Using clay mineralogy and micropalaeontological observations to unravel Neogene climate variations in Northern Arabia

The Zagros Mountain belt in Iran comprises an extensive record of Mesozoic and Cenozoic deposits uplifted by collision of the Arabian and Iranian continental plates. This area has been shown to provide important information to decipher the climatic history of Western Asia, especially regarding the evolution of deserts in Mesopotamia and Northern Arabia. In a high-resolution geological record of 84 samples, we combine clay mineralogy and micropalaeontological observations to refine the late Neogene climate record in Northern Arabia. We investigated a 2.6-km-thick sedimentary profile from the Simply Folded Zagros Mountain belt spanning ca. 10.2 myrs from the late Middle Miocene (Serravallian) to earliest Pleistocene (Gelasian) comprising floodplain sediments and saline mudstones with a strong aeolian contribution. The clay fraction < 2 µm reveals palygorskite, smectite, illite, chlorite and the zeolite mineral erionite. Using clay mineralogy, arid conditions during the Serravallian and early Tortonian in Northern Arabia are identified, followed by semi-arid conditions during the late Tortonian. Clay mineralogy suggests a transition to sustained arid conditions during the early Messinian and provides a more complex picture on a previously proposed period of hyperaridity (NADX) in Northern Arabia penecontemporaneous with the apex of the Messinian salinity crisis.


Introduction
Clay minerals form a main constituent of soils and result from the weathering of primary minerals such as feldspar or micas (Allen & Hajek 1989). The exact composition of a clay mineral assemblage developing in soils is dependent on a variety of factors such as the composition of parent rocks, weathering conditions (temperature, precipitation, drainage of soils) and geomorphology and relief of the landscape. Furthermore, the clay mineral assemblages can be altered by subsequent transport and diagenetic processes (Allen & Hajek 1989;Chamley 1989;Singer 1980). In palaeoclimatic studies, other proxies such as stable oxygen isotopes, magnetic susceptibility and micropalaeontology should be employed alongside the analysis of the clay mineral assemblage to overcome factors not directly related to climatic parameters (John et al. 2012;Singer 1984b). Using multiproxy approaches, the analysis of clay minerals has been used to reconstruct changes in climate and weathering conditions across a wide range of marine and terrestrial settings worldwide (Bolle et al. 2000;Chamley et al. 1986;Li et al. 2000). In areas once covered by the Neotethys ocean such as southern Iran, the clay mineral fraction of Cenozoic and recent sediments was found to comprise smectite, palygorskite, illite, chlorite and kaolinite (Hashemi et al. 2013). However, detailed high-resolution studies of the chronostratigraphic distribution of climate-sensitive clay minerals such as palygorskite in western Iran and Northern Arabia are missing (Hojati & Khademi 2011

Study aims
This study aims to use clay mineralogy to refine palaeoclimatic trends for the late Neogene in Northern Arabia between 12.6 Ma and 2.4 Ma. New data on clay mineralogy of 84 sediment samples is compared to previously published data of fluctuations in magnetic susceptibility, highly soluble salts and sulphate taken from the same samples by Böhme et al. (2021). While soluble salts can be used to reconstruct palaeoclimate in arid and hyper-arid environments, they can be sensitive to post-depositional leaching (Ewing et al. 2006;Rosenthal et al. 1987). Clay mineralogy will hereby help to independently test the robustness of the previously employed soluble salt-based proxy as samples from the combined geological profile were taken at two different profiles, which vary in recent mean annual precipitation. This way the influence of differing post-depositional leaching of highly soluble salts can be tested. As the ratio of smectite/illite + chlorite was suggested as indicative for soil available moisture in palygorskite bearing recent soils by Hashemi et al. (2013), we will try to establish its utility for reconstructing palaeo-aridity along a sedimentary profile. Comparison of different proxies is expected to increase our understanding of long-term responses of clay mineral associations to complex changes in climatic forcing and detrital parent material composition in semi-to hyper-arid environments.

Previous work
Based on clay mineralogical analysis of rock samples from the Cretaceous to Quaternary, Khormali et al. (2005) reconstructed a long-term warm and arid climate from the Eocene to Pliocene in southern Iran. It is suggested that climate in the Afro-Eurasian domain has been affected by the retreat of the Neotethys ocean since the Late Miocene (Zhang et al. 2014). Using thresholds of soluble nitrate, chlorite and bromide as well as end-member modelling of grain size-distributions along a sedimentary profile, Böhme et al. (2021) suggested transient periods of Arabian hyperaridity at 8. 75, 7.78, 7.5 and 6.25 Ma as well as a sustained hyper-arid period between 5.6 Ma and 3.3 Ma (Neogene Arabian Desert Climax, NADX). While transient periods of hyperaridity and NADX match with significant Ponto Caspian low stands, periods of Ponto Caspian high stands corresponded well to a more humid (semiarid) climate in Mesopotamia (Böhme et al. 2021). NADX was initiated at peak of desiccation of the Mediterranean during the Messinian Salinity Crisis (MSC, stage 2) at 5.59 Ma (Manzi et al. 2013). Hyperaridity correlates with the > 2 myr long separation of the Caspian Sea from the Black Sea basin. Hereby, sea levels of the Mediterranean Sea fell by > 1500 m, furthermore by two steps at 5.6 Ma and 5.38 Ma Caspian Sea level dropped by at least 200 m (Roveri et al. 2014;Ryan 2009;van Baak et al. 2016). A possible mechanism to account for changes in aridity is the influence of Paratethys shrinkage, which shifted climate conditions from temperate to continental in central Asia, by establishing the Siberian Pressure High in winter (Ramstein et al. 1997). This strengthening and expanding Siberian Pressure High blocked the moisture-carrying westerlies from reaching Western Asia and Northern Arabia leading to increased aridity in that region during sea level low stands (Böhme et al. 2021;Perşoiu et al. 2019).

Overview of formation conditions of common clay minerals
Smectites are common in many soils of temperate regions, and they are stable in poorly drained environments where leaching of Si and bases are restricted (Borchardt 1989). Smectites are also an important constituent of soils in semi-arid and arid climate zones such as in aridisols in the USA (Dregne 1976;Wilson 1999), Iraq (AI-Rawi et al. 1969), Saudi Arabia (Aba-Husayn et al. 1980) or Israel, where Singer (1984a presented evidence for neoformation of smectite in saprolites. In southern Iran, Khormali and Abtahi (2003) found most smectite in semi-arid and arid soils to have originated from transformation of palygorskite and illite. The transformation from palygorskite generally occurred at P/ET (ratio of mean annual precipitation to evapotranspiration) > 0.4.
Palygorskite is a fibrous clay mineral that can be formed under semi-arid or arid climate conditions (Singer 1989). Its authigenic formation by chemical precipitation has generally been reported from evaporative basins such as lakes and shallow saline lagoons, soils or in open oceans by hydrothermal alteration of basaltic glass or volcanic sediments (Al-Juboury 2009;Callen 1984;Chamley 1989;Millot 1970;Singer 1979). The conditions of palygorskite formation and stability are characterized by alkaline fluids with high Mg and Si activities (Singer 1980(Singer , 1989. Under these conditions, the neoformation of palygorskite is possible from illite and smectite with a loss of K and Al and a relative increase in Si and Mg (Suárez et al. 1994). While arid conditions favour the formation of palygorskite in soils, it is often associated with pedogenic carbonates in semi-arid regions (Al-Juboury 2009). Palygorskite and gypsum content were also shown to have a significant correlation in saline-alkaline soils in southern Iran (Khormali & Abtahi 2003). Shallow lakes and intra-montane lagoons in the post-Neotethyan era (Oligocene-Miocene) provided a suitable evaporative environment for extensive palygorskite formation in central and southern Iran (Khademi & Mermut 1998;Khormali et al. 2005) and Iraq (Al-Juboury 2009) with the first palygorskite detected in the late Palaeocene.
Illite describes clay-sized micaceous minerals similar to muscovite usually containing more Si, Mg, H 2 O and K than ideal muscovite. Also, illite usually constitutes a mixture of micaceous minerals of different origins (Chamley 1989). Illite in soils is often inherited from parent rocks such as shales, siltstones, limestones, loess and a variety of alluvial sediments or may result from alteration of coarser muscovite particles during pedogenesis (Allen & Hajek 1989). According to Singer (1988), illite may also form pedogenically in the surface horizon of aridic and semi-aridic soils from desert dust rich in K-bearing minerals during wetting-drying cycles. Juvenile desert dust deposits resulting from aridic weathering of bedrocks contains only low to moderate amounts of illite (typically in the Middle East), whereas mature dust originating from aridic soils and loess deposits having undergone repeated cycles of deflation and deposition contain relatively high illite contents (typically in Central Asia). Like illite, chlorite in soils is generally inherited and is markedly unstable in pedogenic environments; it is commonly reported in relatively unaltered substrata (Allen & Hajek 1989).
Kaolinite originates from weathering in near-surface environments. Its formation is most pronounced where weathering is intense, as in the humid tropics, with alternate wet and dry seasons. Kaolinite also occurs in humid temperate zones, in areas of unimpeded drainage (Allen & Hajek 1989;Tardy et al. 1973). While kaolinite is the dominant clay mineral in late Cretaceous sedimentary rocks in southern Iran, it has gradually disappeared by the Eocene reflecting more arid climate conditions (Khormali et al. 2005).
Erionite is not a clay mineral but a zeolite, it has been described to form from weathered volcanic glass in saline and alkaline soils and lakes (Hay 1964;Surdam & Eugster 1976) as in Cappadocia, Central Turkey, where volcanoclastic tuffs were deposited in a lacustrine environment in the late Miocene-Pliocene.

Geological setting
The formation of the SE-NW trending Zagros Mountain belt can be related to the collision of the Arabian and Eurasian plates, during the closure of the Neotethys ocean (Alavi 1994;Berberian 1995). While some models suggest that the collision of Arabia and Eurasia occurred prior to 18 Ma already, a collision between ~ 11.2 and 5 Ma is indicated by zircon provenance analysis (Zhang et al. 2017). The belt is divided from the NE to the SW into four zones: (1) the Sanandaj-Sirjan metamorphic zone, (2) the Imbricated Belt dominated by thrusting, (3) the Simply Folded Belt characterized by folding, and (4) the Mesopotamian foreland basin with buried folds, extending to the SE into the Persian Gulf (Colman-Sadd 1978;Falcon 1974;Homke et al. 2004) (Fig. 1).
The study area near the Iraq-Iran border is part of the Push t-e Kuh Arc in the Simply Folded Zagros belt where syntectonic deformation of foreland sediments started between 8.1 and 7.2 Ma and was active for at least 5 Ma (Böhme et al. 2021;Homke et al. 2004). The studied sections comprise the Gachsaran Formation, lower Agha Jari Member, Lahbari Member and Bakhtyari Formation. The continuous profiles of up to 3 km are exposed in the two syncline-anticline structures of Changuleh and Zarrinabad (Homke et al. 2004) (Fig. 1). All samples from this study were taken at georeferenced sampling points previously sampled by Homke et al. (2004) for magnetostratigraphy, thus giving them a robust temporal control. The lithological succession in the study area described by Böhme et al. (2021) is summarized as follows:

Lithology
The Gachsaran Formation (equivalent to Fatah Fm. in Iraq, lower Fars Fm. in Syria) comprises well-bedded evaporites, metric sized red silty clay beds and green to brown sandstones 10 cm to 2 m thick (Homke et al. 2004). The sandstone beds show symmetrical wave ripples on top as well as mud flat-type trace fossils (faecal pellets and cf. dactyloides). The mudstones contain pedogenic gypsum concretions, gypcrete and redox mottling suggesting fluctuating sea levels between shallow marine shoreface and terrestrial backshore depositions. The top of Gachsaran Fm. is marked by a sandy 0.25-m limestone containing the bivalve Clausinella amidae.
The Agha Jari Formation, covering wide areal extension of the Mesopotamian foreland basin (equivalent to the Injana Fm. in Iraq, upper Fars Fm. in Syria) is characterized by a 2.4-km-thick depositional profile of fine-clastic to fluvial sediments in the study area and has been dated to range from the late Middle Miocene (12.3 Ma) to the base of the Pleistocene (~ 2.5 Ma) by magnetostratigraphy (Homke et al. 2004). It comprises two members: the lower Agha Jari Member and the Lahbari Member. In the study area, the lower part of the Agha Jari Mb. (12.3-11.5 Ma) is characterized by greyish to reddish silty clays rich in pedogenic gypsum concretions with interbedded sandstone channels. At 11.5 Ma, the fine clastic sediments show a reduction of pedogenic gypsum concretions and the occurrence of rubified and lessivized palaeosol horizons. At 11.5 Ma, the fluvial sandstones thicken to > 5 m showing first cross-bedding structures. At 10.8 Ma, the thickness of the distinctly cross-bedded fluvial sandstone strata increases to > 10 m with heavy mineral composition, indicating a (palaeo-) Tigris origin and a palaeo-flow direction pointing towards the southeast (Homke et al. 2004). The fine-grained calcareous palaeosols show intense reddish-green mottling as well as root halos. At 9.7 Ma, sandstone bodies are thin, with calcareous rhizocretes occurring between 9.78 and 6.9 Ma. A thickening of cross-and through-bedded sandstone channels is observed at 8.8 Ma as well as mud-ball erosion. Higher soil moisture within the alluvial plains is indicated between 8.75 and 7.5 Ma by Mn-staining on clay cutans, root halos, gastropods at 8.78 Ma and mass occurrence of small charophyte gyrogonites at 8.69 and 8.56 Ma. At 8.4 Ma, the thickness of sandstones decreases. At 6.25 Ma a, 1-m-thick mudstone horizon showing the efflorescence of salt marks the first transient phase of hyperaridity. It is followed by a 60-m unit of fluvial sandstones dated between 6.15 and 5.95 Ma, marking the youngest deposits of the lower Agha Jari.
The following Lahbari Member consists of beige saline mud sediments without visible palaeo-soil development. The sediments are rich in leachable sulphate, nitrate and chloride and contain infrequent channels of very fine-grained fluvial sandstones of less than 5-m thickness. While fluvial sandstones are almost absent between 5.59 and 5.1 Ma, their frequency increases after 5.1 Ma. At 3.5 Ma, small-scale interfingering channels contain cobble-sized limestone clasts with a general coarsening upward trend in conglomerates. Böhme et al. (2021) suggested a higher dust accumulation Zarrinabad anticline and from 9.02 to 2.4 Ma at Changuleh synclineanticline structure. Map modified from Homke et al. (2004) rate during the Lahbari Mb is by grain-sized end-member modelling as well as a more westerly source of the sediments by heavy mineral composition. A possible sediment source for the Lahbari Member is also found in the softer Gachsaran and Agha Jari deposits regionally uplifted above the proposed low-angle blind thrust, generating the Push-e Kush Arc rise (Emami et al. 2010).
The transition to the Bakhtiari Formation is marked by the appearance of the first boulder-sized clasts in conglomerates just after the beginning of the Pleistocene at 2.5 Ma documenting the progression of the Zagros Mountain Flexure Front (Emami et al. 2010;Homke et al. 2004).

Materials and methods
The 84 evenly distributed mudstone samples were taken from silty horizons along the 2600-m-thick sedimentary profile. Samples from 12.6 to 9.38 Ma were taken at the Zarrinabad anticline and from 9.02 to 2.4 Ma at the Changuleh syncline-anticline structure. As this profile was previously analysed regarding the grain size distribution and the geochemistry of soluble salts, the age model and sample number are identical to the samples used in the study of Böhme et al. (2021). Wherever palaeosols occurred, samples were taken from the B-horizon. Clay minerals sampled from the foreland sediments can be both of detrital and pedogenic origin; thus, the clay mineral record is influenced by the composition of eroded parent material and weathering conditions at time of deposition (Singer 1984b).
For each sample, 5 g of sediment was dried at 110 °C in a drying cabinet and then carefully dispersed by submerging the bottom of the sample in a porcelain crucible in deionized water to gently disaggregate the mineral particles using the swelling properties of the smectitic clay minerals (procedure known as "unterschichten"). Some samples with a higher carbonate and gypsum content required a repetition of this gentle disaggregation process. The samples were then treated with a Dr. Hielscher UP 400S ultrasonic processor (200 W, 50% cycle-time, 30 s) to obtain a homogeneous suspension and to destroy soft agglomerates. Afterwards, sand and potential micro-fossils were removed with a 63-µm sieve (Lehmann et al. 2004). The fraction < 63 µm of each sample was left to flocculate in a 2 l beaker, caused by the cations from the solvable minerals still existing in the sample, with the electrolyte-rich clean water on top being decanted after one night. The decantation process was repeated several times to remove the remaining cations until a stable suspension started to develop. After that, the suspension was stabilized with 25 ml of 0.1 M sodium pyrophosphate solution, and clay minerals < 2 µm were separated from silt by repeated sedimentation in Atterberg columns. Unfortunately, coarse-grained crystalline calcium phosphate hydrate crystallites precipitated in the suspension, which formed by reaction of sodium pyrophosphate with calcium cations from dissolving gypsum in the sample. They were removed from the clay suspension by additional sieving using a sieve with 63-µm mesh size and decantation of the suspension from the sediment of the crystallites. The clay was then concentrated from the suspension using a 0.8-µm vacuum filter to a 40-ml volume. Thereafter, 1 ml of the clay mineral suspension was transferred onto 2 glass slides each and air-dried as oriented clay samples for XRD analysis, with one of the samples being re-measured after treatment with ethylene glycol vapour in an exicator for at least 72 h. A Bruker D8 advance diffractometer with a Cu-sealed tube running at 40 kV/20 mA, a Göbel mirror parallel beam optics, a 0.2-mm divergence slit, a fixed knife edge to suppress air scatter and a 1D-VǺNTEC 1-detector in scanning mode with a step size of 0.008° 2θ and 360 s/step was used for the XRD analysis resulting in a measurement time of approximately 2 h for the range from 2°2 Theta to 55°2 Theta. Mineral identification was performed using the 2006 PDF-4 database from the International Centre for Diffraction Data-Joint Committee of Power Diffraction Standards (ICDD-JCPDS). As there are six or more mineral phases present in the clay-sized fraction and mineral texture, mineral grain orientation, reflex overlap as well as swelling of smectites and background extraction can lead to significant errors and uncertainties in exact clay mineral quantification (Brindley & Brown 1980). A comparison of the integrated reflex intensities was not deemed suitable for the task of clay mineral quantification along the sedimentary column due to the clay minerals on the samples being oriented as well as multiple broad reflex overlaps (Fig. 2). As an alternative, a semi-quantitative approach was chosen by comparing the height only of the main reflex of glycolated oriented samples (001 chlorite; 002 smectite, illite; 110 palygorskite; 100 erionite) (Johns et al. 1954). The following correction factors were applied to account for the broad 002 reflex overlap of smectite with 001 chlorite, 110 palygorskite with 002 illite and 110 palygorskite with 100 erionite. The correction factors were chosen by careful comparison of the relative overlap of intensities for each mineral. They are chlorite, − 0.25 *smectite; illite, − 0.25*palygorskite; erionite, − 0.1*palygorskite. For the estimation of the kaolinite content, a selection of samples was additionally heated at 550 °C for 3 h. The resulting percentage ratios of intensities of clay minerals thus do not represent the exact weight percentage of each clay minerals in the sample, but their comparison can be used to reflect the relative change in clay mineral ratios between different samples. The ratio of smectite/(illite + chlorite) was then used to assess changes of soil available moisture. To check for a potential correlation with concentration of clay minerals with previously analysed data of soluble salts and sulphate, correlation plots and Spearman's correlation were obtained using the software JMP.
A Phenom XL C2 desktop SEM with a 15 kV acceleration voltage was used to analyse charophyte (green algae) fossils as well as pedogenic concretions at University of Tübingen.
Clay mineralogy was compared to bulk magnetic susceptibility previously published as supplementary information by Böhme et al. (2021). Magnetic susceptibility was obtained for all samples using an MFK-1 AGICO Kappabridge on weighted portions of the samples.

Results
The clay fraction of the studied samples consists of highly variable proportions of smectite, illite, chlorite, palygorskite as well as the zeolite mineral erionite (Fig. 2) and was plotted next to the corresponding magnetic susceptibility, salt and sulphate data for better comparability (Fig. 3). Small amounts of kaolinite were present in some samples albeit quantification was not possible due to a varying overlap of the 001 intensity of kaolinite with the 002 reflex of chlorite upon heating to 550 °C. Erionite was only detectable in the clay fraction < 2 µm by XRD. It was not identifiable by follow-up SEM analysis probably due to its small size. Some samples of the fraction < 2 µm also contained accessory quartz, calcite as well as dolomite. Bivariate plots for individual clay minerals, Na + , Cl − and SO 4 2− are shown in Fig. 4, and Spearman's correlation coefficients among clay minerals are in Table 1. Among the clay minerals, palygorskite exhibited a moderate negative correlation with illite and smectite while the correlation with chlorite was weak and statistically insignificant. The correlation among illite, chlorite and smectite was also found to be weak. Bivariate plots showed that samples obtained from the Zarrinabad section of the lower Agha Jari Fm. exhibited Na + , Cl − and SO 4 2− concentrations that were generally one or two orders of magnitude lower than samples obtained from the Changuleh section of the lower Agha Jari Member Formation. Clay mineralogy at the same time did not show a distinct clustering between same samples obtained from the two sampling localities. In terms of clay mineralogy and soluble salt geochemistry, samples from the Lahbari Member clustered distinct from those from the lower Agha Jari Member, with higher palygorskite, Na + , Cl − and SO 4 2− concentrations than samples from the Gachsaran Formation and lower Agha Jari Formation.

Gachsaran Formation
The lowermost sample dated to 12.6 Ma, which is the only one belonging to the Gachsaran Formation, contains illite, erionite as well as minor smectite and chlorite and accessory dolomite. Magnetic susceptibility is the lowest of the whole profile. The presence of small gyrogonites of Chara sp. is indicative of freshwater or brackish conditions at the time of sediment deposition (Soulié-Märsche 2008).

Lower Agha Jari Member
The six stratigraphic lowest samples from 12.02 to 11.2 Ma comprise smectite, illite, chlorite and often erionite as well as accessory quartz, while palygorskite is absent. First traces of palygorskite are present at 10.88 Ma; its content increases to a first maximum at 8.75 Ma and then decreases to a new minimum at 7.1 Ma (Fig. 3B). Later on, palygorskite shows a long-term increase towards the onset of the Lahbari Member. While smectite contents are generally high throughout the lower Agha Jari Member, long-term maxima can be Fig. 2 Three exemplary x-ray diffractograms of three glycolated clay mineral samples representing clay mineral assemblages from semi-arid (black curve), arid (blue curve) and hyper-arid (red curve) conditions. Background has been subtracted of all three samples for better comparability of reflex intensities distinguished around 10 and 8 Ma. The amount of smectite starts to generally decrease after 7 Ma to reach a minimum at the onset of the Lahbari Member (Fig. 3A). Illite shows a long-term decrease from 12.02 to 7.25 Ma (Fig. 3C). Its percentage is then temporarily elevated at around 7 Ma and decrease again at the onset of the Lahbari Member. The intra-sample variation of illite is relatively high. While chlorite content (Fig. 3D) is generally lower than illite, a strong intra-sample variability throughout the lower Agha Jari Member is noted too. Erionite occurs rather unsystematically in low amounts in many samples between 12.02 and 7.25 Ma (Fig. 3E). Between 7.1 and 5.5 Ma, the strong variation as well as a total increase of erionite is clearly noticed. The resulting ratio of smectite/(illite + chlorite) has both a strong variability in between samples but also on a longer time-scale (Fig. 3H). After gradually increasing until 9.54 Ma, it drops moderately for about 0.75 myrs. Between 8.75 and 6.9 Ma, it reaches on average the highest values throughout the sedimentary column and then sharply drops, to remain relatively low until the onset of the Lahbari Member at 5.59 Ma with a minimum at 6.25 Ma. The initial increase in palygorskite between 10.88 and 8.75 Ma coincides with an increase of magnetic susceptibility (Fig. 3G) of the sediment. Magnetic susceptibility then fluctuates around a median of 2 −7 SI until the onset of the Lahbari Member. Small gyrogonites of Chara sp. were observed in samples at 10.25 Ma and 9.51 Ma. The presence of dwarfed gyrogonites of Chara sp. at 8.69 and 8.56 Ma was previously described by Böhme et al. (2021). The gyrogonites at 8.56 Ma were now identified to likely comprise Chara vulgaris, Chara globularis and Chara sp. (Fig. 5d-h). Three gyrogonites showing strong dissolution etching of calcite at the surface were identified as Nitellopsis obtusa at 7.5 Ma (Fig. 5ij). SEM analysis revealed the overgrowth of palygorskite fibres onto calcified charophyte thalli at 8.56 Ma (Fig. 5a-c). Palygorskite overgrowth can also be seen on pellet-shaped carbonate concretions with possible halite pseudomorphs at 7.1 Ma (Fig. 6). At 8.78 Ma, tube-shaped aggregates of gypsum likely resembling rhizocretes, similar to those reported by Khalaf et al. (2014) in a fluvial playa, were observed (Fig. 7a).

Lahbari Member
The base of the Lahbari Member at 5.59 Ma is generally characterized by higher palygorskite and lower smectite contents than the lower Agha Jari Member. The clay mineral assemblage of the Lahbari Member can be divided into three intervals: (1) 5.59-4.6 Ma; (2) 4.44-4.33 Ma; (3) 4.25-2.4 Ma. During the first interval, the amount of smectite is the lowest of the entire sedimentary column, while the content of palygorskite is very high. Illite and smectite are also comparatively low, and smectite/(illite + chlorite) reaches the lowest values of all the investigated samples. The second interval is characterized by an increase in illite, smectite, chlorite and smectite/(illite + chlorite) ratio, while the amount of palygorskite is cut in half. In the third interval, the amounts of illite and chlorite are again reduced, while palygorskite sharply increases. Smectite and smectite/ (illite + chlorite) are slightly higher than in the second interval. Erionite is generally rare in the Lahbari Member in comparison to the lower Agha Jari Member. Magnetic susceptibility sharply decreases at the onset of the Lahbari Member, then increases until 4.7 Ma and then moderately decreases until 4.25 Ma. During the third interval, it stays high again. A single gyrogonite of Chara sp. was observed in 10 g of sediment from 5.50 Ma additionally to several ostracod fragments ( Fig. 5k-l). Fragments of ostracods were also found at 3.55 Ma. Fine-grained gypsum comprises a major component of the sand fraction of the Lahbari Member, a potential gypsum rhyzocrete measuring around 1 mm in diameter being observed in one sample at 4.44 Ma (Fig. 7b).

General discussion
While clay minerals provide an integrated record of overall climate impact, they do not provide direct indications of climate parameters. Levels relatively rich in chlorite, illite, palygorskite and quartz correspond to relatively dry periods, while humid periods are dominated by more stable clay minerals such as kaolinite. Smectite can indicate a climate with contrasting seasons and a pronounced dry season (Singer 1984b). The distinction between detritic and authigenic clay minerals is problematic as extra-climatic factors such as parent rock material, differentiation during transport, topography and diagenetic changes can influence clay mineral assemblages (Singer 1984b). As detrital clay minerals will have been transported and repositioned by aeolian and fluvial processes along the catchment area of the Mesopotamian foreland basin, the palaeoclimate record of this study will therefore be indicative of palaeoclimatic impact in a larger integrated area. In areas exceeding 300 mm of annual precipitation, palygorskite is unstable and can weather to other clay minerals such as smectite (Paquet & Millot 1972), making the presence of palygorskite a good indicator for relatively dry climate. As palygorskite can only form from smectite and illite in the presence of alkaline fluids rich in Mg and Si (Singer 1989), palygorskite might not form in arid soils lacking sufficient Mg and Si. In a recent study, Hashemi et al. (2013) found parent rock material composition to be the most important factor on clay mineral distribution in different moisture regimes in gypsiferous palygorskite-rich soils of various geological origins in the semi-arid to arid Fars Province in Iran. Besides the influence of parent material onto the clay mineral assemblage, it was demonstrated that the ratio of smectite/(illite + chlorite) increased with soil moisture. Hereby, the lowest ratio of smectite/(illite + chlorite) was found in soils with aridic moisture regime (mean 0.38) and the highest ratio in a xeric moisture regime (mean 1.26, max 2.12). We assumed that the correlation between soil moisture and smectite/(illite + chlorite) was indicative of changes in the palaeo-soil moisture regime in this study because it was demonstrated across pedons formed on parent material from various geological origins; however, changes in detrital parent material will have an effect on clay mineral assemblages.

Clay transformation
In our study, we observed an almost linear anticorrelation across all samples between palygorskite and illite suggestive of neoformation of the former at the expense of the latter (Table 1, Fig. 4). This relationship also seems to hold true between palygorskite and smectite albeit to a lesser extent. Similar neoformation and transformation between clay minerals were reported in recent arid Iranian soils by Khormali and Abtahi (2003). Further detailed geochemical studies will be helpful in getting a deeper insight into the geochemical controls of palygorskite genesis along this sedimentary profile.

Magnetic susceptibility
The changes in clay mineralogy were also compared to changes in magnetic susceptibility. The magnetic susceptibility of soils is affected by geological parent material composition and soil forming processes (Dearing et al. 1996). In a study of subtropical Chinese soils, a trend of mafic volcanic rocks > felsic rocks > metamorphic rocks > sedimentary rocks was shown for magnetic susceptibility (Shenggao 2000). Magnetic susceptibility of soils was analysed by Sanjari et al. (2021) along a xeric-aridic climolithotoposequence in central Iran. Hereby, it was shown that magnetic susceptibility values were increased with depth in soils with igneous parent material due to primary ferrimagnetic particles, but the opposite trend was found in soils with a sedimentary origin indicating a stronger influence of pedogenic magnetic particles. Increasing weathering in more developed soils of igneous parent material causes the destruction of inherited primary magnetic minerals through time, decreasing magnetic susceptibility. However, formation of pedogenic ferrimagnetic particles in soils with sedimentary parent material increased magnetic susceptibility with the soil development (Fine et al. 1989;Sanjari et al. 2021;Shenggao 2000). Looking at the complete sedimentary profile, we could not establish a clear positive correlation between enhanced weathering, indicated by a high proportion of smectite or kaolinite compared to illite and chlorite or geological indicators such as manganese staining or redox mottling that would correspond to strongly elevated magnetic susceptibility by pedogenic ferrimagnetic particles in the sediments. Therefore, we assume the contribution of inherited primary magnetite grains to be larger than the contribution of pedogenically formed ferrimagnetic minerals such as magnetite or maghemite.

Gachsaran Formation and lower Agha Jari Member
The clay mineral suite between 12.6 and 9.54 Ma likely reflects a gradual transition from arid to semi-arid climate conditions. A more intense transformation of illite to smectite and starting at 10.88 Ma palygorskite indicates increased amounts of soil available water. The enhanced chemical weathering of illite to smectite might also be indicating a climate with increasingly contrasting seasons and a pronounced dry season (Singer 1984b). An increase in mafic input sourced from tectonically exhumed, allochthonous slices the Kermanshah radiolarite-ophiolite complex in the high Zagros is regionally observed in Zagros foreland sediments from the Dezful Embayment to the SE of this study area in the middle Tortonian (Etemad-Saeed et al. 2020). Thus, the co-increase in palygorskite and magnetic susceptibility after 10.88 Ma can best be explained by enhanced input of more mafic parent material rich in magnetite and Mg-rich minerals, providing the necessary magnesium for palygorskite formation. Fine-grained sediments of this section also show indicators of soil development such as redox mottling and carbonate rhizocretes (Böhme et al. 2021), which might partially contribute to the increase in magnetic susceptibility in palaeosols. However, a contemporaneous increase in sedimentation rate along the floodplain between 12.8 and 9.5 Ma from 13 to 30.5 cm/ka (Homke et al. 2004) would indicate the accelerated subsidence of soils giving less time for intense pedogenic formation of ferrimagnetic minerals. The sample taken at 9.78 Ma giving an excessively high smectite/(illite + chlorite) ratio of 7.7 appears to be a statistical outlier probably due to complete lack of illite potentially due to transformation to palygorskite (encircled sample in Fig. 3H). Between 9.51 and 8.78 Ma, a drop in smectite/(illite + chlorite) ratio suggests a slight intermittent aridification which is corroborated by field observations of a decrease in fluvial channel thickness and therefore fluvial run-off rate. This aridification is roughly contemporaneous (I) smectite/(illite + chlorite) ratio as indicator for soil moisture availability with higher ratios indicating higher moisture (encircled sample represents a statistical outlier due to lack of illite). Right columns: new long-term aridity interpretation based on clay mineral ratios beside soluble salt-based interpretations of Böhme et al. (2021). Horizontal red line marking different sampling localities at Zarrinabad and Changuleh syncline-anticline structures, note the lack of soluble salts at Zarrinabad sampling site with no apparent change in clay mineralogy ◂ with a temporary Paratethys low stand which might have led to decreased precipitation in Northern Arabia (Böhme et al. 2021). The general increase in soil available moisture reflected in the highest average high smectite/(illite + chlorite) ratios recorded throughout the profile between 8.75 and 7.25 Ma is also noted by field observation of pedogenic features such as manganese staining, plant-root halos and a thickening of fluvial channels (Böhme et al. 2021 Fig. 3 E-H). As the continuous presence of palygorskite and lack of kaolinite excludes humid climate conditions, a semi-arid climate conditions appear to be most probable. While two clay mineral samples between 7.1 and 7.04 Ma indicate more arid conditions at the beginning of the Messinian already, a permanent drop in smectite/(illite + chlorite) ratio is only observed after 6.9 Ma. No systematic change in magnetic susceptibility is observed during this transition indicating a stronger signal of inherited magnetite grains than pedogenically formed magnetic minerals in the sediments. A possible trigger for the aridification indicated at 6.9 Ma could be provided by the Intra-Maeotian event, leading to a significant Eastern Paratethys base-level drop (Palcu et al. 2019

Lahbari Member
The Lahbari Member samples from 5.59 to 4.6 Ma show a further decrease in the average smectite/(illite + chlorite) ratio; the dominance of palygorskite indicates highly evaporative, arid conditions. Because gypsiferous soils can provide buffered alkaline media with necessary anions and cations for palygorskite crystallization (Birsoy 2002;Owliaie et al. 2006), the observation of very high palygorskite content is possibly linked to an increase in sulphate and gypsum ( Fig. 3 B, H, Fig. 4). Elevated gypsum, halite and palygorskite content in the samples might also be partially derived from other evaporative sediments due to a strong increase in aeolian dust input (Böhme et al. 2021). This detrital material could possibly be regionally sourced from tectonically exposed evaporitic sediments of the Gachsaran Formation within the folded Zagros (Emami et al. 2010). The short reduction in palygorskite and increase in smectite and illite between 4.44 and 4.33 Ma might reflect unfavourable geochemical palygorskite formation conditions, a decrease in aeolian dust input or potential short-term cooling leading to a reduction in chemical weathering. A general increase from 4.33 to 2.4 Ma in the smectite/(illite + chlorite) ratio suggests a slight raise in soil water availability compared to the lower parts of the Lahbari Member. However, it might also reflect changes in parent materials due to the progression of the Zagros Mountain Front Flexure, characterized by the Bakhtiari Formation starting at 2.5 Ma (Homke et al. 2004).

Comparison with soluble salt data
The extremely low salt and sulphate concentrations of the samples taken in the Zarrinabad section of the geological profile (12.6-9.38 Ma) (Fig. 4) are highly indicative of post-depositional leaching processes, rendering the previous palaeoclimatic interpretation in Böhme et al. (2021) in this part of the profile by soluble salt geochemistry as semiarid unreliable. The very low salt and sulphate concentrations are in disagreement with the frequent observation of palygorskite in the same samples, which has been shown to form in evaporative soils at high pH after the increase of Mg/Ca ratio by initial gypsum precipitation (Khademi & Mermut 1998). Arid conditions are also strongly suggested by a very low smectite/(illite + chlorite) ratio and an absence of kaolinite in the samples lacking palygorskite between 12.02 and 11.2 Ma. As clay mineralogy in the Zarrinabad Sect. (12.6-9.38 Ma) of the lower Agha Jari Fm does not cluster distinctively different from the Changuleh Sect. (9.02-5.6 Ma), clay minerals do not seem to have been affected by post-depositional leaching unlike soluble salts and sulphate. The observation of post-depositional leaching at the Zarrinabad sampling site (150-250 mm/a annual precipitation) is probably caused by comparatively higher mean annual precipitation than at the Changuleh sampling site (85-150 mm/a annual precipitation) (Nikpour et al. 2022). Fig. 6 a,  The transient hyper-arid events suggested by soluble salt geochemistry at 8.75, 7.78 and 7.50 Ma, which should correspond to Paratethys low stands are not clearly discernible in the clay mineral paragenesis. This might be due to short duration of the events, other factors affecting regional climate or a stronger overprint of parent material of the weathered clays. However, the transient hyper-arid event at 6.25 Ma suggested by soluble salt geochemistry is also reflected in the clay mineral assemblage. While a continuous signal of enhanced aridity after 6.9 Ma is indicated by clay minerals and in sulphate data, it does not show in soluble salts possibly due to post-diagenetic leaching of highly soluble salts. In the lower Lahbari Member, the clay mineral assemblage is very rich in palygorskite with a low smectite/ (illite + chlorite) ratio which is congruent with the previous interpretation of hyperaridity during NADX based on the geochemistry of soluble salts and palynology (Böhme et al. 2021). Nevertheless, it is not possible to draw a clear threshold between merely arid and hyper-arid conditions using clay mineralogy.

Erionite
As erionite is formed by the weathering of volcanic glass under alkaline-saline conditions (Hay 1964;Surdam & Eugster 1976), its occurrence in the sediments along palygorskite supports the presence of alkaline or saline-alkaline soil solutions or ponds at time of deposition. It is unclear whether the volcanic glass found in the analysed sediments is related to reworked volcanoclastic material or syndepositional ashfall events.

Charophytes
As biota have a much quicker, seasonal, response time to environmental change than clay mineral assemblages, which should take longer timeframes to respond to new equilibrium conditions (cf. Hillier & Pharande 2008), they can give an inside into sub-annual climatic changes. The observation of small charophyte gyrogonites of Chara vulgaris and Chara globularis along palygorskite cutans on calcified charophyte thalli in a sediment sample at 8.56 Ma as well as the occurrence of Chara sp. along ostracod fragments in saline mudstones at 5.5 Ma suggests a strong seasonal fluctuation in surface water availability and salinity. While extant charophytes require at least 3 months of fresh to brackish water conditions to complete their growth cycle (Soulié-Märsche 1991), authigenic palygorskite formation usually occurs in alkaline or alkaline-saline shallow lakes or lagoons with elevated Mg and Si activities (Draidia et al. 2018;Singer 1979). Suitable conditions to explain both the small size of the observed charophyte gyrogonites (Böhme et al. 2021;Vicente et al. 2016) and the authigenic palygorskite formation could be created by occasional seasonal desiccation of shallow fresh ponds within the floodplain under highly evaporative conditions resulting in a contemporaneous increase in alkalinity and salinity. Therefore, palygorskite found on charophyte thalli at 8.56 Ma would only have precipitate after charophytes had succumbed to a stressful, increasingly saline environment. While Chara vulgaris is commonly found in shallow floodplain ponds, the presence of gyrogonites of Nitellopsis obtusa in arid environments is usually an indicator of permanent, oligohaline freshwater bodies of 4-12 m depth (Kröpelin & Soulié-Märsche 1991;Soulié-Märsche 1993). The presence of three gyrogonites of Nitellopsis obtusa at 7.5 Ma showing calcite dissolution in weakly consolidated, carbonate-poor, fine-grained floodplain sediments without sedimentological indications palaeolake or oxbow development such as laminations or marl-banks can possibly be interpreted as allochthonous redeposition of gyrogonites of older lake sediments. An autochthonous presence of N. obtusa in this sample also seems implausible by the soluble salt geochemistry, as it has previously been classified to belong to a transient hyper-arid climate based on an elevated halite content of 1.07 wt %.

Conclusions
Clay mineralogy of the lower Agha Jari Fm suggests arid climate conditions in Northern Arabia during the Late Serravallian and early Tortonian, which were not indicated in a previous soluble salt-based study due to post-depositional leaching of sediments in one part of the stratigraphical section of the previous investigation. While conditions during the late Tortonian are generally suggested to be semi-arid, clay mineralogy argues towards aridification during the early Messinian. A sharp transition in clay mineralogy is visible at the onset of the Lahbari Mb at 5.59 Ma, where high palygorskite content and low smectite/(illite + chlorite) ratio characterise a phase of hyper-arid sediment deposition, that was previously proposed based on soluble salt chemistry. The onset is contemporaneous with the apex of the Messinian salinity crisis. Clay mineralogy of the mid-late Pliocene suggests an increase in soil moisture with arid climate conditions. While magnetic susceptibility data provided useful indirect information on changing parent material mineralogy, further detailed geochemical studies would be helpful for establishing better control on the total influence of parent material composition and detrital clay minerals onto the clay mineral paragenesis. As transient hyper-arid phases proposed based on soluble salt geochemistry at 8.75, 7.78 and 7.50 Ma were not sufficiently resolved in changing clay mineralogy, it is indicated that soluble salts could be more sensitive in distinguishing short-term aridity thresholds between semi-arid, arid and hyper-arid conditions than clay mineralogy. However, we also found that soluble salt geochemistry should always be employed along with other proxies such as clay mineralogy in palaeoclimate studies of continental sediments due to the possibility of post-depositional leaching of individual samples or entire sections. The geographical distribution of erionite was found to be much larger than previously known. High resolution palynological and micropalaeontological studies along this reference profile are warranted to better resolve inconsistencies between the clay and soluble salt-based climate reconstruction of individual samples and to enhance the understanding of temperature and precipitation variations and the associated evolution of the semi-arid to hyper-arid ecosystems.

Acknowledgements
The authors are thankful for fruitful discussions that greatly helped improving the manuscript with Haytham El Atfy (University of Tübingen). Ingeborg Soulié-Märsche (Université de Montpellier) is thanked for invaluable help with the identification of charophytes. Support at the SEM was provided by Tatiana Miranda and Martin Ebner (University of Tübingen). Anette Flicker (University of Tübingen) is acknowledged for technical support during sample preparation in the XRD lab. Sample photography was kindly provided by Agnes Fatz (University of Tübingen). Martin Gross (Naturkundemuseum Graz) is thanked for help with the identification of ostracods.
Funding Open Access funding enabled and organized by Projekt DEAL.

Conflict of interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.