A new positive relationship between pCO₂ and stomatal frequency in Quercus guyavifolia (Fagaceae): a potential proxy for palaeo-CO₂ levels

Abstract

Background and Aims The inverse relationship between atmospheric CO₂ partial pressure (pCO₂) and stomatal frequency in many species of plants has been widely used to estimate palaeoatmospheric CO₂ (palaeo-CO₂) levels; however, the results obtained have been quite variable. This study attempts to find a potential new proxy for palaeo-CO₂ levels by analysing stomatal frequency in Quercus guyavifolia (Q. guajavifolia, Fagaceae), an extant dominant species of sclerophyllous forests in the Himalayas with abundant fossil relatives.

Methods Stomatal frequency was analysed for extant samples of Q. guyavifolia collected from17 field sites at altitudes ranging between 2493 and 4497 m. Herbarium specimens collected between 1926 and 2011 were also examined. Correlations of pCO₂–stomatal frequency were determined using samples from both sources, and these were then applied to Q. preguyavaefolia fossils in order to estimate palaeo-CO₂ concentrations for two late-Pliocene floras in south-western China.

Key Results In contrast to the negative correlations detected for most other species that have been studied, a positive correlation between pCO₂ and stomatal frequency was determined in Q. guyavifolia sampled from both extant field collections and historical herbarium specimens. Palaeo-CO₂ concentrations were estimated to be approx. 180–240 ppm in the late Pliocene, which is consistent with most other previous estimates.

Conclusions A new positive relationship between pCO₂ and stomatal frequency in Q. guyavifolia is presented, which can be applied to the fossils closely related to this species that are widely distributed in the late-Cenozoic strata in order to estimate palaeo-CO₂ concentrations. The results show that it is valid to use a positive relationship to estimate palaeo-CO₂ concentrations, and the study adds to the variety of stomatal density/index relationships that available for estimating pCO₂. The physiological mechanisms underlying this positive response are unclear, however, and require further research.

INTRODUCTION

Palaeoatmospheric CO₂ (palaeo-CO₂) concentration estimates provide important palaeoenvironmental information in geological time and a baseline reference to understand future climatic change. Atmospheric CO₂ concentration has been hypothesized to be a primary determinant of global climate change; periods of low atmospheric CO₂ concentrations witnessed major glaciations, whereas those with higher CO₂ concentrations had warmer conditions (Retallack, 2001; Kürschner et al., 2008; Lunt et al., 2008; Lacis et al., 2010; Smith et al., 2010). This CO₂–temperature relationship is conspicuous during the Quaternary and has also been confirmed for other time periods: for example, the Paleocene–Eocene thermal maximum (PETM) was a brief but intense interval of global warming associated with elevated atmospheric CO₂ concentration (Zachos et al., 2005). In addition, CO₂ levels play a crucial role in affecting the ecology and physiology of plants.

To understand the relationship of CO₂, climate change and ecological function of CO₂, many attempts have been made to estimate palaeo-CO₂ throughout the Phanerozoic (Pagani et al., 1999; Pearson and Palmer, 2000; Berner and Kothavala, 2001; Berner, 2006; Tripati et al., 2009; Seki et al., 2010). A comparatively reliable method is to measure the CO₂ composition of air locked in glacial ice (Petit et al., 1999; Lüthi et al., 2008). However, this method can only be applied for the past 800 000 years because of the absence of older glacial ice (Lüthi et al., 2008). Pre-ice core CO₂ concentration estimations rely on numerous independent palaeobotanical and geochemical proxies and biogeochemical models derived from palaeobotanical and geochemical proxies, such as geochemical models (Berner and Kothavala, 2001; Berner, 2006), the δ¹³C in palaeosols (Ekart et al., 1999; Myers et al., 2012), δ¹³C of the organic remains of phytoplankton (Seki et al., 2010), the δ¹¹B and B/Ca ratio of marine carbonate (Pearson and Palmer, 2000; Tripati et al., 2009; Seki et al., 2010) and stomatal frequency [expressed as stomatal density (SD) or stomatal index (SI)] in fossil leaves (e.g. Royer et al., 2001; Kürschner et al., 2008). Among these approaches, the δ¹³C in palaeosols method and the stomatal frequency method are terrestrial-based proxies, and the others are marine-based proxies. However, estimates of the palaeo-CO₂ concentration made using these different approaches are quite variable (Royer et al., 2001a; Beerling and Royer, 2011). Thus, more research efforts focused on a single time period using different proxies is required. This is especially true for the mid-Miocene climatic optimum [18–15 million years ago (Ma)] and the middle to late Pliocene (3·6–2·6 Ma) (Beerling and Royer, 2011) because both of them were globally warm and relatively recent and may be comparable with the Earth’s immediate future with increasing greenhouse gases. Fortunately, there are abundant fossils from these two time periods (van der Burgh et al., 1993; Kürschner et al., 1996; Kürschner et al., 2008; Retallack, 2009; Stults et al., 2011) and we also found many oak fossils, providing an ideal opportunity to study the CO₂–temperature relationship during these warm climate intervals by estimating the palaeo-CO₂ concentration using the stomatal frequency method.

Generally, the stomatal frequency method is based on the inverse correlation between atmospheric CO₂ partial pressure (pCO₂) and leaf stomatal frequency which is species specific and observed in many C₃ plants (Woodward, 1987; Kürschner et al., 2001; Royer, 2001; Beerling and Royer, 2002a; Kouwenberg et al., 2003). The method has been widely used to estimate palaeo-CO₂ levels by applying the correlation to closely related plant fossils. Numerous genera and species have been used, such as Ginkgo (Retallack, 2001; Royer et al., 2001b; Beerling and Royer, 2002a; Retallack, 2009; Smith et al., 2010), Metasequoia (Royer et al., 2001b; Doria et al., 2011), other conifers (Passalia, 2009; Steinthorsdottir and Vajda, 2013), cycads (McElwain et al., 1999; Haworth et al., 2011b), Quercus (van der Burgh et al., 1993; Kürschner et al., 1996), Lauraceae (McElwain, 1998; Greenwood et al., 2003; Kürschner et al., 2008) and Betula (Finsinger and Wagner-Cremer, 2009). The precision of identifying nearest living relatives (NLRs) of fossil species and the accuracy of the correlation between stomatal frequency and pCO₂ can have profound effects on the estimates because the relationship is species specific. Royer (2001) summarized the stomatal frequency of 176 previously published C₃ plant species and showed that a majority of the species had inverse correlations; some had no significant relationship; only ≤12 % had a positive correlation; and species from the same genus may have inconsistent relationships (McElwain et al., 1995; Rundgren and Beerling, 1999; Eide and Birks, 2004; Finsinger and Wagner-Cremer, 2009; Haworth et al., 2010b). Almost all the previous studies have been based on an inverse relationship between atmospheric pCO₂ and stomatal frequency to estimate palaeo-CO₂ levels. Recently, a positive relationship between atmospheric pCO₂ and stomatal frequency in Typha orientalis was used to estimate palaeo-CO₂ levels during the Plio-Pleistocene transition (Bai et al., 2014). These studies indicate that it is essential to establish the specific stomatal frequency–pCO₂ relationship of a fossil’s NLR before using the relationship to estimate palaeo-CO₂ levels. To achieve this, many studies have used three primary approaches to investigate the response of stomatal frequency to CO₂ variation: (1) study leaves collected over an extended period of time because atmospheric CO₂ concentration has increased from approx. 280 to 390 ppm over the last 150 years; (2) study leaves from different locations along an altitudinal gradient because atmospheric pCO₂ declines as barometric pressure decreases with increasing altitude; and (3) greenhouse experiments in which stomatal frequency can be counted in response to elevated CO₂ concentrations. The first and third approaches have been widely used (Woodward, 1987; van der Burgh et al., 1993; Royer et al., 2001b; Greenwood et al., 2003; Kouwenberg et al., 2003; Haworth et al., 2011a), although the first approach may be somewhat constrained by the availability of historical specimens. The third approach misrepresents the potential for proportional population changes within a gene pool, and it does not consider taxonomic differences in plant generation times (Royer, 2001), and thus may fail to reflect long-term, genetic responses to slow changes in the environment of plants (Woodward, 1988; Beerling and Chaloner, 1993; McElwain and Chaloner, 1995). The second method has been used less frequently (McElwain, 2004; Kouwenberg et al., 2007) because of the difficulty of collecting one species over a long altitudinal gradient.

Oak plants of Quercus guyavifolia H. Lév. [= Q. pannosa Hand.-Mazz. (Flora of China), Q. guajavifolia H. Lév. (Flora of China, Volume 4, page 375)] is a dominant species in the sclerophyllous forests along the steep altitudinal gradients in the Qinghai-Tibet Plateau and Hengduan Mountains. There are also abundant Q. preguyavaefolia Tao (Zhou, 1992) fossils in the late-Cenozoic strata of this region (Zhou, 1999). For example, the Longmen flora (Su et al., 2013) and the Fudong flora (Tao, 1986; Huang et al., 2013) in south-western China, both from the Sanying Formation of the Pliocene, are dominated by Q. preguyavaefolia fossils; the Namling flora in Tibet from the Wulong Formation (middle Miocene, 15 Ma) is the earliest recorded occurrence of Q. preguyavaefolia fossils (Li and Guo, 1976; Spicer et al., 2003). Together, these fossils provide ideal material to estimate the atmospheric CO₂ concentration history of the late Cenozoic and to study the CO₂–temperature relationship during warm climate intervals in the mid-Miocene climatic optimum and the middle to late Pliocene. In this study, we chose Q. guyavifolia (the NLR of Q. preguyavaefolia fossils, Fig. 2) in order to (1) determine how the stomatal frequency of Q. guyavifolia responds to decreasing pCO₂ (increasing altitude) and to generate calibration curves of stomatal frequency vs. atmospheric pCO₂; (2) test whether samples collected along an altitudinal gradient provide results consistent with results from historical herbarium specimens; (3) estimate the late-Pliocene atmospheric CO₂ concentration using two contemporaneous Q. preguyavaefolia fossils; and (4) compare CO₂ levels estimated using the stomatal frequency of Q. guyavifolia with previous estimates. This is the first study to use both extant field collections from along an altitudinal gradient and historical herbarium specimens to establish a specific stomatal frequency–pCO₂ relationship.

Fig. 2.

Comparisons of leaf morphology of extant Quercus guyavifolia and fossil Q. preguyavaefolia. (A, B) Branches of extant Q. guyavifolia. (C, D) Cleared leaves of extant Q. guyavifolia. (E, F) and (G, H) are leaf fossils from the Hunshuitang flora and the Qingfucun flora respectively. Scale bars = 1 cm.

MATERIALS AND METHODS

Collection of extant altitudinal material, historical herbarium specimens and fossil samples

Extant sun and shade leaves of Quercus guyavifolia were collected from five individuals at each of 17 sites at elevations ranging from 2493 to 4497 m (the altitudinal range of Q. guyavifolia distributions is approx. 2500–4500 m), i.e. pCO₂ 22·695–29·134 Pa (Supplementary Data Table S1; Fig. 1). Sun and shade leaves were collected because light intensity affects these two types of leaves differently and has a positive effect on the stomatal frequency (Kürschner, 1997; Royer, 2001; Lake et al., 2002; McElwain, 2004; Kouwenberg et al., 2007). Sun leaves were collected from outer branches; shade leaves were collected from within and beneath canopies. To account for the high natural population variability in stomatal frequency (Poole and Kürschner, 1999; Beerling and Royer, 2002a), four sun and four shade leaves were collected from each of five Q. guyavifolia trees at each site.

Fig. 1.

The locations of 17 sites (black points) where extant field samples of leaf materials of Quercus guyavifolia were collected and two sites (purple stars) where fossil materials of Q. preguyavaefolia were collected. (A) The study area. (B) Locations where extant and fossil leaf materials were collected. (C) Detailed location map showing 16 of the 17 collection sites in the boundary (grey line) region between Yunnan and Sichuan Provinces.

Herbarium samples of Q. guyavifolia from the Herbarium of the Kunming Institute of Botany, Chinese Academy of Sciences (KUN) were examined and they spanned the time period 1926–1995. Specimens from similar altitudes (range 2800–3300 m) were chosen (Supplementary Data Table S2) so as to limit stomatal frequency variation that might be caused by factors associated with changing altitude. Sample HS006 collected in the field extends the time period to 2011, i.e. pCO₂ 21·207–26·539 Pa (Table S2).

Fossil leaves of Q. preguyavaefolia (Fig. 2E–H) were collected from two different late-Pliocene floras – the Hunshuitang and the Qingfucun floras (Fig. 1). The Hunshuitang flora collection site is located 25 km north-east of Kunming, central Yunnan Province, south-western China (25°06'N, 102°57'E, altitude 2102 m). The Hunshuitang flora belongs to the Ciying Formation from the late Pliocene (Yunnan Bureau of Geology and Mineral Resources, 1978, 1990; Jiang et al., 2003). The Qingfucun flora, located in Weixi county, north-western Yunnan (27°05'N, 99°21'E, altitude 2476 m), is from the Sanying Formation, late Pliocene (3·6 Ma; Li et al., 2013). We analysed five fossil specimens of Q. preguyavaefolia from the Hunshuitang flora and four from the Qingfucun flora.

Stomatal analysis of extant samples collected along an altitudinal gradient

Mature leaves (1–2 years of leaf growth, mostly 2 years) were chosen for cuticle preparation. Leaf fragments of Q. guyavifolia were macerated using a 1:1 solution of glacial acetic acid and 30 % H₂O₂ at 80°C for approx. 3 h after the thick trichomes were removed. The lower cuticles without mesophyll tissue were stained using safranin O solution and then mounted in glycerine on glass slides (Stace, 1965; Poole and Kürschner, 1999). Cuticle images were taken using a light microscope (Leica DM 1000) attached to a Leica DFC 295 camera. The size of the images for stomatal and epidermal cell counts was 468 × 351 µm² (approx. 0·1643 mm²). In addition, cleared leaves of Q. guyavifolia (Fig. 2C, D) were made following the protocol of Hickey and Wolfe (1975) to compare the leaf morphology of extant and fossil material.

Stomatal density (SD) is measured as the number of stomata per mm² of leaf. Stomatal index (SI) is the proportion of stomata to the total number of epidermal cells and measured as:

$$\text{SI}\left(\%\right)=\left[\text{SD}/\left(\text{SD}+\text{ED}\right)\right]\times \text{1}00(\text{Salisbury},\text{1927})$$ (1)

where ED is epidermal cell density.

Stomatal and epidermal cells were counted using the software package ImageJ version 1.42q (http://rsb.info.nih.gov/ij). Samples came from five individuals at each site; four leaves were taken from each individual; three microscope fields were counted per leaf. The leaves of Q. guyavifolia are hypostomatous (He et al., 1994) so the images of stomatal and epidermal cell counts were all made on the abaxial surface. This resulted in 60 counts (5 individuals × 4 leaves × 3 counts) for each site. Stomatal and epidermal cells were counted in intercostal areas and restricted to the mid-lamina region to minimize variability (Poole et al., 1996). The SD and SI for each site were calculated as the mean of 60 counts per site.

Stomatal analysis of historical herbarium material

The experimental protocols for cuticle preparation of herbarium samples were the same as for extant field material. Three mature leaves were collected from each historical herbarium specimen; five microscope fields were counted per leaf, thus 15 counts were made for each specimen.

Stomatal analysis of fossil material

Fossil leaf fragments were treated successively with 20 % HCl, 40 % HF and 20 % HCl again to remove calcareous and siliceous materials, and then macerated using 3·5 % NaClO solution for 10 min to 1 h until they became white or translucent. After removing the mesophyll tissue, the lower cuticles were stained using safranin O and mounted in glycerine on glass slides (Ye, 1981; Kerp, 1990; Leng, 2000). Cuticular images were taken using a light microscope (Leica DM 750) linked to a Leica DFC 295 camera. The size of the images was 298 × 223 µm² (approx. 0·0665 mm²). Five to ten cuticular images were counted for each fossil; a separate mean of the counts was calculated for the two fossil sites. The cuticles examined in fossil leaves were from the same part of the leaf as those for the extant leaves (intercostal area near mid-lamina). All cuticular slides were deposited at the Laboratory of Environmental Change and Its Impact on Plants Group in the Kunming Institute of Botany, Chinese Academy of Sciences.

Calibration curves

Calibration curves of stomatal frequency vs. pCO₂ for sun and shade leaves were constructed based on extant samples. Atmospheric pCO₂ used in the calibration curves were calculated from the elevation range using eqn (2) (Beerling and Royer, 2002a, derived from Jones, 1992):

$$elev(p_2)=-\ln \left(\frac{p_2}{p_1}\right)\times \frac{R\times T}{(M_A\times g)},$$ (2)

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 m³mol⁻¹K⁻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. As altitude increases, pCO₂ decreases from 29·134 to 22·695 Pa (Supplementary Data Table S1). The calculated atmospheric pCO₂ should have no significant difference from the field atmospheric pCO₂ (McElwain, 2004).

Calibration curves were constructed using linear regression analysis, using R version 2.14.1 (http://www.R-project.org). To determine if sun and shade leaves should be analysed separately or combined, differences in slopes and y-intercepts of their constructed curves were tested by analysis of covariance using SPSS Statistics version 18.0 (http://www.spss.com.cn).

Calibration curves of stomatal frequency vs. pCO₂ were also constructed using historical herbarium materials. Historical levels of atmospheric CO₂ at sea level were obtained from Etheridge et al. (1996) and from the CO₂ Now website (http://co2now.org/). Using this information, atmospheric pCO₂ at the sites was calculated using eqn (2).

Palaeo-CO₂ estimate

The stomatal frequency of the fossil material was analysed and applied to the calibration curves prepared using the extant field materials and historical herbarium samples, respectively, to estimate palaeo-CO₂ levels during the late Pliocene. Envelopes of uncertainty were obtained after propagating uncertainties in the calibration function and fossil leaf measurements. Due to differences between sun and shade leaves, it was necessary to construct morphotype-specific calibration curves for each type of leaf. Sun leaves are characterized by straight to rounded epidermal cell walls (Fig. 3A, B) whereas shade leaves show a pronounced undulation of the epidermal cell walls (Fig. 3C, D) (Kürschner, 1997), and on this basis fossils from the Hunshuitang flora (Figs 2E, F and 3E, F) and the Qingfucun flora (Figs 2G, H and 3G, H) were all assessed to be sun leaves. Sun leaves of historical herbarium samples were also chosen according to their epidermal features. This necessarily determined that the calibration curve for CO₂ estimation was prepared exclusively using sun leaves of Q. guyavifolia; the calibration curve generated using shade leaves will be used in future work for Q. preguyavaefolia fossils from different sites. We then compared our atmospheric CO₂ concentration estimates from the Hunshuitang and Qingfucun floras with other published results of atmospheric CO₂ concentration during the late Pliocene (van der Burgh et al., 1993; Kürschner et al., 1996; Pearson and Palmer, 2000; Tripati et al., 2009; Seki et al., 2010).

Fig. 3.

Images of the cuticle of sun (A, B) and shade (C, D) leaves of extant Quercus guyavifolia and Q. preguyavaefolia fossils from the Hunshuitang (E, F) and Qingfucun (G, H) floras. Scale bars = 50 µm. Black arrows indicate the undulant epidermal cell walls in shade leaves.

RESULTS

Stomatal frequency of extant Quercus guyavifolia and calibration curves

The calibration curves show a significant (P ≤ 0·01) positive linear relationship between stomatal frequency and atmospheric pCO₂ for both sun and shade leaves in extant field samples of Quercus guyavifolia collected along an altitudinal gradient (Supplementary Data Table S1; Fig. 4).

Fig. 4.

Relationship between stomatal frequency (A, B, stomatal density; and C, D stomatal index) and CO₂ partial pressure of Quercus guyavifolia sun (A, C) and shade (B, D) leaves. Error bars are ±1 s.d. The solid line indicates the best fit in classical regression analysis. Dashed lines are 95 % confidence limits.

The slopes of the calibration curves constructed by sun and shade leaves, respectively, are not different (P > 0·05). The SD and SI of shade leaves were 1·4 and 7·8 %, respectively, lower than those of sun leaves. There was a significant difference in intercept of SI–pCO₂ curves between sun and shade leaves (P < 0·05), but not for SD–pCO₂ curves (P > 0·05), indicating that the SI of shade leaves was lower than that of sun leaves, but the SD was not.

Stomatal frequency of historical herbarium materials and calibration curves

There was no significant relationship between SD and atmospheric pCO₂ for historical herbarium materials, but a significant (P < 0·05) positive linear relationship between SI and pCO₂ was found (Fig. 5).

Fig. 5.

Relationship between stomatal frequency (A, stomatal density and B, stomatal index) and CO₂ partial pressure of Quercus guyavifolia historical herbarium specimens. Error bars are ±1 s.d. The solid line indicates the best fit in classical regression analysis. Dashed lines are 95 % confidence limits.

Palaeo-CO₂ estimate of the late Pliocene

The relationship between SI and atmospheric pCO₂ for both extant specimens collected along an altitudinal gradient and historical herbarium materials were compared (Fig. 6). Results from both sources showed a significant positive linear relationship between SI and pCO₂ (Fig. 6).

Fig. 6.

Correlation curves constructed using the SI of extant field collections along an altitudinal gradient and historical herbarium samples (see key), and comparison of estimated palaeo-pCO₂ values (fossils from the Hunshuitang and Qingfucun flora, as indicated).

Fossils from the Hunshuitang flora had a mean SD of 516 ± 91 mm⁻², and a mean SI of 6·69 ± 0·76 % (Table 1). Using the SD and SI of fossil material, the palaeo-CO₂ concentration was estimated using the equations in Fig. 4A and 4C, respectively, which were both derived from extant sun leaves (Table 1). The palaeo-CO₂ concentration was also estimated using the equation in Fig. 5B which was derived from sun leaves of historical specimens (Table 1; Fig. 6).

Table 1.

Estimates of CO₂ partial pressure (pCO₂) derived by using the calibration curves of stomatal frequency and pCO₂ in Quercus guyavifolia sun leaves from both extant field samples (Fig. 4A, C) and historical herbarium samples (Fig. 5B)

Sample	Age	No. of fossils	Total image counts	SF (mean ± s.d.)	pCO2-E (mean ± s.e.)	pCO2-H (mean ± s.e.)
Hunshuitang flora	Late Pliocene	5	46	SD: 516 ± 91 mm−2	31·46 ± 2·79 Pa (310·51 ± 27·51 ppm)	–
				SI: 6·69 ± 0·76 %	21·01 ± 3·47 Pa (207·36 ± 34·25 ppm)	18·60 ± 5·53 Pa (183·52 ± 54·63 ppm)
Qingfucun flora	Late Pliocene, 3·6 Ma	4	34	SD: 496 ± 24 mm−2	30·59 ± 1·57 Pa (301·92 ± 15·47 ppm)	–
				SI: 7·79 ± 0·1 %	23·84 ± 2·57 Pa (235·26 ± 25·37 ppm)	21·15 ± 3·84 Pa (208·77 ± 37·87 ppm)

SF, stomatal frequency; pCO₂-E, estimates of CO₂ partial pressure from extant field samples; pCO₂-H, estimates of CO₂ partial pressure from historical herbarium samples.

Fossils from the Qingfucun flora had a mean SD of 496 ± 24 mm⁻², and a mean SI of 7·79 ± 0·1 % (Table 1). Palaeo-CO₂ levels of the late Pliocene were estimated by applying the equations in Fig. 4A and 4C to the SD and SI, respectively, of the fossils (Table 1). The palaeo-CO₂ was also estimated using the equation in Fig. 5B to compare with the estimates from the extant field samples (Table 1; Fig. 6).

DISCUSSION

Positive relationship between stomatal frequency and pCO₂ in Quercus guyavifolia

We found a positive relationship between stomatal frequency and atmospheric pCO₂ in Q. guyavifolia sun and shade leaves collected along an altitudinal gradient, which is consistent with the pattern from the historical herbarium samples. It is an unusual and interesting phenomenon because most other species show an inverse relationship (Woodward, 1987; van der Burgh et al., 1993; Woodward and Kelly, 1995; Beerling and Royer, 2002b; Royer, 2003; Kouwenberg et al., 2007; Franks and Beerling, 2009). However, several studies also report that in some species stomatal frequency increases with atmospheric CO₂ concentration (Ferris and Taylor, 1994; Royer, 2001). In particular, some Quercus species such as Q. rubra and Q. robur grown in climate-controlled greenhouses show an increase in SD at elevated CO₂ (Dixon et al., 1995; Atkinson et al., 1997). Although greenhouse results are not necessarily reliable, these results indicate to a certain extent that a positive relationship between stomatal frequency and atmospheric pCO₂ in other species of Quercus may be not unexpected. Recently, this positive correlation between SI and atmospheric pCO₂ in historical herbarium specimens of Typha orientalis has been used to estimate atmospheric CO₂ during the Plio-Pleistocene transition (Bai et al., 2014). This is supportive evidence of the reliability of the positive relationship used to estimate palaeo-CO₂.

The positive relationship detected between stomatal frequency and atmospheric pCO₂ in Q. guyavifolia was observed from both the altitudinal collection and the historical herbarium samples. These two independently derived results showed the same pattern. The altitudinal samples were collected in Yunnan and Sichuan provinces; the historical samples were collected from similar altitudes in northern and north-western Yunnan province. To test if this unusual, positive relationship between pCO₂ and stomatal frequency is affected by other environmental factors, the relationships between stomatal frequency and other climatic factors (mean annual temperature, mean annual precipitation, annual mean relative humidity) were tested by simple linear regression analysis. The result showed that only atmospheric pCO₂ significantly correlates to stomatal frequency of Q. guyavifolia (Figs 4 and 5) rather than the mean annual temperature (P > 0·05 for both the altitudinal collection and the historical herbarium samples), mean annual precipitation (P > 0·05 for both the altitudinal collection and the historical herbarium samples) and annual mean relative humidity (P > 0·05 for both the altitudinal collection and the historical herbarium samples), confirming that the positive relationship between stomatal frequency and pCO₂ in Q. guyavifolia is determined by atmospheric pCO₂. Together with evidence from other studies (Dixon et al., 1995; Atkinson et al., 1997; Bai et al., 2014), we propose that the positive relationship between stomatal frequency and pCO₂ in Q. guyavifolia is reliable and this relationship can be used as a basis to estimate palaeo-CO₂ levels.

The physiological mechanism underlying the positive stomatal response to pCO₂ is probably complicated. Stomata play a central role in the uptake of photosynthetic CO₂ and water loss from the leaf. Both physiological (stomatal aperture change) and morphological (SD change) strategies can be used by plants to regulate gas exchange (Haworth et al., 2013). The positive stomatal response to pCO₂ may be contributed by multiple factors. One of the factors is possibly leaf nitrogen content. Previous studies found that elevated CO₂ concentration can increase leaf nitrogen content in jack pine and white birch seedlings (Zhang and Dang, 2005). Plants of Q. pannosa (= Q. guyavifolia) and Q. aquifolioides grown at lower altitudes (higher atmospheric pCO₂) have higher leaf nitrogen content (or higher nitrogen allocation in the photosynthetic system), stomatal conductance, photosynthetic rate and carboxylation rate (Zhang et al., 2005; Feng et al., 2013). High leaf nitrogen content significantly increases carboxylation capacity (Rogers et al., 1998; Pérez et al., 2011) and consequently results in a decrease in the ratio of the intercellular to atmospheric CO₂ concentration and an increase in δ¹³C values (Sasakawa et al., 1989; Cordell et al., 1999), and therefore an increase of SD (Qiang et al., 2003).

Differences in stomatal frequency between sun and shade leaves

Our results confirm that distinguishing sun leaves from shade leaves is necessary when using the stomatal method to estimate palaeo-CO₂ levels (Kürschner, 1997). Stomatal frequency of both sun and shade leaves in Q. guyavifolia decreased with decreasing pCO₂. However, SI of shade leaves was lower than that of sun leaves although the SD was not different between sun and shade leaves. This corroborates results from previous studies which showed that stomatal frequency of sun leaves is higher than that of shade leaves (Kürschner, 1997; Wagner, 1998; Royer, 2001; Kouwenberg et al., 2007) resulting from the positive effect of light intensity on stomatal frequency (Lake et al., 2001, 2002; Kouwenberg et al., 2007). Kürschner (1997) reported that the SD of Q. petraea sun leaves was about 45 % higher than that of shade leaves, and up to 60 % higher in fossil leaves of Q. pseudocastanea. Kouwenberg et al. (2007) also observed a higher SD and SI in sun leaves compared with shade leaves in both Q. kelloggii (up to 38 % higher) and Nothofagus solandri.

Comparison of stomatal density and stomatal index

The SD and SI gave different estimates for palaeo-CO₂. The SI is a more precise proxy for palaeo-CO₂ estimation than SD because SI removes the effect of other environmental factors such as temperature, water stress and humidity on the size and/or spacing of epidermal cells, which will result in higher or lower SDs (Salisbury, 1927; Kürschner et al., 1996; Kürschner, 1997; Royer, 2001; Sun et al., 2003; McElwain, 2005; Kouwenberg et al., 2007; Haworth et al., 2010a). Here we showed that the SI of historical herbarium samples had a significant positive response to atmospheric pCO₂ but SD did not, confirming that SD is more variable than SI and therefore not as reliable as SI for palaeo-pCO₂ estimation. However, in practice, many fossil leaves are not well preserved and epidermal cells are difficult to identify. Thus SI analysis is impossible and SD becomes the sole option but may give rise to error. Fortunately, our fossils were well preserved and SI could be accurately calculated. Therefore, our results are probably more precise than those derived only by SD. We also used SD for palaeo-pCO₂ calculation in order to compare with the results derived from SI (Table 1). Our comparison of palaeo-pCO₂ levels estimated from SD and SI provides an example of overpredicting pCO₂ levels using SD.

Late-Pliocene atmospheric CO₂ levels

As a test of the applicability and reliability of the positive relationship of stomatal frequency and pCO₂ in Q. guyavifolia as a proxy for palaeo-pCO₂ levels, palaeo-pCO₂ was estimated using Q. preguyavaefolia fossils. Our results derived using the SI calibration data set indicate late-Pliocene atmospheric CO₂ levels of approx. 210 and 240 ppm derived using extant field collections and approx. 180 and 210 ppm derived from herbarium samples, which are consistent with most other estimates. Beerling and Royer (2011) compiled 370 estimates of Cenozoic atmospheric CO₂ levels obtained using different protocols; they show that since the Miocene most atmospheric CO₂ estimates were lower than present-day levels of approx. 390 ppm. We compared our palaeo-CO₂ estimates during the late Pliocene with those based on stomatal frequency of other species (terrestrial-based proxy) and other proxies such as alkenone and boron (marine-based proxies) for the same time period (Fig. 7). All of the estimates indicate palaeo-CO₂ levels between approx. 190 and approx. 430 ppm for the late Pliocene, with the majority indicating that atmospheric CO₂ levels in the late Pliocene were lower than at present. However, even atmospheric CO₂ estimates derived from the same source (terrestrial-based or marine-based proxies) are still quite different. Therefore, atmospheric CO₂ estimates to date have not been sufficient to account for the warm climate interval during the late Pliocene. More research is required to obtain more precise estimates of palaeo-CO₂ during this period.

Fig. 7.

Estimates of palaeo-CO₂ concentration during the late Pliocene derived using different methods. The palaeo-CO₂ estimates derived from alkenone (see key; black lines above and below represent uncertainties) and boron (purple) are cited from Seki et al. (2010) modified by Beerling and Royer (2011); alkenone (blue lines above and below represent uncertainties; Zhang et al., 2013); alkenone (dark green lines above and below represent uncertainties; Badger et al., 2013); boron data (orange; Bartoli et al., 2011); stomata data (van der Burgh et al., 1993; Kürschner et al., 1996); stomata data (Stults et al., 2011); stomatal data (red) and stomatal data (purple) are palaeo-CO₂ levels estimated from field collections along an altitudinal gradient and from historical herbarium samples respectively (this study). Error bars represent uncertainties (see Materials and Methods). One of our fossils is 3·6 Ma, and the other is unknown (sometime during the late Pliocene). The period between the two vertical dashed lines is the late Pliocene (3·6–2·588 Ma) (International Chronostratigraphic Chart; version 2013). The blue horizontal dashed line indicates recent levels of atmospheric CO₂ concentration (390 ppm).

In addition, our comparatively low atmospheric CO₂ values may be related to elevations in the fossil sites. It is generally accepted that the Qinghai-Tibet Plateau reached its current height in the late Miocene and retained it (Spicer et al., 2003); previous studies have also showed that western Yunnan had approached its highest altitude before the late Pliocene (Sun et al., 2011). The palaeoaltitude of the Xianfeng flora, about 60 km away from the Hunshuitang locality, was about 1936 m in the late Miocene (Jacques et al., 2014). This means that the altitudes of our fossil sites in the late Pliocene were probably similar to present-day altitudes (approx. 2000–2500 m). If the palaeoaltitudes of the two sites are taken into account, atmospheric CO₂ levels during the late Pliocene should be approx. 270 and 320 ppm derived by extant field collections and approx. 240 and 280 ppm derived by herbarium samples. Thus, these results are very similar and correspond well to other estimates, confirming that the positive relationship between stomatal frequency and atmospheric pCO₂ in Q. guyavifolia is reliable as a proxy for estimating palaeo-CO₂ levels.

Actually, a correction for pCO₂ at the elevation is necessary to obtain more accurate results. Clearly, when attempting to reconstruct palaeo-pCO₂ using the stomatal frequency of fossils, the pCO₂ estimates will be influenced by the elevation at which the now-fossilized plants were growing. Thus, to obtain more precise pCO₂ estimates it is essential to apply a correction factor for pCO₂ related to altitude. This should also be done for samples collected along an altitudinal range (McElwain, 2004) and for historical herbarium samples (Greenwood et al., 2003) when constructing the stomatal frequency–pCO₂ curve. Unfortunately, few previous studies have made this correction on either extant specimens, herbarium specimens or fossils. Therefore, we suggest that future studies should incorporate a correction factor related to altitude for all specimens and samples.

Recently, there has been much debate about whether the linear relationship between stomatal frequency and pCO₂ continues when the ambient CO₂ concentration rises beyond approx. 500 ppm, because the stomatal frequency of some plant species may lose sensitivity at these high levels (Woodward and Bazzaz, 1988; Kürschner et al., 1997; Beerling and Royer, 2002a, b; Beerling et al., 2009; Haworth et al., 2011a). Nevertheless, the pCO₂ estimate reported in this study, and the stomatal frequency–pCO₂ training sets are applicable because atmospheric CO₂ levels during the Pliocene remained below 500 ppm according to our results and those of previous estimates (Fig. 7).

Conclusions

We have shown a significant positive relationship between atmospheric pCO₂ and stomatal frequency in Q. guyavifolia that can be used as a proxy to estimate late-Cenozoic palaeo-CO₂ concentrations. This is the first study in which both field samples collected along an altitudinal gradient and historical herbarium samples of a single species have been used to estimate palaeo-CO₂ concentration. In addition, a positive relationship between stomatal frequency and pCO₂ has seldom been used to estimate palaeo-CO₂ concentration. Our estimated palaeo-CO₂ concentration provides new independent data for late-Cenozoic CO₂ estimates derived using vascular land plants. There are, however, three sources of variation. First, sun and shade leaves give different results of stomatal frequency, confirming that it is necessary to analyse sun and shade leaves separately when using the stomatal method to estimate palaeo-CO₂. Secondly, the estimates derived from SD and SI using sun leaves were also different, so more precise estimates will be possible when we find and analyse fossils that have well-preserved shade leaves in addition to sun leaves. Thirdly, our research concluded that atmospheric CO₂ levels in the late Pliocene were approx. 180–240 ppm. Although these results are consistent with other studies reporting CO₂ levels lower than modern atmospheric concentrations, when more appropriate fossil material becomes available then much more accurate estimates will be possible.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxford journals.org and consist of the following. Table S1: location, altitude, pCO₂, stomatal density and stomatal index of Q. guyavifolia sun and shade leaves where extant field samples were collected. Table S2: collection time, location, altitude, pCO₂, stomatal density and stomatal index of Q. guyavifolia sun leaves from historical herbarium specimens.