C₄ grasses at ambient and elevated CO₂ partial pressure. Funct. Plant Biol. 28, 1207–1217 (2001).

    CAS  Google Scholar 

72. Gifford, R. M. & Morison, J. I. L. Photosynthesis, water use and growth of a C₄ grass stand at high CO₂ concentration. Photosynth. Res. 7, 77–90 (1985).

    Article  CAS  PubMed  Google Scholar 

73. Hager, H. A., Ryan, G. D., Kovacs, H. M. & Newman, J. A. Effects of elevated CO₂ on photosynthetic traits of native and invasive C₃ and C₄ grasses. BMC Ecol. 16, 28 (2016).

    Article  PubMed  PubMed Central  Google Scholar 

74. Hager, H. A., Ryan, G. D. & Newman, J. A. Effects of elevated CO₂ on competition between native and invasive grasses. Oecologia 192, 1099–1110 (2020).

    Article  ADS  PubMed  Google Scholar 

75. Hamerlynck, E. P., McAllister, C. A., Knapp, A. K., Ham, J. M. & Owensby, C. E. Photosynthetic gas exchange and water relation responses of three tallgrass prairie species to elevated carbon dioxide and moderate drought. Int. J. Plant Sci. 158, 608–616 (1997).

    Article  Google Scholar 

76. Hunt, H. W., Elliott, E. T., Detling, J. K., Morgan, J. A. & Chen, D.-X. Responses of a C3 and a C₄ perennial grass to elevated CO₂ and temperature under different water regimes. Glob. Chang. Biol. 2, 35–47 (1996).

    Article  ADS  Google Scholar 

77. Johnson, S. N., Lopaticki, G. & Hartley, S. E. Elevated atmospheric CO₂ triggers compensatory feeding by root herbivores on a C₃ but not a C₄ grass. PLoS ONE 9, e90251 (2014).

    Article  ADS  PubMed  PubMed Central  Google Scholar 

78. Kellogg, E. A., Farnsworth, E. J., Russo, E. T. & Bazzaz, F. Growth responses of C₄ grasses of contrasting origin to elevated CO₂. Ann. Bot. 84, 279–288 (1999).

    Article  Google Scholar 

79. Kgope, B. S., Bond, W. J. & Midgley, G. F. Growth responses of African savanna trees implicate atmospheric [CO₂] as a driver of past and current changes in savanna tree cover. Austral Ecol. 35, 451–463 (2010).

    Article  Google Scholar 

80. Knapp, A. K., Hamerlynck, E. P. & Owensby, C. E. Photosynthetic and water relations responses to elevated CO₂ in the C₄ grass Andropogon gerardii. Int. J. Plant Sci. 154, 459–466 (1993).

    Article  CAS  Google Scholar 

81. LeCain, D. R. et al. Root biomass of individual species, and root size characteristics after five years of CO₂ enrichment on native shortgrass steppe. Plant Soil 279, 219–228 (2006).

    Article  CAS  Google Scholar 

82. LeCain, D. R. & Morgan, J. A. Growth, gas exchange, leaf nitrogen and carbohydrate concentrations in NAD-ME and NADP-ME C₄ grasses grown in elevated CO₂. Physiol. Plant. 102, 297–306 (1998).

    Article  CAS  Google Scholar 

83. LeCain, D. R., Morgan, J. A., Mosier, A. R. & Nelson, J. A. Soil and plant water relations determine photosynthetic responses of C3 and C4 grasses in a semi-arid ecosystem under elevated CO₂. Ann. Bot. 92, 41–52 (2003).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

84. Lee, T. D., Barrott, S. H. & Reich, P. B. Photosynthetic responses of 13 grassland species across 11 years of free-air CO₂ enrichment is modest, consistent and independent of N supply. Glob. Chang. Biol. 17, 2893–2904 (2011).

    Article  ADS  Google Scholar 

85. Lee, T. D., Tjoelker, M. G., Ellsworth, D. S. & Reich, P. B. Leaf gas exchange responses of 13 prairie grassland species to elevated CO₂ and increased nitrogen supply. New Phytol. 150, 405–418 (2001).

    Article  CAS  Google Scholar 

86. Maherali, H., Reid, C. D., Polley, H. W., Johnson, H. B. & Jackson, R. B. Stomatal acclimation over a subambient to elevated CO₂ gradient in a C₃/C₄ grassland. Plant Cell Environ. 25, 557–566 (2002).

    Article  CAS  Google Scholar 

87. Manea, A., Leishman, M. R. & Downey, P. O. Exotic C₄ grasses have increased tolerance to glyphosate under elevated carbon dioxide. Weed Sci. 59, 28–36 (2011).

    Article  CAS  Google Scholar 

88. Marks, S. & Clay, K. Effects of CO₂ enrichment, nutrient addition, and fungal endophyte-infection on the growth of two grasses. Oecologia 84, 207–214 (1990).

    Article  ADS  PubMed  Google Scholar 

89. McGranahan, D. A. & Yurkonis, K. A. Variability in grass forage quality and quantity in response to elevated CO₂ and water limitation. Grass Forage Sci. 73, 517–521 (2018).

    Article  CAS  Google Scholar 

90. Morgan, J. A., Knight, W. G., Dudley, L. M. & Hunt, H. W. Enhanced root system C-sink activity, water relations and aspects of nutrient acquisition in mycotrophic Bouteloua gracilis subjected to CO₂ enrichment. Plant Soil 165, 139–146 (1994).

    Article  CAS  Google Scholar 

91. Morgan, J. A., LeCain, D. R., Read, J. J., Hunt, H. W. & Knight, W. G. Photosynthetic pathway and ontogeny affect water relations and the impact of CO₂ on Bouteloua gracilis (C₄) and Pascopyrum smithii (C₃). Oecologia 114, 483–493 (1998).

    Article  ADS  CAS  PubMed  Google Scholar 

92. Morgan, J. A., Lecain, D. R., Mosier, A. R. & Milchunas, D. G. Elevated CO₂ enhances water relations and productivity and affects gas exchange in C₃ and C₄ grasses of the Colorado shortgrass steppe. Glob. Chang. Biol. 7, 451–466 (2001).

    Article  ADS  Google Scholar 

93. Morgan, J. A. et al. CO₂ enhances productivity, alters species composition, and reduces digestibility of shortgrass steppe vegetation. Ecol. Appl. 14, 208–219 (2004).

    Article  Google Scholar 

94. Newman, Y. C., Sollenberger, L. E., Boote, K. J., Allen, L. H. & Littell, R. C. Carbon dioxide and temperature effects on forage dry matter production. Crop Sci. 41, 399–406 (2001).

    Article  Google Scholar 

95. Pallett, N. The Effects of Elevated CO₂ on C₄ Panicoid Grass Drought Tolerance. MSc Thesis, Rhodes Univ. (2018).

96. Pastore, M. A., Lee, T. D., Hobbie, S. E. & Reich, P. B. Strong photosynthetic acclimation and enhanced water-use efficiency in grassland functional groups persist over 21 years of CO₂ enrichment, independent of nitrogen supply. Glob. Chang. Biol. 25, 3031–3044 (2019).

    Article  ADS  PubMed  Google Scholar 

97. Pastore, M. A., Lee, T. D., Hobbie, S. E. & Reich, P. B. Interactive effects of elevated CO₂, warming, reduced rainfall, and nitrogen on leaf gas exchange in five perennial grassland species. Plant Cell Environ. 43, 1862–1878 (2020).

    Article  CAS  PubMed  Google Scholar 

98. Paterson, E., Rattray, E. A. S. & Killham, K. Effect of elevated atmospheric CO₂ concentration on C-partitioning and rhizosphere C-flow for three plant species. Soil Biol. Biochem. 28, 195–201 (1996).

    Article  CAS  Google Scholar 

99. Polley, H. W., Johnson, H. B., Mayeux, H. S. & Brown, D. A. Leaf and plant water use efficiency of C₄ species grown at glacial to elevated CO₂ concentrations. Int. J. Plant Sci. 157, 164–170 (1996).

    Article  CAS  Google Scholar 

100. Quirk, J., Bellasio, C., Johnson, D. A., Osborne, C. P. & Beerling, D. J. C₄ savanna grasses fail to maintain assimilation in drying soil under low CO₂ compared with C₃ trees despite lower leaf water demand. Funct. Ecol. 33, 388–398 (2019).

     Article  Google Scholar 

101. Reich, P. B. et al. Do species and functional groups differ in acquisition and use of C, N and water under varying atmospheric CO₂ and N availability regimes? A field test with 16 grassland species. New Phytol. 150, 435–448 (2001).

     Article  CAS  Google Scholar 

102. Rudmann, S. G., Milham, P. J. & Conroy, J. P. Influence of high CO₂ partial pressure on nitrogen use efficiency of the C₄ Grasses Panicum coloratum and Cenchrus ciliaris. Ann. Bot. 88, 571–577 (2001).

     Article  CAS  Google Scholar 

103. Runion, G. B., Prior, S. A., Capo-chichi, L. J. A., Torbert, H. A. & van Santen, E. Varied growth response of cogongrass ecotypes to elevated CO₂. Front. Plant Sci. 6, 1182 (2016).

     Article  PubMed  PubMed Central  Google Scholar 

104. Seneweera, S. P., Ghannoum, O. & Conroy, J. High vapour pressure deficit and low soil water availability enhance shoot growth responses of a C₄ grass (Panicum coloratum cv. Bambatsi) to CO₂ enrichment. Funct. Plant Biol. 25, 287–292 (1998).

     Google Scholar 

105. Seneweera, S. P., Ghannoum, O. & Conroy, J. P. Root and shoot factors contribute to the effect of drought on photosynthesis and growth of the C₄ grass Panicum coloratum at elevated CO₂ partial pressures. Aust. J. Plant Physiol. https://doi.org/10.1071/PP01007 (2001).

106. Sionit, N. & Patterson, D. T. Responses of C₄ grasses to atmospheric CO₂ enrichment. Oecologia 65, 30–34 (1984).

     Article  ADS  PubMed  Google Scholar 

107. Smith, S., Strain, B. & Sharkey, T. Effects of CO₂ enrichment on four Great Basin grasses. Funct. Ecol. 1, 139–143 (1987).

     Article  Google Scholar 

108. Tooth, I. M. & Leishman, M. R. Post-fire resprouting responses of native and exotic grasses from Cumberland Plain Woodland (Sydney, Australia) under elevated carbon dioxide. Austral Ecol. 38, 1–10 (2013).

     Article  Google Scholar 

109. Tooth, I. M. & Leishman, M. R. Elevated carbon dioxide and fire reduce biomass of native grass species when grown in competition with invasive exotic grasses in a savanna experimental system. Biol. Invasions 16, 257–268 (2014).

     Article  Google Scholar 

110. Volin, J. C., Reich, P. B. & Givnish, T. J. Elevated carbon dioxide ameliorates the effects of ozone on photosynthesis and growth: species respond similarly regardless of photosynthetic pathway or plant functional group. New Phytol. 138, 315–325 (1998).

     Article  CAS  PubMed  Google Scholar 

111. Wand, S. J. E. & Midgley, G. F. Effects of atmospheric CO₂ concentration and defoliation on the growth of Themeda triandra. Grass Forage Sci. 59, 215–226 (2004).

     Article  Google Scholar 

112. Wand, S., Midgley, G. & Stock, W. Response to elevated CO₂ from a natural spring in a C₄-dominated grassland depends on seasonal phenology. Afr. J. Range Forage Sci. 19, 81–91 (2002).

     Article  Google Scholar 

113. Wand, S. J. E., Midgley, G. F. & Musil, C. F. Physiological and growth responses of two African species, Acacia karroo and Themeda triandra, to combined increases in CO₂ and UV-B radiation. Physiol. Plant. 98, 882–890 (1996).

     Article  CAS  Google Scholar 

114. Wand, S. J. E., Midgley, G. F. & Stock, W. D. Growth responses to elevated CO₂ in NADP-ME, NAD-ME and PCK C₄ grasses and a C₃ grass from South Africa. Funct. Plant Biol. 28, 13–25 (2001).

     Google Scholar 

115. Watling, J. R. & Press, M. C. How does the C₄ grass Eragrostis pilosa respond to elevated carbon dioxide and infection with the parasitic angiosperm Striga hermonthica? New Phytol. 140, 667–675 (1998).

     Article  CAS  PubMed  Google Scholar 

116. Weller, S. L., Florentine, S. K., Mutti, N. K., Jha, P. & Chauhan, B. S. Response of Chloris truncata to moisture stress, elevated carbon dioxide and herbicide application. Sci. Rep. 9, 10721 (2019).

     Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

117. Wilsey, B. J., Coleman, J. S. & McNaughton, S. J. Effects of elevated CO₂ and defoliation on grasses: a comparative ecosystem approach. Ecol. Appl. 7, 844–853 (1997).

     Google Scholar 

118. Wilsey, B. J., McNaughton, S. J. & Coleman, J. S. Will increases in atmospheric CO₂ affect regrowth following grazing in C₄ grasses from tropical grasslands? A test with Sporobolus kentrophyllus. Oecologia 99, 141–144 (1994).

     Article  ADS  PubMed  Google Scholar 

119. Xiao, L., Liu, G. & Xue, S. Elevated CO₂ concentration and drought stress exert opposite effects on plant biomass, nitrogen, and phosphorus allocation in Bothriochloa ischaemum. J. Plant Growth Regul. 35, 1088–1097 (2016).

     Article  CAS  Google Scholar 

120. Xu, Z. et al. Effects of elevated CO₂, warming and precipitation change on plant growth, photosynthesis and peroxidation in dominant species from North China grassland. Planta 239, 421–435 (2014).

     Article  CAS  PubMed  Google Scholar 

121. Yu, J., Sun, L., Fan, N., Yang, Z. & Huang, B. Physiological factors involved in positive effects of elevated carbon dioxide concentration on Bermudagrass tolerance to salinity stress. Environ. Exp. Bot. 115, 20–27 (2015).

     Article  CAS  Google Scholar 

122. Ziska, L. H., Hogan, K. P., Smith, A. P. & Drake, B. G. Growth and photosynthetic response of nine tropical species with long-term exposure to elevated carbon dioxide. Oecologia 86, 383–389 (1991).

     Article  ADS  CAS  PubMed  Google Scholar 

123. Hedges, L. V., Gurevitch, J. & Curtis, P. S. The meta-analysis of response ratios in experimental ecology. Ecology 80, 1150–1156 (1999).

     Article  Google Scholar 

124. Bracken, M. B. in Effective Care of Newborn Infants (eds Sinclair, J. C. & Bracken M. B.) 13–20 (Oxford Univ. Press, 1992).

125. Hadfield, J. D. MCMC methods for multi-response generalized linear mixed models: the MCMCGLMM R package. J. Stat. Softw. 33, 1–22 (2010).

     Article  Google Scholar 

126. R Core Team. R: a Language and Environment for Statistical Computing https://www.R-project.org/ (R Foundation for Statistical Computing, 2023).

127. Forrestel, E. J. Biogeographic Influences on Grassland Community Structure and Function. PhD Thesis, Yale Univ. (2015).

128. Plummer, M., Best, N., Cowles, K. & Vines, K. CODA: convergence diagnosis and output analysis for MCMC. R News 6, 7–11 (2006).

     Google Scholar 

129. Lenth, R. emmeans: Estimated marginal means, aka least-squares means. R package version 1.9.0 https://CRAN.R-project.org/package=emmeans (CRAN, 2023).

130. Walker, A. P. et al. Integrating the evidence for a terrestrial carbon sink caused by increasing atmospheric CO₂. New Phytol. 229, 2413–2445 (2021).

     Article  CAS  PubMed  Google Scholar 

131. Jacob, V. et al. High safety margins to drought-induced hydraulic failure found in five pasture grasses. Plant Cell Environ. 45, 1631–1646 (2022).

     Article  CAS  PubMed  Google Scholar 

132. Osborne, C. P. & Sack, L. Evolution of C₄ plants: a new hypothesis for an interaction of CO₂ and water relations mediated by plant hydraulics. Phil. Trans. R. Soc. B 367, 583–600 (2012).

     Article  CAS  PubMed  PubMed Central  Google Scholar 

133. Wei, C., Tyree, M. T. & Steudle, E. Direct measurement of xylem pressure in leaves of intact maize plants. A test of the cohesion-tension theory taking hydraulic architecture into consideration. Plant Physiol. 121, 1191–1206 (1999).

     Article  CAS  PubMed  PubMed Central  Google Scholar 

134. Staver, A. C., Wigley-Coetsee, C. & Botha, J. Grazer movements exacerbate grass declines during drought in an African savanna. J. Ecol. 107, 1482–1491 (2019).

     Article  Google Scholar 

135. Trollope, W. S. W. & Potgieter, A. L. F. Estimating grass fuel loads with a disc pasture meter in the Kruger National Park. J. Grassl. Soc. S. Afr. 3, 148–152 (1986).

     Article  Google Scholar 

136. Wigley-Coetsee, C. & Staver, A. Grass community responses to drought in an African savanna. Afr. J. Range Forage Sci. 37, 43–52 (2020).

     Article  Google Scholar 

137. Clayton, W. D., Vorontsova, M. S., Harman, K. T. & Williamson, H. GrassBase—the online world. Kew http://www.kew.org/data/grasses-db.html (2006).

138. Kattge, J. et al. TRY plant trait database—enhanced coverage and open access. Glob. Chang. Biol. 26, 119–188 (2020).

     Article  ADS  PubMed  Google Scholar 

139. Adler, P. B. et al. Functional traits explain variation in plant life history strategies. Proc. Natl Acad. Sci. USA 111, 740–745 (2014).

     Article  ADS  CAS  PubMed  Google Scholar 

140. Atkin, O. K. et al. Global variability in leaf respiration in relation to climate, plant functional types and leaf traits. New Phytol. 206, 614–636 (2015).

     Article  CAS  PubMed  Google Scholar 

141. Baruch, Z. & Goldstein, G. Leaf construction cost, nutrient concentration, and net CO₂ assimilation of native and invasive species in Hawaii. Oecologia 121, 183–192 (1999).

     Article  ADS  CAS  PubMed  Google Scholar 

142. Catford, J. A., Morris, W. K., Vesk, P. A., Gippel, C. J. & Downes, B. J. Species and environmental characteristics point to flow regulation and drought as drivers of riparian plant invasion. Diversity Distrib. 20, 1084–1096 (2014).

     Article  Google Scholar 

143. Craine, J. M. et al. Global patterns of foliar nitrogen isotopes and their relationships with climate, mycorrhizal fungi, foliar nutrient concentrations, and nitrogen availability. New Phytol. 183, 980–992 (2009).

     Article  CAS  PubMed  Google Scholar 

144. Craine, J. M., Lee, W. G., Bond, W. J., Williams, R. J. & Johnson, L. C. Environmental constraints on a global relationship among leaf and root traits of grasses. Ecology 86, 12–19 (2005).

     Article  Google Scholar 

145. Craine, J. M. et al. Global diversity of drought tolerance and grassland climate-change resilience. Nat. Clim. Change 3, 63–67 (2013).

     Article  ADS  Google Scholar 

146. Diaz, S. et al. The plant traits that drive ecosystems: evidence from three continents. J. Veg. Sci. 15, 295–304 (2004).

     Article  Google Scholar 

147. Domingues, T. F., Martinelli, L. A. & Ehleringer, J. R. Ecophysiological traits of plant functional groups in forest and pasture ecosystems from eastern Amazônia, Brazil. Plant Ecol. 193, 101–112 (2007).

     Article  Google Scholar 

148. Fonseca, C. R., Overton, J. M., Collins, B. & Westoby, M. Shifts in trait-combinations along rainfall and phosphorus gradients. J. Ecol. 88, 964–977 (2000).

     Article  Google Scholar 

149. Fortunel, C. et al. Leaf traits capture the effects of land use changes and climate on litter decomposability of grasslands across Europe. Ecology 90, 598–611 (2009).

     Article  PubMed  Google Scholar 

150. He, J.-S. et al. Leaf nitrogen:phosphorus stoichiometry across Chinese grassland biomes. Oecologia 155, 301–310 (2008).

     Article  ADS  PubMed  Google Scholar 

151. Kattge, J., Knorr, W., Raddatz, T. & Wirth, C. Quantifying photosynthetic capacity and its relationship to leaf nitrogen content for global-scale terrestrial biosphere models. Glob. Chang. Biol. 15, 976–991 (2009).

     Article  ADS  Google Scholar 

152. Kazakou, E., Vile, D., Shipley, B., Gallet, C. & Garnier, E. Co-variations in litter decomposition, leaf traits and plant growth in species from a Mediterranean old-field succession. Funct. Ecol. 20, 21–30 (2006).

     Article  Google Scholar 

153. Kerkhoff, A. J., Fagan, W. F., Elser, J. J. & Enquist, B. J. Phylogenetic and growth form variation in the scaling of nitrogen and phosphorus in the seed plants. Am. Nat. 168, E103–E122 (2006).

     Article  PubMed  Google Scholar 

154. Kleyer, M. et al. The LEDA Traitbase: a database of life-history traits of the Northwest European flora. J. Ecol. 96, 1266–1274 (2008).

     Article  Google Scholar 

155. Maire, V. et al. Global effects of soil and climate on leaf photosynthetic traits and rates. Global Ecol. Biogeogr. 24, 706–717 (2015).

     Article  Google Scholar 

156. McDonald, P. G., Fonseca, C. R., Overton, J. M. & Westoby, M. Leaf-size divergence along rainfall and soil-nutrient gradients: is the method of size reduction common among clades? Funct. Ecol. 17, 50–57 (2003).

     Article  Google Scholar 

157. Meziane, D. & Shipley, B. Interacting components of interspecific relative growth rate: constancy and change under differing conditions of light and nutrient supply. Funct. Ecol. 13, 611–622 (1999).

     Article  Google Scholar 

158. Onoda, Y. et al. Physiological and structural tradeoffs underlying the leaf economics spectrum. New Phytol. 214, 1447–1463 (2017).

     Article  CAS  PubMed  Google Scholar 

159. Pakeman, R. J. et al. Relative climatic, edaphic and management controls of plant functional trait signatures. J. Veg. Sci. 20, 148–159 (2009).

     Article  Google Scholar 

160. Peco, B., de Pablos, I., Traba, J. & Levassor, C. The effect of grazing abandonment on species composition and functional traits: the case of dehesa grasslands. Basic Appl. Ecol. 6, 175–183 (2005).

     Article  Google Scholar 

161. Prentice, I. C. et al. Evidence of a universal scaling relationship for leaf CO₂ drawdown along an aridity gradient. New Phytol. 190, 169–180 (2011).

     Article  CAS  PubMed  Google Scholar 

162. Reich, P. B., Oleksyn, J. & Wright, I. J. Leaf phosphorus influences the photosynthesis-nitrogen relation: a cross-biome analysis of 314 species. Oecologia 160, 207–212 (2009).

     Article  ADS  PubMed  Google Scholar 

163. Shipley, B. & Vu, T.-T. Dry matter content as a measure of dry matter concentration in plants and their parts. New Phytol. 153, 359–364 (2002).

     Article  Google Scholar 

164. Tucker, S. S., Craine, J. M. & Nippert, J. B. Physiological drought tolerance and the structuring of tallgrass prairie assemblages. Ecosphere 2, art48 (2011).

165. Wang, H. et al. The China Plant Trait Database: toward a comprehensive regional compilation of functional traits for land plants. Ecology 99, 500 (2018).

     Article  PubMed  Google Scholar 

166. Wright, I. J. et al. The worldwide leaf economics spectrum. Nature 428, 821–827 (2004).

     Article  ADS  CAS  PubMed  Google Scholar 

167. Bates, D., Mächler, M., Bolker, B. & Walker, S. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48 (2015).

     Article  Google Scholar 

168. Liu, Y. et al. Evaluating the responses of net primary productivity and carbon use efficiency of global grassland to climate variability along an aridity gradient. Sci. Total Environ. 652, 671–682 (2019).

     Article  ADS  PubMed  Google Scholar 

169. Simpson, K. J. et al. C₄ photosynthesis and the economic spectra of leaf and root traits independently influence growth rates in grasses. J. Ecol. 108, 1899–1909 (2020).

     Article  CAS  Google Scholar 

170. Danabasoglu, G. et al. The Community Earth System Model version 2 (CESM2). J. Adv. Model. Earth Syst. 12, e2019MS001916 (2020).

     Article  ADS  Google Scholar 

171. Levis, S., Badger, A., Drewniak, B., Nevison, C. & Ren, X. CLM crop yields and water requirements: avoided impacts by choosing RCP 4.5 over 8.5. Clim. Change 146, 501–515 (2018).

     Article  ADS  Google Scholar 

172. Portmann, F. T., Siebert, S. & Döll, P. MIRCA2000—global monthly irrigated and rainfed crop areas around the year 2000: a new high-resolution data set for agricultural and hydrological modeling. Global Biogeochem. Cycles https://doi.org/10.1029/2008GB003435 (2010).

173. Bonan, G. B. et al. Improving canopy processes in the Community Land Model (CLM4) using global flux fields empirically inferred from FLUXNET data. J. Geophys. Res. 116, G02014 (2011).

     ADS  Google Scholar 

174. De Kauwe, M. G. et al. A test of an optimal stomatal conductance scheme within the CABLE land surface model. Geosci. Model Dev. 8, 431–452 (2015).

     Article  ADS  Google Scholar 

175. Shi, M., Fisher, J. B., Brzostek, E. R. & Phillips, R. P. Carbon cost of plant nitrogen acquisition: global carbon cycle impact from an improved plant nitrogen cycle in the Community Land Model. Glob. Chang. Biol. 22, 1299–1314 (2016).

     Article  ADS  PubMed  Google Scholar 

176. Li, F., Zeng, X. D. & Levis, S. A process-based fire parameterization of intermediate complexity in a dynamic global vegetation model. Biogeosciences 9, 2761–2780 (2012).

     Article  ADS  Google Scholar 

177. Li, F., Levis, S. & Ward, D. S. Quantifying the role of fire in the Earth system—part 1: improved global fire modeling in the Community Earth System Model (CESM1). Biogeosciences 10, 2293–2314 (2013).

     Article  ADS  CAS  Google Scholar 

178. Umair, M., Kim, D. & Choi, M. Impact of climate, rising atmospheric carbon dioxide, and other environmental factors on water-use efficiency at multiple land cover types. Sci. Rep. 10, 11644 (2020).

     Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

179. Lawrence, D. M., Koven, C. D., Swenson, S. C., Riley, W. J. & Slater, A. G. Permafrost thaw and resulting soil moisture changes regulate projected high-latitude CO₂ and CH₄ emissions. Environ. Res. Lett. 10, 094011 (2015).

     Article  Google Scholar 

180. Denager, T. et al. Point-scale multi-objective calibration of the Community Land Model (version 5.0) using in situ observations of water and energy fluxes and variables. Hydrol. Earth Syst. Sci. 27, 2827–2845 (2023).

     Article  ADS  Google Scholar 

181. Oloruntoba, B. J., Kollet, S., Montzka, C., Vereecken, H. & Hendricks Franssen, H.-J. High resolution land surface modelling over Africa: the role of uncertain soil properties in combination with temporal model resolution. Hydrol. Earth Syst. Sci. 29, 1659–1683 (2025).

     Article  ADS  Google Scholar 

Download references

Author notes

1. These authors contributed equally: Kimberley J. Simpson, A. Carla Staver

Authors and Affiliations

1. Plants, Photosynthesis and Soil, School of Biosciences, University of Sheffield, Sheffield, UK

   Kimberley J. Simpson, James A. King, Maria Val Martin & Colin P. Osborne

2. Botany Department, Rhodes University, Makhanda, South Africa

   Kimberley J. Simpson, Nita C. M. Pallett, Sarah L. Raubenheimer & Brad S. Ripley

3. Department of Ecology and Evolutionary Biology, Princeton University, Princeton, NJ, USA

   A. Carla Staver

4. Department of Biological Sciences, University of Cape Town, Cape Town, South Africa

   William J. Bond

5. Scientific Services, South African National Parks, Skukuza, South Africa

   Corli Coetsee

6. School of Natural Resource Management, Nelson Mandela University, George, South Africa

   Corli Coetsee

7. Earth System Science Department, Stanford University, Stanford, CA, USA

   Adam F. A. Pellegrini

8. Institute for Global Change Biology, School for Environment and Sustainability, University of Michigan, Ann Arbor, MI, USA

   Sarah L. Raubenheimer

Contributions

K.J.S., A.C.S., J.A.K., B.S.R., M.V.M. and C.P.O. conceptualized the study. K.J.S., A.C.S., J.A.K., C.C., N.C.M.P., M.V.M. and C.P.O. developed the methodology. K.J.S., A.C.S., J.A.K., M.V.M. and C.P.O. performed the investigation. K.J.S., A.C.S., J.A.K. and M.V.M. conducted visualization. K.J.S., A.C.S., B.S.R., M.V.M. and C.P.O. acquired funding. A.C.S., M.V.M. and C.P.O. provided supervision. K.J.S., A.C.S. and C.P.O. wrote the original draft of the manuscript. K.J.S., A.C.S., J.A.K., W.J.B., C.C., N.C.M.P., A.F.A.P., S.L.R., B.S.R., M.V.M. and C.P.O. reviewed and edited the manuscript.

Corresponding author

Correspondence to Kimberley J. Simpson.

Competing interests

The authors declare no competing interests.

Peer review information

Nature thanks David Ellsworth and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions