CO₂ Dynamics and Transport Mechanisms Across Atmosphere–Soil–Cave Interfaces in Karst Critical Zones

Abstract

Cave systems serve as key interfaces connecting surface and underground carbon cycles, and research on their carbon dynamics provides a unique perspective for revealing the mechanisms of carbon transport and transformation in karst critical zones. In this study, we established a multi-factor monitoring framework spanning the atmosphere–soil–cave continuum and associated meteorological conditions, continuously recorded cave microclimate parameters (temperature, relative humidity, atmospheric pressure, and cave winds) and CO₂ concentrations across atmospheric–soil–cave interfaces, and employed stable carbon isotope (δ¹³C) tracing in Mahuang Cave, a typical karst cave in southwestern China, from 2019 to 2023. The results show that the seasonal amplitude of atmospheric CO₂ and its δ¹³C is small, while soil–cave CO₂ and δ¹³C fluctuate synchronously, exhibiting “high concentration-light isotope” signatures during the rainy season and the opposite pattern during the dry season. Cave CO₂ concentrations drop by about 29.8% every November. Soil CO₂ production rates are jointly controlled by soil temperature and volumetric water content, showing a threshold effect. The δ¹³C response exhibits nonlinear behavior due to the combined effects of land-use type, vegetation cover, and soil texture. Quantitative analysis establishes atmospheric CO₂ as the dominant source in cave systems (66%), significantly exceeding soil-derived contributions (34%). At diurnal, seasonal, and annual scales, carbon-source composition, temperature and precipitation patterns, ventilation effects, and cave structure interact to control the rhythmic dynamics and spatial gradients of cave microclimate, CO₂ levels, and δ¹³C signals. Our findings enhance the understanding of carbon transfer processes across the karst critical zone.

1. Introduction

The Earth’s critical zone refers to the interactive zone of surface layers (lithosphere–pedosphere–biosphere–hydrosphere–atmosphere), closely related to the human living environment, extending from tree canopies and soil to aquifers. It emphasizes the interactions among hydrological processes, geochemical processes, and biological processes. The Karst Critical Zone specifically denotes the Critical Zone formed on carbonate bedrock. It exhibits a spatially heterogeneous structure, characterized by soil at the surface and water below. The karst critical zone is a vital terrestrial ecosystem component. Its complex dual-permeability structure, with interconnected surface–subsurface flow paths, increases variability in carbon flux measurements. This complexity creates a natural laboratory for studying subsurface carbon spatiotemporal dynamics [1].

Cave systems function as key conduits for CO₂, making their migration dynamics a critical process in the carbon cycle of Karst Critical Zones. This process exhibits complex spatiotemporal dynamics and kinetic characteristics, driven by the synergistic interactions among the surface atmosphere, soil/lithosphere, and subsurface atmosphere. The surface atmosphere provides the driving force for soil CO₂ generation through temperature and precipitation. Fractures and pores within the soil/lithosphere not only serve as pathways for soil CO₂ diffusion but also facilitate water–rock interactions through their hydrogeological connectivity. This process carries dissolved CO₂ into cave systems while influencing cave humidity and temperature. Temperature and pressure gradients between the surface and subsurface atmospheres drive cave ventilation and air exchange, thereby regulating cave CO₂ concentrations [2,3,4]. Therefore, high-resolution tracking of CO₂ in space and time across the karst critical zone, paired with δ¹³C signatures, sharpens our view of carbon cycling and adds interpretive power to speleothem-based paleoclimate archives [5,6,7,8], to better reveal the mechanisms for transporting climate information [4,9].

As the chief engine of carbonate dissolution and reprecipitation, CO₂ governs how karst landscapes evolve and how they respond to shifting climate; its ambient level in air, soil pores, and cave chambers sets the pace and magnitude of these reactions [10]. The current global atmospheric CO₂ concentration is approximately 424 ppm and the δ¹³C values about −8‰ (https://gml.noaa.gov/, accessed on 5 March 2025). Global topsoil (0–20 cm) CO₂ concentrations typically range from 1000 to 10,000 ppm, whereas in deep karst soils (>50 cm), CO₂ levels can reach up to 50,000 ppm. Typically, soil δ¹³C-CO₂ values range from −26‰ to −12.5‰ [11]. Cave CO₂ concentrations are usually higher than those in the open atmosphere, ranging from near-atmospheric levels to 10,000 ppm, and are often subject to seasonal and interannual variability, with most caves showing similar characteristics of “high in summer and fall, low in winter and spring”, but this seasonal trend is more regionally expressed rather than globally universal [10,12,13,14,15,16,17,18], and abrupt changes also occur under heavy rainfall events [5,17]. Classical theories hold that soil respiration in the vadose zone constitutes the primary source of cave CO₂, thereby driving subsequent processes of carbon transport and sediment formation [19]. Soil CO₂ can enter caves in several ways: First, as a gas that migrates downward through pipe fissures in the surficial karst zone, or is drawn into caves as cave air outflow, leading to a decrease in cave air pressure [13,15]. Second, it is dissolved in seepage water to form carbonic acid, which is transported along bedrock fissures and enters the cave as drip or groundwater [12,17]. In addition, subsurface carbon sources, such as deep-rooted plant respiration [20], decomposition of organic matter in the deep vadose zone [21], and biotic respiration in bedrock fractures [22] also contribute significantly to cave CO₂. Cave ventilation controls both the outflow and removal of CO₂ [20,23,24]. In most cases, better-ventilated caves have lower CO₂ concentrations [25,26]. When cave ventilation is enhanced, the CO₂ concentration decreases, and the δ¹³C-CO₂ values may be light or heavy, depending on the δ¹³C value of the outside atmospheric air [27]. Cave depth, ventilation patterns, and hydrological conditions affect CO₂ mixing and retention times, which in turn modulate the spatial partitioning of δ¹³C [28]. Kinetic fractionation effects during cave microclimate and precipitation can also lead to δ¹³C excursions [7,15,17].

However, current studies mainly focus on short-term observations in limestone caves, while long-term monitoring of dolomite caves and the application of stable carbon isotope tracer techniques remain lacking. In addition, insufficient attention has been paid to the nonlinear responses and threshold effects arising from the synergy of soil temperature (Ts) and soil volumetric water content (VWC). Furthermore, the understanding of the soil–cave CO₂ transformation pathway, as well as the coupling mechanism between seasonal CO₂ dynamics and the carbon cycle driven by ventilation, remains inadequate. To bridge these knowledge gaps in underexplored aspects, this study presents a comprehensive, long-term monitoring program conducted in a dolomite cave (Mahuang Cave). The research is designed to contribute beyond existing studies by: (1) providing a globally representative long-term dataset from a dolomite system, capturing dynamics fundamentally distinct from those in limestone caves; (2) identifying nonlinear thresholds and synergistic effects of Ts and VWC to reveal variation patterns and driving mechanisms of soil CO2 concentrations; and (3) implementing a multi-scale coupling analysis that integrates stable carbon isotopes (δ13C) with concurrent cave microclimatic data to elucidate the soil–cave CO₂ transformation pathway and ventilation-driven coupling mechanisms.

In this study, we constructed a three-dimensional monitoring index system of “atmosphere-soil-cave”, and present the atmosphere–soil–cave air CO₂ and its δ¹³C value data from 2019 to 2023 in Mahuang Cave, including high-resolution (1 min interval) diurnal variations in August (warm season) and December (cold season) of 2019, as well as variations in the cave microclimate (temperature, relative humidity, atmospheric pressure, and cave winds) and meteorological environment outside the cave. This integrated set up enables the multi-scale analysis mentioned above. The study aimed (1) to reveal the dynamic change patterns of atmosphere–soil–cave air CO₂ and its δ¹³C and (2) to elucidate the sources and transport mechanisms of CO₂ across karst critical zone interfaces, with a specific focus on quantifying threshold behaviors, ventilation controls and source contribution.

2. Site Description

The study area is located in the Shuanghe Cave National Geopark (28°08′00″ N to 28°20′00″ N; 107°02′30″ E to 107°25′00″ E,), Wenquan Town, Suiyang County, Zunyi City, Guizhou Province, southwest China (Figure 1a). Shuanghe Cave is the largest cave complex in China, with a proven connecting length of 437.1 km (as of 2024), which is currently the longest dolomite cave in the world. The caves are mainly developed in the Cambrian middle-upper Loushanguan Formation (Є_2–3 ls) and Ordovician lower Tongzi Formation (O₁t) strata (Figure 1b), with the lithology dominated by dolomite and dolomitic chert, which is typical of dolomite worldwide. The region is characterized by a karstic low-middle mountainous landscape with elevation of 600–1700 m. Karst springs and underground rivers are widely developed in the region.

Mahuang Cave is the most important first-level branch cave in the cave system of Shuanghe Cave, and we chose it as the monitoring object (Figure 1c). It is a typical horizontal single-entrance natural cave, 1188 m long and entering at 720 m a.s.l., has had its portal shape, roof thickness, SW-trending development, passage zonation, morphology and sediment types fully described in our previous work [29]. The surface C3 natural vegetation of the cave area is dominated by the middle-subtropical evergreen and deciduous broad-leaved mixed forest, which is widely distributed, and the main C4 crop plants are corn and sorghum. The overlying land-use types and soil mechanical composition of Mahuang Cave are listed in

Table 1. Basic information on soil sample collection for different land-use types overlying Mahuang Cave [30].

No.	Type of Land-Use	Altitude (m)	Vegetation Coverage (%)	Vegetation Condition	Soil Mechanical Composition (%)
					Clay (<2 μm)	Silt (2~50 μm)	Sand (>50 μm)
S1	Cultivated land	962	43	Corn, Sorghum	7.57	84.01	8.41
S2	Shrubland	850	67	Shrub, Weed	10.59	84.31	5.1
S3	Shrub-grassland	760	56	Natural secondary shrub, Weed	10.05	83.17	6.78

.

3. Materials and Methods

3.1. Soil CO₂, Atmospheric CO₂ Monitoring, and δ¹³C Sample Collection

Three soil monitoring points were set up in different land-uses and vegetation types in the overlying area of Mahuang Cave (Figure 1c), and the vegetation and soil mechanical composition of each point are shown in Table 1. Three 3 cm diameter “L”-shaped PVC tubes were implanted at each point to monitor the air CO₂ concentration in the soil at depths of 20, 40, and 60 cm. The surface of the silicone tube was drilled with holes and wrapped with gauze to prevent soil blockage and ensure full contact with the soil. The end of the tube that protruded from the ground was sealed with an anticorrosive plastic plug to avoid atmospheric influence. The GT-903-CO₂ pump suction carbon dioxide detector (Shenzhen KORNO Electronic Technology Co., Ltd., Shenzhen, China) was used to manually read and determine the soil CO₂ concentration once a month with a measurement range of 50,000 ppmv and resolution of 1 ppmv, detection accuracy ≤ 300 ppmv, linearity error ≤ 100 ppmv, response time ≤ 20 s.

To reduce the influence of different soil occurrence layers on CO₂, an additional 40 cm deep PVC pipe was buried specifically to collect soil CO₂ samples and a portable gas sampling pump (Beijing Shi’An Technology Instrument Co., Ltd., Beijing, China) and Devex aluminum-plastic composite film were used to store approximately 1 L of soil CO₂ for δ¹³C analysis. The original residual gas in the gas sampling bag was removed using a gas pump to prevent contamination before sampling. Soil temperature (Ts) and volumetric water content (VWC) at the 5 cm depth were monitored using a W.E.T. Sensor (Model WET-2, Delta-T Devices Ltd., Burwell, Cambridge, UK), which provided measurements with an accuracy of ±3%. About 1 L of atmospheric CO₂ was stored in Devex aluminum-plastic composite film gas bags near the soil points for δ¹³C measurements. Measurements of atmospheric wind speed, temperature, relative humidity, air pressure, and altitude were conducted with a Kestrel-4500 portable weather station (Nielsen-Kellerman Company, Boothwyn, PA, USA). The instrument offers resolutions of 0.1 m/s, 0.1 °C, 0.1%, 0.1 kPa, and 1 m, and accuracies of ±3%, ±1.0 °C, ±3%, ±0.15 kPa (at 25 °C), and ±15 m, respectively. Recorded wind speeds were consistently close to 0 m/s and therefore considered negligible.

3.2. Delineation of Cave Section Zones, Cave CO₂ Monitoring, and δ¹³C Sample Collection

A total of 15 cave air CO₂ monitoring points were set up according to the developmental direction of Mahuang Cave, structure of the cave passage, ventilation, and sediment landscape (Figure 1d). Three cave section belts were divided according to the characteristics of the cave passage: Profiles #1–4 mark the near cave section, #5–11 the transition zone, and #12–15 the deep zone. Cave dimensions (length, width, and height) were measured using a Leica laser rangefinder (Leica DISTO, Leica Geosystems AG, Heerbrugg, Switzerland). The cave microclimate (temperature, relative humidity, barometric pressure, elevation, and cave wind speed) was monitored using a portable weather station, Kestrel-4500, Nielsen-Kellerman, Boothwyn, PA, USA.

Cave CO₂ measurements were made using a Telaire-7001 portable infrared CO₂ m (Telaire Corporation, Goleta, CA, USA) with an external HOBO data auto-recorder (Onset Computer Corporation, Cape Cod, MA, USA) with a resolution of 1 ppm. The measurement range was 0–10,000 ppm with a measurement accuracy of ±50 ppm. Standard (380 mg/L) gas was used for calibration prior to monitoring. To minimize the effect of human respiration on the cave air and ensure the accuracy of the measurements, monitoring proceeded sequentially from the exterior to the interior of the cave, and the instrument was operated at a distance of more than 3 m from people.

Cave air CO₂ was sampled at fixed points #4, #8 and #15, which was stored in Devex aluminum composite film air bags (~1 L) for δ¹³C analysis. To minimize the impact of human breath on cave air, the samplers held their breath and positioned themselves 2–3 m away from the air pump inlet while collecting samples inside the cave.

3.3. Data Collection Instructions

Manual monthly monitoring of indicators inside and outside Mahuang Cave was conducted from January 2019 to December 2023. To ensure comparability, measurements were consistently taken between the mid-to-late portion of each month, with site visits occurring from 10:30 to 15:30. Additionally, high-frequency automated monitoring at 1 min intervals was carried out during two distinct periods: 17–22 August 2019 (rainy season) and 28 December 2019–2 January 2020 (dry season). No rainfall was observed before or after either automated monitoring session. Due to the outbreak of the “COVID-19” coronavirus and the subsequent control measures implemented in 2020, monitoring could not be carried out for six months (specifically February and March 2020, February and October 2021, and March and September 2022) and the missing data were analyzed and mapped in Figure 2 (located in Section 4.1) and Figure 3g (located in Section 4.2.1) using the sample bar linkage processing. Carbon isotope data for individual months are missing because of instrument failure. Meteorological data, including temperature, precipitation, precipitation days, and relative humidity, were obtained on a monthly basis from the Suiyang County meteorological station via the Weather and Meteorological Service website (https://www.qweather.com/).

3.4. Carbon-Isotope Analysis/Date Calculations

(1) All atmospheric, soil, and cave air δ¹³C-CO₂ samples were measured at the State Key Laboratory of Environmental Geochemistry, Institute of Geochemistry, Chinese Academy of Sciences (CAS); the measurements were conducted using a MAT 252 Gas Isotope Ratio Mass Spectrometer (Finnigan MAT, San Jose, CA, USA). All results are reported as Vienna Pee Dee Belemnite (VPDB) given on a thousand-part standard, relative to the international standard. The δ¹³C values were calculated using the following formula [31]:

$$\delta ={\displaystyle \frac{{{\displaystyle R}}_{sample}-{{\displaystyle R}}_{standard}}{{{\displaystyle R}}_{standard}}}\times 1000‰$$

(1)

where δ¹³C denotes the abundance value (‰) of the isotope of the sample, R_sample denotes the isotope ratio of the sample, R_standard denotes the isotope ratio of the standard, and the deviation of the test results was better than ±0.2‰. Each sample was measured five times, and the results were averaged; parallel measurements were conducted for every set of four samples to guarantee an internal accuracy within ±0.1‰.

(2) Using the Keeling-plot approach and the stable carbon isotope composition (δ¹³C) of CO₂, the contributions of different CO₂ sources can be identified and quantified, the model is represented as [12]:

$${\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{cave} \mathrm{air}} = {\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{atm}}\times {\mathrm{F}}_{\mathrm{atm}}+ {\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{light}}\times \left(1-{\mathrm{F}}_{\mathrm{atm}}\right)$$

(2)

where ${\mathrm{F}}_{\mathrm{atm}}$ is the mole fraction of the CO₂ deriving from the atmospheric end member calculated as

$${\mathrm{F}}_{\mathrm{atm}}={\mathrm{X}}_{\mathrm{atm}}/{\mathrm{X}}_{\mathrm{soil} \mathrm{air}}$$

(3)

where X is the gaseous CO₂ concentration, with the light end member’s isotopic composition being the only unknown.

We merely assume that cave air is formed solely by the linear mixing of two end-members—soil and atmospheric CO₂, without considering the contributions of drip water degassing and aqueous CO₂ (aq) to the cave air. Mathematically, the mixing model can be derived as

$${\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{cave}}={ \mathrm{F}}_{\mathrm{soil}}\times {\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{soil}}+\left(1 -{\mathrm{F}}_{\mathrm{soil}}\right)\times {\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{atm}}$$

(4)

Contribution ratio formula:

$${\mathrm{F}}_{\mathrm{soil}}={\displaystyle \frac{{\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{cave}} - {\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{atm}}}{{\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{soil}} - {\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{atm}}}}$$

(5)

$${\mathrm{F}}_{\mathrm{atm}}=1-{\mathrm{F}}_{\mathrm{soil}}$$

(6)

where ${\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{cave}}$: Measured δ¹³C value of cave CO₂. ${ \mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{soil}}$: The isotopic composition (δ¹³C) of the soil CO₂ end-member, represented by the y-intercept (b) in the Keeling plot model [32], because CO₂ undergoes kinetic fractionation (+4.4‰) during diffusion from the soil into the cave air [33], the actual δ¹³C value of soil CO₂ must be obtained by correcting the Keeling plot intercept b, i.e., ${\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{soil}} = \mathrm{b} + 4‰$. ${\mathsf{\delta }}^{13}{\mathrm{C}}_{\mathrm{atm}}$: The δ¹³C value of atmospheric CO₂ (measured or background; this study uses the measured mean). ${\mathrm{F}}_{\mathrm{soil}}$: The fractional contribution of soil-derived CO₂ (0 ≤ ${\mathrm{F}}_{\mathrm{soil}}$ ≤ 1).

(3) In this study, virtual temperatures inside and outside the cave were used to estimate the cave air density and thus determine the air buoyancy of the cave, which in turn indicates the cave ventilation pattern [34]:

$$Tv=T\times (1+0.6079rv-0.3419rc)$$

(7)

where $T$ is the temperature (°C), $rv$ is the water vapor mixing ratio, and $rc$ is the carbon dioxide mixing ratio.

$Tv$ values were used in http://fisicaaplicada.ugr.es/pages/tv/!/download, accessed on 1 March 2025, provided in an Excel template for precise calculations [35]. The required input parameters included the temperature, relative humidity, CO₂ volume percentage, and atmospheric pressure. The temperature gradient between the cave and the external atmosphere governs their density contrast, serving as an indicator of airflow direction and a reflection of ventilation conditions. The virtual temperature difference $\Delta Tv$ between the inside and outside of the cave is calculated using the following equation:

$$\Delta Tv=T{v}_{\mathrm{int}}-T{v}_{\mathrm{ext}}$$

(8)

where when $\Delta Tv$ > 0, positive buoyancy arises from greater external air density, inducing density-driven inflow. When $\Delta Tv$ < 0, negative buoyancy occurs due to higher internal density, triggering outflow and inhibiting ventilation [36,37].

4. Results

4.1. Atmospheric Parameters and Cave Microclimate

The average monthly temperature in the area where Mahuang cave is located ranges from 4.8 to 27.4 °C, with a mean value of 15.74 °C. The temperature is highest in July and August and lowest in December and January (Figure 2d). The majority of the annual precipitation (76.26%) occurred between April and September, with an average of 12.55 d of precipitation per month (Figure 2e), with zero precipitation in August, November, December 2022, and January 2023. The annual mean atmospheric pressure was 934 Pa (Figure 2b), and the relative air humidity was 79% (Figure 2c). The atmospheric parameters (red dashed line) were closely aligned with the cave microclimate trends, reflecting a pronounced rain-heat synchronous pattern in the region.

The microclimate parameters of Mahuang Cave exhibit pronounced seasonal dynamics. Wind speeds were generally low but higher in summer and lower in winter (Figure 2a), with predominant outflow. Cave air pressure averaged 932 Pa—similar to external atmospheric pressure (934 Pa)—yet peaked in winter and dropped in summer, contrasting with seasonal variations in both cave and external temperature, and cave relative humidity (Figure 2b). Cave relative humidity averaged 93%, significantly exceeding the external mean of 79%; it remained high from spring to fall, often causing fog and wall condensation, while decreasing in winter (Figure 2c). Cave temperature, averaging 14.08 °C annually, followed external temperature trends (15.74 °C), rising in summer and fall and falling in winter and spring (Figure 2d).

Figure 2. Variation in microclimate parameters in Mahuang Cave (a–d) indicate cave wind speed, air pressure, relative humidity, and temperature; red dotted lines indicate atmospheric parameters outside the cave (wind speed is negligible), (e) precipitation and days (light cyan area indicates rainy season).

Figure 2. Variation in microclimate parameters in Mahuang Cave (a–d) indicate cave wind speed, air pressure, relative humidity, and temperature; red dotted lines indicate atmospheric parameters outside the cave (wind speed is negligible), (e) precipitation and days (light cyan area indicates rainy season).

4.2. Atmospheric, Soil, and Cave Air CO₂ Concentrations and Its δ¹³C Values

4.2.1. Atmospheric CO₂ Concentrations and Its δ¹³C Values

The atmospheric CO₂ concentration outside the Mahuang cave varied from 303 to 542 ppm, with an annual average of 365 ppm (Figure 3b). The atmospheric δ¹³C-CO₂ varied from −16.86 to −9.14‰, with a mean value of −10.88‰. The mean value of the four seasons was summer (−11.10‰) < spring (−11.06‰) < fall (−10.79‰) < winter (−10.52‰) and the overall variation was relatively minute (Figure 3b). During 2022, anomalously negative values were observed in July and August (−15.22‰, −16.86‰), which was presumed to be affected by the high temperature and low rainfall (26.5 °C and 71.3 mm rainfall in July; 27.4 °C and no rainfall in August). However, it cannot be ruled out that there might have been special activities in the area during the gas collection period that emitted anomalous gases, which requires further analysis.

Figure 3. Variations in CO₂ and δ¹³C values in the atmosphere–soil–cave system. (a) Monthly mean atmospheric temperature and precipitation. (b) Atmospheric CO₂ concentration and its δ¹³C values. (c) Soil temperature (T_S) and volumetric water content (VWC). (d) Soil CO₂ concentration (calculated from the mean values of S1, S2, and S3 at their respective 20–40–60 cm depths). (e) δ¹³C value of soil CO₂. (f) δ¹³C value of cave CO₂. (g) Cave CO₂ concentration (missing values are connected by spline). Light cyan indicates the rainy season.

Figure 3. Variations in CO₂ and δ¹³C values in the atmosphere–soil–cave system. (a) Monthly mean atmospheric temperature and precipitation. (b) Atmospheric CO₂ concentration and its δ¹³C values. (c) Soil temperature (T_S) and volumetric water content (VWC). (d) Soil CO₂ concentration (calculated from the mean values of S1, S2, and S3 at their respective 20–40–60 cm depths). (e) δ¹³C value of soil CO₂. (f) δ¹³C value of cave CO₂. (g) Cave CO₂ concentration (missing values are connected by spline). Light cyan indicates the rainy season.

4.2.2. Soil Temperature and Volumetric Water Content, CO₂ Concentrations and δ¹³C Values

Soil environmental parameters showed that soil temperature (2.67–33.43 °C) and volumetric water content (17.07–64.17%) (Figure 3c) had seasonal synchronized changes with regional air temperature and precipitation (Figure 3a). Notably, summer and fall rainfall from 2020 to 2021 was abundant and intensive, during which the soil volumetric water content continued to be high (>50%).

Soil CO₂ concentration monitoring data showed significant spatial and temporal heterogeneity (Figure 3d). Vertical distribution: The range of CO₂ concentrations in the 20–60 cm soil layer at each monitoring point was S1 (1324–32,471 ppm), S2 (753–32,083 ppm), and S3 (1236–50,000 ppm), respectively, with an increasing trend with increasing depth. The mean value of the land-use type differences was as follows: shrub-grassland (S3, 16,460 ppm) > cultivated land (S1, 12,120 ppm) > shrubland (S2, 10,568 ppm). Seasonal dynamics: the typical high rainy season (>10,000 ppm in summer) and low dry season (<2000 ppm in winter) were observed, with an extreme high value in June 2022 (precipitation of only 12.3 mm) and the lowest value in the continuous dry period from October 2022 to March 2022 (precipitation of almost zero).

The δ¹³C values of CO₂ at sites S1, S2, and S3 ranged from −25.57‰ to −14.68‰, −26.89‰ to −17.23‰, and −23.58‰ to −15.71‰, with mean values of −21.48‰ (S1, cultivated land), −23.87‰ (S2, shrubland), and −23.34‰ (S3, shrub-grassland), respectively, and the trend of the changes was generally consistent (Figure 3e). The seasonal averages were spring (−23.72‰) < fall (−22.71‰) ≈ summer (−22.66‰) < winter (−22.23‰). During the period of scarce rainfall from August 2022 to March 2023, the soil δ¹³C-CO₂ values were overall heavy.

4.2.3. Cave Air CO₂ Concentrations and Its δ¹³C Values

The cave CO₂ concentrations (318–982 ppm) had similar seasonal trends to soil CO₂, however, with significantly lower absolute values (Figure 3g). Their seasonal means were ordered as summer > fall > spring > winter, with two anomalies of concern: first, the concentration in April 2019 (863 ppm) was abnormally high, twice as high as the same period in 2020–2023 (417, 425, 413, and 382 ppm, respectively). The second was the large and sudden drop in concentration in November each year, with the monthly average dropping sharply from 563 ppm in October to 395 ppm in November. The mean concentration exhibited a slight increasing gradient, rising from the near cave section (532 ppm) through the transition zone (544 ppm) to the deep zone (548 ppm).

The δ¹³C-CO₂ variations in #4, #8, and #15 range from −18.12‰ to −8.32‰, −20.05‰ to −7.96‰, and −18.35‰ to −9.54‰, with mean values of −13.39‰, −13.38‰, and −13.63‰, respectively (Figure 3f). The δ¹³C-CO₂ exhibits a trend of enrichment toward the cave entrance: #4 ≈ #8 > #15. Except for the lightest value of #8 in April 2019 (−20.05‰), δ¹³C-CO₂ was gradually light from April, reached the lightest value in August, then gradually became heavier, and maintained the heavy value from December to February. The mean values for the four seasons were as follows: summer (−16.28‰) < spring (−13.31‰) ≈ fall (−13.22‰) < winter (−10.54‰).

4.3. Short-Term (Diurnal) Changes in Cave Air CO₂

Rainy-season records (Figure 4a) show evening CO₂ peaks at #2, #4, #9 (17:00–19:00) that taper overnight to dawn lows (05:00–06:00); amplitudes increase inward (687–796, 703–780, 728–869 ppm). Temperature follows the reverse trend—warmer toward the entrance (14.55–16.35 °C, 14.60–15.29 °C, 14.51–14.80 °C)—while relative humidity stays near saturation (95–100%, 96–100%, 99–100%).

Dry-season traces (Figure 4b) run markedly lower: CO₂ tops 385–432, 369–406, 368–404 ppm at #2, #4, #9, peaking 09:00–10:00 and dipping 15:00–16:00, with the entrance ward site richest. Temperature climbs inward 9.88–11.09 °C → 10.32–10.96 °C → 10.81–11.35 °C, mirrored by humidity 76–88% → 83–92% → 88–93%.

Diurnal CO₂ cycles in Mahuang Cave showed strong daytime maxima and nighttime minima during the rainy season, frequently interrupted by sharp midday perturbations. In contrast, dry-season levels remained stable with minimal variation. The system responded more rapidly to disturbances during warm periods, and fluctuations were most pronounced at sensors located near the cave entrance.

5. Discussion

5.1. Effects of Different Land Uses on Overlying Soil CO₂

5.1.1. Soil Temperature (Ts) and Volumetric Water Content (VWC) Affect Soil CO₂ Production

As key environmental drivers, Ts and VWC collectively govern soil CO₂ production and efflux, modulating both biotic and abiotic components of soil carbon cycling [38]. Overlying soil CO₂ concentrations in Mahuang Cave were generally higher in the rainy season and lower in the dry season (Figure 3d), which is similar to the observation of seasonal CO₂ changes in overlying soils of most caves [12,15,17,39,40]. We correlated the monthly average CO₂ concentration with Ts and VWC at different profile depths (20–40–60 cm) at three soil sites overlying Mahuang Cave and obtained the following patterns:

(I)

Relationship between Ts and CO₂ concentration (Figure 5a): The soil CO₂ concentration was low (<10,000 ppm) when the Ts was between 8 and 12 °C. In the range of 15–34 °C, the soil CO₂ concentration showed an overall increasing trend, especially reaching the highest value (>30,000 ppm) at approximately 26 °C. This is because Ts is one of the key factors influencing soil respiration rates. Low temperatures inhibit microbial activity, thereby reducing soil respiration rates and decreasing O₂ consumption and CO₂ release; conversely, optimal temperatures promote microbial activity, thereby increasing soil respiration rates. Studies have shown that the appropriate temperature for soil microbial activity is 25–37 °C, and temperatures that are extremely low or high inhibit microbial activity, leading to difficulties in organic matter decomposition and affecting the rate of soil CO₂ metabolism [41].

(II)

Relationship between Ts and CO₂ concentration in soil layers at different depths (Figure 5a): A significant positive correlation was observed between Ts and CO₂ concentration in the 20–40–60 cm soil profile, and the correlation slightly increased with increasing soil depth (R² = 0.20, 0.23, 0.30, respectively). Field observations showed that the mean values of soil CO₂ in the 20–40–60 cm profiles were 11,250, 13,803, and 14,115 ppm, respectively, which is consistent with the soil CO₂ gas diffusion model, that is, the soil CO₂ content gradually increased with increasing depth [42]. Surface Ts is affected by several factors and is more likely to change with atmospheric temperature than deeper soils, while deeper soils are more stable in temperature change relative to the surface layer owing to the combined effects of delayed heat transfer, permeability, biological activities, and own thermal insulation of soil. In addition, as soil depth increases, porosity is reduced, gas diffusion rates decline, and CO₂ concentrations become more stable.

(III)

Relationship between soil VWC and soil CO₂ concentration (Figure 5b): Soil CO₂ production began when the VWC approached 18% vol (minimum value < 2000 ppm). In the range of 37–58% vol, soil CO₂ was higher (maximum value > 30,000 ppm). After the VWC exceeded 60% vol, the soil CO₂ showed a decreasing trend. This indicates that VWC has an important effect on soil CO₂ concentration and that there is a threshold effect, this finding aligns with prior research examining the relationship between soil air CO₂ and VWC in the karst mountainous areas of Guizhou [43]. It has been shown that VWC is the main factor limiting soil microbial activity, and an appropriate VWC significantly enhances soil respiration. When the VWC is close to 18% vol, it starts to stimulate microbial activity in the soil and promotes soil CO₂ production [44,45]. A soil VWC of 25–45% vol is optimal for microbial activity, producing substantial amounts of highly concentrated soil CO₂ [46]. When the VWC is >60% vol (lower threshold for deep soil ~80% vol), soil permeability decreases, microbial activity diminishes, and aerobic microbial activity and soil respiration are inhibited, which reduces the amount of CO₂ produced and released, soil CO₂ fluxes decline rapidly as water impedes gas transport [47].

(IV)

Relationship between soil VWC and CO₂ concentration in soil layers at different depths (Figure 5b): CO₂ concentration in the 20–40–60 cm soil profile was positively correlated with soil VWC; however, the correlation gradually weakened with increasing depth (R² = 0.25, 0.17, 0.05, respectively). The presence of higher CO₂ values in deeper soils (40 cm and 60 cm) despite high VWC indicates a complex relationship that may be related to seasonal dryness and humidity, rainfall events, and diffusion coefficients. During the high temperature and rainy season, plant growth is vigorous, soil organic matter and apoplastic materials decompose rapidly, microorganisms and plant roots are active, and CO₂ production is large and easily accumulates in the soil, especially during the plant growing season, when the soil CO₂ concentration increases by 5–10% [48]. The VWC of the deep soil layer is generally higher than that of the surface layer, and the dissolved CO₂ in the surface soil solution is vertically transported downward, resulting in a higher accumulation of CO₂ in the deep soil layer than in the surface layer [49]. The seasonal variability of CO₂ concentration in the topsoil layer is large, whereas seasonal variability of CO₂ concentration in the subsoil layer (e.g., below the clay layer) is relatively less and sometimes insignificant [50]. This is because topsoil has a small capacity and well-developed pores, and CO₂ released by microbes and root respiration can diffuse and escape rapidly, whereas subsoil has a large capacity and small porosity, and CO₂ diffusion is slow due to the limitation of the clay layer [51]. Precipitation can directly increase the concentration of CO₂ in the topsoil layer, while simultaneously reducing the soil’s gas diffusion coefficient. After a rainfall event, the topsoil pores become saturated with water, decreasing the effective porosity available for gas diffusion. When VWC exceeds 40% vol, the diffusion coefficient declines sharply, causing CO₂ to accumulate [52].

Figure 5. Overlying Ts (a) and VWC (b) versus 20–40–60 cm soil CO₂ concentration in Mahuang Cave.

Figure 5. Overlying Ts (a) and VWC (b) versus 20–40–60 cm soil CO₂ concentration in Mahuang Cave.

5.1.2. Soil CO₂ Concentration and Its δ¹³C Under Different Land-Use Types

The stable isotope composition of carbon is an important tracer for exploring the source of CO₂ [53]. Based on this, one hypotheses can be proposed: If soil respiration (including plant root respiration and microbial respiration) is one of the main sources of soil CO₂, then CO₂ produced by soil respiration will be enriched in ¹²C compared to atmospheric CO₂, with a relatively low ¹³C content. Therefore, the δ¹³C is negatively correlated with CO₂ concentration; that is, the higher the CO₂ concentration, the lower (more negative) the δ¹³C value.

Further analysis of the linear regression results between soil CO₂ concentration and its corresponding δ¹³C value showed (Figure 6) a significant linear negative correlation (p < 0.05) between soil δ¹³C value and CO_2. Among them, the most significant correlation was found in S2 (R² = 0.55), and the regression equation was y = −2.14x − 21.19, indicating that the higher the soil CO₂ concentration, the smaller (more negative) the δ¹³C value at S2. However, despite the statistically significant linear relationship between S1 and S3 (p < 0.05), the amount of variance explained by the model was limited (R² values of 0.02 and 0.05, respectively). This shows that the relationship between soil CO₂ concentration and δ¹³C values is not a simple linear relationship but is also influenced by other complex factors.

The mean value of soil CO₂ in the 40 cm profile of Mahuang Cave was S3 (shrub-grassland, 16,562 ppm) > S1 (cultivated land, 12,478 ppm) > S2 (shrubland, 12,368 ppm), and the mean value of δ¹³C was S1 (−21.48‰) > S3 (−23.33‰) > S2 (−23.87‰) (Figure 3d,e). It was hypothesized that this might be caused by factors, such as land-use mode, vegetation type, soil aeration, and organic matter content. S1 (cultivated land): fertilization measures increased soil organic matter content, while human seasonal tillage changed the soil structure and destroyed soil aeration, so that the CO₂ produced was mixed more with atmospheric exchanges, which resulted in a relatively heavier soil δ¹³C value. S3 (shrub-grassland): the root system is densely developed, respiration is strong, anthropogenic interference is low, the deep soil is compact, and CO₂ is not easily emitted. Monitoring showed that the S3 site had a multi-year Ts of 18.7 °C and VWC of 47%, and the high humidity may reduce soil aeration and impede CO₂ diffusion, causing it to accumulate to form a high concentration of CO₂ [54]. S2 (shrubland): underground root biomass is low, and CO₂ produced by respiration is low. Coupled with the loose surface soil, CO₂ exchanges with the atmosphere more rapidly, resulting in a surface layer of CO₂ Loss of sink. In terms of soil mechanical composition (Table 1), S2 was dominated by clay and chalk grains (combined 94.9% weight), with few sand grains, and was a clay loam, the CO₂ emission rate of which, was usually significantly lower than that of sandy loam [55], so that the CO₂ concentration of S2 soil was lower than that of S1 and S3.

Compared to the typical limestone cave—Liangfeng Cave—the mean atmospheric CO₂ concentration in the area was 456 ppm, with δ¹³C-CO₂ averaging −11.1‰. Soil CO₂ concentrations at 30 cm and 60 cm ranged from 3288 to 5774 ppm and 4856 to 8384 ppm, respectively [56]. The overlying vegetation is dominated by C3 shrubland, whose soil δ¹³C-CO₂ averages −22.42‰ at 30 cm and 60 cm depths [57]. As illustrated in Figure 3, although both Mahuang Cave (dolomite) and Liangfeng Cave share common features such as increasing soil CO₂ concentration with depth and higher values during the rainy season compared to the dry season, yet Mahuang Cave displays significantly greater soil CO₂ concentrations and slightly lighter δ¹³C-CO₂ values. These differences reflect the combined effects of soil physicochemical properties, vegetation composition, and land-use practices atop contrasting lithologies, it indicates that soil CO₂ distribution in karst regions exhibits both overall consistency and regional variability.

Overall, soil CO₂ production was predominantly controlled by the synergistic interaction of Ts and VWC, both exhibiting distinct threshold effects. Furthermore, the relationship between soil CO₂ concentration and its δ¹³C signature was not a simple linear function, but rather a nonlinear coupling modulated by multiple factors including land-use patterns, vegetation types, and soil texture characteristics.

5.2. Carbon Sources and Dynamic Transformation Mechanisms of Cave CO₂

5.2.1. Cave CO₂ Source Analysis

Sources of cave CO₂ generally include (1) anthropogenic fluxes, which is, airflow from human respiration and movement [23] and (2) natural fluxes, external atmospheric CO₂ inputs from ventilation effects [58], CO₂ directly diffused from surface karst and overlying soil or water-mediated into the cave [5,13,14], soil respiration [59], indirect CO₂ fluxes from degassing of underground rivers and drips, microbial decay, or endogenous processes of organic matter in cave sediments [60], and deep geologic forcing (e.g., CO₂ release from geothermal activity on fault fractures) [7,61]. We did not consider anthropogenic fluxes or deep geothermal sources in this study, as Mahuang Cave is a natural karst formation without detectable fault structures.

To quantify the contribution of each factor to cave CO₂, the variables selected for analysis included monthly measurements of external atmospheric parameters (CO₂, temperature, and precipitation), soil CO₂ concentration, and cave microclimate conditions (temperature, relative humidity, barometric pressure, and wind speed). The Kaiser–Meyer–Olkin (KMO) value of the selected sample data was 0.805 (>0.50) and passed Bartlett’s test of sphericity (p < 0.05), indicating that the PCA model was statistically significant and well suited for PCA. In the PCA plot, the arrows indicate the loadings of each variable in the principal component space, the direction of the arrow indicates the relationship between the variable and the principal component, and the length of the arrow indicates the magnitude of the variable’s contribution to the principal component. The longer the arrow, the greater the value of the variable’s loading on the principal component and the stronger the correlation with it. The angular separation between arrows reflects the strength of the correlation between variables: an acute angle indicates a strong positive correlation, while variables pointing in opposite directions are negatively correlated [62].

Outcomes appear in Figure 7, where the eigenvalues of the first two principal components are greater than one. The first principal component (PC1) explained 53.8% of the variance, and the second principal component (PC2) explained 15.6% of the variance, with a cumulative contribution of 69.4%. (i) Quadrant 1: Atmospheric precipitation, cave relative humidity, and soil CO₂ had positive loadings on PC1 and PC2, indicating that they were positively correlated with these two principal components are positively correlated. These variables exhibit relatively high loadings, indicating to a certain extent that they are the primary factors jointly influencing the variations in cave CO₂ concentrations. (ii) Quadrant 2: Atmospheric CO₂ has high positive loadings on both PC1 and PC2 and a strong positive correlation with the principal components, and it is one of the factors contributing the most to cave CO₂. Cave air pressure had loadings close to zero on both PC1 and PC2, suggesting a small contribution. (iii) Quadrant 4: Atmospheric temperature, cave temperature, and cave wind speed exhibit positive loadings on PC1 and negative loadings on PC2. This indicates a positive/negative correlation and the relative importance of these variables in changes in cave CO₂ concentration. For instance, increased cave wind speed may enhance cave ventilation, thereby reducing cave CO₂ concentration (negatively correlated with PC2). Increases in atmospheric and cave temperatures may enhance biological processes, such as soil respiration and cave microbial activity, thereby increasing cave CO₂ concentrations (positively correlated with PC1). The angles of atmospheric and cave temperatures were highly overlapping, suggesting a strong correlation and that changes in cave temperatures may lag behind changes in atmospheric temperatures by a small amount.

5.2.2. Soil-Atmosphere Driven CO₂ Transformation in Caves

In fact, the CO₂ concentration in Mahuang Cave (318–982 ppm) is lower than that in some typical limestone caves with significant seasonal ventilation, such as Liangfeng Cave (417–707 ppm; [56]); Bunker-Emst Cave (408–811 ppm; [63]) and Obir Cave (400–1500 ppm; [12]), Blue Spring Cave (400–1600 ppm; [6]), Cueva Larga Cave (600–1800 ppm; [36]), and St Michael’s Cave (600–8000 ppm; [21]) The mean air δ¹³C-CO₂ value of Mahuang Cave (−13.45‰) was similar to those of St Michael’s Cave (−13‰) and Liangfeng Cave (−12.86‰). Soil, cave CO₂, and δ¹³C values in Mahuang Cave were characterized by high values in the rainy season and low values in the dry season (Figure 3f), which, combined with the results of the PCA (Figure 7), suggests that cave δ¹³C during the rainy season may have inherited the seasonal signal of soil δ¹³C, whereas the winter δ¹³C bias may have been driven by cave ventilation mixing. Therefore, we hypothesized that the seasonal variation in the cave δ¹³C signal is a result of the interaction between soil dynamics and cave ventilation.

For the validation of this hypothesis, the Keeling Plot End-member Model was applied to analyze how CO₂ is produced and diffused in both the soil and surface karst zones. The Keeling Plot [64] is based on a simplified two-end-element mixing model that assumes that cave CO₂ concentrations and isotope ratios are the result of the proportional mixing of the background atmosphere with CO₂-rich, light-isotope soil end elements (Equations (2)–(6)). The model captures two key parameters: (1) the intercept value of the linear regression line on the δ¹³C-CO₂ axis, which represents the isotopic signature of the end elements of soil-sourced CO₂. (2) The location of the data points relative to the two end elements, which quantify the proportionate contribution of atmospheric CO₂ to cave air [12,32].

Figure 8 illustrates the surface atmospheric, soil, and cave CO₂ and their δ¹³C compositions in the Mahuang Cave. (i) Linear fitting of the discrete sampling data shows strong robustness (R² = 0.89). The y-axis intercept of −22.93‰ yields an actual δ¹³C value of −18.53‰ for soil-respired CO₂ after applying the +4.4‰ kinetic fractionation correction (Equation (6)). This value is significantly enriched (less negative) compared to the characteristic range for C3-vegetation-dominated soil respiration (−25‰to −27.7‰; [7,33]. isotopic enrichment suggests that cave air CO₂ is not primarily derived from C3 vegetation respiration, but rather reflects mixing with additional carbon sources, potentially including C4 vegetation inputs, atmospheric CO₂ contributions, or geological carbon fluxes, primarily referring to the CO₂ degassing from drip water. The study site does not exhibit significant gaseous fluxes from hydrothermal or magmatic processes. (ii) The spatial distribution of the data points revealed that the warm season data were closer to the soil end-members, reflecting the predominance of biogenic CO₂ and confirming that cave CO₂ mainly originates from root and microbial respiration in the soil above the cave [7,65]. The isotopic composition of cave CO₂ in winter can be up to 5.9‰ heavier than that in summer, and the cave δ¹³C data are shifted toward the atmospheric end-elements and even partially overlap, indicating changes in mixing ratios due to enhanced ventilation, confirming that the δ¹³C of cave air in the cooler season may be strongly influenced by cave ventilation [4]. Equation-derived estimates indicate 34% soil-derived and 66% atmospheric CO₂ contributions (Equations (2)–(6)).

In summary, integrated analyses using PCA, Keeling plots, and binary mixing models demonstrate that CO₂ in Mahuang Cave is predominantly atmospheric in origin (~66%), with soil respiration representing a secondary source (~34%).

5.3. Dynamics and Transport Mechanisms of Carbon in Karst Critical Zones

5.3.1. Controlling Factors of Carbon Dynamics in Soil–Cave Systems

Annually, the cave functions in a respiratory manner, accumulating and storing CO₂ in the warmer months before releasing it to the exterior environment through enhanced ventilation in the cooler seasons [66]. The exchange between caves and the external atmosphere is a complex and dynamic process influenced by internal biogeochemical processes [67], cave ventilation conditions [8,24], seasonal climatic variations [66], long-term geological and environmental changes [68], and other factors. As Mahuang Cave is a natural cave, this study does not consider anthropogenic factors and focuses on analyzing the controlling factors of CO₂ dynamics in soil–cave systems, specifically: (1) carbon-sources (atmosphere/soil), (2) temperature-precipitation, and (3) ventilation effects.

(1) Carbon-source

By calculating the correlation heat map between the monitored parameters in Mahuang Cave (Figure 9), it was found that soil P_SCO₂ had a significant positive correlation with cave P_CCO₂ (r = 0.61, p < 0.05), indicating that soil CO₂ has a significant regulatory effect on cave CO₂ concentrations [20]. Meanwhile, cave δ¹³C-CO₂ was significantly negatively correlated with both cave P_CCO₂ and soil P_S CO₂ (r = 0.87, p < 0.05; r = 0.63, p < 0.05; Figure 9). Specifically, the cave interior and overlying soil change synchronously: rainy season (high CO₂ concentration, light δ¹³C), dry season (low CO₂ concentration, heavy δ¹³C). In summer, cave δ¹³C (mean − 16.28‰) was relatively close to soil δ¹³C (mean − 22.66‰); in winter, cave δ¹³C (mean − 10.54‰) was heavier than soil δ¹³C (mean − 22.23‰) and very close to the atmospheric δ¹³C (mean − 10.52‰) outside the cave (Figure 3b). This finding further confirms the dual atmosphere-soil drivers of cave CO₂ transformation.

(2) Temperature-precipitation

Correlation analysis (Figure 9) showed that P_CCO₂ in Mahuang Cave was significantly and positively correlated with the atmospheric temperature outside the cave (r = 0.76, p < 0.05), suggesting that the air density difference or barometric pressure gradient caused by the thermal gradient between the subterranean and external environments may influence the seasonal changes in cave P_CCO₂. Atmospheric precipitation was significantly and positively correlated with both soil P_SCO₂ and cave P_CCO₂ (r = 0.45, p < 0.05; r = 0.56, p < 0.05). The possible mechanisms are as follows: on the one hand, high rainfall in the warm and humid seasons enhances surface evaporation and both plant root respiration and microbial respiration, generating a high concentration of soil CO₂. In contrast, when karst water drips from the cave roof or seeps out of the fissure pores under capillary force, CO₂ is released by degassing due to changes in pressure and temperature inside the cave, and partially escapes and accumulates in the cave [28,67]. During periods of heavy rainfall or storms, water, clay, silt, and a small amount of calcareous cement fill the pore space, and cave gases are retained for a longer period of time without being easily degassed, resulting in an increase in cave P_CCO₂ in the warm season [69].

The overall trend of changes in Mahuang Cave was similar during the monitoring period from 2019 to 2023, but there were three interesting abrupt change phenomena.

(I)

A significant abrupt decrease in cave CO₂ levels was observed in November: The cave CO₂ concentration showed a relative decrease of 29.8% from October (563 ppm) to November (395 ppm), which is close to the atmospheric CO₂ level outside the cave (winter average value of 409 ppm) (Figure 3g). This was attributed to the decrease in air and soil temperatures outside the cave in November, weakening of soil respiration, decrease in CO₂ release, decrease in precipitation and soil moisture, increase in diffusion capacity of CO₂ in the soil, and escape of more CO₂ to the atmosphere instead of through the fissures into the cave [70]. From October to November, the average virtual temperature difference increased from 0.1 to 3.6 °C (Figure 10), enhanced ventilation introduced cold external air, which displaced and diluted the high CO₂ concentration in the cave, resulting in a rapid decline.

(II)

Abnormally high cave CO₂ values (negative δ¹³C) in April 2019: The average cave CO₂ value of all monitoring sites in April 2019 was 863 ppm, which was almost twice the value of the same period from 2020 to 2023 (417, 425, 413, and 382 ppm, respectively) (Figure 3g). Correspondingly, the average value of soil CO₂ (S, 40 cm) was as high as 18,231 ppm in April 2019, while it was only 12,123 ppm in the same period of the other four years (Figure 3d). This was due to the precipitation bias of 154.4 mm (cave δ¹³C-CO₂ value of −18.38‰) in April 2019, compared to the average precipitation value of 65.9 mm (cave δ¹³C-CO₂ value of −12.59‰) in the same period from 2020 to 2023 (Figure 3a,f). This shows that precipitation is one of the main factors affecting soil CO₂ yield, cave CO₂, and its δ¹³C changes on the interannual scale [71], and a study in Furong Cave, China, also found that precipitation affects soil CO₂ more than temperature on the interannual scale [17].

(III)

Anomalously heavy δ¹³C values during the period of sparse precipitation: The trends of cave δ¹³C-CO₂ values and soil δ¹³C-CO₂ values were generally consistent, but during the period of extremely sparse (almost 0 mm) precipitation from October 2022 to March 2023, the δ¹³C-CO₂ values of the two values were significantly anomalous compared with the values for the same period in other years (with roughly comparable temperature conditions) (Figure 3e,f). For example, the mean winter soil δ¹³C-CO₂ value during this period was −19.76‰, compared to the mean value of −22.67‰ for the same period in other years. This suggests that on an interannual scale, cave CO₂ and its δ¹³C-CO₂ values may inherit the characteristic signals of overlying soil CO₂ and respond sensitively to changes in local atmospheric temperature and precipitation [17].

(3) Ventilation effects

The virtual temperature difference ΔTv (Equations (7) and (8)) between each monitoring point in the cave and the outside of the cave was calculated on a month-by-month basis, and the results are shown in Figure 10.

When ΔTv > 0, the air buoyancy difference is positive, which mainly occurs in winter (December, January, and February) and peaks in December 2020 (11.22 °C). At this time, the cave maintains a higher internal temperature (T_in > T_out) and lower air density (ρ_in < ρ_out) compared to the external environment, resulting in a state of positive ventilation that facilitates active air exchange [72]. The inflow of cold, dense external air through the lower entrance drives a ventilation process: it displaces the warm, humid cave air, forcing it out through the upper reaches and thereby reducing the cave CO₂ concentration (Figure 3g). Mixing of cave CO₂ with atmospheric CO₂ occurred, which shifted the cave δ¹³C toward the atmospheric δ¹³C end element, with heavy δ¹³C values (Figure 8).

When ΔTv < 0, the air buoyancy difference is negative, especially in summer (June, July, and August), where the temperature difference is large, and in July 2020, it reaches the valley value (−14.52 °C). At this time, T_in < T_out, ρ_in > ρ_out, the cold airflow inside the cave partially flows out from the bottom of the cave, and the warm and hot air outside the cave cools and sinks after flowing along the top of the cave and stagnating in the cave. At the same time, high humidity in summer decreases air density, increases viscosity, deteriorates mobility, slows down or even stagnates [73], inhibits airflow exchange, and creates a restrictive ventilation pattern. Combined with high soil CO₂ inputs and cave air accumulation, warm-season caves have high CO₂ concentrations and low δ¹³C values (Figure 3f).

When ΔTv ≈ 0, the virtual temperature difference fluctuates above and below 0 °C in spring (March, April) and autumn (October, November). The air density inside and outside the cave does not differ much, the airflow exchange is not obvious, and it is in the transition stage between active and restrictive ventilation, which is similar to the phenomenon observed in the Shihua Cave in Beijing [74].

The virtual temperature difference in the cave also has spatial variation, and its extreme value increases with distance from the cave entrance (Figure 10). The virtual temperature difference ranged from −11.56 to 6.45 °C in the near cave section, −12.98 to 7.82 °C in the transition zone, and −12.63 to 10.35 °C in the deep zone. Studies have shown that there is strong ventilation and frequent exchanges with the atmosphere in the vicinity of cave entrances [75]. Based on the isotopic balance theory, the spatial expression of air δ¹³C values in Mahuang Cave is #4 ≈ #8 > #15 (Figure 3f).

Short-term (diurnal) monitoring (Figure 4a,b) reveals that summer CO₂ contrasts shrink from #2 to #4 to #9, whereas winter shows the reverse rank, underscoring the hypersensitivity of the warmer months and the leading role of the entrance sector. This seasonal pattern mirrors the larger warm-season cave–air thermal gap (Figure 2d) and the corresponding ΔTv deficit (Figure 10); sharp transients intensify toward the portal, confirming heat advection as the prime driver of cave-air temperature [29].

Figure 10. Variation in the virtual temperature difference at the Mahuang Cave monitoring site. Red upconvex ΔTv > 0; cyan downconvex ΔTv < 0.

Figure 10. Variation in the virtual temperature difference at the Mahuang Cave monitoring site. Red upconvex ΔTv > 0; cyan downconvex ΔTv < 0.

5.3.2. Spatial Variations of Carbon Isotope Transport in Karst Critical Zones

The magnitude of air CO₂ content in the karst system of Mahuang Cave is soil CO₂ > cave CO₂ > atmospheric CO₂. The soil CO₂ content was 26 times higher than the cave CO₂ content and 38 times higher than the atmospheric CO₂ content. δ¹³C-CO₂ values are atmospheric δ¹³C > cave δ¹³C > soil δ¹³C (Figure 11).

The special conditions and stability of caves as a unique ecosystem not only provide important clues for studying the history of caves but also serve as indicators for documenting local changes in climate and vegetation [76]. The transport of CO₂ and its δ¹³C signal in karst critical zone occurs as follows: first, CO₂ in the air dissolves in rainwater to form carbonate; subsequently, rainwater passes through the vegetation and soil layers, absorbing more soil CO₂, forming soil water, and lowering the pH to promote carbonate dissolution; next, soil water carries dissolved CO₂ and carbonate to infiltrate into cave roofs to form dripping water; and finally, owing to the low PCO₂ value, the CO₂ in the dripping water degasses and escapes, leading to calcium carbonate precipitation and the formation of cave sediments, which completes the transmission and recording of δ¹³C signals [19].

The dissolution of carbonate rocks is governed by multiple factors, including water pH, temperature, flow velocity, path length, contact area, residence time, and the degree of fracturing within the rock formation. These factors collectively control the intensity of water–rock interactions, thereby determining the proportion of carbon incorporated into DIC from carbonate dissolution (typically exhibiting δ¹³C values close to 0‰). This process causes the δ¹³C of DIC to shift toward more positive values relative to the initial soil CO₂ derived from C3 vegetation (δ¹³C ≈ −25‰ to −27.7‰; [33]). The degree of CO₂ outgassing from drip water is modulated by cave air pCO₂, drip surface area, flow rate, and ventilation intensity. Degassing is an isotopic fractionation process in which the lighter ¹²C isotope preferentially escapes from DIC as CO₂, leaving the remaining liquid phase (which ultimately forms calcite deposits) with a heavier δ¹³C value. Conversely, the released CO₂ exhibits a lighter δ¹³C value than its source DIC. Consequently, the δ¹³C value of cave-air CO₂ can be regarded as a composite signal reflecting both mixing and fractionation processes: it inherits the initial value from soil CO₂, undergoes a “heavy-fractionation” effect through water–rock interactions, and finally experiences a “light-fractionation” filtering effect during drip degassing.

The Mahuang Cave area is primarily composed of dolomite and dolomitic limestone, which are characterized by slightly slower dissolution rates than calcite under usual conditions. Their contribution to the dissolved inorganic carbon (DIC) pool is likely minor relative to that of soil CO₂. However, this does not imply that their isotopic effects can be disregarded, particularly when compared to atmospheric and soil CO₂ sources. Given the thick overlying bedrock at Mahuang Cave, the complex water flow pathways, and the difficulty in precisely constraining water residence times, the cumulative effects of water–rock interactions and their potential contribution to drip water degassing processes warrant careful evaluation. These aspects require further investigation in future studies. Systematic monitoring of cation concentrations (e.g., Mg²⁺/Ca²⁺ ratios), DIC concentrations, δ¹³C values, CO₂ partial pressure, drip surface area, and drip rates at different drip sites, combined with high-resolution isotopic records from cave sediments (e.g., stalagmites), will help comprehensively elucidate the controlling factors of cave air CO₂ and its δ¹³C composition. Such efforts will allow precise quantification of the contributions of dolomite dissolution and degassing to the cave carbon cycle, thereby improving the reliability of stalagmite isotopes as paleoclimate proxies. These objectives will be the focus of our future high-frequency monitoring and systematic research.

In this study, we focused only on the variations, sources, and influence mechanisms of CO₂ and its δ¹³C in the atmosphere, soil, and cave air in Mahuang Cave. However, the linkages between the CO₂ contents of various gases, cave water bodies (including drips, pools, and underground rivers), and cave sediments in karst critical zone are extremely complex. The migration and transfer process of δ¹³C signals between them, as well as the scientific issues of the process information recorded by these indicators and the response of regional hydrological conditions and climate assessment, still need to be further explored in depth in future cave monitoring.

6. Conclusions

Based on a five-year monitoring campaign (2019–2023) at Mahuang Cave—a representative dolomite cave in SW China—the main findings of this study are summarized as follows:

(1) Cave passage geometry governs the spatial heterogeneity of microclimate parameters by regulating air circulation. High-resolution monitoring revealed that cave microclimate and CO₂ exhibit pronounced day–night fluctuations and marked spatial gradients from the entrance to the deep zone.

(2) Soil CO₂ production shows a threshold effect and land-use dependence. CO₂ production reaches peak productivity at a soil temperature of approximately 26 °C and a volumetric water content of 37–58% vol. with significant variations across land-use types (shrub-grassland > cultivated land > shrubland). The nonlinear relationship between CO₂ concentration and δ¹³C indicates complex microbial regulation by environmental factors.

(3) The Keeling model—a dual-source CO₂ mixing method—reveals climate-sensitive seasonal patterns. Quantitative analysis shows that cave air derives from atmospheric (66%) and soil-respired (34%) sources, with seasonal variations in CO₂-δ¹³C signatures indicating ventilation changes. The synchronous response of CO₂ and δ¹³C to abrupt temperature/precipitation shifts suggests their potential as climate indicators.

(4) Carbon transmission mechanisms require further high-resolution investigation. While this study establishes a preliminary framework for carbon transmission in dolomite caves, the precise mechanisms governing δ¹³C signal migration across the atmosphere–vegetation–soil–cave interfaces remain inadequately constrained, particularly in quantifying the effects of water–rock interactions and drip water degassing processes on carbon isotopic composition under complex hydrogeological conditions, and represent a critical direction for further study.

In summary, this study has preliminarily revealed the transmission and transformation mechanisms of carbon in dolomite caves, providing important modern-process evidence and monitoring paradigms for the carbon cycle in the karst critical zone and paleoclimate reconstruction.