Elevated Concentrations of Carbon Dioxide (CO₂) on the Harbechy Plateau (Moravian Karst) Reveal a Gas-Rich Soil Layer (GRSL)

Abstract

Precipitation leaches soil organic matter (SOM), transporting it downward where it accumulates at the soil–bedrock interface. Intensive agriculture, particularly tillage, accelerates this process. Microbial decomposition of SOM generates CO₂, forming a gas-rich soil layer (GRSL)—a phenomenon long hypothesized but never directly confirmed until now. Drilling on the Harbechy Plateau (Moravian Karst) revealed a GRSL with a thickness of ~0.8 m, CO₂ concentrations averaging 1.5–3 vol. % (peaks of 4–6 vol. %), and isotopic signatures (δ¹³C) indicating a mix of biogenic (−25‰) and atmospheric (−8‰) CO₂. These findings necessitate re-evaluation of carbon cycling models in karst agroecosystems.

1. Introduction

CO₂ links soil respiration and photosynthesis as part of the global carbon cycle [1]. Beyond its biological role, CO₂ influences weathering processes [2] and regulates natural water pH [3]. In soils, CO₂ originates from root respiration [4] and SOM microbial decomposition [5,6]. The main elements of SOM are C, H, O; minor elements are N, P, S [7]. For example, the chemistry of decomposition of polysaccharide (cellulose) is described by the summary equation:

$${\left(C_6{H}_{10}{O}_5\right)}_n+6n O_2\to 6n{ CO}_2+5n H_{2}O$$

(1)

The CO₂ pathway of this process (Equation (1)) is cellulose → aldehydes/ketones → carboxylic acids → CO₂ [8,9]. For lignin, the pathway is: lignin → phenols → aldehydes → humic substances → carboxylic acids → CO₂ [10]. CO₂ production in soil exhibits an exponential dependence on temperature and a linear relationship with soil moisture [11]. CO₂ drives carbonate dissolution, governing speleothem formation [12] and terrestrial-to-ocean calcium transfer. Forward/reverse reactions controlling speleothem growth/dissolution are

$${CaCO}_3+{CO}_2+H_{2}O \rightleftharpoons {Ca}^{2+}+2{ HCO}_3^{-}$$

(2)

Subprocesses covered in Equation (2) are discussed in more detail in [13] and references therein.

The epikarst hydraulic barrier supports the formation of perched aquifers [14] and may trap SOM, facilitating GRSL formation. Recent soil drilling down to the soil–bedrock interface (to a depth of 2 to 3 m) proved the GRSL with a CO₂ concentration up to 6 vol. %. We hypothesize that the moderate temperatures (8–9 °C, at >1.5 m depth) derived from the mean annual temperature in the studied region facilitate SOM degrading to CO₂. CO₂ accumulates at the soil–bedrock interface, forming the GRSL. Supporting evidence includes (1) elevated CO₂ levels in Harbechy Cave lying beneath the studied soils [15], (2) CO₂ concentration gradient in soil profile, (3) discrepancies between drip-water-modeled CO₂ concentrations [16,17,18] and topsoil-measured CO₂ [19].

The goal of this work is (1) to characterize the GRSL through field and laboratory measurements, (2) to find conditions necessary for the formation of the GRSL, (3) to detail a conceptual model, and (4) to summarize the consequences of GRSL creation. To meet the objectives of the work, we chose a multi-instrumental approach using methods such as electric resistivity tomography (ERT), soil drilling (SD), X-ray diffractometry (XRD), scanning electron microscopy with energy dispersive X-ray analysis (SEM/EDXA), field IR spectrometry, gas chromatography, analysis of stable isotopes, and hydrogeochemical modeling.

2. Site of Study

The Moravian Karst (MK) is by far the most extensive karst area in the Czech Republic [20]. It is built mainly from middle Devonian to Lower Mississippian limestones. These limestones were deposited in two formations. The Eifelian to Frasnian Macocha Fm. is made up of more than 1000 m of limestone and dolomite deposits of the cyclic reef of Vilémovice Lst. and lagoonal Lažánky Lst. These older shallow-water deposits are overlain by Famenian to Viséan Líšeň Fm. They comprise calciturbidites of the Hády-Říčka Lst. and more shallow-water Křtiny Lst. Granitoid rocks of the Brno massif (proterozoic age) together with mostly Givetian basal clastics form the basement of this limestone sequence. This paraautochthonous unit is thrust over Variscan flysch units (so-called Culmian facies). Younger strata, which form a patchwork of small erosional relicts, complete the sedimentary evolution of the Moravian Karst. These include Jurassic sandstones and limestones, Cretaceous sediments of the Rudice beds, and several cycles of Miocene deposits. The youngest sediments that cover much of the older deposits are Pleistocene (Middle and Upper) loess and loess loam deposits. These sediments usually fill older valleys, karrens, and other karst phenomena. The area focused on this study, i.e., the Harbechy-Vilémovice plateau, is built mainly of the Vilémovice Lst. of the Macocha Fm. Only a small portion of the plateau in the west is built from the underlying Lažánky Lst. and in the east from overlying Křtiny Lst. Limestones are gently dipping toward the east and are folded into NNE-SSW trending, almost upright folds with wavelengths of hundreds of meters. The fault network is dominated by two systems of faults (NE-SW and NW-SE), which were crucial for underground water migration, karstification, and cave generation. These two systems are apparent from the sinkhole distribution. The Harbechy Cave lies on a NE-SW-orientated series of 15 sinkholes. A map of the study site with part of the Harbechy Plateau is shown in Figure 1. ERT profiles and positions of individual soil boreholes are also indicated.

Based on Köppen [21], MK, as part of the Czech Republic, belongs to the humid continental climate zone Dfb. Quitt [22] further divided this climate into seven climate groups, with MK itself falling into groups MT3 and MT5 (moderately warm areas) and CH7 (cold climate area). According to the Nature Conservation Agency of the Czech Republic [23], the MK can be divided into three parts, where the southern part, the central part, and the northern part show mean temperatures of 8.4 °C. 7.7 °C, and 6.5 °C and precipitation of 550 mm, 600 mm, and 700 mm, respectively. Temperatures could be slightly higher on karst plateaus. Data from the Macocha Weather Station on the neighboring Macocha Plateau (2 km as the crow flies with almost the same altitude, mean of 499.3 m for Harbechy Plateau vs. mean of 500.4 m for Macocha Plateau) show a mean temperature T_(average) = 10.2 °C with a total precipitation of 490 mm, calculated for the period between 2019 and 2025.

3. Methods

A multi-instrumental approach was applied for the characterization of well air and the conditions of GRSL formation. The hierarchy of the used method is depicted in Figure 2.

3.1. Electrical Resistivity Tomography

Five ERT profiles (P13–P17) were measured using an ARES II system [24]. P13 (106.5 m length, 1.5 m step) imaged the subsurface relief to a depth of 20 m, while P14–P17 (17.75 m length, 0.25 m step) resolved shallow vadose zone geomorphology.

3.2. Soil Drilling

Soil drilling is still the main method for directly verifying the presence of gas in the soil cover. The soils and loess layer were drilled through to the underlying limestones with a percussion drilling set with gouges (1 m long, 40 and 75 mm in diameter) with an RD-32 connection (Royal Eijkelkamp, Nijverheidsstraat 9, 6987 EN Giesbeek, The Netherlands) to a depth of about 3 m (down to the limestone bedrock). Coordinates (S-JTSK Krovak EastNorth; WKID 5514) of the resulting boreholes are shown in

Table 1. Coordinates of the ERT profiles and boreholes in Figure 1.

	X	Y	Z	WGS
P14	−587,451.50	−1,143,616.89	485.08	49.35860111	16.72924
P14	−587,455.96	−1,143,620.85	485.23	49.35856146	16.72918
P15	−587,463.04	−1,143,613.18	485.17	49.3586233	16.72907
P15	−587,455.96	−1,143,620.85	485.23	49.35856146	16.72918
P15	−587,448.61	−1,143,628.76	485.38	49.35849773	16.72929
P16	−587,464.15	−1,143,632.08	485.59	49.35845324	16.72909
P16	−587,478.85	−1,143,647.42	485.50	49.35830208	16.72891
P17	−587,479.08	−1,143,632.27	485.51	49.35843732	16.72888
P17	−587,463.58	−1,143,646.88	485.80	49.35832145	16.72912
PS1	−587,471.31	−1,143,639.42	485.61	49.35838079	16.72900
PS2	−587,470.62	−1,143,638.71	485.59	49.5838780	16.72901
PS3	−587,470.55	−1,143,640.16	485.62	49.3583749	16.72901
PS4	−587,472.08	−1,143,640.08	485.60	49.5837416	16.72899
PS5	−587,472.05	−1,143,638.82	485.60	49.35838545	16.72899
PS6	−587,469.23	−1,143,637.31	485.63	49.35840164	16.72902
PS7	−587,469.15	−1,143,641.52	485.68	49.35836407	16.72903
PS8	−587,473.52	−1,143,637.53	485.53	49.35839559	16.72897
PS9	−587,473.34	−1,143,641.56	485.61	49.35835972	16.72897
PS10	−587,455.97	−1,143,620.86	485.26	49.35856136	16.72918
PS11	−587,457.36	−1,143,619.42	485.24	49.35857291	16.72916
PS12	−587,457.39	−1,143,622.22	485.36	49.35854784	16.72916
PS17	−587,655.67	−1,143,761.03	486.12	49.35711779	16.72665
PS18	−586,072.57	−1,139,866.72	504.42	49.39344700	16.74265
PS25	−586,349.49	−1,140,723.66	484.62	49.38552100	16.74010
HAPR3-1	−587,477.42	−1,143,645.52	485.45	49.35832100	16.72892

P14 to P17 are ERT profiles (Figure 1). PS1 to PS25 are wells (boreholes) (Figure 1). PS17-1 (not shown in Figure 1) is the well near sinkhole 17 (SH17).

.

3.3. Gas Analysis

For detailed analysis of air from boreholes, the gas chromatography method was chosen for its accuracy and precision. The samples of air were taken in some wells, in addition to the annulus at the entrance to the cave, all at depths of 1 to 3 m. Samples were collected into Tedlar bags using a portable Ecoprobe-5 (RS Dynamics, LLC Technoparkstrasse 1, 8005 Zurich, Switzerland). Subsequently, all samples were analyzed in the laboratory by gas chromatography on HP 5890A instruments with TCD detection (He and H₂) or an Agilent 7890A with TCD and FID detection (CO₂, Ar, O₂, N₂, CO, CH₄, and hydrocarbons C2 to C6).

3.4. Organic Carbon in Drill Cores

An organic carbon analysis was performed from the bottom of the drill core to check the organic carbon content (SOM) in the deeper parts of the soil profile. Samples were oxidized in an oxygen stream at 1350 °C after previous removal of carbonates by dissolution in HCl (in 1:2 and 1:1 dilutions). Elemental composition was determined using a Metalyt CS 1000S (ELTRA) instrument with an infrared detector (IR).

3.5. Stable Isotopes

The values of δ¹³C-CO₂ were verified in soil air to find possible sources of the carbon. The carbon isotopes were determined on a Delta V Advantage GC IsoLink (ThermoFisher. Bremen, Germany). The gas sample was injected into a gas chromatograph. After separation in the column, it passed through a conversion reactor (in the case of CO₂ without temperature), followed by removal of water with a Nafion desiccant. The CO₂ was then passed through a ConFlo IV to the IRMS. The measurement was calibrated using reference gases. Data were evaluated using ISODAT software; Std used: carbon dioxide: UN1956 50% mol/mol, XO2AI50CA58C002 Airgas and Air Liquide company certified value: −40.1 ± 0.3‰ vs. VPDB.

3.6. XRD Analysis

X-ray diffraction was used to determine the composition or degree of transformation of the loess as a soil-forming substrate. Two soil samples labeled Harbechy (dark) and Harbechy (light) were analyzed by powder X-ray diffraction (XRD). Analyses were performed using a Panalytical X’Pert PRO MPD apparatus with a cobalt anode and an RTMS detector (X’Celerator) in conventional reflection geometry. Measurement parameters of a representative part of the samples: step size: 0.033° 2θ, time per step: 320 s, measured angular range: 4–100° 2θ, total scan duration: 2 h.

To accurately identify clay and related minerals, orientated mounts were created from the clay fraction (<2 μm) of the sample obtained by ultrasonic dissociation in an aqueous suspension and centrifugation [25]. The orientated mounts were dried in air at 25 °C and subsequently subjected to saturation with ethylene glycol vapors at a temperature of 60 °C and fired at 400 °C with an isothermal hold of 45 min. The oriented mounts were analyzed three times using XRD in the angular range 3.7–50° 2θ: (1) after drying in air, (2) after saturation with ethylene glycol vapors, and (3) after firing. Measurement parameters of the orientated mounts: step: 0.033° 2 Θ, time per step: 100 s, measured angular range: 3.7–50° 2θ. The acquired data were processed using the Bruker DIFFRAC plus EVA 2 and Topas 4 software. Quantitative phase analysis was performed using the Rietveld method.

4. Results

4.1. Electrical Resistivity Tomography (ERT)

Processed ERT profiles are shown in Figure 3. Dark areas indicate low resistivity (typical for wet sediments), while light areas correspond to high resistivity (typical for impermeable limestones). The longest profile, P13, orientated in a southwest–northeast direction and measured at 1.5 m intervals, shows the terrain structure down to a depth of 20 m. The shorter ERT profiles, P14, P15, P16, and P17, with a measuring step of 0.25 m, extend to a depth of 4 m. These profiles were designed to detail imaging of the subsoil.

4.2. Soils

The A-horizon of the soil profile, about 0.4 m deep, is associated with long-term agricultural activity, especially intensive plowing and fertilization. The subhorizon below (the confluent B- to C-horizons) with a thickness of 2–4 m is formed by loess, the original rock substrate.

Particles of the loess prepared from a drill core of Harbechy soil are shown in Figure 4. Spot analyses of the particles are in

Table 2. EDX analysis of loess: spot analysis of individual particles and mean composition from sample area (wt. %).

	Na	Mg	Al	Si	K	Ca	Ti	Mn	Fe	O
Spot Analysis	n	0.89	6.56	32.49	0.79	0.49	0.26	n	11.32	47.21
	0.29	2.78	38.37	0.31	0.23	n	n	n	8.92	49.1
	0.25	0.64	7.13	34.46	0.92	0.64	n	0.72	6.58	48.65
	2.5	0.82	11.63	28.93	2.44	0.7	n	n	5.83	47.16
	n	1.05	13.88	24.91	5.1	0.91	0.46	n	8.21	45.48
	n	0.55	7.67	35.67	1.08	0.56	n	n	4.82	49.65
	n	1.22	14.97	25.12	4.94	0.55	0.35	n	6.73	46.12
Area 50 × 50 µm	n	1.06	10.74	28.39	2.07	0.64	0.53	n	10.05	46.51

n—below detection limit.

. The qualitative phase compositions of two analyzed loess samples are nearly consistent. On the contrary, the quantitative phase analysis showed significant differences; see the details in

Table 3. Quantitative XRD phase analyses of loesses (drill cores, Harbechy Plateau, MK).

Minerals/Samples		Harbechy Loesses—Dark	Harbechy Loesses—Light
Clay minerals	Smectite	38.9	13.5
	IIllite and mica structures	18.6	11.2
	Kaolinite	10.4	2.7
	Chlorite	0.4	2.1
Oxides	Anatase	1.1	0.3
	Goethite	13.9	7.9
	Hematite	0.3	0.1
	Quartz	12.2	45.8
Carbonates	Calcite	0.6	0.1
Silicates	K-feldspar	3.1	6.9
	Plagioclase	0.8	9.5
	Sum	100.3	100.1

The mineral content is given as weight percentages.

. The qualitative identification of clay and related minerals was achieved through an analysis of an orientated mount of the clay fraction, as demonstrated in Appendix A.2. The profile of the diffraction line, which shows maximums at 14.7 and 12.9 Å after air drying and expanding to approximately 16.5 Å after ethylene glycol vapor saturation, suggests two distinct smectite structures. Montmorillonite structural models were used for their quantification. Upon firing, these reflections collapsed to about 10 Å. Reflections at 8.5, 5.53, and 3.32 Å, appearing after glycocolation, are also attributed to smectite. The diffraction lines at 10.0 and 4.98 Å, whose positions remain unaffected by any procedure carried out, belong to the basal planes of mica, probably illite. The reflections at 7.17 and 3.58 Å, with stable positions, correspond to the basal planes of kaolinite. In the Harbechy loess dark sample, the reflection at 7.09 Å, attributed to chlorite, becomes discernible with magnification. Considering the limited visibility of chlorite in the diffractograms of orientated mounts, as opposed to the records of a representative sample portion, it is apparent that most chlorite particles exceed 2 μm in size. Reflections at 3.34 and 4.25 Å are linked to quartz, present as a particle within the clay fraction. Diffractograms of randomly oriented representative portions of the analyzed samples are illustrated in Appendix A.1, while diffractograms for orientated mounts are shown in Appendix A.2.

4.3. Drill Cores

Descriptions of drill cores with explanations are in Appendix A.3. The lower part of the drill core analyzed (loess sample SP03) contained only a low organic matter content with a concentration of 0.12 ± 0.024 wt. %.

4.4. Underground Water

In general, an alternative way to test CO₂ concentrations in soil/epikarst is to reconstruct the concentrations from the water composition. Faimon et al. [16], Peyraube et al. [17,18], Milanolo and Gabrovšek [26], and Pracný et al. [27] established this method. In the calculation, virtual CO₂ is gradually added to degassed karst water until equilibrium is achieved again. For these purposes, drip water was sampled in the Harbechy Cave Gallery located below the studied area. The chemical composition of the water is presented in

Table 4. Composition of leaking/dripping water in Harbechy Cave. Gallery.

		Concentrations	Method
pH		6.8	Potentiometry
Sodium	Na	3.97 × 10−4	ICP-OES
Potassium	K	<1 × 10−5	ICP-OES
Ammonium ions	NH4+	7.22 × 10−6	Spectrophotometry
Calcium	Ca	6.74 × 10−3	ICP-OES
Magnesium	Mg	1.65 × 10−4	ICP-OES
Sulfates	S	4.69 × 10−4	ICP-OES
Chlorides	Cl	8.47 × 10−5	Spectrophotometry
Nitrites	NO2−	7.15 × 10−7	Spectrophotometry
Nitrates	NO3−	8.57 × 10−3	Spectrophotometry
Phosphates	P	3.23 × 10−6	Spectrophotometry
Alkalinity		5.88 × 10−3	Volumetric

All concentrations are expressed as mol/L units.

. As expected in karst water, the main components are calcium and alkalinity (HCO₃⁻), although the extreme nitrate content is also worth mentioning.

4.5. Composition of Soil Air in Wells

CO₂ concentrations measured in the topsoil of Harbechy Plateau (to a depth of 0.3–0.4 m) show standard values below 1 vol. %. which is a value typical for the entire Moravian Karst region [19]. The composition of the borehole air is shown in Figure 5. It shows the distribution of the main components in the boreholes on a logarithmic scale. In some places, the CO₂ concentration significantly exceeds the concentration in the topsoil. Another interesting feature is the fluctuating ratios of O₂ and N₂.

The CO₂ concentrations directly measured in the air of boreholes with a handheld IR spectrometer during the period 2022 to 2025 (12 to 35 individual measurements) are given in Figure 6. The median of values ranges from 1.5 to almost 3 vol. %. Maxima exceeds 4 vol. % (except for the PS18 well). The maximum in the HAPR3-1 borehole reaches 6 vol. %.

The results of comparative measurements with gas chromatography are presented in Figure 7a. Some medians of CO₂ concentrations are less than 1 vol. %, the remaining values range from 1 to 2 vol. %. In the case of the PS6-3 well, the median is 3 vol. %. Carbon isotopes were determined in the same air samples (Figure 7b). δ¹³C-CO₂ values are between −20 and −26‰.

The oxygen and nitrogen concentrations in the borehole air determined with chromatography are presented in Figure 8. O₂ concentrations are 9–24 vol. %, while N₂ concentrations 75–88 vol. %. The gases are negatively correlated.

5. Discussion

5.1. Geophysical and Soil Characterization

The ERT profiles (Figure 2) reveal a complex paleotopography beneath the Harbechy Plateau, with sediment-filled depressions (deepest: ~5 m near sinkhole SH5) and variable limestone bedrock morphology. High-resistivity zones indicate compact limestone, while low-resistivity areas confirm 4 m-thick sedimentary cover (loess-dominated). Historical resistivity profiling by Hašek (in Přibyl [15]) mapped a 50 m-deep paleo-relief, consistent with dye-tested drainage into Suchý/Pustý žleb. Our findings align with this author but reveal finer-scale paleotopography due to higher-resolution methods.

The soil is classified as haplic anthrosol (loamic, protocalcaric) [28] reflecting:

(1)

Long-term agricultural use (potentially since the Neolithic);

(2)

A 0.4 m-thick dark Ap soil horizon formed by tillage and fertilization;

(3)

Loess substrate (2–4 m thick; 0.1–0.4 wt. % of calcite; smectite/kaolinite-dominated, Table 4). Subrounded loess particles (Figure 3) suggest short transport, while low SOM implies localized CO₂ sources feeding the GRSL.

5.2. Water Chemistry

The modeling of water indicates it is almost in equilibrium with calcite (saturation index of calcite SI_calcite = −0.04) (

Table 5. Saturation indices (SI) and ion activity products (IAP) for infiltrated water (Harbechy Cave), calculated using the PHREEQC code [29].

Mineral Phase		SI	log IAP	log KT
Anhydrite	CaSO4	−1.79	−6.13	−4.34
Aragonite	CaCO3	−0.19	−8.44	−8.25
Calcite	CaCO3	−0.04	−8.44	−8.4
CO2(g)	CO2	−1.38	−2.62	−1.24
Dolomite	CaMg(CO3)2	−1.82	−18.49	−16.67
Gypsum	CaSO4.2H2O	−1.53	−6.13	−4.6
Halite	NaCl	−9.14	−7.6	1.54
Hydroxyapatite	Ca5(PO4)3OH	−2.06	−3.87	−1.82

CO_2(g) means PCO_2(w) = 10^−1.38, where PCO_2(w) is the partial pressure of CO₂ in water [26,27]. IAP—ion activity product; K_T—temperature-dependent equilibrium constant.

). The partial pressure of CO₂ in water (which represents the virtual partial pressure of CO₂ in cave air, at which water would be in equilibrium), PCO_2(w) = 10^−1.38, exceeds cave air PCO_2(cave) = 10^−1.42 (

Table 6. Results of modeling of water composition in PHREEQC [29]. The CO_2(g)form is the logarithm of the partial pressure of CO₂ at which the composition of the leaking water was formed.

Mineral Phase		SI	log IAP	log KT
Anhydrite	CaSO4	−1.79	−6.13	−4.34
Aragonite	CaCO3	−0.16	−8.4	−8.25
Calcite	CaCO3	0	−8.4	−8.4
CO2(g)form	CO2	−1.42	−2.66	−1.24
Dolomite	CaMg(CO3)2	−1.74	−18.42	−16.67
Gypsum	CaSO4:2H2O	−1.53	−6.13	−4.6
Halite	NaCl	−9.14	−7.6	1.54
Hydroxyapatite	Ca5(PO4)3OH	−1.85	−3.67	−1.82

The CO_2(g)form in this case is PCO_2(form) = 10^−1.42.

), suggesting slight degassing [12].

The geochemical modeling found the partial pressure, PCO_2(form) = 10^−1.42, at which the composition of infiltrated water was formed (Table 6). Under normal conditions, this pressure corresponds to a concentration of 3.8 vol. %. This concentration matches the measurements in the well. The principle of the modeling is based on virtually adding/subtracting CO₂ into/out the water/solution until equilibrium is reached.

5.3. CO₂ Dynamics

The hydrogeochemical modeling of drip water reveals the near-equilibrium with calcite (SI_calcite = −0.04) and elevated CO₂ partial pressure (PCO_2(form) = 10^−1.42, equivalent to 3.8 vol. %), consistent with borehole measurements (1.5–6 vol. %). The CO₂ gradient (Figure 9) reveals two features: (1) a sharp discontinuity at a depth of 1.7 m marking the GRSL boundary, and (2) a 0.8 m-thick layer with a baseline of 3 vol. % CO₂, defining the GRSL itself.

The isotopic signatures (δ¹³C = −25‰ to −8‰) reflect C3 plant respiration, with minor geogenic input (likely carbonate dissolution). Low O₂/CO₂ ratios (11–24 vol. %) suggest incomplete oxidation of SOM.

5.4. Isotopic Signatures

The isotopic composition of CO₂ is shown in the Keeling plot, where δ¹³C-CO₂ is plotted against the reciprocal CO₂ concentration (Figure 10). This relationship, derived from the law of mass conservation [30], represents the mixing line of the two end members: atmospheric CO₂ (δ¹³C = −8.0 ± 0.1‰, [CO₂] = 420 ppm; [31,32]) and the local biogenic source (δ¹³C = −25.0 ± 0.5‰, [CO₂] = 35,000 ppm; [33,34]).

The points on the line indicate that the soil air is a mixture of the end points. The points above the line (PS6, PS10, PS5, SP00, and PS17-b) indicate a small fraction of abiogenic CO₂ probably coming from acidic dissolution of carbonates [6]. Acidity could come from nitrification [35]. Points below the mixing line (PS17-a1, PS17-a2) suggest the presence of another source, e.g., carbon from methanogenesis (δ¹³C = −55 to −90‰, [36]) or CO₂ from cellulose decomposition (δ¹³C = −27 to −32‰, [37]).

The N₂ and O₂ concentrations complement each other (Figure 8). However, O₂ concentrations show significant deviations from the standard value of about 21 vol. %. Generally, 1 mol of O₂ is consumed at the formation of 1 mol of CO₂, see Equation (1). Therefore, under ideal conditions (oxidation led to complete CO₂ formation), the sum of O₂ and CO₂ concentrations is 21 vol. %. However, in this study, this sum moves between 11.12 and 24.02 vol. % with a mean value of 19.82 ± 3.01 (SD). The low sum of CO₂ and O₂ indicates that not all oxygen was consumed for CO₂ production. Some oxygen probably binds to oxidized organic molecules during SOM degradation (see the pathways of oxidation of cellulose and lignin in the Section 1).

5.5. Climatic Influences

CO₂ concentrations within the borehole air exhibit considerable instability, as evidenced by the dispersion of data in Figure 5 and Figure 6. As the conditions in the deeper soil strata, especially temperature, remain relatively stable, consideration of external influences appeared necessary. The temporal evolution of CO₂ concentrations was systematically analyzed in correlation with the progression of the climatic variables coming from the Macocha Weather Station (Figure 11a).

The trends in CO₂ concentrations weakly correlate with trends in external temperature (positively) and soil moisture (negatively) (Figure 11), see the peaks (Mar–Jun) in temperature (T), soil moisture (SM) (Figure 11a), and CO₂ concentrations (Figure 11b). Agricultural remediation (reduced tillage/grass planting [38]) can exacerbate CO₂ accumulation by limiting soil ventilation.

Despite the significant variability in the CO₂ data (Figure 11b), certain indications of influence are apparent. Peak CO₂ concentrations are observed between mid-March and June, roughly aligning with periods of minimal precipitation, reduced soil moisture, and elevated temperatures. On closer analysis, it is observed that the highest temperatures shift to early September, a trend like that observed in the minimum soil moisture levels. From a causal perspective, this shift is unlikely to reflect a strong relationship. The impact of precipitation on 5 June 2024 and 17 May 2024 appears more plausible. However, the expected similar impact on 3 March 2025, 4 April 2025, and 6 May 2025 is minimal, if present at all.

5.6. Conceptual Model

The conceptual model of GRSL formation is in Figure 12. GRSL formation is driven by SOM accumulation at the soil–bedrock interface, with CO₂ fluxes (production jₚ, efflux j_ef, diffusion j_d, and water transport jᵥ) governing concentration dynamics. The relatively low concentration of SOM found in the drill cores indicates that the main source of SOM and CO₂ could be localized in a lesser area, e.g., in a single depression. From there, CO₂ likely spreads to the wider environment.

Generally, the concentration of CO₂ in the GRSL, c_CO2, is proportional to the sum of all mass fluxes.

$${c_{CO}}_2\approx k (j_p-j_{ef}-j_d-j_w)$$

(3)

where j_p, j_ef, j_d,, and j_w are fluxes associated with CO₂ production, efflux, diffusion through the bedrock, and transport of dissolved CO₂, respectively. k is a proportionality coefficient. Note that the fluxes out of the soil have negative signs. It should be emphasized that any change in one of the fluxes will lead to a change in CO₂ concentration. Therefore, all factors driving the single CO₂ fluxes also influence the soil concentrations.

5.7. Implications

To our knowledge, this is the first report about CO₂-rich layers in soil. That is why the direct comparison with similar works is limited. So, we contextualize our findings within adjacent research areas. We can compare our results primarily with the work of Benavente et al. [39] and Vadillo et al. [40], who dealt with CO₂ measurements in the karst zone near Nerja Cave (Malaga, southern Spain, <1 km from the Mediterranean shore). The authors report for shallow soils CO₂ concentrations in the range of 710 to 6900 ppmv (alternatively 340 to 3200 ppmv). These values are in the range that was also measured in Moravian Karst soils [19]. In deeper wells, Benavente et al. [39] measured CO₂ concentrations in the range of 2.5 to 5.5 vol. %, which roughly agree with results our study. However, compared to our study, the values were measured in a much deeper vadose zone (14 to 22 m). The isotopic composition of δ¹³C-CO₂ determined by Benavente et al. [39] ranged from 10.5‰ (atmosphere) to 21.5‰ (soil) or 22.0‰ (boreholes). Some of our samples (PS17-a1, PS17-a2) showed somewhat isotopically lighter CO₂, δ¹³C-CO₂ up to −26.3‰ and approaching the value of −27.7‰ reported by Cerling [34] for soil-respired CO₂ based on the C3 photosynthetic pathway. However, the average of our values, δ¹³C-CO₂ = −23.1 ± 2.17 (SMODCH) ‰, approaches the value of −23.3‰ resulting from Cerling’s diffusion model [34].

The method of soil/epikarstic CO₂ calculation from seepage water hydrogeochemistry was verified due to agreement of calculated concentrations with those measured in the GRSL. The method was generally designed because of the discrepancy between water composition and commonly measured soil CO₂ concentrations ([16,17,26]). Atkinson [41] did pioneer work when he was one of the first to point out the discrepancy between the hydrochemistry of karst springs and CO₂ concentrations in soils. He called the CO₂ calculated from the hydrogeochemistry of the water “subsurface CO₂”.

5.8. Environmental Notes

Elevated carbon dioxide concentrations on the Harbechy Plateau are consistent with an agricultural land reclamation that involved the cessation of tillage and the planting of selected grasses [38]. However, this study shows that the remediation method mentioned above appears problematic. Cessation of tillage reduces soil ventilation, potentially exacerbating CO₂ accumulation despite decreased SOM input. Given the extensive history of land cultivation, including tillage and fertilization over several centuries, it is reasonable to assume that organic matter was leaching from the topsoil by rainfall and subsequently accumulated at the soil–bedrock interface with reduced permeability. Although carbon dioxide production occurs similarly to before remediation, the cessation of tillage likely led to a reduction in typical soil ventilation during the winter months. Furthermore, the formation of continuous grasslands may hinder the diffusion of carbon dioxide into the atmosphere, thereby reducing the efflux of CO₂. The altered ratio of CO₂ production/efflux would lead to elevated CO₂ concentrations. We plan to conduct an in-depth study of the problems in the future.

6. Conclusions

• Validation of the GRSL Hypothesis
  • Over the past five decades, the hypothesis of a gas-rich soil layer (GRSL) that controls the composition of percolating water in the karst system has been developed. This study bridges the gap between drip-water-derived CO₂ (PCO₂(form) = 10^–1.48, i.e., 3.8 vol. %) and topsoil measurements.
• Field Evidence from Harbechy Plateau
  • The haplic anthrosol (loamic, protocalcaric) overlying Harbechy Cave. The GRSL was identified via: (a) drilling (direct detection at the soil–bedrock interface, 2–4 m depth), (b) ERT (confirmed undulating limestone bedrock topography), (c) CO₂ gradients (average concentrations of 1.5–3 vol. %, peaking at 4–6 vol. % in a layer 0.8 m thick), and (d) leaking water composition.
• Isotopic Insights
  • δ¹³C signatures reveal that GRSL CO₂ is a mixture of (a) biogenic (δ¹³C = −25‰), (b) atmospheric (δ¹³C = −8‰), and (c) minor abiogenic sources (likely geogenic).
• Climate Decoupling
  • No significant correlation was found between the GRSL CO₂ fluctuations and the weather station data, suggesting that carbon dioxide dynamics are buffered by soil processes.
• Broader Implications
  • The GRSL represents a previously overlooked carbon pool in karst agroecosystems, with potential impacts on (a) better understanding of karst hydrogeochemistry and karst process, (b) carbon cycling models (especially under land use change), and (c) speleothem formation (through altered seepage water chemistry).
  • This study sheds light on the origin of high CO₂ concentrations involved in calcite–CO₂−water interactions. However, many questions still await clarification. We need to examine in detail other influences: climatic, local (nature of the soil, agricultural use), and others. It is necessary to examine in detail the influence of fertilization in relation to isotopic composition and concentration, including the establishment of limits.
  • We believe that this work establishes a baseline for future detailed studies.