Middle Miocene climate evolution in the northern Mediterranean region (Digne–Valensole basin, SE France)

During the Middle Miocene, the Earth shifted from a warm state, the Miocene Climatic Optimum (MCO; 16.9–14.7 Ma), to a colder state associated with the formation of extensive and permanent ice sheets on Antarctica. This climatic shift, the Middle Miocene Climatic Transition (MMCT; 14.7–13.8 Ma) strongly affected the composition and structure of major biomes, ocean circulation, and precipitation patterns. Although Middle Miocene climate dynamics are well documented in marine records, our knowledge of terrestrial climate change is not well constrained. Here we present a long-term (23–13 Ma) stable (δ¹³C, δ¹⁸O) and clumped (Δ₄₇) isotope record of soil carbonates from a northern Mediterranean Alpine foreland basin: the Digne–Valensole basin (DVB), France. Δ₄₇-derived soil carbonate formation temperatures indicate a highly dynamic dry season temperature pattern that is consistent with multiple periods of reorganization of atmospheric circulation during the MCO. We propose that changes in atmospheric circulation patterns modified the seasonality of precipitation and, ultimately, the timing of pedogenic carbonate formation. Consequently, Δ₄₇ soil carbonate temperature data record the combined effects of long-term regional temperature and carbonate formation seasonality change. The data are consistent with the existence of a proto-Mediterranean climate already during certain MCO time intervals. Following the MMCT, the stable and clumped isotope record displays pronounced cooling after 13.8 Ma accompanied by a rather large (−5.0 %) decrease in soil water δ¹⁸O values. Our northern Mediterranean foreland basin climate record shares strong similarities with time-equivalent records from the terrestrial European mid-latitudes and the global oceans and enhances our understanding of the circum-Alpine Middle Miocene terrestrial climate dynamics.

1 Introduction

The Miocene Climatic Optimum (MCO; 16.9–14.7 Ma) represents a warm but dynamic climatic interval and was followed by a transition to a colder state, the Middle Miocene Climatic Transition (MMCT; 14.7–13.8 Ma). The MMCT was associated with the inception of major, permanent Antarctic ice sheets (e.g., Flower and Kennett, 1994; Holbourn et al., 2005) and is considered to prelude one of the major cooling steps in Earth's Cenozoic climate history (Flower and Kennett, 1994; Holbourn et al., 2007). Global cooling had profound impacts on the distribution of major biomes (Jiménez-Moreno and Suc, 2007), global ocean circulation and ice sheet dynamics (e.g., Holbourn et al., 2014; Zachos et al., 2001), and continental temperature and precipitation patterns (e.g., Methner et al., 2020; Pound et al., 2012).

The MCO corresponds to one of the warmest intervals of the Neogene period (Flower and Kennett, 1994; Holbourn et al., 2022; Zachos et al., 2001). With higher global temperatures and moderately higher atmospheric CO₂ concentrations ($\sim \mathrm{400}-\mathrm{600}\mathrm{ppm}$; Steinthorsdottir et al., 2021a) than today, it bears characteristics of future global warming scenarios (Longman et al., 2022; Steinthorsdottir et al., 2021b). Refining the understanding of climate variability during the MCO may, therefore, underpin projections of anthropogenic climate change, particularly for the (circum-)Mediterranean region, which has been suggested to become highly vulnerable to intensified future droughts (e.g., Lionello, 2012).

Despite the intense study of Earth's Miocene climate history, our knowledge of Middle Miocene continental climate dynamics remains fragmentary. Efforts in reconstructing European continental climate conditions during the Middle Miocene focused on central (e.g., Jiménez-Moreno et al., 2008; Methner et al., 2020; Mosbrugger et al., 2005) or southern Europe (e.g., Abels et al., 2009; Jiménez-Moreno and Suc, 2007) and highlight possible latitudinal differences in (hydro-)climate (Fauquette et al., 2007; Jiménez-Moreno and Suc, 2007).

Well-dated paleosol sequences provide excellent archives to study long-term continental (hydro-)climatic changes (e.g., Quade et al., 2007), soil formation (Sheldon and Tabor, 2009a), and soil water characteristics within soil profiles (Zamanian et al., 2021). They also allow us to estimate climate parameters such as mean annual precipitation (MAP), mean annual temperature (MAT) (Gillot et al., 2022; Retallack, 2005), and soil temperatures during pedogenic carbonate formation through clumped isotope thermometry (e.g., Kelson et al., 2018; Meijer et al., 2024; Methner et al., 2020). Clumped isotope (Δ₄₇) thermometry allows us to reconstruct the formation temperature of (pedogenic) carbonate, independent of the oxygen isotopic composition of the parental fluid (Eiler, 2011). By simultaneously obtaining carbon (δ¹³C) and oxygen (δ¹⁸O) isotopic compositions in addition to Δ₄₇ values of pedogenic carbonates, it is possible to constrain the oxygen isotopic composition of soil water (δ¹⁸O_w) assuming equilibrium fractionation (Kim and O'Neil, 1997). This approach was successfully applied to paleosols to document long-term continental temperature change, temperature seasonality, and changes in moisture sources (e.g., Kelson et al., 2018; Meijer et al., 2024; Methner et al., 2020; Rugenstein et al., 2022). Embedded within a solid geochronological framework, such paleosol records in terrestrial basins may hence provide continental paleoclimatic information with a temporal resolution that permits direct comparison with marine proxy records (e.g., Methner et al., 2020).

Here, we present δ¹⁸O, δ¹³C, and Δ₄₇ data from pedogenic carbonate nodules of the Digne–Valensole basin (DVB; France, northern Mediterranean region) from ca. 23–13 Ma. Today, the DVB is located at an intermediate latitude between the well-studied Middle Miocene localities in southern (Madrid and Calatayud–Daroca basins, Spain; e.g., Abels et al., 2009) and central (e.g., North Alpine Foreland Basin (NAFB), Switzerland; Kälin and Kempf, 2009; Methner et al., 2020) Europe.

2 Geological and climatic settings

2.1 Geology of the Digne–Valensole basin

At the intersection of Mediterranean, semi-continental, and mountainous climates, the modern DVB – as part of the western Alpine foreland basin – is characterized by an altered Mediterranean climate (“climat méditerranéen altéré”; Joly et al., 2010). The latter differs from the typical Mediterranean climate by a higher frequency of “cold” days and a greater annual precipitation amount (800–950 mm) distributed non-uniformly across the year (Joly et al., 2010).

Mediterranean climate is defined as semi-arid, with dry and hot summers that contrast mild and wet winters. During the winter months, cool and wet air masses from the North Atlantic Ocean deliver moisture to the western Alpine foreland, including the DVB (e.g., Seager et al., 2019). During the summer months, a shift in prevailing wind patterns leads to progressively drier conditions. Moisture from the Mediterranean Sea is episodically transported northward to the DVB region by southerly–southwesterly winds. The Eocene to Pliocene Digne–Valensole basin represents a deformed segment of the Alpine foreland basin system that extends along a narrow SW–NE-trending belt west of the Western Alps (Fig. 1a and b). The DVB molasse currently covers ca. 1500 km² (Fig. 1a) and is bordered by the Digne thrust front in the north and east and the Durance strike-slip fault in the west (Fig. 1b; Ford et al., 1999).

Figure 1 (a) Map showing southern central Europe. The location of the Digne–Valensole basin (DVB) in SE France is indicated by the red rectangle. (b) Geological map of the DVB and study area (red rectangle) at the margin of the Digne thrust front (after Gillot et al., 2022). (c) Locations of the Les Granges, Les Ruines, and Beynes sections within the DVB (after Cojan and Gillot, 2022).

Combined advancement of the Alpine orogenic front and variations in sea level (Bauer, 2006; Bialkowski, 2002; Cojan et al., 2013) affected the Late Eocene to Pliocene basin fill. The overall ca. 3000 m thick Tertiary basin succession (the “Digne Molasse”; Cojan et al., 2013, and references therein; Gigot et al., 1974) is composed of alternating continental to shallow marine deposits with lateral variations in facies and stratigraphic thickness and internal unconformities (Gigot et al., 1974). The Digne Molasse (Gigot et al., 1974) is divided into four formations, which include, from base to top, (1) the Chattian Molasse Rouge (thickness of up to 500 m), which is characterized by alluvial and fluvial facies; (2) the Chattian–Burdigalian Molasse Grise or Molasse Intermédiaire (thickness of several hundreds of meters), which consists of deltaic and distal floodplain deposits; (3) the Burdigalian–Tortonian Molasse Marine (thickness up to 1600 m), dominated by tidal and coastal deposits and characterized by several transgression–regression cycles; and (4) the Miocene–Pliocene Molasse Jaune (thickness up to 300 m), made up of coarse-grained alluvial sedimentary rocks. Most of the material in this study originates from soil horizons within the Molasse Grise and Molasse Marine formations, along with a few samples from the Molasse Rouge formation.

2.2 Middle Miocene climate variability

The Miocene (23.03–5.33 Ma) period is characterized by major environmental change, including cool periods associated with the glaciation of Antarctica and the Northern Hemisphere that were interrupted by warm events, such as the Miocene Climatic Optimum (MCO; 16.9–14.7 Ma; Holbourn et al., 2014, 2015; Jiménez-Moreno et al., 2005; Mudelsee et al., 2014). Well-dated oceanic records provide numerous records of Miocene climate (e.g., Flower and Kennett, 1994; Holbourn et al., 2014, 2015; Mudelsee et al., 2014; Westerhold et al., 2020; Zachos et al., 2001). Following protracted Cenozoic cooling, a stepwise trend of increasing temperatures accompanied by short-term cooling events is documented from megafloras for the Early Miocene in central Europe (Early Aquitanian; Mosbrugger et al., 2005). The MCO reflects the warmest episode of the Neogene with minimum global ice volume (Holbourn et al., 2007, 2014, 2015; Jiménez-Moreno and Suc, 2007; Mudelsee et al., 2014; Steinthorsdottir et al., 2021b). The drivers of MCO warming and its geographic and temporal variability during the MCO are still debated. The onset of the MCO coincides with the 17–16 Ma Columbia River Basalt eruptions in western North America (Longman et al., 2022; Steinthorsdottir et al., 2021b). Combined with increased fluxes of CO₂ in the atmosphere caused by volcanic arc or rift activity and decreased continental weatherability, the Columbia River Basalt eruptions may have triggered the transition into the warm MCO (Longman et al., 2022, and references therein). Although the MCO was generally an interval of intense warmth, high-resolution benthic foraminiferal δ¹⁸O records also indicate high-amplitude climate variability, particularly towards its end (15.7–14.7 Ma; Holbourn et al., 2007, 2014, 2022).

While the mid-latitude N–S temperature gradient in western Europe was probably reduced during the Middle Miocene (Jiménez-Moreno and Suc, 2007), continental paleoclimate reconstructions of the MCO emphasize a strong semi-arid to humid latitudinal gradient with annual precipitation amounts lower than today (i.e., herpetological data from Böhme, 2004; mammal diversity from Costeur and Legendre, 2008; pollen records from Jiménez-Moreno and Suc, 2007). Southern mid-latitude locations, such as southern central Spain, likely experienced semi-arid climatic conditions with open environments (subdesertic conditions; Urban et al., 2010), whereas more northern regions, e.g., central eastern France and Switzerland, were characterized by a more humid climate (1000–1500 mm yr⁻¹; Böhme, 2004) and pronounced precipitation seasonality (Botsyun et al., 2022; Costeur and Legendre, 2008, and references therein). The existence of a Middle Miocene latitudinal aridity gradient stretching from southern to northern mid-latitudinal Europe may have resulted from a trade wind system, which was characterized by prevalent northeasterly winds. Orography may have contributed to the zonal precipitation pattern: trade winds produced summer precipitation upwind and dryness downwind of the east–west-oriented European orogens (Pyrenees and Alps; Böhme, 2004). The demarcation between westerlies and trade winds was likely positioned north of 47° N paleolatitude (Böhme, 2004). Paleoclimate modeling suggests this pattern had ceased by the Late Miocene with a shift in European atmospheric circulation towards westerly dominated winds (Quan et al., 2014).

After the MCO, significant cooling led to the renewed expansion of the Antarctic ice sheet and deep-water cooling during the MMCT (14.7–13.8 Ma; Holbourn et al., 2014; Miller et al., 1991). The latter had a profound impact on marine and terrestrial realms, e.g., the global reorganization of ocean circulation patterns (Holbourn et al., 2022), the increase in foraminifera taxa, the decline in crustacean diversity (Steinthorsdottir et al., 2021b, and references therein), the gradual disappearance of subdesertic plants in southwestern Europe (Jiménez-Moreno and Suc, 2007), and the extinction of central European alligators and giant turtles (Böhme, 2004). The MMCT also impacted terrestrial temperature and precipitation patterns with enhanced temperature seasonality and a shift towards a westerly dominated wind regime over central Europe by the Tortonian (Methner et al., 2020; Quan et al., 2014).

2.3 Present-day Mediterranean climate

The main characteristics of the present-day climate in the circum-Mediterranean region – a marked seasonality between mild, maritime, wet winters and warm to hot, dry summers (Lionello, 2012; Seager et al., 2019) – have been recorded as early as the Pliocene (Suc, 1984). The timing of the onset of the characteristic marked seasonality is, however, still unclear. Typically, North Atlantic storm tracks deliver precipitation to the region during the winter months. Below-freezing temperatures rarely occur during these months, except in mountainous areas. During the summer season, mid-latitude storm tracks weaken due to the dominance of the Azores High pressure system, resulting in warm to hot and dry summers (Seager et al., 2019; Ulbrich et al., 2012).

Being influenced by both a mid-latitude climate variability steered by the North Atlantic Oscillation (NAO) and subtropical conditions due to the descending branch of the Hadley cell, the overall temperature and precipitation patterns in the circum-Mediterranean region reflect the interaction of many individual drivers, including orography, latitude, and land–sea distribution (Ulbrich et al., 2012).

3 Methodology

3.1 Pedogenic carbonate horizons and sampling strategy

Stratigraphic correlations between the investigated sections permit the construction of a ca. 500 m long composite section (Châteauredon composite section; BCR) that consists of three correlated sections: Beynes (BE), Ruines (RU), and Les Granges (GR; Fig. 1c; Bialkowski et al., 2006; Cojan et al., 2013; Gillot et al., 2022). Based on bio- and δ¹³C chemostratigraphy, the BCR covers the Early to Middle Miocene from ca. 23 to ca. 13 Ma (Bialkowski et al., 2006; Cojan et al., 2013). Due to its proximity to the Miocene shoreline, the stratigraphy of the BCR consists of alluvial, lacustrine, fluvial, marine, and coastal facies with intercalations of continental deposits including numerous paleosol horizons. More than 90 % of the paleosol horizons can be classified as calcisols with irregularly shaped and globular, hard micritic pedogenic carbonate nodules dispersed in a fine-grained mudstone, shale, or siltstone matrix (Cojan and Gillot, 2022). The pedogenic carbonate nodules formed on floodplain alluvium and are found in isolated accumulation (Bk) horizons. The nodules range from a few millimeters to several centimeters in diameter and are commonly spheroidal with irregularities (see Fig. 2c).

Figure 2 Field photographs. (a) Beynes section: paleosols that developed in distal alluvial fan deposits (channelized conglomerates; modified after Cojan and Gillot, 2022). The white arrow indicates the younging direction of stratigraphy orientation. (b) Les Granges section: stacked horizons of pedogenic carbonate-rich paleosols that contain vertical rhizoconcretions and are overlain by lacustrine limestone. Hammer for scale. (c) Pedogenic carbonate nodules from the Les Granges section.

In this study, 16 pedogenic carbonate samples from different Bk horizons within the BCR were analyzed for δ¹³C, δ¹⁸O, and Δ₄₇ paleothermometry. Individual pedogenic carbonate nodules are from Bk horizons of carbonate-rich mature paleosols and correspond to the sampling horizons of Bialkowski (2002) and complement the existing δ¹³C record (Bialkowski et al., 2006; Figs. 2 and S1 in the Supplement). Nodules were sampled 50–100 cm below the preserved soil top, which at times corresponds to a truncation surface. Therefore, the formation depth of the carbonate nodules was at least 50 cm (Bialkowski, 2002). Two pedogenic carbonate nodules come from the fine-grained alluvial succession of the Beynes section, which is dominated by pink to reddish marls of the Molasse Rouge. A total of 14 pedogenic carbonate nodules originate from the transgression–regression cycles of the Molasse Grise and the Marine Formation of the Ruines and Granges sections (Table 1 and Fig. S3 in the Supplement).

Table 1 Soil carbonate δ¹⁸O_carb, δ¹³C_carb, and T(Δ₄₇) values and calculated oxygen isotopic ratios of soil water δ¹⁸O_w.

3.2 Age model

The initial age model for the BCR composite section relies on integrated biostratigraphy and chemostratigraphy (δ¹³C values of pedogenic carbonate; Bialkowski et al., 2006; Cojan et al., 2013). The age model is based on the correlation of trends in pedogenic carbonate δ¹³C values (n=76) with trends in marine reference δ¹³C records (compilation of ODP–DSDP sites; Bialkowski et al., 2006, and references therein). Each correlated time interval of the BCR section was validated by biostratigraphy (micromammal faunas and gastropod fossils) and palynological analysis (dinocysts, spores, and pollen, including a distinct Avicennia peak; Châteauneuf et al., 2006). The initial age model adopted the timescale of Berggren et al. (1995).

Here, we adopt and update the initial age model by tying the δ¹³C curve of the BCR composite section to the more recent benthic deep-sea foraminifera carbon isotope curve from Westerhold et al. (2020) by using the QAnalySeries software (Kotov and Paelike, 2018). We use the same tie points (e.g., positive or negative δ¹³C excursions) between the continental and reference marine records as Bialkowski et al. (2006) and Cojan et al. (2013). For intervals with limited tie points and matching trends, the correlation was based on the best fit provided by the software assuming uniform sedimentation rates and ensuring similar trends with the reference marine δ¹³C curve (Fig. S4; Tables S2 and S3 in the Supplement).

Age uncertainties in the DVB range from 100 ka to 1 Ma (Fig. 3; Bialkowski, 2002). The age uncertainties in the DVB are constrained by the correlation of the DVB biostratigraphy with the biostratigraphy of the well-correlated magnetostratigraphic NAFB and the well-described biostratigraphy (MN zones) of the Aquitaine basin (Bialkowski, 2002). Given that individual carbonate nodules in the paleosols may form over hundreds to thousands of years (Gile et al., 1966; Kelson et al., 2020), the uncertainty in the age model is larger than the time integrated by the individual carbonate nodules.

Figure 3 Soil carbonate δ¹⁸O, δ¹³C (Bialkowski, 2002), and T(Δ₄₇) (this study) values. Oxygen isotopic ratios of soil water (δ¹⁸O_w) were calculated with δ¹⁸O_carb measured in this study; the timing of intervals I–IV is indicated.

3.3 Carbonate stable isotope analysis and Δ₄₇ thermometry

Clumped isotope thermometry (Δ₄₇) relies on the determination of the temperature-dependent excess abundance of rare carbon and oxygen isotope bonds (¹³C–¹⁸O), the so-called clumping, within the carbonate crystal lattice. The isotopic clumping of carbonate molecules is dependent on the temperature of carbonate formation but independent of the δ¹⁸O value of the fluid from which the carbonate precipitated (Ghosh et al., 2006; Eiler, 2011).

Isotope analyses were performed at the Goethe University–Senckenberg BiK-F Joint Stable Isotope Facility (Frankfurt, Germany). Measurements were performed on a Kiel IV Carbonate Device coupled to an MAT 253 Plus IRMS. For a detailed method description, refer to the Supplement. Prior to measurements, carbonate nodules were cut and subsequently powdered with a dental drill.

Samples were analyzed in a variable number of replicates ($n=\mathrm{8}-\mathrm{18}$) using between 90 and 190 µg of material according to carbonate content. Samples were analyzed together with the ETH-1–ETH-4 carbonate standards (Bernasconi et al., 2018). Raw intensity data were corrected for the negative pressure baseline using the $m/z_{\mathrm{47.5}}$ signal and scaling factor method (Bernecker et al., 2023; Fiebig et al., 2021) applied to ETH-1 and ETH-2. Standardization was carried out by utilizing the D47crunch module (Daëron, 2021) in pooled sessions mode with ETH standards as carbonate anchors (Δ_{47 (I-CDES)}, Bernasconi et al., 2021). The universal temperature calibration of Anderson et al. (2021) is used for temperature conversion.

Stable isotope results of pedogenic carbonates are normalized relative to nominal values of ETH standards (Bernasconi et al., 2018) and reported in per mille deviation relative to V-PDB (δ¹³C_carb, δ¹⁸O_carb). δ¹⁸O_w was calculated based on our obtained carbonate formation temperature (T(Δ₄₇)) and δ¹⁸O_carb data and on the oxygen isotope equilibrium fractionation equation of Kim and O'Neil (1997). δ¹³C_carb and δ¹⁸O_carb errors are reported as 1 standard deviation ( ±1 SD; see tables in the Supplement), whereas Δ₄₇ errors are reported as 1 standard deviation, fully propagated 1 standard error (1 SE), and fully propagated 2 standard error (2 SE; Tables 1 and S1 in the Supplement).

4 Results

Complete results of stable and clumped isotope measurements of the pedogenic carbonate nodules are presented in Table S1. The analyses of pedogenic carbonate nodules of the same horizon (see Methodology) complement data presented in Bialkowski (2002) (Table 1). The following sections discuss the results from analyses carried out in this study.

4.1 Stable (δ¹⁸O_carb and δ¹³C_carb) and clumped (Δ₄₇) isotope data

δ¹⁸O_carb values of paleosol carbonate samples from the Digne–Valensole basin range from −7.02 ‰ to −4.66 ‰ (V-PDB), and δ¹³C_carb values range from −9.90 ‰ to −6.15 ‰ (V-PDB). The Δ_{47 (I-CDES)} values of pedogenic carbonate range from 0.565 ‰ to 0.619 ‰, which translates to a temperature range of 35.2–16.8 °C. External standard errors (1 SE) for 8–18 replicate measurements range from 0.01 ‰ to 0.02 ‰, which corresponds to temperature uncertainties covering 3.5–6.0 °C. Calculated soil water δ¹⁸O_w values based on corresponding δ¹⁸O_carb and Δ₄₇ values range from −6.29 ‰ to −0.51 ‰ (V-SMOW).

Due to potential diagenetic alteration, measurements from one sample (98GR22) are not included in the interpretation of the results and the Discussion (Table S1 and Fig. S2 in the Supplement). This sample yields a clumped isotope temperature (T(Δ₄₇)=39.2 °C), which is slightly higher than plausible on the Earth's surface (see Li et al., 2024). Its δ¹⁸O_carb and δ¹³C_carb values fall within the range of δ¹⁸O_carb and δ¹³C_carb values observed in the other micritic nodules of this section. However, as a note of caution, we rejected this sample from further interpretation, even though the inclusion of the sample in our analysis would not change our interpretation and conclusions.

4.2 Paleosol temperature reconstruction

We report stable and clumped isotope data of our DVB record within four intervals, which closely follow Miocene stages (Fig. 3). Interval I (Aquitanian, 22.3–21.1 Ma) yields relatively similar T(Δ₄₇) values with an average temperature of 27.6 °C (n=2; SD=1.6 °C), while data from Interval II (Burdigalian, 19.6–16.7 Ma) show positive trends in δ¹³C_carb (from ca. −7.4 ‰ to −6.9 ‰) and δ¹⁸O_carb values (from ca. −6.6 ‰ to −6.0 ‰) and an average Δ₄₇ temperature of 32.0 °C (n=4; SD=3.5 °C; Fig. S3). Interval III (Langhian–Early Serravallian, 16.0–13.8 Ma) is characterized by high-amplitude and rapid temperature variations ranging from 20.6–34.7 °C (n=7; SD=5.3 °C; Fig. 3), with corresponding changes in the δ¹⁸O_w record. Overall, δ¹³C_carb and δ¹⁸O_carb values follow a positive trend from 18.6–14.5 Ma (Fig. S3).

Interval IV (Serravallian, 13.4–13.2 Ma) is characterized by the two lowest Δ₄₇ temperatures in our record of 20.5 and 16.8 °C (13.4 and 13.2 Ma, respectively; average 18.6 °C; n=2; SD=2.6 °C).

In conclusion, we observe the following patterns in our Miocene record. The Aquitanian data reflect similar temperatures averaging 27.6 °C. During the Burdigalian and the Langhian, we observe three key elements in our data: (a) Δ₄₇ temperatures decrease from ca. 33 °C (18.3 Ma) to 20 °C (15.0 Ma), (b) Δ₄₇ temperatures increase to ca. 34 °C at the end of the Langhian and the start of the Serravallian, and (c) Δ₄₇ temperatures decrease at 13.8 Ma to the lowest measured Δ₄₇ temperatures. The rapid large magnitude decrease in Δ₄₇ temperature from 34.3–16.8 °C between 13.8 and 13.2 Ma is accompanied by a 1.5 ‰ decrease in δ¹⁸O_carb values from −5.5 ‰ to −7.0 ‰, which corresponds to a decrease in δ¹⁸O_w values of 5.1 ‰ from −1.0 ‰ to −6.1 ‰.

5 Discussion

5.1 Pristine stable isotope values

We consider the obtained Δ₄₇ temperatures pristine and to reflect pedogenic carbonate growth temperatures for the following reasons: (1) our Δ₄₇ temperatures align with the range of temperatures found in the upper layers of Earth's subsurface, which are mainly controlled by the MAT (Molnar, 2022; Mosbrugger et al., 2005); (2) within the reported samples there is no evidence of (localized) fluid flow and associated recrystallization (Quade et al., 2013, and references therein), which is attested by thin sections of nodules that reveal a dominantly micritic texture (Fig. S1); (3) the δ¹⁸O_carb values fall into the range attributed to unaltered pedogenic carbonates (Tabor and Myers, 2015); and (4) the burial depth of the sampled paleosol sequences in the DVB was relatively shallow (<500 m; Ford et al., 1999), and there is no overall trend to higher temperatures within the more deeply buried parts of the section.

However, one pedogenic carbonate sample (99GR22) contains sparry calcite veins. This sample may have been affected by diagenetic alteration, and its relatively high Δ₄₇ temperature (39.2 °C; Table S1; Fig. S2) may not reflect pedogenic carbonate formation temperatures but rather carbonate formation during diagenesis. Consequently, the sample was not included in the Discussion.

5.2 Seasonality of soil carbonate precipitation

We observe large (up to 17.5 °C) variations in T(Δ₄₇) soil carbonate formation temperatures in the DVB record. Owing to the fact that the sampling depth of carbonate nodules does not exceed 1 m below the preserved surface of the paleosol, we rule out the possibility that changes in depth associated with fluctuations in precipitation amount significantly affect the measured T(Δ₄₇) (Kelson et al., 2020; Molnar, 2022). Such large temperature variations cannot be explained by regional or global terrestrial temperature change alone (e.g., Fauquette et al., 2015; Mosbrugger et al., 2005; Westerhold et al., 2020).

In situ pedogenic carbonate formation occurs within soil profiles in the presence of soil water in semi-arid to sub-humid climates. Pedogenic carbonates most commonly form during the warm and dry season when plants are active and evapotranspiration attains a maximum (Quade et al., 2007, 2013; Zamanian et al., 2016). As such, carbonate precipitation responds to interannual changes in soil moisture availability and is mainly controlled by the interplay between soil water availability, soil temperature, and soil CO₂ levels in the presence of sufficient Ca²⁺. Net calcite accumulation occurs when (re-)precipitation of calcite exceeds calcite dissolution during seasonal wet and dry cycling (Zamanian et al., 2016). This implies that changes in the timing and/or duration of the pedogenic carbonate growth season will affect Δ₄₇ temperatures. Because soil drying is an important requirement for pedogenic carbonate formation, a change in the timing of the dry season would inevitably result in a shift in the pedogenic carbonate formation period (Fig. 4). Consequently, recorded Δ₄₇ temperatures may vary depending on the timing of rainfall seasonality and hence seasonality of carbonate formation (Fig. 4a and b; Gallagher et al., 2019; Kelson et al., 2020). In accordance with other studies of pedogenic carbonate T(Δ₄₇) from central Europe (e.g., Methner et al., 2020), we suggest that a seasonal change in the timing of annual carbonate precipitation may have played an important role in controlling the observed temperature pattern, i.e., a change in pedogenic carbonate precipitation from the cool(er) to the warm(er) season or vice versa. Presently, summer temperatures average ca. 19 °C and winter temperatures average ca. 2 °C in the DVB (IAEA/WMO, 2024). Soil temperatures are typically higher than air temperatures due to solar radiation heating (Molnar, 2022; Quade et al., 2013). At shallow depths (<30 cm), soil temperature exhibits significant daily fluctuations, but these diurnal variations diminish with depth. Beyond 300 cm, soil temperature becomes nearly constant throughout the year, converging toward the MAT (Molnar, 2022; Quade et al., 2013; Sheldon and Tabor, 2009). A hypothetical instantaneous shift from warm- to cold-season carbonate precipitation or vice versa would therefore presently account for a change of up to 17 °C in soil carbonate formation temperatures.

Figure 4 Present-day climate conditions for three locations in Europe: (a, c) Aix en Provence and Brest (France) and (b) Debrecen (Hungary). Variations in seasonality of precipitation may drive changes in pedogenic carbonate formation temperatures. Depending on the seasonality of pedogenic carbonate formation, soil temperatures reconstructed from pedogenic carbonate should reflect (a) warm (>MAT) or (b) cool (≤MAT) season temperatures. In panel (c), conditions for the formation of pedogenic carbonate are not met because of the lack of contrasting rainfall seasonality and temperature throughout the year; hence no pedogenic carbonate is formed at this location. The δ¹⁸O_w values are retrieved from the GNIP stations near the locations. (d) Influence (cross: large; circle: moderate) of the interplay between soil temperature and soil water deficit on the formation of pedogenic carbonate. Temperature-dominated (a) and water-deficit-dominated (b) scenarios and a scenario with no pedogenic carbonate formation (c).

Based on the above, we suggest that changes in the Δ₄₇ soil carbonate formation temperatures throughout the Middle Miocene are best explained by warming or cooling during the MCO and MMCT, respectively, in combination with modifications of the intra-annual timing of soil carbonate formation.

A change in seasonality of carbonate precipitation explaining the large observed temperature variations in our Δ₄₇ record is supported by the large variations in calculated δ¹⁸O_w during the MCO and MMCT of up to 3.8 ‰ and 5.0 ‰, respectively. These δ¹⁸O_w variations may reflect changes in atmospheric circulation patterns and/or a change in the season of moisture delivery, which would both be associated with different moisture sources. Shifts in the seasonality of pedogenic carbonate formation could potentially be associated with variations in δ¹³C_carb values. We suggest that δ¹³C of pedogenic carbonates would decrease with rising temperatures, primarily due to increased biological activity and CO₂ production. However, isotopic fractionation during carbonate precipitation moderates this relationship, making it a complex interaction between biological and geochemical processes. However, we would expect lower δ¹³C_carb values during time periods associated with cooler temperatures and high precipitation amounts. This agrees with the significantly low δ¹³C_carb values during the MMCT in our record (Fig. 3). Consequently, a change in the seasonal timing of precipitation in the DVB is likely to be linked with a change in moisture source(s) over the region.

A shift in precipitation seasonality over western (this study) and central Europe (Methner et al., 2020) throughout the Middle Miocene requires a profound reorganization of atmospheric circulation. The distribution of fossil snakehead fish at the start of the MCO suggests moist summer conditions which may be reached by a poleward shift in the Hadley circulation (Böhme, 2004). This may have resulted in a southeastward transport of air masses over the North Sea basin and the Paratethys and along the Alpine chain transporting moist air over central Europe during the summer months (Böhme, 2004), which is supported by tree fall directions and the eccentricity of tree growth rings (Utescher and Bruch, 2000). After the MCO, a permanent shift from a northeasterly to a westerly dominated wind regime (similar to the present-day Mediterranean-type climate) over the northern Mediterranean region is suggested to have occurred at the latest before the Tortonian (ca. 12 Ma; Quan et al., 2014).

The data presented here do not permit the identification of drivers and mechanisms of changes in atmospheric circulation, but we consider it likely that an alternation between northeasterly winds and westerlies or latitudinal variability in the main westerlies track throughout the MCO contributed to modifying the pattern of wet and dry seasons and hence the temperature fluctuations observed in our Δ₄₇ soil carbonate data.

We suggest that the DVB MCO and MMCT record reveals two distinctly different climatic patterns. During the Langhian, the climate is marked by a dry, cool season. Soil-drying and pedogenic carbonate formation during late fall and/or early winter (Fig. 4b) are reflected in relatively low T(Δ₄₇) and δ¹⁸O_w values. In this “dry winter” scenario, rainfall occurred mostly during warmer (summer) months and was possibly delivered by moist air masses from the North Atlantic Ocean. We hypothesize that a transition to a climate similar to the present-day Mediterranean-type climate (“proto-Mediterranean climate”), characterized by hot and dry summers (Fig. 4a), occurred during the Middle Miocene (at ca. 16.5 and 14.2–13.8 Ma; Fig. 5). This implies that, already during certain time intervals of the MCO, a climate similar to the present-day Mediterranean climate existed. In such a climate state, the drying of soils during the summer season led to dominant pedogenic carbonate formation in early summer, with associated high T(Δ₄₇) and δ¹⁸O_w values. Finally, our data are consistent with a return to the “dry winter” scenario in the early Serravallian (13.4–13.2 Ma), resulting in relatively low Δ₄₇ temperatures and δ¹⁸O_w values. We therefore propose that high T(Δ₄₇) (>30 °C) and high δ¹⁸O_w values ($>-\mathrm{2}\mathrm{‰}$) during the MCO and MMCT resulted from a Mediterranean moisture source (“proto-Mediterranean climate”), whereas lower Δ₄₇ temperatures and lower δ¹⁸O_w values ($<-\mathrm{2}\mathrm{‰}$) reflect predominance of a North Atlantic moisture source.

5.3 Middle Miocene northern Mediterranean and central European terrestrial paleoclimate

Irrespective of the change in the seasonality of the soil carbonate formation, three main elements characterize our paleosol temperature data and include (1) the transition into the MCO (17.3–16.7 Ma) during the Burdigalian, which is reflected by an increase in Δ₄₇ temperatures and the highest temperature of the record (35.2 °C); (2) overall cooling from the onset to the end of the MCO; (3) a large temperature increase during the MMCT; and (4) a distinct (>15 °C) and rapid (ca. 600 ka) cooling phase accompanied by a significant decrease in δ¹⁸O_w (−5.0 ‰) after ca. 13.8 Ma, which coincides with global post-MMCT cooling.

In the following, we compare the DVB record with time-equivalent terrestrial data from the northern Mediterranean region and central Europe (Fig. 5).

Figure 5 Reconstructed Miocene mid-latitude terrestrial temperatures from western and central Europe. Temperature reconstructions based on carbonate clumped isotope thermometry from the Digne–Valensole basin are compared to T(Δ₄₇) data from the NAFB (Methner et al., 2020), mid-latitude European mean annual temperatures (MATs) and warm-month temperatures (WMTs) inferred from terrestrial paleofloral data (Fauquette et al., 2015; Mosbrugger et al., 2005) and herpetofauna and climofunction data from paleosols (Ivanov and Böhme, 2011; Gillot et al., 2022), and modeled MAT and WMT values (Botsyun et al., 2022) for the paleogeographic position of the DVB. Timing for the MCO and MMCT is indicated, as are the two intervals of proto-Mediterranean climate in the DVB region.

During the Aquitanian (22.3–21.1 Ma) and Middle Burdigalian (18.6–17.3 Ma), T(Δ₄₇) values of the DVB region are significantly (up to 10 °C) higher than estimated mean annual temperatures (MATs) for the DVB derived from pollen (Fauquette et al., 2015) and paleosol chemistry (Gillot et al., 2022). During the same time interval, our T(Δ₄₇) values are also up to 10 °C higher than MAT reconstructed from central European megafloras (fruits, seeds, and leaves) but in good agreement with estimated warm-month temperatures (WMTs) (Mosbrugger et al., 2005; Fig. 5).

At the onset of the MCO, our DVB T(Δ₄₇) record shows a warming trend similar to that observed in the T(Δ₄₇) record from the Swiss NAFB paleosols (Methner et al., 2020).

At ca. 15 Ma, we observe colder Δ₄₇ temperatures that align particularly well with MAT values derived from herpetological data (Ivanov and Böhme, 2011), pollen records (Fauquette et al., 2015), and climate models (Botsyun et al., 2022). This T(Δ₄₇) value in the DVB suggests a switch from a dry-season summer bias towards conditions indicating a “dry winter” scenario with late fall/early winter soil carbonate formation (see Sect. 5.2). Conversely, during the subsequent warming trend leading into the onset of the MMCT (ca. 15.5–14.0 Ma), we observe T(Δ₄₇) values that converge towards temperatures above the paleobotanical WMT values and hence a return to dry summers and warm-season-biased temperatures .

Finally, we attribute the large-magnitude temperature decrease after ca. 14 Ma to post-MMCT continental cooling. The climate change after 14 Ma is expressed in an impoverishment in plant diversity in western Europe (Jiménez-Moreno and Suc, 2007) and led to the regional extinction of temperature-sensitive tetrapods in central Europe, such as alligators, giant turtles, and chameleons (Ivanov and Böhme, 2011). The temperatures in the DVB at 13.5 Ma, following the post-MMCT cooling, are equivalent to the NAFB Δ₄₇ soil temperatures and herpetological, botanical, and model-based MAT (Fig. 5; Botsyun et al., 2022; Fauquette et al., 2015; Ivanov and Böhme, 2011; Methner et al., 2020). We can only attempt to decipher the contributions of seasonal temperature change and overall regional cooling to the ca. 17.5 °C decrease in soil carbonate formation temperatures visible in the DVB Δ₄₇ record. Assuming that the temperature change of ca. 14 °C (20.6–34.7 °C) that is observed between 14.9 and 14.2 Ma reflects only the change in carbonate formation season without any major regional temperature change, we posit any fluctuations in carbonate formation temperatures larger than 14±2 °C must be the result of additional global warming or cooling. We therefore propose that the remaining 3–4 °C temperature decrease after 13.8 Ma reflects the effect of regional post-MMCT cooling. This value is comparable to the drop in the minimal summer temperature of>7 °C between 14.0 and 13.5 Ma described in herpetological data for central Europe (Böhme, 2003). While the analytical error (1 SE) for the clumped isotope temperatures is on the order of a few degrees, the concurrent major decreases in δ¹³C_carb, δ¹⁸O_carb, and δ¹⁸O_w values within a relatively short timescale (<600 ka) suggest considerable climatic change. Therefore, the significant decrease (17.5 °C) in Δ₄₇ temperatures and associated decrease in δ¹³C_carb, δ¹⁸O_carb, and calculated δ¹⁸O_w values at or after 13.9 Ma support the interpretation that regional cooling (3–4 °C) superimposed on a component of changing climate seasonality is accountable for the decrease in T(Δ₄₇) (Methner et al., 2020).

5.4 Continental soil temperatures related to marine records

Global climate change would not only lead to a decrease (or increase) in marine and continental temperatures, but may also significantly impact atmospheric circulation, resulting in altered weather patterns and ocean–atmosphere interactions (Holbourn et al., 2005, 2007; Methner et al., 2020) similar to changes observed during anthropogenic warming. Changes in temperature gradients can weaken or shift trade winds, disrupting weather systems and ocean currents (Comas-Bru et al., 2016). Throughout the MCO and MMCT, our northern Mediterranean terrestrial soil Δ₄₇ temperatures show significant similarities with marine temperature records (Fig. 6). Because the rapid and large changes in measured soil carbonate T(Δ₄₇) during the Middle Miocene are too large to be explained by regional/global cooling (or warming) alone, we suggest that the changes in soil Δ₄₇ temperatures result from the combined effect of changes in seasonality of rainfall and regional/global cooling (or warming).

Figure 6 Clumped-isotope-based reconstruction of terrestrial temperatures compared with global marine data, within the timing of the MCO, MMCT (Westerhold et al., 2020), and Mi events (Miller et al., 2020). (a) Δ₄₇ temperatures from the Digne–Valensole basin (in green) and benthic oxygen isotopic curve (Westerhold et al., 2020). (b) TEX₈₆-based sea surface temperatures of ODP Site 608, North Atlantic (Super et al., 2018). (c) Bottom-water temperatures from Cramer et al. (2011) in dark gray and Lear et al. (2015) in light gray.

The MCO is characterized as a period of sustained global warmth marked by high-amplitude climate variability (Fig. 6a) (Holbourn et al., 2007, 2014, 2015; Steinthorsdottir et al., 2021b). Our terrestrial record agrees with global records and indicates overall high temperatures with pronounced temperature variability throughout the MCO.

Following a period of rapid warming and/or polar ice melting at 16.9 Ma at the onset of the MCO, the Middle Miocene was globally marked by fundamental changes in the global carbon reservoir, resulting in δ¹⁸O_benthic and δ¹³C_benthic increases (Fig. 6a) (Holbourn et al., 2015, 2022; Westerhold et al., 2020) and cooling in sea surface (Fig. 6b) and bottom-water temperature (BWT; Fig. 6c) from 16.5 to ca. 16.0 Ma (Cramer et al., 2011; Lear et al., 2015; Super et al., 2018). Those results align well with the cooling phase recorded in our T(Δ₄₇) between circa 16.7 and 15.6 Ma.

Within age uncertainty, our coolest Δ₄₇ temperature within the MCO coincides with δ¹⁸O_benthic excursion Mi3a, which is associated with significant changes in global temperature and ice-volume history (Fig. 6a and c; Holbourn et al., 2007; Miller et al., 2020).

The subsequent T(Δ₄₇) warming trend in the MMCT from 14.9 until 14.2 Ma in our continental record is also observed in benthic records (Fig. 6a and c; Cramer et al., 2011; Westerhold et al., 2020). Finally, our rapid post-MMCT cooling (after ca. 13.8 Ma) is associated with a major decrease in δ¹³C_carb (−4.0 %), δ¹⁸O_carb (−2.5 %), and δ¹⁸O_w (−5.0 %) values and parallels the stabilization of the East Antarctic Ice Sheet, which is expressed in benthic δ¹⁸O excursion Mi3, associated sea level decline, and a decrease in BWT (Holbourn et al., 2015; Miller et al., 2020; Sosdian et al., 2020; Steinthorsdottir et al., 2021b; Westerhold et al., 2020). It also coincides particularly well in time with cooling in continuous marine successions from the eastern Mediterranean region at 13.82±0.03 Ma (Abels et al., 2005).

Results of our study documenting the paleosol Δ₄₇ temperatures of the Digne–Valensole basin suggest the region was affected by the Middle Miocene climate dynamics similarly to the global oceans, in addition to significant and rapid changes in temperature and seasonality as a result of a reorganization in atmospheric circulation.

6 Conclusions

Our Early to Middle Miocene terrestrial T(Δ₄₇), δ¹³C_carb, δ¹⁸O_carb, and δ¹⁸O_w record for the northern Mediterranean Digne–Valensole basin (SE France) shows the following:

1. An abrupt decrease in Δ₄₇ temperature (17.5 °C within 600 ka) occurs at 13.8 Ma, which we attribute to post-MMCT cooling whose timing is in agreement with global marine (Holbourn et al., 2022; Westerhold et al., 2020) and Mediterranean (Abels et al., 2005) records. The decrease in Δ₄₇ temperatures associated with a major decrease in δ¹³C_carb (−4.0 %), δ¹⁸O_carb (−2.5 %), and δ¹⁸O_w (−5.0 %) values within less than 600 ka may be best explained by a large seasonality component and overall cooling on the order of 3–4 °C.

2. During the MCO, DVB Δ₄₇ temperatures are higher than other reconstructed European mid-latitudinal average warm-month temperatures (Ivanov and Böhme, 2011; Mosbrugger et al., 2005).

3. Similarly to the observations from the Swiss NAFB record (Methner et al., 2020), we suggest that variations in Δ₄₇ temperatures record the combined effect of global warming and cooling and changes in the annual distribution of rainfall, which affects the timing of pedogenic carbonate formation. The latter would result from the reorganization of atmospheric circulation in the circum-Alpine region.

4. We suggest that a proto-Mediterranean climate characterized by hot and dry summers already existed during certain time intervals of the MCO.

5. We identify an increase in soil Δ₄₇ temperatures at 15 Ma, which coincides with warming trends observed in global marine data (Cramer et al., 2011; Westerhold et al., 2020).

6. We observe a cooling phase in our Δ₄₇ temperatures between circa 16.7 and 15.6 Ma, which coincides with cooling in global sea surface and bottom-water temperatures (Cramer et al., 2011; Lear et al., 2015; Super et al., 2018).

7. The coolest recorded temperature during the MCO in our record coincides with the δ¹⁸O_benthic excursion Mi3a (14.8 Ma) and associated transient glacial expansion (Miller et al., 2020).