Towards a Better Understanding of Subsurface Processes in the Evolution of Sandstone Tablelands—Patterns and Controls of Contemporary Sand Removal from Sandstone Caprock, Stołowe Mountains Tableland, SW Poland

Abstract

This paper reports the results of two-years’ monitoring of sand evacuation from a sandstone tableland through fissure systems and discusses the findings in the context of geomorphic evolution of tablelands, especially addressing the role of subterranean processes. A field experiment using specially designed sand collectors and involving sampling in approximately monthly intervals was carried out at six sites in SW Poland, representing two adjacent but contrasting settings: a mesa and a cuesta front. Data about sand deposition were then analyzed against precipitation data from a station located next to the sites. Sediment volumes deposited during the monitoring period were considerably different between the sites, with those at the mesa much higher than those at the cuesta. This is attributed to strong structural control influencing groundwater circulation pathways and the size of underground drainage systems, which were much smaller next to the cuesta front. Relationships between denudation and precipitation are complex, although the role of very high rainfall events appears clear, especially for the mesa. In general, precipitation in excess of 70 mm during a few consecutive days typically resulted in a considerable outflow of sand. This study highlights the role of mechanical underground erosion in sandstone tablelands, long neglected, and quantifies the denudation process.

1. Introduction

There is a general agreement that the geomorphological evolution of sedimentary tablelands, capped by more resistant beds (caprock), proceeds via parallel escarpment retreat, which may lead to the fragmentation of once-larger plateaus into an array of residual mesas, buttes and pinnacles e.g., [1,2,3,4,5]. In a classic scenario [6] (see elaborate discussion in [7]), the process is facilitated by surface erosion of the subcaprock slope, whose resultant oversteepening causes rock-slope collapses of the overlying caprock unit and hence the backwearing of the entire escarpment, e.g., [8,9,10]. However, if porous and thus highly permeable sandstone acts as caprock and overlies strata of lower permeability, rock slope retreat is often controlled primarily by processes operating at depth, with subsurface water flow as a driving agent [11]. Its roles are diverse and largely dependent on local geological, geomorphological and environmental circumstances. On one hand, the emergence of water at the base of a cliff results in seepage weathering [12], whose sustained activity leads to cliff undercutting and rock-slope collapses, i.e., the phenomenon known as sapping, widely recognized in the U.S. Southwest [11,13,14]. The saturation of rocks at the contact between permeable and impermeable units, as well as plastic deformations of the latter, are also known to be major causes of a variety of large-scale slope instabilities, including rotational landslides (e.g., [15,16]), block slides [17], and toppling phenomena [18]. On the other hand, if subsurface water flow is confined to the well-developed network of vertical fractures cutting sandstone caprock, non-catastrophic processes often dominate, with in situ disintegration of marginal parts of sandstone plateaus and the gradual subsidence of large, cuboidal sandstone blocks into voids created by the removal of weathered rock [19,20]. As a result, once-continuous cliff lines transform into rock cities and ruiniform relief, with criss-crossing corridors and slots separating the remaining bedrock compartments [21,22,23,24,25,26,27].

As demonstrated above, the complexity of processes contributing to the retreat of escarpments in sandstone-capped tablelands seems to be increasingly well recognized. Yet, specific controls standing behind these processes are insufficiently understood [28]. In the majority of studies conducted so far, various dating techniques have been applied to study the time frames, rates of escarpment retreat and sandstone caprock disintegration, utilizing the potential of talus flatirons (e.g., [29,30,31]), boulder aprons [32,33], and cliff-foot sand deposits [34]. The results of dating have then usually been compared with paleoclimatological data, and on this basis, environmental controls of the changing dynamics of geomorphic processes responsible for escarpment retreat across millennia have been inferred, e.g., [31,32,33,35,36]. At the same time, however, patterns and controls of contemporaneous processes involved in the retreat of sandstone-capped escarpments, except rockfalls (e.g., [37,38,39,40]), have rarely been investigated in detail, especially with respect to subsurface phenomena. For example, little is known about the contemporary activity of sapping or the efficiency of fracture-guided underground erosion. Regarding the former, a lack of direct monitoring results in controversies and divergent views on the activity of seepage weathering and the resultant cliff collapses under the present-day environmental regime [see reviews in [12,28,41]. In the case of the latter, it remains virtually unknown what external factors facilitate subsurface removal of sand—a key process in the development of sandstone rock cities and ruiniform relief [23,25,42,43]. In fact, the evidence of the process at work is limited to non-systematic observations of fresh sand sheets, likely appearing after heavy rainfall events [20,24]. The precipitation thresholds beyond which the residual sand produced by weathering in the system of fissures in caprock becomes mobilized have so far not been recognized at all. Likewise, the volumes of mass that may become removed in certain conditions across a year have not yet been calculated, according to our knowledge. Similarly, the spatial variability of contemporary underground erosion remains vague, and hence, the influence of caprock properties and topographic conditions on the efficiency of the sediment supply system could have been reconstructed solely on the basis of depositional landforms formed throughout the entire Holocene [34].

In this paper, we attempt to fill the gap in knowledge outlined above, focusing on patterns and controls of contemporary sand removal in the Stołowe Mountains tableland, Middle Sudetes, SW Poland. The essential role of underground erosion in the disintegration of quartz sandstone caprock has already been recognized [16,24,34], making the area particularly suitable for this kind of research. However, the implications are wider, and the findings reported in this paper may contribute to a better understanding of escarpment evolution, also inspiring analogous studies in other settings. Our study is based on a two-year field experiment involving monitoring of sand removal with monthly resolution, which helps to recognize meteorological controls of underground erosion. To study geological and geomorphological controls of the process, the experiment was conducted in two different parts of the study area, with dissimilar characteristics of the caprock and topography of the escarpments.

2. Study Area

2.1. Geological and Geomorphological Context

The Stołowe Mountains tableland represents structure-controlled relief, developed upon a thick succession of Upper Cretaceous clastic sedimentary rocks of shallow marine origin, where the attitude of beds is generally close to horizontal. The stepped topography, with planar surfaces at various altitudes separated by steep escarpments, reflects an alternation of more-resistant sandstones and less-resistant mudstones, marls and other finer-grained deposits in the lithostratigraphic column [44,45] (Figure 1). The uppermost storey, the subject of particular interest in this study, consists of a thick (up to 50 m) caprock of Upper Turonian/Lower Coniacian quartz sandstone, underlain by approximately 100 m thick series of finer-grained rocks. The altitudes within the caprock reach 922 m a.s.l. at Mt. Szczeliniec Wielki. The sandstone shows large-scale planar or cross-bedding, typically sparsely jointed in an orthogonal pattern [46]. It supports vertical cliffs, locally as high as 30 m, which constitute a significant part of the perimeter of the remnant plateaus and isolated tabular hills (mesas), whose nearly level or gently sloping upper surfaces follow major sub-horizontal bedding planes. However, in other places, sandstone cliffs are replaced by blocky taluses or steep slope sections with scarce outcrops, pointing to different pathways of caprock disintegration [20]. Common features along cliff lines are narrow clefts and slots due to the widening of primary vertical joints, localized vertical zones of heavily broken (disintegrated) sandstone, and sandy cones attached to the base of the cliff, with apices at joint-aligned fissures. The slopes below the cliffs are typically covered by large cubic blocks of sandstone, up to 10 m in length or so. As demonstrated in [19], the genesis of those blocks is related with in situ caprock disintegration as well as mass wasting processes (rockfall, sliding, toppling).

Whereas the landform diversity of the Stołowe Mountains and their geological controls is reasonably well known, including DEM-based quantification of various relief features [44,45], functional relationships within the regional geomorphic systems are insufficiently understood. The stepped topography of the tableland was long explained in simple terms of rockfall-driven escarpment retreat, followed by pedimentation [47]. Later on, Pulinowa [48] directed attention to chemical denudation within the fine-grained series and other subsurface processes contributing to rock mass loss. The view implying the dominance of catastrophic rockfalls was challenged by Duszyński and Migoń [19], who argued for gradual, largely non-catastrophic disintegration of caprock slopes. The long-term retreat of the sandstone-capped escarpment has been recently quantified to occur at a mean rate of 4 m/kyr, but with significant variability related to changing environmental conditions [32]. Subsequent studies provided further observational evidence consistent with significant removal of sand taking place in the subsurface, mainly in the form of subsidence landforms and closed depressions [24,49]. However, the processes themselves were still evaluated qualitatively rather than quantitatively.

Relating to the broader context, the Stołowe Mountains are part of a large area underlain by the Upper Cretaceous clastic succession, known as the Bohemian Cretaceous Basin, and extend from east Germany across northern Czechia to Poland [50]. The area shares many geomorphological characteristics with the Stolowe Mountains, such as the occurrence of plateaus, cliffed escarpments and residual tabular hills (mesas) [5,51,52,53,54]. Cliffs and mesas are also typical for many sandstone tablelands worldwide [55,56]. Nevertheless, we are not aware of any field-based experiments focused on mechanical denudation similar to ours.

2.2. Contemporary and Past Environment

The contemporary regional climate is humid temperate. On the upper plateau of the Stołowe Mountains, the mean annual temperature is 3–4 °C, and precipitation is around 900 mm, distributed evenly throughout the year [57]. Snow cover may be present from mid-November to early April, occasionally until May in deep sheltered clefts, but the actual number of days with snow cover in a year varies considerably: between 50 and more than 150 days [58]. Episodes of heavy rainfall which generate overland flow occur, but they are generally not very common, with a frequency of up to a few times per year.

Prior to the onset of the Holocene, the region was typified by a cold climate, with permafrost [59], but the tableland was not sufficiently high to provide conditions for the development of local glaciation. Thus, geomorphic processes at work were likely those typical for periglacial environments, including frost weathering, solifluction and surface wash. Considering the presence of permafrost, it may be also hypothesized that the efficacy of geomorphic processes related to water presence and flow in the subsurface was considerably retarded. The hydrogeological systems were unblocked during the transition to the Holocene, resulting in accelerated caprock disintegration and escarpment retreat [32]. However, no evident diagnostic periglacial landforms or deposits were recognized in the tableland, probably due to very strong lithostructural control [45].

2.3. Study Sites

2.3.1. Białe Skały

Białe Skały (White Rocks) is the name given to the 1.5 km long western section of the northern escarpment of the Narożnik Plateau (Figure 1A), where tall sandstone cliffs and spurs are abundant, although not continuous. The escarpment is approximately 80 m high, with rock cliffs developed in the quartz sandstone accounting for the uppermost 15–20 m. Beneath the sandstone, within the middle and lower slope, the finer-grained members of the sedimentary succession occur, mainly mudstones. These, in turn, are covered by sandy slope deposits, whose thickness may reach 3 m or more [20,60]. The sandstone cliffs are dissected by ravines and troughs and apparently more weathered, with abundant weathering microrelief (arcades, horizontal recesses, honeycombs). Locally, their dissection is so dense that small rock cities have developed. The upper sandstone slab is not horizontal, but tilted to the south, with the resultant inclination of the plateau surface at an angle of approximately 5–10°. Consequently, the escarpment bears similarities to a cuesta, with the escarpment face and the backslope inclined in the opposite directions.

Landforms indicative of subsurface evacuation of sand derived from sandstone disintegration are rare. Small, closed depressions are present above the escarpment rim, and roofed slots (fissure caves) locally extend from the cliffs into the rock massif. In addition, geophysical surveying strongly suggested the existence of deep zones of caprock disintegration [27]. Sandy cones are present, too [20]. However, deep boulder-filled troughs, blind corridors and closed plazas above the rim of the escarpment, within the cuesta backslope, are absent.

Three sand collectors were installed at the Białe Skały locality, with two of them (SC-BS-1 and SC-BS-2) close to each other, within a small rocky courtyard in the eastern part of the massif (Figure 2A), and one (SC-BS-3) in the central part, at the foot of the north-facing cliffs (Figure 2B). Both the SC-BS-1 and the SC-BS-2 sand collectors were installed between sandstone crags up to 12 m high, bounding the rocky courtyard from the north and south. These crags are the protruding sections of the plateau extending further to the west and, together with more isolated sandstone towers in the vicinity, give rise to classic ruiniform topography [27]. The sand collectors were installed at the outlets of fissures striking at 20–30°, perpendicular to the general course of the cliffs, beneath the basal overhangs in the rock walls < 1.5 m deep. The width of the fissures outlets is c. 30 cm at SC-BS-1 and c. 10 cm at SB-BS-2, and the floor of both is covered with sand. In front of the fissures, at the foot of cliffs, small sandy cones are developed, with fresh sand accumulations on top. The upper surfaces of crags are densely overgrown by birch (Betula L.), Norway spruce (Picea abies L.) and blueberry shrubs (Vaccinium L.), with very acidic soils (pH 3.2–3.6), rich in organic matter. Organic soils developed within the fracture-guided depressions in the plateau surface are locally very thick, even in excess of 1 m [27]. The overall geomorphic situation of the SC-BS-3 sand collector is different (Figure 2A,B and Figure 3). The cliffs are higher, reaching up to 20 m, and show greater continuity than in the eastern section of the Białe Skały massif. Nevertheless, the marginal part of the plateau is also characterized by a sequence of alternating spurs and recesses. The SC-BS-3 sand collector was installed at the outlet of a fissure cutting one of the most distinctive spurs. The fissure, striking at 120°, penetrates deep into the massif; its outlet is 20 cm wide, and its floor is covered by substantial amounts of sand. Likewise, the recess neighboring the spur from the west is filled with huge volumes of sandy and blocky material.

2.3.2. Szczeliniec Wielki

Mt. Szczeliniec Wielki is a distinctive mesa and, at the same time, an isolated remnant of the uppermost morphological level of the Stołowe Mts. tableland (Figure 1B). It has an oblong shape, approximately 630 m in length and 330 m in width, covering an area of nearly 0.17 sq km [61]. The top part of the mesa is built of a horizontal slab of quartz sandstone, approximately 50 m thick, underlain by a thick series of finer-grained rocks, mainly mudstones; however, the contact itself is not exposed, but covered by talus. The total height of the mesa in relation to its footslope is up to 150 m. In contrast to Białe Skały, sandstone rock cliffs are almost continuous along the perimeter of the mesa, locally reaching more than 30 m in height, with only a few narrow (10–30 m wide) zones of disintegration into boulder chaos, which allow for access to the mesa top. The top part itself represents ruiniform relief of considerable diversity, with individual crags (tors) up to 12–13 m high, rock labyrinths with corridors following the main joint directions, and boulder mantles [24,61].

In the context of this paper, landforms evidencing the role of subterranean water movement and sand removal are of particular interest. The previous geomorphometric studies using a high-resolution (1 m × 1 m) DTM revealed that almost 90% of the mesa top surface is drained via subsurface flows, and closed depressions acting as sinks are abundant [24]. Geomorphological mapping, in turn, showed the presence of blind troughs, both boulder-filled and largely boulder-free, tilted blocks, deep clefts, especially near the northern margin of the mesa, roofed slots, and thick boulder accumulations (“boulder chaos”) in topographic hollows. All these features have been interpreted as collective geomorphic evidence of sandstone breakdown into sand and subsequent evacuation of sand grains by subterranean water flows along the network of fissures [24], aided by further in situ disintegration of sandstone by the mechanism of arenization [27]. Sandy cones attached to the sandstone cliffs at their base represent complementary evidence of subterranean sand removal and show exit points of the most efficient transport routes [20].

Three sand collectors (SC-SzW-1, SC-SzW-2 and SC-SzW-3) were installed close to one another in the southeastern section of the mesa’s rim (Figure 2C). There, the subvertical cliff lines are approximately 30 m high and show a zig-zag course following the orthogonal jointing pattern. At the foot of this section of cliff lines, some of the most distinctive sandy cones in the Stołowe Mountains have developed, with fresh sand sheets on top, indicative of contemporary activity of underground erosion [24]. The SC-SzW-1 sand collector was installed on the apex of a sandy cone that attains the height of 4 m and is situated directly at the outlet of a vertical fissure c. 20–25 cm wide and striking at 40° (Figure 2C and Figure 4). The most recent study employing OSL single-grain dating revealed that this cone records sedimentation lasting c. 7.3 ka [34]. The SC-SzW-2 and SC-SzW-3 sand collectors were installed more to the east, on the apex part of one of the biggest cones found in the Stołowe Mountains (c. 6 m high), developed between two sections of cliff lines crossing at right angles. The SC-SzW-2 sand collector was installed at the outlet of a c. 70 cm wide fissure, striking at 130° and partly roofed by boulders sandwiched in between the side walls. The floor of the fissure is entirely covered by substantial amounts of sand. Field inspection showed that while the fissure ends blindly, it is connected at its end with another one, much narrower, striking at 40°. This fissure, penetrating deeper into the massif, likely supplies the system with the sediment. The SC-SzW-3 sand collector was installed at the outlet of a very narrow fissure, c. 5–10 cm wide and striking at 40°. The near-margin plateau section immediately above the cliff lines is characterized by a bare rock surface, with deep fissures and scattered detached sandstone compartments, often of cuboidal shape.

3. Materials and Methods

3.1. Methodological Design

The design and execution of the study involved several steps. At the initial stage, monitoring sites were selected after field reconnaissance, followed by the design and installation of sand collectors. Then, the two-year period of observations was decided, including intervals of data collection in the field at roughly monthly frequency. Parallel to the regular inspection of field sites and measurements of sand volumes deposited in the collectors, meteorological data were collected. Termination of the monitoring period allowed us to begin the analytical part of the study, involving the recognition of temporal patterns of sand removal and quantification of denudation versus time. In the last step, meteorological and lithostructural controls on sand removal were analyzed.

3.2. Monitoring of Contemporary Sand Removal

3.2.1. Selection of Localities

The decision to install sand collectors, i.e., boxes collecting sand grains washed out from fissure outlets, at two different localities (with three study sites within each) reflected the intention to investigate whether the volume of sand removed from fissures is related to the structural and topographic setting. The mesa of Szczeliniec Wielki and the cuesta of Białe Skały are situated reasonably close to each other (c. 2.5 km; Figure 1) and at similar altitudes to minimize potential influence of spatial variability of meteorological conditions. Significant differences in geological and topographical conditions between the two localities included the following features: caprock thickness, dip of the strata, the proportion of bare rock surface exposed within the plateau, as well as the complexity of the fissure system.

Table 1. Location of sand collectors and the basic geological and geomorphological properties of the caprock system at each study site.

Locality	Study Site	Latitude	Longitude	Altitude, m.a.s.l.	Fissure Outlet Width [cm]	Caprock Thickness [m]
Białe Skały	SC-BS-1	50°27′32.43″ N	16°21′37.93″ E	776.2	30	8–14
	SC-BS-2	50°27′32.27″ N	16°21′37.85″ E	775.6	10	8–14
	SC-BS-3	50°27′38.41″ N	16°21′26.31″ E	786.8	20	22–26
Szczeliniec Wielki	SC-SzW-1	50°28′55.33″ N	16°20′38.54″ E	872.2	23	30–31
	SC-SzW-2	50°28′55.45″ N	16°20′38.98″ E	876.7	70	30–31
	SC-SzW-3	50°28′55.43″ N	16°20′39.05″ E	875.9	5–10	30–31

summarizes basic information about each study site.

3.2.2. Construction of Sand Collectors

To construct sand collectors, transparent plastic boxes of a volume of 61 l were used (Figure 3 and Figure 4). They were covered with plastic lids to prevent delivery of mineral or organic material other than that originating from the fissure. On one of the shorter sides of each box, rectangular holes were cut, through which the end parts of plastic conduits (0.5 to 1.0 m long) of rectangular cross-section (7.5 cm × 15 cm) were inserted. The opposite side of each conduit was then connected with the fissure outlet. If the fissure was narrower than the conduit (Table 1), additional plastic elements (such as smaller pipes, plastic mats, etc.) were used to ensure direct inflow of water with any solid material towards the box. By contrast, if the fissure was wider than the plastic conduit, artificial dams made of plastic and strengthened by mounting foam (and glued to the rock surface, if needed) were constructed to prevent water flow in a different direction than to the sand collector (Figure 3 and Figure 4). These elements were fitted and glued together to ensure tightness of the whole construction. The upper parts of the inlets of the conduits were additionally protected by plastic nets. In the beginning of the experiment, each sand collector was equipped with a time-lapse camera. Using metal and wood brackets and rods, the position of the sand collectors was stabilized. During the two-year monitoring period, each sand collector remained in place and was not subjected to any displacements that could influence its operation.

3.2.3. Data Collection

The purpose of the field experiment was to monitor the volume of sand removed from the caprock by subsurface water flow and deposited in the sand collectors. Monitoring (the term subsequently used to describe the day when the volume of sand removed was measured) was carried out in approximately monthly intervals during the two-year period. The installation of all the sand collectors was completed on the 30 June 2022, and the first monitoring was conducted on the 2 August 2022. The last monitoring took place on the 7 August 2024.

During each monitoring, water collected in the sand collectors was removed manually using plastic bottles and bowls. The remaining small amounts of water, which could not be removed by these tools, were drained with a paper towel. Organic particles—including soil, leaves and small pieces of wood—were removed manually, too. The volume of the remaining grains of sand was then measured. To do so, sand was collected with a silicone spatula and placed in a container with a volumetric scale. If the amount of sand was greater than the container’s volume, the procedure was repeated until the whole sand collector was emptied. Several times in the winter season, water in a sand collector was partly frozen. In such a situation, a block of ice was removed. No grains of sand were found within the ice. After each measurement, the interior of the sand collector was cleaned, and its top was covered with a plastic lid again until the next monitoring.

Thus, the database available at the end of the experiment contained 144 records (24 observation periods × 6 sand collectors). These may be called “denudation variables”. The following were used in the subsequent analysis (

Table 2. Volumes of sand collected at each study site during the field experiments (all data expressed in ml except the number of reactions). Bold highlights the highest yield at a site. Data for the observation period 1, in italics, are provided but are considered neither in the calculation of the total yield and the number of reactions, nor in subsequent statistical analysis because of possible disturbance during the installation of the sand collectors.

Observation Period	SC-BS-1	SC-BS-2	SC-BS-3	SC-SzW-1	SC-SzW-2	SC-SzW-3	BS in Total	SzW in Total	All Sites
1 (30 June 2022–2 August 2022)	8	45	10	170	10	2	63	182	245
2 (3 August 2022–3 September 2022)	4	2	2	1980	70	6800	8	8850	8858
3 (4 September 2022–7 October 2022)	2	2	2	13	2	15	6	30	36
4 (8 October 2022–3 November 2022)	0	0	0	0	0	0	0	0	0
5 (4 November 2022–29 November 2022)	0	0	0	0	0	0	0	0	0
6 (30 November 2022–30 December 2022)	0	0	0	0	0	0	0	0	0
7 (31 December 2022–2 February 2023)	0	0	0	0	0	0	0	0	0
8 (3 February 2023–6 March 2023)	0	0	0	0	0	0	0	0	0
9 (7 March 2023–5 April 2023)	0	0	0	0	0	0	0	0	0
10 (6 April 2023–6 May 2023)	2	4	0	0	0	2	6	2	8
11 (7 May 2023–27 May 2023)	0	0	0	0	0	0	0	0	0
12 (28 May 2023–6 July 2023)	0	0	0	0	0	3000	0	3000	3000
13 (7 July 2023–7 August 2023)	0	0	0	4	800	950	0	1754	1754
14 (8 August 2023–2 September 2023)	20	10	45	10	0	1025	75	1035	1110
15 (3 September 2023–12 October 2023)	0	0	0	0	0	0	0	0	0
16 (13 October 2023–21 November 2023)	0	0	20	0	0	0	20	0	20
17 (22 November 2023–5 January 2024)	30	5	32	500	2	4800	67	5302	5369
18 (6 January 2024–28 January 2024)	0	0	0	0	0	0	0	0	0
19 (29 January 2024–29 February 2024)	0	0	40	0	0	560	40	560	600
20 (1 March 2024–5 April 2024)	0	0	0	10	0	0	0	10	10
21 (6 April 2024–9 May 2024)	0	0	10	0	0	0	10	0	10
22 (10 May 2024–29 May 2024)	0	0	40	0	0	35	40	35	75
23 (30 May 2024–4 July 2024)	0	0	100	0	0	0	100	0	100
24 (5 July 2024–7 August 2024)	0	20	230	5	5	330	250	340	590
Total yield	58	43	521	2522	879	17,519	622	20,918	21,540
Number of reactions	5	6	10	7	5	10	21	22

): volumes of sand deposited in each collector, the total volume of sand collected at each locality (e.g., for three sites at Białe Skały and Szczeliniec Wielki, respectively, considered together) and the overall total deposition at all sites. Each case of sand delivery to the collector recorded during monitoring was termed a “reaction”, and the number of “reactions” during each monitoring was also noted. It was decided not to include observational data from the first monitoring in the statistical analysis to avoid potential effects of ground disturbance during the installation of the sand collectors. However, these data are reported in Table 2.

During the first two observation periods (the term used later in the paper to describe the time span from the day following one monitoring until the day of the next monitoring, approximately a month), the sand collectors were equipped with standard time-lapse cameras. The idea was to photograph the exact moments of sand delivery to the collectors. For this reason, the cameras were programmed to take one photograph every two hours. Unfortunately, although the major parts of the cameras were covered by string bags to protect them from water, some devices stopped working after only a few days and the others after c. two months. The likely reason was high humidity inside the collectors covered by plastic lids, which can be destructive for electronic devices. Nevertheless, one of the cameras at the Szczeliniec Wielki locality worked properly during the torrential rainfall episode in August 2022 and precisely recorded the moment of sand expulsion.

3.3. Meteorological Data

Information on precipitation patterns in the study area was acquired from the publicly available record for the meteorological station situated in the village of Karłów (roughly in between the study sites, c. 0.8 km south of the Szczeliniec Wielki site and c. 2 km northwest from the Białe Skały site; see the locality in Figure 1). The meteorological station, which serves as a part of the Kłodzko County Flood Protection System [62], is automated and records precipitation every hour. The raw precipitation data covering the entire field-experiment period (from the 30 June 2022 to 2 August 2024) were aggregated and used in statistical analysis.

Precipitation variables retrieved from the raw dataset included the following, all related to each observation period: frequency of days with precipitation [%]; max daily precipitation [mm]; max total precipitation within an episode [mm]; length of episode with the highest precipitation [days]; longest precipitation episode [days]; length of wave with the highest precipitation [days]; longest precipitation wave [days]; mean daily precipitation [mm]; and max hourly precipitation intensity [mm/h] (Table S1). For the purpose of this paper, the term “episode” is used for any number of consecutive days with precipitation, regardless of the actual precipitation within those days. In turn, “wave” applies to longer periods of nearly continuous precipitation, with singular days without precipitation allowed. The subsequent statistical analysis revealed a very high correlation between maximum total precipitation during “episodes” and “waves” (close to 0.95), and hence, the latter variable was excluded from further analysis.

3.4. Statistical Analysis

For each denudation and precipitation variable under consideration, a Ljung–Box test was performed to assess the independence of observations, allowing for valid statistical inferences in terms of correlation analysis. As none of the variables was autocorrelated, classical correlation coefficients were applied. Different correlation coefficients were used (Pearson’s, Spearman’s and Kendall’s) according to the characteristics of variables, such as the normality of their distributions (assessed with a Shapiro–Wilk test), the presence of outliers (identified on the basis of interquartile range), linear vs. non-linear relationship and the number of tied-ranks. R package (version 4.4.2) was used for all calculations.

4. Results

4.1. Precipitation Patterns

During the whole monitoring period, precipitation patterns generally followed the typical annual variability for the region [57], with higher total precipitation recorded in summer months (observation periods 2 and 3 (August and September 2022), 13 and 14 (July and August 2023), as well as 23 and 24 (June and July 2024); Figure 5). However, while in the autumn 2022 and winter 2022/2023 (observation periods 4–8), the overall precipitation was significantly lower than in the summer months of that year, in 2023, the situation was different. The total precipitation values recorded in observation periods 16 and 17 (October–December 2023) were the second (193 mm) and the third (174 mm)-highest for the whole monitoring period (the highest total precipitation, 218 mm, was recorded for observation period 13). Individual observation periods may be divided into more and less humid, considering the percentage of days with precipitation. There were 12 periods with more than 50% days with precipitation, with the maximum being 72.5% (period 16). The interval from observation period 13 to 19 was the wettest, with only one period having a percentage of days with precipitation less than 50.

A large portion of the total precipitation was delivered during several consecutive days (“episode”, as defined in Section 3.2). In ten observation periods, the maximum total precipitation within an episode exceeded 50 mm, and in five, it exceeded 70 mm, whereas in two cases, it even exceeded 90 mm (Figure 5). The episodes were of a varying length (from 1 to 9 days). Interestingly, total precipitation in excess of 90 mm occurred during relatively short episodes, in four and three days for observation periods 2 and 13, respectively. Maximum daily precipitation ranged from 7.6 mm to 65 mm (observation period 13). The daily value of 30 mm was exceeded in four other cases, during observation periods 2 (51.5 mm), 14 (30.5 mm), 16 (49 mm) and 17 (38.4 mm).

The maximum hourly precipitation intensity ranged from 2 mm/h (observation period 19) to 25 mm/h. In general, the highest values of precipitation intensity (>20 mm/h) were observed in summer months, in association with torrential rainfall during storm events. Intensities between 13 and 20 mm/h were recorded in both summer (four cases) and autumn (two cases). Interestingly, in observation period 17 (December 2023), characterized by the second greatest total precipitation, the third-highest maximum total precipitation within the episode and the third-greatest contribution of days with precipitation within the whole observation period, the maximum rainfall intensity recorded was merely 5.4 mm/h.

4.2. Spatial and Temporal Variability of Contemporary Sand Removal

Among 138 individual records, no evidence of sand evacuation (no reaction) was noted in 83 cases (60%), including nine observation periods with no deposition in any of the six collectors (Table 2). The longest time interval of continuous inactivity spanned observation periods 4 to 9 (October 2022–March 2023). During another six observation periods, only one sand collector yielded a measurable volume of sand. On the other hand, in only three observation periods, each sand collector trapped material, with the deposition in five out of six collectors in another two periods. The general reactivity of the underground sand removal systems may therefore be considered low.

Considering both localities (Białe Skały and Szczeliniec Wielki) together, the following observations emerge. The overall activity of denudation systems is similar (21 reactions at Białe Skały and 22 reactions at Szczeliniec Wielki). However, at each locality, one sand collector showed much higher yields than the other two. At Białe Skały, 10 reactions were recorded in SC-BS-3, whereas at Szczeliniec Wielki, SC-SzW-3 showed reactions 10 times. Moreover, at each locality and considering the entire duration of the field experiment, each of these collectors yielded much more sand than the other two (Białe Skały: 521 mL in total versus and 58 mL and 43 mL; Szczeliniec Wielki: 17,517 mL in total versus 2522 mL and 879 mL). Thus, the efficacy of one system far exceeded the efficacy of the remaining two. Furthermore, in each case and regardless of the total volume, there was one particular observation period during which considerable sand removal occurred. The relative yield during such a period ranged from 39% (SC-SzW-3) to 91% (SC-SzW-2) of the total. If the next-most-efficient reaction is added, the respective percentage exceeds 70% in all cases but SC-SzW-3, where the two most efficient reactions account for 66%. Thus, the nature of the systems’ reaction is highly episodic, with the key role of singular events of sand delivery. These key events did not match in time, although at Szczeliniec Wielki, two sand collectors yielded maximum output in observation period 2, whereas at Białe Skały, a similar situation was observed in observation period 24. Also of note are observation periods 12 and 15, when significant volumes of sand were collected at SC-SzW-3, with no sand deposition at all in any of the remaining collectors. Thus, the local denudation system at SC-SzW-3 clearly emerges as the most efficient among all investigated.

The database also shows considerable differences between the Białe Skały and Szczeliniec Wielki localities. Considering the total volume collected, the sand collectors at Szczeliniec Wielki yielded 35 times more material (20,918 mL versus 622 mL). There were several cases of significant reactions at Szczeliniec Wielki, with no or very limited reaction at Białe Skały (e.g., observation periods 2 and 13). For other observation periods, the reactions were synchronous (e.g., periods 17 and 24), even though the total yields at Białe Skały were much smaller. In only four observation periods, the sediment yield at Białe Skały was higher than at Szczeliniec Wielki (periods 16, 21, 22, 23), but the total volumes for the three collectors at each locality were small (including no yield at some collectors), except period 23. The database also suggests an increasing activity of the SC-BS-3 site towards the end of the field experiment.

5. Discussion

5.1. Sand Removal Versus Rainfall Patterns

Correlation analysis revealed three types of relationships between the efficacy of sand evacuation through the underground fissure systems (denudation) and precipitation characteristics (

Table 3. Correlation coefficients for pairs of variables under consideration. Statistically significant correlation coefficients are given in bold. Asterisks indicate level of statistical significance: * p ≤ 0.05, ** p ≤ 0.01.

	var_11	var_12	var_13	var_14	var_15	var_16	var_17	var_18	var_19
var_1	0.247	0.400 *	0.474 **	0.417 *	0.379 *	0.146	0.233	0.428 *	0.228
var_2	0.187	0.410 *	0.490 **	0.385 *	0.319	0.092	0.129	0.449 **	0.334 *
var_3	0.263	0.346 *	0.406 *	0.407 *	0.358 *	0.180	0.161	0.459 **	0.307
var_4	0.230	0.374 *	0.438 **	0.204	0.252	0.040	0.110	0.387 *	0.346 *
var_5	0.222	0.501 **	0.550 **	0.061	0.203	0.133	0.153	0.491 **	0.368 *
var_6	0.181	0.468 **	0.547 **	0.259	0.310	0.343 *	0.380 *	0.514 **	0.363 *
var_7	0.273	0.399 *	0.467 **	0.458 **	0.406 *	0.175	0.195	0.509 **	0.306
var_8	0.147	0.406 *	0.492 **	0.215	0.261	0.304	0.349 *	0.460 **	0.332 *
var_9	0.227	0.481 **	0.607 **	0.365 *	0.412 *	0.359 *	0.388 *	0.536 **	0.408 **
var_10	0.385	0.494 **	0.788 **	0.540 **	0.451 **	0.300	0.433 *	0.705 **	0.560 **

Rows: Variables related to sand deposition (var_1 to var_9—deposition in ml): var_1—SC-BS-1; var_2—SC-BS-2; var_3—SC-BS-3; var_4—SC-SzW-1; var_5—SC-SzW-2; var_6—SC-SzW-3; var_7—Białe Skały (BS); var_8—Szczeliniec Wielki (SzW); var_9—all sites; var_10—number of collectors with sand deposition. Columns: Variables related to precipitation patterns: var_11—frequency of days with precipitation [%]; var_12—max daily precipitation [mm]; var_13—max total precipitation within an episode [mm]; var_14—length of episode with the highest precipitation [days]; var_15—longest precipitation episode [days]; var_16—length of wave with the highest precipitation [days]; var_17—longest precipitation wave [days]; var_18—mean daily precipitation [mm]; var_19—max precipitation intensity [mm/h].

). These include (a) statistically significant correlations with precipitation characteristics for all denudation variables; (b) no such correlations for all variables; (c) statistically significant correlations for the Białe Skały (BS) locality, with no correlations for the Szczeliniec Wielki (SzW) locality, or vice versa.

The strongest correlations occurred between denudation variables and maximum precipitation within an episode (var_13), followed by correlations with the mean daily precipitation (var_18). Statistically significant correlations also occurred with the maximum daily precipitation within an observation period (var_12), although the levels of significance were mostly p ≤ 0.05. In contrast, no statistically significant correlations were found between any denudation variables and the frequency of days with precipitation within a period (var_11), whereas only one site at Szczeliniec Wielki (SC-SzW-3) showed a significant correlation with the duration of precipitation waves (var_16, var_17). As a consequence, for var_17, correlations for all sites at SzW and all sites in total also proved statistically significant. Nevertheless, precipitation itself, even spread over a prolonged period (“wet periods”), appears to be an insufficient factor to initiate widespread subterranean transport of sand. Rather, the underground denudation systems require a high amount of precipitation within just a few days to be activated.

Several precipitation variables show partial correlations with the denudation variables (Table 3). Two of them, the length of an episode with maximum precipitation (var_14) and the maximum length of an episode (var_15), correlate with the volumes of sand at BS, but not at SzW. In contrast, rainfall intensity (var_19) correlates much better with the volume of sand at SzW, whereas at BS, only one site showed a statistically significant correlation. Nonetheless, if all the individual study sites are considered together, the correlations are statistically significant.

A closer examination of individual cases reveals further interdependencies between variables, but also certain unexpected occurrences. It is also becoming evident that each site shows a great deal of individuality. The highest sediment yields at each site were associated with episodes with the total precipitation in excess of 70 mm. However, if these episodes are ranked according to the total rainfall, apparent inconsistencies emerge. The highest daily rainfall recorded during the experiment (65 mm) and the most-rainy episode (110.5 mm in three days) did not generate any sand removal effect at BS and inconsistent reactions at SzW (the highest volume of sand removed in the SC-SzW-2 system, a moderate one in SC-SzW-3, and a marginal one in SC-SzW-1). Likewise, the second-most-rainy episode (96.2 mm in 4 days) produced a negligible effect at BS, whereas at SC-SzW-1 and SC-SzW-3, maximum yields for the entire experiment were recorded, 1980 mL and 6800 mL, respectively (Figure 6A,B). Reactions at BS sites were much stronger during the next two episodes, with total precipitation of 85.3 mm and 82.1 mm, but the fundamental difference with the others was the length of these episodes, eight days in each case. This is consistent with the results of the correlation analysis as reported above. The fifth-most-efficient episode (72.6 mm) does not fit the pattern, however, since it was modest in duration (5 days) and yet produced the highest yields, at two, among the BS sites, whereas the reactions at SzW were minor, despite it being the site of the second-highest rainfall intensity recorded during the experiment (23.6 mm/h).

The data also show that precipitation around 30 mm within an episode is the threshold to generate a clear reaction of the underground denudation systems, although exceptions occur. For instance, quite high precipitation in observation period no. 16 (65 mm, the sixth highest) produced a reaction at one site only. Finally, most maximum yields at a site are clearly associated with precipitation events of the highest intensity.

Photographs taken by the time-lapse camera inside the SC-SzW-1 sand collector soon after the beginning of the experiment confirm the statement above. The precipitation episode that occurred on 13–15 August 2022 was of too low an intensity to enable sand delivery, and only water emerging from the fissure system was collected inside the plastic box (Figure 6C,D). Until 6 p.m. of 20 August 2022, nothing changed (Figure 6E). Then, when rainfall of 25 mm/h intensity occurred, the sand collector was filled with foamy water (likely resulting from rapid, turbulent flow inside the fissure system), with the right side of the box having traces of sand delivery (Figure 6F). The photograph taken in the daylight a day later (Figure 6G) confirms that the sand collector was filled with material at that moment (in Figure 6B, sand deposited in the collector SC-SzW-1 is visible after the removal of water during the monitoring on 3 September 2022). The fissure supplying the collector SC-SzW-3 reacted to the event in the same way but delivered greater than three times more material (Figure 6A).

These observations can be summarized as follows:

• To become activated, underground denudation systems require precipitation episodes with at least 30 mm in a few consecutive days, although significant reactions are recorded if the respective precipitation is 70 mm.
• High intensity of precipitation enhances the efficacy of sand removal systems, a statement particularly valid for Szczeliniec Wielki (Figure 6). In contrast, sustained delivery of atmospheric water, rather than short-term intensity, appears more important for Białe Skały.
• The frequency of days with precipitation and their distribution within an observation period are irrelevant for underground denudation.

Notwithstanding these generalizations, the database includes many exceptions, and these are of two kinds. First, high yields were noted at individual sites despite low precipitation, and second, weak reactions followed precipitation well above the average. The possible reasons for these are discussed in the next Section 5.2.

In conclusion, observational data supported by statistical treatment permit us to consider the precipitation factor as important, but not dominant, and fails to explain each case of considerable sand removal from the underground systems. This, in turn, suggests the role of geological and local geomorphological factors, explored in the next section.

5.2. The Role of Local Geological and Geomorphological Factors

The most evident feature of the dataset is the huge discrepancy in sediment outflows recorded at both localities (Białe Skały and Szczeliniec Wielki). Although very local differences in extreme precipitation amounts cannot be excluded, it is unlikely that this hypothetical factor alone could explain the much higher denudation at Szczeliniec Wielki. In contrast, these differences are consistent with dissimilar structural characteristics at these localities (Table 1), which in turn influence relief and groundwater circulation patterns (Figure 7). Szczeliniec Wielki is a tabular mesa, with a negligible dip of caprock strata and a well-developed system of underground drainage, as confirmed by a large proportion of internally drained areas within the mesa top, reaching 90% of its area [24]. Open clefts and closed depressions are numerous, providing pathways for rain- and meltwater to enter subterranean drainage systems. At site SC-SzW-3, the subsurface system is so well developed that water was continually emerging from the fissure some 16 h after the end of the large precipitation episode in the beginning of January 2024 (Video S1). Although the exact size of these systems feeding the fissures with sand collectors cannot be determined, it is reasonable to assume that it is not limited to these fissures themselves and extends into the massif.

Białe Skały, in turn, represents a cuesta face, with the strata dipping into the slope. The topography is adjusted to the dip and cliff faces give way to surfaces inclined in the opposite direction than the cuesta front faces [27]. Therefore, any underground flow along bedding planes would be away from the cliff face (and sand collectors), unlike at Szczeliniec Wielki, and the resultant feeding systems are not very likely to be large (Figure 7). In addition, there is little evidence of the development of rock cities and the existence of deep clefts immediately behind the study sites (sand collectors). Thus, catchment areas at both localities appear fundamentally different in terms of size and transmissivity, which is primarily related to structural control.

In addition to differences in total sand yield, denudation systems at Białe Skały show different reactions to precipitation patterns than those at Szczeliniec Wielki, with limited immediate effect of considerable rainfall at the former (Figure 7). At Szczeliniec Wielki, high rainfall apparently quickly reaches the fissure system, and the resultant subterranean flows are capable of entraining sand grains earlier liberated by weathering (probably mostly by arenization—see [27]). Mechanical erosion of weaker parts along fractures is also possible. These sand-laden flows quickly emerge at the exits from fissures, building up sandy cones [34] and leaving sand in our sand collectors. In contrast, at Białe Skały, reaction to torrential rainfall is small to non-existent, which suggests an important role of buffers and temporary storage of water between the cuesta top surface and the exits. This storage is likely of two kinds. First, the cliff tops and plateau surface at Białe Skały are covered with soils containing organic horizons several tens of cm thick (up to 1 m in extreme cases), able to absorb significant amount of rainwater [27]. At Szczeliniec Wielki, in particular in the marginal parts of the mesa, soil cover is less developed, and the proportion of bare rock is considerable. Second, the porosity of sandstone at Białe Skały is up to 20% [27], which makes it capable of storing substantial amounts of water. During the field experiment, we saw only one situation (on the 4 January 2024, after the very rainy observation periods 16 and 17) in which the rock was saturated to such a degree that a new drainage pathway was created (Video S2). It was formed as a pipe within one of recesses next to the cliff lines, with a discharge of 0.75 l/s and fresh sand sheet deposited beneath. Simultaneously, all fissures at Białe Skały showed significant reactions, supplying the collectors with sand. Otherwise, intergranular flow occurs, and little free water is available in the system to carry the sand material outside the caprock. The opposite situation is at Szczeliniec Wielki, where sandstone is less porous and drained via discontinuities, supporting the efficient removal of material.

Different response times of sand removal following strong rainfall may be also related to the pace of sand liberation from the rock. Although porous sandstone can store substantial amounts of water, prolonged conditions of this kind can have disruptive effects on the rock itself, associated with wetting-weakening [63] and the resultant decrease of the tensile strength [36]. The process requires time, and this may explain why the denudation system at Białe Skały shows poor correlation with intensive rainfall of short duration. However, observational data also reveal significant sediment outflows at Szczeliniec Wielki, apparently not connected with extreme precipitation. This suggests that in this locality, large yields are products of quick circulation in the marginal parts of the mesa, where the buffering effect is small, whereas moderate and low yields result from longer circulation, involving various delays inherent to the system.

There also seems to be a sort of inverse relationship between the efficacy of a denudation system and the width of the feeding fissure zone. This is evident at Szczeliniec Wielki, where the highest volumes of sand were recorded at SC-SzW-3, which is a fissure only 5–10 cm wide. The widest fissure (SC-SzW-2), in turn, produced the least material, 20 times less than its neighbor. In specific situations, such as the one following a larger rainfall event in early January 2024, the flow of water emerging from a fissure was only noted towards the SC-SzW-3 sand (Video S1), whereas the other two fissure systems showed no activity at that time. The relationships described above are less clear at Białe Skały, with the most efficient system (SC-BS-3) being drained by a fissure of moderate width. Yet, in this case, we fall short of providing a plausible explanation.

In addition, the position of sand collectors with respect to the fissure system may explain differences between individual sites. At Szczeliniec Wielki, fissures supplying collectors SC-SzW-1 and SC-SzW-2 generally share the same catchment area (Figure 2). Higher yields recorded at SC-SzW-1 suggest a dominant subsurface flow in the SE direction, with only local supply to the SC-SzW-2 collector, which showed negligible reaction (e.g., observation periods 2 and 17; see Table 2). The opposite situation was recorded during observation period 13, which could have resulted from a temporary blockage of the fissure feeding the SC-SzW-1 collector and flow diversion towards SC-SzW-2.

The final factor to consider is material availability. Its role is suggested by both relatively high yields noted at individual sites, despite low precipitation, and only weak reactions that followed precipitation well above the average. Thus, considerable yields at Szczeliniec Wielki during observation period 2 (Figure 6) could also mean the removal of much of the sand stored in the fissures until that time, and hence, the next precipitation episode did not produce any substantial deposition. Then, after nearly a year, a considerable volume of material was mobilized from SC-SzW-3 (observation period 12), even though precipitation was modest, whereas the yields in the next two periods were smaller, despite higher precipitation. Similarly, at SC-SzW-1, the highest yield was recorded at the beginning of the experiment (Figure 6B), and none of the subsequent yields was comparable, despite several episodes of significant precipitation.

5.3. Paleoenvironmental Implications and Significance for the Geomorphic Evolution of Sandstone Tablelands

For decades, quartz sandstones and quartzites have been perceived as some of the most insoluble rocks on Earth; hence, mechanical weathering and erosion were traditionally considered to be the sole contributors to the origin of sandstone landforms (see [25,64,65,66] for review). Later, in the 1970s and 1980s, various lines of evidence, particularly from the fields of petrography and geochemistry, were used to argue that quartz sandstones and quartzites are subject to corrosion, and many individual landforms such as deep shafts, caves, block fields, towers and spires, as well as their assemblages at a larger scale—rock cities and ruiniform relief—are in fact products of solutional weathering, sharing many morphological and functional characteristics with the “classic” limestone karst [42,67,68,69]. At the same time, interest in mechanical processes declined. Only in the last decade, Bruthans et al. [21,22] emphasized the role of physical erosion again, but their views were based on studies of friable sandstones and non-cemented interlocked sands, particularly prone to mechanical breakdown into individual grains. For solid, lithified sandstone, the issue remained puzzling, especially regarding the efficacy of slow, mechanical removal of material from the caprock. As already pointed out by Jennings [42], this issue is crucial, since without the physical removal of detached grains of sand, the underground conduits would become blocked. Thus, “(…) a plentiful supply of flowing water is necessary (…)” [65] (p. 84). Yet, according to our knowledge, until now, no quantification of the contribution of mechanical denudation in the evolution quartz sandstone/quartzite-capped tablelands has been conducted, except some rather old studies on rockfalls and other mass movements [37,70].

From the short-time, present-day perspective, sandstone-capped mesas and escarpments appear as landforms of apparent stability, subject to almost no geomorphic change, with the exception of occasional rockfalls (almost unknown in the Stołowe Mountains [71]). This is the likely reason why a number of authors have argued that more efficient processes leading to escarpment retreat were associated with environmental regimes different than the contemporary one, either more humid [72,73], or cold, periglacial [48]. Our systematic monitoring, though limited to two years, quantitatively confirmed that sand removal from caprock does operate in present-day conditions.

The results of our study stay in line with the general concept that the evolution of tablelands capped by quartz sandstone proceeds along vertical fractures cutting caprock, in consequence producing cave passages in the subsurface [74] or corridors of ruiniform relief at the surface [24]. Among the key ingredients is the prolonged contact with water within the pore spaces and tight cracks, which supports efficient dissolution of cement (the essence of the arenization concept) [75], thus freeing individual grains of sand. These, in turn, are subject to removal in the zones where the water flow is the most efficient, i.e., along the fissures.

The collected data might suggest that only small volumes of material are removed outside the caprock under contemporary environmental conditions. Yet, if these values are extrapolated into longer time intervals, these volumes of material are no longer insignificant. For example, the sediment yield from the fissure feeding the SC-SzW-3 sand collector was almost 18 l per two years, which gives the average annual value of c. 9 l. In the time span of 1000 years, this would account for 9 m³, while in 10,000 years, 90 m³ of sand would be removed from the caprock through just one fissure system. This value might still appear low, but the marginal parts of sandstone-capped tablelands do not decay as a whole, but transform into ruiniform relief and boulder chaos through the preferential erosion of fracture zones [19,20]. Thus, to make the process effective, only a fraction of caprock-forming sandstone must be removed, enabling the separation of more massive compartments. The outlet of the fissure supplying the SC-SzW-3 sand collector is connected with the apex of a large sandy cone, c. 6 m high. Its volume, calculated according to the methodology developed by [20], is c. 116 m³. The comparison of this figure with the sediment yield of 9 m³/1000 years reveals that the sedimentation would have lasted almost 13,000 years. This is surprisingly close to the results of OSL dating of sand inside the cones in the Stołowe Mountains, which showed that the onset of cone build-up could have even been at the beginning of the Holocene [34]. On the other hand, a significantly smaller sediment yield at the SC-SzW-1 site observed during the experiment period (c. 1.35 l/year), if extrapolated in a similar way (1.35 m³/1000 years, thus, 13.5 m³/10,000 years), would not match the volume of the cone (53.5 m³) situated at the outlet of the feeding fissure (this particular cone was subject to OSL dating by Duszyński et al. [34], showing the age of c. 7.3 ka for the basal unit). Assuming constant sand delivery rate throughout the Holocene, however, does not seem realistic, and it is possible that in the past, the sedimentation rate at this site may have been higher. The cliff-foot sandy cones at Białe Skały are significantly smaller than the ones at Szczeliniec Wielki [20], which remains in agreement with the much-smaller sediment yield at this locality demonstrated in this study.

Finally, our study confirms that the availability of water percolating through the caprock is essential for the geomorphic evolution of sandstone tablelands. Recently, Duszyński et al. [32] showed that much higher rates of escarpment retreat of the mesa of Szczeliniec Wielki characterized the Pleistocene/Holocene transitional period, with the unblocking of subsurface drainage due to the degradation of permafrost as one of the possible reasons. Similarly, the pace at which sandy cones were up-building throughout the Holocene varied, with higher sedimentation rates during warmer and more humid periods [34]. The crucial role of water in efficient sandstone decay was also demonstrated in the experimental work by Bruthans et al. [36], according to whom its presence in pore spaces leads to serious reduction of tensile strength and greater susceptibility to erosion due to freeze–thaw activity.

6. Conclusions

In this paper, we focused on the long-neglected problem of the significance of mechanical subsurface erosion in the evolution of sandstone tablelands from the present-day perspective. According to our knowledge, this is the first research which directly demonstrated that sandstone-capped mesas and plateaus under humid temperate climate lose mass through the removal of sand via underground conduits. While in the last decades, much effort has been directed at recognizing the role of the weathering phenomena, the dissolution of cement and quartz grains boundaries in particular, the second crucial component of the arenization concept—the mechanical removal of detached grains of sand—seems to have remained overlooked. Although singular events of sand removal were occasionally reported [20,24], the incidental nature of the process made it very hard to observe. In this paper, we offered the methodology to overcome this problem, involving systematic, monthly monitoring of selected sites. This approach enabled us to better understand the meteorological and geological/geomorphological controls behind mechanical denudation.

At the investigated sites, underground sand removal from sandstone caprock occurs up to several times a year, and the major control of its efficacy is precipitation. Yet, neither the frequency of days with precipitation nor their distribution within an observation period play any role in initiating sand evacuation. Contrarily, large precipitation in a few consecutive days is of special significance, with a minimum of 30 mm to allow for flows capable of entraining and removing sand grains, whereas a threshold of approximately 70 mm results in significant reactions of the underground denudation system. In addition, we showed different reaction patterns at two localities subject to monitoring. At the tabular mesa of Szczeliniec Wielki, significant amounts of removed sand are associated with very high precipitation intensity. At the cuesta front of Białe Skały, underground systems responded mainly to the sustained delivery of atmospheric water. However, the total yield of sand at Szczeliniec Wielki was 35 times higher than at Białe Skały, which is associated with local geological and geomorphological controls. In particular, the size of underground drainage systems is much smaller next to the cuesta front, whereas the higher porosity of sandstone at Białe Skały additionally supports the storage of substantial amounts of water before any reaction occurs. Moreover, soils with thick organic horizons, widely present at Białe Skały (but less so at Szczeliniec Wielki), absorb significant amount of rainwater, causing the delayed reaction of the underground denudation system.

Our study stays in line with recent findings [27,34], according to which water percolating through the sandstone caprock is essential for its ongoing disintegration. While the sediment yields observed during this two-year experiment might seem insignificant, their values extrapolated into longer time intervals appear sufficient to significantly contribute to the transformation of once-continuous cliff lines into ruiniform relief and boulder chaos through preferential erosion along fracture zones. Possibly, in more humid periods, when the water availability in underground system is greater, the process operates at an even higher pace. It remains an open question how representative are our results, given the wide range of climatic and lithostructural settings. Analogous monitoring studies carried out elsewhere may shed further light on the relationships and causal linkages suggested here. In addition, one sand collector has been left in the field and will continue to be monitored.