Fossil seed fern Lepidopteris ottonis from Sweden records increasing CO 2 concentration during the end-Triassic extinction event

The end-Triassic event (ETE), a short global interval occurring at the end of the Triassic Period (~201.5 Ma), was characterized by climate change, environmental upheaval, as well as widespread extinctions in both the marine and terrestrial realms. It was associated with extensive perturbations of the carbon cycle, principally caused by the volcanic emplacement of the Central Atlantic Magmatic Province in relation to the break-up of Pangea. The correlated change in atmospheric CO 2 concentrations ( p CO 2 ) can be reconstructed with the stomatal proxy, which utilizes the inverse relationship between stomatal densities of plant leaves (here stomatal index (SI), which is the percentage of stomata relative to epidermal cells) and p CO 2 . Fossilized Lepidopteris leaves are common and widespread in Triassic strata, thus offering great potential for high-resolution p CO 2 reconstructions. A dataset of leaf cuticle specimens belonging to the seed fern species Lepidopteris ottonis from sedimentary successions in Skåne (Scania), southern Sweden, provided the possibility of p CO 2 reconstruction at the onset of the ETE. Here, we tested the intra-and interleaf variability of L. ottonis SI, and estimated the p CO 2 during the onset of the ETE. Our findings confirm L. ottonis as a valid proxy for palaeo-p CO 2 , also when using smaller leaf fragments. Importantly, the statistical analyses showed that the SI values of abaxial and adaxial cuticles are significantly different, providing a tool to distinguish between the two sides and select cuticles for analysis. Reconstructed p CO 2 increased from ~1000 pre ETE to ~1300 ppm at the onset of the event, a significant increase of ~30% over a relatively short time period. The p CO 2 recorded here is similar to previously published estimates, and strongly supports the observed pattern of elevated p CO 2 at the onset of the ETE.

The ETE is causally linked to the Central Atlantic Magmatic Province (CAMP) and subsequent large-scale changes to the global carbon cycle.This large igneous province was the probable cause of a warming-driven ecological turnover, due to enormous release of CO 2 into the atmosphere, which is consistently reflected by organic and inorganic carbon isotope data across the ETE (e.g., Hesselbo et al., 2002;Guex et al., 2004;Akikuni et al., 2010;Whiteside et al., 2010;Bacon et al., 2011;Schaller et al., 2011;Dal Corso et al., 2014;Yager et al., 2017;Schobben et al., 2019;Fujisaki et al., 2018), and by stomatal density data (McElwain et al., 1999;Bonis et al., 2010a;Steinthorsdottir et al., 2011).However, the exact tempo of CAMP outgassing events is yet to be fully resolved (see review by Marzoli et al., 2018), bringing the need for pCO 2 reconstructions into focus.The Earth system during the latest Triassic (Rhaetian Age) experienced greenhouse climate conditions, with pCO 2 at ~1000 ppm according to stomatal proxy analysis (McElwain et al., 1999;Bonis et al., 2010b;Steinthorsdottir et al., 2011), implying pCO 2 levels ~2.5 times higher than the present level of ~417 ppm (Tans and Keeling, 2014;May 2020, NOAA data: www.esrl.noaa.gov).During the ETE, pCO 2 levels have been reported to double to ~2000-2500 ppm, before returning to pre-extinction values in the Early Jurassic (Hettangian) (Steinthorsdottir et al., 2011), with a similar trend demonstrated by stable isotopes of organic and inorganic carbon from palaeosols (Cleveland et al., 2008;Schaller et al., 2011Schaller et al., , 2012Schaller et al., , 2015)).
The aim of this study is to test the applicability of L. ottonis as a pCO 2 proxy using the stomatal ratio method.Although the pCO 2 proxy has previously been applied to L. ottonis fossil leaves from Germany (Bonis et al., 2010b), here we analyze a large dataset of leaves from the coalbearing deposits of Skåne (Scania), southernmost province of Sweden, and further develop previous studies and datasets.If Lepidopteris can unequivocally be shown to be a reliable pCO 2 proxy, it would cement its position as an important fossil for pCO 2 reconstructions across the entire Triassic in general, and during the onset of the ETE in particular, due to its abundance and wide stratigraphic range throughout the entire period and its proliferation at the ETE.Here we statistically test the utility of L. ottonis as a pCO 2 proxy and present a new pCO 2 reconstruction from the Rhaetian.

Geological setting
The Mesozoic sediments southwest of the Fennoscandian Border Zone typically range between 100 and 1000 m in thickness, lie disconformably on largely Palaeozoic sedimentary rocks and form part of the Central European Basin, which includes the Danish Basin (Norling and Bergström, 1987;Ahlberg et al., 2003a;Vajda and Wigforss-Lange, 2009).The Upper Triassic-Jurassic sedimentary successions of Skåne were deposited in various environments such as alluvial, deltaic, coastal plain or shallow shelf conditions, in several separated sedimentary basins along the Fennoscandian Border Zone (Ahlberg et al., 2003a).
The lower to middle Rhaetian Vallåkra Member is the basal member of the Höganäs Formation and consists mainly of mudstone with interbedded sandstone and conglomerate (Sivhed, 1984).It represents a transition from the Norian red beds of the Kågeröd Formation to the mid-to late Rhaetian Bjuv Member (Sivhed, 1984;Ahlberg et al., 2003a).
The succession spanning the ETE in Skåne is historically referred to as the informal "Mine beds" (Fig. 1) as both coal and clay were extracted by shaft mining and strip mining during the earlier half of the 20th century (Lundblad, 1950).It correlates to the formal Bjuv Member, which is the middle member of the Höganäs Formation (Sivhed, 1984;Lindström and Erlström, 2006).These successions include two main coal beds: the upper coal seam A and the lower coal seam B. Coal seam A constitutes the top of the Bjuv member (Lundblad, 1950;Sivhed, 1984).This coal seam is in some places succeeded by non-marine carbonaceous mudstones ("roof layer" sensu Lindström and Erlström, 2006) and coarse poorly sorted sandstones ("Boserup beds" sensu Troedsson, 1951).These represent the T-J palynofloral transition beds (Lindström and Erlström, 2006) and the basal Helsingborg Member which is otherwise characterized by minor coal seams interbedded in sand-and mudstone, followed by mainly marine sand-and mudstones, indicating a marine transgression during the Latest Triassic to Early Jurassic time interval (Troedsson, 1951;Sivhed, 1984).

Stratigraphic range of the seed fern Lepidopteris
During the Mesozoic, seed ferns were an important element of the terrestrial ecosystems and one of the more common elements was the genus of leaf foliage Lepidopteris belonging to the Peltaspermales (Taylor et al., 2009;Kustatscher and van Konijnenburg-van Cittert, 2013) which, so far, comprises two species in the Northern Hemisphere: L. stuttgardiensis and L. ottonis (Schimper, 1869).Lepidopteris is abundantly documented from both hemispheres, but is lacking from the Lower Triassic European fossil record (Kustatscher and van Konijnenburg-van Cittert, 2013).However, Lepidopteris has been identified from lowermost Triassic successions in Gondwana, as one of the first seed plant groups to establish itself after the end-Permian mass extinction (Retallack, 1980(Retallack, , 2002;;McLoughlin et al., 1997;Mays et al., 2020;Vajda et al., 2020).During the Late Triassic, it became one of the most important seed ferns globally before it widely went extinct presumably at the TJB (Lundblad, 1950;Kustatscher and van Konijnenburgvan Cittert, 2013).Although this range might be considerably extended for Gondwana as a relictual occurrence of Lepidopteris was documented from the Early Jurassic in Patagonia, Argentina (Elgorriaga et al., 2019).Hence, Lepidopteris fossil leaves are abundantly present in Triassic deposits in both hemispheres, making Lepidopteris an ideal candidate as a global palaeo-pCO 2 proxy for this important period of climate and environmental change.

The Lepidopteris ottonis dataset
Leaf cuticles of the species Lepidopteris ottonis (Göppert) Schimper, 1869 from the uppermost Rhaetian of the Höganäs Formation of Skåne, southern Sweden, were examined in this study.The material encompasses 98 specimens of pinna and pinnule fragments which were collected by the Swedish palaeobotanists Erdmann, Nathorst, Halle and Lundblad from the localities Bjuv (56 and Höganäs (56 • 12 ′ 06" N,  12 • 33 ′ 20 ′′ E); Fig. 1) between 1868 and 1965.The cuticles were prepared and curated by Vörding in 2004 and are now hosted at the Swedish Museum of Natural History (Naturhistoriska riksmuseet), Stockholm, Sweden.
Based on macrofossils (e.g., Kustatscher and van Konijnenburg-van Cittert, 2013), Lepidopteris ottonis has a bipinnate frond architecture, pinnules with confluent bases, and partially to entire dentate margins.The frond has intercalary pinnules along the rachis between the pinna.Characteristics that distinguish this species from L. stuttgardiensis are the "blisters" or "scales" along the rachis.Swedish specimens of L. ottonis have been described by Nathorst (1878), Antevs (1914) and Lundblad (1950) and an overview of the European occurrences of this species has been provided by Bonis et al. (2010b, and references therein).
The Skåne dataset studied here consists of fragments of different parts of the fronds, which are of exceptional preservation (Figs.2A, 3).They show differences in anatomical traits and preservation between the adaxial (upper) and abaxial (lower) cuticles.Most importantly, the leaves have stomata on both sides but with more numerous stomata on the abaxial than on the adaxial surface (Fig. 2B-C).Stomata on both surfaces look identical, with four to eight papillate subsidiary cells, which vary in size and cover the stomatal pits, obscuring the guard cells.The lower cuticle generally displays a stronger venation, which is mostly invisible on the upper cuticle.The upper cuticle is often papillate, which is seldom the case for the lower cuticle.The preservation of the cuticles is usually pristine, with clearly visible cell walls and stomata.Venation, papillation and cell size can differ markedly from specimen to specimen, and therefore distinguishing between upper and lower cuticles could sometimes be difficult, especially when studying smaller fragments.

Sample treatment
Standard palaeontological procedures for extracting cuticle from fossil leaves were applied (according to Vörding, 2008, andVörding andKerp, 2008): Lepidopteris ottonis leaves were recovered from the sediments, macerated through oxidation with Schulze's reagent consisting of 30% HNO 3 and a few KClO 3 crystals and then treated with 10% potassium hydroxide (KOH).Lastly, the upper (adaxial) and lower (abaxial) cuticles were separated, permanently immersed in glycerine jelly and mounted on microscope slides.

The stomatal proxy for palaeo-atmospheric pCO 2 reconstruction
Stomata are small pores on leaf surfaces, piercing the waterproof cuticle outer layer to facilitate photosynthesis.During this process, CO 2 is taken up through the stomata, whilst oxygen is expelled as a waste product; however, water in the form of water vapor is also inevitably lost when the stomata are open for gas exchange.The empirically established inverse relationship between the frequency of stomata of most woody plants and pCO 2 in the atmosphere constitutes a strong proxy for palaeoclimate analyses (Royer, 2001;Beerling and Royer, 2011).The reason for this relationship between stomatal frequency and pCO 2 , which was first reported by Woodward (1987), is the physiological response of plants, seeking to preserve water when CO 2 is abundantly available in the atmospehere around them, by decreasing the number of stomata (Lake et al., 2002;Woodward, 1987;McElwain, 1998; M. Slodownik et al.Rundgren, 2013;McElwain and Steinthorsdottir, 2017).
Most living and fossil woody plants have hypostomatous leaves (stomata present only on the abaxial leaf surface) (Muir, 2015) and the abaxial cuticle is thus solely involved in gas-exchange and utilized for pCO 2 reconstruction by default.Some plants howeversuch as L. ottonis have amphistomatous leaves, with stomata present on both the abaxial and adaxial cuticle surface (Muir, 2015).Even when leaves are amphistomatous, stomata are generally more numerous on the abaxial than the adaxial cuticle (as is the case here).The costs and benefits of amphistomaty, and the interplay between abaxial and adaxial surfaces in amphistomatous leaves, are not well understood (Muir, 2015;Richardson et al., 2017;Drake et al., 2019;Xiong and Flexas, 2020).Therefore, it is important to distinguish between abaxial and adaxial cuticles when applying the stomatal proxy.Here we developed statistical tools to rigurously select and use abaxial cuticles only.
Stomatal frequency is typically quantified as: 1) stomatal density (SD), which is the number of stomata per unit leaf area; and 2) stomatal index (SI), is the percentage of stomata relative to the number of epidermal cells + stomata per unit leaf area.Epidermal cell size, and thus SD, is influenced by various additional environmental factors aside from pCO 2 , such as irradiation levels, water availability, salinity, and soil nutrients (McElwain and Chaloner, 1995).The influence of these extraneous factors can mostly be eliminated by using SI, which is therefore considered a more reliable proxy for pCO 2 than SD.The equation for SI (%) is: SI [%] = [SD / (SD + ED)] * 100, with ED = epidermal cell density per unit leaf area (Salisbury, 1928).
In order to interpret the SI measurements for palaeo-pCO 2 reconstructions, the fossil SI data must be calibrated with those of modern plants grown under known or controlled pCO 2 levels.A prerequisite is that the SI of the ancient and modern plants respond to environmental changes in a similar way.If possible, the nearest living relative (NLR) should be used, with the assumption that these closely related plants respond to CO 2 congruently (McElwain and Chaloner, 1995).If no NLR is available (which is mostly the case for pre-Cenozoic fossils) or differ significantly in their modern adaptations, the nearest living equivalent (NLE) can be used, which should either live in a similar ecosystem and/ or have anatomical analogues to the target fossil (McElwain, 1998;McElwain and Chaloner, 1995).
Here we use the stomatal ratio (SR) approach to reconstruct pCO 2 (see 3.3.1.),which is a simple, semi-quantitative method to estimate pCO 2 (e.g., McElwain and Chaloner, 1995;McElwain, 1998;Steinthorsdottir et al., 2011;McElwain and Steinthorsdottir, 2017).However, two other methods for palaeo-pCO 2 reconstruction using the stomatal proxy have been developed to date, but could not be tested within this study: transfer functions (e.g., Kürschner et al., 2008;Barclay and Wing, 2016) and gas exchange modelling (e.g., Franks et al., 2014;Konrad et al., 2017).Since no NLR exists for Lepidopteris ottonis, consequently no appropriate transfer functions are available for this species.In addition, although gas exchange modelling might be considered ideal to reconstruct pCO 2 using the extinct L. ottonis, the necessary model inputs of guard cell width and stomatal pore length could not be measured, because these features are completely obscured by subsidiary cells.Furthermore, coeval stable carbon isotope ratios need to be included in this approach, and these were not available.However, it has repeatedly been shown that results based on these three methods usually display a remarkable inter-method consistency (Li et al., 2019;McElwain and Steinthorsdottir, 2017;Steinthorsdottir et al., 2011Steinthorsdottir et al., , 2016Steinthorsdottir et al., , 2019aSteinthorsdottir et al., , 2019bSteinthorsdottir et al., , 2020)).

The stomatal ratio method
The stomatal ratio (SR) is the quotient of the SI of the NLE and the SI of the fossil plant.The pCO 2 was obtained with the Carboniferous standardization established by McElwain and Chaloner (1995) and Chaloner and McElwain (1997), which is believed to be the most appropriate for Mesozoic SR (McElwain and Chaloner, 1995;Steinthorsdottir et al., 2011;Mays et al., 2015;Steinthorsdottir and Vajda, 2015).The SR pCO 2 calibration is expressed by the equation: Because Lepidopteris largely went extinct at the ETE, it does not have a NLR, nor an obvious NLE.This problem was solved by Bonis et al. (2010b) by cross-calibration to co-occurring fossil leaves of Ginkgo taeniatus, which was inferred to respond to pCO 2 similar to the extant Ginkgo biloba.As a result, the L. ottonis SI values were corrected by subtracting 1.95, which was the average difference of SI of the two species in the same bed (Bonis et al., 2010b).We followed this protocol and cross-calibrated the SI values of L. ottonis from Skåne using the G. taeniatus data from Bonis et al. (2010b).As an additional test, the cooccurring cycadalean Ctenis nilssonii (NLE is Zamia furfuraceae) from bed L1 ("α-bed") from Skåne (McElwain et al., 1999) was used to correct the L. ottonis SI values.The SI of C. nilssonii was 6.5 (McElwain et al., 1999) and 8.6 of L. ottonis (this study), resulting in a subtraction of 2.1 for the correction of L. ottonis.The SR was calculated with SI NLE /SI L. ottonis, cor- rected , with SI Ginkgo biloba = 11.33 (Bonis et al., 2010b) and SI Zamia furfuraceae = 10.12)(McElwain et al., 1999).

Image analysis for stomatal proxy palaeo-pCO 2 reconstruction
High magnification images of the cuticle surfaces were photographed with a Leica DFC310FX camera mounted on a Leica DM 2000 transmitted light microscope with the associated software (Leica Application Suite v.3.8).For this purpose, transmitted light microscopy was used at 200× magnification.Microscope photographs of the cuticle were taken evenly distributed across each leaf, but excluding leaf margins and veins, and avoiding overlaps.SI data was obtained using the software ImageJ (Abràmoff et al., 2004).A square of 0.09 mm 2 was digitally engraved on every image and the stomata and epidermal cells (excluding subsidiary cells) within this square were counted.To account for cells and stomata that fell only partially within the field of view, we counted the partial cells and stomata along two sides of the square field of view and disregarded them on the other two sides of the square, according to standard protocol (Poole and Kürschner, 1999).Cumulative SI means analysis showed that seven fields of view per specimen, were optimal to reach a statistically stable value.

Statistically testing L. ottonis' intra-and inter-leaf SI variation
In order to evaluate Lepidopteris ottonis as a palaeo-pCO 2 proxy and to determine pCO 2 levels of the late Rhaetian of Skåne, the following analyses were conducted: 1.In order to test the use of cuticle fragments for determination of the average SI of the stratigraphic units, the intra-pinna SI variation was evaluated.For this purpose, the SI of the abaxial cuticle of four to six pinnules of three exceptionally large and well-preserved pinnae (pilot specimens) were compared statistically with a 'one-way analysis of variance' (one-way ANOVA), applying a 5% confidence interval (α = 0.05) using the data analysis tools of Microsoft Excel v.2019.Thereby, the null hypothesis (NH) was defined as "all pinnules of a single pinna have the same SI".Hence, a p-value higher than 0.05 supports the NH, which would suggest that the inter-pinnule variability is low enough to use isolated pinnules for valid SI estimates.On the other hand, a p-value lower than 0.05 rejects the NH and indicates a statistically significant difference of the mean SI between at least two different pinnules, which would mean that single pinnules cannot provide reliable SI estimates.
2. To test for differences in SI between the adaxial vs abaxial cuticles, a paired sample t-test ('paired two sample for means' on Microsoft Excel v.2019) was conducted with a 5% confidence interval on the pilot specimens.The NH is "both adaxial and abaxial cuticles of one leaf have the same mean SI", i.e., that the upper and lower leaf surface have the same mean SI.A p-value higher than 0.05 would confirm the NH, indicating that both of the cuticle SI values are statistically similar, and thus the study of either cuticle surface would provide a valid SI.In contrast, a p-value lower than 0.05 would reject the NH indicating that abaxial and adaxial cuticles are significantly different, and the leaf parts should therefore be counted and grouped separately without mixing the data.
3. For investigations of the pCO 2 changes within the Skåne succession, the SI means of 6-9 abaxial samples (depending on the available specimens) of each stratigraphic bed (L1− 3) were analysed.The variation between the SI of specimens within each bed was tested with the one-way ANOVA as outlined above.Thereby, the NH was "all samples from the same bed have the same SI", presuming a relatively stable pCO 2 during the deposition.Hence, a rejection of the NH, indicating significant variability of SI, would suggest that the morphology of Lepidopteris cuticles are not only dependent on pCO 2 but also other factors, which could obscure pCO 2 -estimates.Cumulative mean analysis was also applied to the SIs of each bed to see how many specimens were necessary to reach a stable value.

SI variation of Lepidopteris ottonis
Difference of the SI values between pinnules of a single pinna of the pilot specimens was shown to be not significant (Fig. 3): the p-values of the ANOVA were much higher than 0.05 (p leaf A = 0.75; p leaf B = 0.36; p leaf C = 0.40), supporting the NH "all pinnules of a single pinna have the same SI".Furthermore, the standard deviations between the mean-SIs of the pinnules of a leaf were small (σ leaf A = 0.3, σ leaf B = 0.27, σ leaf C = 0.21).Consequently, the ANOVA test and standard deviation showed small inter-pinnule variability, and thus smaller leaf fragments or single pinnules can be used for reliable SI analysis.
The SI analysis of the adaxial cuticles revealed values about two to three times lower than the respective abaxial cuticles (Fig. 4).The t-test confirmed the difference of the SI between the abaxial and adaxial cuticle by rejecting the NH with a p-value of 0.014.Hence, only the abaxial cuticle must be used for reconstructing pCO 2 through geological time, and combined abaxial and adaxial cuticle counts should be avoided.These findings provide a statistical method to differentiate between abaxial and adaxial cuticle fragments, when this is not possible by observing morphological traits.
ANOVA tests of the variation of SI between leaves within a bed showed a generally high variability in three of four beds.The test resulted in a p-value of about 0.177 for L1 and hence the Null-Hypothesis was supported.In contrast, the NH was rejected for L2 (p ≈ 0.016), L3a (p ≈ 0.001) and L3b (p ≈ 5.748*10 − 5 ).Nevertheless, despite the variability recorded, the data showed a clear development towards lower SI from the stratigraphically older to younger beds (Table 1, Fig. 6).In particular, the SI-mean of the oldest bed L1 was 8.60%, whereas the stratigraphically younger beds showed lower SImeans of 7.58%, 7.65% and 7.22% (L2, L3a and L3b, respectively).The calculation of cumulative means (Fig. 5) showed, that at least five specimens were needed to reach the flat curve and hence a stable SI value.A flat curve for the SI values is clearly visible for L1 and L3b, whereas it remained ambiguous for L2 and L3a because only six Lepidopteris specimens were available from each bed.

Atmospheric CO 2 estimates for the latest Rhaetian
Calibration of the palaeo-pCO 2 levels were estimated with the semiquantitative stomatal ratio method (McElwain and Chaloner, 1995) using the equation of Bonis et al. (2010b) to correct L. ottonis' SI values, as described in the methods (Table 1, Fig. 6).For the leaves derived from L1, the lowermost of the observed beds, pCO 2 was estimated ~1022 ppm, using the Carboniferous standardization (Carb) and the crosscalibration with Ginkgoales.pCO 2 increased significantly based on the SI of the leaves from the overlying beds (L2 and L3), where the values remained relatively stable: between 1193 and 1295 ppm (Carb).The cross-calibration of Lepidopteris with Cycadales resulted in 88 to 104 ppm lower pCO 2 estimates from all beds, but showed the same trend of increasing pCO 2 with time (Table 1).

Discussion
The stomatal proxy-based Lepidopteris ottonis pCO 2 record presented here supports the findings of Bonis et al. (2010b), thus identifying Lepidopteris as a useful taxon for palaeo-pCO 2 reconstruction.The present study furthermore strengthens the status of Lepidopteris as a pCO 2 proxy by providing statistical tools to overcome potential weaknesses in stomatal analyses, such as distinguishing between abaxial and adaxial cuticles.Like the results of Bonis et al. (2010b) and Vörding (2008), our analysis showed only minor SI variations between the pinnules of a pinna.Additionally, we could statistically confirm our findings with the one-way ANOVA.This supports the validity of using cuticle fragments, since SI values do not change significantly within a pinnule, opening up the possibility of using even small cuticle fragments macerated from bulk samples, thus greatly increasing the amount of available Lepidopteris cuticle material for analysis.Our results confirmed the necessity of distinguishing the adaxial and abaxial cuticle for the SR method, because utilizing a mix of the cuticle surfaces would overestimate palaeo-pCO 2 (Vörding, 2008).
The one-way ANOVA revealed a relatively wide distribution of L. ottonis SIs within three of the four horizons (L2, L3a, L3b), that may have been caused by additional environmental factors apart from pCO 2, or may reflect natural variability, as suggested by Bonis et al. (2010b).Therefore, we recommend using the cumulative mean approach to ensure that a stable SI is achieved for subsequent Lepidopteris pCO 2 reconstructions, with at least five specimens per bed and seven area counts per specimen.
The pCO 2 estimations provided here were produced with a single, semi-quantitative stomatal proxy method, and should be taken with caution.Ideally, future Triassic and ETE Lepidopteris pCO 2 reconstructions should test these results using additional methods and larger datasets.We note that the SI correction with fossil Ginkgoites taeniatus (as per Bonis et al., 2010b) and the new correction using Ctenis nilssonii resulted in highly comparable pCO 2 estimates, but there remains a lack of evidence of the assumed parallel response to changing pCO 2 of these fossil plant species.To test this assumption, additional fossils of G. taeniatus and/or C. nilssonii in co-occurrence with L. ottonis from at least two or more beds would be required.Regardless, there is a clear trend of significantly increasing pCO 2 over the stratigraphic interval of ~14 m between the oldest and the youngest beds in the studied sedimentary succession.
Reconstructed pCO 2 using the G. taeniatus correction resulted in ~1022 ppm in the late middle Rhaetian and a maximum of ~1290 ppm just below the TJB, recording an increase of ~268 ppm.Previously published studies using the SR method reported similar pCO 2 values Table 1 pCO 2 reconstruction with the stomatal ratio method using the SI of Lepidopteris ottonis corrected to Ginkgoales (Bonis et al., 2010b) andCycadales (McElwain et al., 1999)  An additional study (Steinthorsdottir et al., 2011) used a large, highresolution database of Ginkgoales, Bennettitales and Coniferales from Greenland and Ireland and showed a pre-boundary estimate of ~1000 ppm, and steeply increasing values to a maximum of ~2200 ppm at both sections near the TJB (also using the Carboniferous standardization of the SR method).A similar trend of elevated pCO 2 has been reported by stable carbon isotope analyses of pedogenic carbonate and organic matter, but indicating an even greater rise (Cleveland et al., 2008;Schaller et al., 2011Schaller et al., , 2012Schaller et al., , 2015)).High resolution studies estimated a sharp increase from ~1000 ppm just before the ETE to ~4000 ppm during the ETE (Schaller et al., 2011(Schaller et al., , 2015)).Hence, our rising pCO 2 estimates generally correspond with other mid to late Rhaetian pCO 2 reconstructions, particularly with those derived from stomatal proxies (see Fig. 7).Our reconstructed pCO 2 probably corresponds with the early stages of the rapid rise in pCO 2 during the ETE but preceding the TJB peak concentrations.

Conclusions
Despite being a species of an extinct plant order, the results herein support Lepidopteris ottonis as a useful proxy for palaeo-pCO 2 reconstructions.We recorded semi-quantitative pCO 2 estimates based on fossil leaves and leaf fragments from end-Triassic extinction (ETE) successions in Skåne, Sweden, using stomatal indices (SI) in the stomatal ratio method with a cross-calibration to a fossil with living relatives.The significant decrease in SI reflects an increase in pCO 2 of ~300 ppm, from ~1000 in the late mid Rhaetian to ~1300 ppm at the onset of the ETE and supports the trends of previously published reconstructions.We detected low intra-pinnule variation in SI allowing the use of small fragments, but significant differences in SI between the abaxial and adaxial leaf cuticles.Furthermore, the variation within each stratigraphic bed suggests the necessity of large sample populations.Due to its abundance as well as extensive chronologic, stratigraphic and geographic ranges, Lepidopteris has demonstrable future potential as an important fossil for high-resolution Triassic pCO 2 reconstructions.

Fig. 3 .
Fig. 3. Interpinnule variation of the stomatal index (SI) of three pilot specimens.Leaf A: S064195-02; Leaf B: SO64168-06; Leaf C: SO64088-01.The squares represent the SI means of each pinnule and the error bars the standard deviation s.Note, that these leaves were not included in the CO 2 reconstructions because they originate from a unit with an uncertain stratigraphic correlation.Scale = 5 mm.

Fig. 4 .
Fig. 4. Comparison of the stomatal index (SI) means between the abaxial and adaxial cuticle of Leaf A: S064195-02, Leaf B: SO64168-06 and Leaf C: SO64088-01.The whiskers represent minimum and maximum SI measured for each leaf.

Fig. 5 .
Fig. 5. Cumulative mean SI of Lepidopteris ottonis specimens from each observed horizon of the Skåne strata.

Fig. 6 .
Fig. 6.Left: Composite stratigraphic log of the Upper Triassic to lowermost Jurassic of Skåne.N. = Norian and K.F.= Kågeröd Formation.Centre: The boxplots show the distributions of the SI of the Lepidopteris ottonis specimens from the beds L1, L2, and combined L3a and L3b; vertical red lines represent the SI means for all specimens within an individual bed.Right: pCO 2 reconstruction in the "Mine strata" from Skåne based on the cross-calibrated SI values; L3a + b is calculated from the average combined SI data of L3a and L3b.
*This is calculated as the average between L3a and L3b.