CO₂ Emission from Caves by Temperature-Driven Air Circulation—Insights from Samograd Cave, Croatia

Abstract

Opposite to atmospheric CO₂ concentrations, which reach a minimum during the vegetation season (e.g., June–August in the Northern Hemisphere), soil CO₂ reaches a maximum in the same period due to the root respiration. In karst areas, characterized by high rock porosity, this excess CO₂ seeps inside caves, locally increasing pCO₂ values above 1%. To better understand the role of karst areas in the carbon cycle, it is essential to understand the mechanisms of CO₂ dynamics in such regions. In this study, we present and discuss the spatial and temporal variability of air temperature and CO₂ concentrations in Samograd Cave, Croatia, based on three years of monthly spot measurements. The cave consists of a single descending passage, resulting in a characteristic bimodal climate, with stable conditions during summer (i.e., stagnant air inside the cave) and a strong convective cell bringing in cold air during winter. This bimodality is reflected in both CO₂ concentrations and air temperatures. In summer, the exchange of air through the cave’s main entrance is negligible, allowing the temperature and CO₂ concentration to equilibrate with the surrounding rocks, resulting in high in-cave CO₂ concentrations, sourced from enhanced root respiration. During cold periods, CO₂ concentrations are low due to frequent intrusions of fresh external air, which effectively flush out CO₂ from the cave. Both parameters show distinct spatial variability, highlighting the role of cave morphology in their dynamics. The CO₂ concentrations and temperatures have increased over the observation period, in line with external changes. Our results highlight the role of caves in transferring large amounts of CO₂ from soil to the atmosphere via caves, a process that could have a large impact on the global atmospheric CO₂ budget, and thus, call for a more in-depth study of these mechanisms.

1. Introduction

The year 2023 was the 12th consecutive year of global atmospheric CO₂ increase by at least 2 parts per million (ppm) per year, and this trend continues in 2024 [1,2]. This increase is resulting from continuously increasing fossil CO₂ emissions that outpace the reduction in land use, land-use change, and forestry emissions (although the uncertainties for this reduction are too large to be considered robust [1]) and ocean uptake. The anthropogenic emissions linked to burning fossil fuels and land-use changes occur additionally due to natural sources in the atmosphere, hydrosphere, and biosphere [3]. While our conceptual understanding of the mechanisms behind natural CO₂ release to the atmosphere is adequate, large uncertainties remain, especially linked to the amount of CO₂ released by root respiration and stored as ground air in the unsaturated zone of karst aquifers [4]. These aquifers occur in regions with carbonate and other evaporative rocks, which cover about 20% of the Earth’s ice-free continental area.

Ground air in limestone rocks has long been known to be an important reservoir of CO₂, although the source of the gas and its interactions with the various reservoirs (soil CO₂, microbial-produced CO₂, degassing from water, precipitation of secondary calcite, etc.) are only partly understood [4,5,6,7,8]. These sources result in cave atmospheres accumulating large amounts of high concentrations of CO₂, where values up to several thousands of ppm are common [9,10,11,12,13,14], and, in particular cases, even up to tens of thousands ppm CO₂. The exact source of this CO₂ is unclear, with soil CO₂ [7] and ground air [5] being the most important ones. Monitoring studies in caves located in different climates have added important new information to our understanding of CO₂ dynamics in caves [15,16,17,18] and the links with external atmospheres. Daily to yearly variations in pCO₂ have been observed, linked to seasonal changes in soil CO₂ production, the influence of cave biota, cave ventilation regimes, and visitors’ influence. No study demonstrated a clear link between the long-term growth of atmospheric CO₂ concentrations and similar dynamics in caves. Such a link is difficult to observe, as the various sources of cave CO₂ and the large values observed in caves (see above) potentially mask the external signal. Additionally, the internal cave environment is highly diverse, shaped by cave morphology, sediments, and hydrology not only in large conduits, but also in dense networks of impassable fissures. These fissures, though inaccessible to humans, contribute significantly to the cave’s atmospheric dynamics [6,14,19,20]. Furthermore, the complex cave morphology creates substantial differences between cave sections, making accurate measurements even more challenging due to the technical limitations of current measuring equipment.

Disentangling the complex interactions between CO₂ sinks and sources in caves and karst areas requires long term observations that are not readily available. Especially missing are data linking CO₂ dynamics with climatic parameters that would help to understand the “export” of CO₂ from caves, a yet poorly understood mechanism. Here, we provide a large dataset of cave climate and CO₂ concentrations (pCO₂) from Samograd Cave (Croatia) and discuss their long- and short-term variations. We show that complex interactions between cave morphology and microclimatic conditions lead to large quantities of CO₂ being exported from karst areas, a yet poorly constrained but potentially important source of CO₂ adding to the global atmospheric carbon budget.

2. Materials and Methods

2.1. Cave Description

Samograd Cave is located on Grabovača hill in Perušić (Cave Park Grabovača, Lika region, Croatia, Figure 1), where the entrance elevation is 684 m above sea level (asl).

The cave features a simple morphology with a single entrance and a unique 346 m long passage, up to 32 m high and 25 m wide. The deepest point of the cave is 52.7 m below the surface; alternatively, if measured from the entrance of the cave itself, at the bottom of the doline (spot measuring point T2 in Figure 1), the cave depth is 40.5 m. Beyond the entrance, the main cave passage (8–17 m wide and 3–11.5 m high) is steeply descending. The bottom of this passage is covered by debris, limestone blocks, and soil derived from the bottom of the entrance doline. The ceiling has a complex topography due to the dense network of collapse and dissolution features (e.g., chimneys, and half tubes). This entrance section is followed by a large (18–33 m wide and 11–22 m high) but short (35 m long) SSW-oriented passage, with a flat to slightly inclined floor, covered by rock debris, while further inside the cave, there are flowstones and large dome-shaped speleothems. The ceiling is characterized by a complex topography controlled by a fault zone, giving the gallery a characteristic A-shaped cross section. The gallery then narrows slightly and slopes steeply towards the lowest point, but the ceiling maintains a similar height with a similar topography (as in previous part), featuring numerous chimneys as well as flat rock surfaces. There are also several fracture zones through which, after heavy rain or snowmelt, a significant amount of percolating water accumulates into the passage, forming small streams on the ground. The passage width at the lowest part of the cave is 11.5–16.5 m, with a maximum height of 32 m. From this flat area, the passage continues towards the next section along an ascending passage, up to 10 m wide and up to 22.7 m high. The ceiling also ascends parallel to the bottom, somewhat milder slope. This section is characterized by partially human-made flat terraces covered with sinter and rock debris. After a narrow section, the cave continues, reaching a widened part (up to 10 m wide and 16.5 m high) which is shaped like a dome (the 4th along the cave, see Figure 1). The final 40 m of the gallery consists of a narrow and steep passage rich in speleothems.

2.2. Methods

Air temperature (T_air) and CO₂ concentration were recorded at ground level at 12 stations (Figure 1 and

Table 1. Spot measurement location distance and vertical distance from the entrance level.

Location	Distance (m)	Horizontal Distance (m)	Vertical Distance (m)	Note
T1	-	-	-	Outside point
T2	0.0	0.0	0.0	Collapsed doline bottom
T12	12.8	11.4	5.6
T10	28.3	25.3	12.8
T3	44.8	40.8	18.4
T11	87.6	82.9	28.4
T4	132.6	126.3	40.5	Lowest point in the cave
T9	159.7	156.4	32.5
T8	170.6	168.6	26.2
T7	180.6	179.8	17.3
T6	187.0	185.7	21.7
T5	202.9	202.5	12.6

) with a handheld AZ 77525 multimeter (AZ Instrument Corp., Taiwan) monthly between March 2021 and February 2024.

The multimeter is equipped with Non-Dispersive Infrared Sensor (NDIR) that is calibrated at 400 parts per million (ppm) in open air. The CO₂ measuring range is 0–9999 ppm with 1 ppm resolution and ±50 ppm accuracy. Air temperature resolution is 0.1 °C and the accuracy is ±0.6 °C. Measurements were performed on a monthly basis at 12 stations on a longitudinal profile from the entrance to the end of the cave as follows (Figure 1): one location at the surface above the cave (T1), one location at the bottom of the collapse doline in front of the entrance (T2), and 10 locations from the entrance towards the end of the cave. All measurements of CO₂ and air temperature were performed at ground level on various surfaces as follows: bare soil (T1), soil covered with sparse vegetation (T2), rock debris, and flowstone (all other locations). Measurements outside the cave (T1 and T2) were always conducted in the shade.

3. Results

The results of the air temperature and CO₂ measurements in Samograd Cave are presented in Figure 2 and summarized below.

3.1. Air Temperature

Air temperature (T_air) dynamics follows a clear annual cycle (Figure 2) outside the cave and at the bottom of the entrance doline, which rapidly decreases in amplitude (Figure 3) with increasing distance from the entrance, being hardly apparent at the final station of the cave.

The maximum T_air values were recorded yearly in July and August outside the cave (T1 and T2), while inside, they were recorded randomly between July and November, but always after the external ones. Notably, in 2023, the cave maxima T_air was registered with an increasing delay with increased distance from the entrance, reaching the most distant locations in November. The minima were recorded in January 2021 and 2022 and February 2023 and occurred simultaneously throughout the cave. The maxima and minima follow opposite trends along the cave, decreasing and increasing, respectively, from the entrance towards the inner parts of the cave, although the minima does show a very slight increasing tendency towards the very final station (Figure 3). These trends are mirrored by the overall temperature amplitude, which decreases with increasing distance from the entrance (Figure 3).

At two stations (T3 and T4), the increasing tendencies of winter temperatures are reversed (Figure 2 and Figure 3), likely as the result of cold air accumulation in topographic depressions along the cave’s floor (Figure 1, see “Discussions” for further details).

3.2. Carbon Dioxide

Carbon dioxide values range between 400 ppm (February and March 2021) and 449 ppm (July 2023) outside the cave and between 400 ppm (March 2021) and 2578 ppm (August 2022) inside the cave. The amplitude of CO₂ concentration (Figure 4) follows a tendency opposed to that of the air temperature, increasing sharply at the first cave station (T12) and continuously rising throughout the cave. Notably, this large amplitude in CO₂ concentrations is the result of very high summer values (Figure 2), while winter values are stable at generally low values close to those outside the cave.

CO₂ concentrations in the atmosphere outside the cave (station T1) are variable, likely due to local influences (e.g., anomalies of the weather systems reaching the cave site). At station T2, located near the bottom of the collapsed doline and still strongly influenced by outside climatic conditions, the CO₂ concentrations have large variations, with maxima exceeding 800 ppm in summer 2021 and 2022. Starting with station T12 (near the cave’s entrance), CO₂ concentrations show a clear bimodal distribution, with low values between November and April and high values between May and October, peaking in August–September (Figure 2). The values start to increase in April, peak in August (2021 and 2022) and September (2023), and decrease abruptly, reaching background values by November. The average values at each cave station increase with distance from the cave entrance, with the highest values recorded at station T5. Minimum pCO₂ values are almost constant for every station (standard deviation (1σ) of 13.38 ppm), while the maxima are more variable, as also indicated by the higher standard deviation (682.85 ppm at 1σ).

4. Discussion

4.1. Cave Climate

The mean air temperature within the cave exhibits a clear seasonal trend, increasing during the warmer months and decreasing during the colder months, which aligns with typical seasonal temperature variations (Figure 2 and Figure 5) in caves [19,20]. Minimum and maximum values also follow a similar pattern, but the standard deviations (Figure 5) are higher during summers than winters, indicating more uniform conditions between the different parts of cave during winter months. The temperature range tends to be larger during the warmer months, indicating larger differences between distinct parts of the cave. During the colder months, the temperature range narrows, reflecting more uniform conditions. These seasonal contrasts are common in caves with similar morphology, where outside influences are transmitted mainly through a single “entry point”, as opposed to caves with multiple entrances and/or complex morphologies, where air circulation patterns tend to be more variable, thus complicating cave air temperature dynamics.

Downsloping caves with a single entrance typically experience convection cells during winter. In this process, cold outside air enters the cave along the floor, warms up as it moves inward, and then ascends along the upper part of the passage towards the entrance [19,20]. Such convection cells can also be observed in Samograd during cold periods when the outside temperature falls below 7 °C. The extent of the cell varies with temperature and has not yet been determined. However, ice formations can be observed on the floor below the entrance. Additionally, cryoclasts are observed along most of the cave, indicating even deeper freezing in the past. During warm months, the convective cell is inactive, and temperatures throughout the cave stabilize, with a distinct thermocline near the entrance. The strength and spatial extent of the convective cells vary in time, as indicated by the correlation coefficients (Figure 6) between external air temperature (station T1) and cave temperatures (stations T2–T12). The correlation is stronger during the late autumn, winter, and early spring months for stations located closer to the entrance, and it quickly diminishes with increasing distance from the entrance as the inflowing cold air quickly warms in contact with the warmer cave surfaces, e.g., [20]. In summer, the correlation completely breaks down, as the cave’s atmosphere is decoupled from the exterior (Figure 6). Notably, T_air changes at the locations located furthest from the cave’s entrances (stations T5–T7) that are not linked to those outside the cave further support the occurrence of stable cave conditions.

4.2. Temporal Variability of pCO₂ Values

pCO₂ values in the cave follow a clear annual cycle, with the minima occurring in winter and the maxima occurring in late summer through early autumn (Figure 2 and Figure 7). The average cave maxima occur yearly in September and the minima between January and March, in contrast to the global values (Figure 7). The closest atmospheric CO₂ monitoring station is at Hegyhátsál (Hungary, ~300 km to the NNE). At this station, atmospheric CO₂ values follow a clear annual cycle, with an annual maximum in December and minimum in August [21]. The minima is linked to vegetation development starting in April, acting as a strong atmospheric CO₂ sink. Interestingly, atmospheric CO₂ concentrations start to decrease in January already, likely as a result of mixing between near-surface CO₂-rich air with CO₂-depleted free tropospheric air [21]. The development of vegetation starting in March–April in the area above the cave likely results in a similar depletion of atmospheric CO₂, but drives up soil CO₂ concentrations (via root respiration), thus providing the source of CO₂ seeping in the cave, driving values up to their maxima (Figure 2 and Figure 4). As vegetation activity decreases in September, root respiration and soil CO₂ values start to decrease, reaching a minimum in winter months, leading to clear seasonal differences inside the cave. These seasonal differences are also likely linked to the dynamic behavior of the cave. During cold periods, the influx of cold external air continuously mixes with the cave atmosphere and dilutes the CO₂ within it. Additionally, CO₂ production in the soil and epikarst is minimal in winter months, leading to the depletion of the surrounding massif’s CO₂ reservoir. As a result, the CO₂ input from the massif is also minimal [5,6,12,13,22,23].

Intriguingly, the average cave CO₂ concentrations, calculated for stations T10 through T5 (Figure 7), exhibit a slight increasing trend throughout the monitoring period, loosely following the global increase, but, due to the short observation period, no definitive conclusions can be drawn. The increasing trend is almost similar for all cave stations except station T2, indicating that despite potential uncertainties induced by our methods, the observation is robust.

pCO₂ values reach maxima throughout the cave with a delay of about two months with respect to T_air maxima (Figure 2), a process we suggest occurs via the following two mechanisms: (1) soil pCO₂ reaching maximum in late summer/early autumn [24,25,26,27], and (2) a potential delay in the transfer of CO₂ from soil to the cave via water [28,29,30]. The increase in pCO₂ values in the cave starts yearly in April, but this increase is slow until June, when values “jump” by several hundreds ppm (Figure 2). The high pCO₂ values are maintained until late September each year, abruptly dropping in late September/early October, and continuing to decrease until January/December. The slow increase in cave pCO₂ in early spring is noticeable mostly in the inner parts of the cave (Figure 1 and Figure 2) and are likely related to the slow release of CO₂ from the rock matrix and cessation of air circulation and the subsequent flushing of the cave with cold, dry, and low pCO₂ air. Likely, these high pCO₂ values result from a large reservoir of ground air rich in CO₂, stored in the limestone matrix above the cave [4,7,31]. It is very likely that, in the absence of cold air inflow during winter, the entire cave would have higher atmospheric pCO₂ values throughout the year.

The primary source of CO₂ in the cave, as supported by the temporal and spatial dynamics of pCO₂, is high soil microbial activity and plant root respiration [8,15], as well as microbial activity in the vadose zone above the cave [4,11,15,32]. CO₂ migrates downward to the vadose zone through advection and diffusion. Maximum values of pCO₂ in the soil and vadose zone occur in late summer/early autumn.

A secondary potential source of high pCO₂ values could be related to CO₂ released by tourists visiting the cave. The number of tourists visiting the cave peaked in August 2021, June 2022, and May 2023, while pCO₂ values peaked in August 2021, August 2022, and September 2023, respectively. The correlation analysis between pCO₂ values and the number of visitors shows low correlation, with no clear link between the two. A positive correlation between the visitor numbers on pCO₂ values has been observed in caves with thousands of visitors per day for extended periods of time, e.g., [18,33], but any increase in pCO₂ values is temporary (hours) and does not impact the long-term (monthly) values [18,33,34,35,36,37]. Given the very low number of visitors in Samograd Cave (monthly maximum below 2200 people), their influence on the cave pCO₂ values is negligible, at best.

Degassing from cave waters, either direct or due to calcite precipitation, a known source of CO₂ in caves, can only provide low amounts of CO₂ in Samograd Cave as no permanent water bodies or abundant calcite precipitating streams are found in the cave. Another potential source of CO₂ in caves is in-cave microbial activity. CO₂ production in caves requires the decay of organic matter transported to the cave from the outside, e.g., [4,5]. However, organic matter is found in small amounts in Samograd Cave (near the entrance area), so we can safely rule out this source as a potential contributor to the overall CO₂ balance in the cave.

4.3. Spatial Distribution of pCO₂ Values

The spatial distribution of pCO₂ follows a clear pattern of increasing values from the entrance to the end of the cave, albeit with notable differences between summer and winter (Figure 8).

Both winter and summer pCO₂ values are low near the cave entrance and progressively increase towards the end of the cave; however, winter values are limited to well below 500 ppm (with the exception of February 2021; pCO₂ 542 ppm), while summer values rise above 2500 ppm. The summer maxima are registered yearly in the station at the end of the cave (T5), but the winter maxima occur more randomly between station T4 and T5 (Figure 1). The high summer pCO₂ values registered throughout the cave support a source of CO₂ from the soil and, possibly, ground air, an inference further supported by the high values registered towards the end of the cave in winter (Figure 8). The cave’s morphology and related air circulation patterns (described above) are very likely responsible for this distribution. Cold air inflow in winter, which effectively cascades e.g., [38,39] down towards the lowest parts of the cave (T4 in Figure 1), dilutes the high CO₂ concentration air (Figure 7). There is an inflow of CO₂ from the surrounding massif into the cave (also indicated by higher values at T7–T9), which is subsequently flushed from the cave once it reaches the active convection cell.

The maximum CO₂ concentration varies across locations, with higher values observed at the inner locations (e.g., T5). This indicates that certain areas of the cave experience periods when CO₂ levels spike significantly, potentially due to less ventilation or other localized factors. The minimum CO₂ concentrations are fairly uniform across locations, indicating a consistent baseline level of CO₂ throughout the cave. The range of CO₂ concentrations (difference between max and min) increases slightly towards the interior of the cave. This suggests that while baseline CO₂ levels are consistent, the variability or fluctuation in the CO₂ levels increases in some locations, particularly deeper into the cave. The spread of CO₂ concentrations around the mean is relatively consistent across locations, with a slight increase toward the more interior locations. Higher standard deviation values towards T5 indicate that the CO₂ levels are more variable in these areas.

The comparison of variability between CO₂ and air temperature was conducted using the Coefficient of Variation (CV, representing the ratio of standard deviation to the mean), calculated from a dataset comprising monthly spot observations for the CO₂ concentration and air temperature between March 2021 and February 2024. The CV was chosen as a measure of variation instead of the standard deviation (SD) because it allows for a comparison of relative variability between variables with different units or scales, providing a normalized measure of dispersion relative to the mean. When comparing the CV calculated from all available data for the CO₂ concentration (N = 360) and air temperature (N = 360) at cave locations (T12 to T5, excluding T1 on the surface and T2 at the collapse doline bottom), the air temperature exhibits a higher range of relative variability (40.8979) than the CO₂ concentration (4.2792,

Table 2. Coefficients of variability for CO₂ and air temperature.

	CO2 (ppm)	Tair (°C)
Max CV	71.4297	44.8979
Min CV	67.1504	4.0119
Range CV	4.2792	40.8860

).

The maximum CV for T_air is lower than that of CO₂, but still indicates substantial variability relative to the mean temperature. This suggests that air temperature can fluctuate considerably under certain circumstances. The minimum CV for T_air indicates that there are conditions within the cave where the air temperature is relatively stable. The range of the CV for T_air is large, indicating a significant difference in relative variability under different conditions or at different locations within the cave.

Generally, CO₂ tends to have a higher CV across most locations within the cave, suggesting that CO₂ levels are more variable relative to their mean. This suggests that CO₂ concentrations are more sensitive to changes in environmental factors such as ventilation, biological activity, or other microclimatic conditions. The CV for air temperature is typically lower than that for CO₂, reflecting the greater stability in temperature across the cave measuring stations. In the most distant cave locations, where ventilation might be reduced, the CV for CO₂ is likely to be higher. This indicates more significant fluctuations in CO₂ concentrations, possibly due to limited air exchange, leading to the accumulation or depletion of CO₂ in response to localized activities or seasonal changes. Locations closer to the cave entrance or areas with more consistent airflow may show a lower CV for CO₂, reflecting more stable CO₂ levels due to regular ventilation. However, even in these areas, the CV for CO₂ might still be higher than that for T_air, indicating that CO₂ is inherently more variable.

The seasonal pattern of high summer and low winter pCO₂ values was further confirmed by performing a regression analysis using surface air temperature as the independent variable and CO₂ levels in cave locations as dependent variables. The regression analysis showed a strong relationship, indicating that the surface air temperature significantly influences CO₂ levels within the cave but only in the winter months. As the surface air temperature decreases, enhanced ventilation occurs, leading to lower CO₂ concentrations. Conversely, higher surface temperatures correspond to reduced ventilation and higher CO₂ levels.

When analyzing the CO₂ concentrations across different cave locations (Figure 8), we observe that CO₂ levels and variability tend to increase towards the interior sections of the cave, particularly around station T5 (Figure 1). This spatial trend can also be understood in the context of the cave’s ventilation patterns.

During colder months, the inflow of fresh air from the surface provides ventilation that reaches the outer sections of the cave more effectively, helping to maintain lower and more stable CO₂ levels. However, further into the cave, particularly in the more enclosed sections, this ventilation effect diminishes, leading to higher CO₂ concentrations and greater variability. In the summer, when this ventilation is absent, CO₂ levels rise more significantly, especially in the interior locations, resulting in the observed higher concentrations and increased variability.

The regression analysis between surface air temperatures and CO₂ concentrations at different cave locations further supports this interpretation. It demonstrates that the cave’s natural ventilation system, driven by temperature differences between the cave interior and the surface, plays a critical role in regulating CO₂ concentrations. In colder months, this ventilation is more effective in reducing CO₂ levels, especially in the outer cave sections, while in warmer months, the lack of ventilation leads to CO₂ accumulation, particularly in the cave’s interior.

In both the temporal and spatial analyses, the variations in CO₂ levels within the cave can be largely explained by changes in ventilation driven by the external air temperature. This natural ventilation system is more active during colder months, leading to lower and more stable CO₂ levels, while during warmer months, the absence of ventilation results in higher CO₂ concentrations and greater variability, particularly in the cave’s interior locations.

A further implication of our study is for the wider scientific community studying past climate variability using δ¹⁸O and δ¹³C values of speleothem calcite as proxies of past climate variables [14,40]. One important prerequisite of such studies is to determine when (during the year) the calcite is precipitated, as this would lead to the knowledge of what climate variable the stable isotopes in the speleothem calcite are registering, e.g., [41]. The high variability of pCO₂ in the cave atmospheres, as shown in this article, results in calcite being potentially precipitated at different times during a year in different parts of the cave. Consequently, the stable isotopic composition of the calcite will register different climate variables in different parts of the cave. Any potential study using these proxies should thus consider the spatial and temporal pCO₂ variability in the targeted cave before embarking on past climate reconstructions. Previous studies analyzing the stable isotopic composition of cave CO₂ have generally focused on understanding what the δ₁₃C value in speleothems is actually registering in terms of climate information. The rapid inflow of outside air was linked to an increase in the calcite precipitation rate and a drop in δ¹³C [42], but other studies [15] have shown that such changes do not occur if the pCO₂ variability is low, thus calling for similar analyses to be performed in every particular case. These could also result in a better understanding of CO₂ fluxes, as shown by Yidong et al. [18]. These authors have used δ¹³C of CO₂ in order to disentangle between the different sources and fluxes in a cave in central China, and their data indicate a clear separation (in terms of stable isotopic composition) of atmospheric and soil air (i.e., −8 ‰_VPDB vs. −25 ‰_VPDB). Consequently, δ¹³C values of CO₂ could be used to track the CO₂ fluxes in and out of caves through their entrances (during ventilation events like the ones described by us), as well as from soil air through the vadose zone through gravitational settling.

5. Conclusions

This study focuses on the variations in pCO₂ and air temperature within a cave system that functions as a convective cell. Measurements were conducted from March 2021 to February 2024 via spot measurements. During winter, cold surface air flows into the cave, lowering the air temperature throughout much of the cave and decreasing CO₂ concentrations. In contrast, during the summer, the cave maintains a cold air pool trapped in its lower part due to the cave’s morphology, leading to equilibrium conditions. The seasonal variability is attributed to the cave air dynamics, where cold air inflow during winter effectively displaces CO₂, and high summer pCO₂ levels result from CO₂ production in the soil and its subsequent seepage into the cave atmosphere through a network of fractures (Figure 9).

The observed delay in the pCO₂ maxima relative to air temperature peaks suggests a complex interaction between soil CO₂ production and its transfer to the cave, potentially involving both air flow and degassing from seeping water. The gradual increase in cave pCO₂ starting in early spring and the significant jump in values by June highlights the influence of the ground CO₂ reservoir within the limestone matrix.

Our data show an increasing trend in both atmospheric (outside the cave) and cave pCO₂ levels from 2021 to 2024. While tempting to strongly emphasize this observation, the uncertainties in measurements could potentially hinder such an approach. The time series is only three years long, but monitoring continues, and we are following this increase as it unfolds. Increasing average global air temperatures with prolonged warm seasons will reduce the number of days with an active convection cell, thus potentially reducing CO₂ export from caves via dynamic processes but increasing in-cave values (also aided by higher biological activity and the production of CO₂ in the soil). This dataset reflects processes in a specific cave; however, many caves with similar geometry exhibit comparable airflow patterns and CO₂ dynamics. In caves where other ventilation mechanisms, such as the chimney or barometric effect, are predominant, different behaviors may be observed. Generally, when external air flows through a karst massif, it mixes with “ground air” and exits enriched with CO₂. Thus, caves act as primary ventilation pathways, allowing some of the CO₂ from the karst massif to be released back to the surface. Additionally, CO₂-depleted air within caves promotes calcite precipitation, leading to CO₂ outgassing that can also return to the surface. While these mechanisms are well recognized, they remain somewhat poorly quantified, requiring long-term observations across diverse settings to refine the data. CO₂ from the vadose zone can be intercepted by groundwater and carried to karst springs, where it may be released into the atmosphere through direct degassing, tufa formation, or uptake by aquatic organisms. This segment of the carbon pathway is also insufficiently understood, highlighting the need for continued monitoring and modeling.

In conclusion, we hypothesize that caves developed in the vadose zone function as conveyor belts, moving CO₂ generated in soil by root respiration to the free atmosphere, potentially affecting our understanding of CO₂ dynamics and also possibly accounting for some missing sources [1]. As atmospheric CO₂ reaches a minima during the vegetation season (August in the Northern Hemisphere), the CO₂ generated in the soil is conveyed to the global atmosphere via caves. As karst areas cover a large fraction of the earth’s surface, our findings could help better understand the yet poorly accounted for CO₂ fluxes from the underground to the atmosphere. Similar studies should be deployed in caves located in karst areas located in different climates and with different hydrologic conditions to test our hypotheses described above.