Sedimentary anatomy and hydrological record of relic fluvial deposits in a karst cave conduit

This case study from the renowned Demänová Cave System in the Carpathians of Slovakia demonstrates that the conventional methods of fluvial sedimentology, combined with an allostratigraphic mapping and speleothem U‐series isotopic dating, can give unprecedented insights into the hydrological history of underground karst conduit. The deposits studied are a relic compound sidebar ranging from gravel to mud and encapsulating the conduit’s hydrological history from the middle Pleistocene to the present time. A succession of 10 allostratigraphic units, time‐constrained by speleothems, are recognized in the sidebar deposits, and the corresponding morphodynamics of an evolving cave‐floor sedimentation are reconstructed in considerable detail. The subterranean river water stages recognized from the deposits, time‐constrained by flowstone layers and stalagmites, correlate with and add to the regional record of climate changes. Two distinct episodes of flow ponding (high‐stage slackwater conditions) are recognized and attributed to the independently documented downstream cave‐roof collapses, probably triggered by the Carpathian post‐orogenic earthquakes. This multidisciplinary study may serve as a useful methodological guide for the analysis of fluviokarstic deposits in speleological research and reconstruction of their hosting cave hydrological history.

Particularly little sedimentological research has focused on cave clastic deposits to reconstruct the hydrological history of karst cave conduits, where the subterranean rivers are constrained by bedrock morphology and are no longer free to operate in their 'normal' manner known from the land-surface alluvial plains. Instead, their behaviour in cave conduits resembles that of rivers in bedrock mountain ravines (Tinkler & Wohl, 1998;Herman et al., 2012). However, these conditions also mean that the river entire hydrological history is confined to and potentially recorded in a single narrow subterranean conduit.
Reconstruction of the hydrological history of underground conduits is crucial to palaeoclimate studies and to the landscape evolution in karstic terranes, as the subterranean drainage limits surface water and may allow few or no surficial rivers or lakes to form, or can make such land-surface features virtually disappear. Many land areas in the world are karstic terranes (Assaad & Jordan, 1994), which renders their hydrological systems highly vulnerable to the impending modern global climate change.
Fluvial deposits in cave conduits are sparsely preserved, but even their local relics can shed important light on the hydrological history of underground river conduitsas is demonstrated by the present case study from the Demänová Cave System of Slovakia. The study employs conventional sedimentological methods, combined with allostratigraphic outcrop mapping and speleothem U-series isotopic dating, to reconstruct in detail the hydrological history of the cave conduit and its changing internal morphodynamics of sediment accumulation and erosion from >350 ka to the present time. The deposits studied are a relic compound sidebar of the subterranean branch of the Demänovka River but are highly informative as a condensed local sedimentary hydrological record. Apart from its regional significance for the history of the Demänová Cave System, this case study may serve as a useful method guide for the sedimentological and hydrological analysis of fluviokarstic deposits in speleological research.

SPELEOLOGICAL SETTING
The Demänová Cave System (DCS), known as the Demänovský jaskynný systém in Slovak, is located in the north-trending Demänová Valley of the Slovakian Low Tatra Mountains of the Carpathian orogenic belt (Fig. 1). Glacial tills indicate that the upper part of the Demänová Valley was glaciated at least twice during the Pleistocene (Vitásek, 1923;Louček et al., 1960;Droppa, 1972). The cave system formed in the Anisian-Ladinian carbonate rocks thrust by the Carpathian (Alpine) tectonism as a nappe over the granitoid crystalline core of the Low Tatra Mountains (Droppa, 1957;Bella et al., 2014a;Gaál, 2016;Gaál & Michalík, 2017). The DCS is a renowned case of a multilevel karstic system formed by allogenic rivers and featured in many speleological textbooks (e.g. Sweeting, 1973;Warwick, 1976;Bögli, 1980;Jennings, 1985;Palmer, 2007). The cave system has a total length in excess of 41 km and includes 11 caves interconnected by looped conduits, distributed over a subsurface depth range of more than 190 m (Fig. 2;Herich, 2017). Droppa (1966Droppa ( , 1972 distinguished nine cave levels in the DCS and correlated them with the surficial fluvial terraces of the Demänovka River and adjacent Váh River in the Liptov Basin. However, it remains disputed whether these cave levels represent the surficial water drainage table or rather a combination of the water table and phreatic adjustments (Bella, 1993;Bella et al., 2014b). Droppa (1966Droppa ( , 1972) assigned tentatively the cave levels, long abandoned by fluviokarstic drainage, to the successive glacial stages of the classic Alpine morphostratigraphic scheme, but the U-series isotope dating subsequently revealed that they were actually older than presumed (Hercman et al., 1997;Bella et al., 2014b).
The present study is from the DCS level still active as a subterranean branch of the modern Demänovka River (Fig. 2) and dating to the middle Pleistocene until Holocene. Relics of cave fluvial deposits have been reported from the DCS (Droppa, 1957;Kojdová & Slíva, 2005), but the present study is the first attempt to decipher the hydrological history and morphodynamic sedimentation pattern of the underground cave system in its latest, post-early Pleistocene conduit (Fig. 2).

METHODS AND TERMINOLOGY
The study uses conventional sedimentological field methods, such as a detailed logging, measuring of bedding attitude and palaeocurrent directions, photographic documentation and a line drawing of bedding architecture on an outcrop photomosaic. Special attention is given to mappable outcrop-scale bedding discontinuities. The descriptive sedimentological terminology, including clast fabric notation, is according to Harms et al. (1975Harms et al. ( , 1982 and Collinson et al. (2006). The concept of allostratigraphic analysis is after NACSN (1983), with an allostratigraphic unit defined as a mappable sedimentary body bounded by bedding discontinuities. Such units represent the main increments of local sediment accumulation.
The U-series dating of speleothems, including chemical separation of uranium and thorium by the chromatographic method with TRU-Resin (Hellstrom, 2003), was conducted at the U-series Laboratory of the Institute of Geological Sciences of the Polish Academy of Sciences in Warsaw. Two different measurement methods were used: mass spectrometry and alpha spectrometry. The samples for mass spectrometry were between 0.1 g and 0.5 g. Mixtures of 233 U/ 236 U/ 229 Th, calibrated by uraninite analysis in secular equilibrium, were used as a chemical procedure and isotopic fractionation control. The isotopic composition of U and Th was measured at the Institute of Geology of the Czech Academy of Sciences in Prague, using a double-focusing sector-field ICP mass analyzer (Element 2, Thermo Finnigan MAT; Thermo Fisher Scientific, Waltham, MA, USA).
The samples for alpha spectrometry were 0.5 to 2.0 g (mainly <1 g). Uranium and thorium were separated by a standard chemical procedure for carbonate samples (Ivanovich & Harmon, 1992), using a chromatographic method with the DOWEX 1 × 8 ion exchanger (Sigma Aldrich, St. Louis, MO, USA). The chemical separation efficiency was controlled by the addition of a 228 Th/ 232 U spike (UDP10030 tracer solution by Isotrac, AEA Technology QSA, Carlsbad, CA, USA) before chemical treatment. The activity measurements were obtained on an Alpha Ensemble spectrometer (EG&G ORTEC/ AMETEK, Oak Ridge, TN, USA) with 1200 mm 2 active area and ultra-low background detectors.
The U-series ages were calculated iteratively from the 230 Th/ 234 U and 234 U/ 238 U activity ratios, and are given herein with an error limit of two standard deviations, which means a confidence level of at least 95.4% (Table 1). The decay constants of Jaffey et al. (1971) for 238 U, Cheng et al. (2013) for 234 U and 230 Th, and Holden (1990) for 232 Th were used. Age errors do not include uncertainties related to the decay constants. Corrected ages were adjusted for detrital contamination indicated by the presence of 232 Th using the typical silicate activity ratio 230 Th/ 232 Th of 0.83 (AE0.42) derived from the 232 Th/ 238 U activity ratio of 1.21 (AE0.6), 230 Th/ 238 U activity ratio of 1.0 (AE0.1) and 234 U/ 238 U activity ratio of 1.0 (AE0.1) (cf. Cruz et al., 2005). The initial value of the 234 U/ 238 U activity ratio (Table 1) was calculated based on the corrected activity ratio and sample age.

STUDY RESULTS
The deposits studied are preserved at the righthand (eastern) margin of the modern subterranean branch of the Demänovka River in the Prízemie segment of the Demänovská Jaskyňa Slobody Cave (Figs 3 and 4A). The deposits have a thickness of around 4 m and a lateral extent of a few tens of metres, including their hanging upstream relics plastered on the cave wall ( Fig. 3, upper right). Despite the limited extent of their outcrop, these relic depositsinterspersed with carbonate speleothemsencapsulate the hydrological and sedimentation history of the host cave conduit since the middle Pleistocene. The present study focuses on the allostratigraphy and sedimentological characteristics of these cave-side deposits, and on their hydrological and regional climatic implications.

Allostratigraphic units
The cave-side deposits studied comprise alternating gravel, sand and mud (Fig. 4B). The sand and mud are siliciclastic, whereas gravel consists of both granitoid and carbonate rock clasts. The clastic deposits include intervening flowstone layers of various thickness and lateral extent, which were dated by the U-series isotopic method (Fig. 3, Table 1). The stacking pattern of deposits was recognized by allostratigraphic mapping of outcrop-scale discontinuities (Fig. 3) and detailed sedimentological analysis (Fig. 4B). The discontinuities are erosional unconformities with a variable relief and related facies changes, bounding ten allostratigraphic units. Nine of the units stand perched in the outcrop wall above the presentday river level (Fig. 3), whereas the youngest, tenth unit are deposits of the modern incised river floor. The following two sections of the text give sedimentological description and morphodynamic interpretation of the consecutive allostratigraphic units.  Droppa, 1966Droppa, , 1972, showing location of the studied Prízemie-Veľký Dóm segment of its lowest underground conduit. Note the projected level of the present-day surface route of the Demänovka River (cf. Fig. 1B). Sedimentological description Unit 1 -This oldest exposed unit, dated to >349 ka by its earliest speleothem cap (log II, Fig. 3), occurs in the sidebar downstream part (see outcrop segment with logs I and II in Fig. 3). Its maximum thickness exposed above the presentday river level is~50 cm (log II), but a hand-drill probing indicates that its base reaches a few more decimetres below the river floor and that its true maximum thickness may be around 80 cm. In the downstream direction, this unit abuts against and covers a bedrock knoll protruding as a low-relief ridge (azimuth~275°) from the cave eastern wall. Similar is the trend of Unit 1, which extends as a mound obliquely (azimuth~290°) from the cave wall into the outcrop (cf. Fig. 3). The deposits of this unit are clast-supported, sand-filled pebble gravel with an admixture of cobbles (local D max = 11 cm to 25 cm; Fig. 4) and with a 'rolling' clast fabric a(t)b(i) (Fig. 5A) indicating local bedload transport towards the WNW. The top of Unit 1 is erosional with an irregular relief, draped by speleothems ranging in age from 349 ka to 305 ka (Fig. 3). Unit 2 -This second unit is sandy and occurs as a speleothem-draped erosional remnant, 15 cm thick, at the site of log IIwith the flowstone drape extending to log I (Fig. 3). The sediment is fine-grained sand, light brownish grey in colour, with a few thin intervening layers of flowstone cementation (Figs 4 and 5B). The basal flowstone layer, dated to 349 ka, has bedding attitude of 175/10°and the upper flowstone layer of 200/30°, indicating an upward-steepening upstream sand accretion with episodes of emergence. The earliest flowstone layer draping this unit's erosional escarpment (Fig. 5B) is dated to 236 ka (Fig. 3, log II). ; the tenth unit is the gravel lag of the incised modern underground river. Note the location (red points) and dates of speleothem samples; sample numbers as in Table 1.  Unit 3 -This next unit extends laterally almost throughout the outcrop section, from between logs I/II to between logs IV/V (Fig. 3), and reaches a thickness of up to 1.1 m above the present-day river level in log III. The depth of its true base in logs III and IV is unknown, where also the unit top is markedly erosional, overlain directly by Unit 5 (Fig. 3). Unit 3 in log II is less than 40 cm thick and overlies an erosional terrace that is a local remnant of the two older units (Fig. 3). The terrace is covered with a lag of pebble gravel (D max = 4 cm) draped by speleothems dated to 182.4 ka and overlain by a package of alternating thin sand and mud layers inclined towards 344°, with a steepening-upward dip angle of up to 20° (Fig. 4). In log IV, the exposed upper part of Unit 3 is 75 cm thick and consists of planar parallel-stratified coarsegrained sand fining upward into ripple crosslaminated fine/medium sand ( Fig. 5C) and overlain by mud (Fig. 4). In the intermediate log III, the parallel-stratified sand varies from coarsegrained to medium-grained with sporadic mud interlayers, and the cross-laminated uppermost part of the unit contains several such thin muddy interlayers (Fig. 4). The basal speleothem, at the top of Unit 2 (Fig. 3, log II), indicates a lower time limit of 182.4 ka, but the exact upper time limit for the deposition of Unit 3 is unknown.
Unit 4 -This unit is preserved between logs II and III as an erosional remnant only 0.30 m thick (Fig. 3), but is very meaningful, as it overlies Unit 3 and yet is the finest-grained, composed of thinly interlayered dark brownish-grey clayey mud and lighter-shade muddy silt (Figs 4 and 5D). The bedding is flat and nearly horizontal, commencing with a thin (5 cm) basal layer of chaotic, massive mud bearing scattered silty mudclasts. This unit clearly post-dates Unit 3 and pre-dates Unit 5 (Fig. 3), but lacks bounding speleothems in the outcrop.
Unit 5 -This next unit has a distinctly erosional base, with a relief exceeding 1 m, and extends laterally throughout the outcrop (Fig. 3). It consists of a light brownish-grey, clast-supported and sand-filled coarse pebble gravel ( Fig. 5E and G; local D max = 5 to 16 cm, Fig. 4) with crude plane-parallel stratification. Imbricate a(t)b(i) clast fabric (140/45°in log II and 113/30°in log III, Fig. 4) indicates local bedload transport towards the north-west. The erosional base of this unit is deepest-incised in log I, where it reaches down to~0.3 m above the present-day river level and directly overlies the flowstone-draped Unit 1 (Figs 3 and 4). The base is least incised in log II, about 1.3 m above the modern river level, and reaches down to less than 1 m above the river level in logs III to V (Figs 3 and 4). This unit clearly post-dates Unit 4, and the oldest speleothems at its top are 102 ka in age.
Unit 6 -This unit is preserved only in log III, as a flowstone-draped discontinuous sand wedge less than 0.3 m thick and pinching out towards log IV (Fig. 3), with bedding attitude 250/13° (Fig. 4). The sand is medium-grained to coarse-grained, with planar parallel stratification, and overlies the flowstone-draped top of Unit 5 dated to between 102 ka and 93 ka  (Fig. 4), showing flowstone-separated units 5 and 6 overlain with a low-relief erosional escarpment by units 8 and 9. Note the limestone blocks (Fb) fallen from the cave roof at the base of Unit 9. Hammer (35 cm) for scale. Allostratigraphic units as in Fig. 3. Flowstone layers are highlighted in light blue. Speleothem samples with dates (red dots) are numbered as in Table 1.  (Figs 3 and 6). The flowstone cover of Unit 6 in log III has an age range of 81.6 to 70 ka (Figs 3 and 6). These dates would then define the time bracket for the deposition of Unit 6.
Unit 7 -This unit extends laterally from log III, where it shows depositional onlap on the flowstone-draped Unit 6, to the upstream end of the outcropwhere it overlies the flowstonecovered Unit 5 and reaches a thickness of 1.2 m in log IV (Figs 3 and 4). The component sedimentary facies show great variation. The lower part of Unit 7 consists of fine-grained to coarsegrained sand with ripple cross-lamination and plane-parallel stratification, including mud interlayers (up to 5 cm thick) and thin flowstone drapes, and with trough-shaped isolated shallow scours filled with a solitary cross-strata set (log IV) or non-stratified pebbly coarse sand (log V, Fig. 4). The upper part of this unit consists of massive (non-stratified) sand, which is finegrained and bears rotated blocks of partly cemented and flowstone-draped stratified sand in log IV ( Fig. 7A and B), and is mediumgrained and rich in mudclasts in log V (Fig. 4). The topmost part of Unit 7, preserved as an erosional relic only in log IV (Fig. 4), is a heterolithic package of thinly interlayered fine-grained sand and mud with a few intervening horizons of flowstone cementation. The onlapping base of Unit 7 between logs IV and III is slightly diachronous, from 70.9 to 70 ka, whereas a flowstone drape near its top in log IV has an age of 41.6 ka (Fig. 3). The time span of Unit 7 is then from 70 ka to at least 41.6 ka.
Unit 8 -This gravelly unit has a basal erosional relief of around 1 m and extends laterally from log II to log V (Fig. 3), with a maximum thickness of 0.7 m in log V and a virtual downstream pinch-out towards log I. The gravel has a sand-filled clast-supported texture (Figs 6 and 7C), varies from granule to pebble grade (D mean = 2 cm, D max = 8 cm) and shows clast imbrication fabric a(t)b(i) trending 152/50° (  Fig. 4), indicating gravel transport towards the north-west. According to the bounding speleothem ages, the minimum time bracket for the deposition of Unit 8 would be between 41.6 and 10.7 ka (Fig. 3).
Unit 9 -This unit is mud-rich and heterolithic, composed of thinly interlayered mud and cross-laminated fine-grained sand (Figs 6 and 7E to G). It extends laterally from log V to log II, reaching a maximum thickness of 1.2 m in log IV (Fig. 3) and marking the highest water stand of the river. The finest-grained deposits of Unit 9, in log V (Fig. 4), include massive (nonlaminated) mud layers with abundant silty mudclasts. The base of this unit shows fallen caveroof limestone blocks (log III, Figs 4 and 6). The heterolithic deposits of Unit 9 form a sigmoidal large-scale bed set of 'microdelta' type (sensu Jopling, 1965), with the bottom-set attitude of 325/6°, foreset attitude of 344/25°and topset attitude of 60/5°( Fig. 7D and H). The topset contains some ripple cross-sets directed obliquely upstream, towards 145°. The top surface of Unit 9 shows desiccation cracks in log V and is covered by a flowstone layer with stalagmites in logs II to IV (Fig. 4), dated to 3.56 ka (Fig. 3). The deposits also contain a few thin flowstone interlayers and carbonate-cemented horizons with dates of 10.7 to 8.9 ka. According to the speleothem dates (Fig. 3), the time bracket for the deposition of Unit 9 would be between 10.7 ka and 3.56 ka, although the top surface of this unit continues to accumulate flowstone today.
Unit 10 -This youngest unit is a gravelly pavement of the incised, modern subterranean Demänovka River (Figs 3 and 4A). It consists mainly of coarse pebbles and small cobbles, subrounded to well-rounded, and has a clast-supported, sand-filled texture. Hand-drill probing indicates gravel thickness of around 0.3 m, with an underlying sand probably representing Unit 3 (cf. Fig. 3). This unit clearly post-dates the latest major incision of the subterranean Demänovka River.

Morphodynamic interpretation
The morphodynamic development of sedimentation in the cave conduit is summarized schematically as an interpretive cartoon in Fig. 8. Unit 1 is older than 354 ka and interpreted as the downstream part of an early gravelly sidebar mound (Fig. 8A) formed in the flow separation zone of the cave conduit bend (cf. Fig. 4A). This bar abutted on a transverse bedrock knoll and buried it (Fig. 3). The overlying Unit 2 has an age range between 354 ka and 236 ka and is interpreted as a sandy scroll-bar accreted obliquely in an upstream direction to the cross-cave protrusion relief of the earlier gravelly sidebar (Fig. 8B). The river water level at this stage was lower and fluctuated, as indicated by the flowstone drapes (Fig. 4B) recording temporal emergence.
Unit 3 recorded an episode of cave-wide erosional reactivation of the river, leaving a gravel lag on the terrace made of units 1/2 in log II and forming a short-lived chute channel near the right-side (eastern) cave wall, while incising its thalweg towards the cave axis and left-side wall (Figs 3 and 8C). The flow then subsided and became confined to the thalweg zone, whereby the abandoned chute and erosional intrachannel terrace emerged and were covered by speleothems including prominent stalagmites (log II, Fig. 3). The time bracket of this development was 236 to 182.4 ka. The water level subsequently rose and a cave-wide river aggradation occurred, with a heterolithic channel-bank levée   Fig. 4A). (B) Stage 2 (349 to 236 ka): a mid-channel sand-gravel bar forms at the river bend, spilling out sand towards the channel right-hand bank. (C) Stage 3 (236 to 182.4 ka): an erosional sand-bed channel shifts to its right-hand cave wall while maintaining also its flow route between the two earlier-formed bars, with flowstone patches and stalagmites formed initially outside these channel branches. (D) Stage 4 (182.4 to >102 ka): cave-wide aggradation of fluvial sand culminates in a flow-ponding episode of mud deposition, attributed to downstream cave-roof collapse (Fig. 9). (E) Stage 5 (>102 to 102 ka): the river break-through occurs and its flow is reactivated, scouring the substrate as two channel branches while spreading gravel over the whole cave width. (F) Stage 6 (102 to 81.6 ka): the river level falls and flow decreases, with speleothem flowstone patches forming on channel banks (G) Stage 7 (81.6 to <41.6 ka): the river channel banks are emerged and collapsing, with the near-wall channel branch increasingly ephemeral, plugged by a slowly advancing mid-channel sand bar. (H) Stage 8 (<41.6 to >10.7 ka): the gravel-bed river is reactivated, with substrate scouring and cave-wide gravel deposition. (I) Stage 9 (10.7 to <8.9 ka): a new episode of flow ponding (high-stage slack water conditions), with progradation of a channel-filling silt-mud microdelta (sensu Jopling, 1965), attributed to the flow blocking by another downstream cave-roof collapse (cf. Fig. 9). (J) Present-day stage of incised river with its channel-floor gravel lag and left-bank sidebar as the youngest depositional unit. Diagrams not to scale; the cave width there is about 25 m and the thickness of preserved terraced deposits is~4 m. Time ranges based on speleothem U-series isotopic dating. built obliquely towards the cave right-hand wall and burying the speleothems (Fig. 4B). This marked sidebar depositional aggradation, between 182.4 and~160 ka, heralded the subsequent stage of pronounced water-level rise and flow ponding recorded by the muddy Unit 4 (Figs 3, 4 and 8D).
Unit 4 recorded pulsating high-stage slackwater conditions, attributed to the abrupt retardation and virtual blocking of the river flow by a barrier of cave roof-collapse breccia in the downstream Veľký Dóm segment of the cave conduit (Fig. 9A). The chaotic basal muddy layer in Unit 4 (Fig. 4) is a mudflow deposit derived from the rapidly submerged earlier channel-bank levée. The ponding stage is poorly constrained by speleothems, but its time bracket is estimated at between~160 and~140 ka. The Veľký Dóm chamber ( Fig. 9A and B) was apparently prone to multiple roof failures. The collapse breccia blocks there bear displaced speleothems post-dating 237 ka, with the younger recognizable collapses post-dating 33 ka and draped by flowstones younger than 13 ka ( Fig. 9C to E).
Unit 5 recorded an abrupt erosional reactivation of a gravel-bed river, attributed to the water eventually breaking through the downstream breccia barrier in the Veľký Dóm transit chamber (Fig. 9A). The local transport directions of the unit's bedload gravel indicate main flow to the north-west. This implies a river flow guided again by the cave conduit bend and conforming to the pre-existing erosional terrace made of Units 1 to 4, which also made the river channel split into two erosive branches: a deeper one at the right-hand (eastern) side of the conduit and a shallower one diverted obliquely towards its left-hand side (Figs 4 and 8E). The time bracket for the deposition of gravelly Unit 5 is estimated as~140 ka until~120 ka.
Unit 6, preserved only in log III (Fig. 4), was apparently deposited in its main volume in the cave near-wall sector around log I, but was subsequently removed by later erosion. The stratified relic of this unit in log III, dipping towards the WSW and pinching out towards log IV (Fig. 3), is a sand wedge spilled over into the abandoned 'hanging' channel of Unit 5 (Fig. 8F) which episodically emerged, as indicated by flowstone drapes (Fig. 4). As an interpretive inference, the main, non-preserved part of Unit 6 in the outcrop sector of logs I and II was probably gravelly to coarse sandy, reaching a cumulative height of up to 1.5 m above the reference level of present-day river to cause eventually the overspill of sand. The net time bracket for stage 6 (Fig. 8F) would be~120 until 70 ka, with the time bracket for clastic sedimentation between 93 ka (youngest flowstone at the top of Unit 5) and 81.6 ka (oldest flowstone at the top of Unit 6; Figs 3 and 6).
Unit 7 is a heterolithic facies assemblage indicating ephemeral flow with episodes of slackwater conditions, emergence and flowstone deposition (Fig. 8G). The main flow of an aggrading river channel at that time (70 ka until 41.6 ka) was apparently near the cave western wall, where the main sedimentary volume of Unit 7 was deposited but is unpreserved. The outcrop sector between logs III and V shows a perched terrace with the pre-existing 'hanging' channelplugged with sediment at the upstream bend, periodically emerged (speleothems) and recording only episodic sand overspill flow of the river thalweg. The deposition of massive sand with mudclasts (log V, Fig. 4) and with flowstone-draped rotated sand blocks (log IV, Fig. 4) indicates liquefied slumps derived from the perched channel's emerged margins. The main sediment body of Unit 7, as well as that of Unit 6, in the outcrop sector of logs I and II was removed by the erosional emplacement of Unit 8 (Figs 3 and 4).
Unit 8 represents a gravelly palaeochannel that first strayed away from the eastern wall towards the cave axis (logs IV-V sector of the outcrop, Fig. 3), and then cut down back towards this wall (logs I-III sector), removing the pre-existing deposits of units 5 to 7 and leaving only a thin gravel lag in logs II and III. The channel apparently reached an incision bypass stage and accumulated only sparse gravel pavement, with clast imbrication indicating bedload transport towards the NNW (Fig. 4). The capping speleothem above Unit 8 is dated to 10.7 ka (Fig. 3), but the high-relief base of this unit is hiding a considerable erosional gap, whereby the latest preserved speleothem (41.6 ka) of the relic Unit 7 below cannot be regarded as a true lower-boundary date for Unit 8 itself. The high water discharge of stage 8 (Fig. 8H) commenced probably not beforẽ 20 ka. Unit 9 represents a mud-rich 'microdelta' that prograded downstream from the cave conduit bend. Its deposition alternated between frictional tranquil flow (Wright, 1977) and slackwater conditions, as it gradually filled in the nearwall scour relief created at the previous stage (Fig. 8I). The formation of this unit indicates flow ponding (high-stage slackwater conditions), attributed to yet another episode of cave-roof collapse and flow damming in the downstream Veľký Dóm segment of the cave conduit (Fig. 9). The evidence of fallen roof-rock blocks at the base of Unit 9 in the Prízemie segment of the cave (Fig. 6, and log III in Fig. 4) supports this interpretation. Internal speleothem drapes (log IV, Fig. 4) indicate water-level fluctuations, with temporal topset emergence and a related emplacement of delta slope-derived mudflows (log V, Fig. 4). The reverse-flow ripple crosslamination in the microdelta topset (log IV, Fig. 4) indicates a reversing water circulation in the river-flow entrapment and separation zone of the cave bend (Fig. 8I). Unit 9 clearly postdates Unit 8 (Fig. 3) and pre-dates the subsequent deep river incision and emplacement of gravelly Unit 10. The youngest speleothem inter-layer~20 cm beneath the top of the resulting erosional terrace (log IV) is dated to 8.9 (AE0.2) ka, which means that the river level still oscillated around 2.5 m above its present level. Almost immediately followed the river deep incision of stage 10 (Fig. 8J), as indicated by the stalagmite growth~2.5 m above the present river level about 60 m downstream from the studied section and post-dating the ponding episode (date sample 29 in Fig. 9A).
Unit 10, much like the earlier Unit 8, recorded another dramatic incision of the subterranean river, this time towards the cave left-hand (western) wall and resulting in the present-day morphology of the Prízemie passage (Fig. 8J). Sidebar undercutting caused local displacement of large flowstone fragments onto the river floor (Fig. 3). This major erosional event is attributed to the river's new episode of breaking through its downstream roof-collapse breccia barrier (Fig. 9). The gravel lag of Unit 10 indicates a temporal end of river local incision and a switch to depositional mode, although the hydraulic profile of the subterranean river will inevitably evolve further by morphological adjustment.

DISCUSSION
The river profile Underground bedrock fluvial systems are generally the locus of the most rapid change in the drainage basin andin regional scalethey control the spatial propagation of base-level change (Tinkler & Wohl, 1998;Ford & Willliams, 2007). The longitudinal profile of the active underground conduit in the DCS, between the river southern inlet and its northern karst-spring outlet (Fig. 2), is notably more concave upward and deeper incised than the corresponding profile of the surface stem of the river (Droppa, 1966(Droppa, , 1972Hochmuth, 1993Hochmuth, , 1997. The DCS inflow zone shows an upstream-rising hydraulic gradient of the underground subhorizontal epiphreatic or looped phreatic conduits, as compared to the surface river, with several ponors and vadose drawdown passages (Bella, 1993). The intersection point of the underground and the subaerial river profiles is in the downstream Veľký Dóm chamber (Fig. 2), where the steadystate earlier incision of the underground river was perturbed by a roof-collapse rubble barrage (Fig. 9A). The barrage broke down the river profile, established a local temporary base level and disturbed underground incision by knickpoint upstream migration (Fig. 2). The underground river stem with a perennial flow has been able to regulate its profile more efficiently than the parallel surface river stem that is periodically dry. Downstream of the river resurgence to the surface, its profile has regulated to the valley base level by another knickpoint (Fig. 2). The karst spring is at an altitude of 789 m, near the mouth of the right-hand side valley, and has a cascade morphology of knickpoint incision by 3 to 4 m. The resurging river there joins the subaerial river stem and its profile. In addition to the two main knickpoints (Fig. 2), both the underground and the resurging surface river profiles show an array of secondary knickpoints in the form of recessing bedrock steps less than 0.5 m in relief (Bella, 2019).
The roof collapse in the Veľký Dóm chamber was a multi-episode phenomenon, although the limited access to the rubble barrage interior ( Fig. 9A and B) allows only the youngest episode to be dated based on speleothems. This youngest pile of rubble, up to 4 m thick, overlies an impact-crashed flowstone layer dated to 33 ka and is covered by a flowstone with stalagmites dated to 13 ka ( Fig. 9D and E). The fallen blocks bear rotated cave-roof flowstone drapes with a wide range of ages (Fig. 9C). The two distinct episodes of flow ponding in the upstream Prízemie passagecommencing around 180 ka and 10 ka, respectively (Fig. 8) were apparently a result of the main successive roof collapses and flow damming stages in the Veľký Dóm chamber (cf. Fig. 9A).
The flow-ponding sedimentary record (units 4 and 9, Fig. 4) indicates fluctuations of the contemporaneous river water level and energy between a sand-carrying weak tranquil flow and mud-depositing slackwater conditions. The barrage at each episode of roof collapse must have been initially an openwork feature, gradually patched with the river bedload gravel and sand, and eventually filled in with mud (Fig. 9E). However, the barrier permeability varied due to the local washout of its interstitial sand and mud by filtrating water flow. As the river fluctuating water level behind the barrier was critically rising in the Veľký Dóm chamber, a sufficient breakthrough of flow eventually occurred to terminate the ponding episode.
The water percolating through the cave roof today is oversaturated with calcium and forming speleothems, but the river water is undersaturated (Motyka et al., 2005) and hence dissolving both carbonate debris and the cave limestone bedrock. Similar hydrochemistry was likely the case during the warmer periods of the Pleistocene and may have contributed to the river breaking through the cave barrage. One can predict that once the barrage is eliminated by washout, corrasion, abrasion and dissolution, the local intersection point (Fig. 2) will vanish and the whole underground to subaerial river profile will eventually be regulated towards a joint equilibrium and the valley downstream general base level.

Speleothem record
The compound-bar stratigraphic succession (Fig. 10) shows seven recognizable generations of speleothems. They record relatively warm and humid land-surface conditions, with a vegetation cover and carbonate bedrock dissolution  Lisiecki & Raymo, 2005). The estimation of active river-floor height relative to its present-day level is based on the maximum height of particular allostratigraphic units in the sidebar outcrop. The estimation of flow power is qualitative, comparing flow over the sidebar (plot in red) with the inferred contemporaneous flow of river thalweg incised beyond the bar area (plot in pink). by percolating water and with the re-precipitation of calcium carbonate as speleothems in the cave (Fairchild & Baker, 2012). These generations are broadly correlative with the warming phases of north-west European climate (see the plot of marine isotope stages, MIS, in Fig. 10; Lisiecki & Raymo, 2005), although there are also some notable discrepanciesprobably due to specific local or regional conditions. For example, the oldest, first speleothem generation commenced its precipitation well before the global warming shown by deep-marine record, while the second generation apparently corresponds to the cool period MIS 8 (Fig. 10).
However, the Scandinavian glaciation at that time reached no farther south than central Poland and has virtually no record in southern Germany (Eismann, 2002). Contemporaneous glaciations in the Tatra Mountains have been suggested (Marks et al., 2016, fig. 5), but the local caves show formation of speleothems at that time. If mountain glaciers did exist in the Carpathians, they must have been short-lived and of limited extent (Szczygieł et al., 2019;Błaszczyk et al., 2020). Speleothems formed contemporaneously in the Alps (Spötl & Mangini, 2007), where recognizable significant cooling occurred at the closure of MIS 8 (ca 240 to 230 ka), indicating short-lived glaciers.
A well-developed generation of three successive speleothems, with thick flowstone layers and stalagmites up to 70 cm high, formed between 236 ka and 182.4 ka, during the warm period of MIS 7 (Fig. 10). Coeval growth of numerous speleothems occurred in Central European caves in the Alps (Spötl & Mangini, 2007;Spötl et al., 2008) and the Tatra Mountains (Kicińska et al., 2017;Szczygieł et al., 2019).
The younger part of the studied succession contains four generations of flowstone layers, dated to 102 to 93 ka, 81 to 70 ka,~41.6 ka and the Holocene, which correspond roughly to the MIS warming phases 5c, 5a, 3 and 1 (Fig. 10). Coeval precipitation of speleothems occurred in other parts of the DCS (Hercman et al., 1997(Hercman et al., , 2006Hercman, 2000;Podgórska, 2019), as well as in the caves of the Tatra Mountains (Głazek, 1984;Hercman et al., 2008) and the Alps (e.g. Moseley et al., 2014). These four youngest generations of speleothems in the present case formed on the emerged bar surface, which means that the river apparent low water stages were due to its thalweg lateral incision away from the bar (Fig. 8).
Somewhat puzzling may seem the lack of speleothems correlative with MIS 5e (Fig. 10), whichaccording to the European pollen recordwas the warmest and most humid episode of MIS 5 (Helmens, 2014), corresponding to the Eemian interglacial. The reason was probably the cave bar submergence by an erosive high-level flow of undersaturated river water (Fig. 8). Speleothems of this age are known from other parts of the DCS (Hercman et al., 2006).

Cave hydrology and regional climatic implications
Based on the speleothems and vertical facies changes, ten stages of sedimentation are recognizable in the Prízemie passage (Fig. 10). Their sedimentary record is allostratigraphic units 1 to 10 (cf. Figs 3 and 8), which include episodes of emergence as well as stratigraphic gaps due to erosion or sediment bypass. The units give a proxy record of the cave-conduit hydrological conditions (Fig. 10). The active river-floor level relative to its present-day position, as reached through the successive stages, is estimated from the maximum measured height of particular sedimentary unit and the height of speleothems (emergence levels) in the sidebar outcrop. These estimates approximate the river minimum water level. The qualitative estimation of relative flowpower changes is based on sedimentary facies.
The vertical organization of facies in the cave sidebar shows a characteristic motif of erosively emplaced gravel lag covered directly by a flowstone layer (Fig. 10), which indicates tunnelwide erosional rejuvenation of powerful flow followed by the river thalweg sideways incision, with the water level fall, sidebar emergence and cave-roof water percolation. The formation of contemporaneous speleothems implies wet landsurface conditions. The flowstone cover, recording sidebar emergence, is overlain by sandy facies alternating with flowstone drapes, which indicates low-power flow and river-floor net aggradation with a fluctuating water level. Aggradation in karstic rivers is generally linked to cool periods (Harmand et al., 2017). These episodes correlate with the periods MIS 10, 8, 6 and 2 (Fig. 10). In the regional context of the Low Tatra Mountains, this cave-fill facies motif may indicate climatic warming to cooling phases. In addition came the meaningful random episodes of flow ponding due to cave-roof collapses, probably recording major regional earthquakes (Fig. 10), with the initial infilling of rubble barrage by fluvial sediments and a subsequent river breakthrough.
Notably, there is no evidence of permafrost formation in the cave tunnel. The deposits show no cryoturbation, such as described from freezing caves (e.g. Luetscher, 2013;Obu et al., 2018), and there are no cryogenic cave carbonates that might indicate temporal formation of ice in the tunnel. The extent of permafrost in the Western Carpathians during the last glaciation is uncertain, and some other parts of the multilevel DCS may have been locally and ephemerally frozen (Orvošová et al., 2014). However, the pollen studies of intermountain peat bogs in the Western Carpathians document the presence of forests during the last glaciation (Jankovská et al., 2002;Jankovská & Pokorný, 2008). The allogenic DCS was located at the mountain glacier front that constantly provided water to the karst system, during the ice-front advance, stillstand and retreat. The concentrated water flux probably transferred heat to the cave by advection (Domínguez-Villar, 2012), as exemplified by the Castleguard Cave in the Canadian Rocky Mountains (see Yonge et al., 2018, fig. 15.32). The latter cave is located below the Columbia Icefield, yet the temperature in the vast part of its interior is above 0°C.
The gravel Unit 1 (Fig. 10) is older than 349 ka, but speleothem cover ages (Fig. 3) indicate that its top was reworked with minor gravel accretion around 349 ka (MIS 9). The bulk of this unit is apparently a cave-side relic record of the early Pleistocene (pre-MIS 10) melting of mountain glaciers. Little is known about these early mountain glaciations in Europe, but it is generally accepted that the Scandinavian Ice Sheet had its greatest southward extent in Central Europe during MIS 16 (676 to 621 ka) and MIS 12 (478 to 424 ka) (see review by Böse et al., 2012). In Poland, the Scandinavian Ice Sheet is considered to have reached the Carpathian foothills during the glaciations Sanian 1 and Sanian 2, correlated with MIS 16 and MIS 12, respectively (Marks, 2011;Lindner & Marks, 2015;Marks et al., 2016). Marks (2011) and Marks et al. (2016) suggested coeval glaciations in the Tatra Mountains, although direct surface evidence of these mountain glaciations in Europe is lacking. Their only proxy relic record may then be the underground fluvial deposits.
Similar is apparently the significance of the gravelly Unit 3 and Unit 5, recording a retreat of mountain glaciers at the warming phases of MIS 7 and MIS 5e, respectively (Fig. 10). The erosional gravel of Unit 3 correlates with the melting of small Alpine glaciers around 230 ka (cf. Spötl & Mangini, 2007). The age frame of Unit 5 is imprecise, but this stage of underground flow rejuvenation correlates with the final peak of penultimate deglaciation in the Central European mountains (cf. Böse et al., 2012), dated from the Alpine speleothems to between~137 ka (Häuselmann et al., 2015) and 134 ka (Spötl et al., 2002;Moseley et al., 2015). This timeframe corresponds well with the erosional emplacement of Unit 5 (Fig. 10). The surface record of mid-Pleistocene mountain glaciations in the Alps and Carpathians is poor, erased by younger erosion (Louček et al., 1960;Droppa, 1972;Schlüchter, 2008). The Scandinavian Ice Sheet supposedly made significant advances in the area of Germany and The Netherlands (Böse et al., 2012), and may have even sneaked into the Czech territory through the Moravian Gate in southern Poland (Marks et al., 2019). On the other hand, coeval cave speleothems in the Tatra Mountains Kicińska et al., 2017) and Outer Carpathians (Gradziński et al., 2012) indicate lack of permafrost and a relatively mild regional climate, perhaps weakly seasonal.
The gravelly Unit 8 correlates with the onset of MIS 1 (Fig. 10). The erosional base of this unit reached destructively flowstone layers dated to 41.6 ka, and its sandy cover contains thin flowstone layers dated to 10.7 ka (Fig. 3). Landforms and surface deposits of the preceding glaciation are common in the Low Tatra Mountains (Louček et al., 1960), but remain there undated. The 10 Be exposure age of glacial maximum moraines in the High Tatra Mountains is between~20 ka (Engel et al., 2015;Makos et al., 2018) and 18 ka (Engel et al., 2017). The oldest Late Pleistocene postglacial speleothems in the DCS have ages between 15 ka and 13.5 ka . The deglaciation phase MIS 1 in the Low Tatra Mountains would then occur between~20 ka and 15 ka, which matches the timeframe of Late Pleistocene glacier demise in the Alps (Ivy-Ochs et al., 2004;Wirsig et al., 2016;Fabbri et al., 2018).
The erosional incision of gravelly Unit 10 ( Fig. 10) recorded the post-glacial onset of Holocene hydrography, combined with the river break through the downstream roof-collapse barrage. This stage of sedimentation in the cave conduit marked the development of the modern river profile, as discussed in a previous section.

Rivers underground
The main independent controls on cave fluvial systems (Fig. 11, top; cf. Ford & Willliams, 2007) are similar to those acting on surface bedrock rivers (Tinkler & Wohl, 1998;Jansen, 2006;Whipple et al., 2013), where barriers can form by gorge-side rockslides. Bedrock river channels are generally considered to be fundamentally different from alluvial river channels, with the latter being shaped principally by water flow and sediment transport processes and the former principally by lithological and structural bedrock controls (Richards, 1982;Baker & Pickup, 1987;Ashley et al., 1988). The range of geological, environmental and geomorphic factors controlling surface alluvial rivers is thus different and wider (Richards, 1982;Schumm, 1985;Tooth et al., 2004;Bridge., 2009;Nicholas et al., 2018). Alluvial rivers have their channels cut in floodplain sediments, potentially mobile, and have the capability to migrate laterally or shift position by abrupt avulsion. The morphology of bedrock river channels, in broad terms, is a direct compromise negotiated between the fluvial forces applied and bedrock resistance offered (Tinkler & Wohl, 1998). The underground rivers may locally behave in an alluvial style in wider passages and large chambers (Ghinassi et al., 2009), but not in the cave-linking narrower conduits that predominate the subsurface extent of karst fluvial systems.
Bedrock rivers, even at quite low stages, typically show greater flow velocities and shear stresses than alluvial rivers (Ashley et al., 1988;Tinkler & Wohl, 1998). In contrast to alluvial river channels, the width of bedrock conduit is nearly invariant and its vertical accommodation space for a rising water level is practically unlimited. The morphodynamic response to hydraulic change at a falling water stage is limited to the appearance of a narrow thalweg zone, its braidedstyle splitting or localized incision. The flow in bedrock systems is unsteady and non-uniform and is commonly referred to as transcritical (Tinkler & Wohl, 1998), because it involves sections that are hydraulically critical, supercritical or subcritical. The transcritical regime is important, as it generates flow perturbations and causes flow superelevation at bends, which in alluvial channels dissipates by overbank spill-out. Fig. 11. Principal controls on an underground fluvial system, with a tentative list of the river hydrological and morphodynamic features recognizable from its cave deposits.
The most specific and crucial dependent control on underground bedrock rivers is the speleogenesis (Ford & Willliams, 2007), with its sinkhole pattern of water and sediment supply and its direct impact on cave conduit morphology (Fig. 11). An allogenic karst system with a hydraulic gradient receives water and sediment mainly from an upstream non-karstic drainage terrain. The sinking river can switch abruptly to a deeper cave conduit through vadose drawdown passages (Ford & Ewers, 1978;Ford, 2000), whereby its underground longitudinal profile evolves independently of the parallel surface river-stem profile (Fig. 2). The underground and surface river stems are joining one another beyond the karst spring zone, where a knickpoint forms and a common new downstream river profile is hydraulically negotiated. It is also speleogensis, combined with random independent perturbations (roof-collapse barrages, bedrock lithological changes and possibly active cross-faults), that determines the underground river profile and its evolution (Fig. 11). Underground channel bends are not simple ingrown natural meanders (Brakenridge, 1985), but features imposed by speleogenesis and bedrock characteristics.
Underground bedrock rivers are an extreme end-member case of the continuum of alluvial and bedrock channels (Brakenridge, 1985;Ashley et al., 1988). Their deposits provide a unique sedimentological opportunity to study the behaviour of a fluvial system arrested in a bedrock confinement, where the river evolution and hydrological history are a response to a limited number of specific controlling factors and are recorded within a single bedrock conduit. As the present case study demonstrates, even the local relic cave deposits provide a whole range of crucial geological information (Fig. 11, lower part)with direct implications for the cave system hydrological history and regional climatic conditions. Particularly interesting will be conceptual considerations as to how the least action principle (LAP), manifested by maximum flow efficiency (Nanson & Huang, 2017), plays its governing geomorphic role in a bedrock-confined underground fluvial system. Furthermore, the existing stream power models for surface alluvial rivers rely heavily on the substitution of water discharge for drainage area (Anthony & Granger, 2007). The style of channel incision and knickpoint migration in sinkhole-fed underground fluvial conduits, controlled by speleogenesis, may differ significantly from that in alluvial rivers and hence calls for a closer analytical consideration. The definition of karst-spring catchment area and boundaries has long been a major problem in the karst hydrology and hydrogeology, similarly begging for a new careful consideration (Bonacci & Andrić, 2015).

CONCLUSIONS
The present study demonstrates that a detailed sedimentological analysis of the relic cave-fill fluvial deposits can be an important contribution to speleological research and to an understanding of the hydrological history of karsticcave river conduits. The study shows that the underground Quaternary fluvial deposits can provide an important proxy record of the regional history of glaciations and climatic changes. From a regional perspective, the study sheds a new light on the history of mountain glaciations and deglaciations in the Demänová Valley of the Low Tatra Mountains, which was a subject of long-time literature speculations and controversies, and which may have regional climatic implications for the whole Central European Carpathians.
This is the first speleological study documenting a direct time link of the cave-fill gravelly facies motif with the climate warming phases, surface catchment deglaciation and rejuvenation of fluvial drainage. Another significant facies motif is that of aggradational sandy deposits indicating lower flow regime and water level fluctuations during cold climate phases, as well as the random slackwater facies motif of flow ponding, related to major cave-roof collapses and indicating regional earthquakes.
The study encourages sedimentological research of underground fluvial systems, which have thus far been little explored and for which the dating of speleothems, facies analysis and knowledge of bedrock rivers prove to be invaluable for both local hydrological and regional climatic reconstructions. reviews with editorial comments from Nick Eyles and Peir Pufahl helped to improve the manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.