Evolutionary History of Atmospheric CO2 during the Late Cenozoic from Fossilized Metasequoia Needles

The change in ancient atmospheric CO2 concentrations provides important clues for understanding the relationship between the atmospheric CO2 concentration and global temperature. However, the lack of CO2 evolution curves estimated from a single terrestrial proxy prevents the understanding of climatic and environmental impacts due to variations in data. Thus, based on the stomatal index of fossilized Metasequoia needles, we reconstructed a history of atmospheric CO2 concentrations from middle Miocene to late Early Pleistocene when the climate changed dramatically. According to this research, atmospheric CO2 concentration was stabile around 330–350 ppmv in the middle and late Miocene, then it decreased to 278–284 ppmv during the Late Pliocene and to 277–279 ppmv during the Early Pleistocene, which was almost the same range as in preindustrial time. According to former research, this is a time when global temperature decreased sharply. Our results also indicated that from middle Miocene to Pleistocene, global CO2 level decreased by more than 50 ppmv, which may suggest that CO2 decrease and temperature decrease are coupled.


Introduction
Carbon dioxide (CO 2 ) is an important greenhouse gas that influences the surface temperature of the Earth [1]. The 5 th report of IPCC concluded [2] that the present positive radiative forcing is unequivocally caused by anthropogenic increases in atmospheric CO 2 concentration and that it influences the climate [3,4]. Estimating the impact of high CO 2 concentration on global environmental systems is the first step to propose solutions for the present global climate change. This impact can be unraveled by a better understanding of the relationship between the paleo-atmospheric CO 2 concentration (paleo-[CO 2 ] atm ) and ancient climate change.
A lot of research has involved the estimation of paleo-[CO 2 ] atm to understand the correlation between CO 2 and global warming. To obtain the paleo-[CO 2 ] atm values three major approaches have been used: (1) geochemical modeling (GCS) [5][6][7], (2) composition measurements of air trapped in ice cores [8], and (3) various proxies (reviewed in [9]). Geochemical modeling (GCS) can reconstruct paleo-[CO 2 ] atm , but for long geological time scales its resolution cannot be fine enough to show the details of paleo-[CO 2 ] atm fluctuation [10]. Ice core analysis is the most reliable method to measure paleo-[CO 2 ] atm directly, but is only applicable after 0.8 Ma [8]. Several CO 2 proxies have been used to estimate paleo-[CO 2 ] atm , such as the carbon isotope composition of phytoplankton, the boron (B) isotope composition of fossil foraminifera, the carbon isotope composition of carbonates in paleosol, and the stomatal parameters of fossil leaves [11]. High resolution records for CO 2 can be obtained from marine sediments with the two former proxies, but these do not directly show the paleo-[CO 2 ] atm . The latter two proxies are terrestrial-based proxies that reflect paleo-[CO 2 ] atm directly, although they rarely provide continuous paleo-[CO 2 ] atm records for a long geological time. Therefore, while there is a consensus on the general tendency of the Cenozoic paleo-[CO 2 ] atm changes, the estimated paleo-[CO 2 ] atm values vary greatly [9]. To understand the paleoclimatic system, it is important to reduce uncertainties in the relationships between paleo-[CO 2 ] atm and past climate [12].
Stomatal parameters (SI (stomatal index) and SD (stomatal density)) are reliable proxies to estimate paleo-[CO 2 ] atm . In particular, SI can provide a robust indicator of terrestrial paleo-[CO 2 ] atm as it is independent of other environmental parameters, such as soil moisture supply, atmospheric humidity and temperature [13]. Many studies have already used the SI of different taxa to estimate paleo-[CO 2 ] atm , such as Metasequoia Miki ex Hu et Cheng [12,14], Ginkgo Linn. [15,16], Quercus Linn. [17,18], Laurus Linn. [17,19,20], Platanus Linn. [17,21], and Typha Linn. [22]. As the relationship between the SI and paleo-[CO 2 ] atm is species-specific even within a single family [23] and the response sensitivities to CO 2 change are different in various taxa [24], it is necessary to select a single modern taxon that has survived for an extended period to reconstruct atmospheric CO 2 over a long geological time.
Metasequoia has exhibited an evolutionary stasis since its appearance in the Late Cretaceous [25], and fossilized Metasequoia can be considered to be conspecific with modern Metasequoia based on the morphology, biochemistry and inferred physiology [26]. Therefore, the paleo-[CO 2 ] atm changes over a long geological time can be determined from a correlation between the SI of Metasequoia needles and the paleo-[CO 2 ] atm concentration [14].
In this study, we use Metasequoia needles from seven localities in China and Japan to reconstruct continuous terrestrial paleo-[CO 2 ] atm changes from the middle Miocene to Pleistocene. Based on the reconstructed paleo-[CO 2 ] atm curve, we discuss the interaction between paleo-[CO 2 ] atm evolution and global environment change since the middle Miocene.

Materials
The fossilized needles of Metasequoia (Fig 1) were collected from one locality in SW China (Sanzhangtian) and six localities in central Japan (Kumagaya, Sennan, Hachioji, Konan, Tokamachi, and Ikoma sites) (Fig 2, Table 1). We confirm that our field study did not involve endangered or protected species and none of the localities which provided samples for this study are in protected areas. The Sanzhangtian locality belongs to the National land of the People's Republic of China, and the Land and Resources Bureau of Zhenyuan County gave permission to collect fossils from this locality. The Japanese sites: Kumagaya, Hachioji, Konan, and Tokamachi are on valley floors which are public space, so no permission was required to conduct sampling. The Sennan and Ikoma sites belong to private owners, who gave permission for sampling.
Metasequoia fossils had previously been reported from all the fossil localities. Their ages were estimated based on stratigraphic studies (Sanzhangtian site), zircon fission-track methods (Kumagaya site), and regional stratigraphic correlation using magnetostratigraphy and calcareous nanoplankton stratigraphy (Sennan, Hachioji, Konan, Tokamachi, and Ikoma sites) ( Table 1). For the samples, two were from the Miocene (Sanzhangtian and Kumagaya), three from the upper Pliocene (Sennan, Hachioji, and Konan), and two from the lower Pleistocene (Tokamachi and Ikoma) ( Table 1). At least six different needles from different branchlets were used in the studies from each site, and the exact amount depends on the total amount of materials at each fossil site ( Table 2).
Voucher specimens from the Sanzhangtian, Kumagaya, Hachioji, Konan, and Tokamachi sites are housed in the Herbarium of Kunming Institute of Botany (KUN), Chinese Academy of Science. Specimens from the Sennan and Ikoma sites are housed in the Graduate School of Horticulture, Chiba University, Japan.

Pretreatment of the fossilized needles
To remove the inorganic compounds adhering to the fossilized needles, the material was first immersed in 10%-25% Hydrochloric Acid (HCl) for two hours, then in 40% Hydrofluoric Acid (HF) for 12 hours, and in 10%-25% HCl for at least one hour. The needles were then rinsed with distilled water and divided into three parts, and the central piece (when available) used to obtain the cuticle.

Cleaned cuticular membrane maceration
For the material from the Sanzhangtian, Kumagaya, and Konan sites, we followed the methods of Kerp [41] to isolate the lower cuticle of the fossilized needles. (1) The specimens were first macerated with 70% Nitric acid (HNO 3 ) for between a few minutes to an hour until they turned yellowish-brown. (2) Once it had been rinsed with distilled water several times, (3) the upper and the lower epidermis were separated using a needle. (4) Then, the epidermis was

Cuticle observation and photography
The separated cuticles of the material from the Sanzhangtian, Kumagaya, and Konan sites were observed using a transmitted light microscope (Zeiss Axio Imager A2) and photographed with a digital camera (Zeiss AxioCam MRc). For the materials from the Sennan, Hachioji, Tokamachi, and Ikoma sites, pretreated fossilized leaves were mounted with water on slides and the lower sides of the needles were directly scanned by a confocal laser scanning microscope (Zeiss LSM710, Imager. Z2, Ar Lasser 488nm). Each field-of-view was larger than 0.03mm 2 [42]. Photoshop (version CS6, Adobe Systems; Mountain View, CA) was used to merge 6-12 serial images that were taken of the same area but at different focal levels. The SI data were used to estimate the paleo-[CO 2 ] atm from the middle Miocene to Pleistocene by using the species-specific, nonlinear negative correlation between atmospheric CO 2 partial pressure and SI (Eq 2) based on Royer et al. [14].

Measurement of SI and paleo-[CO 2 ] atm concentration
Paleo À ½CO 2 atm ¼ SI À 6:672 0:003883 Â SI À 0:02897 Equation 2 The significant differences between the mean variance of the SI from different ages were statistically tested using the two tailed one-way ANOVA with the "LSD" option in IBM SPSS Statistics (Version 20.0).

Results
Fossilized Metasequoia needles from the early late Miocene Kumagaya site had the lowest SI value (SI = 9.80 ± 0.65) and those from the middle Miocene Sanzhangtian site had the second  Table 3, for more details see S1 Table). The SI of the Pliocene and Pleistocene samples were higher (SI mean = 15.2-17.9) than the SI of the Miocene samples. The SI of the samples from the middle Early Pleistocene Ikoma site had the highest SI value (SI = 17.9 ± 1.9) and give out the lowest CO 2 Table 3, for more details see S1 Table). The significant differences between the mean variance of the stomatal index from different fossil localities were statistically tested (F = 54.016, p<0.001) by one-way ANOVA with the "LSD" option in SPSS Statistics (Version 20.0). The result showed there was no significant difference between the SI data from Sanzhangtian locality (middle Miocene) and Kumagaya locality (late Miocene), but the SI data of these two localities were significantly different from the SI data from late Pliocene and Pleistocene localities. SI data of Konan locality (Late Pliocene) was significantly different from all other localities, but no significant difference has been detected among Sennan (Late Pliocene), Hachioji (Late Pliocene), Tokamachi (middle Early Pleistocene) and Ikoma localities (latest Early Pleistocene) ( Table 4). In general, the values of the middle Miocene [CO 2 ] atm estimated from marine proxies are lower than those from terrestrial records. Boron/Calcium (B/Ca) ratios of surface-dwelling foraminifera give a paleo-[CO 2 ] atm of ca. 420 ppmv during the MCO that declined gradually to ca. 200 ppmv in the earliest late Miocene [47]. B isotope (δ 11 B)-based paleo-[CO 2 ] atm from ODP761 changed from ca. 400 ppmv in the MCO to ca. 280 ppmv in the late middle Miocene [48]. A stable paleo-[CO 2 ] atm curve with slight changes around 210 ppmv from the MCO to late Miocene was drawn based on phytoplankton δ 13 C alkene analysis [49,50]. The paleo-  [51]. This was related to climatic cooling in the later late Miocene [18,52]. When using B/Ca [53] and phytoplankton [49,50] from marine proxies, they showed fluctuating values that were mostly less than 300 ppmv (Fig 4B). The estimated paleo-[CO 2 ] atm values for 10-9 Ma (351 ppmv) from this work are almost the same as the value from ca. 8.5 Ma from Quercus leaves [18]. Our data showed little change between the middle Miocene (334 ppmv) and the early late Miocene (351 ppmv) that confirmed the stable paleo-[CO 2 ] atm condition during this time as indicated by the phytoplankton record [49,50].

Late Pliocene to Pleistocene paleo-[CO 2 ] atm change
In most of the previous research, paleo-[CO 2 ] atm values are distributed between 200 and 400 ppmv during the Pliocene to Pleistocene (Fig 5A). B/Ca and B data have been used to determine the paleo-[CO 2 ] atm of this period, as there is a lack of data from the stomatal method. The paleo-[CO 2 ] atm curve based on B/Ca from surface-dwelling foraminifera exhibited a peak of ca. 300 ppmv at ca. 3.4 Ma, this decreased to 181 ppmv at ca. 2.9 Ma, and then increased to 332 ppmv at ca. 1.4 Ma [47]. The downward shift in its fluctuation range was observed in the Early Pleistocene (Fig 5A), and the lowest value of 188 ppmv was recorded in the last glacial maximum (0.02Ma) [47]. The paleo-[CO 2 ] atm recorded in the B isotopes indicates a higher level than that in B/Ca record during the Late Pliocene, that is, ca. 340 ppmv at ca. 3.4 Ma and ca. 400 ppmv at ca. 3.0 Ma [54]. However, it decreases to the same level (ca.  Table) along with our data. Vertical error bars: standard deviation of paleo-[CO 2 ] atm values, and horizontal error bars: standard deviation of ages of materials. The current atmospheric CO 2 concentration (390 ppmv) is indicated by the horizontal dashed line. (b) SST records for the last 3.5 Ma from southern South China Sea [56]. (c) Global oxygen isotopes of benthic foraminifera shells [57]. The vertical color bands in (a), (b), and (c) indicate the periods considered by this research, and same period is marked by the same color.  2 ] atm based on the SI of Typha at the Plio-Pleistocene boundary (2.65 Ma) exhibits a much higher value (534 ppmv) than the other results [22].
Our data showed that the paleo-[CO 2 ] atm was maintained in the range between 280 and 285 ppmv in the Pliocene and Pleistocene (Figs 4B and 5A), which is about 150 ppmv lower than the results estimated from B isotopes [54], and about 70 ppmv higher than the results estimated from the B/Ca proxy [47]. Our data are consistent with the results estimated from Quercus stomata [18], but are much lower than the data estimated from Typha from sediment at the Plio-Pleistocene boundary [22]. While different proxies [22,47,54] have recorded fluctuations accompanying climate changes (Fig 5B), the paleo-[CO 2 ] atm value of this study stabilized at around 280 ppmv. Seiki et al. concluded that the Pliocene CO 2 levels determined by numerous methods agreed well with each other [9,54]. The present research suggests that some disagreements still remain in the results between our stomatal data and B, B/Ca records in the Pliocene, while the Pleistocene proxies give more consistent CO 2 levels (than the Pliocene).

Paleo-[CO 2 ] atm change and late Cenozoic climatic deterioration
The overall climate cooling reconstructed for the past 20 Ma has generally been attributed to changes in CO 2 concentration in the atmosphere [46,58]. According to the marine oxygen isotope record, global temperature peaked at around 16 Ma (middle MCO) (Fig 4A), and the later part of the middle Miocene is characterized by climate cooling with expansion of the East Arctic ice sheet [59,60]. However, the middle Miocene paleo-[CO 2 ] atm reconstructed in this study (around 334 ppmv) was just slightly lower than the present level, which was also the level maintained during the late Miocene (around 354 ppmv). That means that before the global temperature decrease, paleo-[CO 2 ] atm had already achieved a stable low level. The Miocene paleo-[CO 2 ] atm estimated based on alkenones also showed that paleo-[CO 2 ] atm was similar during middle Miocene and late Miocene [61]. The δ 13 C record from foraminifera and B/Ca ratios in the foraminifera suggest that paleo-[CO 2 ] atm decreases were apparently synchronous with major episodes of glacial expansion during the middle Miocene [53,62,63], but this synchronization was not observed in our data. This study supports the view that Miocene climate change was not only influenced by paleo-[CO 2 ] atm changes, but also by increases in seasonality and ocean circulation changes [50,64,65], and these accelerated the cooling in the late middle Miocene that also acted to decrease the paleo-[CO 2 ] atm [62]. Also, climate sensitivity to paleo-[CO 2 ] atm may have been greater than previously thought [66]. The impact of high latitude vegetation on Earth's albedo may have also played an important role in the Earth's energy budget in the Miocene [67].
After termination of the mid-Pliocene warmth at ca. 2.9 Ma, cooling trends continued until the onset of major expansion of the Northern Hemisphere ice sheet at ca. 2.7 Ma, which culminated at ca. 2.5 Ma in the earliest Pleistocene [68][69][70]. However, present results show that the lower paleo-[CO 2 ] atm level started around 2.8-3.0 Ma and lasted until the late Early Pleistocene. Therefore, we consider that the transition to the icehouse world was possibly induced by a decrease of the paleo-[CO 2 ] atm , which already dropped to their lowest levels during the complete Cenozoic before the major expansion of the Northern Hemisphere ice sheets. During the Pliocene to Pleistocene, our data are very stable, but the global temperature estimated from the marine oxygen isotope record [56,57] shows drastic fluctuations (Fig 5). However, our middle and late Miocene data are significantly higher than our Pliocene and Pleistocene data. The oxygen isotope record confirms that the temperature in the Pliocene and Pleistocene was much lower than that of the middle and late Miocene [44,46]. Therefore, we can conclude that the decrease of paleo-[CO 2 ] atm level is coupled with temperature decrease during middle Miocene to Pleistocene.

Conclusions
We used the stomatal index of Metasequoia Miki ex Hu et Cheng as a proxy to reconstruct the paleo-[CO 2 ] atm evolution from the middle Miocene to late Early Pleistocene for the first time.
Our results indicate that: (1) From middle to late Miocene the atmospheric CO 2 level stabilized around 350 ppmv which is slightly lower than today. (2) The CO 2 level during the Pliocene to Pleistocene was similar to the pre-industrial level and no fluctuation can be detected by this research. (3) The Pleistocene CO 2 level estimated by different proxies agree well with each other. (4) From middle Miocene to Pleistocene, when the global temperature decreased sharply, the global CO 2 level decreased by more than 50 ppmv, which may suggest that CO 2 decrease and temperature decrease are coupled.
Supporting Information S1