Part of the research leading to these results was funded by a grant of the Romanian Ministry of Education and Research, CNCS - UEFISCDI, project number PN-III-P4-ID-PCE-2020-2282 (ECHOES), by the EEA Grants 2014-2021, under Project contract no. 3/2019, and by the SMIS 2014+ 120009 project (PI M. Vlaicu). We would like to thank the mayoralty of Mangalia and the Romanian Aeronautical Authority for the approval of UAV flights for photogrammetry. We would also like to thank two anonymous reviewers for their helpful comments.

1. Introduction

1The occurrence of oolitic-lumachellic Sarmatian limestone, as well as the presence of mesothermal groundwater rich in H₂S, has created conditions for hypogene karst development on the southern Dobrogea Plateau. There are few cave occurrences in this region, of which Limanu and Movile are the most important. Limanu Cave is a two-dimensional maze with important human impact onto its passage morphology and a total length exceeding 3,500 m (Dumitrescu et al., 1965; Drăgușin et al., 2019). Movile Cave on the other hand is small, only ~200 m long, actively forming under the influence of hypogene processes (Sârbu and Lascu, 1997; Sârbu et al., 2019). More than the caves, the most pervasive karst form is represented by dolines, resulting from both dissolution and collapse (Lascu et al., 1995; Constantinescu and Constantin, 2001). A handful of very large collapse dolines are located west and north of the town of Mangalia, the largest (>1 km in diameter) being the Mangalia Swamp.

2Apart from having a sedimentary infill of more than 70 m thick (Constantinescu, 1995), the ~300-m diameter Obanul Mare collapse doline is surrounded by a large number of small hillocks of uncertain origin. They form a unique landscape in Romania and similar morphology was neither observed in neighboring Bulgaria, where the same geological setting exists. Lascu et al. (1995) proposed that it represents a residual relief, and Onac and Drăgușin (2017) suggested these hillocks might be the remnants of an unroofed maze cave. Although the surface terrain appears very similar to a small-scale labyrinth karst, it is obvious that its origin might be different from that of the landforms in humid environments (Brook and Ford, 1978; Haryono and Day, 2004).

3Our study stems from the hypothesis that if there were a maze cave at Obanul Mare and its ceiling collapsed, then the hillocks should represent the former walls separating the passages. After being unroofed, these inner walls should have become subject to subaerial weathering and their profiles would have tended towards equilibrium, sloping in the direction of the former passages or voids. As a result, the aspect of these slopes should be perpendicular to the general orientation of local fractures along which cave passages have developed. We aim to test this hypothesis by quantifying the aspect of these hillocks based on a high-resolution digital surface model and explore its relationship with local and regional fault orientations, as well as with the orientation of the Movile cave passages. Local structural lineaments could be extracted from surface landforms such as dolines and valleys, as they commonly develop along faults and other fractures, and the long axis of dolines corresponds in many cases to local fracture orientations (Day, 1976; Deike, 1989; Öztürk et al., 2018; Siska et al., 2016; Valle, 1968; Williams, 1972).

4Two-dimensional maze caves develop close to the water table level, where all available discontinuities are enlarged (Palmer, 1991). Identifying former maze networks close to the Black Sea shore can lead to a better understanding of past water tables and their relationship with sea level changes and local tectonic activity.

5Finally, our study could bring light onto the hypogene processes that generated the karst relief of this region, as well as bringing more information on the conditions that shaped the chemoautotrophically-based ecosystem thriving in the underground sulphidic environment of Southern Dobrogea.

2. Materials and methods

6The tectonic setting of Dobrogea is dominated by a system of crustal faults (Peceneaga-Camena, Capidava-Ovidiu, and the Intramoesian Fault), with a general NW-SE orientation (fig. 1). In southern Dobrogea, the Rasova-Costinești, North Mangalia, and Mangalia faults are strike-slip faults of similar orientation (Zamfirescu et al., 1994; Zamfirescu et al., 2010). According to Dinu et al. (2005), between the Capidava-Ovidiu and the Intramoesian strike-slip faults there is a horst and graben structure defined by two fault systems, oriented WNW-ESE and NNE-SSW (fig. 1).

7Movile Cave is located on the eastern side of the Obanul Mare collapse doline (fig. 2), at some 18-20 m below the hillocky surface. It has a dry level consisting of ~200 m of passages, and a short submerged level. It was discovered during prospection works for an industrial project and is accessible through an artificial shaft (Lascu et al., 1995; Sârbu et al., 2019; Sârbu and Lascu, 1997). It is comprised of a main, large passage, oriented WNW-ESE, which is blocked towards the WNW by a boulder accumulation. The side passages are mostly of reduced volumes and, as in the case of the main passage, exhibit morphological features such as corrosion pockets and cupolas. The cave provides access to the local, mesothermal aquifer from which chemoautotrophic microorganisms oxidize sulphur, methane and nitrogen compounds to form the base of a rich ecosystem isolated, at present, from the surface (Kumaresan et al., 2014; Brad et al., 2021). According to Sârbu et al. (2019), 51 invertebrate species were identified within the cave, of which 34 species are endemic, making Movile Cave and the local hydrothermal aquifer a biological hotspot. The paleontological study of Știucă and Ilinca (1995) showed that the sediments within the cave contain Late Pleistocene rodent fossils, as well as fish and frog remains. Later, Engel et al. (1997) identified fluvial deposits in the main passage, indicating an earlier fluvial phase in the cave evolution. At present, the most important geomorphic process is condensation corrosion (Sârbu and Lascu, 1997). According to Horoi (1994) the oolithic limestone walls display a stratification that goes from clayey insoluble remains at the interface with the atmosphere, to an oolithic mass with no cement, and finally to the unaltered limestone, a setting that is in agreement with the notion of ghost rock karstification (see the review of Dubois et al., 2014).

8A high-resolution surface model was obtained by performing photogrammetric flights using a DJI Phantom─4 unmanned aerial vehicle, covering an area of ~18 km² west and north of the town of Mangalia, as well as above Limanu cave. Images were collected at a speed of 5 m/s, with a vertical angle of 90, and a side and forward overlap of 70%. To increase horizontal and vertical accuracy, ground control points were collected in the field using a Trimble GeoXH 2008 differential GPS. The root mean square error for all the points was 0.21 m. High-resolution digital surface models (DSM) were generated by using structure from motion photogrammetry algorithms implemented in the Drone2Map™ software (fig. 3).

9Using the ArcGIS 10 software package, we extracted the area occupied by hillocks at Obanul Mare and resampled the DSM at a resolution of 50 cm. On the newly obtained terrain model, we determined the aspect of each pixel and extracted its value using the Raster to Point tool. The aspect map is displayed on 45° bins (fig. 4), where North (N) is defined by values between 337.5° and 22.5°, North-East (NE) between 22.5° and 67.5° and so forth (tab. 1). Further, considering that slopes facing one direction and their counterparts (180° opposite) would be perpendicular to the same fracture direction, we summed the values of opposing aspects and expressed them as N-S, NE-SV, E-W, and NW-SE. If we consider slope facets to be perpendicular to hypothetical fracture lines, then adding 90 degrees to each aspect class would return values for the orientation of these hypothetical fracture lines.

10We referred to the structural features extracted from the DSM under the general term of ”fractures”, with no distinction between joints, faults and other alike, since this task goes beyond the purpose of this paper. However, the term “fault” was used whenever we referred to well-documented such tectonic features (Zamfirescu et al., 1994; Dinu et al., 2005).

11From the high-resolution DSM near Mangalia, we extracted contour lines at 1-m intervals, allowing us to identify surface karst features such as dolines and fluvial valleys, which are represented in Figure 2. An oblique view is used for a better perspective on the hillock morphology (fig. 3). The fluvial valleys and doline long axes identified on the DSM and topographic map (scale 1:25,000) were therefore extracted as line features. Their orientation values were binned in 22.5° intervals, similar to those of the aspect. For each 22.5° bin, the total length of drawn lines with orientations pertaining to that specific bin was summed and expressed as percentage of the total length of these lines. In addition to that, fault lines close to our study area (fig. 1) were digitized from Zamfirescu et al. (1994) and split into 1 km long segments. The orientations of these segments were calculated as for the local alignments and expressed as percentage of their total.

12We surveyed Movile Cave using a modified Leica Disto X310 laser distance meter with an integrated compass and clinometer (Heeb, 2009). The length and orientation of cave passages (fig. 5) were extracted from the map as lines and treated similarly to the local lineaments (doline axes and valleys). We also surveyed Limanu cave in its entirety, and we present here some 2/3 of its total length, which are relevant for the present study. The passage orientations of Limanu Cave are not used in this study due to the lack of detailed information regarding the distinction between passages that were entirely created by human activity and those that retain their initial morphology and orientation.

13Above Limanu cave, the presence of trees made the observation of the relief difficult on the DSM. We extracted contour lines at 1 m intervals based on the DSM and we manually corrected them combining the original data with field observations.

3. Results and discussion

14Figure 6 shows the orientation of hypothetical fractures alongside orientation frequencies of local fractures, and regional fault lines, on 22.5° bins. The frequencies of aspect values, calculated for each degree (fig. 6D) and show that the terrain generally faces E and W, and ENE and WSW, the two directions encompassing a total of 30% of all the values. The E-W and WNW-ESE orientations of regional fault lines (comprising 53% of the total) correspond to N Mangalia and Rasova - Costinești faults. Local fracture line orientations show more noise (fig. 6B). The E-W and WNW-ESE orientations that would pertain to minor fractures parallel to the Rasova-Costinești, North Mangalia and Mangalia major faults comprise 19% of the total, while the NNE-SSW-oriented fractures amount to only 7%. The N-S fractures sum up to 18%. The NNW-SSE orientations, accounting for 26% of the total, complemented by NW-SE with an additional 27% of the occurrences are the most abundant and could be normal faults associated with the strike-slip fault system.

15Where hypothetical fractures (i.e., fractures derived from orientation of the downward sloping terrain) are concerned (fig. 6C), the results are even noisier, probably due to the fact that the intervening surface erosion that modelled the hillocks has created different terrain aspect not related to the initial inner walls of a maze cave. The frequency of the orientation values of these hypothetical fractures spread between 10% and 15%. The E-W and WNW-ESE orientations related to the main strike-slip faults amount to 25%, whereas the NNW-SSE and N-S amount to 30%. The correlation coefficient between hypothetical and local fracture lines is 0.8 (p-value = 0.02).

16The most important orientation of Movile cave passages is NNW-SSE (18%), followed by WNW-ESE (16%), the latter being represented mostly by the main passage. The N-S orientation is surprisingly underrepresented, with only 5% of the total passage length. The correlation plot between hypothetical fractures and Movile cave passage orientations (fig. 6F) shows an outlier for the N-S orientation, but if this outlier is omitted from the calculation, the correlation factor becomes 0.7 (p-value = 0.05).

17A strong correlation between hypothetical and local fractures indicates a high probability that the space between the hillocks formed along all available local fractures. As maze caves, too, usually form along all available local fractures, there is a high likelihood that the space between the hillocks represents former cave passages.

18Similar, but less abundant hillocks appear on the southern and western sides of the nearby Blebea collapse doline, although those on the western side have been levelled by the construction of terraces. Moreover, we recognized similar morphology outside of Limanu Cave and on a small area within the village of Limanu (fig. 2).

19At Limanu cave, where the overlaying rock is only a few meters thick, two passages are open to the surface and offer access to the cave (fig. 7). Other passages usually end with collapsed blocks, sediment, and soil accumulation, due to their closeness to the surface. Outside, their extremities correspond to depressions that we consider to be resulted from the collapse of these passages. These depressions are separated by small hillocks that correspond to the walls between the passages. Their position close to steep slopes prevented the formation of closed depressions except for one instance on the northern side of the hill. As seen in Figure 8, these depressions and hillocks resemble those from around Obanul Mare, although they are smaller, but their presence offers strong support for the hypothesis of a former maze cave at Obanul Mare. Moreover, they are found in the same altitude range as those from Obanul Mare (~15-25 m), implying that Limanu and the former cave from Obanul Mare might have been part of a cave network formed by a single water table.

20Although this relief resembles the labyrinth karst, this region did not benefit from an increased pluvial regime in the recent geological past. Increased rainfall amounts and temperatures were present in the region surrounding our site during the early Pliocene. Marine sediments from the SW Black Sea, originating from Anatolia, contain fossil pollen taxa of subtropical vegetation from before 4.3 million years ago (de Leeuw et al., 2018; Feurdean and Vasiliev, 2019). During the same period, a few hundreds of kilometers west of our site, climate reconstructions based on fossil plant assemblages indicate that the temperature was some 3-4°C warmer than today. At the same time, precipitation was as high as 1,500 mm, compared to 700 mm at present (Ticleanu, 1995; Iamandei and Diaconu, 2013).

21Regarding the age of the karst in Dobrogea, Nicod (1977) suggested that its formation, both at the surface and at depth, took place during the wet periods of the Pliocene. Lascu et al. (1995) described, in higher detail, three karst levels found near Mangalia and proposed that the lowermost of them, found at around 150-180 m depth, was formed during the Messinian Crisis, when the Black Sea level dropped dramatically.

22Even if these hillocks would have formed in a wet climate, their preservation for millions of years with intervening glacial and interglacial phases would have been difficult. At the least, their high density suggests their horizontal dimensions did not change greatly since their formation.

23Although the origin of the deep karst conduits could be assigned to the Messinian Crisis or a similarly dramatic marine regression, outstanding questions still remain, regarding the origin of an extensive karst network developed very close to the present day surface. It is also interesting to learn if such a karst network was formed by the sole action of the sulphidic water or were there periods when it was mixed with meteoric or even sea water. Recognizing that an extensive karst network existed in the area, as resulting from our work, is a first step in quantifying the evolution of this karst region. Our ongoing studies are tackling these questions using a range of investigative techniques such as remote sensing, geophysical investigations, dating of karst sediments, and stable isotope analyses.

4. Conclusion

24In this study we investigated the relationship between geological structure and the hillocky relief around Obanul Mare collapse doline, near the town of Mangalia, SE Romania. This morphology was hypothesized to represent the weathered pillars that were separating the passages of a former maze cave whose ceiling collapsed. After becoming exposed to subaerial erosion, the slopes of these pillars became oriented towards the supposed former cave passages (and hypothetical fracture lines).

25A high-resolution digital surface model allowed us to automatically quantify slope aspect of this hillocky terrain and to identify the most pervasive orientation that would be in relation to the orientation of former cave passages. From the same digital surface model, we extracted doline and fluvial valley axes, which usually form along local fracture lines.

26The comparison with regional fault lines shows that local lineaments fit to the shear structural assemblage characteristic for southern Dobrogea. Furthermore, hypothetical fractures and local lineaments show a good correlation, with a coefficient of 0.8 (p-value = 0.02), indicating that the hillocks were delimited between possible fracture lines. Moreover, a strong correlation coefficient (0.7) is obtained in relation to the passages of Movile Cave.

27We also showed that there is a good morphological resemblance between the hillocky terrain at Obanul Mare and that resulting from the collapse of passage extremities from the nearby Limanu cave. These arguments strengthen the hypothesis that the hillocky terrain is actually the remnant of a former two-dimensional maze cave, exposed to surface by ceiling collapse, in a tabular structure.

*Corresponding author: Tel: +40 213 188 132
virgil.dragusin@iser.ro (Drăgușin V.)