In Anatolia, eight pedogenic carbonate sites were sampled within the Central Anatolian Volcanic Province (CAVP) and two sites each adjacent to the Tuz Gölü and Ecemiş fault zones, as well as the Ermenek basin. Two additional sites were sampled within the Adana basin near the Mediterranean Sea (Fig. 1b, Table 1; see Table S1 and Text S1 for detailed site information). Site elevations from this study range between ca. 110 and 1650 m a.s.l. and the sampled soil carbonates range in age from ca. 10 Ma to modern (Fig. 1b; Table S1). Modern pedogenic carbonate samples consist of friable to weakly consolidated calcareous nodules typically 1 to 3 cm in diameter, from the B horizon of well-drained soils with a pink to red hackly appearance. Miocene to Pleistocene carbonate subsoil horizons developed within floodplain deposits. They comprise moderately consolidated to hard calcareous nodules ranging from 3 to 8 cm in diameter, or representative domains of dm-scale nodular carbonate layers. Whenever possible, pedogenic carbonate nodules and in some cases casts of taproots and fibrous roots, were collected from Bk horizons of cream to red-colored subsurface soils. Sampling choices were based on the availability of pedogenic carbonates in outcrops, which were taken over several months of fieldwork. Where identifiable, pedogenic carbonate was sampled at a minimum soil depth of 30 cm to reduce the influence of atmospheric CO₂ (Cerling and Quade, 1993). Representative pictures of sampled outcrops and levels are shown in Fig. S1.

Powdered material from pedogenic carbonate samples was extracted with a diamond-tip dental drill. Only micritic parts devoid of clasts were drilled for analysis. A total of 447 δ13C values were obtained for pedogenic carbonate samples from sixteen sites using a Thermo V mass spectrometer at the Institute of Geology (University of Hanover, Hanover, Germany) and a Thermo MAT 253 mass spectrometer at the joint Goethe University-Senckenberg BiK-F Stable Isotope Facility (Frankfurt, Germany). Data were collected during multiple sessions in both labs. δ18O and δ13C values were evaluated simultaneously and are reported in Table S1. The δ18O values that were obtained during the measurements were published in Meijers et al. (2025) and interpreted with respect to the surface uplift history of Anatolia. To assess intra- and inter-nodule variability, multiple measurements were performed on select single pedogenic carbonate nodules which were cross-sectioned from three sites: 10TG55 (n = 16, five nodules, three to four analyses per nodule), 10FD (n = 48, 25 nodules, of which three nodules with six to 13 analyses per nodule), and 12C047–052 (n = 24, 16 nodules, of which two nodules with five measurements each; Table S1). Powdered carbonate samples were digested in 100 % H₃PO₄ and analyzed as CO₂ in continuous flow mode using above mentioned mass spectrometers interfaced with a Thermo GasBench II. Analytical procedures followed those of Spötl and Vennemann (2003), which includes corrections for scale, linearity, and drift. Raw isotopic ratios were calibrated against an in-house standard (Carrara marble; δ13C = 2.01 ‰ (V-PDB), δ18O = -1.74 ‰ (V-SMOW)), a synthetic Merck standard (δ13C = -35.69 ‰ (V-PDB), δ18O = 12.38 ‰ (V-SMOW)) and international carbonate reference material (NBS18). The in-house Carrara marble standard was weighed in for four different sample sizes for the linearity correction. Final isotopic ratios are reported against Vienna Pee Dee Belemnite (V-PDB), with analytical uncertainties that are typically better than 0.07 ‰. In the text, we refer to average δ13C values with 1σ standard deviations; averages, 1σ, and 2σ standard deviations, as well as medians, and 25th and 75th percentiles are displayed in Tables 1 and S1. Pedogenic carbonate that formed in equilibrium with soil-respired CO₂ is enriched by ca. 14 ‰–17 ‰ in δ13C compared to CO₂ derived from root respiration and microbial decomposition of organic matter (e.g., Tipple and Pagani, 2007).

Age constraints for the paleosols containing the analyzed pedogenic carbonates from the CAP and the Tuz Gölü area are based on radiometrically dated ignimbrite intercalations of the CAVP (Aydar et al., 2012; Özsayin et al., 2013). The stratigraphic framework for the sampled horizons is based on the ignimbrite stratigraphy and the published geochronological data (see Text S1 and Table S1). A composite section of the area of the southern CAVP where six sites were sampled shows the stratigraphic relationship between the sampled intervals (Fig. S1). The Quaternary age for site 11AD is based on the 1 : 500 000 geological map of Adana (Ulu, 2002). Modern soil carbonate horizons were identified based on field relationships and assigned a Holocene age (5 ± 5 ka).

A reconstruction of paleobotany-derived climatic parameters of Anatolia (Fig. 2b, c) is based on the coexistence approach (Mosbrugger and Utescher, 1997). All data are listed in Table S2.

The compilation of 12 Ma to Holocene soil carbonate δ13C values from Türkiye and Greece (Fig. 2a; Table S3) includes data from Böhme et al. (2017), Lepetit (2010), and Quade et al. (1994). The compilation of 12 Ma to Holocene soil carbonate δ13C values from Asia and Africa (Fig. 3d, e) was retrieved from Fox et al. (2018). Datasets lacking age constraints were excluded. We also excluded elevated δ13C values from one multiproxy study on the Qaidam basin, which are attributed to low soil respiration rates in response to increased aridification (Zhuang et al., 2011). As such, the δ13C values may not be only interpreted in terms of variations in the dominating photosynthetic pathways (see also Sect. 4.1). Studies included in the Africa compilation (Fig. 3e): Aronson et al. (2008), Bestland and Krull (1999), Cerling and Hay (1986), Cerling et al. (1988, 1991, 2003, 2011), Kingston (1992), Levin et al. (2004, 2011), Plummer et al. (1999, 2009), Quade et al. (2004), Quinn et al. (2007), Sahnouni et al. (2011), Sikes (1994), Sikes et al. (1999), Sikes and Ashley (2007), WoldeGabriel et al. (2009), Wynn (2000, 2004), Wynn et al. (2006). Studies included in the Asia compilation (Fig. 3d): An et al, (2005), Behrensmeyer et al. (2007), Ding and Yang (2000), Ghosh et al. (2004), Kaakinen et al. (2006), Passey et al. (2009), Quade et al. (1994), Quade and Cerling (1995), Sanyal et al. (2004), Yao et al. (2010).

The 12 Ma to Holocene δ13C record from Anatolia shows large changes between the four time intervals (see Sect. 3 Results) and covers a wide range from -9.5 ‰ to 3.4 ‰ (Fig. 2). This variation in δ13C values (up to 12.9 ‰) may reflect temporal changes in soil respiration rates, plant water stress, and/or the dominating photosynthetic pathway (C₃ vs. C₄) of biomass in Anatolia. We interpret obtained δ13C values (and associated δ18O values; Meijers et al., 2025) as primary (i.e. reflecting the isotopic ratios upon the formation of the carbonates), because the values are within a range typical of soil carbonates (e.g., Cerling and Quade, 1993). Additionally, the paleoaltimetry study by Meijers et al., (2025) that is based on the δ18O values associated with the δ13C values of this study includes two dual clumped isotope (Δ47, Δ48) temperatures. First, results for the ca. 5.4 Ma pedogenic carbonate sample (from CAVP site 10CKK, see also this study) and lake carbonate sample (from the Adana basin in southern Turkey) plot within error of the dual clumped isotope equilibrium line, which implies that their isotopic compositions are devoid of significant kinetic biases. Second, the obtained Δ47 temperatures of 15.6 ± 3.3 °C (CAVP) and 24.1 ± 3.5 °C (Adana basin) show that the carbonates precipitated under near-surface conditions rather than at elevated diagenetic temperatures. More evidence for the primary origin of the sampled soil carbonates comes from the δ18O and δ13C values and thin sections of lake carbonate that was sampled in the same or similar fluvio-lacustrine host lithologies, which show no evidence of diagenetic alteration (Meijers et al., 2018, 2020). Moreover, all sites within the Central Anatolian Plateau interior were buried only 170 m or less (Table S1), except for Mustafapaša (ca. 9.9 ± 0.8 Ma; burial depth ca. 400 m). Low burial depths result from relatively low sedimentation rates (ca. 25–80 m Myr⁻¹; Table S1) and rapid drainage integration and incision of the basins within maximal 2.5 Myr after the latest deposition of the soil carbonate host lithologies (Brocard et al., 2021; Meijers et al., 2020).

Under conditions of low soil CO₂ production, soil carbonates tend to form at shallower depths and may incorporate varying proportions of atmospheric CO₂ (Cerling, 1984; Cerling and Quade, 1993), resulting in higher δ13C values (Caves et al., 2016; Licht et al., 2020). Particularly in dry ecosystems, C₃ plants may yield increased δ13C values (Kohn, 2010). Although Anatolia is presently characterized by a climate with dry summers, all available paleobotanic datasets from the region suggest subhumid climatic conditions during the Late Miocene to Pliocene (Table S2). We therefore interpret the variations in δ13C values of our 10 Ma to recent soil carbonate record to reflect significant changes in the relative contribution of C₃ and C₄ components of vegetation.

The large range of soil carbonate δ13C values of over 10 ‰ at ca. 9.9 Ma in Anatolia is consistent with a heterogeneous vegetation cover that includes both C₃ and C₄ plants, albeit with a dominance of C₃ vegetation given an average δ13C of -5.2 ‰. Pedogenic carbonate δ13C values between 7.1 and 4.9 Ma, which average 4.0 ‰ more positive compared to δ13C values at 9.9 Ma indicate central Anatolian floodplain environments dominated by C₄ vegetation (Fig. 2a). The dominance of C₄ biomass between 7.1 and 4.9 Ma is consistent with published CAP δ13C values of pedogenic carbonate from this time interval (6.7 to 4.8 Ma; Fig. 2a; Lepetit, 2010), some of which were obtained from the same stratigraphic intervals. For the time intervals from 3.9 to 1.4 Ma and the Holocene our δ13C averages are significantly lower – by nearly 7 ‰ – than those from the 7.1 to 4.9 Ma interval, and by ca. 2.5 ‰ compared to the 9.9 Ma interval. This indicates the presence of floodplains dominated by C₃ vegetation after 3.9 Ma.

δ13C values of ca. 9.3 to 5.3 Ma fossil equid tooth enamel (Rey et al., 2013), and ca. 10 to 6.5 Ma pedogenic carbonate δ13C records (Böhme et al., 2017; Quade et al., 1994; Table S3) from the Aegean region (Greece) indicate C₃ vegetation. However, phytoliths, as well as δ13C values of published pedogenic carbonate and fossil herbivore tooth enamel, imply the presence of C₄ vegetation between ca. 9 and 7 Ma in the CAVP (Kayseri-Özer et al., 2017; Lepetit, 2010), Pikermi and Samos (Aegean, Greece; Fig. 1b; Böhme et al., 2017; Quade et al., 1994), and Maragheh (Bernor et al., 2016; Biasatti et al., 2015; Strömberg et al., 2007). On Crete and Cyprus (Greece) ca. 5 ‰ variations in δ13C values of leaf waxes (long-chain n-alkanes produced by terrestrial plants), superimposed on an overall 4 ‰ increase in δ13C values between ca. 7 and 6 Ma, indicate the expansion of C₄ vegetation (Butiseacă et al., 2022; Mayser et al., 2017) in response to Messinian climate cycles. Collectively, our data in combination with published datasets indicate that the geographic extent of C₄ expansion in the Eastern Mediterranean region during the Late Miocene and earliest Pliocene may not have been restricted to Anatolia, but extended into the Aegean and the Iranian plateau. After 3.9 Ma (late Early Pliocene), δ13C values from Anatolian soil carbonates are similar to those derived from the Aegean (Fig. 1b; Quade et al., 1994) and indicate vegetation dominated by C₃ biomass, which is the observed dominant vegetation type in Anatolia and the Aegean region today (< 10 % C₄; Still et al., 2003).

Pedogenic carbonates from ca. 10 Ma to Holocene floodplain deposits in Anatolia indicate that the region has been characterized by rainfall seasonality since the Late Miocene, as their formation requires periodic soil drying (e.g., Zamanian et al., 2016). Furthermore, the dominance of C₄ vegetation between 7.1 and 4.9 Ma in Anatolia, as reconstructed from our pedogenic carbonate δ13C values, suggests that open-habitat grasslands characterized portions of the landscape. This is supported by (a) paleobotanical (macrofossils, pollen, and spores) data that suggest the introduction of steppe elements in Anatolia between 9 and 6 Ma (Denk et al., 2018), (b) the (Middle-)Late Miocene rise of open-habitat grasslands in Anatolia and nearby Greece and Iran reconstructed from phytoliths (Strömberg et al., 2007), and (c) the presence of the open-environment adapted Pikermian chronofauna in aforementioned regions (Eronen et al., 2009). However, macrofossil, pollen, and spore records from Anatolia, Greece, and Bulgaria (Denk et al., 2018) also sketch Late Miocene landscapes covered by evergreen needleleaf forests and mixed forests. As such, Denk et al. (2018) reject a cohesive savannah biome. We propose that the results from the paleobotanical and paleontological studies, as well as our stable isotope-based vegetation reconstructions, can be reconciled by heterogeneous Late Miocene Anatolian landscapes with largely interconnected forested as well as savannah-like environments (see also Fortelius et al., 2019).

The paleobotanical, paleontological, and soil carbonate records were all retrieved from ca. 11 to 4 Ma low-relief floodplains where fluvial and lacustrine deposits accumulated. Similar to today, the low-relief areas were interrupted by relict mountain ranges and local fault-controlled relief, but were not connected to marine basins. The fluvial and lacustrine records are currently accessible as a result of rapid latest Miocene to Pliocene drainage integration of the Anatolian plateau and subsequent river incision (Brocard et al., 2021; Meijers et al., 2020). As such, our soil carbonate δ13C records are biased towards partially preserved and incised intermontane floodplains and underrepresent vegetation dynamics of (potentially) forested and eroding topographic highs. Simultaneously, wind-blown pollen from forested montane areas are partially preserved in the fluvial and lacustrine records. Our δ13C soil carbonate record therefore highlights the importance of multi-proxy ecosystem reconstructions and solidifies evidence for a highly variable vegetation cover in Anatolia during the Late Miocene.

Located at the crossroads of Africa, Asia, and Europe, the soil carbonate δ13C record of Anatolia bears the potential to identify paleoenvironmental dynamics specific to the spread of C₄ vegetation in the Old World as well as to changing paleoclimatic conditions in the Mediterranean region. We compare the Anatolian soil carbonate δ13C record with available leaf wax, fossil tooth enamel, and soil carbonate δ13C records that constrain the initial expansion of C₄ vegetation (Polissar et al., 2019) and its rise to dominance in the Old World grasslands during the Late Miocene (Fig. 3d–f). We conclude that the onset of C₄ expansion in Anatolia during the Late Miocene (at the latest at ca. 9.9 Ma) is roughly coeval with the expansion of C₄ ecosystems in NW and E Africa and predates its rise in other Asian and African regions (Fig. 3f). However, C₄ vegetation dominated Anatolian floodplains and potentially parts of the Aegean by ca. 7.2 Ma, which roughly coincides with the start of the rapid rise to dominance of C₄ vegetation in southern Asian grasslands between 7.8 and 7.2 Ma (Fig. 3c, d; e.g., Behrensmeyer et al., 2007; Quade and Cerling, 1995; Tauxe and Feakins, 2020) and predates the rise to dominance of C₄ vegetation in the grasslands of the southern East African Rift during the Pliocene (Fig. 3e; e.g., Cerling et al., 2011; Lüdecke et al., 2016).

The rise to dominance of C₄ vegetation in East Asian grassland ecosystems is hypothesized to have occurred in response to declining atmospheric pCO₂ levels that led to Late Miocene Cooling (LMC, ca. 7.5 to 5.5 Ma; Fig. 3b; Herbert et al., 2016) when low pCO₂ provided an evolutionary advantage for C₄ over C₃ vegetation despite decreasing temperatures (e.g., Polissar et al., 2019; Wen et al., 2023). Worldwide, LMC is manifested in a sharp drop in sea surface temperatures (SSTs; Fig. 3b; Herbert et al., 2016) and its onset is accompanied by a sharp decrease in δ13C values of benthic foraminifera (Fig. 3a; Westerhold et al., 2020), which attests to a profound change in the global carbon cycle and consequently ocean circulation patterns (Holbourn et al., 2018). Climatic reconstructions from Anatolia based on the coexistence approach (Table S2) indicate a ca. 2–3 °C decrease in mean annual and an up to 10 °C decrease in cold month mean temperatures during the Late Miocene in the CAP interior, although uncertainties in both reconstructed temperatures and ages are large (see Fig. 2b, c). Because the rise to dominance of C₄ vegetation in Anatolian floodplains and possibly the Aegean occurred simultaneously with its rise in southern Asian ecosystems and LMC (Fig. 3c, d) we suggest that it was caused by drivers that go beyond regional environmental changes such as surface uplift of the CAP since ca. 11 Ma (Fig. 2d; Meijers et al., 2018).

Whereas C₄ grassland expansion in southern Asia, the Chinese Loess Plateau, and Arabia coincided with aridification (Huang et al., 2007; Wen et al., 2023), the (gradual) expansion of C₄ vegetation in northern and eastern Africa was not driven by aridification (Crocker et al., 2022; Polissar et al., 2019). The latter also appears to hold for Anatolia and the Aegean, as indicated by Anatolian paleobotanical datasets suggesting subhumid conditions (e.g., Kayseri-Özer, 2017; Table S3) and mesic environments in Anatolia and the Aegean instead (Denk et al., 2018).

Around ca. 8.0–6.6 Ma, the Pikermian chronofauna peaked in terms of geographic extent (Fig. 2d), including large parts of Europe and Central Asia (Eronen et al., 2009). We suggest that the spread of C₄ grasslands, which started before 9.9 Ma, and their dominance in floodplain environments by 7.2 Ma led to the expansion of the hypsodont Pikermian chronofauna in Anatolia. A similar process is observed in southern Asian grasslands, where combined soil and fossil mammal tooth enamel δ13C values show that long-term climate forcing changed the vegetation structure (C₃ to C₄) between 8.5 and 6.0 Ma and led to the disappearance of most mammalian lineages that fed on C₃ vegetation (Badgley et al., 2008).

A marked decrease in our pedogenic carbonate δ13C values by approximately 7 ‰ from 4.9 to 3.9 Ma (Fig. 2a) implies a major overturn of Anatolian and potentially Aegean ecosystems. This transition led to the reemergence of landscapes dominated by C₃ vegetation, a feature that persists to the present day. Plant leaf wax δ13C values and pollen data from a sediment core in the Gulf of Aden reveal a brief reversal in the trend of increasing C₄ vegetation contributions between 4.9 and 4.6 Ma (Feakins et al., 2013). On and around the Chinese Loess Plateau, variations in C₄ and C₃ biomass have been linked to the strengthening and weakening of the East Asian summer monsoon (e.g., An et al., 2005; Passey et al., 2009; Yao et al., 2010). However, these fluctuations are not persistent over time. Therefore, the rapid and persistent Pliocene switch to C₃ vegetation is a feature that is so far unique to Anatolia. By contrast, southeast China (Fig. 3f) and regions in the New World (see e.g., Polissar et al., 2019) have been colonized by C₄ vegetation only after the return to C₃-dominated vegetation in Anatolia. Given that the return to C₃ vegetation is geographically limited we assess potential causes for C₄ decline in Anatolia within a regional framework.

The Messinian Salinity Crisis (MSC) in the Mediterranean basin (5.96–5.33 Ma; e.g., Roveri et al., 2014) had the potential to alter the regional hydrologic cycle and hence vegetation. However, considering that the MSC is relatively short-lived and predates the decline of C₄ grasslands in Anatolia between 4.9 and 3.9 Ma, it is unlikely to be the main driver of the prolonged changes in the vegetation structure of the Eastern Mediterranean region.

During the Early Pliocene, regions adjacent to Anatolia also experienced significant environmental changes. In Central Europe, increased aridity led to the disappearance of open woodland, broad-leaf evergreen, and humid sclerophyllous taxa (Mosbrugger et al., 2005), some of which found refuge in Anatolia and the Eastern Mediterranean region (Eronen et al., 2009). Concurrently, Eastern Europe experienced a transition to cooler, drier conditions, leading to the expansion of grasslands and an enhanced fire regime since ca. 4.4 Ma (Feurdean and Vasiliev, 2019). The divergence in climate between a drier Central Europe and a more humid Eastern Mediterranean region (Kovar-Eder, 2003) has been attributed to the deflection of the westerlies from Central Europe to the Mediterranean region (Eronen et al., 2009). Its timing coincides with strengthening of the Atlantic Meridional Overturning Circulation (AMOC) in response to shoaling of the Central American Seaway (CAS) around 4.8–4.0 Ma (e.g., Haug and Tiedemann, 1998). We therefore surmise that the increase in SSTs over the North Atlantic associated with AMOC strengthening (Karas et al., 2017) caused hydroclimatic changes over Europe and the Mediterranean region.

Regions with prevalent C₄ vegetation today are not only characterized by high growing season temperatures but also by warm season precipitation. In contrast, Anatolia presently experiences hot, dry Mediterranean summers and receives most of its precipitation from October through May (Schemmel et al., 2013; Türkeş and Erlat, 2005). Notably, leaf morphologies of Miocene species of evergreen oak (Quercus sect. Ilex) from southwest Türkiye and Aegean islands resemble modern Himalayan members, which suggests (summer-)humid climatic conditions (see Denk et al., 2018, but also Denk et al., 2014, and Velitzelos et al., 2014). Additionally, Tortonian (ca. 11.6–7.2 Ma) paleoclimate simulations of Europe indicate humid to subhumid summer conditions in the northern Mediterranean region (Quan et al., 2014). We therefore propose that the shift from warm to cold season precipitation and the emergence of a Mediterranean-style climate drove the demise of C₄ grasslands and a return to a C₃-dominated environment in Anatolia and the Aegean during the Early Pliocene (4.9 to 3.9 Ma). Such a higher proportion of precipitation falling during the cool season has been shown to result in a strong and rapid (3–4 years) effect on the relative abundance of C₃ and C₄ vegetation during the 1930s Dust Bowl and was successfully reproduced in experimental studies in modern grasslands in the Great Plains (United States; Knapp et al., 2020).

The Early Pliocene demise of C₄ biomass occurred simultaneously with significant large mammal faunal turnover (Huang et al., 2019) and the vanishing of open-landscape adapted large mammals from Anatolia (MN14; Eronen et al., 2009), which formed the last stronghold of the Pikermian chronofauna (Eronen et al., 2009). Our study therefore identifies long-term vegetation dynamics and indicates that Early Pliocene climate change profoundly and irreversibly transformed vegetation structures and drove a turnover in mammalian populations.