Paleoenvironmental conditions and drainage evolution of the central Anatolian lake system (Turkey) during late Miocene to Pliocene surface uplift

The world’s largest orogenic plateaus, the Puna-Altiplano and the Tibetan Plateau, are characterized by internally draining basins that persist under arid conditions in the rain shadow of orographic barriers at the plateau margins. Protracted rapid shortening and high surface uplift rates under arid conditions prevent fluvial systems from defeating the internally drained plateau regions, thereby providing a way for internally drained conditions to develop and persist (e.g., Sobel et al., 2003; Strecker et al., 2007).

At a smaller scale, the Pontide and Tauride mountains (Fig. 1A) border the semiarid Central Anatolian Plateau (CAP, Turkey; average elevation ∼1 km; Fig. 1A) to its north and south, respectively. Both mountain belts exhibit humid conditions on their windward sides toward the Black Sea and Mediterranean Sea. The onset of the rise of the CAP after ca. 11 Ma (Meijers et al., 2018b) was followed by surface uplift of the Pontide and Tauride Mountains after ca. 8 Ma (Cosentino et al., 2012; Yıldırım et al., 2011; Meijers et al., 2018b). Following a period of compression, central Anatolia has been under an extensional tectonic regime roughly since the Miocene-Pliocene boundary (ca. 5.3 Ma; Jaffey and Robertson, 2005; Rojay and Karaca, 2008; Fernández-Blanco et al., 2013; Özsayin et al., 2013). Today, the CAP is characterized by a bimodal drainage system. The CAP region roughly west of the Tuz Gölü fault (Fig. 1A) is characterized by the internally drained Konya closed catchment (Fig. 1A), which includes the highly evaporative Tuz Gölü (Turkish for “Salt Lake”). The remainder of the CAP is externally drained, by means of rivers flowing into the Black Sea, Mediterranean Sea, and Persian Gulf (Fig. 1A).

Active continental sedimentation has occurred in numerous central Anatolian basins since the late Oligocene (Fig. 1C; e.g., Leng et al., 1999; Ocakoğlu, 2002; Önal et al., 2004; Lüdecke et al., 2013; Mazzini et al., 2013; Meijers et al., 2016, 2018a, 2018b). Paleoenvironmental conditions of fluvio-lacustrine sedimentation and of the development of the central Anatolian lake system can presently be retrieved from the now-incised former depocenters.

Here, we present a comprehensive body of sedimentological data and stable isotope values (δ¹⁸O and δ¹³C, N = 665, including 564 new δ¹³C values) from 13 fluvio-lacustrine sections within the southern CAP (Fig. 1A, Table 1) that aims at understanding the hydrology and paleoenvironmental setting of the former depocenters. Ages of deposition are derived from radiometric and magnetostratigraphic ages (e.g., Meijers et al., 2018a, 2018b) and combined with data from (mammal) biostratigraphy (e.g., The NOW Community, 2017). The presented study includes observations and stable isotope data from one section sampled in an early Miocene basin and twelve sections sampled in late Miocene to Pliocene basins of the southeastern CAP. Based on our observations and stable isotope data of the late Neogene depocenters, we discuss lake distribution and connectivity, the paleoenvironmental setting, and the switch from internal to external drainage during times of a changing tectonic regime, surface uplift, and the westward tectonic escape of the Anatolian microplate.

The Pontide and Tauride Mountains that border the CAP locally exceed elevations of 3 km, and their steep slopes contrast with the subdued topography of the plateau interior. Owing to its position within the Africa-Eurasia convergence zone, Turkey consists of an amalgamation of geologic terranes. The Pontide Mountains are part of a Late Cretaceous to Paleogene fold-and-thrust belt with a basement of Variscan origin (Okay and Tüysüz, 1999). The Tauride Mountains are underlain by Pan-African basement (Kröner and Şengör, 1990) of the Anatolides-Taurides and are a fold-and-thrust belt of latest Cretaceous to Eocene age (Monod, 1977; Demirtaşlı et al., 1984; Özgül, 1984). The Anatolides-Taurides are separated from the Pontides by the Izmir-Ankara-Erzincan suture zone (Fig. 1B). The basement below most of the CAP consists of the Central Anatolian Crystalline Complex (Fig. 1B), which exposes metamorphic, ophiolitic, and igneous rocks (e.g., Whitney and Hamilton, 2004).

The Neogene sedimentary cover of the CAP includes marine and continental basins. After the deposition of the youngest marine rocks on the plateau interior during the middle Miocene (ca. 12–11 Ma; Poisson et al., 2016; Ćorić et al., 2012; Ocakoğlu et al., 2002), fluvio-lacustrine conditions prevailed. The continental sedimentary rocks are commonly intercalated with volcanic rocks (tuffs, ignimbrites, basalt) that are derived from a number of volcanic provinces. These include the middle Miocene and younger Yamadağ, Galatia, and Central Anatolian volcanic provinces (Fig. 1C; e.g., Wilson et al., 1997, Gürsoy et al., 2011, Aydar et al., 2012), which have in part a well-developed volcanic stratigraphy (Central Anatolian volcanic province ignimbrites between ca. 9 and 2.5 Ma; Aydar et al., 2012).

The presence of steep and narrow pre–late Miocene paleorelief in central Anatolia is evidenced by the results of sedimentary provenance studies in the region surrounding the Ecemiş fault segment of the Central Anatolian fault (Fig. 1A; Jaffey and Robertson, 2005). Extremely steep, deeply incised middle Miocene paleovalleys are found buried under continental series many hundreds of meters thick in the Gürün (Fig. 1C; Önal et al., 2004) and Dikme basins (Fig. 1C; Ocakoğlu, 2002), as well as along the Aksu River paleovalley. In the Mut Basin, paleovalleys are concealed by hundreds of meters of estuarine and marine sediments (Bassant et al., 2005; Eriş et al., 2005). Palynological assemblages with mountain-derived flora further suggest that these steep-relief areas locally attained substantial elevations (Akgün et al., 2007). Regional surface uplift of the CAP has occurred since ca. 11 Ma, and the development of orographic barriers along the Pontide and Tauride Mountains after ca. 8 Ma shaped the present-day Anatolian topography and climate (Figs. 2B, 2C; e.g., Yıldırım et al., 2011; Cosentino et al., 2012; Meijers et al., 2018b; Öğretmen et al., 2018).

The modern CAP is cross-cut by a number of transtensional and transpressional faults, most prominently the North Anatolian, East Anatolian, Central Anatolian, and Tuz Gölü faults. The North Anatolian and East Anatolian faults facilitate the present-day westward escape of the Anatolian microplate (Figs. 1A, 1B; McKenzie, 1976). Following distributed shortening during the late Miocene, the sedimentary basins of the CAP have recorded extensional deformation since ca. 5.3 Ma (i.e., the Miocene to Pliocene transition; e.g., Jaffey and Robertson, 2005; Rojay and Karaca, 2008; Fernández-Blanco et al., 2013; Özsayin et al., 2013).

The Pontide Mountains and Tauride Mountains orographic barriers that separate the CAP interior from the Black Sea and Mediterranean Sea, respectively, presently control climatic conditions in Anatolia (Fig. 1A; Schemmel et al., 2013). Mean annual precipitation (MAP) decreases from ∼1000−1500 mm on the windward sides of both ranges to 300−500 mm within the plateau interior. Mean annual temperatures (MATs) along the Black Sea coast of 13 °C are lower than MAT of 20 °C along the Mediterranean coast, while MAT averages 4−10 °C within the plateau interior.

During the early to middle Miocene, central Anatolia was characterized by a humid, subtropical climate (Akgün et al., 2007). A review of floral data sets indicates MATs of 17−21 °C from the late early Miocene to the late middle Miocene, after which MATs dropped through the late Miocene to 13−17 °C during the early Pliocene (Kayseri-Özer, 2017). Meanwhile, MAPs ranged between 1000 mm to 1400 mm from the late early Miocene to the Pliocene, with an intermittent period of lower values of ∼900 mm during the late to latest Miocene (Kayseri-Özer, 2017).

Vegetation reconstructions based on phytolith data indicate a mosaic of open-habitat grassland, woodland, and forest vegetation in central Anatolia during the Miocene (Strömberg et al., 2007). Phytolith and palynofloral records further indicate that open-habitat grasslands became increasingly dominant during the late Miocene (Akgün et al., 2007; Strömberg et al., 2007), consistent with mammal dental ecometrics and faunal similarity pattern data (Kaya et al., 2018). Along with the dominantly open-habitat grasslands, paleovegetation reconstructions indicate the presence of a large variety of other vegetation types covering central Anatolia during the Messinian (ca. 7.2–5.3 Ma), such as mixed mesophytic forests, broad-leaved evergreen forests, evergreen needleleaf forests, and xeric open woodlands or steppes, as well as areas with high percentages of zonal herb components (Kayseri-Özer, 2017; Denk et al., 2018). The abundance of mid-elevation conifers in the late Miocene (post–9 Ma) to Pliocene pollen record of the Çankırı Basin is consistent with elevations of ∼940–2000 m (Fig. 1C; Yavuz et al., 2017). During the Pliocene, the relative abundance of the sclerophyllous, legume-like and zonal herb components increased over Anatolia; a relative increase of broad-leaved components in central Anatolia with respect to western Anatolia has been interpreted to result from higher central Anatolian topography (Kayseri-Özer, 2017).

During the late Miocene, central Anatolia formed part of the Old World savannah paleobiome, which harbored the Pikermian chronofauna (Kaya et al., 2018). The post-Miocene demise of the Old World savannah paleobiome in the Anatolian region was possibly caused by an increase in humidity (Kaya et al., 2018), which may be reflected in the increase of MAP across the Miocene-Pliocene boundary (Kayseri-Özer, 2017).

Presently, the highly evaporative Tuz Gölü is one of the few remaining lakes in central Anatolia. Continental sedimentation within the Tuz Gölü Basin has been ongoing since the Tortonian (Fernandez-Blanco et al., 2013). The Tuz Gölü Basin is presently the terminal lake of the Konya closed catchment (Fig. 1A), with its subdued topography and elevations of ∼850–1100 m. The Konya closed catchment is fed by surface and ground water from the Tauride Mountains (Üstün et al., 2015). During wetter periods, in particular during interglacials, it hosted extensive “pluvial lakes”, the last of which remained in existence well into the Holocene; (e.g., Leng et al., 1999; Jones et al., 2007; Dean et al., 2013). The Konya closed catchment is separated from the eastern CAP by the Tuz Gölü fault (Fig. 1A). The eastern CAP (i.e., the portion east of the Tuz Gölü fault) attains overall higher elevations and hosts only a handful of small, closed and commonly fault-bounded evaporative basins in which lacustrine deposition continues (Lake Seyfe, Lake Tuzla, and the Sultan Marshes; Figs. 1A, 1C). The last episode of widespread lacustrine deposition occurred during the early Pliocene (Fig. 2C; Meijers et al., 2018b). River incision subsequently led to drainage of these former lacustrine areas toward all surrounding terminal sediment sinks, namely the Mediterranean Sea (via the Seyhan River), the Persian Gulf (via the Euphrates River), and the Black Sea (via the Kızılırmak River; Fig. 1A).

Twelve (12) middle Miocene to Pliocene continental sedimentary sections (Fig. 3) within the southern CAP were sampled for carbon (δ¹³C) and oxygen (δ¹⁸O) stable isotopes. Levels sampled for stable isotopes include lacustrine carbonates as well as marls, and calcareous claystones and siltstones. We evaluate the distribution and age range of the sampled basins as well as their δ¹³C and δ¹⁸O data alongside the stable isotope results obtained from the Kangal Basin (Kumarlı section; Meijers et al., 2018a). The age interval covered by each section is presented in Figure 4. For the chronologies of the sampled Miocene to Pliocene fluvio-lacustrine sedimentary rocks, we refer to Meijers et al. (2018a, 2018b). In summary, the chronologies are based on radiometric ages, magnetostratigraphy, and (mammal) biostratigraphy. The indispensable chronologies allow us to interpret the sedimentological observations and stable isotope results within a reliable temporal framework.

Twelve (12) sections (Fig. 3) were logged in the field by determining the rock type of each bed or sequence of beds (including clast type for coarse-grained sedimentary rocks as well as for ignimbrites). Stratigraphic thicknesses were measured using a TruPulse laser range finder (except for the Gürün and Hacıbekirli sections, for which a measuring stick was used) and dip corrected to obtain true stratigraphic thicknesses. Periodic recalibration to a reference base point indicates a cumulative thickness error of <1 m over the entire thickness of each section. The stratigraphic nomenclature (i.e., the named geologic formations and members used in the Results section) is based on the 1:100,000-scale geological maps of the General Directorate of Mineral Research and Exploration of Turkey (MTA, 2019).

A total of 564 samples were taken from carbonate-bearing levels of 12 measured sections. All δ¹⁸O results were reported in summary fashion in Meijers et al. (2018b). δ¹³C and δ¹⁸O results for the Kumarlı section (N = 101) were reported in Meijers et al. (2018a). Here we plot all δ¹³C and δ¹⁸O data alongside their corresponding stratigraphic logs (Fig. 5) with the ultimate aim of establishing common patterns of lake development across the evolving CAP. After drilling the finest component of the whole-rock samples in the laboratory, sample powders were digested in orthophosphoric acid and analyzed as CO₂ in continuous-flow mode using a Thermo MAT 253 mass spectrometer interfaced with a Thermo GasBench II at the Joint Goethe University– Senckenberg BiK-F Stable Isotope Facility, Institute of Geosciences, Goethe University Frankfurt, Germany. The analytical procedures follow Spötl and Vennemann (2003). Raw isotopic ratios were calibrated against NBS 18 and NBS 19 carbonate reference materials and against a Carrara marble in-house standard. Final δ¹⁸O and δ¹³C values are reported against Vienna standard mean ocean water (V-SMOW) and Vienna Peedee belemnite (V-PDB), respectively, with analytical uncertainties that are typically <0.10‰ (δ¹⁸O) and <0.07‰ (δ¹³C). The uncertainties are based on the standard deviations of the reference materials over the course of the measurements. Carbonate contents were derived from standard to sample total peak area comparison and uncertainties are <10% absolute (see Supplemental Table S1¹). Covariance between δ¹³C and δ¹⁸O was determined using the Pearson correlation coefficient (Table S1). The significance of the calculated covariance is determined by the (N-dependent) critical value. The critical value decreases asymptotically with N, for which we set the significance level to 95% (see the “Critical values” sheet in Table S1). Only intervals that display a significant covariance at the 95% level are listed in the Results section. Standard deviations of all obtained δ¹³C and δ¹⁸O values for each section refer to 1σ uncertainties.

All sections are characterized by shallowly dipping to horizontal bedding planes. For the locations of the sections, stratigraphic logs, age constraints, and δ¹³C and δ¹⁸O values, we refer to Figures 1 and 3–6, Table S1 (^{footnote 1}), and the Supplemental KML file (^{footnote 1}). For detailed information about the sampled age intervals of each section, we refer to Meijers et al. (2018a; 2018b). To ease the retrieval of information for each section, we describe the sedimentology and the stable isotope results of the sections below in alphabetical order.

Ürgüp Formation

Logged interval: 104 m; sampled interval: 20.0–93.5 m, N = 40

The lowest part of the Akmezar section (0–20 m) consists of mafic volcanoclastic tuffs (Göbü tuff; 6.5 ± 0.3 Ma, ⁴⁰Ar/³⁹Ar on plagioclase; Meijers et al., 2018b) and detrital volcanic tuff that contains olivine and feldspar phenocrysts as well as serpentinite clasts. The 20–93.5 m interval consists of an alternation of sandstones, claystones, marls, (argillaceous) carbonates, and soils with a prismatic structure. Following an unexposed interval between 93.5 and 102 m, the top of the section is capped by the ca. 2.6 Ma Valibabatepe ignimbrite (2.52 ± 0.49 Ma and 2.73 ± 0.08 Ma, ⁴⁰Ar/³⁹Ar on plagioclase; Aydar et al. [2012] and Higgins et al. [2015], respectively). The normal magnetic polarity of the ignimbrite elsewhere in the Central Anatolian volcanic province (Piper et al., 2013) very likely constrains its emplacement to within chron C2An.1n (3.032–2.581 Ma; Hilgen et al., 2012).

The mean δ¹³C value is –7.1‰ ± 1.8‰ and the mean δ¹⁸O value is 21.9‰ ± 0.7‰ for the sampled interval (73.5 m, N = 40). The interval from 58 to 93 m displays a significant negative covariance between δ¹³C and δ¹⁸O.

Tuzköy Formation and Ürgüp Formation (Kışladağ Limestone Member)

Logged and sampled interval: 130 m, N = 38

The Bayramhacı section is an alternation of sandstones, lacustrine limestones, argillaceous limestones, sandy limestones, claystones, and cross-bedded pumice gravel. Minor soft-sediment deformation structures with convolute bedding occur at ∼13 m (Fig. 7B) and between 30 and 42 m (10–30 cm thick; Fig. 7C). At ∼32 m, the section contains a cherty level. Between 46 m and 92 m, large convolute structures tens of meters across (Figs. 3, 7A) consist of deformed lake sediments and parts of the overlying pumiceous, nonwelded ignimbrite sheet. The deformed lake sediments consist of ripoff masses of up to 5 m thick that are embedded into the main ignimbrite sheet. Samples BH29 to BH31 were retrieved from such embedded lake sediments. An ash layer in the lower part of the section yielded an age of 9.96 ± 0.08 Ma (⁴⁰Ar/³⁹Ar on biotite; Meijers et al., 2018b).

The 130-m-thick sampled interval (N = 38) has a mean δ¹³C value of –0.2‰ ± 1.7‰ and a mean δ¹⁸O value of 25.5‰ ± 2.4‰. δ¹⁸O and δ¹³C variability below the slumped and disturbed interval (∼9‰ and ∼6‰, respectively) is higher than above the slumped interval (∼6‰ and ∼4‰, respectively).

Peçenek Formation and Kışladağ Limestone Member

Logged and sampled interval: 30.2 m, N = 22

The Boğazlıyan section consists primarily of (cross-bedded) sandstones, claystones, and silty claystones. The upper 9 m of the section are formed by limestones, which are in places nodular. 75 m below the base of the section, the Kızılkaya ignimbrite is exposed. The Kızılkaya ignimbrite was dated near the section at 5.02 ± 0.20 Ma (⁴⁰Ar/³⁹Ar on sanidine; Özsayin et al., 2013).

The mean δ¹³C value is –5.9‰ ± 1.1‰ and the mean δ¹⁸O value is 22.8‰ ± 1.4‰ for the 30.2 m sampled interval (N = 22). δ¹³C values show an overall decreasing trend from the base of the section to the top, from approximately –5 ‰ to –7 ‰, whereas δ¹⁸O values remain relatively constant. The relatively large 1.4‰ standard deviation for δ¹⁸O stems from one sample with δ¹⁸O = 28.4‰ (at 22.4 m); omitting this sample reduces the average δ¹⁸O value and uncertainty to 22.5‰ ± 0.7‰.

Ürgüp Formation

Logged and sampled interval: 76 m, N = 49

The Büyükkünye section comprises two subsections (each 38 m thick; Fig. S1 [^{footnote 1}]) that are separated by an oblique unconformity of uncertain origin (tectonic or erosional). The lower subsection consists of unconsolidated conglomerates with well-rounded limestone cobbles and boulders, as well as claystones, calcareous sandstones, and paleosols. A pumiceous felsic ash layer within the first subsection was dated at 5.55 ± 0.08 Ma and 5.60 ± 0.19 Ma (⁴⁰Ar/³⁹Ar on amphibole and plagioclase, respectively; Meijers et al., 2018b). The upper subsection consists of travertine and dolomitic, clayey, oncolitic and chalky limestones, as well as a limited number of marl and claystone layers.

Nearby outcrops show that both sections were previously overlain by the now-eroded ca. 2.6 Ma Valibabatepe ignimbrite (Aydar et al., 2012; Higgins et al., 2015).

The lower subsection (N = 13) has a mean δ¹³C value of –7.0‰ ± 0.6‰ and a mean δ¹⁸O value of 22.9‰ ± 0.3‰. The upper subsection (N = 36) has a mean δ¹³C value of –4.1‰ ± 1.7‰ and a mean δ¹⁸O value of 22.8‰ ± 0.6‰. The higher standard deviations indicate higher variability within the upper, carbonate-rich units. The upper subsection displays a significant negative covariance between δ¹⁸O and δ¹³C.

Gürün Formation (Terzioğlu Member)

Logged interval: 175.2 m; sampled interval: 30.0–116.2 m and 160.4–175.2 m, N = 44

The Gürün section is located within the Gürün Basin and consists of two white and beige lacustrine carbonate successions between three basalt flows (Karadağ Volcanics; Önal et al., 2004). From the bottom to the top, these basalts were dated at 19.90 ± 0.40 Ma, 17.79 ± 0.15 Ma, and 17.53 ± 0.14 Ma (⁴⁰Ar/³⁹Ar on whole rock; Meijers et al., 2018b). The lacustrine succession locally consists of marls and dolomite, and in some places contains cherts.

The lowest sampled interval (30.0–116.2 m) yields a mean δ¹³C value of 6.0‰ ± 1.8‰ and a mean δ¹⁸O value of 30.8‰ ± 2.6‰. The upper sampled interval (160.4–175.2 m) yields a mean δ¹³C value of –0.3‰ ± 1.6‰ and the mean δ¹⁸O value of 23.8‰ ± 0.7‰. Overall, the section displays a significant positive covariance between δ¹³C and δ¹⁸O.

Ürgüp Formation (Kışladağ Limestone Member)

Logged interval: 94 m; sampled interval: 55.5–93.8 m, N = 21

The sampled Güzelöz section starts just above the ca. 5 Ma Kızılkaya ignimbrite (e.g., Özsayin et al., 2013; Aydar et al., 2012) with a 6-m-thick pumiceous lahar (25–31 m) that reworks the underlying ignimbrite. It is overlain by an ∼40-m-thick interval of (cross-bedded) sandstones, silty mudflows, reworked ignimbrite, and conglomerate. Within this interval, paleosols alternate with mudflows and calcarenites. Toward the top of the section (71–94 m), the presence of chalky, oolitic and massive limestones alternating with unexposed intervals suggests a lower-energy lacustrine depositional environment.

The mean δ¹³C value is –2.7‰ ± 2.8‰ and the mean δ¹⁸O value is 21.5‰ ± 0.7‰ for the 38-m-thick sampled interval (N = 21). δ¹³C and δ¹⁸O display significant negative covariance.

İnsuyu Formation (Kızılbayır Member and Katrandedetepe Member)

Logged and sampled interval: 77.4 m, N = 98

The Hacıbekirli section is an alternation of limestones, sandy, silty, or clayey limestones, rare (cross-bedded) sandy limestones, and (uncemented) conglomerates, as well as marls, claystones, clayey sandstones, and calcareous sandstones. Travertine occurs toward the top of the section, from which no stable isotope data are presented here. A tuffite in a laterally equivalent section (37.742806°N, 34.740091°E) yielded ⁴⁰Ar/³⁹Ar ages of 6.5 ± 1.1 Ma (total fusion, plagioclase) and 6.22 ± 0.12 Ma (plateau age, amphibole; Meijers et al., 2018b). We rely on the latter for stratigraphic correlation of the Hacıbekirli section. The beds of the Hacıbekirli section are gently tilted to the WNW (302°/12°, dip direction and dip angle).

The δ¹³C and δ¹⁸O values (N = 98) average −5.4‰ ± 2.3‰ and 25.1‰ ± 2.4‰ and show a significant positive covariance throughout the section.

Incesu Formation

Logged and sampled interval: 176 m, N = 76

Located within the Sivas Basin, the stratigraphy of the Haliminhanı section consists of conglomerates, (cross-bedded) sandstones, siltstones, marls, claystones, and limestones. Some levels contain rootlets and cherts. The Haliminhanı section has been extensively studied for mammal stratigraphy, which indicates a MN11–MN12 age (e.g., The NOW Community, 2017; Kaya and Kaymakçı, 2013), i.e., 8.9–7.4/6.8 Ma (Hilgen et al., 2012).

The 76 δ¹³C and δ¹⁸O values from the Haliminhanı section do not systematically change from the base to the top of the section, and neither does the internal variability of the δ¹³C and δ¹⁸O data. The mean δ¹³C value is −2.5‰ ± 1.9‰ and the mean δ¹⁸O value is 22.1‰ ± 0.8‰.

İnsuyu Formation

Logged interval: −0.6 to 25.6 m; sampled interval: −0.6 to 17.1 m, N = 33

The Ibrala section was sampled in a quarry in the Karaman Basin. It consists almost entirely of lacustrine limestones, in particular chalky, massive, oncolitic and nodular limestones interbedded with thin marly layers. Some layers contain stromatolites. Higher-energy channels scouring the succession are filled with marls and oncolites. The section overlies a sequence of marine, oyster-bearing, nearshore and estuarine deposits of Serravallian age (13.82–11.62 Ma; Ćorić et al., 2012). The gradual, conformable transition from the marine to the continental environment therefore provides a maximum age of ca. 11 Ma (Tortonian) for the sampled continental rocks.

The 33 δ¹³C and δ¹⁸O values average −3.8‰ ± 2.3‰ and 24.1‰ ± 0.5‰, respectively; which indicates the rather high variability in the range of δ¹³C values (ca. −7‰ to −1‰) in comparison to the δ¹⁸O values. Overall, δ¹³C and δ¹⁸O tend to increase toward the top of the section. δ¹³C and δ¹⁸O show significant positive covariance.

Kızılırmak Formation and Kozaklı Limestone Member

Logged interval: −12 to 53.6 m; sampled interval: 0–53.6 m, N = 36

The Kozaklı section was sampled along a road cut near the town of Kozaklı. Large intervals of the section are unexposed. Its base is defined by the top of a 7.6 ± 0.5 Ma ignimbrite (⁴⁰Ar/³⁹Ar on plagioclase; Meijers et al., 2018b). The lowest exposed interval (0–10.5 m) consists of claystones that include rootlets, fossils, and gypsum rosettes as well as rare sandstone and limestone layers. The central exposed part of the section (18–26.6 m) includes (calcareous) claystones, (clayey) limestones, and layers rich in organic matter. The upper part of the central part of the section consists of an alternation of (silty) claystones and (cross-bedded) sandstones, including a layer that contains (unsampled) pedogenic carbonate. The top of the section (49.8–53.6 m) consists of clayey limestones that turn into (micritic, oncolitic) limestone beds (beds <10 cm thick) and massive limestones toward the very top. An additional age for the deposits in the sampled basin comes from a nearby mammal locality (MN14, 5.3–5.0 Ma, locality Kozaklı; The NOW Community, 2017).

Carbonate δ¹³C values (N = 36) vary widely (between −7‰ and +6‰), which is expressed in the high 1σ standard deviation of the average, −2.0‰ ± 4.3‰. The top of the section that consists dominantly of limestone yields significantly higher δ¹³C values compared to the rest of the section. The average δ¹⁸O value is 22.8‰ ± 1.2‰, and δ¹⁸O does not change significantly throughout the section. The lower part of the section (0–18.3 m) displays a significant positive covariance between δ¹³C and δ¹⁸O. The upper part of the section (18.3–53.6 m) displays a significant negative covariance.

Kangal Formation and Etyemez Member (member of Kangal Formation), overlain by the Göbekören Basalt

Logged and sampled interval: 0–135 m, N = 101

The Kumarlı section is located within the Kangal Basin. A detailed description of the Kumarlı section and the isotopic results are provided in Meijers et al. (2018a). The Kumarlı section consists of an alternation of (carbonaceous) claystones, (clayey) limestones, chalk, and marls, as well as rare conglomerate. Magnetostratigraphy, whole-rock basalt ⁴⁰Ar/³⁹Ar dating, and mammal stratigraphy constrain sediment deposition to between ca. 6.6 and 4.8 Ma (Meijers et al., 2018a).

The average δ¹⁸O value is 23.0‰ ± 1.1‰, and the average δ¹³C value is −7.6‰ ± 0.9‰ (N = 101). δ¹³C and δ¹⁸O display a significant positive covariance.

İç Anadolu group, Kozaklı Limestone Member

Logged and sampled interval: 89 m, N = 72

The Özvatan section consists entirely of chalky, micritic, and rarely dolomitic or oncolitic limestones, in places intercalated with marls. Parts of the section are covered in scree, which likely conceals marly intervals. The age of the section is based on a mammal assemblage dated as MN14 (5.3–5.0 Ma, locality Iğdeli; The NOW Community, 2017) in a time-equivalent section.

The 72 δ¹³C and δ¹⁸O values average −2.1‰ ± 0.9‰ and 20.1‰ ± 0.3‰, respectively. δ¹³C values roughly increase toward the middle part of the section, followed by a rough decrease toward initial values. δ¹³C and δ¹⁸O display a significant negative covariance throughout the section.

Ürgüp Formation

Logged and sampled interval: 96 m, N = 35

The Taşhan section was sampled starting 5 m below the highstand watermark of the Yamula reservoir on the Kızılırmak River near Taşhan village. The section starts with an alternation of cross-bedded siltstones and sandstones that are in places calcareous and include gravels of ophiolitic cherts, followed by an alternation of claystones and dominantly (massive, clayey or nodular) limestones that are covered by an ignimbrite. The ignimbrite (25.5–37.0 m) was dated at 7.51 ± 0.07 Ma and 7.52 ± 0.14 Ma (⁴⁰Ar/³⁹Ar, amphibole and plagioclase, respectively; Meijers et al., 2018b). The top of the ignimbrite is reworked into overlying coarse fluvial sediments. Above an unexposed interval, the top of the section (71–96 m) starts with a cross-bedded conglomerate that contains angular chert and mudball pebbles, followed by unconsolidated sands and massive red clays, and eventually (massive, nodular) carbonates, which are in places dolomitic and crumbly or nodular. The overlying 120 m of the sediment succession was not sampled and is overlain by a basalt flow, which is in turn separated by an erosional unconformity from the overlying Valibabatepe ignimbrite (2.52 ± 0.49 Ma and 2.73 ± 0.08 Ma, both ⁴⁰Ar/³⁹Ar; Aydar et al., 2012; Higgins et al., 2015).

The 35 δ¹³C and δ¹⁸O values average −0.6‰ ± 2.0‰ and 22.3‰ ± 0.5‰, respectively. The δ¹³C values tend to be higher in the upper, dominantly limestone portion of the section. No trend can be observed in the δ¹⁸O values.

The sampled sections belong to a series of depocenters that have been only moderately incised since sediment deposition ceased, such that the overall thickness of exposed sediment is potentially limited compared to the total sediment thickness. Most sections were sampled (nearly) to the top of sedimentary exposure (except for sections TS and HA), whereas the deepest sedimentary rocks filling each basin were typically not accessible (with the exception of section IB-KAR). Therefore, our results only document isotopic trends toward the end of sedimentation in each of the depocenters.

The time intervals captured within the sampled sections individually span ∼0.2–2.5 m.y. Age constraints are most robust within sections with ⁴⁰Ar/³⁹Ar ages from multiple levels (the GU section covers ∼2.5 m.y.) and by the KU section which was dated using magnetostratigraphy. The latter constrains the duration of sedimentation of the sampled KU section to ∼1.8 m.y. A compilation of well-dated Neogene Anatolian fluvio-lacustrine sections yields an average sedimentation rate of 7.5 cm/k.y. for Oligocene to Miocene Anatolian continental basins (Meijers et al., 2018b). According to this rate, the shortest section (IB-KAR, 17.7 m) still covers a time interval of >200 k.y.

Changes in lake carbonate δ¹³C values generally result from changes in biogenic productivity (e.g., Li and Ku, 1997). Because ¹²C is preferentially partitioned into lake organic matter, an increase in biogenic productivity results in a relative increase of ¹³C in dissolved CO₂, which is taken up by carbonates that precipitate from lake water. Therefore, an increase in δ¹³C values generally reflects an increase in lake productivity. Further, an increase in evaporation generates higher δ¹⁸O values, because ¹⁶O is preferentially removed from the lake water during evaporation.

Based on information about the sediment and water supply (which is linked to precipitation and evaporation) as well as the potential accommodation space (which is linked to subsidence) of a lake, Carroll and Bohacs (1999) designed a lake basin classification. The tripartite lake basin classification categorizes lakes as either overfilled, balanced fill, or underfilled. From a stable isotope perspective, lake hydrology can be categorized as open or closed. The former refers to lakes that maintain a stable water level, which is controlled by a surface outlet. Conversely, a closed lake is characterized by limited surface or underground outflow, meaning that most water exits the lake through evaporation. Under such circumstances, lake levels respond to fluctuations in the amount of riverine input and evaporation. A closed lake hydrology is indicated by a positive covariance between δ¹³C and δ¹⁸O, when evaporation and lake productivity are generally governed by changes in temperature during carbonate formation. A hydrologically closed lake corresponds to an underfilled lake in the classification of Carroll and Bohacs (1999). Hydrologically open lakes correspond to balanced fill and overfilled lakes.

Data from five sections exhibit a significant positive covariance between δ¹³C and δ¹⁸O values, which points to closed lake hydrology: sections GU, IB-KAR, HC, KU, and the lower part of KZK (Figs. 4, 6; Table S1 [^{footnote 1}]). Within the data sets from the remaining sections, we observe either no covariance between δ¹³C and δ¹⁸O (sections BH, TS, BO) or a negative covariance (sections GO, OV, BK, AZ, HA, and the upper part of KZK; Figs. 4, 6; Table S1), which indicates open lake hydrology. Open as well as closed lake conditions are observed rather randomly across the study region, irrespective of their age or location. We hence do not discern any spatiotemporal trend across our regional data set.

Akmezar (AZ). The lower part (0–20 m) of the Akmezar section consists of the Göbü tuff (6.5 ± 0.3 Ma; Meijers et al., 2018b). The section is covered by the ca. 2.6 Ma Valibabatepe ignimbrite (above 102 m; Aydar et al., 2012; Higgins et al., 2015), atop an unexposed interval between 93.5 and 102 m. Where exposed in the vicinity of the section, the sole of the ignimbrite overlies thermally heated evolved paleosols, which indicates a time gap between deposition of the topmost part of the section and the emplacement of the ignimbrite. The ignimbrite forms a regionally extensive mesa, which has been incised by the Zamantı River. The section spans the entire thickness of underlying sediments that were exposed by river incision along the river valley. Ignimbrite occurrences on the valley floor ∼30 km farther downstream the Zamantı River show that the valley had already been incised close to its current depth when the ignimbrite was emplaced and draped the valley and its surrounding terraces (Fig. S2 [^{footnote 1}]). Based on a number of sections with multiple age constraints, the estimated average sedimentation rate in central Anatolia is 7.5 cm/k.y. (Meijers et al., 2018b). The youngest sedimentary rocks at the top of the Akmezar section are therefore estimated to be ca. 5.3 Ma. The estimated time gap of 2.7 m.y. between the youngest sedimentary rocks at top of the section and ignimbrite emplacement therefore brackets the time span (5.3–2.6 m.y.) during which incision of the Zamantı River started.

Bayramhacı (BH). The convolute bedding at ∼13 m and between 30 and 42 m (Figs. 7B, 7C) was produced on a flat-lying lake bed and resembles structures produced by ground shaking (seismites). We interpret the slumped interval between 46 m and 92 m (Figs. 2, 7A), however, as deformation resulting from the emplacement of the ignimbrite onto the soft lake sediments. The large-scale convolute bedding may be the result of rapid gravity-driven sediment mobilization during the emplacement of the massive pumiceous ignimbrite.

Based on the age of the ash layer near the base of the section (9.96 ± 0.08 Ma, ⁴⁰Ar/³⁹Ar on biotite; Meijers et al., 2018b), the regional sedimentation rate of 7.5 cm/k.y. (Meijers et al., 2018b), and assuming that the slumped interval (46–92 m) was deposited instantaneously, the 130-m-thick section most likely covers the time interval from 10.1 to 9.0 Ma. The large pumiceous ignimbrite occurring 35 m above the dated ash layer within the section is most likely the Kavak or Zelve ignimbrite. The most recent ages obtained for the Kavak ignimbrite range from 10.5 to 9.1 Ma, and for the Zelve ignimbrite, 9.9–9.1 Ma (Lepetit et al., 2014; Aydar et al., 2012). Given the proposed spatial extent of the Kavak and Zelve ignimbrites (Le Pennec et al., 1994), the ignimbrite within the Bayramhacı section is most likely the Zelve ignimbrite.

Boğazlıyan (BO). The Boğazlıyan section comprises the topmost beds of a much thicker succession that contains two interbedded ignimbrites of 6.81 ± 0.24 Ma and 5.02 ± 0.20 Ma (Özsayin et al., 2013). If we assume that the sedimentation rate after the deposition of the second ignimbrite did not change compared to the 8.0 cm/k.y. sedimentation rate between the two ignimbrites, we find that the Kışladağ Limestone Member at the top of the section was deposited ca. 3.7 Ma.

The sedimentary rocks of the Boğazlıyan section were deposited over the footwall of the Tuz Gölü fault, which presently acts as a normal fault under a NE-SW extensional regime (Kürçer and Gökten, 2012; Özsayin et al., 2013; Yıldırım, 2014). Prior to 6.8 Ma, and possibly until the Pliocene, the fault acted as a strike-slip fault with a minor reverse component (Özsayin et al., 2013). Changeover from a compressional to an extensional stress regime of the Tuz Gölü fault occurred by the start of the Pliocene (Özsayin et al., 2013). The sedimentary rocks of the BO section are perched ∼400 m above the actively aggrading basin floor of the Tuz Gölü Basin in the hanging wall of the fault. The 400 m represents the minimum vertical separation across the fault since deposition of section BO. Sedimentary rocks exposed at section BO indicate a low-energy depositional environment, and therefore the Tuz Gölü fault was unlikely to have had a surface expression during deposition of the sediments.

Büyükkünye (BK). The Büyükkünye section consists of two subsections (Fig. S1 [^{footnote 1}]) juxtaposed across an unexposed erosive or tectonic contact. One subsection, which we interpret to be in a lower stratigraphic position, contains clastic sedimentary rocks, paleosols, and a 5.6 Ma pumiceous ash layer (Meijers et al., 2018b). The other subsection consists entirely of lacustrine limestones with travertine intervals. The rock types of the lower subsection indicate sedimentation under high- to low-energy fluvial depositional conditions, allowing for the preservation of the tuff layer, with intermittent periods of ceased sedimentation that allowed for soil formation. The limestone beds of the upper subsection indicate carbonate precipitation under persistent lacustrine conditions, with little evidence for nondepositional intervals, and little clastic input. The unexposed contact between the two sections may be the steep flank of a paleovalley incised in the sediments of the clastic section, or a steeply dipping fault with a normal component with the clastic section in the footwall. The former explanation seems less likely, given that it would be very difficult to maintain a steep paleovalley in soft clastic sediments during deposition adjacent to lacustrine carbonate without any downslope transport and dispersal of clastic sediment into the carbonates. We therefore assume that the two subsections were offset by a fault after deposition. A minimum age for both sections is provided by the ca. 2.6 Ma Valibabatepe ignimbrite (Aydar et al., 2012; Higgins et al., 2015), which caps the hill that exposes the two subsections. In the surrounding region, the ignimbrite flowed across an already differentiated landscape of flat-lying, unincised former Pliocene sediment depositional surfaces, as well as already deeply incised narrow stream valleys (see, e.g., Fig. S2 [^{footnote 1}]). However, the areal extent of the dissected landscape was not as extensive as today. By analogy to section AZ, the end of fluvio-lacustrine sedimentation was likely followed by several million years of nondeposition prior to the emplacement of the Valibabatepe ignimbrite.

Gürün (GU). Unlike sections from other basins presented in this study, the Gürün Basin was affected by substantial post-depositional folding (Önal et al., 2004). In its lower part, the GU section yields the highest δ¹⁸O values of all sampled sections. The positive covariance between δ¹³C and δ¹⁸O indicates closed lake conditions. Across the basalt that separates the lower and upper sampled lacustrine sequence, a large decrease of 6‰–7‰ occurs in both δ¹³C and δ¹⁸O values. The lower δ¹⁸O and δ¹³C values in the upper part of the section are indicative of a more positive water balance alongside a decrease in biogenic productivity. Given that the isotopic shift in both carbon and oxygen directly followed lava emplacement, we conclude that the flow likely modified the lake’s volume and/or dimensions. The change in isotopic composition can therefore be explained not by climatic factors, but rather by modifications to lake connectivity and openness.

Spore and pollen fossils of the Çayboyu Member of the Gürün Formation indicate that the Gürün Basin fill is of middle Miocene age (Önal et al., 2004). However, ⁴⁰Ar/³⁹Ar chronology (three whole-rock ages, Burdigalian; Meijers et al., 2018b) of the younger Terzioğlu Member indicates that the basin is older, and of early Miocene age. The GU section is therefore the only section that is not late Miocene to Pliocene in age.

Güzelöz (GO). Published stable isotope geochemistry results from the Güzelöz section (six samples; Göz et al., 2014) fall within the range of the δ¹³C and δ¹⁸O values presented here. Assuming that sedimentation at section GO occurred with an average regional sedimentation rate of 7.5 cm/k.y. (Meijers et al., 2018b), sedimentation at section GO would have continued well into the Pliocene (until ca. 3.5 Ma), therefore likely making it one of the youngest sampled sections.

Hacıbekirli (HC). The beds of the Hacıbekirli section slant slightly to the northwest (302°/12°, dip direction and dip angle). The sequence forms part of the northern limb of the Tauride Mountains antiformal arch (Radwany et al., 2017). Final sedimentation at section HC is estimated to have occurred at ca. 5.7 Ma, based on an average regional sedimentation rate of 7.5 cm/k.y. (Meijers et al., 2018b). In the absence of detectable fanning of bedding dip at the scale of the outcrop, it appears that any syndepositional tilting over the duration of deposition of the measured section is small compared to the finite amount of post-depositional tilting. This means that the distal northern limb of the Tauride arch continued to develop after ca. 5.7 Ma, through broadening and/or amplification.

Haliminhanı (HA). The age of the Haliminhanı section is constrained to 8.9–7.4/6.8 Ma (Hilgen et al., 2012) by mammal stratigraphy, which indicates an MN11–MN12 age (e.g., The NOW Community, 2017; Kaya and Kaymakçı, 2013). Atop the sampled stable isotope section, another 160 m of unsampled stratigraphy is exposed, and terminates with the deposition of less-clastic lacustrine carbonates. If deposition continued at the regional rate of 7.5 cm/k.y. (Meijers et al., 2018b), fluvio-lacustrine sedimentation in this part of the Sivas Basin would have continued for another 2.1 m.y., thereby possibly continuing into the Pliocene.

Ibrala (IB-KAR). The Ibrala section is the only section where fluvial and lacustrine rocks directly and conformably overlie nearshore and estuarine marine rocks. The Serravallian age (13.82–11.62 Ma; Ćorić et al., 2012; Hilgen et al., 2012) of the marine rocks therefore marks the switch from marine to continental sedimentation in the southwestern portion of the CAP (see also Meijers et al., 2018b).

Kozaklı (KZK). δ¹³C and δ¹⁸O values in the Kozaklı section change from positive covariance to negative covariance upsection, thereby indicating a shift from closed to open lake conditions. Paleolake Kozaklı is the sole Anatolian lake within our study that displays this shift. The ignimbrite at the base of the section was dated at 7.6 ± 0.5 Ma (Meijers et al., 2018b). Using an average sedimentation rate of 7.5 cm/k.y. (Meijers et al., 2018b), the change in lake hydrology occurred at ca. 7.4 Ma. Assuming the same rate, fluvio-lacustrine sedimentation at Kozaklı may have continued until ca. 6.8 Ma.

Kumarlı (KU). Magnetostratigraphy and an ⁴⁰Ar/³⁹Ar age from the top of the section indicate that fluvio-lacustrine sedimentation in the Kangal Basin must have ceased by ca. 4.8 Ma (Meijers et al., 2018a).

Özvatan (OV). The (upper Miocene–)lower Pliocene lacustrine carbonates of the Özvatan section are interfingered with alluvial fan deposits that are derived from higher-elevation areas of the Central Anatolian Crystalline Complex (Fig. S3 [^{footnote 1}]). Initially restricted to the basin sides, the alluvial fan deposits must have prograded across the lacustrine domain, thereby depositing a thin, extensive layer of clastic sediments over the lacustrine unit.

Taşhan (TS). Similar to the HA section, the Taşhan section continues above the sampled interval for another ∼100–140 m (2–3 km southwest of section TS) and is capped by a basalt. Using the regional 7.5 cm/k.y. sedimentation rate (Meijers et al., 2018b), sedimentation possibly continued until ca. 5.4–4.9 Ma. The Valibabatepe ignimbrite (ca. 2.6 Ma; Aydar et al., 2012; Higgins et al., 2015) unconformably covers the top of the section, which suggests a significant amount of time between final fluvio-lacustrine sedimentation and ignimbrite emplacement.

In spite of a great variety of rock types and fluctuating depositional environments, sedimentation in the investigated sections almost invariably finishes under limestone-forming lacustrine conditions (Fig. 3). All sections sampled in upper Miocene–Pliocene rocks (except for section TS) display a terminal phase of limestone deposition just before sedimentation ceased and incision began.

It would be difficult to hold climate change responsible for the final phase of lacustrine limestone deposition, because the onset of final lacustrine sedimentation occurred asynchronously over the various sampled basins (Figs. 3, 4). Alternatively, because limestones form protective caps that prevent the erosion of underlying layers, the systematic termination of sedimentation with the deposition of relatively resistant layers could be regarded as preservation bias. In such a case, any clastic layer deposited after limestone deposition would have been eroded away. This scenario is highly unlikely, however, because (1) the contrast in erodibility between clastic and limestone layers is not sufficient for erosion to perfectly strip off a soft cover without incising the limestones and leaving residual patches of the overlying cover along drainage divides, and (2) thin, continuous clastic levels of at most a few meters thick are locally found between the cap carbonates and overlying ignimbrites. Therefore, we regard the trend of final limestone deposition within most basins as a true, diachronous signal, independent of climate change. We argue that the process leading to the cessation of sedimentation in the intramontane basins is also responsible for the deposition of the cap carbonates.

Although the sections are widely distributed throughout the study area, some lie in close proximity and their depositional ages overlap. Deposition, therefore, may have occurred within the same lake. The size, shape, and extent of the lakes are difficult to constrain. It is unclear whether the cap carbonate mesas of the present-day CAP are the scarce remnants of much wider lakes, or the well-preserved remnants of much smaller lakes, for which evidence may be qualitatively deduced from the age constraints in Figure 4. Additionally, because the upper Miocene to Pliocene lakes occupied inward-draining basins, their size and extent probably fluctuated rapidly and widely in response to changes in water balance. One way to test whether separate sections of similar age belong to a single lake is to compare their isotopic record, and to test whether they share similarities in δ¹⁸O and δ¹³C values and covariance. While similarities are a strong argument for sedimentation within a common lake, the reverse is not true, as wide, shallow lakes can maintain perennial spatial variations in average annual δ¹⁸O values (Shi et al., 2017; Wen et al., 2016). While inconclusive, a discussion on the extent and connectivity of the former lakes can be found in the Supplemental Material (File S1 [^{footnote 1}]).

The sampled former depocenters evolved during the rise of the Anatolian plateau and its northern and southern margins (Yıldırım et al., 2011; Meijers et al., 2018b) between 11 and 5 Ma (Cosentino et al., 2012; Meijers et al., 2018b). The rise of the CAP and its southern margin is expressed by a 3‰−4‰ decrease of the lowest 25% of δ¹⁸O values of the lacustrine carbonates, interpreted to reflect a similar decrease in δ¹⁸O of precipitation as a result of plateau uplift (Meijers et al., 2018b). Paleoprecipitation estimates based on the coexistence approach indicate MAPs of 1000−1400 mm from the late early Miocene to the Pliocene, with a period of lower values of ∼900 mm during the late to latest Miocene (Kayseri-Özer, 2017). The paleoprecipitation estimates therefore indicate a (sub)humid Anatolian climate during late Miocene surface uplift. Eventually, the rise of the Tauride and Pontide Mountains generated a rain shadow, which led to semiarid conditions within the plateau interior that contrast with the humid conditions observed along the Mediterranean and Black Sea coasts. The present-day rain shadow is expressed by low δ¹⁸O values of surface waters along the immediate leeward flanks of both mountain belts; higher δ¹⁸O values in the plateau interior result from evaporation (Schemmel et al., 2013). Today, the remaining lakes of the CAP experience highly evaporative conditions, resulting in δ¹⁸O values of lacustrine carbonate of up to 36‰ in close-to-modern lake carbonates (late Pleistocene to Holocene; e.g., Leng et al., 1999; Roberts et al., 2001; Jones et al., 2006, 2007; Dean et al., 2015).

The δ¹³C and δ¹⁸O values obtained from the former depocenters show little variability in a number of cases (as expressed by standard deviations ≤1.1‰ for sections KU and OV and the lower part of section BK), and individual records display no significant trend toward higher or lower values with time (sections KU and OV and the lower parts of sections BK, as well as sections AZ, BH, HA, TS), in spite of coeval changes in sediment composition. We therefore conclude that lake hydrology was relatively stable over extended periods of time: ∼1.8 m.y. for section KU, and arguably >1.0 m.y. for section HA. Such stability is remarkable, given that the studied basins evolved during a tectonically active period, namely the formation of the CAP and its southern and northern margins. Additionally, no increasingly evaporative trend (as expressed in modern lake δ¹⁸O values >36‰) is observed within the Miocene to early Pliocene lake carbonates during surface uplift of the CAP and its southern margin, which one would expect during the formation of a rain shadow. The highest δ¹⁸O values of up to 35‰ occur within the oldest sampled carbonates (ca. 20–17.5 Ma, section GU; Figs. 4, 5). During surface uplift (11–5 Ma), only carbonates at the top of section HC (ca. 6 Ma) range up to 33‰. Besides one δ¹⁸O value of ∼28‰ within section BO, all other δ¹⁸O values reported here, as well as within the previously published ca. 10 Ma Tuğla section (Çankırı Basin; Fig. 1C; Mazzini et al., 2013), remain <27‰. Late Miocene surface uplift therefore seemingly did not generate intense aridification of the CAP interior during the late Miocene to early Pliocene, although this is a trend that may be obscured in hydrologically open lakes (Fig. 4) where through-flowing waters could counter an aridification trend.

We explain the absence of an aridification trend (i.e., a temporal trend toward higher δ¹⁸O values in the sampled sections) by the relatively humid climate of Miocene to Pliocene Anatolia. Under humid conditions, the water balance of the lakes during carbonate precipitation could have continued to remain positive, thereby explaining the absence of δ¹⁸O values indicative of high evaporation. Similarly, the absence of an aridification trend by the time of the full establishment of a Tauride Mountains orographic barrier at 5 Ma may be explained by the counteracting increase in MAP into the Pliocene (Kayseri-Özer, 2017).

Summer drought has been recorded on the Iberian Peninsula since the late Pliocene (ca. 3.4 Ma; Jiménez-Moreno et al., 2010). Similarly, grassland expansion in Anatolia since the late Miocene (Akgün et al., 2007; Strömberg et al., 2007) may not have led to pronounced seasonality in the eastern Mediterranean until the late Pliocene. In concert with high MAP, reduced seasonality in Anatolia during late Miocene to Pliocene surface uplift may have allowed year-round delivery of moisture to the Anatolian lake system and the absence of an aridification trend in lake records.

The Çankırı Basin at the northern end of the CAP interior seems to be an important exception in terms of absence of aridification. Here, hypersaline lake conditions resulted in the deposition of evaporites within vast areas during the late Miocene and early Pliocene (Karadenizli, 2011; Mazzini et al., 2013), thereby clearly marking (semi)arid conditions.

An environmental event that could have affected the climate of the CAP is the Messinian Salinity Crisis (MSC). During the MSC, dramatic climatic changes occurred within the Mediterranean and Paratethys basins as a result of restricted connectivity of these basins with the Atlantic Ocean. The δ¹³C and δ¹⁸O values of section KU, however, indicate that the MSC did not cause significant changes in the hydrology of lake Kangal (Meijers et al., 2018a).

A compilation of the youngest marine sedimentary rocks (Meijers et al., 2018b) reveals that the switch to nonmarine sedimentation likely occurred at ca. 12–11 Ma in the southern portion of the CAP (compare the ca. 13 Ma and ca. 8 Ma paleogeographic sketches in Figs. 2A, 2B). This provides a maximum age for the onset of lacustrine sedimentation in the continental basins that have since covered much of the CAP. Paleogeographic information (Figs. 2A, 2B) suggests that the Anatolian lake system expanded during surface uplift of the CAP interior.

Lacustrine sedimentation in the southeastern CAP ended at or around the Miocene-Pliocene boundary (ca. 5 Ma; Fig. 4), with the exception of sections BO and GO, where sedimentation continued until ca. 3.7 Ma and ca. 3.5 Ma, respectively (Zanclean-Piacenzian boundary, i.e., the early to late Pliocene boundary). Sedimentation in the northern CAP (Çankırı Basin; Fig. 1C) also continued until ca. 3.6 Ma (Kayseri-Özer, 2017), whereas sedimentation within a few large depocenters in the present-day Konya closed catchment area likely continued well into the late Pliocene to Pleistocene and even up to the present day in its northern portion (lake Tuz Gölü; Figs. 1A, 1C; see, e.g., Fernandez-Blanco et al., 2013). In summary, lacustrine sedimentation ceased between the late Miocene and early Pliocene east of the Tuz Gölü fault (Figs. 1A, 1C), while it continued west of the fault. Former depocenters of the eastern CAP are now incised by rivers that drain toward the Mediterranean Sea, the Black Sea, and the Persian Gulf.

In the late Miocene–early Pliocene, numerous lakes occupied the CAP. According to the carbonate isotope record, most lakes were hydrologically open, with the exception of two lakes (Fig. 4) located near the southern margin of the plateau, as well as in the Çankırı Basin near the northern margin of the plateau (ca. 10 Ma; Fig. 1C; Mazzini et al., 2013). Of the two closed lakes (sections HC and KU) within the plateau interior, one former lake (section HC) is presently located within the Konya closed catchment. The open lakes drained either to the surrounding seas or to terminal lakes within the plateau interior.

Incision of the lake outlets remained limited, allowing lacustrine sedimentation under open lake conditions over more than a million years in most basins. Lake Kozaklı is the only lake that experienced a switch from closed to open hydrology at ca. 7.5 Ma. Inception of overflow at the time did not lead to any substantial incision of the outlet, allowing for the lake to exist for an additional ∼700 k.y. First, limited outlet incision requires very low topographic gradients within the plateau interior. Second, if the lakes drained to the surrounding seas, then the steepening of the plateau margins resulting from plateau uplift would have resulted in the incision of the plateau margin drainage and the progressive propagation of incision upstream toward the plateau interior. The fluvio-lacustrine basins of the plateau interior would eventually have been dissected and the lakes would have drained away. It is therefore likely that the open lakes instead drained into a single or multiple terminal lakes, such as Lake Kangal, thereby maintaining a high base level within the plateau interior.

Yet, between ca. 5 Ma and ca. 3.6 Ma (i.e., during the early Pliocene), most of the lakes in the eastern CAP drained away and their floors were incised (Figs. 2C, 4). If a major shift from closed lakes to open lakes was observed at the time, one could argue that lake overflow would have been the cause of drainage integration across the plateau interior and their subsequent incision. Yet, open lake conditions persisted for long periods of time.

The establishment of connections between the internal drainage and the surrounding seas, in particular the Mediterranean Sea after 5.5 Ma, as evidenced by provenance studies in the Adana Basin (Fig. 1C; Radeff et al., 2017), and possibly over the same period with the Persian Gulf and Black Sea, would have allowed for incision to propagate into the plateau interior, thereby ending lacustrine conditions in the southeastern CAP after the Miocene (Fig. 2C). The Zamantı River, a tributary of the Seyhan River that drains the southeastern CAP, would have then started reaching the Mediterranean coast in the Adana Basin at 5.5 Ma. Incision would have propagated upstream along the Zamantı River and its tributaries, thereby reaching section AZ at 5.3 Ma and section BK at 4.9 Ma. Some 10–30 km downstream of these sections, the Zamantı River had already reached its current depth of incision by the time of deposition of the ca. 2.6 Ma Valibabatepe ignimbrite.

Sedimentation east of the Tuz Gölü fault likely ended (ca. 3.7 Ma, section BO) slightly before sedimentation in the Çankırı Basin (at ca. 3.6 Ma), when the Kızılırmak River led to the incision of sediment fills of the former depocenters in both areas. Their incision marked the end of the late Miocene to early Pliocene phase of drainage integration and lake disappearance. Since the early Pliocene, little has changed in the drainage system of the southern CAP, when it was composed of an internally draining Konya closed catchment in the west and a dissected plateau in the east.

The combination of established chronologies with sedimentological characterization and stable isotope analyses for 13 basins in the southern CAP reveals important clues about the development of the Anatolian lake system during the late Miocene to Pliocene. Fluvio-lacustrine sedimentation of substantial duration occurred under fairly constant environmental and climatic conditions, during a time of the formation of the Anatolian microplate, surface uplift of the CAP and its northern and southern margins, and an overall change from a compressional to an extensional geodynamic setting.

• (1) Open lake hydrology persisted in the fluvio-lacustrine basins of the Central Anatolian Plateau interior during surface uplift of the plateau. Surface uplift did not result in coeval incision of these basins, as one would expected as a result of an increase in the topographic gradient between the plateau interior and its margins. This suggests that the plateau interior was disconnected from marine basins and occupied by inward-draining lakes.

• (2) Inverse covariance of δ¹³C and δ¹⁸O from eight sections suggest that basin-capping lacustrine carbonates were deposited in open lakes, with the exception of four sections located along the northern flank of the Tauride Mountains, where positive δ¹³C and δ¹⁸O covariance suggests deposition in closed lakes. Data from one section display a hydrological switch from closed to open lake conditions.

• (3) Consequently, the hydrologically open lakes were likely draining into terminal lakes, such as paleolake Kangal, or various paleolakes occupying the present-day Konya closed catchment.

• (4) The δ¹³C and δ¹⁸O values from 13 Miocene to Pliocene fluvio-lacustrine sections indicate slow changes in environmental conditions during lake basin sedimentation that occurred during surface uplift of the Central Anatolian Plateau.

• (5) The δ¹⁸O data do not yield increasingly high values with time, which indicates the absence of major aridification during the early phases of surface uplift of the Central Anatolian Plateau and its margins. This may have been caused by prominent westerly storm tracks permitting moisture transport onto the evolving plateau. Minor aridification may also be obscured in our data set by an overall wet subtropical climate with adequate moisture availability and a reduced seasonality.

• (6) The onset of river incision roughly coincides with a switch from contractional to extensional deformation in central Anatolia. Compression may have led to the creation of new, tectonically induced relief across the subdued topography of the plateau interior, thereby generating a new partitioning of the depocenters, which modified the water balance and promoted drainage integration across the eastern Central Anatolian Plateau.

• (7) The last continental sedimentation was of lacustrine nature (i.e., carbonates, marls) in most of the sampled basins. Given that final sedimentation occurred at different times, it is unlikely that climatic drivers were responsible for final lacustrine sedimentation.

The authors acknowledge support through the U.S. National Science Foundation Continental Dynamics (CD) program (EAR-1109762, “Central Anatolian Tectonics [CD-CAT]” to Whitney); Côme Lefebvre, Tamás Mikes, Ahmet Peynircioğlu, Seçkin Şiş, Levent Tosun, and Jane Willenbring are thanked for field assistance; and Jens Fiebig is thanked for laboratory support at the joint Goethe University–Senckenberg BiK-F Stable Isotope Facility. We would like to thank Faruk Ocakoğlu and Joel Saylor for constructive and helpful reviews.