Global CO₂ rise leads to reduced maximum stomatal conductance in Florida vegetation

Abstract

A principle response of C3 plants to increasing concentrations of atmospheric CO₂ (CO₂) is to reduce transpirational water loss by decreasing stomatal conductance (g_s) and simultaneously increase assimilation rates. Via this adaptation, vegetation has the ability to alter hydrology and climate. Therefore, it is important to determine the adaptation of vegetation to the expected anthropogenic rise in CO₂. Short-term stomatal opening–closing responses of vegetation to increasing CO₂ are described by free-air carbon enrichments growth experiments, and evolutionary adaptations are known from the geological record. However, to date the effects of decadal to centennial CO₂ perturbations on stomatal conductance are still largely unknown. Here we reconstruct a 34% (±12%) reduction in maximum stomatal conductance (g_smax) per 100 ppm CO₂ increase as a result of the adaptation in stomatal density (D) and pore size at maximal stomatal opening (a_max) of nine common species from Florida over the past 150 y. The species-specific g_smax values are determined by different evolutionary development, whereby the angiosperms sampled generally have numerous small stomata and high g_smax, and the conifers and fern have few large stomata and lower g_smax. Although angiosperms and conifers use different D and a_max adaptation strategies, our data show a coherent response in g_smax to CO₂ rise of the past century. Understanding these adaptations of C3 plants to rising CO₂ after decadal to centennial environmental changes is essential for quantification of plant physiological forcing at timescales relevant for global warming, and they are likely to continue until the limits of their phenotypic plasticity are reached.

Land plants play a crucial role in regulating our planet's hydrological and energy balance by transpiring water through the stomatal pores on their leaf surfaces. A fundamental response of C3 plants to increasing atmospheric CO₂ concentration (CO₂) is to minimize transpirational water loss by reducing diffusive stomatal conductance (g_s) and simultaneously increasing assimilation rates (1). The resulting increased intrinsic water-use efficiency (iWUE: the ratio of assimilation to g_s) improves the vegetation's drought resistance and reduces the cost associated with the leaf's water transport system like leaf venation (2, 3). On a regional to global scale, decreasing rates of transpiration concurrently affect climate through reduced cloud formation and precipitation (4) and with this exert a physiological feedback on climate and hydrology on top of the radiative forcing of increasing CO₂ (5–7). In the light of continuing anthropogenic climate change, it is therefore imperative to determine how plants adapt to rising atmospheric CO₂.

During their 400 million year history, land plants have been exposed to large variations in environmental conditions that prompted genetic adaptations toward mechanisms that optimize individual fitness. Over this period, plant adaptation to CO₂ is apparent as periods with high CO₂ favored species with few relatively large stomata and low g_s, whereas periods with low CO₂ (as at present) favored species with many relatively small stomata and higher g_s (8). Moreover, decreasing CO₂ after ≈100 million years likely triggered the evolutionary development of a more extensive leaf vein network in angiosperms, giving them the advantage of potentially higher g_s than gymnosperms with low vein density (9). At shorter timescales, plants have the ability to adjust their phenotype to optimize gas exchange. In response to short (seconds to hours) perturbations in CO₂, plants open and close their stomata (10, 11), whereas in response to CO₂ changes at decadal to centennial timescales, plants adjust leaf stomatal density (D) and/or maximum stomatal dimensions (a_max) (12–15). This process of epidermal structural adaptation is in part controlled by a signaling mechanism from mature to developing leaves, optimizing stomatal density and size to the changed environmental conditions (16). These epidermal characteristics determine the anatomical maximum stomatal conductance to water vapor (g_smax, mol·m⁻²·s⁻¹) of fully opened stomata and can be calculated as (8, 17):

in which stomatal density [D (number of stomata·m⁻²)], the size of the fully opened stomata a_max (m²), and depth of the stomatal tube l (m) are the determining variables. The diffusivity of water vapor d_w (m²·s⁻¹) and the molar volume of air v (m³·mol⁻¹) are constants. Values of a_max and l are derived from the stomatal pore length L (m). Maximum stomatal conductance to CO₂ is g_smax/1.6 (18).

The most comprehensive analyses of plant adaptation to elevated CO₂ in (semi)natural environments are available from free-air carbon enrichments (FACE) growth experiments (19). Although decreases in D of C3 plants did occur in some studies (20), the observed reduction in g_s was found to be caused by instantaneous adaptation only (21). Apparently, the run-time of these growth experiments of <5 y might be too short to trigger statistically significant epidermal structural adaptation (22). Consequently, the subtle adaptation of vegetation to continuously increasing CO₂ can only be elucidated from material covering periods long enough to deduce quantifiable structural adaptation. Because CO₂ has already increased by ≈100 ppm over the past 150 y, historical leaves preserved in sediments and stored in herbarium collections offer an excellent opportunity to study the adaptation of g_smax to the gradual rise in CO₂.

Because the leaf epidermal properties D and a_max are also influenced by other environmental factors such as light, temperature, and water availability (23–25), it is necessary to use leaf material from plants that grew under conditions in which the global CO₂ rise is the dominant variable factor. This prerequisite is met in Florida, where the vegetation has been exposed to the global ≈100 ppm CO₂ increase under near constant average growth season temperatures and precipitation rates over the past 150 y (Fig. 1). Moreover, structural adaptation to increasing CO₂ by decreasing D has already been demonstrated for a number of Florida forest taxa (28, 29).

Fig. 1.

Locations of leaf material collection sites in Florida: state counties covered by herbarium material (gray) and subfossil leaf fragment sites (black stars). Florida averaged spring (March, April, and May) temperature and cumulative precipitation (http://www.ncdc.noaa.gov) and global atmospheric CO₂ concentration [Siple station (26); Mauna Loa (27)] from 1880 A.D. to present are given. The various sites are situated approximately at sea level. Black lines in the temperature and precipitation graphs are long-term means of ≈21 °C and ≈260 mm, respectively.

Here we present a high-resolution historical record of nine C3 species that adapted g_smax to the 100 ppm rise in CO₂ since approximately 1880 A.D. Species studied are the woody angiosperms Acer rubrum (Aceraceae), Myrica cerifera (Myricaceae), Ilex cassine (Aquifoliaceae), Quercus laurifolia (Fagaceae), and Quercus nigra (Fagaceae), the conifers Pinus elliottii (Pinaceae), Pinus taeda (Pinaceae), and Taxodium distichum (Cupressaceae), and the fern Osmunda regalis (Osmundaceae). This selection embraces species with deciduous and evergreen leaf types, growing in wet to well-drained sites in upper to lower canopy layers (Table S1). The cuticle material analyzed originates from subfossil leaf fragments retrieved from well-dated young peat deposits (26, 30) as well as herbarium and modern material collected from various sites in Florida (Fig. 1).

In the present study we aim to quantify how these species have adapted g_smax in response to the industrial rise in CO₂. Moreover, the present selection includes multiple angiosperm and coniferous species, of which the leaves are characterized by a high and low leaf vein density, respectively. Because angiosperms have invested more in an elaborate leaf hydraulic system (31), we raise the hypothesis that they will benefit more by reducing g_smax stronger to rising CO₂. The data we present here allow for quantification of plant physiological adaptation at timescales relevant for anthropogenic climate change and provide much-needed data for parameterization and validation of climate models that include these physiological feedbacks.

Results

A consistent decrease in g_smax over the anthropogenic rise in CO₂ is observed in all species studied (P < 0.05) (Fig. 2). The inferred g_smax values of these nine species range between ≈0.4 (mol·m⁻²·s⁻¹) to ≈4 (mol·m⁻²·s⁻¹), with the highest values for the angiosperm canopy species Q. nigra, Q. laurifolia, and A. rubrum and the lowest values for the fern O. regalis. Despite the large differences in absolute values of g_smax between species, relative sensitivities in g_smax over ≈100 ppm CO₂ rise are highly comparable, with a mean slope of −34% (±12%) per 100 ppm (Table 1). The weakest responses occur in P. elliottii and Q. laurifolia, with a relative sensitivity of only −17% and −18% per 100 ppm, whereas P. taeda shows the strongest sensitivity in g_smax, with −55% per 100 ppm. Despite these differences in response rate, the total change exceeds the maximum intrinsic variability quantified as the root mean square error (RMSE) in all species (Table S2). The CO₂-induced phenotypic decrease in g_smax on decadal timescales resembles evolutionary g_smax adaptation over geological timescales (32), reflecting the permanent attempt of plants to optimize individual fitness.

Fig. 2.

Species-specific relation between g_smax and CO₂ over the past 150 y. Symbols are average g_smax (mol·m⁻²·s⁻¹) for each species per CO₂ level (ppm) studied (n = 160), and accompanying year A.D. (species names and abbreviations are given in Table 1). Solid lines show linear regressions between CO₂ and g_smax for each species, r² and relative sensitivity are given in Table 1. The functions and RMSE for each species are provided in Table S2.

Table 1.

Relative sensitivity of g_smax, D, and a_max to CO₂ increase for the species sampled (intercept, 100% at 280 ppm CO₂), with r² of the linear regressions used

		Average gsmax		Average stomatal density D		Average pore size amax
Species	Species code	Relative sensitivity (%·ppm−1)	r2	Relative sensitivity (%·ppm−1)	r2	Relative sensitivity (%·ppm−1)	r2
Acer rubrum	Ar	−0.41*	0.45	−0.29*	0.30	−0.27*	0.40
Ilex cassine	Ic	−0.30*	0.36	−0.26*	0.38	−0.13	0.04
Myrica cerifera	Mc	−0.36*	0.49	−0.31*	0.40	0.01	<0.001
Osmunda regalis	Or	−0.42*	0.24	−0.27	0.09	−0.31*	0.31
Pinus elliottii	Pe	−0.17*	0.36	−0.23*	0.55	0.13	0.15
Pinus taeda	Pt	−0.55*	0.54	−0.42*	0.56	−0.25	0.25
Quercus laurifolia	Ql	−0.18*	0.21	−0.09	0.13	−0.14	0.07
Quercus nigra	Qn	−0.37*	0.61	−0.28*	0.44	−0.21	0.18
Taxodium distichum	Td	−0.33*	0.58	−0.35*	0.52	0.06	0.06

*Statistical significance for the regression as well as the change, with P < 0.05.

As on geological timescales (8), combined values of D and a_max on which the calculation of g_smax is based here are negatively correlated and follow a power law relationship in which high values of D are accompanied by low a_max values, and vice versa (Fig. 3). For individual species, however, D and a_max are confined to specific ranges forming clusters distributed along this power law, where significant negative correlations are also apparent in five out of nine individual clusters (P. elliottii, T. distichum, Q. laurifolia, M. cerifera, and O. regalis; Table S3). This implies that the clusters represent the phenotypic plasticity of the various species, showing adjustments of both D and a_max that occurred in response to the complex of environmental perturbations to which the sampled vegetation was exposed, including CO₂.

Fig. 3.

The measured stomatal density [D (mm⁻²)] and pore size [a_max (μm²)] of nine common species in Florida (n = 667) (species names and abbreviations as in Table 1). The clusters depict a phenotypical range of D and a_max for each species under changing conditions of the past 150 y. Approximate lower limits are D = ≈20 mm⁻² and a_max = ≈15 μm². Multiple combination of D and a_max can lead to the same g_smax value (mol·m⁻²·s⁻¹), indicated by the black curved lines.

Within the total dataset, the most prominent difference exists between the angiosperm clusters with many small stomata that display large diversity in D and the conifers and fern clusters with few large stomata that display large diversity in a_max. The position of individual species on this power law curve likely represents their different evolutionary history (33, 34), with an earlier design for conifers and ferns and a more innovative design for angiosperms. Nevertheless, different combinations of D and a_max can lead to the same g_smax (Fig. 3) and the same decrease in g_smax in response to rising CO₂ (Table 1).

Testing the CO₂ sensitivity of D and a_max individually, we observe that the plastic response of D is always negative and more pronounced than in a_max (Table 1). This consistent decrease of D under rising CO₂ has already been reported for the angiosperm and fern species in our dataset (26, 27). We now complement the range of species known to reduce D in response to rising CO₂ by including the conifers P. elliottii, P. taeda, and T. distichum. Over the sampled CO₂ rise in CO₂, the relative sensitivity in D varies from maximal −42% per 100 ppm in P. taeda to minimal −9% per 100 ppm in Q. laurifolia (Table 1) (P < 0.05 for all but O. regalis and Q. laurifolia, with P = 0.12 and P = 0.10, respectively). The total change in D exceeds the maximum intrinsic variability quantified as the RMSE in all species except O. regalis and Q. laurifolia (Table S4). These rates are broadly consistent with decreases in D reported for European tree species grown under anthropogenic CO₂ increase (12, 13, 15, 35).

Focusing on the changes in a_max over the sampled CO₂ increase, weak and unidirectional relations are observed. Significant relations were only found for A. rubrum and O. regalis, which show reductions in a_max of −27% and −31% per 100 ppm, respectively (Table 1). Moreover, the changes in our a_max data series only exceed the RMSE for five of the species studied (Table S5). This variable response is different from changes in a_max to anthropogenic CO₂ rise reported earlier for two European tree species, for which a weak increase was observed (13, 36). From these observations it is apparent that D is highly sensitive to rising CO₂, whereas changes in a_max are variable between species and seem to be governed independently.

Because it is hypothesized that the different leaf structures, in particular leaf vein density, of angiosperms and conifers (31) result in different epidermal structural responses to rising CO₂, we compared the general relative sensitivities of the two plant groups in our dataset. Results show that coniferous species seem to respond with a significantly stronger decrease in D (slope, −35% per 100 ppm) than angiosperms (slope, −27% per 100 ppm) (Fig. 4A and Table S6). Conversely, angiosperms respond with an apparent but not significantly stronger decrease in a_max (slope, −15% per 100 ppm) compared with the conifers (slope, −7% per 100 ppm). Conifers also display a much larger range of variability, indicated by the broader confidence interval (Fig. 4B). Despite these differences, a highly comparable overall decrease in g_smax to a rise of CO₂ from preindustrial to present in angiosperms (slope, −33% per 100 ppm) and conifers (slope, −37% per 100 ppm) emerges from this combination (Fig. 4C). Summarizing, our data show that both angiosperms and conifers exhibit a similar response in g_smax to the anthropogenic rise in CO₂.

Fig. 4.

Relative sensitivity in D (A), a_max (B), and g_smax (C) of the grouped angiosperm species (black line, black dots) and coniferous species (gray line, gray triangles) over the sampled CO₂ increase since the industrial revolution. Dashed lines depict 95% confidence intervals for angiosperms (black short dashed lines) and conifers (gray long dashed lines). (SE, r², and P values given in Table S6). Only D is significantly different between angiosperms and conifers (P < 0.001).

Discussion

The presented data reveal that the nine species from Florida reduce their g_smax in response to the industrial CO₂ rise via D and a_max adaptation within their phenotypic plasticity. This likely represents the plants’ adaptation to increase iWUE by optimizing carbon gain to water loss (11, 37). We demonstrate that adaptation of g_smax is achieved by species-specific strategies to alter D and/or a_max. The overall decrease in g_smax is predominantly the result of a general and significant reduction of D in response to rising CO₂ in all species, whereas a_max seems to adapt to other environmental conditions as well, because no consistent relation with CO₂ was observed. However, the importance of including D as well as a_max in the reconstruction of g_smax is emphasized by the generally improved correlation of g_smax with CO₂, compared with D and a_max separately (Table 1). The observed change in a_max opposes the positive relation between pore size and CO₂ found over geological timescales (8). This discrepancy can be explained by considering that on the timescale studied here plants adapt within their phenotype and not genotype to reduce g_smax, which is most efficiently done by reducing rather than increasing pore size. This suggests that plants can and do adapt to changing conditions by fine-tuning D and a_max plastically to optimize their individual fitness.

Despite the consistent trend observed in g_smax, considerable variability characterizes the individual D and a_max data series, and consequently g_smax, because climatic and site-specific environmental factors such as light, temperature, and water availability affect D and a_max as well (22, 24, 25). Even though the long-term mean temperature and precipitation in Florida have not changed over the past 150 y, strong interannual temperature and precipitation fluctuations (Fig. 1) caused by the El Niño/Southern Oscillation (ENSO) and Atlantic Multidecadal Oscillation teleconnections (38, 39) may in part have caused D and a_max variability. Indeed, short-term changes in epidermis morphology in Q. laurifolia have been linked to ENSO-tied winter precipitation (25). Together with D and a_max diversity throughout the canopy and even within the same leaf (40), these environmental factors produce substantial scatter in the data. Consequently, sampling on low temporal resolution might explain the lack of evidence for CO₂-induced iWUE adaptation as in herbarium studies covering 2 to 5 selected years only (41). The present study therefore emphasizes the necessity of sufficiently high-resolution as well as multidecadal data series to elucidate the long-term subtle response of g_smax to changing CO₂.

The large variation in reconstructed g_smax values reflect the difference in leaf vascular architecture, whereby the high vein density typical for angiosperms allows for high g_smax and the low vein density in ferns and conifers is reflected by low g_smax (31, 42). The differences in the leaf hydraulic systems between angiosperms and conifers are also expressed in their position on the power law relation between D and a_max. Angiosperms reach high g_smax with numerous small stomata, and conifers reach lower g_smax with fewer large stomata (8). These findings can be placed against an evolutionary background, where ferns and conifers evolved in a higher-than-present CO₂ world, in which lower g_smax would be perfectly sufficient to maintain high photosynthesis. The late Cretaceous drop in CO₂ likely triggered the expansion of the leaf-vascular network in angiosperms (9), allowing them to attain higher photosynthesis rates than conifers and ferns but at the cost of high carbon and transpirational water loss (3). This water loss in angiosperms might be minimized as small stomata are faster to close than large stomata under desiccating conditions (43). Moreover, a consequence of the associated high water loss is the resulting evaporative cooling, which maintains an optimal leaf temperature (44). Our data thus show that species-specific g_smax is determined in part by evolutionary adaptation to conditions in which they evolved.

When exposed to decadal variability the species studied adapt within the limits of their phenotypic plasticity, by adjusting D and a_max. Despite the large differences in D and a_max between species, even within the same genus they all exhibit highly comparable adaptation of g_smax to increasing CO₂. Comparing the general adaptation of the angiosperms and conifers as groups, however, a different strategy to reduce g_smax was observed, depending on their position on the power law curve. Although only the relative change in D between angiosperms and conifers is significantly different, the tendency towards an opposite response in D and a_max does illustrate that variable adaptations lead to the same reduction in g_smax. These results can be explained by the different position on the power law curve, whereby species reduce g_smax most efficiently by changing either D or a_max to get the steepest gradient in g_smax (Fig. 2). Because the construction of an extended vascular network is coupled to high carbon costs (3, 42), it was hypothesized before that angiosperms reduce g_smax more than conifers and ferns. However, our data show a highly comparable sensitivity to the industrial CO₂ rise in all groups sampled and thereby demonstrate the underlying principle that plants generally optimize their leaf structure in response to rising CO₂, apparently irrespective of their leaf architecture.

Having discussed the responses and possible underlying mechanisms, the potential further development of g_smax under future increasing CO₂ can be evaluated. The iWUE responses measured in short-term growth experiments over below-present to present CO₂ levels are also found to be comparable in angiosperms, conifers, and ferns, but the trends diverge from present to elevated CO₂, where the response in conifers and ferns levels off (45). Our results of structural adaptation from ≈280 ppm to 387 ppm CO₂ does not bear any evidence for a diverging response between plant lineages. Whether any g_smax off-leveling will occur under continuing CO₂ enrichment, and at what CO₂ concentration this will happen, should be estimated by modeling exercises incorporating adaptation within the species-specific phenotypic plasticity (37).

In conclusion, our results point to a common mechanism in C3 plants to reduce maximum stomatal conductance via adjustment of stomatal density and pore size within the limits of their phenotypic plasticity on a decadal timescale. As atmospheric carbon dioxide concentration is rising, plants can and do reduce water loss by reducing maximal stomatal conductance while maintaining carbon uptake (3, 31). Further decreases in stomatal conductance have been observed at CO₂ rising above present levels in FACE short-term experiments (21) and in fossil leaves over geological timescales (8). Both lines of evidence, however, fall beyond or below the timescales of the projected rate of continuing CO₂ increase, which is likely to surpass the time needed for adaptation via natural selection. Consequently, the adaptation within the phenotypic plasticity is likely to constrain epidermis structural adaptation in the near future when phenotypic response limits are reached (35, 37). Current increase in CO₂ and the coinciding reduction in plant transpiration already results in increased continental run-off (46), and climate models predict surface temperature increases arising from reduced evaporative cooling (6, 7). The mechanisms of optimization of carbon gain to water loss described here could be used to better estimate this physiological forcing for the past and future CO₂ but should be considered within the framework of species-specific phenotypic plasticity (37).

Materials and Methods

Sample Preparation and Analysis.

The leaf fragments were treated in 4% sodium hypochlorite (NaClO₂) at 40 °C for several minutes up to 24 h, after which the stomata-bearing abaxial cuticle could be peeled off from the mesophyll, dyed with saffranine, and mounted in glycerine jelly. Because Pinus has an approximately equal amount of stomata on the abaxial as well as the adaxial surface, the entire cuticle was processed. Standardized, computer-aided analysis of the epidermal properties was performed on Leica Quantimet 500C/500+ and AnalySIS image analysis systems. Stomatal density (D; number of stomata·m⁻²) was measured on 5–10 alveoles of each leaf sample and averaged. Because of different epidermis cell patterning, Pinus is measured with the stomatal rows running diagonally in the image. Pore length (L; μm) is determined by averaging measurements of ≈25 stomata for each sample. Data are available upon request.

Calculating g_smax to Water Vapor.

To determine the stomatal conductance to water vapor g_smax (mol·m⁻²·s⁻¹), the equation provided by Franks and Farquhar (17) is applied, using a two-way end correction accounting for the diffusion shells (8) (Eq. 1). Maximum pore surface area a_max (m²) is defined as an ellipse and quantified as π·L²/8, with L being stomatal pore length (m). Stomatal pore depth l (m) is assumed to be equal to the guard cell width of the stomata when the guard cell is fully inflated (8). Quantification of l follows from the significant positive linear relations between pore length and guard cell width for each species, with exception of P. taeda, for which a constant value is taken (Table S7). Values used for gas constants d and v are those for 25 °C. For the determination of the long-term relative sensitivities of the measured D and a_max, and consequent g_smax, the regressions are performed on values averaged per sampled year.

Statistical Analyses.

The significance of the observed regressions presented here is tested in three steps, with P values of <0.05 considered statistically significant. First, the significance of each regression plotted through the data series was tested. Second, using a Student t test on the slopes of these regressions, it was determined whether the observed changes were significantly larger than 0. Finally, to test whether the average responses in D, a_max, and g_smax were significantly different between angiosperms and conifers, a t test (two samples assuming unequal variances) was performed on the pooled data of each group within the CO₂ interval of 360–387 ppm.