Abstract
Plants with the C4 photosynthesis pathway typically respond to climate change differently from more common C3-type plants, due to their distinct anatomical and biochemical characteristics. These different responses are expected to drive changes in global C4 and C3 vegetation distributions. However, current C4 vegetation distribution models may not predict this response as they do not capture multiple interacting factors and often lack observational constraints. Here, we used global observations of plant photosynthetic pathways, satellite remote sensing, and photosynthetic optimality theory to produce an observation-constrained global map of C4 vegetation. We find that global C4 vegetation coverage decreased from 17.7% to 17.1% of the land surface during 2001 to 2019. This was the net result of a reduction in C4 natural grass cover due to elevated CO2 favoring C3-type photosynthesis, and an increase in C4 crop cover, mainly from corn (maize) expansion. Using an emergent constraint approach, we estimated that C4 vegetation contributed 19.5% of global photosynthetic carbon assimilation, a value within the range of previous estimates (18–23%) but higher than the ensemble mean of dynamic global vegetation models (14 ± 13%; mean ± one standard deviation). Our study sheds insight on the critical and underappreciated role of C4 plants in the contemporary global carbon cycle.
Similar content being viewed by others
Introduction
C4 is one of the three photosynthetic pathways for terrestrial plants1 and is reported to account for 18–23%2,3,4 of global photosynthesis. C4 plants also drive wildfire dynamics in tropical and subtropical ecosystems5. C4 plants first evolved in the low atmospheric CO2 environment of the Oligocene Epoch, roughly 24–35 million years ago6. They developed distinct biochemical and anatomical characteristics to enrich CO2 concentration at the site of Rubisco carboxylation in leaves, thereby reducing photorespiration and enhancing carbon-fixation rates7. These characteristics produce different climate sensitivities in C4 plants compared to more prevalent C3 plants4,8, and thus are expected to cause a shift in C4 plant distributions and their contribution to global photosynthesis under contemporary and future climate change8,9,10.
Many previous studies have examined C4 plant responses to multiple environmental factors. A consensus is that since most C4 species originated in lower atmospheric CO2 concentrations11,12, they are expected to benefit less from rising CO2 concentrations compared to C3 plants. Meanwhile, higher temperatures are expected4 and reported13,14,15 to favor C4 over C3 photosynthesis, because the affinity of O2 to Rubisco relative to CO2 becomes stronger with increasing temperature and also due to differing solubilities of CO2 and O2 with increasing temperature. This should produce an advantage for the carbon concentrating mechanism of C4 species, especially under high temperatures16. Hence C4 species are characteristic of tropical and subtropical ecosystems. Correspondingly, since C4 photosynthesis is less limited by CO2 than C3 photosynthesis, it achieves a higher photosynthetic quantum yield and photosynthetic rates under high light, especially under high temperatures17. C4 species also should have a carbon assimilation advantage in arid environments18,19 due to their higher water use efficiency (i.e., less water loss through stomata for equivalent carbon gain) than C3 species, though under humid conditions this advantage could be limited9. Contemporary climate change, such as elevated CO2, rising temperatures, and changing rainfall patterns, can therefore lead to temporal and spatial shifts in the relative advantages of C4 to C3 photosynthesis. For instance, the differential response to a changing environment has been linked to observed woody plant encroachment in tropical Africa, where an increase in precipitation and elevated atmospheric CO2 levels are hypothesized to have caused a net decrease in C4 grassland distribution20,21. However, we currently lack a consensus on how the relative advantages of C4 to C3 photosynthesis change at the global scale, as regional studies have reported contrasting results and different driving factors—such as increased C4 grass distribution due to increased temperature22, decreased distribution due to elevated CO214 or no overall trend23. Understanding of how climate change has impacted C4 vegetation constitutes a major challenge due to its role in global photosynthesis and the terrestrial carbon cycle.
C4 vegetation overwhelmingly consists of natural grasses and crops using the C4 pathway. One prominent approach to estimate the distribution of C4 natural grasses is based on the crossover-temperature model, which predicts that a particular month is determined to favor C4 grasses over co-occurring C3 grasses when the mean daytime air temperature is >22 °C and precipitation in that same month is ≥25 mm2,24,25. This approach is based on each pathway’s relative carbon assimilation as a function of temperature, and thus the crossover temperature is dependent on atmospheric CO2 concentration with higher crossovers at higher CO2 levels. A few efforts to model C4 vegetation distribution have further incorporated the seasonality of precipitation26,27,28,29, or mean annual temperature and precipitation8,22,26, but so far they are only validated and applied at the regional scale. Some dynamic global vegetation models (DGVMs) allow adjustment of C3 and C4 grass distribution based on the difference between simulated C3 and C4 photosynthesis or the difference between their net primary productivity30, or based on the simulations from bioclimate distribution models in each time step, with the baseline C4 map acquired from remote sensing land cover classifications31,32. Some cohort-based DGVMs further consider competition for resources33 and disturbances34 when simulating C4 distributions. In general, current estimates of the distribution of C4 vegetation adopt a wide range of assumptions and generate rather different results10.
Uncertainty in global C4 grass distribution is further exacerbated by the lack of ground observations for validation and then for model extrapolation, since previous models often relied on either local datasets26,27 or literature reviews of C4 grass presence and absence2 for validation. This issue has become less prominent recently as some studies have used continental scale (i.e., North America) C4 plots25 and 13C isotopic records23,35 to validate C4 grass distribution models. Meanwhile, the distribution of C4 crops has been collated and estimated in some open datasets36,37,38,39. These datasets are based on Food and Agriculture Organization (FAO) census and national reporting of the harvested area for major C4 crops (i.e., maize, sorghum, millet, sugarcane), which comprised 24% of the global harvested area37, and are supplemented by total cropland area change from FAO and remote sensing40. In particular, the Land-Use Harmonization dataset version 2 (LUHv2) is the principal gridded land use dataset for the assessment of global carbon budgets36 and future climate change in CMIP641, in which C4 crop area over time is explicitly reported. Changes in global C4 crop distribution and the related contribution to global photosynthesis have yet to be evaluated, except for a few studies that have examined the C4 crop distribution for specific years2,42,43,
Here we quantify the global C4 vegetation distribution (including natural grasses and crops) and its contribution to global photosynthesis, as well as examine changes in C4 vegetation distribution over the past two decades. To do so, we use photosynthetic optimality theory to estimate the relative advantage of C4 to C3 photosynthesis over the global land surface, and then use the estimated difference in combination with observations to infer global C4 grass distribution. The optimality model includes a wide array of selective drivers for C4 grass distribution - CO2, temperature, light, aridity, nitrogen, and their interactions44 (see Methods), which are advances over previous crossover-temperature approaches which include CO2, temperature, and a precipitation threshold. The optimality model estimates the optimal leaf photosynthetic rate for C3 and C4 plants, along with optimal stomatal conductance and root/shoot carbon allocation based on growing season climate, with a target to maximize carbon gain with minimized water loss44. We further take advantage of multiple open-access databases of C4 species richness and coverage (i.e., the global TRY database45, a dataset for the contiguous United States (the DG dataset)23 and a subset of the Nutrient Network (NutNet)46), global grassland fraction maps from remote sensing (i.e., as the majority of C4 plant cover is non-woody47), in combination with the optimality model simulations to acquire data-constrained estimates of C4 grass distribution for the past 20 years. Meanwhile, we obtain and examine C4 crop distribution using multiple open datasets36,39. We further use an emergent constraint technique—a method to infer an unobservable variable from an observable variable based on the large spread of estimates of both variables from DGVMs (see Methods) –to estimate the contribution of C4 plants to global photosynthesis. By quantifying how C4 vegetation distribution and photosynthesis have changed over recent decades, our study improves understanding of historical changes in terrestrial photosynthesis and the global carbon cycle.
Results
C4 photosynthetic advantage and C4 grass coverage
We found a strong positive relationship between the observed C4 grass coverage (i.e., the % of grassland area covered by C4 grass species) and the relative advantage of C4 photosynthesis (AC4) over C3 photosynthesis (AC3) estimated by the optimality model (denoted as the AC4/AC3 - C4 coverage relationship hereafter; see Methods; Fig. 1b). With the increase in modeled AC4/AC3, the observed C4 coverage increased and then gradually plateaued. When AC4/AC3 = 1, C4 accounts for only 5.2% of grassland cover; when AC4/AC3 = 2.5, C4 coverage approaches 100% (Fig. 1b). Across global non-woody regions, AC4/AC3 ranged from 0.5 to 2.5, with a mean of 1.9 (Fig. 1a). Importantly, we obtained C4 coverage observations from multiple sources (i.e., TRY and DG; see Methods), which have different geographic representations and spatial resolutions. However, the AC4/AC3 - C4 coverage relationships are similar when using different observations (Fig. 1b), affirming the robustness of the relationship for C4 coverage estimation. Using the relationship between AC4/AC3 and C4 coverage (Fig. 1b), and the global AC4/AC3 estimated from the optimality model (Fig. 1a), we estimated the global C4 grass coverage (% of grassland covered by C4; Fig. 1c). We found C4 grass coverage followed a clear climatic gradient, and tended to be greater under warmer conditions (Fig. 1d).
a the ratio of C4 to C3 photosynthesis estimated by the optimality model (AC4/AC3) over global non-woody regions; (b) the relationship between observed C4 coverage (% of grassland) and estimated C4/C3 photosynthetic ratio by the optimality model. C4 coverage observation obtained from difference sources (i.e., TRY, DG datasets; please see methods); gray shaded area indicates the uncertainty range for the relationship between AC4/AC3 and C4 coverage (i.e., 95% confidence interval). The black line represents the regression using both the TRY and DG datasets, while the red and blue dash lines represent the regression using either the TRY or the DG dataset. c C4 grass coverage (% of grassland) over the globe, which can be regarded as the potential C4 area abundance when grassland covers 100% of the land surface; (d) C4 coverage in a climate space of mean annual temperature (MAT: °C) and mean annual precipitation (MAP: mm/yr).
The global distribution of C4 vegetation
After predicting the C4 grass coverage (% of grassland covered by C4 grasses; Fig. 1c), we overlaid a global grassland fraction map from remote sensing (see Methods and Fig. S1) to estimate the actual C4 natural grass area abundance (% of the land surface covered by C4 grasses; Fig. 2a). From 2001 to 2019, C4 natural grass accounted for 14.8 ± 1.3% (mean ± one standard deviation) of the non-frozen land surface area (Fig. 2a), while C4 crops accounted for 2.8 ± 0.3% (Fig. 2c). The total estimated C4 area abundance was 17.5 ± 1.4% (Fig. 2e). There were several C4 natural grass hotspots (i.e., >30% C4 area abundance) across continents (Fig. 2a): the Great Plains in North America, the savannas in Southern Brazil, the savannas in Africa, the grasslands in Central Asia, and Northern Australia. Meanwhile, we found the main C4 crop zones were in central North America, the Sahel region, and the west coast of India (Fig. 2c). The disagreement between remote sensing-based grassland fraction maps (Fig. S4), along with the uncertainty in the AC4/AC3 - C4 coverage relationship (Fig.1b), incurred uncertainties in the C4 natural grass distribution (Fig. 2b)—the uncertainty typically ranges between 1 and 3% of the land surface area, though in regions like Australia the uncertainty could be as high as 6–7% (Fig. 2b).
The area occupied by (a) C4 natural grasses, (c) C4 croplands and (e) all C4 vegetation (unit: % of the land surface). The uncertainties of the area abundance of (b) C4 natural grasses, (d) C4 croplands and (f) all C4 vegetation (unit: % of the land surface).
The changes in C4 vegetation distribution
Based on our simulation of C4 natural grass distribution for the past two decades and the C4 cropland distribution from the LUHv2 dataset (see Methods), we found the overall area of C4 vegetation decreased from 17.7 ± 1.4% (mean ± one standard deviation) in 2001–2005 to 17.1 ± 1.4% in 2015–2019, as a net effect of a decrease in C4 natural grasses from 15.0 ± 1.3% to 14.2 ± 1.3%, and an increase in C4 crops from 2.6 ± 0.3% to 3.0 ± 0.3% (Fig. 3b). The change in C4 shows large spatial heterogeneity (Fig. 3a). In particular, C4 natural grass area decreased all over the globe, except for the central Europe and parts of the western U.S. (Fig. 3c). C4 crop area increased in most parts of the world, except for central North America where there was the largest decrease, and Europe where there were slight decreases (Fig. 3e).
Spatial distributions of changes in (a) total C4 vegetation, (c) C4 natural grasses and (e) C4 crops from 2001 to 2019; b The changes in the total area of C4 vegetation, C4 natural grasses and C4 crops, in percentages of global vegetated land surface; d the synergies of changes in C4 natural grasses and C4 crops, where ++ means both C4 natural grasses and C4 crops area abundance increased, − − means both decreased, + − means C4 natural grasses increased and C4 crops abundance decreased, − + means the opposite; f the drivers for the change in C4 natural grasses and C4 crops area abundances. Climate drivers include atmospheric CO2 concentration, air temperature (Tair), vapor pressure deficit (VPD) and soil moisture (SM) during the growing seasons. In (b), the uncertainty for C4 crop is 10% of the C4 crop area reported, the uncertainty for C4 natural grass area is the combination of the uncertainty in remote sensing-based grassland fraction and the uncertainty of the AC4/AC3 - C4 coverage relationship (Fig. 1b), and uncertainty for C4 vegetation is the combination of uncertainties of C4 crop and C4 natural grass areas.
The increase in C4 crop area was often at the expense of decreasing C4 natural grasses, as we found that in regions where there were both C4 crops and C4 natural grasses, more than 50% of the regions showed C4 natural grasses decreased but C4 crops increased. Meanwhile, only 14% showed that C4 crops and C4 natural grasses increased simultaneously, 28% showed they both decreased, and only 7% of the region showed that C4 natural grasses increased and C4 crops decreased (Fig. 3d). Our attribution analysis suggested that elevated CO2 was the dominant reason for the decrease in C4 natural grass distribution, while the impacts of temperature and water stress (i.e., soil moisture and vapor pressure deficit) were positive (i.e., increase C4 coverage) over the study period (Fig. 3f). Most of the increase in C4 crops, as we analyzed from another independent dataset on major C4 crop distributions32, came from the expansion of maize in South America and eastern Europe (Fig. 3f; Fig. S9; see Methods).
The contribution of C4 vegetation to global photosynthesis
The changes in the C4 area can cause associated changes in total C4 photosynthesis, thus impacting global carbon cycle dynamics. Current ensemble of DGVMs predicted that C4 vegetation contributed from 2% to 40% of global photosynthesis, on 7% to 23% of the global vegetated land surface area (Figs. S6, S7). The large spread of model estimates indicates the various assumptions adopted in C4 vegetation distribution and potentially the different parameterizations for C4 photosynthesis in DGVMs. Despite these wide inter-model variations, it is possible to infer emergent constraints on C4 vegetation contributions to the carbon cycle. To quantify the contribution of C4 photosynthesis to global photosynthesis, we established an emergent constraint (p < 0.01) between the DGVM-simulated occupied area and percentage contribution of C4 natural grasses and crops to global photosynthesis, respectively. We found that with a 1% increase in area, C4 natural grass contribution to global photosynthesis increased by 1.10% (Fig. 4a), while the contribution of C4 crops increased by 1.16% (Fig. 4b). We also conducted a grid cell-level emergent constraint analysis and acquired similar ranges of slopes (Fig. S10). The lower coefficient of emergent constraint for C4 grass (i.e., 1.10) than the coefficient for C4 crop (i.e.,1.16) suggests that croplands tend to have a higher ecosystem photosynthetic rate compared to grasslands over the same area.
The emergent constraints between the percentage of area occupied and the percentage of global photosynthesis contributed by (a) C4 natural grasses and (b) C4 crops, based on the estimates from an ensemble of the DGVMs; c changes in emergent constraint coefficients (i.e., the slopes of the linear regressions in (a, b) from 2001 to 2019 for C4 natural grasses and C4 crops). The uncertainties in (a, b, c) were quantified as one standard error (i.e., SE) by bootstrapping models when getting emergent constraint; d The contributions of C4, C4 natural grasses and C4 crops to global GPP from 2001 to 2019, whereas the uncertainty was quantified as one SE by bootstrapping the uncertainty range of C4 areas and the uncertainty range of the emergent constraint coefficients.
We further explored the changes in the coefficients of emergent constraints over the past two decades. We note that for C4 grasslands and C4 croplands, the coefficients all slightly decreased in the past two decades. The coefficient of C4 crops decreased from 1.16 to 1.15, while the coefficient of C4 grasses decreased from 1.11 to 1.10 from 2001 to 2019 (Fig. 4c). Interestingly, the coefficients are all greater than 1, highlighting that the per unit area photosynthetic rate of C4 is generally higher than that of the remaining C3 vegetation. With the likely increase of global photosynthesis in recent decades, the decreasing coefficients of C4 indicated that C4 photosynthesis increased at a slower pace than other (mostly C3) vegetation. By applying the estimated area of C4 to the annual emergent constraint coefficients (Fig. 4c), we found that the global C4 natural grass contribution to photosynthesis decreased from 16.5 ± 1.5% (mean ± one standard deviation) in 2001–2005 to 15.5 ± 1.5% in 2015–2019, the C4 crop contribution to photosynthesis increased from 3.0 ± 0.3% to 3.4 ± 0.4%, and in total the C4 contribution to global GPP decreased from 19.7 ± 1.9% to 19.0 ± 1.9% (Fig. 4d). The reported value is greater than the ensemble mean reported by the DGVMs (14 ± 13%), and is within the range of previously modeled values (18–23%2,3,4).
Discussion
In this study, we estimated the global C4 vegetation distribution and quantified the changes in C4 vegetation distribution and photosynthesis over the past two decades, using an optimality photosynthesis model, photosynthetic pathway records from global/regional databases and remote sensing observations. On average, from 2001 to 2019, C4 plants occupied 17.5 ± 1.4% of the global vegetated surface and contributed 19.5 ± 1.9% of global photosynthesis, within the range of previous estimates (i.e., 18–23% for photosynthesis) but are greater than the estimates from the ensemble mean of DGVMs (13 ± 8% for area and 14 ± 13% for photosynthesis). C4 total area and C4 contribution to global photosynthesis both decreased over this period (i.e., 0.6% of the land surface and 0.7% of global GPP), which resulted from the increases in C4 crop area and its contribution to global photosynthesis, and the decreases in C4 natural grass area and its contribution to global photosynthesis.
Our study suggests that the decrease in C4 natural grass distribution was primarily driven by elevated CO2, in accordance with previous theoretical and experimental works that showed C4 advantage in carbon assimilation over C3 decreased with rising CO248,49,50. There is also evidence showing many grassland and savanna areas have been invaded by C3 woody species, with increased atmospheric CO2 proposed as a major driver for the encroachment51,52. The decrease in the emergent constraint coefficients demonstrated that the impact of elevated CO2 on C4 and C3 were included in most DGVMs (Fig. 4c). Evidence for the historical expansion of C4 over geological time scales seems to support our conclusion, in particular for Africa where CO2 dominates the C4 grassland expansion or decline50,53,54, while in Central Asia55, Australia56, central China57 and central US grasslands58 reports show hydroclimatic change impacted C4 grass distribution. Our results also show that the changes in soil moisture and VPD over the past two decades largely induced positive impacts on C4 grass distribution, though globally their impacts were unable to offset the negative changes driven by elevated CO2 (Fig. 3f; Fig. S8).
To assess the distribution of C4 vegetation, we noted that there were three types of abundance used in previous literatures: for modeling practice, we often used the area abundance, that is, the percentage of the land surface occupied by C4 plants. However, the area abundance was challenging to observe over large scales, and it was rare to have direct observations of large-scale area abundance other than approximations from remote sensing32. Most observations of C4 presence and cover are at the species level and, therefore, observation-based studies often report the relative species richness of C4 plants13,26. A few studies reported biomass abundance, i.e., the percentage of local biomass contributed by C4 plants, and biomass abundance is often much higher than the species abundance27,29, echoing some studies suggesting that species abundance should be used with caution to infer biomass abundance and productivity56. In this study, we developed a conversion factor to translate C4 species richness into C4 area abundance, using plot-level concurrent measurements of both from NutNet. We found 1% increase in C4 species richness led to 1.51 ± 0.15% in C4 area abundance (Fig. S3). The conversion factor enabled us to translate global observations of C4 species richness into C4 coverage to develop the AC4/AC3 - C4 coverage relationship in our study (Fig. 1b, S1).
Compared to the previously estimated C4 vegetation distribution from a crossover-temperature model2 (Fig. S5), our study provided a similar area occupied by C4 vegetation (20.5 million km2 versus 21.1 million km2). We find a lower estimate for Africa, where our estimate of C4 area abundance showed a more nuanced gradient compared to the estimate from the crossover-temperature model (Fig. S5). We suspect this is partly due to their difference in relating C4 photosynthetic advantage to C4 grass coverage (% of grassland covered by C4 grasses) – while our study used the AC4/AC3 - C4 grass coverage relationship to gradually adjust C4 grass coverage, the crossover-temperature model assumes all grasslands in the pixel are C4 grassland as long as the monthly climate satisfies the crossover criteria (e.g., mean daytime air temperature is >22 °C and precipitation in that same month is ≥25 mm). Therefore, for regions where monthly climate meets the crossover criteria (e.g., sub-Sahel Africa), the crossover-temperature model tends to estimate 100% C4 grass, however, it was not the case for the optimality approach in those regions (Fig. 1c). Meanwhile, we estimated considerably higher C4 grass cover in Central Asia, which is consistent with the reported prevalence of C4 species in the region26,59 – the high abundance in these inland regions were potentially due to harsh environments characterized by high maximum temperature and aridity levels, which favor the growth of C4 plants over C3. We also note that our approach may underestimate C4 grass distribution in some mesic savanna ecosystems—such as the longleaf pine savannas in the Southeastern US—where a C4 understory exists beneath a C3 canopy60,61. The underestimation is likely because the optimality model predicted no photosynthetic advantage for C4 plants over C3 under low light conditions in understory (Fig. 1a), and the remote sensing products reported low grassland fraction in the region (Fig. S4).
In our study, we estimated the annual distribution of C4 grasses using the growing season mean climate; however, many locations have seasonal shifts between C4 and C3 grass dominance depending on seasonal climate variations4. For example, in the grasslands of southeast Australia, a recent study suggests C4 dominance is the highest in summer when there is high temperature and low precipitation, while other seasons have more C3 vegetation28. Therefore, for modeling the seasonal variation of carbon fluxes from seasonal changes in C4 grass distribution, the crossover-temperature approach based on monthly climate and weighted by a vegetation index like NDVI could be more useful2. We acknowledge that the distribution of our observations was not uniform across the globe, with North America being better represented compared to other regions (Fig. S2). As an additional test to validate our estimation of C4 vegetation distribution, we compared the C4 grass coverage estimated in our study (Fig. 1c) with the C4 coverage estimated from isotopic measurements and remote sensing in Australia62 (Fig. S11). This validation demonstrates a strong agreement (r = 0.69, p < 0.01) between the two independent estimates, affirming the robustness of our estimates in under-sampled regions.
We also need to highlight that the optimality approach was based on the assumption that C4 grass distribution is determined by the photosynthetic advantage of C4 compared to C3, and the photosynthetic advantage is largely dependent on local climate44. While this assumption is also adopted by the crossover-temperature model, it neglects the role of grass phylogeny in determining C4 grass distributions. Some studies have suggested that grass clades (i.e., Pooideae for C3, and PACCMAD for C3 and C4) are perhaps more critical than photosynthetic pathway for determining C4 and C3 grass distributions, at least along temperature gradients19,63,64. This lack of consideration on phylogeny in C4 grass distribution models may impact predictions of future distributions, as C3 grasses from certain clades have less competitive disadvantage compared to C4 grasses in a warmer world.
Other environmental changes that can impact C4 grass distribution include fire and nitrogen deposition. Fire can influence C4 coverage and productivity either through creating open canopies for light capture by C4 plants or as a characteristic of semi-arid environments that provide a photosynthetic advantage for C465. Recent woody plant encroachment suggests fire had a central role51,66 in the formation of grasslands and the rise of C4 dominant grasslands in late Neogene67 and late Miocene68. For recent decades, since globally the trend of fire occurrence is still very uncertain with strong regional variations in the trend69, we were unable to quantify its impact on C4 grass distribution. Neither the crossover-temperature C4 model, nor the optimality model we used, incorporates the role of fire in C4 dynamics at present. However, since our approach used annual grassland fractional maps based on satellite remote sensing, the impacts of fire at the annual scale were implicitly considered. Some DGVMs have incorporated fire-relevant processes, however, we found they estimated lower C4 grass abundance (i.e., 9.6 ± 6.7%) than those models that do not include fires (i.e.,11.8 ± 3.2%). Additionally, since C4 plants have higher photosynthetic nitrogen use efficiency than C370, anthropogenic nitrogen deposition71 might have impacted the relative advantage of C4 to C3 photosynthesis. Previous studies have suggested C4 plants tend to have a higher photosynthetic rate than C3 across a spectrum of nitrogen supply - meaning C4 plants have photosynthetic advantage on infertile soils and the advantage will be enhanced by increased nitrogen availability, which can be used to increase C4 leaf area72. In this study, we used a data-driven product of leaf nitrogen content and remote sensing leaf area index to simulate C4 grass distribution (see Methods), which may have implicitly accounted for the effect of nitrogen deposition or limitation on C4 photosynthesis.
In this study, we used LUHv2—the primary dataset used in current global carbon cycle modeling and climate forecasting36,41—to examine the changes in C4 cropland and reported an increase in C4 cropland area. However, we note a key source of uncertainty in this dataset: the historical simulation of C4 crop distribution used a constant fraction of C4 crop cover for each global grid cell, based only on observations circa 200041. Therefore, the increase in C4 crop area in LUHv2 could just reflect an increase in all croplands rather than real C4 expansion. To reduce the uncertainty caused by this issue, we used another cropland dataset39, which dynamically simulated the area of 17 major crop types (including main C4 crops maize, millet and sorghum) based on annual FAO census of crop harvested areas. This analysis confirmed our results regarding the increase in C4 cropland mainly due to the expansion of maize (Fig. 3f), with a similar spatial pattern reported (Fig. S9). However, the sum of the three main C4 species only increased by 0.1%, relatively lower than what we see from the LUHv2 dataset (i.e., 0.4%). We can conclude there was an increase in C4 cropland, though the magnitude of increase should be subject to further examinations.
In conclusion, we used a combination of plant photosynthetic pathway records, remote sensing, and an optimality-based photosynthesis model to estimate the global C4 coverage and the magnitudes of C4 photosynthesis and their variations over the past two decades. We infer that C4 vegetation covered on average 17.5% of the global land surface over the period from 2001 to 2019, while C4 grass cover decreased due to elevated CO2 and C4 crop cover increased because of corn (maize) expansion. We predict that C4 photosynthesis accounted for 19.5% of the global total photosynthesis, with an increased contribution from C4 crops and a decrease from C4 natural grasses during this period. Our study offers an updated and more observationally constrained estimate of C4 vegetation distribution and photosynthesis, thereby improving our understanding of potential future C4 changes and enhancing the quantification of the global carbon budget.
Methods
The overarching framework
The distribution of C4 vegetation overwhelmingly consists of C4 natural grasses and C4 crops. To estimate the C4 natural grass distribution, we first used an optimality photosynthesis model44 to simulate the optimal photosynthetic assimilation rates of C4 and C3 plants (noted as AC3 and AC4, respectively) using 0.5 × 0.5 degree gridded historical climate (i.e., CRU-JRA2020), soil73 and leaf nitrogen content74. We calculated the ratio of AC4 to AC3, and established a statistically significant (p < 0.01) relationship between AC4/AC3 and the observed C4 coverage from multiple databases—the TRY database45 and the DG dataset25 based on an assumption that larger AC4/AC3 indicates higher C4 grass coverage (% of grassland covered by C4 grasses). Using the AC4/AC3 - C4 coverage relationship, we estimated the C4 grass coverage (a.k.a. potential C4 abundance when grasses cover 100% of the land surface) from estimated AC4/AC3 for the globe. We lastly overlaid the C4 coverage map to a global map of grassland fraction from remote sensing to acquire actual C4 grass abundance (% of the land surface covered by C4). The workflow is presented in Fig. S1. Meanwhile, for C4 crop distribution, we directly used the estimates from the LUHv2-2019 dataset, in which C4 crop distribution is estimated from FAO survey and satellite remote sensing.
Processing observational C4 records
We acquired 61,588 georeferenced records of photosynthetic pathways from the TRY database (last accessed 2022 June), among them, there were 2269 records of C4. The time range of the record covers roughly the past 50 years. We first removed the woody species from the records, based on species names and an index table from the TRY database (https://www.try-db.org/TryWeb/Data.php#3), as our study aimed to examine C4 grass distribution and the global cover of C4 forests is not extensive47. After the step, we kept 13,919 records for non-woody species, among which 1963 were C4. We further removed 82 records that belong to major C4 crops (i.e., maize, sugarcane, millet, and sorghum) and kept 1881 records. We then aggregate these records to 10 × 10 degree cells, in each cell we calculate the species richness of C4 (i.e., number of C4 species/total number of herbaceous species, the numbers were derived from the available records in the TRY database). The gridded values of C4 species richness would be further used to constrain the optimality model to estimate global C4 grassland coverage. Here we use the large-size grid cell to make sure there were enough samples in each cell to acquire a meaningful estimate of C4 abundance – in this analysis, each cell should have at least 50 species (i.e., C3 and C4 in total). We used 1619 (out of 1881) C4 species records in this aggregation step, and obtained 23 10 × 10 degree cells for the analysis (Fig. S2).
Note here the C4 abundance from the TRY database was species richness, not equal to the area abundance that is often used in DGVMs. To acquire C4 grass coverage (% of grassland covered by C4) from C4 species richness (% of grass species that is C4), we used an open dataset from the global nutrient network (NutNet) that has paired C4 species richness and C4 area abundance (% of the land surface covered by C4) to infer their relationship46 (Fig. S3). The dataset includes species-specific coverage records as well as the grass species richness data collected in 25 m2 plots across 34 sites. Each site has between 1 and 6 control plots. We only used the data from the control plots, excluding plots that underwent nutrient addition treatments. To avoid the uneven distribution of data samples, we grouped the paired observations by their C4 species richness, and for each species richness we get a mean C4 area abundance and the standard deviation of the C4 grass coverage. We then conducted 1000 linear fittings (i.e., with an intercept of 0, since C4 grass coverage should be 0 when C4 species abundance is 0), and for each fitting we used randomly sampled C4 grass coverage values (i.e., based on mean and the standard deviation) value against C4 species richness values. The slopes of the linear regressions represented a conversion factor between C4 species richness and C4 grass coverage (Fig. S3).
In addition to the TRY database that has a global representation, we also used a gridded C4 grass coverage data compiled for the contiguous United States (denoted as the DG dataset)23. The DG dataset provides C4 grass coverage (% of grassland) aggregated at a 100 km resolution grid, which was sampled from roughly 40,000 plots over the past 40 years. Please note that the DG dataset only surveyed C4 grass species. We used the DG dataset and the TRY database to establish the relationship between AC4/AC3 and C4 grass coverage (Fig. 1b).
Processing cropland and land use data
We used a gridded C4 crop distribution from the LUHv2 dataset (version: LUHv2-GCB2019)36. It was estimated based on the FAO census and national reporting of >170 major crop types (including the main C4 types), supplemented by the total cropland area collated by HYDE3.240 which came from FAO census and remote sensing products. The C4 crop fraction of each grid cell was only acquired based on observations circa 200037 and the fraction was kept constant over the study period. The most recent versions of the LUHv2 dataset have been used in CMIP6 for the IPCC AR6 report and the global carbon budget. Since the LUHv2 dataset did not contain an estimate of uncertainty, we relied on an independent study that compared four different land use products (including LUHv2) and reported the uncertainty of cropland estimation between products was about 10%75. We thus used 10% to represent the uncertainty range of the C4 cropland area.
In addition to LUHv2, we used another open dataset reporting the area of 17 main crop types from 1961 to 201439. Unlike the LUHv2 dataset which almost exclusively relied on observations circa 2000 to quantify the C4 crop fraction, this other dataset used annual FAO census records for crop area fraction estimation, including those of the three primary C4 crops: maize, millet, and sorghum. We used this dataset to examine the changes in global C4 croplands and compare to the values obtained from the LUHv2 dataset.
The optimality model for C4 and C3 photosynthesis
We used optimal C3 and C4 photosynthesis models to simulate optimal C3 and C4 photosynthesis44. The soil-plant-air water continuum was incorporated in C3 photosynthesis models76 and C4 photosynthesis models77 to examine interactions of CO2, water availability, light and temperature. The model considered optimal stomatal resistance and leaf/fine-root allocation to maximize the carbon gain regarding water loss, and successfully predicted the ancient distribution of C4 species in Oligocene and Miocene44.
In the current study, we improved the modeling processes through the following aspects. (1) We used different parameters for C3 and C4 species specifically to better represent the diversity of C3 and C4 species variability (Table S2 in the Supplementary Note). (2) We considered the effects of nitrogen availability and optimal nitrogen allocation between C3 and C4 species. Specifically, we adjusted the maximum carboxylation rate (Vcmax) and maximum electron transport rate (Jmax) values using optimal Jmax/Vcmax ratio (i.e., 2.1 for C3 and 5.0 for C4, which were supported by both measurements and theoretical modeling)78,79 and available leaf nitrogen content for C3 and C4 respectively. (3) Since a large majority of C4 species are herbaceous, when we modeled closed canopy biomes (e.g., those pixels dominated by tree and shrubs), we used estimates of understory photosynthetic active radiation (PAR) to model the relative advantage of the herbaceous species. A full model description and the parameterization is in the Supplementary Note.
Using the models, we were able to calculate the optimal assimilation rates for C3 and C4 (i.e., AC3 and AC4) over the globe at the 0.5-degree resolution (i.e., dependent on the spatial resolution of climate input), where the relative advantage of C4 to C3 is defined as AC4/AC3. The simulation was conducted at an annual time step and there was no need for model initialization. When establishing the relationship between AC4/AC3 and C4 grass coverages (Fig. 1b) from the DG and TRY datasets, we aggregated the simulations from 0.5-degree to 1-degree (approximately 100 km at the equator) and 10-degree. As the relationships derived from both 10-degree (i.e., TRY) and 1-degree (i.e., DG) data were similar, we assumed that the relationship is scale-independent. Consequently, we applied it to 0.5-degree estimates of AC4/AC3 to infer global C4 grass coverage. We also assumed that the relationship between AC4/AC3 and C4 grass coverage was time-invariant.
Running the optimality model
We used annual growing season average soil water potential, vapor pressure deficit (VPD), 2 m daytime air temperature (Tair), photosynthetic active radiation (PAR) and leaf nitrogen content as inputs for the optimality photosynthesis model. The growing season was defined using the MODIS phenology product (MCD12Q2)80.
Tair was acquired from the CRU-JRA2020 dataset. VPD was estimated using the specific humidity and air temperature from CRU-JRA2020. Soil water potential was estimated from soil texture properties from soil grids and global soil water content datasets, using the Clapp & Hornberger equation81. The global soil water content datasets came from GLEAM v382. To avoid extreme low soil water potential that does not allow plant growth in the optimality model, we set the minimal soil water potential to −3 MPa. The leaf nitrogen content was acquired from a machine learning upscaled leaf traits product74. PAR was also acquired from the CRU-JRA2020 dataset, which is a reanalysis from CRU83 and JRA84. We directly used PAR for ‘open’ ecosystems (i.e., grasslands, savannas), however, for dense forests and shrublands we used understory PAR (i.e., as C4 grasses often exist in understories), which was derived from PAR and multi-year average MODIS LAI85 (i.e., assume they are overstory LAI) following a radiation gradient mandated by the Beer’s Law.
We ran the photosynthetic optimality models multiple times in the process. We first modeled the growing season AC4/AC3 in the study period using the climatology of the variables mentioned from 2001 to 2019. To model the growing season AC4/AC3 from 2001 to 2019, for each year we used the 20-year climatology (i.e., 20 years before the target year) of the driving variables. In addition, we also conducted simulation for four scenarios, in which we replaced the climate input for 2001 simulation with the CO2, Tair, VPD and soil moisture from 2019, respectively. Then we used AC4/AC3 to estimate the C4 grass distribution for each year or for each scenario. By calculating the difference between the C4 grass distribution of four scenarios and the C4 grass distribution in 2001, we quantified the contribution of CO2, Tair, VPD and soil moisture to the changes in C4 grass distribution.
Remote sensing estimates of global grassland fraction
Multiple remote sensing products provide information on grassland distributions. Some directly provide continuous fraction values (i.e., GLC86 at 100 meter and Dynamic World87 at 10 meter) and some provide categoric information on grassland and savannas (i.e., MODIS88 at 500 meter and ESA-CCI89 and 300 meter). For the former, we can directly calculate the grassland fraction value at 0.5-degree resolution; for the latter, we assign the grassland/savanna type pixel to 100% and others to 0% grassland, and then obtain the mean value for each 0.5 grid cell. We found that those four estimates of grassland fraction vary considerably (Fig. S4). Based on a visual comparison of the four estimates (i.e., GLC, Dynamic World, MODIS and ESA-CCI) against vegetation map estimates90, we found Dynamic World and ESA-CCI substantially underestimate grassland fraction. We therefore used only MODIS and GLC estimates of grassland fractions in our study.
The MODIS grassland fraction product is available from 2001 to 2019. The GLC product is only available from 2015 to 2019. To extend the GLC product back to 2001, we employed a random forest approach to estimate GLC estimates based on surface reflectance, climate, and soil type and extrapolate it to 2001 (i.e., the training accuracy is 99% and the validation accuracy is 95%). We used the average of MODIS and GLC estimates to represent the grassland fraction. To quantify the uncertainty of the approach, for each pixel we bootstrapped 1000 times between the MODIS estimate and the GLC estimate, and use the one standard deviation of these 1000 values to represent the uncertainty in grassland fraction.
Dynamic Global Vegetation Models (DGVMs)
We used 11 DGVMs participating in the global carbon project91(Table S1) in our study. Though all of the 11 DGVMs provided simulations for C4 natural grasses, only 7 of them have simulations for C4 crops (Table S1). We established an emergent constraint between C4 area and C4 photosynthesis contribution using the estimates from the model ensemble. We used the S3 scenario (i.e., considering elevated CO2, climate change and land use change) of model simulations in our analysis.
Emergent constraint approach
The emergent constraint technique is widely used in climate and modeling communities to infer unobserved quantities of interest in land surface processes92,93. The underlying assumption is that although there is a large spread in the model estimates of an observed variable X and an unobserved variable Y across models, the relationship linking the two is tightly constrained across models. Based on the strong and robust relationship across models between X and Y, observations of X can be used to generate a constraint on unobserved Y. This approach has been termed ‘emergent’ because the functional relationship cannot be diagnosed from a single model, but rather emerges from the spread of the model estimates. The emergent constraint identified in this study links the contribution of C4 grasses/crops to total GPP to the percentages of area covered by C4 grasses/crops as estimated from the ensemble of DGVM simulations.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
The global C4 vegetation distribution map is available at https://zenodo.org/records/10516423. The CS C4 map was acquired from https://daac.ornl.gov/cgi-bin/dsviewer.pl?ds_id=932. The CRU TS4.02 climate data is available at https://crudata.uea.ac.uk/cru/data/hrg/, the soil moisture data can be downloaded from https://www.gleam.eu/#datasets. The global dataset of leaf photosynthetic pathway was acquired from the TRY database https://www.try-db.org/TryWeb/Home.php, by selecting those records with the field “photosynthesis pathway (traitID: 22)”. The DG dataset was obtained from the supporting information of https://onlinelibrary.wiley.com/doi/10.1111/jbi.13061. The subset of the observations from the nutrient network (NutNet) are accessible at https://portal.edirepository.org/nis/mapbrowse?packageid=edi.1037.2.
Code availability
The code for analysis is available at https://github.com/lxzswr/C4distribution/ and https://zenodo.org/records/10516423. The code of optimality photosynthesis model is available at https://github.com/zhouhaoran06/C3C4OptPhotosynthesis-. All maps in the study were generated with the assistance of the ‘M_Map’ package in Matlab, the instruction of which is available at https://www.eoas.ubc.ca/~rich/map.html.
References
Lambers, H., Chapin, F. S. & Pons, T. L. Plant Physiological Ecology. (Springer New York, 2008). https://doi.org/10.1007/978-0-387-78341-3.
Still, C. J., Berry, J. A., Collatz, G. J. & DeFries, R. S. Global distribution of C3 and C4 vegetation: Carbon cycle implications. Glob. Biogeochem. Cycles 17, 6-1–6–14 (2003).
Lloyd, J. & Farquhar, G. D. 13C discrimination during CO2 assimilation by the terrestrial biosphere. Oecologia 99, 201–215 (1994).
Ehleringer, J. R., Cerling, T. E. & Helliker, B. R. C4 photosynthesis, atmospheric CO2, and climate. Oecologia 112, 285–299 (1997).
Randerson, J. T. et al. Fire emissions from C3 and C4 vegetation and their influence on interannual variability of atmospheric CO2 and δ 13 CO2: C4 FIRE EMISSIONS. Glob. Biogeochem. Cycles 19, n/a-n/a (2005).
Sage, R. F. The evolution of C 4 photosynthesis. N. Phytol. 161, 341–370 (2004).
Christin, P. & Osborne, C. P. Tansley review: the evolutionary ecology of C4 plants. N. Phytol. 204, 765–781 (2014).
Collatz, G. J., Berry, J. A. & Clark, J. S. Effects of climate and atmospheric CO2 partial pressure on the global distribution of C 4 grasses: present, past, and future. Oecologia 114, 441–454 (1998).
Baldocchi, D. The grass response. Nature 476, 160–161 (2011).
Still, C. J., Cotton, J. M. & Griffith, D. M. Assessing earth system model predictions of C4 grass cover in North America: from the glacial era to the end of this century. Glob. Ecol. Biogeogr. 28, 145–157 (2019).
Pagani, M., Zachos, J. C., Freeman, K. H., Tipple, B. & Bohaty, S. Marked decline in atmospheric carbon dioxide concentrations during the paleogene. Science 309, 600–603 (2005).
Edwards, E. J. et al. The origins of C4 grasslands: integrating evolutionary and ecosystem science. Science 328, 587–591 (2010).
Bremond, L., Boom, A. & Favier, C. Neotropical C3/C4 grass distributions - present, past and future. Glob. Change Biol. 18, 2324–2334 (2012).
Morgan, J. A. et al. C4 grasses prosper as carbon dioxide eliminates desiccation in warmed semi-arid grassland. Nature 476, 202–205 (2011).
Teeri, J. A. & Stowe, L. G. Climatic patterns and the distribution of C4 grasses in North America. Oecologia 23, 1–12 (1976).
Ehleringer, J. & Björkman, O. Quantum yields for CO 2 Uptake in C3 and C4 plants: dependence on temperature, CO2, and O2 concentration. Plant Physiol. 59, 86–90 (1977).
Ehleringer, J. R. Implications of quantum yield differences on the distributions of C3 and C4 grasses. Oecologia 31, 255–267 (1978).
Young, H. J. & Young, T. P. Local distribution of C3 and C4 grasses in sites of overlap on Mount Kenya. Oecologia 58, 373–377 (1983).
Edwards, E. J. & Still, C. J. Climate, phylogeny and the ecological distribution of C4 grasses. Ecol. Lett. 11, 266–276 (2008).
Sage, R. F. A portrait of the C4 photosynthetic family on the 50th anniversary of its discovery: species number, evolutionary lineages, and Hall of Fame. J. Exp. Bot. 67, 4039–4056 (2016).
Stevens, N., Lehmann, C. E. R., Murphy, B. P. & Durigan, G. Savanna woody encroachment is widespread across three continents. Glob. Change Biol. 23, 235–244 (2017).
Epstein, H. E. et al. The relative abundance of three plant functional types in temperate grasslands and shrublands of North and South America: effects of projected climate change. J. Biogeogr. 29, 875–888 (2002).
Griffith, D. M., Cotton, J. M., Powell, R. L., Sheldon, N. D. & Still, C. J. Multi‐century stasis in C3 and C4 grass distributions across the contiguous United States since the industrial revolution. J. Biogeogr. 44, 2564–2574 (2017).
Wei, Y. et al. The north american carbon program multi-scale synthesis and terrestrial model intercomparison project - Part 2: environmental driver data. Geosci. Model Dev. 7, 2875–2893 (2014).
Griffith, D. M. et al. Biogeographically distinct controls on C3 and C4 grass distributions: merging community and physiological ecology: climate disequilibrium in C4 grass distributions. Glob. Ecol. Biogeogr. 24, 304–313 (2015).
Wang, R. & Ma, L. Climate-driven C4 plant distributions in China: divergence in C4 taxa. Sci. Rep. 6, 27977 (2016).
Winslow, J. C., Hunt, E. R. & Piper, S. C. The influence of seasonal water availability on global C3 versus C4 grassland biomass and its implications for climate change research. Ecol. Model. 163, 153–173 (2003).
Xie, Q. et al. Satellite-observed shifts in C3/C4 abundance in Australian grasslands are associated with rainfall patterns. Remote Sens. Environ. 273, 112983 (2022).
Paruelo, J. M. & Lauenroth, W. K. Relative abundance of plant functional types in grasslands and shrublands of North America. Ecol. Appl. 6, 1212–1224 (1996).
Fox, D. L. et al. Climatic controls on C4 grassland distributions during the neogene: a model-data comparison.Front. Ecol. Evol. 6, 147 (2018).
Krinner, G. et al. A dynamic global vegetation model for studies of the coupled atmosphere-biosphere system. Glob. Biogeochem. Cycles 19, 1–33 (2005).
Bonan, G. B., Levis, S., Kergoat, L. & Oleson, K. W. Landscapes as patches of plant functional types: an integrating concept for climate and ecosystem models: plant functional types and climate modELS. Glob. Biogeochem. Cycles 16, 5–1-5–23 (2002).
Smith, B. et al. Implications of incorporating N cycling and N limitations on primary production in an individual-based dynamic vegetation model. Biogeosciences 11, 2027–2054 (2014).
Longo, M. et al. The biophysics, ecology, and biogeochemistry of functionally diverse, vertically and horizontally heterogeneous ecosystems: the ecosystem demography model, version 2.2 – Part 1: model description. Geosci. Model Dev. 12, 4309–4346 (2019).
Cotton, J. M., Cerling, T. E., Hoppe, K. A., Mosier, T. M. & Still, C. J. Climate, CO2, and the history of North American grasses since the Last Glacial Maximum. Sci. Adv. 2, e1501346 (2016).
Chini, L. et al. Land-use harmonization datasets for annual global carbon budgets. Earth Syst. Sci. Data 13, 4175–4189 (2021).
Monfreda, C., Ramankutty, N. & Foley, J. A. Farming the planet: 2. Geographic distribution of crop areas, yields, physiological types, and net primary production in the year 2000: global crop areas and yields in 2000. Glob. Biogeochem. Cycles 22, n/a-n/a (2008).
Yu, Q. et al. A cultivated planet in 2010 – Part 2: the global gridded agricultural-production maps. Earth Syst. Sci. Data 12, 3545–3572 (2020).
Jackson, N. D., Konar, M., Debaere, P. & Estes, L. Probabilistic global maps of crop-specific areas from 1961 to 2014. Environ. Res. Lett. 14, 094023 (2019).
Goldewijk, K. K., Beusen, A., Doelman, J. & Stehfest, E. Anthropogenic land use estimates for the Holocene - HYDE 3.2. Earth Syst. Sci. Data 9, 927–953 (2017).
Hurtt, G. C. et al. Harmonization of global land use change and management for the period 850–2100 (LUH2) for CMIP6. Geosci. Model Dev. 13, 5425–5464 (2020).
Powell, R. & Still, C. Biogeography of C3 and C4 vegetation in South America. An. XIV Simpósio Brasileiro Sensoriamento Remoto 14, 2935–2942 (2009).
Powell, R. L., Yoo, E.-H. & Still, C. J. Vegetation and soil carbon-13 isoscapes for South America: integrating remote sensing and ecosystem isotope measurements. Ecosphere 3, art109 (2012).
Zhou, H., Helliker, B. R., Huber, M., Dicks, A. & Akçay, E. C4 photosynthesis and climate through the lens of optimality. Proc. Natl. Acad. Sci. USA 115, 12057–12062 (2018).
Kattge, J. et al. TRY plant trait database – enhanced coverage and open access. Glob. Change Biol. 26, 119–188 (2020).
Carroll, O. et al. Nutrient identity modifies the destabilising effects of eutrophication in grasslands. Ecol. Lett. 25, 754–765 (2022).
Sage, R. F. & Sultmanis, S. Why are there no C4 forests? J. Plant Physiol. 203, 55–68 (2016).
Sage, R. F. & Kubien, D. S. The temperature response of C3 and C4 photosynthesis. Plant, Cell Environ. 30, 1086–1106 (2007).
Raubenheimer, S. L. & Ripley, B. S. CO2 ‐stimulation of savanna tree seedling growth depends on interactions with local drivers. J. Ecol. 110, 1090–1101 (2022).
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: response of african savanna trees to CO2. Austral Ecol. 35, 451–463 (2009).
Bond, W. J., Midgley, G. F. & Woodward, F. I. The importance of low atmospheric CO2 and fire in promoting the spread of grasslands and savannas: FIRE, LOW CO2 and trees. Glob. Change Biol. 9, 973–982 (2003).
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. Change Biol. 18, 675–684 (2012).
Polissar, P. J., Rose, C., Uno, K. T., Phelps, S. R. & deMenocal, P. Synchronous rise of African C4 ecosystems 10 million years ago in the absence of aridification. Nat. Geosci. 12, 657–660 (2019).
O’Mara, N. A. et al. Pleistocene drivers of Northwest African hydroclimate and vegetation. Nat. Commun. 13, 3552 (2022).
Shen, X. et al. Increased seasonality and aridity drove the C4 plant expansion in Central Asia since the Miocene–Pliocene boundary. Earth Planet. Sci. Lett. 502, 74–83 (2018).
Murphy, B. P. & Bowman, D. M. J. S. Seasonal water availability predicts the relative abundance of C3 and C4 grasses in Australia. Glob. Ecol. Biogeogr. 16, 160–169 (2007).
Wang, H. et al. Asian monsoon rainfall variation during the Pliocene forced by global temperature change. Nat. Commun. 10, 5272 (2019).
Knapp, A. K. et al. Resolving the Dust Bowl paradox of grassland responses to extreme drought. Proc. Natl. Acad. Sci. USA 117, 22249–22255 (2020).
Toderich, K. et al. C3/C4 plants in the vegetation of Central Asia, geographical distribution and environmental adaptation in relation to climate. in Climate Change and Terrestrial Carbon Sequestration in Central Asia (eds. Lal, R., Suleimenov, M., Stewart, B., Hansen, D. & Doraiswamy, P.) 33–63 (Taylor & Francis, 2007). https://doi.org/10.1201/9780203932698.ch3.
Noss, R. F. Forgotten Grasslands of the South: Natural History and Conservation. (Island Press, 2013).
Pau, S., Griffith, D. M., Zampieri, N. & Costanza, J. Longleaf pine savannas reveal biases in current understanding of savanna biogeography. Glob. Ecol. Biogeogr. 32, 2047–2052 (2023).
Munroe, S. E. M. et al. A vegetation carbon isoscape for Australia built by combining continental-scale field surveys with remote sensing. Landsc. Ecol. 37, 1987–2006 (2022).
Edwards, E. J. & Smith, S. A. Phylogenetic analyses reveal the shady history of C 4 grasses. Proc. Natl. Acad. Sci. USA 107, 2532–2537 (2010).
Pau, S., Edwards, E. J. & Still, C. J. Improving our understanding of environmental controls on the distribution of C3 and C4 grasses. Glob. Change Biol. 19, 184–196 (2013).
Lehmann, C. E. R., Archibald, S. A., Hoffmann, W. A. & Bond, W. J. Deciphering the distribution of the savanna biome. N. Phytol. 191, 197–209 (2011).
Venter, Z. S., Cramer, M. D. & Hawkins, H. J. Drivers of woody plant encroachment over Africa. Nat. Commun. 9, 1–7 (2018).
Scheiter, S. et al. Fire and fire‐adapted vegetation promoted C4 expansion in the late Miocene. N. Phytol. 195, 653–666 (2012).
Karp, A. T., Behrensmeyer, A. K. & Freeman, K. H. Grassland fire ecology has roots in the late Miocene. Proc. Natl. Acad. Sci. USA 115, 12130–12135 (2018).
Bowman, D. M. J. S. et al. Vegetation fires in the Anthropocene. Nat. Rev. Earth Environ. 1, 500–515 (2020).
Long, S. P. Environmental Responses. in C4 Plant Biology 215–249 (Elsevier, 1999). https://doi.org/10.1016/B978-012614440-6/50008-2.
Vitousek, P. M. et al. Human alteration of the global nitrogen cycle: sources and consequences. Ecol. Appl. 7, 737–750 (1997).
Ripley, B. S., Abraham, T. I. & Osborne, C. P. Consequences of C4 photosynthesis for the partitioning of growth: a test using C3 and C4 subspecies of Alloteropsis semialata under nitrogen-limitation. J. Exp. Bot. 59, 1705–1714 (2007).
Hengl, T. et al. SoilGrids250m: Global gridded soil information based on machine learning. PLoS ONE. 12, e0169748 (2017).
Moreno-Martínez, Á. et al. A methodology to derive global maps of leaf traits using remote sensing and climate data. Remote Sens. Environ. 218, 69–88 (2018).
Bayer, A. D. et al. Uncertainties in the land-use flux resulting from land-use change reconstructions and gross land transitions. Earth Syst. Dynam. 8, 91–111 (2017).
Farquhar, G. D., von Caemmerer, S. & Berry, J. A. A biochemical model of photosynthetic CO2 assimilation in leaves of C3 species. Planta 149, 78–90 (1980).
Caemmerer, S. V. Biochemical models of leaf photosynthesis. (CSIRO Publishing, 2000).
Zhou, H., Akçay, E. & Helliker, B. Optimal coordination and reorganization of photosynthetic properties in C 4 grasses. Plant Cell Environ. 46, 796–811 (2023).
Pignon, C. P. & Long, S. P. Retrospective analysis of biochemical limitations to photosynthesis in 49 species: C4 crops appear still adapted to pre‐industrial atmospheric CO2. Plant Cell Environ. 43, 2606–2622 (2020).
Friedl, M., Gray, J. & Sulla-Menashe, D. MCD12Q2 MODIS/Terra+Aqua Land Cover Dynamics Yearly L3 Global 500m SIN Grid V006. (2019) https://doi.org/10.5067/MODIS/MCD12Q2.006.
Clapp, R. B. & Hornberger, G. M. Empirical equations for some soil hydraulic properties. Water Resour. Res. 14, 601–604 (1978).
Martens, B. et al. GLEAM v3: satellite-based land evaporation and root-zone soil moisture. Geosci. Model Dev. 10, 1903–1925 (2017).
Harris, I., Osborn, T. J., Jones, P. & Lister, D. Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset. Sci. Data 7, 109 (2020).
Kobayashi, S. et al. The JRA-55 reanalysis: general specifications and basic characteristics. J. Meteorol. Soc. Jpn. 93, 5–48 (2015).
Myneni, R. B. et al. Global products of vegetation leaf area and fraction absorbed PAR from year one of MODIS data. Remote Sens. Environ. 83, 214–231 (2002).
Tsendbazar, N. et al. Towards operational validation of annual global land cover maps. Remote Sens. Environ. 266, 112686 (2021).
Brown, C. F. et al. Dynamic world, near real-time global 10 m land use land cover mapping. Sci. Data 9, 251 (2022).
Sulla-Menashe, D. & Friedl, M. A. User Guide to Collection 6 MODIS Land Cover (MCD12Q1 and MCD12C1) Product.
ESA. Land Cover CCI Product User Guide Version 2. Tech. Rep. https://www.esa-landcover-cci.org/?q=webfm_send/84 (2014).
Lehmann, C. E. R. et al. Functional diversification enabled grassy biomes to fill global climate space. http://biorxiv.org/lookup/doi/10.1101/583625 (2019). https://doi.org/10.1101/583625.
Friedlingstein, P. et al. Global Carbon Budget 2019. Earth Syst. Sci. Data 11, 1783–1838 (2019).
Cox, P. M. Emergent constraints on climate-carbon cycle feedbacks. Curr. Clim. Change Rep. 5, 275–281 (2019).
Eyring, V. et al. Taking climate model evaluation to the next level. Nat. Clim. Change 9, 102–110 (2019).
Acknowledgements
X.L., T.W.S., J.T. and R.Z. are supported by the NUS Presidential Young Professorship awarded to X.L. (A-0003625-03-00), a Tier II research grant from the Ministry of Education (A-8001551-00-00) and a Singapore Energy Center Core project (A-8000179-00-00). NGS acknowledges support from the National Science Foundation (DEB-2045968) and Texas Tech University. T.K. and N.G.S. also acknowledge support from LEMONTREE (Land Ecosystem Models based On New Theory, Observations, and Experiments) project, funded through the generosity of Eric and Wendy Schmidt by recommendation of the Schmidt Futures program. T.K. acknowledges additional support from a NASA Carbon Cycle Science Award 80NSSC21K1705, and the RUBISCO SFA, which is sponsored by the Regional and Global Model Analysis (RGMA) Program in the Climate and Environmental Sciences Division (CESD) of the Office of Biological and Environmental Research (BER) in the U.S. Department of Energy (DOE) Office of Science. C.J.S. and D.M.G. are grateful for support from the National Science Foundation (Award 1926431). We thank the TRY database and the global nutrient network for making their data available to support the study. X.L. thanks Prof. Graham Farquhar, Prof. Belinda Medlyn, Prof. Shuli Niu and Prof. Anthony Walker for sharing their insights on the study through several personal communications.
Author information
Authors and Affiliations
Contributions
XL conceived the idea. XL and HZ designed the study and carried out the analysis. XL, HZ, DG, TK, NS, SS and CS participated in discussions at various stages. SS provided the TRENDY v9 DGVMs simulations. JT, TS, RZ and NS aided in data collection and analysis. All authors contributed to writing.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Jaideep Joshi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
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 license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license 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 license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Luo, X., Zhou, H., Satriawan, T.W. et al. Mapping the global distribution of C4 vegetation using observations and optimality theory. Nat Commun 15, 1219 (2024). https://doi.org/10.1038/s41467-024-45606-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-024-45606-3
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
