Microfacies analysis

Six thin sections were cut parallel to the growth direction from locations with visible discontinuities during deposition, and those that differed in the macroscopic characteristic of calcite were used to determine the microscopic description. The main goal of the microscopic observation was to study growth mechanisms and microscopic features such as banding, fabric type, microdiscontinuities and inclusions. The microscopic observations were obtained using the Nikon Eclipse LV100POL microscope from the Institute of Geological Sciences at the Polish Academy of Sciences (Warsaw, Poland). The analysis and characterization of the speleothem fabrics were based on the methodology proposed by Turgeon and Lundberg (2001)1 and Frisia (2015). Based on the microscopic observations, a microstratigraphic log was constructed for selected parts of the flowstone; its construction used a series of codes to identify fabrics similar to those proposed by Muñoz-García et al. (2016) and Frisia (2015). The applied code scheme begins under very stable conditions and is represented by columnar compact (C), which is followed by columnar compact with laminae (Cl), columnar open (Co), columnar elongated (Ce), and columnar microcrystalline (Cm). The log concludes with columnar with possible dissolutions and hiatuses (Cd), which indicates the presence of the most unstable stage and corrosion and destruction conditions in the flowstone.

U-series dating

Series of 0.5–1.7 g calcite samples were drilled along the growing layers. The average thickness of the drilled sample was 2 mm. After the thermal decomposition of organic matter, the samples were spiked with a ²³³U-²³⁶U-²²⁹Th spike before further chemical treatment. The calcite sample was dissolved in nitric acid. Uranium and thorium were separated from the carbonate matrix using a chromatography method with TRU Resin. Standard samples and blank samples were prepared simultaneously for every series in the studied samples. Mass abundances of ²³⁶U, ²³³U, ²³⁴U, ²³⁸U, ²²⁹Th, ²³⁰Th and ²³²Th were measured by a double-focusing sector field ICP mass spectrometer (Element 2, Thermo Finnigan MAT) in an ICP-MS laboratory at the Institute of Geology (Czech Academy of Sciences, Prague, Czech Republic). The measurement results were corrected to account for background and chemical blanks. The final results were reported as isotopic ratios. U-series ages were iteratively calculated from the ²³⁰Th/²³⁴U and ²³⁴U/²³⁸U activity ratios using the following decay constants (yr^–1): λ₂₃₈ = (1.55125 ± 0.0017) ·10^–10 (Jaffey et al., 1971), λ₂₃₄ = (2.826 ± 0.0056) ·10^–6 (Cheng et al., 2000), λ₃₂ = (4.95 ± 0.035) ·10^–11 (Holden, 1990), and λ₂₃₀ = (9.1577 ± 0.028) ·10^–6 (Cheng et al., 2000).

Oxygen and carbon stable isotope compositions

Samples for the δ¹⁸O and δ¹³C analyses were acquired at a 4.5 mm resolution. The analyses were carried out at the Laboratory for Isotope Dating and Environmental Research at the Institute of Geological Sciences, Polish Academy of Sciences (Warsaw, Poland). Oxygen and carbon stable isotope compositions were determined using a Thermo Kiel IV carbonate device connected to a Finnigan Delta Plus IRMS spectrometer in a dual inlet system. CO₂ from calcite was extracted using orthophosphoric acid (density: 1.94 g/dm³) at 70°C. International standard NBS 19 was analysed in every ten samples. The isotope ratios were reported as delta (δ) values and expressed relative to the Vienna Pee Dee Belemnite standard. Measurement precisions of 0.07%o and 0.03%o were identified for oxygen and carbon, respectively (at 1 standard deviation).

Isotopic record correlations

For age estimation, we used the OIS method. OIS is based on the correlation between the local δ¹⁸O record and the global reference δ¹⁸O record (Imbrie et al., 1984). As a reference δ¹⁸O record with an established chronology, we used the global LR04 δ¹⁸O record (Lisiecki and Raymo, 2005). The LR04 curve was built as a combined record from 57 various and globally distributed foraminifera records. The LRO4 record is considered to be a record that reflects global changes in the δ¹⁸O composition of marine water. It also meets all of the necessary conditions for a reference curve: it is complete and detailed, it includes the desired time interval, it is controlled by the same factors, and it is reliably prepared. Before the correlations were calculated, data from both records were normalized. Normalization eliminates the effect of amplitude changes and the effect of data value shifts. Both of these effects are connected with local paleoenvironmental conditions. The next problem is the various reactions of caves and oceans to global changes, which are revealed by the same factors. During glaciation, ocean water becomes enriched in heavy oxygen isotopes, while meteoric water is enriched in light oxygen isotopes. For speleothems, the δ¹⁸O composition of meteoric water is an important factor. Therefore, we can expect that δ¹⁸O signals in the Głęboka Cave will have a negative correlation with marine signals. To solve this problem, we inverted the Głęboka Cave δ¹⁸O record for intercorrelation purposes. Correlation was accomplished using the modified GenCorr software (Pawlak and Hercman, 2016). GenCorr seeks an optimal correlation (i.e., the greatest similarity) between two or more records using different correlation procedures depending on the a priori age information. The basic procedure assumes that there is no knowledge of the record ages and therefore attempts to calculate correlations in a completely free mode. In contrast, comprehensive age information calculated using an age-depth model (with a continuous chronology) allows for the identification of optimal correlations within the limits of the model’s confidence bands. However, in most typical situations, only partial information about the age of a record is available (e.g., few dating results). This type of information may be used as a chronological benchmark during the correlation process. Three types of benchmarks are implemented in GenCorr: (i) fixed benchmarks, which are known ages at specified points in the profile; (ii) age information at specified points (e.g., dating results), which is described by a probability distribution (e.g., normal, uniform or skewed distribution); and (iii) age information that is described by an age limit (e.g., older than 0.5 Ma). The age assigned for a fixed benchmark remains constant throughout the entire correlation procedure, while the ages for non-fixed benchmarks may change according to their distributions. Calculations for points located between benchmarks are required to meet only the following basic assumptions: (i) the ages are in accordance with the ages for the benchmarks and (ii) the results agree with their stratigraphic positions.

Using a genetic algorithm as a tool for correlation is one typical example of an AI technique that solves optimization problems. The final results depend on the initial populations and parameters of the genetic algorithm. The goal of the correlation process is to identify the most similar position in the studied records (i.e., identify the global optimum). To minimize the local optimum problem, it is recommended to perform several independent correlations using a random initial population, as the results choose the group of individual results with the best heuristic value (Pawlak and Hercman, 2016).

Calcite fabrics

A microscopic analysis of the calcite crystal appearances and the identification of the original texture features of the studied material shows that approximately 90% of the observed flowstone is composed of elongated calcite crystals, which range from 0.2 mm to 3 mm wide and 2 mm to 30 mm long. Overall, the elongated crystals are perpendicular to the growth surface in each layer. Typically, the crystals are flush with each other; they are subparallel with one another and extinguish light at the same angle for the entire length of the crystals during microscopic observation (Fig. 4A). The surfaces of the crystals are usually jagged or serrated (i.e., characteristic protruding rhombohedral terminations; Fig. 4B). The colour of the crystals varies between clear, pale yellow and brown. The detrital material builds only thin, dark laminae. Other parts of the crystal are almost free of detrital material. Inclusions are visible within the crystals, which have extensions that are subparallel to the direction of crystal growth surface (Fig. 4C). The characteristics mentioned above are similar to those described in the works of Frisia (2015), Turgeon and Lundberg (2001), and Dziadzio et al. (1993), which indicates that columnar fabrics are also highlighted in those works. The described material often contains zones that are characterized by the occurrence of alternating white and brown laminae of different thicknesses. These laminae form sets, which are located within larger, flatter terminated calcite crystals (Fig. 4D). The brown colour of the laminae is caused by the higher content of detrital material, while the white laminae are composed mainly of pure calcite. These features are similar to those described for laminated fabrics (Dziadzio et al., 1993) or columnar compact fabrics with visible lamination (Frisia, 2015).

Fig. 4

Photomicrographs for different types of calcite fabrics observed in thin sections. A – Columnar fabric (compact and open); B – Columnar with hiatuses, columnar fabrics and protruding rhombohedral terminations (C_rt) underlined with a higher content of detrital material; C – Columnar fabric with a faint lamination, which has a visible characteristic elongation of the inclusions parallel to the direction of the calcite crystal growth; D – Columnar compact fabric with faint and flat laminae; E – A large discontinuous surface with a blocky crystal fabric that gradually transforms into a columnar fabric; F – A degradation surface that is also visualized by the change in the direction of crystal growth; G – Initial crystalline stage, which precipitates after discontinuity surface with transitional basal crystals; H – Columnar fabric of the microcrystalline type; I – Dark layers (i.e., condensed dark laminae) with possible dissolutions and/or hiatuses.

During petrographic observations, the occurrence of several types of columnar fabrics was observed: elongated – competitive growth with the incomplete coalescence of crystals and a length to width ratio greater than 6:1, which sometimes shows lateral overgrowth; compact – when the crystals form a compact aggregate, and the intercrystalline porosity is not discernible; open – characterized by the presence of linear inclusions or pores; microcrystalline – composed of highly irregular intercrystalline boundaries, has a uniform extinction, and is punctuated by inter- and intracrystalline microporosity (Frisia, 2015).

Other characteristic zones are related to discontinuities in the surfaces. Those zones contain darker layers that are enriched with detrital material. They are often accompanied by visible signs of corrosion in older layers. These layers are essential for the formation of small calcite crystals, which are shaped by small blocks surrounded by quartz, micas and clay materials. Initially, many small crystals grow in different directions gradually with distance from the nucleation surface, then the size of the crystal increases significantly, and their orientation becomes optically uniform (Fig. 4E). This means that the crystals grew in accordance with the rule of competitive growth.

Additionally, large degradation surfaces appeared in external regions that had several thin sections; this is where the destruction and subsequent overgrowth of younger crystals occurred, which formed in different directions than the original crystals (Fig. 4F).

As mentioned previously, a microstratigraphic log was constructed for several parts of the flowstone using series of codes to identify fabrics similar to those proposed by Muñoz-García et al. (2016) and Frisia (2015). According to these authors, columnar compact fabric, (C) – 1, corresponds to the formation of crystals under a stable growth regime and a warm climate with a persistent water film, which enables the undisturbed crystallization of large, well-formed crystals. Next in the profile is columnar compact with laminae, (Cl) – 2, which often alternates with columnar compact fabrics; its occurrence is probably associated with seasonal changes in local environmental conditions. During the dry seasons, a smaller amount of detrital material is available, and the crystallization of pure calcite occurs, while during the wet seasons, an increase in meteoric water flow causes a higher clay material content. The observed flat terminations of the crystals formed from a very thin film of water. The flat terminations of the crystals facilitate the capture of clay materials. Further along in the log is the columnar open, (Co) – 3, which forms from a thicker film of fluid than that of compact fabrics and with faster water supply and less efficient degassing (Frisia, 2015). Another type is the columnar elongated, (Ce) – 4, which is usually formed under rather stable conditions, when water delivery rate does not change significantly, but higher Mg/Ca ratios in the drip water than in the previous columnar fabrics. The occurrence of this fabric can be indicative of interglacial periods, when high flows would prompt degassing, increase the precipitation rate and promote the incomplete coalescence of crystals and the development of linear inclusions (Turgeon and Lundberg, 2001). The fifth fabric is columnar microcrystalline, (Cm) – 5, which has highly irregular intercrystalline boundaries, a uniform extinction, and is punctuated by inter- and intracrystalline microporosity. Columnar microcrystalline is typical in stalagmites in temperate regions that are characterized by significant seasonal contrasts in temperature, vegetation activity and autumnal rainfall. Its development also requires seasonal changes in cave ventilation, as a less efficient exchange between cave and atmospheric air occurs when the inflow of colloidal particles from the soil is greater. The coding scale presented in our work ends with a columnar fabric that has possible dissolutions, hiatuses and zones related to discontinuities and detrital layers, (Cd) – 6, which indicates the presence of the most unstable stage and corrosion and destruction conditions in the flowstone.

As a result of the calcite fabric analysis, it was possible to prepare brief descriptions of the conditions that could lead to the formation of these layers (Frisia, 2015; Turgeon and Lundberg, 2001) and the observed depths of their occurrences in the flowstone (Table 1). The presented profile is not complete because only 6 thin sections were prepared to document the most typical fragments of the flowstone, and some transitions had macro-scopically different calcite fragments.

U-series results

For two samples (277 and 278, Table 2), the ²³⁰Th/²³²Th activity ratios were below 20, indicating significant allogenic thorium contamination. These samples were taken near layers that were visibly enriched in detrital materials, which was probably the source of detrital thorium.

The U content in all samples, except for the two contaminated samples, was relatively low (below 0.1 ppm), which was typical for speleothems from the Kraków-Częstochowa Upland. Only two samples (333 and 334, Table 2), which were located in the upper part of the flowstone profile (depths of 107 and 74 mm, respectively; Table 2), had ages within the method limit, but they had large uncertainties. The other samples had ²³⁰Th/²³⁴U activity ratios near 1, which suggests that the samples exceeded the age limit of the U-series method (>500 ka). Furthermore, the ²³⁴U/²³⁸U activity ratios for those samples (except the lowest one) remained in an isotopic disequilibrium state. This suggests that the measured sample ages were not older than 1.2 Ma. In summary, the results of the U-series dating only suggested that the studied flowstone crystallized in a time interval between 0.5 and 1.2 Ma, and only the basal layer may be older than 1.2 Ma.

Oxygen and carbon isotopic compositions

A total of 116 samples have been analyzed for the oxygen and carbon isotopic compositions, and as a result, we obtained the δ¹⁸O and δ¹³C records (Fig. 5). The isotopic composition records show significant changes during flowstone sedimentation. The δ¹⁸O values differ from –8.74 to –6.45%o, while the δ¹³C values differ from –10.80 to –5.88%. The flowstone from the Głęboka Cave was divided into four sections and three discontinuity zones. The δ¹⁸O record has the strongest variability in the H3 zone, where its amplitude is 0.8%, while the H2 and H1 zone variability is ca. 0.5%. In contrast, a visible shift in the δ¹³C record shows the strongest value in the H2 discontinuity (0.8%); for the H1 and H3 discontinuities, the amplitude of δ¹³C is ca. 0.6%.

Fig. 5

Oxygen (A) and carbon (B) stable isotope compositions in the flowstone from the Głęboka Cave. The vertical lines indicate visible breaks in the flowstone deposition.

Chronology estimation based on the oxygen stratigraphy method

The age of the studied flowstone is outside of the U-series limit. Speleothems from the Kraków-Częstochowa Upland are characterized by a low uranium content, which is why there are problems with the application of the uranium-lead method regarding dating.

Therefore, the isotopic stratigraphy method was used to establish a more detailed chronology. Six dated points from the profile (Table 2) were utilized as temporal benchmarks. For the oldest point, an age older than 500 ka was assumed. For the younger layers, the time benchmarks were based on the dating results (samples 333 and 334, Table 2), or it was assuming that these ages had a uniform distribution from 500 ka to 1,200 ka. The δ¹⁸O record from the Głęboka Cave flowstone was correlated with the LR04 record (Lisiecki and Raymo, 2005) using GenCorr software (Pawlak and Hercman, 2016). The correlation results (Fig. 6A) show that the δ¹⁸O record from the Głęboka Cave fits the δ¹⁸O Atlantic record from 935 ± 5 ka to 470 ± 5 ka. GenCorr splits the correlated proxy into three parts, which are separated by deposition breaks (marked by grey sections in Fig. 6A). The age-depth model for the studied flowstone was constructed on the basis of the correlation results (Fig. 6B). The solid and horizontal black lines in Fig. 6B mark the depths of the hiatuses detected in the flowstone profile during the macroscopic observation. The solid grey sections in the age-depth graph mark the hiatuses detected during the correlation procedure. Two hiatus positions detected by correlation are in agreement with hiatuses H1 and H3 that are observed in the sample. According to the correlation results, the time interval for hiatus H1 ranges from 505 ka to 540 ka, and the time interval for hiatus H3 ranges from 760 ka to 790 ka. The H2 hiatus was not detected by correlation, which suggests a short duration deposition break that was connected with a large supply of detritus. The age–depth model shows rapid changes in the sedimentation rate. From 875 ka to 790 ka, the sedimentation rate is slow. It increases after hiatus H3. The time period between 740 ka and 620 ka is characterized by the fastest sedimentation rate, which comprises approximately 50% of the flowstone profile that crystallized during this period of time. The sedimentation rate slows down after 620 ka and remains slow until 570 ka. After 570 ka, the H1 sedimentation rate increases again. After the hiatus, the H1 sedimentation slows down and finally ends at approximately 470 ka.

Fig. 6

Oxygen stratigraphy results for the analysed flowstone. (A) 1 – δ¹⁸O record from the Głęboka Cave flowstone; 2 – discontinuity surfaces; 3 – the foraminifera δ¹⁸O record from the LR04 profile (Lisiecki and Raymo, 2005); (B) Age-depth model.

Chronology and sedimentation rate

One of the major objective of the present work was to obtain a chronology of the cave flowstone profile. A first attempt was made using U-series dating. However, the results showed that the studied flowstone age was out of the range of the ²³⁰Th/U method. On the other hand, the ²³⁴U/²³⁸U ratios dated the profile between 1,200 ka and 500 ka. Finally, isotopic stratigraphy was used as a tool to develop a more detailed chronology. The results of the isotopic stratigraphy correlation placed this flowstone profile in the time interval between 935 ± 5 ka and 470 ± 5 ka. The flowstone from Głęboka Cave was divided by three hiatuses. Two of them, H1 and H3, were also detected by numerical correlation. The detection of two hiatuses by numerical correlation confirmed the reliability of the calculated age-depth model. The deposition break time for H2 may be too short for detection by correlation at this record’s resolution.

The speleothem’s growth rate varies from approximately 0.4 mm up to approximately 5 mm/ka and is in range of the typical values reported for flowstones (Ford and Williams, 1989). The growth rate is positively correlated with temperature, thickness of water film and calcium ion concentration in seepage water (Fairchild and Baker, 2012). The fragment located between depths of 110 mm and 340 mm has the faster sedimentation rate and comprises ca. 50% of the whole flowstone profile. This fragment has been crystallized for over 120 ka, from 740 ka to 620 ka, which was during the late MIS 18, MIS 17 and MIS 16. MIS 17 is considered to be a relatively weak interglacial period, while MIS 16 is one of the strongest glacial periods over the last 800 kyr (Lang and Wolff, 2011). Calcite fabrics from this part of the flow-stone also indicate conditions that are favourable for speleothem growth (constant water delivery and thick water film; Table 1). On the other hand, breaks in deposition (H1 and H3), which are determined both in the macroscopic and numerical analyses, are dated during weak interglacial periods (MIS 13 and MIS 19). The fast growth rate during MIS 16 and the deposition breaks during the interglacials suggests the crucial role of precipitation amount (and water film thickness) in controlling the growth rate. The dominant role of aridity in the cessation of speleothem growth has also been described for younger sediments in other regions of Europe (Stoll et al., 2013).

Climate change recorded in the speleothems

Results of the isotopic measurements show high variability in the environmental conditions during the formation of the studied flowstone. The time range of the isotopic record spans from MIS 24 to MIS 12 (Fig. 7A). The oldest record in the flowstone is characteristic of a cold period with low vegetation (MIS 24). However, this period was warm and humid enough to enable the start of flowstone crystallization. The final part of MIS 22 was characterized by relatively higher temperatures and a rather smooth transition into MIS 21 (e.g., Hernández-Almeida et al., 2013). Additionally, the results of the microfacies analysis confirm the variability of the environmental parameters during the analyzed time. Initially, the observed columnar open fabric is later replaced by columnar compact and sometimes with faint laminae (Fig. 7B). This probably indicates the transition from a thicker film of fluid with faster water supply rate during a short period of time (potentially connected with thawing conditions) to a thin film of water with a relatively low water delivery rate and enhanced ventilation (Kendall and Broughton, 1978; Frisia, 2015). Since approximately 850 ka, the flowstone record has shown a gradual transition into a period of warmer temperatures with lush vegetation (i.e., a gradual increase in the value of δ¹⁸O and a decrease in δ¹³C); this transition corresponds to MIS 21. The MIS 20 is weakly visible in the isotopic records, while MIS 19 is correlated with significant discontinuities (Fig. 7). A hiatus (H3) and the corrosion of an older part of the flowstone are clearly visible in the thin section. After this episode, the δ¹⁸O values show a gradual increase, while δ¹³C values are low. This indicates improved climatic conditions with high soil activity, developed vegetation and warmer temperatures (Fig. 7C). After the hiatus episode, the thin section shows a transition towards an elongated columnar fabric, which is associated with relatively turbulent flow. Turbulent flow prompts degassing, increases the calcite precipitation rate and promotes the incomplete coalescence of crystals and the development of linear inclusions (Turgeon and Lundberg, 2001). Then, a short episode of rapid changes occurs near 720 ka. This episode corresponds to a later stage in MIS 18 and is recorded as a rapid increase in δ¹³C compared with a rapid decrease in δ¹⁸O. MIS 18 is considered to be a glacial period that is divided by warmer episodes (e.g., Alonso-Garcia et al., 2011). In comparison with other colder episodes, MIS 18 is characterized by relatively high temperatures and humidity, which is reflected in the fast deposition rate. In Poland, as evidenced by pollen and paleosoil records, this period is included in the Podlasian interglacial (Marks et al., 2016). After this episode of rapid climate depletion, the isotopic record shows a constant improvement in climatic conditions. The obtained age-depth relation shows a noticeable increase in the deposition rate. It is a period of higher soil activity and vegetation development with conducive conditions for the formation of cave deposits. Two thin sections correspond to this period. The first thin section is characterized by the presence of columnar microcrystalline and elongated fabric, and the other is mainly composed of columnar compact (Fig. 7B). By realizing that the development of columnar microcrystalline is typical in stalagmites from temperate regions characterized by marked seasonal contrasts in temperature, precipitation and vegetation activity (Frisia, 2015), it can be assumed that there has been an improvement in the conditions and transitions towards stable growth regimes and warm climates with persistent water films, which enable the undisturbed crystallization of large, well-formed crystals specific to columnar compact fabric. These results contradict earlier studies on the Middle Pleistocene glaciation range in Poland (Marks et al., 2016), which suggest that the studied area was glaciated during the so-called Sanian 1 glaciation. Even if we consider possibility that this region had nunatak areas or an ice-free oasis, the carbon isotopic signal suggests the development of vegetation different from that in tundras. The period with a relatively fast deposition rate terminated at approximately 620 ka. The period with a slower deposition begins with a simultaneous increase in δ¹³C and δ¹⁸O values, which is evidence of relatively high temperatures and the deterioration of conditions suitable for soil and vegetation development. This leads to episodes of depositional breaks (hiatus H2) at approximately 565 ka at the end of MIS 15. This hiatus is short and is not identified by the age-depth model. Next, a cessation in deposition is dated from approximately 540 ka to approximately 505 ka (MIS 13). Again, the period preceding the break in deposition, as derived from the isotopic record, is relatively warm but arid. Similar conclusions can be drawn from the analysis of petrologic features in the flowstone. The part of the microstratigraphic log corresponding to this period records the transition from the Cc to the Cm fabric, which can be interpreted as a deterioration in sedimentary conditions connected with the hiatus. The fabric characteristics under destructive conditions are divided by fabric type connected with the formation of transitional basal crystals. The last thin section has similar characteristics. It is a record of alternating conditions, including the initial growth of the fabric, the episodes of cessation, and the crystallization or even the destruction of the flowstone. Such characteristics during this period result in the slowest growth rate in the top part of the flowstone (Fig. 7B). That is probably an evidence of the deterioration of conditions and interruptions in the feedwater source. It may also be connected to local perturbations in the seepage route between the epikarst and the cave. Anyway during that period the crystallization of the examined flowstone was finally ended.

Fig. 7

The variability of δ¹⁸O (B) and δ¹³C (C) isotopes, with marked discontinuities in the surfaces, along with oxygen isotope stages (A), insolation at 50°N (D) and a fabric log (E) with marked as ellipsoids locations of thin sections on the timescale.. 1 – discontinuities in the surfaces; 2 – boundaries of oxygen-isotope stages (Lisiecki and Raymo, 2005).

In summary, the analyzed record contains a clearly visible succession of marine isotopic stages from MIS 24 to MIS 12 (Lisiecki and Raymo, 2005; Martin-Garcia et al., 2015). The detected ages of the hiatuses in the flowstone from the Głęboka Cave correspond to MIS 13, MIS 15 and MIS 19. The hiatuses correspond to periods of deteriorating sedimentary conditions and, consequently, the cessation of flowstone crystallization and its partial erosion. Clearly, all recorded hiatuses occur during interglacial periods, which suggests that a major factor controlling the flowstone growth rate is the precipitation amount. The smaller role of temperature in this process confirms that the relative strength of cold periods in the Middle Pleistocene is significantly weaker than that during the last 400 ka. Eventual cooling does not cause flow-stone growth cessation; it even favours its deposition by increasing the humidity of the climate, at least in Central Europe.