Quaternary environmental evolution in the South Carpathians reconstructed from glaciokarst geomorphology and sedimentary archives

Abstract The Carpathian island-type glaciokarst has a great potential of preserving signals of past environments, archived in cave deposits like speleothems and clastic infills. We present here the geomorphology and structural control of several relict alpine caves and the surrounding glaciated marble karst in the Făgăraș Mountains. Four truncated and partially unroofed caves remained on the ridge-top of Mușeteica Mountain, above the glacial cirque, while a ponor cave that developed on the cirque bottom could be related to the Last Glacial Period. Structural measurements and cave morphology showed that the conduits formed at the intersection of foliation planes and tectonic fractures on the NE-SW and NW-SE directions. Cave development reflects three speleogenetic stages: 1) texture- and fabric-controlled dissolution and distension; 2) structurally-controlled breakdown; and 3) truncation, unroofing, and cave infilling with sediments. Slow diffuse dissolution was typical for the ridge-top caves, whereas M1 Cave developed by pressure flow. Further, we report the first U Th speleothem ages, related to the evolution of alpine caves and island glaciokarst in the South Carpathians during the Middle and Late Pleistocene. Dating results show a minimum estimated age of ~560 ka for the ridge-top caves, and that speleothem deposition met optimal conditions only during warmer periods, largely corresponding to interglacials. Stable carbon isotope values in speleothems range between −9.96‰ and −4.11‰, indicating the presence of plant and soil organic activity at the time of deposition. In total, five speleothem growth phases were distinguished during the last ~560 ka. We excavated the sediment infill of a ridge-top doline down to a 2-m depth. Radiocarbon dating revealed that it was deposited during the Late Holocene, and preliminary pollen analysis identified a plant assemblage dominated by grasses. Using the relationships between karst development, glaciation, and cave sedimentary archives, we present a time slice chronology of alpine landscape evolution at >560 ka, ~400 ka, ~330 ka, the Last Glacial Period (70–12 ka), and the Late Holocene. Our geomorphological, isotopic, and geochronological results also support the existing hypothesis that the South Carpathians may have experienced at least two glacial phases during the Pleistocene. Glacial erosion rate during the Last Glacial Period, and most likely during the penultimate glaciation, averages around 0.6 mm yr−1.


INTRODUCTION
Glacial activity leaves geomorphological imprints on carbonate rock bodies isolated at high altitudes, resulting in what is known as the island type glaciokarst .
Glaciated karst is a complex landscape, sensitive to climate change and capable of preserving unique evidence of past environmental change in alpine regions (Žebre and Stepišnik, 2015).
The presence of speleothems, the most used paleoenvironmental archive found in karst, often indicate the existence of periods with optimal environmental conditions for their deposition, which at mid-and high-latitudes are restricted mostly to warm periods (McDermott, 2004).
The initiation of cave formation itself, as resulting from a complex feedback between geology, geomorphology, climate, hydrology, and organic activity, is difficult to date but can be constrained using speleothem or clastic sediment deposition events (Polyak et al., 1998;Sasowsky, 1998;Stock et al., 2006;Häuselmann et al., 2015). The most common dating technique uses uranium series in speleothems (Richards and Dorale, 2003;Dorale et al., 2004;White, 2007;Scholz and Hoffmann, 2008), but trapped charge techniques such as OSL have also been used (Constantin et al., 2014).
In a glaciated environment, caves can be either truncated or completely removed by glaciation (Mais, 1999;Klimchouk et al., 2006), or be buried and preserved by infilling with glacial till (Cooper and Mylroie, 2015). They can also preserve morphological traits and sediments that could offer information useful for the reconstruction of mountain uplift (Meyer et al., 2011).
Alpine caves and their associated sedimentary deposits brought more light onto the Quaternary J o u r n a l P r e -p r o o f 9 (H min ) and the highest (H max ) altitude within a glacial cirque (e.g., Federici and Spagnolo, 2004). We calculated the cirque height as the maximum difference in altitude on each profile, then we used the average of these values and the time lapse between the beginning and the end of the Last Glacial Period (LGP) to estimate the maximum glacial erosion rate (E g ) within the cirque, according to the formula below.

( ⁄ ∑ ) ⁄
where n is the number of calculated height range values (here, n=5), is the value of each calculated item (i.e., the cirque height), and T is the estimated LGP time lapse. The duration of the LGP was estimated at ~59 ka, according to Lisiecki and Raymo's (2005) MIS 5-4 age boundary of ~71 ka, and the deglaciation age for the Făgăraș Mountains of ~12 ka provided by Kuhlemann et al. (2013). The maximum glacial erosion rate is assumed to have occurred along the lowermost points on the cirque floor, and gradually decreased up the headwall towards the crest.

Structural measurements and orientation analysis
We performed a comparative orientation analysis of cave passages and structural features (joints, faults and fractures). Results were plotted on rose diagrams with a bin width of 10° and on equal-area projection spheres on Schmidt stereonet.
Three sites ranging in size from 50 to 100 m 2 were selected for microtectonic measurements on marble outcrops in the vicinity of caves. We sampled 50─60 joints/site and measured the joint azimuth and dip. The best fit great circles of joint families were compared to stereographic projections of cave passages in order to compare each joint family with cave conduit data.

Journal Pre-proof
Fault and fracture sets were digitized from various sources at different scales, and plotted against each other on rose diagrams (half circles) to explore the regional and local controls of structural assemblages on cave development. Over 500 fault and fracture segments were extracted in the central Făgăraș Mts from CNES/Airbus satellite images (courtesy of Digital Globe). Documented faults (Dessila-Codarcea et al., 1968;Schuster, 1977) were plotted against the manually-extracted fault set. A supplemental set of fractures and fault segments was extracted from the high-resolution DSM of Mușeteica glacial cirque, and compared to those mentioned above.
Orientation data of underground and surface karst features were either collected during field survey or extracted from the DSM. Cave passages, ridge-top dolines, and shaft entrances were analyzed together, to better constrain the cave system's structural arrangement. The lengths of cave passages rather than frequencies were considered meaningful for analysis, allowing for better highlighting the cave development on preferential orientations.

Cave morphology
We surveyed the M1, M3-R2, M4, and M5 caves using a modified Leica™ Disto X310 range finder. The processed data were used to extract segments of cave passages, and calculate their morphometry and orientations. We then calculated the pattern morphometric indices (Table   1), synthesized by Piccini (2011), suitable for characterizing the geometry and morphometry of short alpine caves typical to stripe karst (Lauritzen, 2001). The focus was placed on indices' reliability for highlighting the most important morphometric features of a cave, and that they should be selected according to cave geometric properties and spatial complexity (Piccini, 2011). We therefore avoided to use indices that would be better fit for maze or horizontal caves. The length of M3-R2 Cave was conventionally calculated as the sum of all its passage segments (n=10), although the lengths of ridge-top branches (n=3), accessible only via the D1 collapse doline (Fig. 6), were not initially added (Tîrlă et al., 2016). However, for analytical purposes these branches will be considered in this study and their lengths cumulated to that of the cave itself. Vertical range coincides with negative drop, since no cave in the Mușeteica system has positive altitude difference.

Journal Pre-proof
Cave morphology was analyzed in relation to the three main genetic groups of erosive features and deposits considered relevant for cave development: erosive, chemical, and breakdown forms (Ballesteros et al., 2015). These were illustrated on the geomorphological maps scale 1:300.  (2008), and at the University of Melbourne (Australia) according to the procedure of Hellstrom (2003).

Speleothem U-Th dating and stable isotope analysis
Samples for low resolution isotope profiles were hand-drilled at 5 mm distance for M3-R2-3 (n=12) and M5-1 (n=27). Powder samples were also taken by hand drilling from three host rock samples from near the ridge-top caves, labeled M5-3, M3-R2-1, and M3-R2-2 (should not be confused with speleothem labeling). They were analyzed at Northumbria University, using a Thermo Delta V Advantage coupled with a GasBench II sample preparation device.

Sediment infill of doline D2 -stratigraphy, 14 C dating and pollen content
The clastic sediment infill of the D2 doline was excavated to a depth of ~2 meters, without reaching the bottom. Stratigraphy was described based on field observations. Bulk samples were taken at 10-cm intervals, on thicknesses of around 2-4 cm, and kept refrigerated in plastic bags. Grain size content of seven global samples was analyzed at the Geotechnics Laboratory of the University of Bucharest, following the standard procedure STAS 1913/5-85. First the samples were weighed and introduced into the oven at 105C for 24 hours, to remove the organic matter. water and lithium carbonate were added in the dry samples and left for 24h, then the cement was separated using the 0.063 mm sieve. The mass of the cement resulting from washing exceeded 10% of the sample and was further analyzed by applying the sedimentation and Sieve-pipette methods for the fine fractions and sieving only for the coarse fractions (Gee and Bauder, 1986). Based on the data obtained, particle size and percentage were determined for each of the following Udden-Wentworth classes: a) larger than 2 mm; 2-0.063 mm; c) 0.063-0.004 mm; d) smaller than 0.004 mm (Wentworth, 1922).
Three samples for optically stimulated luminescence dating were taken by hammering a 20 cm long plastic tube into the sediment and were analyzed at the Babes-Bolyai University in between the acid steps. At the end of the procedure, the pH was set to neutral adding ultrapure water, followed by sample drying at 60C. The organic carbon was then collected for the graphitization step. The 14 C ages were calibrated using the Oxcal online program (Bronk Ramsey, 2009), based on the IntCal13 calibration curve of Reimer et al. (2013).
One random 2-cm³ sediment sample (M6-16, from ~140 cm deep) was selected and prepared in order to extract and assess the quality of preservation of pollen and spores, following the procedure described in Moore et al. (1991). Pollen and microspore identifications were made under 400x magnification using the descriptions and identification keys in Moore et al. The height range is 2100 to 2380 m, and the headwall height is 120 m. It is a valley-head cirque with an area of 0.49 km 2 , comparably larger than the Râiosu cirque to the east (0.12 km 2 ), and the small, nested cirques (0.02-0.08 km 2 ) hanging on the western and northern slopes of Mușeteica Mountain at ~2000 and ~2200 m a.s.l.
The values used for calculating the average glacial erosion rate are given in Table 2. The five topographic profiles across the Mușeteica glacial cirque along which we calculated the cirque height are illustrated in Figure 4.  Journal Pre-proof of the cirque headwall. These channels end with imbricated triangular marble scree cones, which partly cover the lateral moraines.

Gravitational landforms
A large, meter-deep mass movement breaks the continuity of the glacial rim to the NW. Field observations show that it is a massive consequent rockslide, seeming to have been driven Surface deposits like talus scree, glacial till and rockslide deposits increase the landscape complexity, but these also cover, destruct or hinder surface karst development. Triangular talus scree deposits with angular marble fragments fringe at the base of snow avalanche channels, developed only on the eastern-facing cirque headwall.
The glaciokarst is divided into two areas: a) the area above the trimline containing the ridgetop caves and collapse dolines; b) the relict glacial cirque with solution and subsidence dolines, ponors, karren-incised roches moutonnées, and dry valleys.
Journal Pre-proof Journal Pre-proof M1 Cave developed preferentially at the contact between marbles and an amphibolite stripe (more easily weathered), with subsequent thin actinolite-and biotite-rich sheets interlayered within the marble. The cave has the steepest profile of all the caves in the area, about 45-53° on the foliation planes and 70-90° on the fault cross-cuts (Fig. 7). It is nearly a subvertical cave with limited development, a feature which is typical for stripe karst (Lauritzen, 2001). In the first section, large chambers are filled with breakdown deposits, which often form the

M3-R2
Cave is very wide and deep, in contrast to its shortness (Table 2) M6 is an unroofed cave passage found after excavating into the D2 ridge-top doline. The doline is filled with more than 2 meters of sediment and seems to continue with a small passage, although exploration work is still needed to access it. The walls at the western end host well-preserved speleothems, while many speleothem fragments can also be found within the clastic infill.
Morphometric indices are typical for alpine small caves (e.g., Piccini, 2011) and have comparable values influenced mostly by the geological conditions (Table 3).  Journal Pre-proof

Fault and fracture system
Three main orientations of the major faults were distinguished: NW-SE, W-E, and NE-SW (Table 4). With small differences, this pattern matches orientations of the medium-scale fault and fracture set extracted from the CNES/Airbus imagery (Fig. 9A).
The NE-SW and N-S oriented fractures are the most frequent, in contrast to the regional   Journal Pre-proof

Types of joints and the joint families
The joints generally form two systematic sets forming an orthogonal network (Fig. 9B), close-fitting a tension joint system (Passchier and Trouw, 2005). Shear joints form conjugate sets that intersect the former under a sharp angle (30-60º). This pattern is better observed on the ridge, near the entrance of M3-R2 Cave. Frost-weathering joints developed on preferred detachment planes of the minerals and have sub-decimetric to sub-centimetric densities. They distinguishable by their high density compared to that of schistosity and shear joints, the latter being strictly controlled by tectonics. The joint sets measured near the caves fall into four families (Fig. 10), of which only one (J1) seems to be common to the entire study area.
Another three joint groups were identified, two on the ridge-top (

Speleothem U-Th ages
Results of U-Th dating are given in Table 5 and plotted in Figure 11. A low concentration of uranium can be seen across all sub-samples of the four speleothems, with an average value below 20 ng/g, a maximum of ~50 ng/g (sample M6-1) and a minimum of ~10 ng/g (sample M3-R2-3). Stalagmite M3-R2-1 seems to have formed during a short, 2000-year period, roughly between 125 ka and 123 ka. The base age of M6-1 flowstone is older than 560 ka, beyond the limit of U-Th dating method, and its top age is 526 (+97/-52) ka. Journal Pre-proof Figure 11. Sampling of speleothem calcite for U-Th dating and stable isotope analysis.
Sample names are bolded, and the ages with corresponding uncertainties are given in ka.

Stable isotope results
In stalagmite M3-R2-1, δ 13 C values range between a minimum of -8.51‰ and a maximum of Journal Pre-proof

Age and pollen content of D2 doline (M6 Cave) infill
The sedimentary infill of ridge-top doline D2 down to the depth of 2.10 m consists in three distinct layers (I-III) separated by brownish-black clayey horizons (Fig. 12) Journal Pre-proof (1-column fitting image)

Relations between karst, glacial and periglacial processes
The last glaciation left recognizable geomorphic evidence in the Făgăraș Mountains (Urdea et al., 2011;Kuhlemann et al., 2013;Mîndrescu, 2016). The crest, headwall and bottom of the Journal Pre-proof the last deglaciation, affected only by erosion and local gravitational unloading.
The higher occurrence of dolines and ponors on the cirque bottom delineates an important area of dissolution and water infiltration in a concentrated flow, and the probable existence of an organized subterranean network. The glacial infill of covered paleodolines would have been subsequently affected by subsidence due to underground drainage, a process often seen in glaciokarst landscapes (Veress, 2016;Veress et al., 2019). During the last glacial period, the fossil ridge-top caves (M3-R2, M4, M5, and M6), located in the area above the trimline, were not directly influenced by glaciers, being subjected only to periglacial processes.
Such caves most probably formed during interglacials, when waters were more aggressive and able to enlarge cave passages . Glaciers do not have a significant impact on dissolution speleogenesis (Audra, 2001;Häuselmann et al., 2002), due to the saturation in Ca 2+ ions of the subglacial waters via incorporation of finely ground calcite eroded by the glacier (Bini et al., 1998), and due to a smaller CO 2 concentration that should have originated only from the atmosphere, in the absence of soil.
Under the influence of cold surface temperatures, ice could have formed in the frontal parts of the caves, contributing to passage enlargement during freeze-thaw cycles. This might explain for example the morphological differences between the front and the back of the M1 Cave.

Structural control
The cirque distribution in the Mușeteica-Buda area is highly controlled by geological Miocene (Mațenco and Schmid, 1999). Subsidiary fracture sets (WNW-ESE and NNE-SSW) are less abundant, but perceptible at regional and local scales.
Still, not all fractures have a tectonic trigger at the origin. Locally, where slope disequilibrium was reached due to glacial lateral sapping, followed by glacier retreat, it prompted massive rockslides oriented conformably to marble foliation.

Speleogenetic stages
Journal Pre-proof J o u r n a l P r e -p r o o f 36 The observed speleogenesis in this area consists of three development stages: 1) texture-and fabric-controlled dissolution on foliation planes and tension fractures; 2) structurallycontrolled breakdown; 3) cave unroofing and sediment infilling.

Texture-and fabric-controlled dissolution (M1 Cave) and distension (ridge-top caves)
The importance of contact planes between banded minerals with different physical and chemical properties in metamorphic rocks as speleoinception horizons plays a similar role with that of bedding planes in sedimentary rocks (Lowe and Gunn, 1997). Mineral Exhumation of the metamorphic rocks, followed by erosion, gradually released the pressure from the overburden, and probably caused the formation of J3 plane-parallel distension joints.
Thereafter, the central role on conduit development seems to have been played by the tension fractures that cross marbles perpendicularly on the S 2 foliation. These fractures correspond to the J1 joint family, which is the dominant joint system in the regional tectonic system (Fig.   9). Evidence of pressure flow can only be observed in the M1 Cave, but none in the ridge-top caves. These caves could have developed by diffuse water infiltration along the joint network, although organized water flow at low discharge could have played a role. Where fissure aquifers were intersected by faults or fractures, these evolved into karst conduits as pressure tubes (M1 Cave).

Structurally-controlled breakdown
In a later stage, enlargement of karst conduits caused breakdown due to decreasing stability of the ceiling, as described by White and White (2000). Breakdown, comparable to rock dissolution, occurred preferentially along the marble-schist contacts. Schist bands on the cave ceilings or ceiling edges were found in M3-R2 and M1 caves, which are larger and their morphology is better highlighted. The process is active and contributes to enlargement of chambers and passages and even to passage unroofing (surface breakdown).
Development of M1 Cave progressed along two main directions: NW-SE and NE-SW, intersecting in a 90º angle, conformably to local fractures geometry. The NW-SE passage depicts mostly a breakdown morphology, which is abruptly replaced by a pressure tube morphology in case of the NE-SW passage (Fig. 4A). This change in cave morphology could be attributed to tectonics: distension caused breakdown, whereas compression (or transpression) favored pressure tubes forming.
The breakdown-dissolution couple has gradually determined subsurface karst development along the marble foliation planes, whereas the surface components are mostly represented by dolines with ponors. M1 Cave is an exception, because it opens out on a roche moutonnée, partially collapsed due to subsurface breakdown.

Truncation, unroofing, and sediment infilling
The entrance of the M3-R2 Cave hangs above the Râiosu cirque on the eastern slope of Mușeteica Mt, indicating it was truncated by slope retreat, which generated the present-day shape of the Râiosu upper basin. As well, morphology of the D2 doline is evidence that former cave passages (M6 Cave) collapsed and were filled with sediment washed in from the surrounding slopes. The sediment infill has preserved the collapse morphology and in situ speleothems.

Journal Pre-proof
All the studied caves except M1 are arranged into a very small, 5,500-m 2 area on the mountain ridge top (Fig. 5). M4 and M5 caves, located very close to each other (~15 m), were undoubtedly truncated as inferred from the existence of in situ flowstones outside their entrance. Given their close position, we propose that these caves are remnant segments of a single network, whose interconnecting passages are probably obstructed by breakdown and

Speleothem growth phases
Overall, the chronology of speleothem growth is generally well constrained given the limitations of the U-Th dating method. The largest age errors given by flowstones M6-1 and M5-1 are due on the one hand to the very low uranium content (between 14 and 51 ng/g), and on the other to the closeness of the U-Th dating limit, of ~500,000 years.
Speleothem formation seems to have occurred during at least five distinct phases (or episodes). The first phase was recorded by flowstone M6-1, which started growing before 560 ka, although this age estimate does not benefit from error assessment. The top of this flowstone has an age of 526 ka with an error defined as +97/-52 ka, which is not unusual for samples of similar age and with such low uranium content. A second phase is identified in Journal Pre-proof flowstone M5-1, which appears to have grown between 440 ±37 ka and 399 ±30 ka, roughly overlapping marine isotope stages 10 to 13 if uncertainties are considered in full. The third phase took place during MIS 9 and is identified at the base of sample M3-R2-3, at 330 ±11 ka. Also, in flowstone M3-R2-3, we can identify the fourth phase around 200 ka. Here, a change in petrography from dark brown to white, opaque calcite is defined by two age estimates of 204 ±6 ka and 200 ±8 ka, during MIS 7a-7c. Taking into account the error estimates, there could have been no hiatus at all between them, or a hiatus of at most 14 ka.
One could speculate that it might have been caused by detrimental environmental conditions for speleothem formation during MIS 7b. A fifth phase is clearly defined by stalagmite M3-R2-1, which grew only during the warmest part of the last interglacial (MIS 5e), between 124.9 ±1.7 ka and 123.6±1.7 ka (Drăgușin, 2013). Moreover, the top of flowstone M3-R2-3 was dated at 97.6 ±26.0 ka, restricting its deposition to sometimes during MIS 5.
M3-R2-3 seems to even span several glacial cycles and we suspect that there should be hiatuses within the sample that cover the intervening glacial periods.
At sites similar to ours, for example at high elevation in the Alps, Spötl et al. (2002) and J o u r n a l P r e -p r o o f 40 11, and 9 (Berstad et al., 2002). Also in N Norway, flowstone LP6 from Laphullet Cave grew mostly during the MIS 11 interglacial Lauritzen and Lundberg, 2004).
This speleothem is very similar with our sample M3-R2-3 with respect to its geomorphological and climatic setting. Laphullet Cave developed on a ridge-top in marble bedrock in an area with mean annual temperature of 2.8°C, near the Arctic Circle, and was repeatedly affected by glacier advance and retreat. can be the result of a mixing between organically derived CO 2 which usually has a δ 13 C value of about -23‰ (in a C3 dominated plant association), and carbonate host rock (around 1‰ in our case), which usually gives a δ 13 C speleothem value of around -11‰ (see for example the review work of McDermott, 2004). The highest values that we measured can be given by a wide range of processes intervening between host rock dissolution and speleothem deposition, but is not the aim of this work to discuss them. The most important information we want to convey is that values of -10‰ cannot be produced in the absence of plant and soil microbial activity specific to warm periods.

Significance of clastic sediment infill
The young age of the sediment infill, inferred both from the direct radiocarbon dating of bulk organic matter but also from the failure of obtaining OSL ages, implies that the collapse of the M6 Cave ceiling took place very recently, during the Holocene or, in any case not during the last glaciation. If the collapse doline generated by ceiling collapse would have been amounts that could be analyzed for a wide range of biogeochemical proxies offers an opportunity that is not available for the study of lake sediments, that relies on coring, and could be a welcomed addition to such type of Holocene archives from high elevation.
Further improvements in pollen and spore retrieval could be achieved by changes in the preparation protocol (e.g., removing the acetolysis step and reducing the HF treatment).

Regional landscape evolution and reconstruction of cave assemblage
The U-Th ages of our speleothems indicate that underground voids developed in the Mușeteica area before ~560 ka. As the sample yielding this age was retrieved from a wall, it also shows that this section of M6 Cave did not suffer from breakdown processes since its formation.
Evolution of the Mușeteica caves in a cold, glaciated environment was most likely dependent on surface erosion (via cirques development and increased lateral migration of the crests under the ice dynamics), coupled with water availability and soil development. The most important effect on surface topography was glacial and periglacial modelling, counterbalanced by mountain tectonic uplift, and to a lesser extent by fluvial erosion and marble dissolution.

Glacial erosion
The morphology of the Mușeteica glacial cirque itself is likely the result of several glacial phases during the Pleistocene. Studies have shown that the necessary time for cirques to form may be between 125 ka and beyond 400 ka (Barr and Spagnolo, 2015, and references therein). Therefore, it is likely that the glaciers have eroded the bedrock at least during the penultimate glaciation, MIS 6 (191-130 ka), and the Last Glacial Period, MIS 4-2 (70-12 ka).

Journal Pre-proof
The exact time of ice formation during the LGP is unknown. Based on the widely accepted fact that speleothems grow mostly during the interglacials/warm periods, and that our youngest speleothem age is 97.6±26.0 ka, we presume that local glaciation should have initiated sometimes between 115 and 70 ka. However, isotopic studies have shown that conditions for ice formation or advance were met mostly towards ~70 ka, given that the last warming peak of the MIS 5 interglacial (5a) was at 82 ka (Lisiecki and Raymo, 2005). The approximate age of deglaciation in the nearby cirques was around 12 ka (Kuhlemann et al., 2013). We can estimate that, over the cumulated intervals of the penultimate and last glaciations (~120 ka), glacial erosion could have removed a maximum of 72 m of bedrock, resulting in an average erosion rate of 0.6±0.02 mm yr -1 . If the Mușeteica glacier melted earlier than 12 ka, then we may allow for a slightly higher value of the glacial erosion rate, probably no more than 0.65 mm yr -1 . Temperate glaciers are able to erode the bedrock at rates between 0.1 and 10 mm yr -1 (Swift et al., 2015, and references therein). In the European Alps, Valla et al. (2011) have shown a minimum mean erosion rate of 1 mm yr -1 . Yet, the Alpine glaciers are better developed and more aggressive compared to former Carpathian glaciers, considering the cirque size of the latter (Mîndrescu and Evans, 2014).

Estimation of dissolution rates
During periods with no glacier presence, the dominant processes which might have lowered the topographic surface were karst dissolution and fluvial erosion. These processes must have contributed considerably less than the glaciers to surface erosion.
Regarding dissolution rates, Bögli (1971) reported values between 0.014 and 0.071 mm yr -1 for the bare karst in Muotatal (Switzerland), while Lauritzen (1990) (Häuselmann, 2008). According to these studies, marble dissolution rates do not exceed 0.1 mm yr -1 in such conditions, much lower than our calculated glacial erosion rate of ~0.6 mm yr -1 .  (Zugrăvescu et al., 1998). Despite the apparent increase of the uplift rates since the Upper Miocene until present, such a large time gap confines the arguments on paleoaltimetry only to rough estimations, and therefore we excluded it from the proposed scenario below.

Sequence chronology of cave development and landscape evolution
The scenario proposed below is based on speleothem ages, and the assumption that the caves must be older than the deposits within. We describe six time periods in the evolution of this landscape during the past >560 ka, illustrated in Figure 14: Here, the former glacier seems to have described a clockwise 70º-90º rotation from south towards the east, imposed by the geological structure of marbles.

CONCLUSIONS
This study provides new geomorphological and geochronological information on the The ridge-top caves and former passages were truncated due to slope retreat under glacial and periglacial erosion, throughout the Middle and Late Pleistocene. We discovered a cave remnant (M6) by excavating the siliciclastic sediment infill of a ridge-top doline.
We reported the first speleothem U-Th dates from the alpine karst in the Romanian Carpathians, that range between ~70 ka and more than 560 ka, spanning the Middle and Late Pleistocene. U-Th dating shows that speleothem deposition was limited mostly to interglacial periods. Stable carbon isotope values of these speleothems have a minimum around -10‰, indicating the presence of sustained activity from plant and soil microorganisms, consistent with environmental conditions of warm periods.
Based on speleothem ages and glaciokarst geomorphology, we proposed a landscape reconstruction scenario comprising six time slices, at >560 ka, 400 ka, 330 ka, 70 ka, 70 to 12 ka, and at present. Each chronological sequence illustrates changes that occurred in cave development, speleothem growth, and glacial/periglacial erosion.
The sediment accumulation we identified in the ridge-top dolines has a late Holocene age, preliminary radiocarbon dating showing it started depositing before ~3 ka. We showed that it preserved pollen from a mostly herbaceous plant association. Combined with other biogeochemical proxies and a possibility for large sample availability, these sedimentary archives can be a welcome complement to lake sediment cored from the region.

Abstract
The Carpathian island-type glaciokarst has a great potential of preserving signals of past environments, archived in cave deposits like speleothems and clastic infills. We present here the geomorphology and structural control of several relict alpine caves and the surrounding glaciated marble karst in the Făgăraș Mountains. Four truncated and partially unroofed caves remained on the ridge-top of Mușeteica Mountain, above the glacial cirque, while a ponor cave that developed on the cirque bottom could be related to the Last Glacial Period.
Structural measurements and cave morphology showed that the conduits formed at the intersection of foliation planes and tectonic fractures on the NE-SW and NW-SE directions.
Cave development reflects three speleogenetic stages: 1) texture-and fabric-controlled dissolution and distension; 2) structurally-controlled breakdown; and 3) truncation, unroofing, and cave infilling with sediments. Slow diffuse dissolution was typical for the ridge-top caves, whereas M1 Cave developed by pressure flow.
Further, we report the first U-Th speleothem ages, related to the evolution of alpine caves and island glaciokarst in the South Carpathians during the Middle and Late Pleistocene. Dating results show a minimum estimated age of ~560 ka for the ridge-top caves, and that speleothem deposition met optimal conditions only during warmer periods, largely corresponding to interglacials. Stable carbon isotope values in speleothems range between -9.96‰ and -4.11‰, indicating the presence of plant and soil organic activity at the time of deposition. In total, five speleothem growth phases were distinguished during the last ~560 ka.
We excavated the sediment infill of a ridge-top doline down to a 2-m depth. Radiocarbon dating revealed that it was deposited during the Late Holocene, and preliminary pollen