The authors are grateful to TOTAL, ENI, SHELL and PETROBRAS for partial funding of this project. U/Th dating conducted in the High-precision Mass Spectrometry and Environment Change Laboratory (HISPEC) was supported by the Science Vanguard Research Program of the Ministry of Science and Technology (108-2119-M-002-012) and the Higher Education Sprout Project of the Ministry of Education, Taiwan ROC (108L901001). Sándor Kele was supported by the National Research, Development and Innovation Office in Hungary (NKFI Alap, KH-125584 project) and by the European Union and the State of Hungary, co-financed by the European Regional Development Fund in the project of GINOP-2.3.2-15-2016-00009 ‘ICER’.

1 – Introduction

1Continental carbonates have received recent attention as palaeoenvironmental and palaeoclimate recorders (Frank et al., 2000; Bertini et al., 2014; Toker et al., 2015; Vignaroli et al., 2016) and as hydrocarbon reservoirs (Lottaroli et al., 2012; Ronchi & Francesco, 2015). In addition, their associated springs and caves attracted a wide range of flora and fauna, including humans that were captured as fossils in the carbonates (e.g. Barker et al., 2007; Cremaschi et al., 2010; Lebatard et al., 2014; Meyer et al., 2017). The deposits thus can provide unique archaeological information on paleo-climate and paleo-environment, and sometimes even faunal and/or human responses to the former. When retrieving and interpreting palaeoclimatological and paleoenvironmental information from continental carbonates, reliable dating is essential. For palaeoclimate studies, dating is of interest on a long-term time-scale, e.g., to verify whether continental carbonates such as travertines form preferentially during interglacials or not (Hennig et al., 1983; Frank et al., 2000; Uysal et al., 2009; Özkul et al., 2013), and on a shorter time-scale, i.e., for variations within the relevant recorded palaeoclimate signature section (e.g. Yan et al., 2012). In addition, dating of travertine veins can be used to better understand past carbon dioxide emissions which is of relevance for the integrity for geological storage sites (Kampman et al., 2012; Berardi et al., 2016; Mancini et al., 2019). Accurate dating of travertine formations associated with river terraces is crucial for quantification of neotectonic deformation and incision rate of rivers (e.g. Ruszkiczay-Rüdiger et al., 2005, 2018). From a palaeoenvironmental point of view, the timing of active precipitation and associated environmental conditions are key and correlation of dated horizons are of high interest. The precipitation rate of continental carbonates, including travertines, is also of prime importance for example when dealing with forward modelling of hydrocarbon reservoir architectures, making dating a decisive factor.

2Pentecost (2005) gave an overview of the main possible dating methods on travertines. For fossil travertines the most widely applied method is U/Th dating, but its application in some cases is controversial because of the complexity related to open-system U-leaching and the presence of detrital Th (Alonso-Zarza & Tanner, 2010). Recent studies (Mallick & Frank, 2002; Prado-Pérez et al., 2013; Özkul et al., 2013), however, claim to be able to circumvent these problems, evidenced by its wide application. The time limits of the ²³⁰Th/U-dating method are from three years to about 600 ka (Edwards et al., 2003), but with the ²³⁸U-²³⁴U decay scheme, this dating range can be extended up to ca. 1.5 million years (Spötl & Boch, 2012; Ricketts et al., 2014).

3Two case studies (fig. 1), the Ballık area (SW Turkey) and the Süttő travertine complex (Gerecse Hills, Hungary), have been selected to evaluate the possibilities and constraints related to U/Th dating of fossil travertine deposits. In addition, this study aims to better constrain the temporal and depositional evolution of travertine in these areas.

2 - Previous age data of the Ballik and Süttő area

2.1 - Ballik (Turkey)

4The Ballık area lies at the northern edge of the Denizli basin, which is known for its numerous travertine occurrences. Özkul et al. (2013) provided an overview of the travertine occurrences in the Denizli basin, including U/Th and thermoluminescence (TL) dating results. Up to 2013 the oldest travertine ages in the Denizli basin were reported at 828 ± 93 ka and 1.11 ± 0.11 Ma by thermoluminescence and electron spin resonance (Engin & Güven, 1997; Engin et al., 1999). Unfortunately, no detailed information on the exact location of the dated samples was provided. Another dating of the Ballık travertines by TL resulted in an age of 490 ± 50 ka (Özkul et al., 2004). The recent study of Lebatard et al. (2014) aimed at dating the Homo erectus found in these travertines. They found much older age estimates for the travertines. Cosmogenic nuclide dating of travertine interfingering with and covering the conglomerates in the Faber quarry was combined with palaeomagnetic measurements. The travertine unit above the marl-conglomerate incursion would have precipitated at roughly 10 m per 100 ka between 1.22 (1.07) Ma and ~1.5 Ma (Lebatard et al., 2014). The oldest travertine is older than the lower conglomerates which have a minimum age of 1.63 Ma (Lebatard et al., 2014).

5For the Denizli basin in general, up to now, no travertines have been dated older than the Quaternary period. Travertine deposits become younger towards the centre of the basin. In the northern part of the basin, precipitation is still ongoing, for example in Pamukkale and Karahayit (Kele et al., 2011). In the Ballık area, underground precipitation is ongoing in the Kaklık cave (Gulecal-Pektas & Temel, 2017).

2.2 - Süttő (Hungary)

6Several, mostly Quaternary, travertine occurrences can be found in the Pannonian Basin (e.g. Kele, 2009; Sierralta et al., 2010). More than 50 travertine occurrences are known in the Gerecse Hills, located predominantly in the northern marginal areas of the Hills (Scheuer & Schweitzer, 1988). The Süttő travertine complex is situated on the slope of the Haraszt Hill, in the Gerecse Mountains (Hungary). The travertine bodies occurring along the fluvial terraces of the River Danube and its tributaries played a key role in correlation of the terrace levels and in estimation of the incision of the river and the uplift rate of the Mid-Hungarian Mountain Range (e.g. Ruszkiczay-Rüdiger et al., 2005; 2018). Pécsi (1959) and Kretzoi and Pécsi (1982) identified six to eight terrace levels and 12 associated travertine horizons along the Danube. The travertine complexes are often covered by loess, dated by luminescence dating (e.g. Novothny et al., 2009; Rolf et al., 2014) thus providing independent information regarding the time interval of the travertine formation. Ruszkiczay-Rüdiger et al. (2005; 2018) summarized the results of the chronological studies and dated the under- and overlying sequences of the travertine occurrences along the terrace levels of the Danube with cosmogenic nuclides. Based on borehole information, the underlying beds of the Süttő travertine complex may correlate with the t_VII terrace level of the Danube River with an estimated Late Pliocene - Early Pleistocene age (Pécsi, 1959). On the other hand, according to the most recent re-evaluation of the borehole data (Csillag et al., 2020), they may belong to the redeposited Upper Pannonian (Újfalu Formation) as well, and may have no connection with the Paleo-Danube river. On the top of the eroded surface of the Süttő travertine body, Late Pleistocene siliciclastics and Riss (=Saalian)-Würmian (=Weichselian) loess was deposited (Novothny et al., 2009; 2011; Rolf et al., 2014) and the whole sequence is covered by Holocene top-soil. Based on the vertebrate contents of the Diósvölgy quarry, Brunnacker et al. (1980) estimated a maximum Pleistocene age. Paleomagnetic investigations of Lantos (2004) suggested that the Süttő travertine complex formed 780 ka before present (still during the Matuyama chron). U/Th dating analyses of Kele (2009) resulted 422 ± 21 ka (Diósvölgy Quarry), 273 ± 65 ka (Hegyhát quarry) and 296 ± 22 ka (Újharaszt quarry). Sierralta et al. (2010) performed paleomagnetic measurements on samples from the Újharaszt and Hegyhát quarries with a resulting unstable paleomagnetic signal but their U/Th dating on samples from the Újharaszt and Hegyhát quarries resulted in Middle Pleistocene ages (235 ± 21 ka and 314 ± 45 ka). According to the fossil remains, most of the fracture-fillings in the area can be correlated with different phases of MIS 5 (130 and 70 ka) which is in line with the optical dating of the covering loess (Pazonyi et al., 2014). The fossil content of the travertine body itself, exposed in the Hegyhát quarry, indicates a Lower to Middle Pleistocene age (~800 ka), whereas the fossil remnants of the Újharaszt quarry suggest a Lower Pleistocene age, that contrasts with the suggested age based on uranium series dating of Sierralta et al. (2010). The fauna of the Cukor quarry indicates an even older, Pliocene/Lower Pleistocene age (Pazonyi et al., 2014). U/Th data of Török et al. (2019) from the Cukor quarry were incorporated in the presented data set.

7In general, travertine occurrences in the Gerecse Hills become younger from the hills towards the River Danube. Active travertine precipitation can still be observed along the lowest terrace level of the Danube (Dunaalmás, Csokonai springs), however, down there, no data exist regarding the precipitation rate.

3 - Theoretical background

8U/Th dating starts from the difference in solubility between parent uranium (high) and daughter thorium (negligible) in groundwater under oxidizing conditions (Ivanovich & Harmon, 1992) producing a disequilibrium in calcite precipitating from that water. In an ideal scenario, all thorium found in freshly precipitated carbonates, like for example travertine, tufa and speleothems, would result from the radioactive decay of uranium. The time since precipitation can be deduced, based on the extent to which secular equilibrium has been restored, unless secular equilibrium is already achieved.

9There are several criteria for a sample to be suitable for dating: 1) a measurable quantity of uranium should be present; 2) no daughter nuclide must be present at the time of deposition or otherwise an appropriate correction should be applied; 3) the sample must have behaved as a closed system to post-depositional migration or addition of the radionuclides used for dating (Ivanovich et al., 1992). The amount of co-deposited ²³⁰Th in the carbonates is expected to be negligible (Harmon et al., 1980; Schwarcz, 1990), since most of it absorbs on clay minerals (Schwarcz, 1990), ferric oxi/hydroxides (Pentecost, 2005) and organic particles. These three phases can, however, contaminate carbonates (Mallick & Frank, 2002). The incorporation of allochthonous particles (also as solid inclusions) could also lead to an overestimation of the age (Geyh, 2008).

10Continental carbonates are characterized by their eye-catching porosities, hence also their potential importance as hydrocarbon reservoir rocks (Ronchi & Francesco, 2015). Large pore volumes, however, yield a great potential for water/rock interaction with open system behaviour (Mallick & Frank, 2002). An adapted sampling strategy is thus essential.

11Petrography should be combined with mineralogical and geochemical analysis to select primary calcite (Livnat & Kronfeld, 1985; Mallick & Frank, 2002). Continental carbonate samples that show obvious evidence of detrital carbonate inclusions should be excluded (Livnat & Kronfeld, 1985) and dense material, that lacks primary porosity is preferred because it consists of almost pure micrite with little filling of cements (Schwarcz, 1980; Mallick & Frank, 2002). Recrystallised samples should also be avoided but are difficult to recognize as pointed out e.g. by Claes et al. (2015) and De Boever et al. (2017). Sometimes they can be distinguished based on their elemental composition and their cathodoluminescence. Recrystallisation from aragonite to calcite will increase the Mg content, while Sr, Ba and Ra may be lost (Sturchio et al., 1994). Mallick and Frank (2002) concluded that:

1. Solely areas of dominant micrite and/or spar abundance should be sampled;

2. The present micrite and spar must exhibit solely weak cathodoluminescence;

3. Sr concentrations of the samples should be within a ± 20 % range of the average value determined at a travertine site or one specific stratigraphic unit;

4. Al, Th and Fe concentrations should be less than 300 µg/g, 40 ng/g and 200 µg/g respectively (the latter would indicate potential contribution by clays or oxi/hydroxides or both).

12²³⁰Th/²³²Th activity ratio is an indicator of detrital input. Prado-Pérez et al. (2013) reported a positive correlation of the ²³⁰Th/²³²Th activity ratio with the calcite content and found a negative correlation of the former with quartz, micrite and thorium content. Thus, during sampling, sparite calcites with low quartz and low thorium content are preferred.

4 - Sampling and applied methods

4.1 - Sampling of lithotypes and lithofacies

13Hand samples were collected from different stratigraphic levels of the quarries from the Kıllık travertine dome (Ballık, Turkey) and the Süttő travertine complex (Hungary). Three to four samples without macroscopic indications of secondary alteration were selected in each quarry or borehole included in this study. Samples were taken at different intervals in the quarries covering the entire exposed succession. The sample suitability was evaluated by petrography, mineralogical and elemental composition as already available from several previous studies (Claes et al., 2015, 2017, 2019; El Desouky et al., 2015; Soete et al., 2015; Mohammadi et al., 2020; Török et al., 2017, 2019).

14The dominant fabrics differ according to lithotype both in case of the Ballık travertine area and the Süttő travertine complex (fig. 2). The Ballık travertines are dominated by peloidal, phyto- and dendritic lithotypes. Peloidal lithotypes are built out of organic-rich micrite clots that are surrounded by cloudy micrite and microsparites (fig. 2A-B). The phyto-lithotype samples are characterized by (possibly cement reduced) plant moulds that are surrounded by dendritic micrites or organic-rich micrites with, in-between, fabrics similar to peloidal travertines. The dendritic micrites form dense, macroscopically white fabrics (fig. 2A & 2C). In addition, a specific sampling on late stage veins and stalactite-like cements was executed in the Ballık area (fig. 2H-I). They consist of alternating dark brown, light brown and white laminae, in which often a sub-lamination (mm- to µm-thick) can be observed. The laminae occur parallel to the fracture wall or the stalactite-like features and have a variable thickness (mm- to cm-thick).

15The massive and the flat laminated travertines are the most dominant lithofacies in the Süttő travertine area (fig. 2D). These lithofacies are predominantly characterized by different kinds of micrite structures (such as clotted and dense micrite) embedded in microsparitic and sparitic calcite cement (fig. 2E). Similar to the Ballık travertine area, micropetrographic observations reveal significant microscale variations in the microtexture.

16Samples of the Cukor quarry (Süttő) are macroscopically white coloured showing sugar-like macrotexture (fig. 2F). Microscopically these samples are built up of clean sparitic calcite crystals that are present in the form of mosaic, bladed and elongated, feather-like structures (fig. 2G). One sample was taken from a brownish crystalline vein similar to the Turkish example shown in figure 2H.

17Samples were selected to be as purely calcitic as possible with absence of visible (petrography) and measurable (XRD) clays or quartz. In addition, as recommended by Mallick and Frank (2002), elemental compositions were used to avoid samples with high detrital contributions. For the selected samples of the Ballık and Süttő area, mostly a background blue or black colour is observed under cathodoluminescence microscopy. Some micrites in the Süttő area showed a dull pink to reddish cathodoluminescence. In general, micrites and sparites never showed any bright cathodoluminescence. The fabrics are similar to that of other travertines, for example to the grouped micritic peloids forming clotted textures as described by Prado-Pérez et al. (2013).

18The continental carbonate samples were cut and micro-drilled in their most homogeneous, least porous parts, i.e. in macroscopically white, micrite dominant or fully sparitic zones. The fully sparitic samples originate from the Cukor quarry (Süttő, Hungary) and from cements around stalactite-like features and veins in the Ballık (Turkey) and Süttő area (Hungary) (fig. 2).

4.2 - U/Th dating

19About 50-200 mg for each subsample was dried for U/Th dating analyses at the Isotope Laboratory of Xi’an Jiaotong University (China) (Turkish samples) and the High-Precision Mass Spectrometry and Environment Change Laboratory (HISPEC) at National Taiwan University (Taipei, Taiwan) (Hungarian samples). Standard chemistry procedures were applied to separate uranium and thorium for dating, as described in Edwards et al. (1987) and Shen et al. (2002). Samples were dissolved in 1 % HNO₃, after which U and Th were co-precipitated with a Fe carrier, separated to form a clean U fraction and a clean Th fraction using an anion exchange resin (BioRad 1-X8, 100-200 µm mesh, 0.6-10.0 ml column volume; Edwards et al., 1987). A triple-spike, ²²⁹Th–²³³U–²³⁶U, isotope dilution method was employed to correct for instrumental fractionation and determine U/Th isotopic ratios and concentrations. Measurements of U/Th isotopic ratios were performed on a multi-collector inductively coupled plasma mass spectrometer (MC-ICPMS), Thermo-Finnigan Neptune-Plus). The instrumentation, standardization and half-lives are reported in Cheng et al. (2000; 2013) and Shen et al. (2002). All U and Th isotopes were measured on a MasCom multiplier behind the retarding potential quadrupole in the peak-jumping mode. The procedures of characterizing the multiplier are described in Cheng et al. (2000). Uncertainties in U/Th isotopic data were calculated offline at 2 σ level, including corrections for blanks, multiplier dark noise, abundance sensitivity, and contents of the four nuclides in spike solution (Cheng et al., 2000; 2013). For MC-ICPMS analyses, like in this study, samples with a ²³⁰Th/²³²Th activity ratio lower than 100-300 should either be corrected for the effect of contamination or rejected (Richards & Dorale, 2003; Woodhead et al., 2006).

5 - Results

20The results of the U/Th analyses are listed in tables 1, 2 and 3. Because of the old ‘ages’ of detrital components it was assumed that their U/Th radionuclides would be in secular equilibrium (Ivanovich et al., 1992). The initial atomic ratio of ²³⁰Th/²³²Th is thus assumed at 4.4 ± 2.2 x 10^-6. This is the value for a material at secular equilibrium, with the bulk earth ²³²Th/²³⁸U value of 3.8. The errors are arbitrarily assumed to be 50%. The ²³⁰Th/²³²Th activity ratios suggest that, in line with the elemental data (Al, Th and Fe concentrations being higher than 300, 40 and 200 µg/g respectively; Claes et al., 2019), thirty-two samples from the Ballık and Süttő travertine areas contained a significant contribution of thorium from detrital sources (²³⁰Th/²³²Th activity lower than 100). Nonetheless, for all samples a detrital correction of the age was applied. Corrected ages range from 118.750 ± 1.349 to older than 550 ka in case of the Ballık area (tabs. 1 & 2) and 120.615 ± 43.152 to 553.810 ± 60.137 ka in the Süttő area (tab. 3). Most ages of the micritic or mixed micrite-sparite Ballık travertine samples plot between 400 and 550 ka or older (tab. 1), whereas in the case of the Süttő area most of the data range between 220 and 400 ka (tab. 3). The six samples of the Ballık travertine that have ²³⁰Th/²³²Th activity ratios higher than100, all fall in the age range of 450-700 ka (error included). All the sparite samples collected from late stage veins and stalactite-like structures result in ages older than 550 ka (tab. 2). In addition, most of these sparite samples show the highest ²³⁰Th/²³²Th activity ratios, indicating the lowest presence of detrital Th. In case of the Süttő travertine area, eight samples have higher ²³⁰Th/²³²Th activity ratios than 100. Their ages range between 196 ± 45 and 554 ± 60 ka. The corrected ages of both areas are plotted on relief maps showing the different quarries studied (figs. 3 & 4). In addition, a plot of ages versus topographic elevation is provided (figs. 3 & 5).

21Age ranges for samples older than the ²³⁰Th age dating limit, i.e. ~500-700 ka (depending on the uranium content), can be calculated if the ²³⁸U-²³⁴U radiogenic system is not yet in a state of secular equilibrium (~1.5 Ma; Ricketts et al., 2014). The samples from the Ballık area in table 2 correspond to such cases and thus U-series model ages can be calculated. Model ages are less precise because the initial δ²³⁴U must be estimated according to the range from the other measurements (table 1). Based on the samples with a ²³⁰Th/²³²Th activity ratio higher than 100, δ²³⁴U ranges between 72 and 309 (errors included). Model ages are further restrained to best estimate ages between 0.6-1.5 Ma, considering the (out of) secular equilibrium of the ²³⁸U-²³⁰Th and ²³⁸U-²³⁴U systems (tab. 2). A more elaborate description of this approach can be found in Ricketts et al. (2014).

6 - Discussion

6.1 - Explaining stratigraphically inconsistent U/Th ages

6.1.1 - Sample and age reliability evaluation

22Samples were selected from a large set from which those with indications of contamination or alteration, as deduced from petrography, as well as the elemental concentrations and mineralogy, were removed. The remaining selected samples from the Ballık and the Süttő area consisted of calcite, in accordance with other studies (Sierralta et al., 2010; Özkul et al., 2013; Khatib et al., 2014; Soete et al., 2015). Petrography confirmed the dominance of micrite or fully sparitic fabrics for the sampled zones and no quartz was observed. In addition, there was an absence of any strong luminescence, indicating only early or minor alteration according to Mallick and Frank (2002), under oxidizing conditions. Independent of the extent to which the samples show indications of contamination, ages were corrected.

23Age data from the Ballık and Süttő continental carbonates seem to scatter over the quarry walls (figs. 3-5). Samples in higher stratigraphic positions sometimes show older ages and are thus contrary to their stratigraphic order. In addition, the oldest ages are found for the sparitic late-stage veins and cements along stalactite-like features, which in principle should display the youngest ages. These apparent inconsistencies could be explained in various ways.

6.1.2 - Complex depositional organization

24The best indication of reliability of the reported ages for layer cake carbonates is provided by the consistency between the radiometric ages and their relative stratigraphic position (Livnat & Kronfeld, 1985). However, spring carbonates like travertines and tufa have an intrinsically whimsical nature of deposition and thus rarely follow a layer-cake depositional model. Depending on the water table, tectonic activity and local hydrological conditions, spring locations and landscape morphology can change. The deposition of the studied travertines was, however, studied in detail (Claes et al., 2015; Török et al, 2017; 2019 ; Mohammadi et al., 2020). The depositional history is easy to follow and along the sampled quarry sections are in line with the classical stratigraphic build-up from old at the bottom of the quarries to younger deposits at their top. Even considering the complex depositional nature of spring-related carbonate deposits, the age data are clearly out of stratigraphic order.

6.1.3 - Tectonically moved travertine packages

25Fault activity is important in the studied regions and could strongly affect the organization of the deposits. Hence, the tectonic history of the Ballık travertines was treated in detail by Van Noten et al. (2013, 2019) whereas structural geological observations from the Süttő travertine area are summarized by Török et al. (2017, 2019). Several observations of clear structural elements re-shaping the landscape and influencing the travertine bodies were unraveled. An important fault was revealed between the Sugar quarry and the rest of the Süttő travertine complex (Török et al., 2019). This not only explains differences in ages but also differences in fabric, i.e. the sugarlike sparitic versus the micritic-sparitic texture of the other Süttő travertines. The other reported scattered ages could, however, not be explained by any of the faults, block tilting or tectonic activities that were documented. A possible exception could be a syndepositional tectonic uplift (see part 6.1.5).

6.1.4 - Contribution of detrital thorium

26Despite careful sampling, the powders can contain non-carbonate constituents with detrital thorium. The ages were all corrected for this initial amount of detrital ²³⁰Th, corresponding to a fraction not formed by the decay from uranium. The error in age (percentage) decreases with age and increases with increasing ²³²Th/²³⁰Th ratio (i.e. the absolute age error becomes larger) (Kaufman, 1993). In our dataset, the samples that show the least indication for contamination, based on the ²³⁰Th/²³²Th ratios, result in the oldest ages, while it is the incorporation of allochthonous particles that is expected to lead to an overestimation of the age (Geyh, 2008). In addition, analyses from Sierralta et al. (2010), who used the leachate-leachate method and an isochron approach to correct for detrital Thorium, also yielded age data out of stratigraphic order. Detrital thorium alone can thus not explain the problem of the scattered ages.

6.1.5 - Open system processes

27Discrepancies between stratigraphy and radiometric ages sometimes relate to a possible opening of the system, i.e. input or loss of uranium or thorium after deposition (e.g. Drysdale & Head, 1994; Horvatinčić et al., 2000; Scholz & Hoffman, 2008). Especially highly porous, and reactive samples allow uranium to leach, resulting in older ages. In addition post-depositional incorporation of uranium into travertines (by the subsequent cementation of carbonate cements) may lead to younger apparent ages (Livnat & Kronfeld, 1985).

28Weathering causes uranium leaching and is more easily induced in high permeable rocks (Prado-Pérez et al., 2013), but in this study none of the samples displayed signs of apparent weathering. Micro-drilling was executed in the least porous, micritic parts of the samples (with exception of the fully sparitic samples), in order to exclude leaching. Based on the age results it is pertinent that diagenetic alteration has nonetheless affected several samples. Micro-scale alteration in the form of micritisation resulted from dissolution by organic acids that form due to the decay of organic matter (Claes et al., 2015; 2017; Török et al., 2017, 2019). The negative correlation of the sample quality index with micrite quantity for U/Th dating, as reported by Prado-Pérez et al. (2013), could be related to this kind of secondary micrites. While the micro-scale dissolution that resulted in micritisation could be the cause of uranium loss (Clark et al., 1992), it occurs very early in the diagenetic process and could even be syndepositional (Claes et al., 2015). Its effect will thus be the biggest in relatively young travertines, i.e. close to the three years limit of the U/Th method. For samples of 100 to 550 ka the effect on the final age dating will be minor. In addition, for the fully sparitic samples, micritisation can easily be excluded or avoided during sampling.

29Post-depositional introduction of uranium could only have resulted from newly formed cements or recrystallisation. The distinction of primary and secondary precipitated fabrics remains one of the main challenges in continental carbonate research (e.g. De Boever et al., 2017). However, the development of undulose extinction affecting micritic fabrics (fig. 6; Claes et al., 2017) indicates that recrystallisation could be an ongoing process related to post-depositional crystal growth, as suggested by Brasier et al. (2011). Sierralta et al. (2010) also reported strong evidence of recrystallisation and cementation by only partial preservation of fossil fragments and micrite grains and replacement by interlocking sparite crystals. For some locations, also cathodoluminescence can provide proof of recrystallisation (Mohammadi et al., 2019), but this was not the case for the studied sites. Although a micro-sampling was executed, a mixing of fabrics cannot be excluded. Based on petrographic similarities between the fabrics of samples showing unequivalent recrystallisation and others of the same and other deposits, the heterogeneous fabrics of the travertines in general show open system characteristics. This is further supported by the high scatter of ²³⁴U/²³⁸U versus ²³⁰Th/²³⁸U activity ratios of coeval samples on the Osmond-Ivanovich (1992) plot (e.g. Sierralta et al., 2010). In case of inclusion of post-depositional growth crystals, associated with an addition of uranium, the dated samples will result in younger ages. Its effect will dominate over that of the early micritization, especially for the ages in the presented dataset, so that the obtained ages can be considered minimum ages.

6.2 - Open system implications for the Ballik travertine ages

30Several of the Turkish micritic travertines approach the U/Th dating limit (tab. 2). In order for the ages of the micritic samples to be consistent with the stratigraphy, a minimum age of 550 ka for the travertines inside the sampled quarries of the domal area is necessary. Further evidence for this minimal age is provided by the sedimentologically youngest samples from fully sparitic late stage veins and from stalactite-like features (tab. 3). These samples are older than the ²³⁰Th age dating limit, but are not yet in a state of secular equilibrium for the ²³⁸U-²³⁴U radiogenic system. Best estimate ages for these samples range between 0.55 and 1.31 Ma. Since these ages can be considered minimum ages, the whole travertine deposit can be considered to be older than 550 ka. The age result of the Kömürcüoğlu quarry (490 ± 50 ka), obtained by Özkul et al. (2004), is in agreement with the ages obtained from micritic travertines in this study. Given that the data are considered minimum ages, they are consistent with the older ages from Lebatard et al. (2014). The Ballık area travertines consequently, not only are the most voluminous, but also represent the oldest travertines reported from the Denizli basin.

6.3 - Open system implications for the Süttő travertine ages

31U/Th dating resulted in scattered ages in the Süttő travertine area. The age data measured and calculated from borehole profiles are strikingly not in line with the stratigraphy. Remarkably, however, an interesting trend can be observed between the individual outcrops and their elevation above sea level (fig. 5). Younger ages are situated in the lower part of the lithological column in most of the cases. The Cukor quarry situated on the top of the Haraszt Hill represents the oldest (min. 554 ± 60 ka) and, as discussed above, represents a separate travertine body in the area. The sample ages from the other outcrops are overlapping, but yield a clear trend of younger ages towards the River Danube, i.e. towards lower topographic elevations. This observation is in line with the terrace stratigraphy of Pécsi (1959) and Kretzoi and Pécsi (1982). It is in this context that a continuous syndepositional uplift throughout the Pleistocene-Holocene, and an associated downward shift of the groundwater system and their springs, could be invoked to explain the anomalies from the law of superposition. Cementation and recrystallisation could take place in a later stage in the relatively lower parts compared to the higher parts of the travertine deposit. These age data thus further support the complexity of travertines, and continental carbonates in general, even after deposition.

32The presented U/Th age data of the Süttő travertine area are in line with the ones of Kele (2009) and Sierralta et al. (2010) and in agreement with optically simulated luminescence (OSL) dating of the overlying loess (Novothny et al., 2011; Rolf et al., 2014). In contrast to the U/Th data set presented in this study, Sierralta et al (2010) excluded only a few samples based on open system conditions. Their remaining samples, however, showed the clearest signs of recrystallisation with ghost structures of fossils and micrite grains being replaced by sparite crystals, and ages after correction still out of stratigraphic order. Due to the introduction of the younger sparite fabrics, similar to this study, the resulting ages can only be considered minimum ages. The age of the Süttő travertine complex was also estimated based on the fossil content by Pazonyi et al. (2014), suggesting a long period of travertine formation, with a maximum interval from 2600 to 780 ka (fig. 7). The underlying beds of the travertine are most probably Upper Pannonian sediments redeposited by slope mass movements during the Pliocene. They have a minimum age of 2.6 Ma suggesting that travertine formation possibly started right after the development of the terrace horizon. The U/Th travertine ages are thus systematically younger compared to other dating methods (see also Ruszkiczay-Rüdiger et al., 2018), in line with the effect of cementation and recrystallisation during open system conditions.

6.4 - Implications for application of u/th dating on continental carbonates in general

33Due to their porous nature and carbonate mineralogy, continental carbonates are extremely susceptible to post-depositional processes. The conditions and processes during deposition and early meteoric diagenesis are very similar, making distinction between primary and secondary fabrics challenging. Pristine fabrics in most deposits will not even be preserved at all. As with regard to dating, a proper paragenesis is key to correct interpretation of the ages. It could be questioned if U/Th dating, as now commonly applied on travertines and tufa, should be ceased. Bad practices, like for example dating only one or few samples to date an entire outcrop, must be avoided at all costs. A further warning criteria to be considered includes every kind of petrographic or chemical signs of recrystallisation, including, but not limited to, heterogeneous micrite-sparite fabrics (as typical for most tufa and travertines), sweeping extinction over micrite fabrics and ghost structures of fossils or grains. Fabrics displaying these criteria will result in minimum ages for the deposits. The extent of bias towards younger U/Th ages strongly depends on the extent that diagenetic processes could affect the deposits and so younger travertines in general will be less affected. Ruszkiczay-Rüdiger et al. (2018), recognized that Late Pleistocene travertines (ages of ~100 ka) were still in line with other dating methods (cosmogenic nuclides, luminescence, paleontology and magnetostratigraphy) while travertines older than the last interglacial showed significantly younger U/Th ages. This significantly reduces the successful dating range of the method but explains its successful use in several young travertine cases (e.g. Antoine et al., 2007; Schulte et al., 2008; Zentmyer et al., 2008; Scotti et al., 2014; Wang et al., 2017). Older age data of the typical travertine fabrics should be critically re-evaluated and re-interpreted in terms of minimum ages. Even considering them as minimum ages might hamper their application since the older the travertines are, the more the resulting ages will be spread over the range of the dating method. As a result, it is often easier to focus on cements and veins consisting of larger sparite crystals. On the one hand because of their, usually, more limited detrital contamination. On the other hand, because of their less complex fabrics, which are less prone to diagenetic alteration due to their lower surface reactivity, cements and veins are subject to a more straightforward evaluation of diagenetic alteration. For verification, several and preferably at least three coeval samples should be included in the sample set. Several studies demonstrated the successful application of U/Th method on larger crystalline fabrics (e.g. Uysal et al., 2007; Karabacak et al., 2019; Yildirim et al., 2020). These cement ages, however, will still only result in minimum ages for the whole deposits and so, if the age of the deposit is the main interest, it should still be combined with maximum ages obtained with other dating methods, often on the surrounding lithologies. As contradictory as it seems, if the age of the travertine deposits is of interest, focus should be on the intercalating and surrounding lithologies or veins, from which the former are dated with dating methods different from U/Th techniques.

7 - Conclusions

34Despite adapted sampling and detrital thorium corrections, age data from the Ballık and the Süttő travertine areas are inconsistent with the classical stratigraphic position of the samples. It is therefore concluded that the micritic and mixed micritic-sparitic travertines behaved as an open system for uranium after deposition. The observation of wavy extinction affecting micritic fabrics is an indication of recrystallisation. Given these arguments and in order for the sample ages to respect their stratigraphic position, the U/Th ages can be considered as minimum ages of the sampled continental carbonates. The openness of the system has also possible implications for other geochemical studies.

35Fully sparitic samples are more suited for U/Th dating. On the one hand because of their, most of the time, more limited detrital contamination. On the other hand, because of their less complex fabrics, which are less prone to diagenetic alteration due to their lower surface reactivity, sparitic samples are subject to a more straightforward evaluation of diagenetic alteration. Analysis of multiple samples from the same study site, including coeval samples, is strongly recommended. These cements and veins, however, can still only be considered as minimum ages for the main travertine formation.

36Only dating based on several methods combining U/Th dating with ages of the surrounding lithologies and based on fossil content can provide the reliable data set for the age estimation of the travertine body that is necessary for its application in palaeoclimate studies.

37Based on the presented data from this study, together with literature data, the Ballık travertines most likely precipitated before 0.55-1.31 Ma ago, when late-stage cementation along fractures and in cavities took place. The formation of the Süttő travertine area started at the earliest around 2.5 Ma with precipitation of the Cukor travertine. With some short interruptions, active travertine formation in the area seems to have been continuous and at its maximum until 120 ka. The younging trend of the Süttő travertines towards the Danube, strongly suggests late cementation and recrystallisation controlled by a lowering of the groundwater table, which ultimately resulted in anomalies with regard to the law of superposition.