Reconstruction of Middle to Late Quaternary sea level using submerged speleothems from the northeastern Yucatán Peninsula

We examined 14 subaerially deposited speleothems retrieved from submerged caves in the northeastern Yucatán Peninsula (Mexico). These speleothems grew during the Middle to Late Quaternary and were dated by 230Th‐U techniques to provide upper depth limits for past sea levels. We report the first relative sea‐level limits for Marine Isotope Stages (MIS) 11 and 6, and present new evidence for sea‐level oscillations during MIS 5 and early MIS 1. For the latter periods, the origin of growth interruptions is evaluated by combining petrographic methods with trace element analyses. The MIS 5c sea‐level highstand probably occurred between 103.94 ± 0.58 ka and 96.82 ± 0.42 ka and must have exceeded ‐10.8 m (relative to present‐day local sea level). The minimum average rate of sea‐level fall over a 9.4 ka‐long period during the MIS 5e/5d transition is calculated from stalagmite and published coral data at 1.74 ± 0.37 m/ka. For the early Holocene, previous discrepancies with respect to a potential multimetre oscillation of local sea level were found to be challenging to reconcile with the existing speleothem data from the area.


Introduction
Reconstructing past sea-level changes is a fundamental component in the study of global climate change (e.g. Kemp et al., 2015;Carlson et al., 2019;Capron et al., 2019). Several archives are used to reconstruct past sea levels, such as warm-water corals, marine terraces and shoreline angles (e.g. Muhs et al., 2012;Hibbert et al., 2016). They contribute to a growing sea-level database, which combines and synergises the results from individual studies (see Düsterhus et al., 2016). Subaerially deposited stalagmites, subsequently submerged by rising sea levels, also provide important constraints on the upper limit of sea-level highstands (e.g. Richards et al., 1994;Onac et al., 2012). Precise 230 Th-U dating allows for the accurate determination of the start and end of speleothem growth interruptions (Scholz and Hoffmann, 2008;Cheng et al., 2013). The cause of speleothem growth cessations may be a change towards an arid climate, changes in local karst hydrology, or submergence (Kaufmann, 2003). While speleothem layers containing marine serpulid worm tubes provide direct evidence of flooding by seawater (Antonioli et al., 2004;Surić et al., 2005;Dutton et al., 2009;Moseley et al., 2015;Miklavič et al., 2018;Dumitru et al., 2019), the absence of these incrustations during a growth stop does not necessarily exclude temporary submergence of the speleothem. In these cases, petrographic and trace element analyses can be valuable tools to distinguish submergence from other causes (Surić et al., 2009;Wainer et al., 2017). A 'dry' hiatus is often the result of a progressive drying of the drip site, which may be associated with an increase in the elemental ratios such as Mg/Ca, Sr/Ca or Ba/Ca (Fairchild and Treble, 2009). In addition, such growth interruptions are often marked by a detrital layer with high concentrations of Al, for example (Fairchild and Treble, 2009). In contrast, a hiatus due to submergence must not necessarily exhibit such a signature since the water prevents deposition of detrital material on the stalagmite surface. In the case of submergence, erosion by water can lead to abrasion of the crystal surfaces, or in the case of a corrosive environment, the dissolution of previously deposited carbonate (Back et al., 1986, Fairchild andBaker, 2012).
The northern Yucatán Peninsula, Mexico, holds the world's most extensive submerged karst systems. They reach depths of more than 100 m below the present sea level and these underwater caves are heavily decorated with speleothems. The region is assumed to be tectonically inactive and stable (Szabo et al., 1978;Zúñiga et al., 2000;Smart et al., 2006) and is located about 1000 km to the south of the maximum extent of the Pleistocene Laurentide ice sheet (Licciardi et al., 1998). This is important as the western North Atlantic-Caribbean region is defined by complex glacio-isostatic adjustment (GIA) processes related to the waxing and waning of the Laurentide ice sheet (Potter and Lambeck, 2004;Creveling et al., 2017). This causes vertical movements of the continental crust that play an important role in relative sea-level changes. Throughout the Quaternary, the Yucatán Peninsula did not reach a GIA-equilibrium state (Potter and Lambeck, 2004) because the equilibrium was interrupted by subsequent glacial cycles (see Horton et al., 2018). Thus, predictions of the future development cannot solely be inferred from the current nonequilibrium state but have to consider the history depending on past relative sea-level reconstructions.
Previous studies provided important first information on different timescales, and highlighted the potential of the Yucatán Peninsula for sea-level reconstruction. Moseley et al. (2013) presented data regarding the evolution of sea level for Marine Isotope Stage (MIS) 5 and calculated rates of sea-level fall between MIS 5e and MIS 5d by combining speleothem data with marine-limiting data from Blanchon et al. (2009). Hering et al. (2018) and Collins et al. (2015b) provided constraints on the early Holocene sea-level rise in the region, while Moseley et al. (2015) documented four submergence events at 8.9, 8.6, 8.4 and 6.0 thousand years (ka) ago by dating marine deposits interlayered in speleothems. These data suggest that rapid sea-level fluctuations of several metres' amplitude occurred between 10 and 8 ka (Moseley et al., 2015;Hering et al., 2018). Other information about past sea-level changes on the Yucatán Peninsula is mostly limited to MIS 5e and the middle to late Holocene, as well as past interglacials (e.g. Szabo et al., 1978;Gabriel et al., 2008;Khan et al., 2017).
We here present several 230 Th-U-dated speleothems from the northeastern Yucatán Peninsula, Mexico, and use petrographic and trace element data to study the origin of growth interruptions in these deposits. Our results underscore the great potential of the Yucatán Peninsula for the reconstruction and spatial delimitation of Quaternary sea-level changes in the Caribbean region.

Study area
The cave systems in the area of Tulum,Quintana Roo,.6°W, Fig. 1) developed in a carbonate platform of about 3,000 m thickness, reaching back to the early Cretaceous (Ward et al., 1985). Tectonic stability during the late Pleistocene is supported by similar ages for corals at about the same elevations as other areas in the Caribbean (Szabo et al., 1978). Furthermore, a compilation of earthquake data shows that the northern Yucatán Peninsula is aseismic (Zúñiga et al., 2000), and evidence based on studies of Cenozoic limestones suggests that little deformation has occurred since their deposition (Smart et al., 2006). Intensive karstification affected the uppermost >100 m of the carbonate sequence composed of almost horizontally layered thick-bedded reef and shallow-marine limestone of Plio-to Miocene age (e.g. Ward et al., 1985).
Groundwater exists as a freshwater lens above saline water intruding from the coast, with a direct connection and high permeability to the Caribbean Sea in the east (Perry et al., 2002). The water table in the Tulum area is known to be roughly equal to the sea level of the nearby coast due to the low hydraulic gradient of 1-10 cm/km (Bauer-Gottwein et al., 2011;Beddows et al., 2007 and references therein). The bathymetry along the coastline shows a moderately sloped ocean floor of no less than 10 m/km (Kjerfve and de Campeche, 1994).
The Holbox fracture zone is the only regional tectonic feature with significant influence on groundwater hydrology (Bauer-Gottwein et al., 2011). This zone crosses the northeastern Yucatán Peninsula in an approximately north-south direction and is located, in the area of Tulum, approximately 10 km inland (Smart et al., 2006). At sites located closer to the coast, the groundwater flows towards the sea, i.e. in an easterly direction, whereas further inland, groundwater flows partly along the Holbox fracture zone in a north to northeast direction (Tulaczyk, 1993;Perry et al., 2002), parallel to the coastline, which agrees with the dominant orientation of caves in this area (Fig. 1).

Sample collection and petrographic methods
Fourteen stalagmites with lengths between 2 cm and 34 cm (Figs A.1 and A.2) were collected in situ by technical divers in the Chan Hol, Naharón, Quelonios and Pat Jacinto caves ( Fig. 1 and Table 1). These speleothems were found at water depths between 0 and 28.7 m. Water depth was measured with a dive computer at the base of each stalagmite with an uncertainty of ±0.1 m. Speleothems were cut along their extension axes and macroscopically examined. The petrography was examined using thin sections with a transmitted-light microscope (Zeiss Axio Imager M2). Subsamples were subject to high-precision 230 Th-U dating and geochemical analyses were conducted at growth interruptions.

Th-U dating
Speleothem growth phases were determined using the highprecision multicollector inductively coupled plasma mass spectrometer (MC-ICPMS) 230 Th-U dating technique. Subsamples of 80-140 mg were cut along the growth axis using a diamond wire saw. The methods of chemical sample preparation, U and Th purification, mass spectrometric analysis and applied corrections followed those of Douville et al. (2010), Arps (2017) and Wefing et al. (2017). Isotope ratios were determined with a MC-ICPMS (Thermo Fisher Neptune plus ) at the Institute of Environmental Physics, Heidelberg University. Calculations of activity ratios and 230 Th-U ages (Table 2 and https://doi.pangaea.de/10.1594/PANGAEA.930526) were performed using the half-lives of Cheng et al. (2013). Ages are quoted as ka before the year 1950 CE, and final errors do not include half-life uncertainties. The correction for the amount of initial 230 Th is of importance for young samples and samples with a high detrital content. For the detrital 230 Th correction a ( 230 Th/ 232 Th) activity (A) ratio of 3.5 ± 1.3 was chosen based on the analysis by Moseley et al. (2013) and secular radioactive equilibrium within the uranium decay chain was assumed.
Where applicable, age extrapolation to a petrographic boundary layer (growth hiatus) relies on a linear regression of the two ages closest to this layer on either side. Due to sampling close to these boundary layers, all extrapolated ages only reflect a minor adjustment as they fall within the statistical uncertainty band of the closest 230 Th-U age.

Trace element analysis
Trace element analyses were performed in order to identify possible geochemical alteration associated with hiatuses in the speleothems. A laser ablation ICP-MS with an ArF excimer laser (NWR193 UC by New Wave Research) coupled to an inductively coupled plasma quadrupole mass spectrometer (Thermo Fisher iCAP-Q) at the Institute of Environmental Physics, Heidelberg University, was used. Laser tracks are parallel to the stalagmites' growth axes and cover transitions across relevant hiatuses. The rectangular laser spot was 15 × 150 µm (with the larger side being vertical to the track) and moved at 15 µm/s along the track. The repetition rate of the laser pulses was 20 s -1 . Each isotope was measured every 100 ms. The applied method closely followed the procedures outlined by Jochum et al. (2012). Background counts were measured with the laser in off mode and subtracted from the raw data. Measured signals of trace elemental isotopes ( 23 Na, 24 Mg, 27 Al, 31 P, 34 S, 57 Fe, 88 Sr, 138 Ba, 232 Th, 238 U) are given relative to the 44 Ca signal of the speleothems, which are assumed to have a constant Ca concentration. Subsequently, signals are standardised against the NIST SRM 612 glass standard using the recommended concentrations of Na and Al published by Evans and Müller (2018) and for all other elements by Jochum et al. (2011). The resulting element-to-Ca ratios are presented as mass ratios. In order to remove major outliers, the data were filtered by applying a floating median of seven data points (equivalent to moving the laser spot by 9 µm). The data will be made available upon request.

Periods of speleothem growth
The growth episodes of 14 speleothems from the Chan Hol (CH), Naharón (NAH), Quelonios (QUE) and Pat Jacinto (PJ) caves fall into four distinct periods within the last~421 ka (Fig. 2). The association of MIS with growth phases is based on the chronology of Lisiecki and Raymo (2005) and the substage numbering is based on Railsback et al. (2015).
Stalagmite PJ10 has a length of 3 cm. Three measurements distributed over 2 cm are not in stratigraphic order but partly overlap in their uncertainty range. Their average age of 421 ± 15 ka places the growth phase into early MIS 11. A relatively low U content of 352 µg/g (average of the three measurements), a low ( 230 Th/ 232 Th) A of <276 and its high age hamper a higher precision of the corrected ages.  Stalagmite NAH11 is 11 cm long and grew between 172.5 ± 1.9 ka and 141.6 ± 0.9 ka. Stalagmite PJ1 has a length of 9 cm and identical base and top ages within uncertainty of 166.8 ± 1.9 ka and 166.5 ± 3.2 ka, respectively. Stalagmite PJ4 measures 2 cm and is dated to 135.8 ± 0.8 ka. Therefore, these three stalagmites grew during MIS 6.
Stalagmite PJ17 (length of 6 cm) has a basal age of 113.43 ± 0.90 ka, which marks the start of growth and coincides with the transition from MIS 5e to MIS 5d. Its top age of 112.70 ± 0.85 ka overlaps with the basal age within the uncertainty range. Stalagmite QUE01 is 11 cm long (Fig. A.1). Its base was measured twice (#9710 and #10215) and replicated within uncertainty. Measurement #10215 was conducted with more sample material that improved the uncertainty and is therefore the one used. Growth started at 109.16 ± 0.29 ka (base age of #10215) and ended at 90.88 ± 0.57 ka (top age). Additional ages indicate a hiatus (Fig. A.1) at 99 mm distance from top (dft), which encompasses 103.94 ± 0.58 ka to 96.82 ± 0.42 ka (duration 7.12 ± 0.72 ka). Petrographic analyses of this hiatus are reported below. Stalagmite PJ8 with a length of only 2 cm yielded ages of 105.7 ± 1.5 ka and 107.7 ± 1.4 ka for its base and top, respectively. Stalagmite NAH07 (length of 6 cm) also has identical base and top ages within uncertainty of 91.98 ± 0.33 ka and 91.57 ± 0.46 ka, respectively. The growth phases of PJ17, QUE01, PJ8 and NAH07 thus correspond to MIS 5 (Fig. 3). Stalagmite PJ16 (length of 6 cm) grew from 70.88 ± 0.40 ka (base age) to 68.30 ± 0.41 ka (top age), which corresponds to the transition from MIS 5 to MIS 4. Stalagmite PJ2 (8 cm) grew from 64.19 ± 0.49 ka (base) to 63.85 ± 0.44 ka (top) and thus corresponds to MIS 4. Speleothem PJ7 is 2 cm long and its base was dated to 14.95 ± 0.09 ka. The top age is compromised by a high detrital contamination (13.54 ± 0.92 ka). Stalagmite NAH14 (15 cm), studied by Warken et al. (2021), grew from 11.04 ± 0.09 ka to 9.52 ± 0.04 ka without interruption. The age of 11.58 ± 0.05 ka was excluded from the age model due to a probable contamination with old carbonaceous material (see Warken et al., 2021). CH8 is a stalagmite of 34 cm length. It grew from 9.40 ± 0.11 ka to 8.91 ± 0.08 ka, with no indication for growth cessations. CH10 (32 cm) has a basal age of 18.61 ± 0.12 ka. The top age of 8.57 ± 0.10 ka is an average of the two measurements 4.5 mm apart and their ages overlap within their uncertainties. A 7.1 kalong hiatus was identified 5 cm below the top between measurements #10223 (17.54 ± 0.07 ka) and #10221 (10.15 ± 0.08 ka) alongside a macroscopic change in colour (Fig. A.1). Further dating in combination with petrographic analyses (below) identified two more hiatuses from 9.74 ± 0.19 ka to 9.26 ± 0.08 ka (duration of 0.48 ± 0.21 ka) and from 9.02 ± 0.06 ka to 8.57 ± 0.10 ka (duration 0.45 ± 0.12 ka).
The detrital correction shifts the final ages of a few subsamples of CH8, CH10 and PJ7 by more than the 2σ statistical uncertainties. Apart from these three stalagmites and the MIS 11 stalagmite PJ10, measured ( 230 Th/ 232 Th) A for all other speleothems are >310, i.e. the ages of these samples are unaffected by the initial 230 Th.

Petrography and geochemical composition of speleothem hiatuses
Macroscopic growth boundaries and hiatuses revealed by 230 Th-U dating were investigated by thin-section petrography and trace element analyses in order to assess whether they reflect submergence events.
Stalagmite QUE01 (Fig. A.1) shows a very thin and slightly lighter layer at the hiatus at 99 mm dft. However, based on plain and cross-polarised light microscopy (Fig. A.4), no distinct break in the crystal growth pattern was identified at this  position. Moreover, the growth of columnar elongated crystals shows no clear interruption across this hiatus. Renucleation of crystals is only observed at the flank of the stalagmite in a small but unconfined range above and below the hiatus where the older layers suggest signs of dissolution (Fig. A.4). Trace element data across this boundary reveal two different patterns (Fig. A.3). Post-hiatus calcite is characterised by higher U/Ca and Mg/Ca and lower Sr/Ca and Ba/Ca ratios compared with pre-hiatus calcite, with a sudden shift (<100 µm) at the boundary (99 mm dft). Ratios of Mg/Ca, Al/Ca, Fe/Ca and Th/Ca are slightly elevated at the hiatus compared with the mean. The magnitude of changes, however, is comparable to the variability before and after the growth stop. These data therefore suggest that no significant increase in detritus is associated with this hiatus. Thus, the growth hiatus at 99 mm is primarily identifiable by an age jump of 7.1 ka and by the trace element data, but is not marked by a clear break in crystal growth nor an enrichment in detrital particles. CH10 shows numerous heterogeneous, thin (a few 100 µm to a few mm) layers in the topmost 5 cm of the stalagmite (i.e. the 10.15 ka to 8.57 ka interval) (Fig. A.1). These layers present a wide variety of crystal structures, colours and detrital content. Nevertheless, abrupt changes and breaks in the crystal structure are only identified in three distinct layers (L1-L3, Figs A.6, A.7). The two younger ones (L1 and L2) are only 0.7 mm apart and their respective contribution to the age gap cannot be resolved by the 230 Th-U chronology. L1 shows no significant increases in trace elements, but L2 and L3 yield elevated to high concentrations of all trace elements (Fig. A.5). Based on these results, the younger cessation period (between 9.02 ± 0.06 ka and 8.57 ± 0.10 ka) is thought to be reflected by hiatus L2 at 9.5 mm, while the third break in crystal growth (L3) marks the older cessation period (between 9.74 ± 0.19 ka to 9.26 ± 0.08 ka) at 26.2 mm.

Yucatán stalagmites as sea-level limiting data points
In principle, the ages of stalagmites provide upper limits for past sea levels, i.e. terrestrial-limiting points. In the following  . Elevation relative to present-day local sea level of dated MIS 5 speleothems plotted against their 230 Th-U ages. Six stalagmites from this study (in red) and stalagmites from Moseley et al. (2013, in black) are from the area of Tulum (Fig. 1). Ages bracketing the sea-level highstand during MIS 5c in Bermuda from Wainer et al. (2017) are shown in green. Relative sea level in Bermuda is expected to differ from that at the Yucatán Peninsula by >10 m due to glacial isostatic adjustment (Potter and Lambeck, 2004 discussion we first investigate the uncertainties associated with these limits and then explore whether observed hiatuses may indeed be linked to cave flooding and hence sea-level highstands.

Dating uncertainty of sea-level limiting data points
A major source of uncertainty is the accuracy of the 230 Th-U ages, which depends on the purity of the carbonate samples indicated by the measured ( 230 Th/ 232 Th) A ratio and the applied correction model to account for initial Th. The value chosen here as initial ( 230 Th/ 232 Th) A of 3.5 ± 1.3 has been found adequate for stalagmites from this region (Moseley et al., 2013, Warken et al., 2021 and is also within the uncertainty equal to the value of 3.96 ± 0.2 derived by an isochron approach for a flowstone from the Chan Hol cave (Stinnesbeck et al., 2020). Based on stratigraphic reasoning, the significantly contaminated sample 9703 (PJ7) provides an indication for an initial ( 230 Th/ 232 Th) A value of 1.5 or higher, while a value of 14.4 ± 4 from another regional study (Moseley et al., 2015) would result in significant age inversions for CH10 and CH8. Based on these considerations we regard all presented 230 Th-U ages to be reliable within the quoted (2σ) uncertainty.

Spatial uncertainty of sea-level limiting data points
Secondly, the position of the stalagmites relative to each other and the sea level defines the validity of our set of speleothems as sea-level limiting data points. This may be affected by GIA processes, for example, or changes in the hydraulic connection between the cave and the sea. Since all caves and stalagmites studied here were collected from an area within 20 km of Tulum (Fig. 1), GIA must have had a similar influence on all speleothems revised here. Consequently, all stalagmites from this, as well as from the studies by Moseley et al. (2013Moseley et al. ( , 2015, are treated as being from the same locality. For comparability within the Caribbean and on a global scale, however, regional GIA has to be considered (Potter and Lambeck, 2004). Due to the peninsula-wide present-day hydraulic gradient of <10 cm/km (Bauer-Gottwein et al., 2011;Beddows et al., 2007), the water table in the cave systems is slightly elevated relative to the Caribbean Sea. These physical properties are assumed to have been constant throughout the past four glacial-interglacial cycles represented by the speleothem deposition discussed here. A lower sea level causes the coastline to advance seawards and increases the distance between sample positions and the ocean. This increases the offset between the water table in the cave and that of the ocean, because the hydraulic gradient is assumed to be constant. For a relative sea level of −28.7 m, which corresponds to the deepest sample of this study, the coastline advanced to the east by about 4 km (inferred from Kjerfve and de Campeche, 1994). Therefore, considering the upper limit for the hydraulic gradient, the maximal additional hydraulic head is 0.4 m. In the direction of the interior of the Yucatán Peninsula, this correction is reduced by a declining hydraulic gradient. Given the high relative uncertainties of such a correction, and the fact that it would be associated with a depth correction of only up to a few tens of centimetres, we have chosen to refrain from a hydraulic head correction (see Moseley et al., 2013). The seasonal variability is no more than 0.2 m (see Collins et al., 2015a). We expect the total systematic uncertainty to be comparable to ±0.5 m exceeding the mere measurement uncertainty of ±0.1 m.

Speleothem growth cessations as evidence for sea-level highstands
Marine submergence as a cause for speleothem growth cessation can be indicated by marine-limiting data from a higher elevation. Additionally, petrographic investigations are used to characterise associated hiatuses and provide a useful tool to argue for their cause. In this study, we investigated growth stops from two different stalagmites (CH10 and QUE01).
Stalagmite QUE01, collected at 10.8 m depth, grew during MIS 5, but its deposition was interrupted between 103.94 ± 0.58 ka and 96.82 ± 0.42 ka. There is evidence for a MIS 5c sea-level highstand of > −10 m at 108.5 ka (Dodge et al., 1983) and > −8 ± 2 m at 107.3 ± 5 ka (Dumas et al., 2006) based on coral growth from Haiti. According to calculations of the isostatic response to deglacial unloading (Potter and Lambeck, 2004), relative sea level in northern Haiti was similar (<2 m difference) to that of Tulum for MIS 5c. These highstands coincide with the QUE01 growth cessation within dating uncertainties. The Haiti corals were collected from depths shallower than −10.8 m (level of QUE01), thus we attribute the growth stop to submergence due to a rise in sea level. QUE01's hiatus is marked by a millimetric and slightly brighter layer as well as a sudden change in trace element composition. Stable growth conditions before and after the hiatus are indicated by elongated columnar crystals since these fabrics require (amongst other factors) a constant drip rate to form (see Frisia, 2015). There is no evidence for the existence of a detrital layer based on the element data (Fig. A.3) or a break in crystal growth based on thin-section analyses (Fig. A.4). The absence of a clear break during a prolonged growth cessation is counterintuitive but might be explained by submergence in the freshwater lens. A very slightly aggressive freshwater environment (Back et al., 1986) could be reflected by the newly formed crystals at the flank of the stalagmite (see Fig. A.4). Uninterrupted stalagmite growth during that time at a higher elevation (Moseley et al., 2013) strengthens the argument for submergence of QUE01 in the freshwater lens caused by a risen sea level. Given the potentially corrosive environment and an associated possible loss of previously deposited calcite, 103.94 ± 0.58 ka is regarded as an upper limit for the start of the hiatus.
In the CH10 stalagmite, two hiatuses from 9.74 ± 0.19 ka to 9.26 ± 0.08 ka and from 9.02 ± 0.06 ka to 8.57 ± 0.10 ka, were identified. Coeval deposition of NAH14 at −17.7 m up until 9.52 ± 0.04 ka is a strong argument against submergence during the older of the two cessation periods (Fig. 4). The hiatus of the younger cessation period is characterised by a clear break in crystal growth and a distinct detrital layer similar to the older hiatus, but corresponding sea-level data are ambiguous (discussed below). Consequently, the cessation periods in stalagmite CH10 cannot be confidently attributed to submergence. Interruptions of stalagmite growth are often associated with dry climatic conditions (Fairchild and Baker, 2012). Changes in the early Holocene precipitation amount have been documented for the Yucatán Peninsula and the western tropical Atlantic region and interpreted to be related to North Atlantic climatic oscillations (Hillesheim et al., 2005;Bernal et al., 2011;Fensterer et al., 2013;Warken et al., 2021). On centennial to millennial timescales, the so-called Bond events, basin-wide cold relapses in the North Atlantic, are recorded by fluctuations in the abundance of hematite-stained, ice-rafted grains and occurred at a recurrence rate of~1500 years (Bond et al., 2001). They are related to a weakening of the Atlantic meridional overturning circulation and a southward shift of the intertropical convergence zone leading to enhanced aridity in the Caribbean region (Knight et al., 2006;deMenocal, 2000;Hillesheim et al., 2005;Bernal et al., 2011;Fensterer et al., 2013). The three CH10 growth phases, from 10.18 ± 0.08 ka to 9.74 ± 0.19 ka, 9.26 ± 0.08 ka to 9.02 ± 0.06 ka and around 8.57 ± 0.10 ka, and the CH8 growth from 9.40 ± 0.11 ka to 8.91 ± 0.08 ka, occurred during Bond events 5-7, while absence of growth in CH8 and CH10 (orange shading in Fig. 4) rather coincides with presumably more humid phases in-between. This observation is counterintuitive to the common interpretation that speleothem growth phases are associated with more humid phases. The high resolution proxy record of speleothem NAH14 suggests that the early Holocene climate on the northern Yucatán Peninsula was characterised by a much more variable climate, which cannot solely be attributed to Bond events (Warken et al., 2021). In their study, pronounced swings between dry and wet conditions on the decadal to centennial scale were identified. In particular, a link of local drought variability to the centennialscale solar Suess-de Vries cycle was identified, with generally drier but very variable conditions occurring during periods of high solar activity and vice versa. However, no clear connection between CH10 growth phases and hiatuses and solar activity was found.
In addition to the possibility of growth cessation due to arid conditions, we alternatively propose that the repeated early Holocene growth interruptions of the CH10 stalagmite could have occurred due to a very variable climate and/or even during pluvial episodes. The high concentration of Al and Fe at the hiatuses suggests the deposition of detrital particles during growth cessation, which could also be interpreted to result from intermittent cave flooding. The sampling site is situated west of the Holbox fracture zone, which separates the groundwater drainage system of Chan Hol from the coastal areas of the Caribbean Sea and therefore diminishes the eastward flow of groundwater into the ocean. This delayed drainage along the Holbox fracture zone may have formed a barrier that could have enabled precipitation-induced cave flooding at Chan Hol during a more pluvial climate. This hypothesis is based on the present-day setting of a step-like increase in the water table and groundwater drainage along the Holbox fracture zone (Tulaczyk, 1993;Bauer-Gottwein et al., 2011).
In summary, the character of the hiatuses in CH10 is very different to the hiatus in QUE01. For QUE01 we suggest prolonged submergence as the cause for the growth cessation.

Middle to late Quaternary sea-level record based on speleothems from Tulum
The 230 Th-U data of subaerially deposited speleothems from the Tulum area improve our understanding of sea-level change in this region of the Yucatán Peninsula and provide upper limits for local sea level over the last four glacial cycles. The hiatus determined in the QUE01 stalagmite between 103.94 ± 0.58 ka and 96.82 ± 0.42 ka is considered as a lower limit because submergence is the suggested cause of cessation.

MIS 12/11 and MIS 6
The mean age of speleothem PJ10 (0 m) is 421 ± 15 ka, i.e. the MIS 12-11 transition. This may be the oldest reported speleothem from the region. Speleothem deposition during MIS 6 yielded the first limit on sea level for this period: Speleothem PJ4 (−28.7 m) constitutes the lowest data point of this record at 135.8 ± 0.8 ka, while NAH11 (−17 m) provides data from 172.5 ± 1.9 ka to 141.6 ± 0.9 ka. Even though the samples analysed here do not cover the transition at the end of MIS 6 and are tens of metres above eustatic sea-level estimates for that period (Spratt and Lisiecki, 2016), they still provide evidence for speleothem growth at different stages of this glacial period. Since the cave system reaches depths of more than 100 m below the present-day sea level, speleothems relevant to delimit sea level during MIS 6 may be recovered in future investigations.

Constraints on MIS 5 sea level in the Caribbean region
Evidence of speleothem growth during MIS 5 is preserved in stalagmites PJ2, NAH07, PJ16, PJ8, QUE01 and PJ17 (Fig. 3) from which the latter two provide new limits of local relative sea level. PJ17 grew during a declining sea level between MIS 5e and MIS 5d and thus provides a new constraint on local sea level at 113.43 ± 0.90 ka to below −11.5 m. . Height relative to present-day local sea level of dated speleothems (CH8, CH10 and NAH14) plotted against their 230 Th-U ages. Vertical orange bars mark the absence of growth in stalagmites CH8 and CH10. 14 C-dated charcoal samples from the Chan Hol cave are from Hering et al. (2018). The three pairs of ages (Oscillation events) mark 230 Th-U-dated calcite in speleothems bracketing flooding events with the ellipses of the respective colour marking the maximum age range of submergence (Moseley et al., 2015). Fluctuations in the abundance of hematite-stained grains (a proxy for drift ice) and the corresponding Bond cycles 5-7 (Bond et al., 2001) are shown. The sea-level reconstruction for the Caribbean side of Mexico from Khan et al. (2017) is plotted for reference [Color figure can be viewed at wileyonlinelibrary.com].
Using marine-limiting coral data from MIS 5e and deeper terrestrial-limiting data from the following interval of a lower sea level (MIS 5d), it is possible to calculate a minimum average rate of sea level fall across the glacial inception. Accordingly, Moseley et al. (2013) calculated a rate of sealevel fall of 1.9 ± 0.4 m/ka between 122.8 ± 0.6 ka and 117.7 ± 1.4 ka based on their oldest speleothem sample (117.7 ± 1.4 ka at −4.9 m) and corals from the Yucatán Peninsula shelf margin about 40 km to the north of our site (Blanchon et al., 2009). They used the average age of 122.8 ± 0.6 ka of samples xD4-2 and xD4-3 (+4.8 m) of Blanchon et al. (2009) for the marine limit and considered a ± 2 m uncertainty for both the corals and the stalagmite. The considered time period and the sea-level change are almost doubled when using the sample PJ17 of our study as the terrestrial limit for this calculation. The resulting average rate of sea-level fall is 1.74 ± 0.37 m/ka (between 122.8 ± 0.6 ka and 113.43 ± 0.90 ka, and +4.8 m and −11.5 m). This is within the uncertainty of the value Moseley et al. (2013) derived for the older part of this time period.
Stalagmite QUE01 grew between 109.16 ± 0.29 ka and 90.88 ± 0.57 ka at a depth of 10.8 m. A hiatus occurred between 103.94 ± 0.58 ka to 96.82 ± 0.42 ka. Based on the discussion above, we suggest that this interruption in growth was caused by submergence during the MIS 5c sea-level highstand (Spratt and Lisiecki, 2016;Moseley et al., 2013;Dumas et al., 2006;Dodge et al., 1983). By accepting the arguments for submergence and considering data from Moseley et al. (2013), a well-confined sea-level history unfolds. Shortly after 107.7 ± 0.9 ka, speleothems became submerged at 11 m depth and at 10.8 m depth at 103.94 ± 0.58 ka. While never exceeding −4.5 m, sea level fell below -10.8 m by 96.82 ± 0.42 ka.  dated uplifted coral-reef terraces at Barbados and inferred that the MIS 5c highstand occurred during a 5 ka period from 103 ka to 98 ka. Accounting for uncertainty, their results are consistent with the QUE01 hiatus from 103.94 ± 0.58 ka to 96.82 ± 0.42 ka.
In a speleothem from Bermuda collected at 1.5 m above today's local sea level, Wainer et al. (2017) detected a growth interruption between 105.9 ± 6.4 ka and 92.0 ± 3.1 ka (Fig. 3). The authors argued that this 13.9 ± 7.1 ka-long interruption was caused by a sea-level highstand. This growth interruption covers the full range of the QUE01 hiatus (7.12 ± 0.72 ka) but is about 7 ka longer. The difference in elevation of the peak sea level at the two localities for this highstand is in accordance with the sea-level gradient modelled for the Caribbean region by Potter and Lambeck (2004). This model suggests a peak MIS 5c sea level at the Yucatán Peninsula at roughly −10 m, which agrees with a sea level of >1.5 m in Bermuda. It is necessary to apply a correction when comparing the local sea level to the eustatic sea level because of the described regional effects of GIA. As the sea level at the Yucatán Peninsula was rising throughout MIS 5c due to this effect, the eustatic highstand might be slightly earlier and/or in the early phase of the time range presented here.
The uncertainties of the timing of the MIS 5c highstand provided by stalagmite QUE01 presented here are about an order of magnitude lower than in previous Caribbean studies Wainer et al., 2017).
Stalagmite NAH14 (−17.7 m) enables a better constraint on the Mexican Caribbean sea-level curve presented by Khan et al. (2017) because active growth took place up until 9.52 ± 0.04 ka and thus falls within the curve's uncertainty range (Fig. 4).
While all Holocene growth phases from this study are considered to represent reliable upper limits for the sea level, we suggest that the two Holocene growth cessations detected in stalagmite CH10 (intense orange shading in Fig. 4 marks time periods with the highest certainty of a lack of growth), collected at −8.5 m, cannot be confidently attributed to marine submergence. As discussed above, these interruptions that dated from 9.74 ± 0.19 ka to 9.26 ± 0.08 ka and from 9.02 ± 0.06 ka to 8.57 ± 0.10 ka, may have been rather caused by climate change (intermittent cave flooding or cessation due to aridification). Moseley et al. (2015) provided evidence for a short-term local sea level of >−6.12 ± 0.1 m between 8.9 ± 0.4 ka and 8.4 ± 0.5 ka, as well as between 8.6 ± 0.1 ka and 8.7 ± 0.5 ka, based on marine layers in stalagmites. For this time range, Hering et al. (2018) presented radiocarbon ages of charcoal collected from fireplaces in the Chan Hol cave at depths of 7.7 m to 10.8 m. These samples document human activity at these depths during the deposition of CH10, and even after deposition ended at 8.57 ± 0.10 ka. The Chan Hol data from Hering et al. (2018) thus provide strong evidence for sea level below −10.8 m between 9.00 and 8.73 ka, and below −10.0 m between 8.16 and 8.00 ka (see Fig. 4). They agree with a sealevel reconstruction of −12 m at 8.1 ka by Collins et al. (2015b) based on 14 C ages of aquatic sediments in the area. Speleothem growth (CH8 and CH10) is found at 8.5 m depth between 9.40 ± 0.11 ka and 8.91 ± 0.08 ka as well as around 8.57 ± 0.10 ka.
Questions remain regarding the position of the local sea level between 9 and 8 ka, i.e. data of <−10.0 m as suggested by Hering et al. (2018) versus >−6.12 ± 0.1 m as reported by Moseley et al. (2015) leaving a gap of 3.9 m. Moseley et al. (2015) found no evidence for local disturbances that would raise the water level at their study site, nor have disturbances been identified by Hering et al. (2018) for their site. The calibrated 14 C ages reported from the Chan Hol charcoal samples would allow for a century-long cave flooding between about 9.0 and 8.7 ka, which corresponds to the younger CH10 hiatus and falls within the uncertainty of the oldest oscillation event recorded by Moseley et al. (2015). While unprecedented and unexpected, a possible scenario is the rise and fall of the sea level by 3.9 m within these 300 years. Since this seems unlikely considering GIA as well as geological and climatic constraints, the studies by Moseley et al. (2015), Collins et al. (2015b) and Hering et al. (2018) do not provide a consistent picture of the early Holocene sea-level evolution. Results from this study add well-dated terrestrial sea-level limits bracketing the 9.0 to 8.7 ka period but cannot reconcile previous studies. In addition, the possible influence of varying precipitation patterns during the highly variable climate of the early Holocene on speleothem growth phases and cave water levels remains to be investigated.

Conclusions
Fourteen speleothems from caves near Tulum, southern Mexico, provide upper limits for the local palaeogroundwater level. As the sampling sites are located close to the Caribbean coastline, our data provide evidence for terrestrial sea-level limits on the northeastern Yucatán Peninsula since the middle Pleistocene.
The oldest sample (PJ10) dates back to the transition from MIS 12 to 11 (421 ± 15 ka), while the three other stalagmites date to MIS 6. These speleothems thus provide the first evidence for speleothem growth during these periods on the peninsula.
Samples dated to MIS 5 extend previous datasets from this period. The onset of speleothem growth at −11.5 m at 113.43 ± 0.90 ka in combination with existing local coral data from MIS 5e allow for the calculation of a minimum rate of 1.74 ± 0.37 m/ka for the sea-level drop during the MIS 5e/5d transition. A growth stop in stalagmite QUE01 is interpreted as a result of submergence in freshwater caused by a risen sea level recording the start and end of the MIS 5c sea-level highstand (103.94 ± 0.58 ka to 96.82 ± 0.42 ka) on the Yucatán Peninsula, which lasted 7.12 ± 0.72 ka. The interpretation of this hiatus as a marine sea-level limit indirectly dated by two bracketing terrestrial-limiting growth phases is based on available sea-level limiting data points and supported by petrographic analyses.
Results from previous studies suggest a sea-level oscillation between 9 and 8 ka by as much as 3.9 m within a few centuries. This is unexpected in both amplitude and rate and while this study adds terrestrial sea-level limits for this time range, more data of higher temporal and spatial resolution are required to clarify this issue. The CH10 stalagmite provides evidence for multiple growth interruptions which are attributed to climate change related to Bond cycles rather than sea-level fluctuations.
The present data attest to the high potential of speleothems from the Yucatán Peninsula as indicators for the timing and duration of sea-level highstands during glacial-interglacial periods reaching back to MIS 12.

Declaration of interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.