Stomatal frequency of Quercus glauca from three material sources shows the same inverse response to atmospheric pCO₂

Abstract

Background and Aims

The inverse correlation between atmospheric CO₂ partial pressure (pCO₂) and stomatal frequency in many plants has been widely used to estimate palaeo-CO₂ levels. However, apparent discrepancies exist among the obtained estimates. This study attempts to find a potential proxy for palaeo-CO₂ concentrations by analysing the stomatal frequency of Quercus glauca (section Cyclobalanopsis, Fagaceae), a dominant species in East Asian sub-tropical forests with abundant fossil relatives.

Methods

Stomatal frequencies of Q. glauca from three material sources were analysed: seedlings grown in four climatic chambers with elevated CO₂ ranging from 400 to 1300 ppm; extant samples collected from 14 field sites at altitudes ranging from 142 to 1555 m; and 18 herbarium specimens collected between 1930 and 2011. Stomatal frequency–pCO₂ correlations were determined using samples from these three sources.

Key Results

An inverse correlation between stomatal frequency and pCO₂ was found for Q. glauca through cross-validation of the three material sources. The combined calibration curves integrating data of extant altitudinal samples and historical herbarium specimens improved the reliability and accuracy of the curves. However, materials in the climatic chambers exhibited a weak response and relatively high stomatal frequency possibly due to insufficient treatment time.

Conclusions

A new inverse stomatal frequency–pCO₂ correlation for Q. glauca was determined using samples from three sources. These three material types show the same response, indicating that Q. glauca is sensitive to atmospheric pCO₂ and is an ideal proxy for palaeo-CO₂ levels. Quercus glauca is a nearest living relative (NLR) of section Cyclobalanopsis fossils, which are widely distributed in the strata of East Asia ranging from the Eocene to Pliocene, thereby providing excellent materials to reconstruct the atmospheric CO₂ concentration history of the Cenozoic. Quercus glauca will add to the variety of proxies that can be widely used in addition to Ginkgo and Metasequoia.

INTRODUCTION

Reconstructing the deep-time dynamics of atmospheric CO₂ concentrations has been the subject of a great deal of attention because this greenhouse gas plays an important role in driving and amplifying global climate change (McElwain et al., 2016). For example, the Paleocene–Eocene thermal maximum (PETM) is well known as an intense interval of global warming linked to a dramatically elevated CO₂ concentration (Zachos et al., 2005), while the Eocene–Oligocene transition is characterized by a rapid temperature drop that was probably associated with a significant CO₂ decrease (Zanazzi et al., 2007; Goldner et al., 2014). Other than measuring ice cores from the past 800 000 years (Lüthi et al., 2008), there is no direct method to determine palaeoatmospheric CO₂ (palaeo-CO₂) concentrations. Pre-ice core CO₂ concentration estimates are achieved by biogeochemical models (Berner and Kothavala, 2001; Berner, 2006) and various independent palaeobotanical and geochemical proxies, such as palaeosols (Ekart et al., 1999; Myers et al., 2012), phytoplankton (Pagani et al., 2005; Seki et al., 2010), marine carbonate (Tripati et al., 2009; Seki et al., 2010) and fossil stomata (Royer et al., 2001b; Kürschner et al., 2008; Franks et al., 2014). These methods have provided numerous palaeo-CO₂ estimates throughout the Phanerozoic Eon (Royer, 2006; Breecker et al., 2010); however, there appear to be considerable discrepancies and large variabilities between estimates obtained by these different approaches (Royer et al., 2001a; Beerling and Royer, 2011).

Because their main function is to exchange gas between plants and the atmosphere, stomata respond directly to atmospheric CO₂ (Lake et al., 2001, 2002; Miyazawa et al., 2006; Mizutani and Kanaoka, 2018); thus fossil stomata may conceal the atmospheric CO₂ in the geological past, especially in the Cenozoic Era (Royer et al., 2001a; Beerling and Royer, 2002a; Steinthorsdottir et al., 2011). It follows that stomata-based methods have been used extensively to estimate palaeo-CO₂ levels. Stomata-based methods include both empirical approaches (van der Burgh et al., 1993; McElwain and Chaloner, 1996; Beerling and Royer, 2002a; Kürschner et al., 2008; Retallack, 2009; Doria et al., 2011; Steinthorsdottir et al., 2011; Barclay and Wing, 2016) and mechanistic models (Wynn, 2003; Konrad et al., 2008; Grein et al., 2011; Franks et al., 2014). Empirical methods are generally based on the close correlation between atmospheric CO₂ partial pressure (pCO₂) and leaf stomatal frequency [expressed as stomatal density (SD) or stomatal index (SI)] that has been observed in many C₃ plants (Woodward, 1987; McElwain, 1998; Kürschner et al., 2001; Royer, 2001; Beerling and Royer, 2002a; Kouwenberg et al., 2003; Barclay et al., 2010; Bai et al., 2015; Steinthorsdottir et al., 2019). Alternative mechanistic models have been developed in recent decades. These models infer palaeo-CO₂ levels from photosynthetic gas exchange and/or water availability measurements (Wynn, 2003; Konrad et al., 2008; Franks et al., 2014). More recently, both empirical stomata-based proxies and mechanistic models have been applied to the same fossil leaves; these studies have shown that these two methods provide comparable estimates of palaeo-CO₂ concentrations (Barclay and Wing, 2016; Montañez et al., 2016; Sun et al., 2017). Since empirical stomata methods are simpler than mechanistic models, because they only require SD or SI measurements from fossil leaves, they remain the most widely used proxy for palaeo-CO₂ levels (McElwain and Steinthorsdottir, 2017).

Woodward (1987) was the first to propose that SD decreases with increasing CO₂ levels, and Woodward and Bazzaz (1988) showed that stomatal frequency responds to atmospheric pCO₂ (Pa) but not to CO₂ mole fraction (μmol mol^–1) or concentration (ppm). Since then, an increasing number of studies have used the stomatal frequency (SF)–pCO₂ relationships to estimate palaeo-CO₂ levels (Kürschner et al., 2001, 2008; Beerling and Royer, 2002a, b; Bai et al., 2015; Hu et al., 2015; Barclay and Wing, 2016; Steinthorsdottir et al., 2019). In empirical stomata methods, the SF–pCO₂ relationship of a fossil’s nearest living relative (NLR) must be established, before the fossil’s stomata can be used to estimate palaeo-CO₂ levels (Royer, 2001; Steinthorsdottir et al., 2016). This is because the SF–pCO₂ correlation is species specific: while the majority of plant species studied to date show an inverse correlation, some have no significant relationship and a minority exhibit a positive correlation (Woodward and Kelly, 1995; Royer, 2001; Haworth et al., 2010b). Materials from three different sources can be used to determine the SF–pCO₂ relationship of an NLR species: (1) experimental plants grown under elevated pCO₂ in greenhouses; (2) historical herbarium specimens collected over an extended period of time; and (3) specimens collected along an altitudinal gradient (Haworth et al., 2010b; Hu et al., 2015). So far, greenhouse and herbarium materials (Woodward, 1987; van der Burgh et al., 1993; Retallack, 2001; Royer et al., 2001b; Greenwood et al., 2003; Kouwenberg et al., 2003; Barclay et al., 2010; Haworth et al., 2011a) have been used much more frequently than altitudinal samples (McElwain, 2004; Eide and Birks, 2006; Kouwenberg et al., 2007; Hu et al., 2015).

However, all three sources of material have inherent limitations. (1) Experimental greenhouse materials may not capture the long-term, genetic responses of plants to slow environmental changes (Woodward, 1988; Beerling and Chaloner, 1993a; McElwain and Chaloner, 1995). Plants may exhibit incomplete phenotypic adaptation to the elevated pCO₂ (Barclay and Wing, 2016) and, because of the limited pCO₂ gradient in the greenhouse, only a trend of stomatal frequency response can be obtained, which is inadequate to construct a calibration curve for palaeo-CO₂ estimates. (2) Use of herbarium materials may be limited by the availability of historical specimens, and the ensuing paucity of data may lead to larger errors. (3) Altitudinal materials are valuable only if the targeted plant species is distributed over a large altitudinal gradient and altitude-induced environmental variations may affect stomatal frequency. Moreover, both herbarium and altitudinal materials only capture sub-ambient to ambient pCO₂, and thus are not particularly useful for estimating palaeo-CO₂ during greenhouse intervals. Indeed, previous studies have shown inconsistent stomatal frequency responses to atmospheric pCO₂ in the same plant species (Beerling and Chaloner, 1993b; Atkinson et al., 1997; Beerling, 1997; Lin et al., 2001; Eide and Birks, 2006), probably reflecting the inherent weaknesses of the material source type used. To reduce their inherent bias and to obtain a reliable correlation, combined use of all three material types is highly advisable. So far, very few studies have attempted this approach. Eide and Birks (2006) used the three material types to investigate the relationship between stomatal frequency and pCO₂ in Betula pubescens but found no clear SF–pCO₂ relationship, leading them to conclude that B. pubescens was unsuitable for palaeo-CO₂ reconstruction. Clearly, more studies that combine all three types of materials are needed.

To date, the most widely used proxies to estimate palaeo-CO₂ levels have been Ginkgo biloba and Metasequoia glyptostroboides, since both species exhibit ideal inverse SF–pCO₂ correlations and have abundant fossil relatives stretching as far back as the Cretaceous Period (Retallack, 2001, 2009; Royer et al., 2001b; Beerling and Royer, 2002a; Quan et al., 2009; Smith et al., 2010; Doria et al., 2011; Wang et al., 2015; Barclay and Wing, 2016). Other species, including other conifers (Passalia, 2009; Steinthorsdottir and Vajda, 2013; Liu et al., 2016), cycads (McElwain et al., 1999; Haworth et al., 2011b), Quercus petraea (van der Burgh et al., 1993; Kürschner et al., 1996), Q. guyavifolia (Hu et al., 2015; Huang et al., 2016), members of the Lauraceae (McElwain, 1998; Greenwood et al., 2003; Kürschner et al., 2008) and Betula species (Finsinger and Wagner-Cremer, 2009), have been used as proxies much less frequently because of their limited number of fossil relatives. Clearly, identification of additional proxies that are sensitive to atmospheric pCO₂ and also have numerous fossil relatives is highly desirable.

The ring-cupped oaks [Quercus section Cyclobalanopsis (Oerst.) Benth. & Hook. f., Fagaceae] (Denk et al., 2017), which today dominate sub-tropical East Asian forests (Zhou, 1993; Xu et al., 2015; Zhu et al., 2016; Deng et al., 2018), have rich fossil records in the Cenozoic sediments of East Asia ranging from the Eocene to Pliocene Epochs (e.g. Huzioka and Takahasi, 1970; Writing Group of Cenozoic Plants of China, 1978; Li, 2010; Shi, 2010; Xing et al., 2013; Hu et al., 2014; Xu et al., 2016; Barrón et al., 2017); therefore, they are excellent potential candidates for reconstructing the historical atmospheric CO₂ concentration of the Cenozoic Era. In this study, we selected Q. glauca Thunb., a dominant species in East Asian sub-tropical forests and one of the fossils of the NLRs of section Cyclobalanopsis, to determine how the stomatal frequency of Q. glauca responds to pCO₂ variation using all three material sources, i.e. seedlings grown in climatic chambers under elevated pCO₂; extant field samples collected along an altitudinal gradient; and historical herbarium specimens. The overarching aim of this study was to determine the suitability of Q. glauca as a proxy for palaeo-CO₂ concentrations.

MATERIALS AND METHODS

Experimental design in the climatic chambers

Quercus glauca seeds were collected from four different altitudes. The altitudes of vouchers DH359, DH358, DH349 and DH360 are 240, 314, 715 and 1940 m, respectively (Supplementary Data Table S1). Seeds were germinated in sandy beds to young seedlings with two or three leaves (Fig. 1A) and then transplanted to pots. Seedlings in pots were grown in four walk-in climatic chambers (Grandcool, Beijing, China) at the Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences (21°55′2.9″N, 101°16′39.7″E, altitude 548 m) with an atmospheric control of ambient air (approx. 400 ppm CO₂) or ambient air with elevated CO₂ concentrations (approx. 700, 1000 and 1300 ppm). All other growth conditions in the four chambers were kept the same: during the day (08.00–18.00 h) the temperature was 25 °C with a light intensity of approx. 300 μmol m^–2 s^–1; during the night (18.00–08.00 h) the temperature was 18 °C and dark; the relative humidity was maintained at 70 %; 300 mL of water was given every 3 d. The ranges of recorded error in CO₂ concentration, temperature and relative humidity around the treatment set points are listed in Supplementary Data Table S2. Only leaves growing from newly developed buds after transfer of the seedlings into the climatic chambers were recorded.

Fig. 1.

Photos of Quercus glauca in experimental and natural field conditions. (A) Young seedlings grown from seeds in sandy beds; (B) plants grown in the chambers for 1 year (photographed by Dr Li Wang); (C) plants in the natural field.

Plants were grown under two types of treatments in the four chambers: (1) plants collected from the same site were grown in different chambers under four CO₂ concentrations from 400 to 1300 ppm, i.e. pCO₂ approx. 38.035, 66.562, 95.088 and 123.614 Pa, respectively, in the four chambers (Supplementary Data Table S3); and (2) plants collected from different altitudes were grown in the same chamber under the same CO₂ concentration (700 or 1000 ppm, i.e. pCO₂ approx. 66.562 and 95.088 Pa; Supplementary Data Table S4). Vouchers DH349 and DH360 were used for treatment 1; vouchers DH359, DH358, DH349 and DH360 were used for treatment 2. Experimental plants grown in the chambers (Fig. 1B) appeared quite healthy when compared with those from the field (Fig. 1C). After approx. 1 year (from 1 January 2013 to 15 January 2014) in the climatic chambers, by which time plants had 8–20 leaves, the uppermost mature leaves, which received full irradiance, were sampled. Twenty-five individuals for each voucher were grown in each chamber and, after 1 year, 5–19 individuals were still alive; 2–4 leaves were collected from each of the surviving individuals; however, only one leaf was collected from some individuals which grew very slowly.

Collection of extant altitudinal samples

Altitudinal samples of Q. glauca were collected from five individuals at each of 14 sites with elevations ranging from 142 to 1555 m, which represents a pCO₂ of 32.886–38.838 Pa (Supplementary Data Table S5; Fig. 2). Four sun and shade leaves were collected from each individual tree to account for stomatal frequency variation (Poole and Kürschner, 1999; Beerling and Royer, 2002a), since light intensity may have a positive effect on stomatal frequency (Royer, 2001; McElwain, 2004; Kouwenberg et al., 2007).

Fig. 2.

Locations of 14 sampling sites (black points; B) in south-eastern China (A) for extant altitudinal samples of Quercus glauca.

Collection of historical herbarium specimens

Quercus glauca historical herbarium specimens that spanned the time period of 1930–2005 were obtained from the Herbarium of Kunming Institute of Botany, Chinese Academy of Sciences (KUN). We selected specimens collected from south-western China at similar altitudes (1000–1680 m) (Supplementary Data Table S6) to minimize altitude-induced error. Extant sample DH343, collected in the field, extended the time period to 2011. These samples represent a pCO₂ of 26.187–33.862 Pa (Supplementary Data Table S6). Two to three leaves for each target specimen were used.

Cuticle preparation and SD/SI counts

Mature leaves were chosen for cuticle preparation, which followed the methods of Stace (1965) and Poole and Kürschner (1999), and were photographed under a light microscope (Leica DM 1000) attached to a Leica DFC 295 camera. To minimize variability, fields of view were concentrated near the mid-lamina region in the intercostals (Poole et al., 1996). The size of the images for SD (number of stomata per mm²) and SI (proportion of stomata to the total number of epidermal cells) counts was approx. 0.1643 mm². The leaves of Q. glauca are hypostomatous (Deng et al., 2014); thus, stomatal and epidermal cell counts were made on the abaxial surface. The software package ImageJ version 1.42q was used for SD/SI counts.

For samples grown in the climatic chambers, three microscope fields per leaf were counted. The stomatal frequency of samples from the same voucher in the same chamber was averaged and the standard deviation was calculated. For extant altitudinal samples, three microscope fields were counted per leaf, resulting in 60 SD/SI counts (5 individuals × 4 leaves × 3 counts) for each of the 14 sites for both sun and shade leaves. For historical herbarium specimens, five microscope fields per leaf were counted, thus 10–15 counts were made for each specimen. Previous studies have showed that in Quercus sun leaves have a trait of straight to rounded epidermal cell walls, whereas shade leaves exhibit a pronounced undulation of the epidermal cell walls (Kürschner, 1997; Hu et al., 2015), and on this basis only sun leaves from the historical herbarium specimens were chosen for SD/SI counts. In fact, we found that sun leaves of Q. glauca were much easier to obtain than shade leaves in the herbarium, possibly because leaves from outer branches were more easily collected.

All cuticular slides were deposited at the Laboratory of Palaeoclimate Change and Plant Evolution Research Group in the Kunming Institute of Botany, Chinese Academy of Sciences. The protocols for cuticle preparation, as well as stomatal analysis for extant altitudinal samples and historical herbarium specimens, are detailed in Hu et al. (2015).

Data analysis

For samples in climatic chambers, the changes in SD and SI values under the two treatment types were illustrated as histograms using R. Levene’s test of equality of error variance was conducted and showed that the error variance of SD or SI was equal across treatment 1 (plants from the same voucher under four pCO₂) or treatment 2 (plants from four different altitudes under the same pCO₂), except for the SD from voucher DH349 in treatment 1. Differences in the SD or SI values across treatment 1 or 2 were tested using one-way analysis of variance (ANOVA) and least significant difference (LSD) tests, which were applied to each of the data sets with equal error variance. For the SD from voucher DH349 with unequal error variance, differences were compared using a non-parametric Kruskal–Wallis one-way ANOVA (k samples).

For extant altitudinal samples, calibration curves of SD or SI vs. pCO₂ for sun and shade leaves were constructed. Atmospheric pCO₂ used in these calibration curves were calculated from the different altitudes using eqn (1) (Beerling and Royer, 2002a, derived from Jones, 1992):

$$elev(p_2)=-\ln \left(\frac{p_2}{p_1}\right)\times \frac{R\times T}{(M_A\times g)}$$ (1)

where p₁ and p₂ are the CO₂ partial pressures (Pa) at sea-level and at the site, respectively; R is the gas constant (8.3144 Pa m³ mol^–1 K^–1); T is the mean annual temperature (K) of the range in elevation; M_A is the molecular weight of air (0.028964 kg mol^–1); g is the acceleration due to gravity (9.8 m s^–2); and elev (p₂) is the elevation (m) of the site. T was obtained by inputting the latitude, longitude and elevation of each site into CLAMP Climate Related Diagnostics, available from the BRIDGE website (https://www.paleo.bristol.ac.uk/ummodel/scripts/html_bridge/clamp_UEA.html). Because stomatal frequency responds to CO₂ partial pressure, the approximate values of pCO₂ in the four climatic chambers were also calculated (Supplementary Data Tables S3, S4) using eqn (1).

Calibration curves of SD or SI vs. pCO₂ for historical herbarium specimens were also constructed. Since the altitudes of the historical herbarium specimens were relatively high, from 1000 to1680 m, atmospheric pCO₂ at these altitudes was calculated by applying historical levels of atmospheric CO₂ at sea level of the collection time and altitudes to eqn (1). Historical atmospheric CO₂ concentrations at sea level before 1958 AD were obtained from Etheridge et al. (1996) and afterwards from the CO₂.Earth website (https://www.co2.earth/).

All calibration curves were generated using simple linear regression analysis, using R version 3.0.2 (http://www.R-project.org). Paired-samples t-tests were conducted to test the difference of the stomatal frequency between sun and shade leaves of the extant altitudinal samples; analysis of covariance was conducted to test the differences in slopes and y-intercepts of the constructed curves of sun and shade leaves as well as historical herbarium specimens by using SPSS Statistics version 19.0 (http://www.spss.com.cn). In this study, significance is defined as P < 0.05.

RESULTS

pCO₂-elevated experiment

Quercus glauca samples that were collected from two sites with different altitudes (vouchers DH349 and DH360) and grown in climatic chambers under CO₂ enrichment of 400–1300 ppm (treatment 1) displayed a reduction in mean SI from 13.1 (14.3) % to approx. 12 % (Fig. 3C, D; Supplementary Data Table S3); however, this inverse response was rather weak and the mean SD showed no apparent significant change with elevated pCO₂ (Fig. 3A, B).

Fig. 3.

Changes in stomatal frequency (A and B, stomatal density; C and D, stomatal index) for Quercus glauca (vouchers DH349 and DH360) under four CO₂ concentration gradients. Error bars represent + 1 s.d. Different letters above the histograms indicate a significant difference.

Quercus glauca samples collected from four different altitudes but modulated under the same pCO₂ for 1 year (treatment 2) displayed a similar SI (Fig. 4C, D; Supplementary Data Table S4). In particular, the SI values of the four vouchers in the chamber with a CO₂ concentration of 1000 ppm (pCO₂ 95.088 Pa) varied only from 12.1 to 12.6 %. The SD was more variable than the SI; the SD of the samples grown under 700 ppm CO₂ concentration (pCO₂ 66.562 Pa) exhibited similar values, but those grown under 1000 ppm CO₂ showed no regular response (Fig. 4A, B).

Fig. 4.

Changes in stomatal frequency (A and B, stomatal density; C and D, stomatal index) for Quercus glauca collected from different altitudes and treated under the same CO₂ concentration for 1 year. Histograms were labelled as the collection numbers and altitudes. Error bars represent + 1 s.d. Different letters above the histograms indicate a significant difference.

Stomatal frequency of Quercus glauca from altitudinal and herbarium samples

The calibration curves show a significant inverse linear correlation between SI and atmospheric pCO₂ for both sun and shade leaves of Q. glauca collected along an altitudinal gradient (Fig. 5D, E; Supplementary Data Table S5). There was a slight decrease of SD with increasing pCO₂ in sun leaves, but this was not significant (Fig. 5A). However, a significant inverse linear relationship between SD and pCO₂ was found in shade leaves (Fig. 5B). Moreover, there was no difference in SI between sun and shade leaves (P = 0.252). Further, the slopes (P = 0.933) and y-intercepts (P = 0.548) of their constructed curves were not different. However, the SD in sun leaves was slightly higher than that of shade leaves (P = 0.031).

Fig. 5.

The relationship between stomatal frequency (A–C, stomatal density; D–F, stomatal index) and CO₂ partial pressure of Quercus glauca sun (A, D) and shade (B, E) leaves of extant altitudinal samples, and from historical herbarium specimens (C, F). Error bars represent ± 1 s.d. The solid line indicates the best fit in a classical regression analysis. Dashed lines are 95 % confidence limits.

Similar to the response of the extant altitudinal samples, the SD and SI of historical herbarium specimens showed a significant inverse correlation with atmospheric pCO₂ (Fig. 5C, F; Supplementary Data Table S6).

There was no significant difference in the slopes and y-intercepts of SD/SI–pCO₂ curves between extant altitudinal samples and historical herbarium specimens (P = 0.497 and 0.171, respectively, for slope and y-intercept comparison of SD–pCO₂ curves; P = 0.969 and 0.441, respectively, for slope and y-intercept comparison of SI–pCO₂ curves). Thus, the SD/SI of extant altitudinal samples (sun and shade leaves) and historical herbarium specimens were combined to generate calibration curves (Fig. 6), which were of higher quality (R² = 0.652 for SD–pCO₂ curve; R² = 0.645 for SI–pCO₂ curve) than the individual curves.

Fig. 6.

Calibration curves of Quercus glauca constructed by combining stomatal frequency (A, stomatal density; B, stomatal index) of extant altitudinal samples (sun and shade leaves) and historical herbarium specimens (see key). Error bars represent ± 1 s.d. The solid line indicates the best fit in a classical regression analysis. Dashed lines are 95 % confidence limits.

DISCUSSION

An inverse response in pCO₂-elevated experiment

Quercus glauca seedlings grown in climatic chambers under four different CO₂ concentrations, ranging from 400 to 1300 ppm (treatment 1), showed an inverse relationship between stomatal frequency and pCO₂, while seedlings collected from different altitudes and grown for 1 year under the same pCO₂ (treatment 2) mostly displayed a similar SI. Both results point to pCO₂ as the main environmental factor controlling stomatal frequency.

Improved SF–pCO₂ relationship derived from three material sources

A significant inverse correlation between stomatal frequency and pCO₂ was found for Q. glauca from the three material sources, namely seedlings grown under elevated pCO₂, extant altitudinal samples and historical herbarium specimens. These results indicate that Q. glauca is sensitive to changes of pCO₂ and is an ideal proxy for palaeo-CO₂ levels. Further, our results confirm that combined use of these three material sources, to investigate the SF–pCO₂ relationship of a plant species, can overcome the limitations inherent to each material source (Woodward, 1988; Beerling and Chaloner, 1993a; McElwain and Chaloner, 1995; Hu et al., 2015; Barclay and Wing, 2016). These limitations have been highlighted by previous studies, using only one material source, which have reported different SF–pCO₂ relationships for the same species. For example, Pinus sylvestris showed a reduction in SD under a pCO₂-elevated treatment (Beerling, 1997; Lin et al., 2001), but Eide and Birks (2006) did not find a statistically significant relationship for both historical herbarium specimens and pCO₂-elevated experiments; Beerling and Chaloner (1993b) showed an inverse SD–pCO₂ correlation for Q. robur using historical herbarium specimens, while Atkinson et al. (1997) reported increased SD in this species under elevated pCO₂. However, until now, only a handful of studies have attempted the combined use of all three material sources. Using this approach, Royer et al. (2001b) and Barclay and Wing (2016) were able to generate high-quality SI–CO₂ inverse curves for Ginkgo and/or Metasequoia; however, they used only one or two field sampling sites to complement the historical herbarium data sets, not a series collected along an altitudinal gradient.

In this study, the three material types of Q. glauca were analysed independently yet produced comparable results. This confirms that pCO₂ is the main factor influencing stomatal frequency not only in historical herbarium specimens but also in extant altitudinal samples. This, in turn, demonstrates that extant field samples collected along an altitudinal gradient are also a reliable, yet hitherto underutilized, material source with great application potential. To date, only a few studies have used extant altitudinal samples to investigate the relationship between stomatal frequency and pCO₂ (McElwain, 2004; Eide and Birks, 2006; Kouwenberg et al., 2007; Hu et al., 2015). This probably reflects the fact that only a limited number of plant species are distributed over a wide enough altitudinal range. Nevertheless, our study clearly demonstrates that combining extant altitudinal samples with historical herbarium specimens can be advantageous, as it expands the range of pCO₂ and thus improves the reliability and accuracy of SF–pCO₂ curves.

Quercus glauca is one of the NLRs of section Cyclobalanopsis fossils which are widely distributed in the strata of East Asia ranging from the Eocene to Pliocene Epochs (Huzioka and Takahasi, 1970; Guo, 1978, 2011; Writing Group of Cenozoic Plants of China, 1978; Zhou, 1999; Xiao et al., 2006; Jia et al., 2009; Xia et al., 2009; Li, 2010; Shi, 2010; Xing et al., 2013; Hu et al., 2014; Jia et al., 2015; Xu et al., 2016; Barrón et al., 2017; Huang et al., 2017; Linnemann et al., 2017; Ding et al., 2018). These successive fossil records provide ideal materials to reconstruct the atmospheric CO₂ concentration history of the Cenozoic Era by applying the stomatal frequencies of closely related fossils to the constructed SF–pCO₂ curves of Q. glauca. Thus, these fossils will considerably increase the range of optimal proxies to estimate palaeo-CO₂ levels, beyond Ginkgo and Metasequoia. Recently, a new positive SF–pCO₂ relationship has been determined in Q. guyavifolia, the NLR of Q. preguyavifolia fossils (Hu et al., 2015) which coexisted with section Cyclobalanopsis fossils in many floras (Xing et al., 2012; Hu, 2013; Xu, 2016). Reconstructing palaeo-CO₂ concentrations using these two coexisting taxa with contrasting responses to pCO₂ (inverse in Q. glauca and positive in Q. guyavifolia) will provide independent results to cross-check the palaeo-CO₂ levels within the same time period.

A weak response to elevated pCO₂

Although the stomatal frequency of Q. glauca grown in climatic chambers showed an inverse response to atmospheric pCO₂, this response was rather weak. For example, voucher DH360 displayed higher SD and SI in the 700 ppm CO₂ treatment than in the 1000 ppm CO₂ treatment (P < 0.05), while voucher DH349 showed similar SD and SI between the 700 and 1000 ppm CO₂ treatments (P > 0.05) (treatment 1, Fig. 3). Additionally, SD of the samples from four different altitudes was higher in the 700 ppm CO₂ treatment than that in the 1000 ppm CO₂ treatment (P < 0.05) (treatment 2); however, there were no difference in SI between the two treatments (P > 0.05) (Fig. 4). These results indicate that while some plants in the chambers responded to the pCO₂-elevated treatment, others did not. Moreover, the SD and SI of these seedlings grown under elevated pCO₂ did not exhibit obviously lower values than those of historical herbarium specimens and extant altitudinal samples; in fact, their SD values (500–550 mm^–2) were similar to those of the extant altitudinal samples collected from a low altitude range (142–552 m, i.e. pCO₂ 37.007–38.838 Pa; Supplementary Data Table S5) and their SI values (12–13 %) were similar to the median value of the SI range for both historical herbarium specimens and extant altitudinal samples (Fig. 6; Supplementary Data Tables S5 and S6).

A possible explanation for the weak response and relatively high SD and SI values in the climatic chambers is incomplete phenotypic adaptation to elevated pCO₂. A previous study also observed relatively high SI in G. biloba grown at 1500 ppm CO₂; additionally, they found malformed stomata and high SI variance among these leaves, suggesting incomplete anatomical adjustment to elevated pCO₂ (Barclay and Wing, 2016). Moreover, it has been demonstrated that plants often need multi-year (at least two growing seasons) pCO₂-elevated treatments for their stomatal frequency to show a response (Royer, 2003; Overdieck and Strassemeyer, 2005; McElwain and Steinthorsdottir, 2017), and Hincke et al. (2016) showed that amplified adjustment of stomatal parameters in Betula nana occurred only in the second year of experimental pCO₂ exposure. Therefore, it is likely that the weak response and relatively high SD and SI values of Q. glauca reported here are due to insufficient exposure (<1 year) of the experimental plants to elevated pCO₂.

In addition, unsatisfactory simulation of natural field conditions within the climatic chambers may also have contributed to the weak response and relatively high SD and SI values observed here. We used a light intensity of 300 μmol m^–2 s^–1, because of technical limitations of our climatic chambers; this light intensity may be too low for Q. glauca, as this species occurs in sub-tropical East Asian forests (Zhou, 1993; Xu et al., 2015). Because of these two potential problems with our experimental design, we excluded the experimental data set from the calibration curves of SF and pCO₂ for Q. glauca.

SD vs. SI and sun vs. shade leaves

Our results confirm that SI is more reliable than SD. Previous work has shown that SD is area dependent and susceptible to environmental factors that affect epidermal cell expansion, such as temperature, water stress and humidity (Kürschner et al., 1996; Royer, 2001; Sun et al., 2003; Haworth et al., 2010a); however, SI can reduce the effect of these environmental factors. It follows that SI is a more precise parameter for investigating the SF–pCO₂ relationship and a more reliable proxy for palaeo-CO₂ estimates (McElwain, 2005; Kouwenberg et al., 2007). Indeed, in this study, we showed that the SI, but not the SD, of sun leaves from extant altitudinal samples had a significant inverse response to atmospheric pCO₂, and that the SI of seedlings in the climatic chambers exhibited a more pronounced response to elevated pCO₂ than their SD. These results confirm that SD varies more than SI and that it is, therefore, less reliable than SI for palaeo-CO₂ reconstruction. It is worth noting, however, that in cases when fossil leaves are not well preserved thus rendering SI analysis impossible, SD remains a viable option for palaeo-CO₂ reconstruction, although it may give rise to error.

Previous studies have shown that in many species the stomatal frequency of sun leaves is higher than that of shade leaves (Kürschner, 1997; Wagner, 1998; Kouwenberg et al., 2007) due to the positive effect of light intensity on stomatal frequency (Lake et al., 2001, 2002). Our study shows that Q. glauca sun leaves had a higher SD than shade leaves but had a similar SI. Since SD is more variable than SI, as also demonstrated by the exceptional lack of significant correlation between pCO₂ and SD of sun leaves (Fig. 5A), we conclude that light intensity has only a negligible effect on the stomatal frequency of Q. glauca. Therefore, it is feasible to combine sun and shade leaves together with historical herbarium specimens to generate SF–pCO₂ curves for this species. When applying these calibration curves to related fossils for estimation of palaeo-CO₂ concentrations, both sun and shade fossil leaves could also be used together. This represents an additional advantage of Q. glauca as a potential proxy for palaeo-CO₂ concentrations. In species where the stomatal frequency of sun leaves differs from that of shade leaves, it is important to distinguish between the two: combining only sun leaves from extant altitudinal samples with sun leaves from historical herbarium specimens results in more accurate calibration curves; it is also necessary to distinguish fossil sun and shade leaves for palaeo-CO₂ estimates.

Conclusions

We have shown a statistically significant inverse correlation between atmospheric pCO₂ and stomatal frequency in Q. glauca using samples from three sources: seedlings grown in climatic chambers under elevated pCO₂, extant altitudinal samples and historical herbarium specimens. These three types of samples were analysed independently, thus compensating for the disadvantages of each individual material type and allowing for cross-validation of different material sources. These three material types show the same response, indicating that Q. glauca is sensitive to atmospheric pCO₂ and is a potential proxy for palaeo-CO₂ levels. The combined calibration curves, which integrated the data from extant altitudinal samples and historical herbarium specimens, showed higher accuracy than the individual curves. Thus, we suggest that samples collected along an altitudinal gradient should be utilized more often to investigate the SF–pCO₂ correlation and that combining both extant altitudinal samples and historical herbarium specimens will improve the reliability and accuracy of the calibration curves and, thus, palaeo-CO₂ estimations.

The numerous Quercus section Cyclobalanopsis fossils from the Eocene to Pliocene Epochs in eastern Asia provide ideal materials to estimate the atmospheric CO₂ concentration history of the middle to late Cenozoic Era. However, although the seedlings of Q. glauca (the NLR of section Cyclobalanopsis fossils) from our pCO₂-elevated experiment showed an inverse SF–pCO₂ relationship, they displayed a weak response and relatively high SD and SI values. This is likely to be due to incomplete phenotypic adjustment to elevated pCO₂ because of too short exposure time (only 1 year) and unsatisfactory simulation of natural field conditions within the climatic chambers. Clearly, longer exposure to elevated pCO₂, >2 years, and better simulation of natural field conditions are recommended for future studies of stomatal frequency in tree species under elevated pCO₂.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: voucher, location and altitude of Q. glauca seedlings in the climatic chambers. Table S2: the treatment set points and range of recorded points in CO₂ concentration, temperature, relative humidity and light intensity in the climatic chambers. Table S3: stomatal density and stomatal index of Q. glauca under four pCO₂ gradients in the climatic chambers. Table S4: stomatal density and stomatal index of Q. glauca collected from different altitudes under the same pCO₂ in the climatic chambers. Table S5: location, altitude, pCO₂, stomatal density and stomatal index of Q. glauca sun and shade leaves where extant altitudinal samples were collected. Table S6: collection time, location, altitude, pCO₂, stomatal density and stomatal index of Q. glauca sun leaves from historical herbarium specimens.

FUNDING

This work was supported by the National Natural Science Foundation of China [grant no. 41702027 to J.J.H.], a Joint Fund from the National Natural Science Foundation of China and Yunnan Provincial Government [grant no. U1502231 to Z.K.Z.] and the NSFC-NERC (Natural Environment Research Council of the UK) joint research program [grant no. 41661134049 to T.S.].