SEARCH

SEARCH BY CITATION

Keywords:

  • carbon dioxide;
  • radon;
  • cave air degassing;
  • entropy of curves;
  • microclimatic monitoring;
  • karst

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Variations of carbon dioxide and radon content of cave air are presented as key parameters to assess the outgassing and isolation processes of a subterranean atmosphere. An exhaustive monitoring in the Castañar cave determined the temporal evolution of CO2 and 222Rn levels over a 12-month period, in order to characterise the mechanisms of these microclimatic processes. Concentrations of both gases show both seasonal variations and short-term fluctuations depending on several climatic factors: the air temperature difference between cave and exterior, cave air pressure, rainfall and anthropic factors including visits and duration of door opening. Over the course of an annual cycle, a cause-effect analysis has been conducted by stationary clustering of time series in terms of entropy of curves. Two opposing patterns of cave microclimate have been distinguished: (1) storage of trace gases in the cave reservoir during the cold-wet season, and (2) CO2 emissions during warm-dry season. The partial water filling of the porous system and fissures of the membranes covering the cave (host rock and soil) is determined by the external relative humidity (controlled by the external air temperature) as well as by rainfalls, which play a key role in confining the cave atmosphere. Copyright © 2009 Royal Meteorological Society


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Carbon dioxide is a key component controlling carbonate karst processes like limestone dissolution and speleothem growth, which are controlled by differences in CO2 partial pressure between the water of the upper epikarst and the cave air. As an example, dramatic changes in the pCO2 and δ13C can be detected in water and cave air, in response to seasonal changes in ventilation, encouraging flushing of the CO2 inside the cave by relatively CO2-poor air during the cold season and a high degree of calcite saturation of drip water (Spölt et al., 2005). Understanding CO2 distribution and dynamics in caves is important for paleoclimatic research using stalagmites because their growth rates partially depend on cave atmosphere pCO2 (Kaufmann, 2003; Spötl et al., 2005; Baldini et al., 2008).

Cave CO2 levels show significant seasonal and spatial variations (Ek and Gewelt, 1985; Bourges et al., 2001; Baldini et al., 2006; Fernandez-Cortes et al., 2006b; Batiot-Guilhe et al., 2007). Evolution of air CO2 concentration within a cave depends on a balance of fluxes between (Faimon et al., 2006) (1) CO2 inputs (diffusion of CO2 from epikarst, degassing of dripwaters, cave organic matter decomposition, anthropogenic flux by visitors breathing and, in some cases, by endogenous processes such as volcanism or magmatism), and (2) CO2 outputs controlled by ventilation, depending on cave geometry and microclimatic relationships between cave and outer atmosphere (air temperature and barometric pressure). Recent studies have demonstrated that for a temperate pasture overlying an accessible cave (Altamira cave, northern Spain), afternoon CO2 emissions in summer are likewise inexplicable in a biological context (light-dependent photosynthesis versus temperature-dependent respiration), but coincide with periods when outgassing is observed inside the cave (Kowalski et al., 2008). However, these studies fail to explain the controlling mechanisms. Understanding the physical processes and mechanisms controlling these anomalous CO2 fluxes from the caves is one of the ultimate challenges for research in micrometeorology. Therefore, geological exchange processes of carbon dioxide such as, carbonate rock dissolution, caves capable of storing large quantities of carbon in both gaseous and aqueous phases, or CO2-degassing by carbonate precipitation forming speleothems and tufa deposits, can not be neglected when interpreting surface-atmosphere exchange of CO2 over carbonate substrates and karst systems. In short, geochemical aspects of carbon cycling have not received sufficient attention during the determination of the global carbon balance, where emphasis has been heavily based towards biological influences.

Radon in caves has been of interest because it may sometimes reach high values (Hakl et al., 1997). Radon concentrations in cave air are influenced by a wide number of factors, including radium content of the rock and sediments, porosity, flow of air and water, atmospheric pressure and earthquakes (Cigna, 2005). Once radon has been incorporated by emanation to fractures and the pore system of the host rock, exhalation is the process determining the radon movement to cave atmosphere. Both diffusion and mass flow play an important role in radon exhalation to the cave. In general, diffusion is the dominant mechanism in intergranular channels, capillaries and smaller pores, while in the larger pores and fractures, mass flow may become important or even dominant. In short, radon, CO2 and others proxy gases can be used as tracers for cave ventilation studies (De Freitas et al., 1982; Duenas et al., 1999; Perrier et al., 2004, 2007; Faimon et al., 2006).

The conventional model of heat transfer by water exchange within a cave consists of a summer regimen with heat flux downwards by diffusion of vapour water, while winter regimen is characterised by a heat flux upwards by convective water transport (De Freitas and Littlejohn, 1987; Perrier et al., 2004; Bourges et al., 2006). For a trace gas within a cave, its distribution is related to cave morphology and air circulation, and its evolution is correlated with microclimatic parameters such as temperature, barometric pressure, rainfall, biogenic activity or carbon dioxide production by breathing visitors. While the mechanistic influence of each microclimatic parameter on CO2 or 222Rn behaviour is known, the difficulty lies in knowing what microclimatic factor prevails during a certain period as well as how the synergetic combination of several factors may affect CO2 or radon levels, or when the dominance of one factor can hide the effects of others.

The aim of this study is to understand the causes and mechanisms of the CO2 and 222Rn variations within a cave atmosphere with very stable temperature and humidity conditions (Castañar de Ibor, Spain), which show an outgassing stage during summer season and a gas recharge phase during wet and rainy periods. Several microclimatic factors have been considered, including air temperature difference between cave and exterior, cave air pressure, rainfall (indicative of infiltration processes) and visiting regime. Long time multivariate series of CO2, 222Rn and microclimatic factors with high-frequency sampling are suitable for data analysis in the Castañar cave, but are generally non-stationary such that straightforward regression analysis cannot be performed. Analysis of entropy of curves is useful for determining the main factor controlling the CO2 and 222Rn variations of cave air through different stationary periods, as well as to identify the main processes involved.

2. Materials and methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

2.1. Study site and microenvironmental monitoring system

Castañar de Ibor is a karstic cave located in the Extremadura Region of SW Spain (Figure 1). It is hosted in Neoproterozoic rocks that form the core of the Ibor Anticline. These rocks are shales and greywackes with interbedded dolostones and magnesites (Alonso-Zarza et al., 2005). The cave developed by the dissolution of the dolomitic beds and the extensive weathering of the shales and greywackes favours collapses which created and enlarged the cave. The texture, fracture system and the mineral composition of the host rock in the different paths of water percolation determine the infiltration rate of meteoric waters and, consequently, the water chemical composition and mineralisation degree of cave waters. The seepage water is rich in Mg, with Ca/Mg ratios oscillating from 0.5 to higher than 1.0 and usually saturated in calcite, dolomite and aragonite (Sanchez-Moral et al., 2006). Castañar de Ibor is a remarkable cave (Natural Monument) with a singular variety of calcite and aragonite speleothems and also moonmilk deposits composed of huntite, dolomite and minor hydromagnesite (Alonso-Zarza and Martin-Perez, 2008). Since its declaration as a Natural Monument in 1997, Castañar cave has limited visitation (i.e. 1508 persons during the year 2004, distributed among 189 groups). The cave entrance is currently a vertical access, 9 m long over an area of 1.5 m2, with a quasi-hermetic trap door installed at the entrance. During periods without visitors the trap door is fully closed. The visitor itinerary does not reach the ‘Lagos’ chamber. The monthly distribution of the number of visitors was very homogeneous throughout the year 2004 (Table I), with an average of 126 people per month. During 2004, the 66% of days, the cave did not receive any visits. The visitors group size ranges from 12 to 14 people, with a maximum size of 15 people per group and 2 groups per day as established by the cave managers. A detailed record of visits (number of visitors, length of stay and duration of door opening during visits) was supplied by cave managers and guides.

thumbnail image

Figure 1. Map and 3D view of Castañar de íbor cave showing spatial distribution of the environmental monitoring stations and depth location. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

Table I. Monthly means of the main environmental parameters during 2004 considering daily data (rainfall and number of visitors correspond to absolute values per month)
MonthExteriorNevadaLagosNo. of visitors
 T ( °C)HR (%)Rainfall (l/m2)T ( °C)HR (%)CO2 (ppm)Rn (Bq/m3)P (mbar)T ( °C)CO2 (ppm) 
  • *

    Data until 14 December.

January 20047.7887.43616.9699.82366339 67796216.973730127
February 20048.0882.08616.9599.83368240 91496416.973713171
March 20049.3274.53116.9599.85368040 93996216.973662168
April 200412.4463.41516.9599.86378038 69196316.973767182
May 200415.1367.47616.9499.87386639 21216.97387882
June 200425.1442.92116.9499.92351626 74316.973537136
July 200426.6927.6416.9399.93342222 31496016.973284100
August 200423.9545.25916.9399.92336422 91295916.973422123
September 200422.2439.0016.9399.92340223 76496116.97357381
October 200415.5269.917416.9499.90358926 82395816.963818122
November 20049.2183.61416.9599.88412133 14196016.964402120
December 2004*6.6683.33016.9599.87414431 82896416.96442196
Year15.2064.654616.9599.88368531 89196116.9737671508

A microenvironmental monitoring system was installed to record the microclimate at three locations: outside the cave entrance, the Nevada chamber and the Lagos chamber (Figure 1). Interior stations recorded every 10 min from December 2003 to December 2004, while the exterior monitoring station recorded hourly measurements throughout the study period. Daily mean data were examined for the simulation of trace gas levels from the analysis of entropy of curves. The network of monitoring stations consists of a datalogger (dataTaker DT50, Grant Instruments Ltd, Cambridge, UK) accepting voltage and current inputs from a 16-channel multiplexer with special sensors designed for the narrow range of measurements expected. The microclimatic parameters measured within cave were air temperature, relative humidity and carbon dioxide content of the air in both chambers, and air pressure and radon content only in Nevada chamber. The outdoor weather station includes a tipping-bucket rain gauge for rainfall measurements and sensors for relative humidity and air temperature (HOBO Weather Station). Cave air CO2 concentrations were measured using a dual wavelength infrared absorption sensor (non-dispersive infrared technology, model 8102, Ventostat) over the range 0–7000 ppm, with an accuracy of 7% or ± 75 ppm. Temperature and relative humidity of the air were measured by a Rotronic humidity and temperature probe HygroClip S3, which combines a resistance thermometer Pt100 1/10 DIN in conformity to IEC 751-DIN 43760 standards (measuring range 0–50 °C, accuracy ± 0.03 °C and resolution 0.01 °C at 0 °C) and a humidity sensor (measuring range 0–100% RH and accuracy ± 0.6%). Temperature and carbon dioxide records were calibrated periodically versus accurate measurements registered within cave with hand-held sensors, previously calibrated in the lab. Air pressure was measured by a Vaisala BAROCAP silicon capacitive, absolute pressure sensor with an accuracy of ± 0.1 mbar (measuring range 800–1100 mbar).

Finally, the 222Rn concentration was measured by means of a Pylon AB5 scintillator-photomultiplier based sensor, which functions as a continuous passive radon detector (CPRD). This radon detector has maximum counting rate of 10 000 cps, with a lowest activity detectable of 24.8 Bq/m3, a sensitivity of 0.041 cpm/Bq/ m3 and an accuracy of ± 4% (operating temperature: −5 to + 50 °C). The nominal background is 0.4 cpm when tuned for optimum performance to measure low levels of radon activity. This equipment was calibrated periodically with a 222Rn calibration standard cell model Pylon 3150 and a standard radioactive source model RNC (Pylon Electronic Inc., Ottawa, Canada) of known activity concentration (Dinh-Chau et al., 2005). Data of uranium content of host rock and speleothems from Castañar cave were determined by Inductively Coupled Plasma Mass Spectrometry (ICP-MS), thanks to ACMElab facilities (Vancouver, Canada).

In order to characterise the porous media of the host rock, mercury intrusion porosimetry (MIP) and helium picknometry techniques were carried out using the Applied Petrology Lab facilities (University of Alicante). A detailed description of these analytical techniques and procedures has been provided by Benavente et al. (2009). From MIP, the connected porosity and pore size distribution were all obtained using an Autopore IV 9500 Micrometrics mercury porosimeter. Total porosity depends on the true and bulk densities. The true density is defined as the ratio of its mass to solid volume, while the dry bulk density of a rock is defined as the ratio of its mass to its volume, including the volume of voids and grains. The true and bulk densities, and consequently the calculation of the total porosity, were obtained using an AccuPyc 1330 Helium pyknometer.

2.2. Statistical analyses

Stationarity is a fundamental requirement for statistical and geostatistical analysis of environmental data (Vanmarcke, 1983; Mateu et al., 2003); a stationary time series has a constant mean and a constant variance, and shows no seasonal variations or irregular fluctuations. For instance, the seasonal trends on weather can disguise other short time oscillations and correlation between microclimate variables. Thus, a traditional regression analysis of complete time series could result in unsatisfactory correlation coefficients due to a non-stationary behaviour, even when the cause-effect relationship is previously known or after detrending process is performed. The technique of entropy of curves offers better results, since this analysis divides a time series into sections where the variance of the series does not vary with the time. The concept of entropy of curves is used in this study based on methodology of Denis and Crémoux (2002) and Denis et al. (2005). This method allows analysing changes in the variability of each time series. More specifically, the entropy of a time signal makes it possible to divide a non-stationary random field, differentiated from others by a change in variance, into subdomains where data are stationary or can be rendered so after trend elimination. In practice, changes in the value of the entropy correspond to changes in the statistical properties of a signal. This property can be used to divide the signal into a set of local stationary sections. In this context the entropy, H(t), is a measure of an uncertainty and variability; the larger the variability, the larger the entropy. Thus, entropy is constant for stationary data and its value gives information about the variability of the signal. For a time series X(t), entropy is expressed as

  • equation image(1)

where the length L(t) is a measurement of the fluctuating amplitude of a signal and can be expressed as (Denis and Crémoux, 2002)

  • equation image(2)

where m|x′| denotes the absolute value of the time derivate of the process X(t), and 〈·〉 denotes mean value.

For practical applications, function L(t) is a cumulative sum of absolute values of first differences, i.e.

  • equation image(3)

From Equations (1) and (2), it is deduced that when L(t) is linear equation image, the entropy is constant and equal to the slope of function L(t). In this context the time series X(t) can be considered stationary of second order. The value of the entropy, H(t), indicates change in the variance and is not affected by a linear or a quadratic trend.

Stationary requirements of each subdomain can be checked by nonparametric testing for trends such as Kendall's test (Hirsch et al., 1982; Hirsch and Slack, 1984). Thus, entropy analyses and Kendall's test are combined to ensure that the detected and localised data set can be called stationary. The test of non-stationarity warns the user that a trend should be removed before a statistical analysis of data sets. After dividing time series into stationary segments, it is possible to analyse, via correlation and regression, the cause-effect relationship between CO2 or 222Rn variations and each of the factors such as cave air pressure, rainfall, visit regime and temperature difference between cave/exterior air (TexteriorTcave). Rainfall largely determines the infiltration rate within cave and the seepage water can be one of the main vehicles for transporting CO2 to the cave atmosphere, by degasification through dripping stalactites. This methodology allows determining the main factor controlling the CO2 and 222Rn variations through different stationary segments. In summary, the methodological plan comprises the following steps:

  • (1)
    Division of time series into stationary intervals using the technique of entropy of curves. In order to analyse the stationary segments (by entropy changes) an automatic data processing has been performed based on the entropy analysis (Frantziskonis and Denis, 2003). The technique of entropy of curves is based on the function L(t) (Denis and Crémoux, 2002), which is a measurement of the fluctuating amplitude of a signal and, for practical application, can be expressed as the cumulative sum of absolute values of first differences. When L(t) is linear, the entropy is constant [equal to the slope of L(t)] and, therefore, the signal can be considered stationary and of second order.
  • (2)
    Definition of constant-entropy segments by combining all intervals from step 1 for each variable, including key parameters (CO2 and 222Rn concentrations) and factors controlling their fluctuations.
  • (3)
    Application of Kendall's test to ascertain the stationarity of the data within each segment. Kendall's test determines whether there is a monotonic (single-direction) trend over time. It is a nonparametric test, identifying trends whether linear or not, and also independent of whether the data follow a normal distribution. The flexibility to incorporate data of different shapes gives the Kendall procedure a distinct advantage over linear regression, where normality of residuals is required. For this study a computer program for the Kendall family of trend test developed by the United States Geological Survey (USGS) (Helsel et al., 2006) has been used. A detailed procedure for calculation of the Tau coefficient for Kendall's test can be found on Helsel and Hirsch (2002) and Helsel and Frans (2006).
  • (4)
    Once the Kendall test is applied, the trend component should be removed from non-stationary series in order to create a stationary residuals set.
  • (5)
    Correlation analyses (cause-effect) of stationary series within each segment. The correlation study is performed by a trend analysis (Kendall's test) where each factor controlling the gases variations is considered as an independent variable, instead of time.
  • (6)
    Simulation of CO2 and 222Rn signals as a function of the predominant control parameter (factor) within each stationary segment and according to Tau correlation coefficients obtained from the previous step.

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

3.1. Prevailing microclimatic conditions

The microclimate of Castañar cave during 2004 and its relationship with external meteorological conditions are shown in Figure 2, and expressed as monthly averages in Table I. Outside Castañar cave a temperate continental climate dominates with an average temperature around 15.2 °C and a strong seasonality (the highest temperature variations reached 43 °C in 2004). The relative humidity is moderately high with an average annual value of 64.6% and a high level of seasonal and short-term fluctuations. The area is characterised by relatively low annual rainfall (546 l/m2 in 2004) with long periods of drought, and maximum rainfall in autumn (174 l/m2 in October).

thumbnail image

Figure 2. Hourly series of the main environmental parameters recorded within Castañar cave during 2004. (a) Evolution of carbon dioxide and radon content of the air. (b) Time series of main factors controlling carbon dioxide and radon content of the air: air temperature difference between cave atmosphere and exterior (TexteriorTcave), air pressure, rainfall (infiltration) and visits (the last two factors are expressed as absolute daily values). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

After following the continuous automatic recording of climatic parameters during an entire annual cycle (2004), it is possible to define Castañar de íbor as a ‘low energy cave’ (Heaton, 1986) with very high microenvironmental stability throughout the annual cycle under natural conditions. The mean values of the temperature and CO2 air concentration are 16.95 °C and 3685 ppm, respectively, in the Nevada chamber, whereas in Lagos chamber these averaged values for air temperature and carbon dioxide are 16.97 °C and 3767 ppm, respectively.

Thermo-hygrometric conditions are very stable over an annual cycle in both areas of the cave (Table I). The maximum temperature variations in the year 2004 were 0.16 °C in the Nevada chamber, and 0.15 °C in the Lagos chamber. The relative humidity of the cave is also very stable, showing monthly mean values higher than 99.80% and a constant state of saturation. The air pressure inside is influenced by the limited connection to the external atmosphere through a single narrow entrance, as well as the particular shape of the cavity, showing an annual average of 961 mbar with a moderate range of oscillation ( ± 39 mbar). Another peculiarity of the microclimatic stability within Castañar cave is the radon concentration (Lario et al., 2006). This cave has a very high level of 222Rn (annual average of 32 246 Bq/m3, with monthly mean values always higher than 20 000 Bq/m3), much higher than the average Rn concentrations found in most caves studied around the world (2500 bq/m3 accord to Hakl et al, 1997; Cigna, 2005). 238U decays through several steps into 226Ra, which decays into 222Rn. Therefore, the concentration of 222Rn depends upon the concentration of these radionuclides in the rock, which varies in a very large range. The uranium content of the shales where the cave is embedded ranges from 2 to 9 ppm, while the speleothem samples ranges from 3.8 to 19.3 ppm. Therefore, both host rock and spelethems significantly could contribute to high radon concentration in the cave atmosphere. The two trace gases (CO2 and 222Rn) show significant parallels in their time series with the lowest values during summer and the highest during winter and early spring. Figure 2 shows an appreciable correlation between the variation of these gases and some environmental factors, including cave/exterior air temperature difference, rainfall and barometric oscillations, as well as the presence of visitors.

3.2. Clustering process and stationarity tests of the factors controlling trace gases variations

Several stationary segments of constant entropy (or potential stationary segments after detrending, Table II) have been chosen for CO2 (7 segments) and 222Rn (8 segments), following automatic data processing based on the combination of entropy analysis and linear regression. The experimental functions L(t) are calculated for CO2 concentrations at each of the monitoring stations (Nevada and Lagos chambers), as well as for the 222Rn level in Nevada chamber, represented together with the time series of these gases (Figure 3). Different factors controlling trace gases variations have been considered depending on the trace gas. For CO2 concentration four factors have been chosen: (1) temperature difference between cave/exterior air (within each chamber; DifT); (2) barometric variations of cave air (Pair) in relation to the external air pressure which can provoke mass flow events; (3) rainfalls (Rf) and 4) Number of visitors (Vt) directly related with production of CO2 by human breathing. By contrast 222Rn is a passive gas whose variation depends only on the exchange of air with the exterior, determined by the air temperature difference between cave atmosphere and exterior (within Nevada chamber; DifT) and the barometric variations of cave air (Pair). Likewise, variations in 222Rn levels could be registered during door opening events (dVt) due to forced ventilation.

thumbnail image

Figure 3. Entropy of curves analysis—functions L(t) for CO2 (a) and radon (b) times series showing evolution of each parameter and different stationary zones defined by constant entropy; from A to H–G (Table II). This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

Table II. Summary of stationarity analyses (entropy values and Kendall's test) performed on selected major stationary segments of constant entropy for CO2 and 222Rn
SegmentClustering for CO2 as key parameterClustering for Rn as key parameter
 CO2(SN)CO2(SL)DifTPairRfVtRnDifTPairdVt
 H(t)KTH(t)KTH(t)KTH(t)KTH(t)KTH(t)KTH(t)KTH(t)KTH(t)KTH(t)KT
  1. DifT, air temperature difference between cave and exterior TexteriorTcave; Pair, cave air pressure; Rf, accumulated rainfall; Vt, accumulated visits; dVt, duration of visits or door opening.

  2. KT index compares the Kendall test coefficient with values given in standard normal distribution tables, so if KT ≤ 1 the variable can be assumed to be stationary (no temporal trend) with a 95% confidence (data in bold). If KT > 1, then temporal trend should be removed and residuals used for correlation analyses.

A8.313.8010.424.431.290.221.630.970.424.004.004.419540.941.310.521.770.9041.202.75
B20.164.9016.234.941.463.022.920.221.046.578.516.8717111.461.600.641.820.0084.202.64
C26.703.9623.484.201.511.450.744.136.114.4414470.731.491.932.920.2277.616.22
D39.090.7629.432.591.982.021.440.342.393.305.963.8111812.831.071.9151.603.19
E27.180.3917.991.431.080.611.621.910.211.703.214.269981.011.630.5675.183.55
F30.042.3624.812.711.120.111.750.149.402.925.292.546640.832.002.491.40054.843.59
G30.083.2120.092.980.820.691.522.500.463.186.473.632722.111.191.411.491.1637.263.16
H8063.911.193.462.040.9244.064.59

Thus, each segment corresponds to a region either stationary (constant entropy) or potentially so after detrending, which can be identified by a linear L(t) function. The corresponding entropy, H(t), has been calculated directly from the slope of the experimental L(t) function. The borders of each stationary region have been established considering the coincident limits of the linear L(t) functions for all key parameters and factors controlling trace gases variations. Some inevitable limits have been established when data lack due to instrument failures. Table II summarises the results from the stationarity analyses, while Tables IIIV summarise the results of the correlation analyses between each of the control parameters and the CO2 and 222Rn levels. The main conclusions from these hard data are summarised and synthesised in the Table VI and Figures 6 and 7, and consequently the arguments put forward in the Section 4 are based on this table and figures.

Table III. Summary of correlation analyses for CO2 in the ‘Sala Nevada’ chamber and each of the control parameters: air temperature difference between exterior and cave (DifT; TexteriorTcave), cave air pressure (Pair), rainfall (Rf) and number of visitors (Vt)
SegmentTemperature difference between exterior and caveCave air pressureRainfallNumber of visitors
 TauKTCO2(SN) = f(DifT)TauKTCO2(SN) = f(Pair)TauKTCO2(SN) = f(Rf)TauKTCO2(SN) = f(Vt)
  1. Some segments with statistically significant correlations with visits and rainfall have been discarded, since they represent conceptual incoherencies as a result of the masked effect of other highly correlated factors.

A0.010.020.16 + 0.06 × DifT0.371.96−2074.5 + 2.16 × Pair−0.411.99−2.14 −2.69 × Rf0.211.00−9.23 + 0.42 × Vt
B−0.110.792.28 −2.21 × DifT0.281.27−1535.9 + 1.60 × Pair0.010.081.34 + 0.04 × Rf0.090.608.43 + 0.23 × Vt
C−0.140.6838.72 −5.59 × DifT0.522.484.74 + 15.67 × Rf−0.060.2940.36 −0.59 × Vt
D0.180.763402.9 −8.80 × DifT0.441.85−17881 + 22.17 × Pair−0.080.333400.3 −1.18 × Rf0.502.123332.9 + 6.45 × Vt
E0.251.173452.8 −10.11 × DifT0.231.053394.7 + 10.21 × Pair−0.121.064608.17 −3.54 × Rf0.030.143393.9 + 0.38 × Vt
F−0.190.57−53 −94 + 6.57 × DifT0.100.29−2843.1 + 2.93 × Pair−0.050.14−72.87 −0.36 × Rf0.080.21−21.23 + 1.02 × Vt
G−0.190.75114.48 −9.86 × DifT0.552.22−68.41 + 10.5 × Pair0.271.06−34.32 + 9.52 × Rf−0.210.85−49.15 −1.51 × Vt
Table IV. Summary of correlation analyses for CO2 in the ‘Sala Lagos’ chamber and each of the control parameters: air temperature difference between exterior and cave (DifT, TexteriorTcave), cave air pressure (Pair), rainfall (Rf) and number of visitors (Vt)
SegmentTemperature difference between exterior and caveCave air pressureRainfallNumber of visitors
 TauKTCO2(SL) = f(DifT)TauKTCO2(SL) = f(Pair)TauKTCO2(SL) = f(Rf)TauKTCO2(SL) = f(Vt)
  1. Some segments with statistically significant correlations with visits and rainfall have been discarded, since they represent conceptual incoherencies as a result of the masked effect of other highly correlated factors.

A−0.150.74−21.04 −2.85 × DifT0.180.86−1889.1 + 1.97 × Pair−0.482.332.16 −5.57 × Rf0.261.26−13.53 + 0.92 × Vt
B−0.030.200.15 −0.46 × DifT0.241.08−1652.8 + 1.71 × Pair0.080.54−2.46 + 0.27 × Rf−0.060.41−4.18 −0.15 × Vt
C−0.180.8620.82 −5.86 × DifT0.512.45−20.62 + 19.21 × Rf−0.120.5625.93 −1.33 × Vt
D0.170.73−7.86 + 4.91 × DifT0.592.50−16 251 + 16.92 × Pair0.080.35−14.98 + 0.98 × Rf0.472.00−46.82 + 3.59 × Vt
E−0.050.2111.29 −1.75 × DifT0.281.281.80 + 8.85 × Pair−0.080.69596.5 −1.75 × Rf−0.060.281.03 −0.71 × Vt
F−0.060.18−88.82 −3.21 × DifT−0.170.504238.4 −4.50 × Pair−0.050.14−72.9 −0.36 × Rf0.110.32−58.2 + 1.79 × Vt
G−0.230.92−104 −9.17 × DifT0.582.32−34.11 + 14.23 × Pair0.572.2940.6 + 20.25 × Rf−0.471.8920.56 −4.53 × Vt
Table V. Summary of correlation analyses for Rn and each of control parameters: air temperature difference between exterior (DifT; TexteriorTcave), cave air pressure (Pair) and duration of door opening during visits (dVt)
SegmentTemperature difference between exterior and caveCave air pressureDuration of door opening
 TauKTRn = f(DifT)TauKTRn = f(Pair)TauKTRn = f(dVt)
A−0.130.4537 087 −195.2 × DifT0.090.31−17 221 + 58.2 × Pair0.291.0338 842 −4.4 × dVt
B0.170.50895 + 72.7 × DifT0.140.39−32 906 + 34.4 × Pair−0.090.27337.9 −1.03 × dVt
C0.191.2539 851 −275.1 × DifT−0.140.64148 690 −112.6 × Pair0.251.6241 344 −8.94 × dVt
D0.541.96−2 390.1 −538.6 × DifT0.391.43−890 + 12.78 × dVt
E0.110.44−1 813.4 + 122.2 × DifT0.441.75546.2 −9.69 × dVT
F0.321.2922 603 + 154.3 × DifT0.371.49−133 700 + 163 × Pair0.361.4622 549 + 5.14 × dVt
G0.331.18−346.4 + 97.4 × DifT0.270.94−341.6 + 75.1 × Pair0.050.18−227.7 + 1.07 × dVt
H0.050.2673.7 + 52.33 × DifT0.100.52−51 207 + 53.5 × Pair0.010.07186.7 + 0.48 × dVt
Table VI. Clustering of the main control factors, environmental conditions and processes prevailing during each period, obtained from the entropy of curves analysis
PeriodSegmentKey factorPrevailing processesCave stage
 CO2222Rn   
  1. DifT, air temperature difference between cave and exterior, TexteriorTcave; Pair, cave air pressure; Rf, accumulated rainfall; Vt, accumulated visits; dVt, duration of visits or door opening.

December–JanuaryA′–AA′–A–BPair• Contribution of CO2-depleted and 222Rn-enriched air from the fissure networkLimited air exchange cave/exterior
   Vt• Forced ventilation by door opening 
February–MayBCPair• Water recharge and CO2 outgassing by seepagesLimited air exchange cave/exterior
    • Gas contribution by air pressure rises in the fissure network 
    • Convective and forced ventilation controlling 222Rn levels 
May–JuneCD–EDifT• Inversion of thermal gradientUnderground air renewal
   Rf• Air exchange cave/exterior through the fissure network 
   dVt• Short-term fluctuations of gases by rainfall events and door opening 
July–SeptemberD–EF–GDifT• High air exchange cave/exteriorUnderground air renewal
   Pair• Barometric control of air from the fissure network 
   Vt• Influence of visiting regime 
October–NovemberF–GHDifT• Trend towards limited air exchange cave/exteriorGases recharge
   Pair• Barometric control of cave atmosphere 
   Rf• Gas mobilisation by infiltration 

The entropy function must be complemented with Kendall's test to ensure that the each detected and localised dataset is stationary. Table II summarises the data from stationarity analyses (entropy values and Kendall's test) performed on selected major stationary segments of constant entropy, considering both key parameters (CO2 and 222Rn concentrations of cave air) and each of the factors controlling trace gas variations.

The entropy values demonstrate higher fluctuations for trace gases than for factors over an annual cycle. The entropy values for CO2 time series are generally higher in the Nevada chamber than in the Lagos chamber, due to proximity to the cave entrance, versus damping of external meteorological oscillations in the Lagos chamber. Comparing simultaneous segments of the time series of both trace gases, the entropy data of CO2 range practically in the same way (increase or decrease) in each chamber, whereas radon levels show an oscillation in H(t) values that coincide with those of CO2 only when both gases begin an upward trend (stationary segments E–F for CO2 and G–H for 222Rn). These results demonstrate that both gases have a similar behaviour during determined periods, which can be useful when they are used as tracers for cave air degassing. Likewise, the dissimilar H(t) variations for CO2 and 222Rn are related to the different source mechanisms for each trace gas.

3.3. Simulation of levels of trace gases from the technique of entropy of curves

Once the detrending process has been performed, a new batch of Kendall's tests is carried out for checking the stationarity of the residuals dataset within each segment. Results showed that all segments passed the stationarity conditions established by the Kendall's test, allowing application of the correlation analyses (cause-effect) within each stationary segment. The correlation study (functional relationship between the trace gases and each factor controlling the levels of gases) is performed again by a trend analysis (Kendall's test) where each factor controlling the trace gases variations is considered as an independent variable, instead of time. Tables IIIV summarise the results of the correlation analyses between each of the control parameters and CO2 level in the Nevada chamber, CO2 level in the Lagos chamber and the 222Rn level, respectively. For periods that failed the previous stationary test (segments with a temporal trend), the correlation analyses were performed using residuals datasets in order to identify reliable relationships (without any trend effect) between parameters. In such cases the correlation functions correspond to the relationships between the variations, after removal of the linear trend, of the trace gases and the control factors. Therefore, results derived from the application of these correlation functions must be added to either downward or upward trend, to obtain the predicted concentration of the trace gas.

The KT index compares the Kendall test coefficient with values given in standard normal distribution tables; thus if KT > 1 either CO2 or 222Rn can be assumed to be correlated with the control parameter with a 95% confidence (data in bold of Tables IIIV). Consideration of these segments that show statistically significant correlations among parameters allows modelling the temporal functional relationship between the trace gases and the factors controlling their concentrations and transport. Some segments with statistically significant correlations with visits and rainfall have been discarded, since they represent conceptual incoherencies as result of the masked effect of other highly correlated factors. For instance, a statistical incoherence corresponds to the high KT index for the inverse relationship between rainfalls (which should increase dripping of water and, consequently CO2-degassing) and the CO2 variations in the ‘Sala Nevada’ chamber (Table III).

Figure 4 shows the percentage of error (or percent deviation) of the simulated signal with respect to the time series of CO2 and 222Rn of the cave air registered during the studied period. The formula for calculating percentage error (deviation) is the simulated value minus the observed value divided by the observed value and multiplied by 100, in absolute value. The simulated signal achieves a high accuracy, with mean percentage errors for CO2 estimation oscillating from 0.83% (Lagos chamber) to 0.92% (Nevada chamber) and maximums below 8% in both cases. The simulated 222Rn signal presents a higher percentage error (mean of 3.82%) due to the intrinsic elevated noise of the measured data (related with accuracy of the radon detector). The maximum percentage error for 222Rn estimation (during early April) coincides with sharp drops of gas levels due to forced ventilation when the cave entrance is open for a long time. Figures 6 and 7 represent the simulation of the mean daily data of CO2 and 222Rn in the cave atmosphere through the period December 2003–December 2004, considering the main factors correlated in each stationary segment of constant entropy. Through some segments, the trace gas concentrations have been modelled using their functional relationship with several factors, and sometime factors from the same segment exert opposing effects on the gas level. In this last case, the simulated gas concentration is calculated as a weighted mean of results from different functional relationship in function of each Tau correlation coefficient (Tables IIIV).

thumbnail image

Figure 4. Percentage of error of the simulated signal with respect to the time series of CO2 and 222Rn of the cave air registered during the studied period

Download figure to PowerPoint

3.4. Porous media properties of the host rock

Figure 5 shows the cumulative mercury intrusion and pore size distribution curves obtained by MIP for the host rock from Castañar cave (shales, chiefly). The host rock has a complex porous media: high porosity and polymodal pore size distribution. The pore size interval characterisation by MIP ranges from 0.002 to 200 µm. The polymodal pore size distribution of the host rock samples a substantial part of the pore spectrum between 0.002 µm and 0.02 µm with a mean radius of 0.14 µm, providing high water absorption rates. The pore size distribution curve reveals the presence of larger pores (>100 µm), probably linked to the microfractures along planes of weakness of the typical thin sheets forming this rock. In short, the porous media is characterised on average by low values of total porosity (14.3%, the bulk density is 2.4 g/cm3 and the true density is 2.8 g/cm3) and a very low connected porosity (4.3%); therefore, the pore fluid displacement is limited and the water filling of the porous system is intense.

thumbnail image

Figure 5. Cumulative mercury intrusion and pore size distribution curves obtained by MIP for host rock (shales) from Castañar cave; r (µm) is the mean radius of the pores on a logarithmic scale

Download figure to PowerPoint

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

Figures 6 and 7 represent the monitored and simulated series. These figures also include the temporal evolution of rainfall (daily data and accumulated values of every stationary segment) and external relative humidity (reciprocal to the air temperature difference between cave and exterior). The key question regards the value of each factor as a predictor of trace gas levels and specifically the cave air degassing. Detailed below is the time evolution of this functional relationship between factors and trace gases from the entropy of curves analysis. Table VI summarises the main control factor, environmental conditions and processes prevailing during each period.

thumbnail image

Figure 6. Simulation of CO2 behaviour within both chambers of the cave, considering the main factors correlated in each stationary zone of constant entropy. Time evolutions of rainfall and external relative humidity are shown to indicate degree of partial water filling of the host rock-soil system. Note that CO2 has been simulated in stationary zones where correlation coefficients (with factors) were significant. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

thumbnail image

Figure 7. Simulation of 222Rn behaviour within both chambers of the cave, considering the main factors correlated in each stationary zone of constant entropy. Time evolutions of rainfall and external relative humidity are shown to indicate degree of partial water filling of the host rock-soil system. Note that Rn has been simulated for stationary zones where correlation coefficients (with factors) were significant. This figure is available in colour online at wileyonlinelibrary.com/journal/joc

Download figure to PowerPoint

During the first part of winter season; from December 2003 until late January 2004 (corresponding to segments A for CO2 and A–B for 222Rn), the external air temperature is considerably lower than air temperature within the cave (the daily average of air temperature difference, TexteriorTcave oscillates between −13.03 and −5.30 °C in both chambers). The accumulated rainfall over the cave catchment is low (37.8 l/m2) but the external relative humidity is higher than 80% most of the time. Under these environmental conditions, the high water content existing in the system of membranes (formed by the soil and the host rock of vadose zone immediately over the cave) leads to decreased exterior-interior exchange. The CO2 level of the cave air drops slightly due to the scarce gas addition from the soil (depleted biological activity during winter season) and the low infiltration rate, reaching an equilibrium gas level for the cave atmosphere. However, the above-mentioned process of cave isolation allows increasing of the radon levels. During this period air pressure variations control the short-term fluctuations of CO2 in the Nevada chamber following a natural downward trend, as is explained later for the next segment (spring season). Visits (123 people) provoke non-accumulative increments of gas levels in both chambers. 222Rn fluctuations are due to the influx of colder air when the cave entrance is open during visits.

After the rain events in late January and spring (130 l/m2 from February to May), seepage water reached the cave and degassing took place depending on the pCO2 of the subterranean air. This process predominates throughout the annual cycle since the pCO2 of dripping water is always moderately higher than the air pCO2 in this cave (Sanchez-Moral et al., 2006). Therefore, during this stage (segment B for CO2 and C for Rn, Figures 6 and 7) with limited surface-atmosphere exchange of air, rainfalls led to a progressive CO2 increase by outgassing air into both chambers. However, the host rock filters the infiltration signal and no correlation was found between rainfall and short-term fluctuations of CO2. The variations of CO2 are clearly influenced by fast barometric oscillations in the cave air. The fluctuations of cave air pressure are transmitted to the air trapped within the network of pores in the host rock, and vice versa. Thus, a rise of air pressure in the fissure network constrains the liberation of this air, which has previously reached a moderate concentration of CO2 as a function of the partial water filling of the porous system and host rock fissures from the vadose zone.

From March to May 2004 the temperature of external air tends to match the cave air temperature and the external relative humidity decreases to 60%. The CO2 and 222Rn levels during this period, however, undergo high fluctuations around daily mean values of 39 845 Bq/m3 for Rn and 3767 ppm for CO2 in both chambers, strongly controlled by the variations of the air temperature difference between the cave atmosphere and the exterior in the case of radon content, as well as by periods when the cave entrance was open. Thus, several sub-sectors can be differentiated for the B and C segments as a function of sharp drops of gas levels due this forced ventilation. The 222Rn level decreases when the duration of door opening is extended due to the inflow, by convective circulation, of external air with lower gas content. This process is favoured when external air temperature decrease relative to the cave air temperature (TexteriorTcave < 0). However, the negative Tau coefficient between air temperature difference and 222Rn levels (Table V for segment C) demonstrates an inverse relationship between both parameters over long time periods. Actually, the upward trend of air temperature outside cave favours evaporation in the soil and the most superficial layers of host rock, decreasing the partial water filling of the porous system and, therefore, increasing the air exchange with the outer atmosphere. Likewise, the short-term decreases in the exterior air temperature are associated with an increase in relative humidity, which provokes a rise of the partial water filling of the soil and most superficial layers of the host rock. This reinforces the isolation of the cave reflected by rises in 222Rn levels. Unfortunately the air pressure data lack in the second part of the segment B (from April to May), and complicate a proper simulation of CO2 levels during these months (Figure 6). However, the high correlation between both parameters during the first part of this segment (from the late February and throughout March) entails that air pressure must not be ruled out for simulating the gas variations during the entire segment B. In the case of the Nevada chamber, variations of CO2 are also due to the presence of visitors (469 people). In short, during this period gas exchange with the outer atmosphere is limited due to an intense water filling process within the network of fissures and pores, as well as due to the prevailing state during which the cave entrance is closed.

The segments of constant entropy denoted as C for CO2 and D–E for Rn (late May to early July 2004) correspond to a period of transition characterised by a clear downward trend in the two gases towards minimum annual levels. The sharp drop in concentrations of both gases correlated clearly with the inversion of the thermal gradient (external air temperature begins to consistently exceed the cave air temperature) and, therefore, with the drop of relative humidity in the external atmosphere. This provokes significant air exchange (cave/exterior) through the soil and host rock membranes. As this thermal effect provokes a perfect downward trend in levels of trace gases (ideal non-stationary series), without fluctuations of any other kind, residuals are practically null and cannot be correlated with short-term fluctuations in the cave-exterior air temperature difference. However, the correlation between trends of this air temperature difference and each trace gas, reaches high negative Tau coefficients (−0.69 for CO2 in Nevada chamber, −0.51 for CO2 in Lagos chamber, and −0.63 for 222Rn). During this period of cave degassing the 222Rn levels depend only on emanation from the host rock and the subsequent exhalation process into the cave air, which is favoured under the prevailing conditions of air exchange cave/exterior (inverted thermal gradient). With regard to CO2, occasional rainfall in June provoked gas increments after the hydric recharge of spring months.

After this period of transition, another segment of constant entropy of trace gases is reached (time-segment D–E for CO2 and F–G for Rn; from July to early autumn 2004). Previously, the CO2 concentration of air in both chambers of the cave undergoes a sudden fall to a mean level of 3400 ppm. These minimum values of CO2 concentrations remain practically constant during the summer, showing short-term fluctuations never exceeding 250–300 ppm. By contrast, the 222Rn content of cave air drops from 40 000 Bq/m3 to the lowest levels of the annual cycle (mean of 22 700 Bq/m3) with short variations between 20 500 and 24 900 Bq/m3 (Figure 7). During the 3-month period corresponding to summer, the external air temperature reaches the highest values relative to the air temperature within cave. Accumulated rainfall over the cave catchment was low (75.8 l/m2) taking into account the long duration of these constant-entropy time segments (73 days). There are many days without rain (89%) and the external relative humidity reaches the lowest annual levels (between 30% and 50%). Under these environmental conditions, evaporation from the soil and from the more superficial layers of the host rock (shale and greywacke) are more intense, as well as decreased water deposition on soil by condensation (dew) due to the low environmental humidity. Hence the degree of water content of the host rock and soil membranes reaches the lowest level, increasing the cave-exterior air exchange through the network of fissures and pores that feeds the cave. This air exchange process could be generated by diffusion of gases from the upper layers or, even, by a slight external air intake, warmer and gas-depleted, through the lower levels of the karst system, which in some cases are geomorphologically closer to the external atmosphere. This is an interesting issue for a future research, since it needs to be checked with new environmental monitoring studies.

During this period of moderate connection between cave and exterior atmosphere, the evolution of CO2 is clearly influenced by barometric conditions. The fast rise of the exterior temperature (increase of cave/exterior air temperature difference), and consequently of air pressure, generates an imbalance relative to the cave atmosphere. As air exchange is more intensive during this period, the barometric equilibrium is reached with the movement towards the cave of the air trapped within the network of fissures and connected pores of host rock and upper soil. The CO2 and 222Rn contents of cave air maintain a moderate level, with mild short-term increments due to the periodic rises of cave air pressure. Finally, CO2 fluctuations are enhanced during the time interval with tourist visits, even in isolated areas as Lagos chamber. As regards the CO2 concentration of air in the Lagos chamber, there is a more active supply of infiltration water from seepages, drip points and flowstones, so that even the scant rainfall events during this summer period contribute to maintain a significant CO2 level by degasification when the infiltration water reaches the cave atmosphere.

Beginning in mid-October the air temperature of the exterior was consistently below the cave air temperature (Figure 2). This weather coincides with the time segments of constant entropy denoted as F and G for CO2 levels, and H for 222Rn level (Figures 4 and 5). Simultaneously, the external relative humidity again reaches a level above 80% and the most intense rainfall events are registered (max 35.8 l/m2 on 20 October), with accumulated rainfall over the cave catchment is 142.6 l/m2 in only 18 days (25.6% of the annual rainfall). The partial water filling of the rock-soil upper membranes increases continuously, again restricting cave-exterior air exchange and raising the CO2 and 222Rn levels. The upward trend in concentrations of both gases is directly proportional to the downward trend in cave-exterior air temperature differences (TexteriorTcave, segment F). In the case of short-term variations of 222Rn, no factor shows noteworthy correlation from Kendall's test. Barometric variations and rainfall events exert the main regulatory control on CO2 fluctuations during October–November, coinciding with the beginning of the period when the cave air temperature is constantly higher than in the exterior. Once the air proportion within the network of fissures and pores has decreased due the intense rainfalls, the barometric rises mainly affect the bigger air-filled voids such as the cave atmosphere. Thus, the fluctuations of cave air pressure still remains a control factor, hindering the inflow of the scarce air trapped within this network of fissures and pores when the air pressure rises, whereas decreases in air pressure have an opposite effect. Barometric pressure changes include both periodic and non-periodic components: periodic changes are due mostly to solar diurnal and semidiurnal tides, while non-periodic changes are due to the general air circulation, which is thermally driven (Ritzi et al., 1991). A similar influence of air pressure changes on the rate of dripping from the stalactites has been described for other caves (Genty and Deflandre, 1998; Baker and Brunsdon, 2003; Fernandez-Cortes et al., 2007). In any event, the heavy rainfalls registered during this short-period (time-segment F) constrain the radon flux in rock and its transport by groundwater through joints and fissures, hindering barometric control. The infiltration process favours the CO2 and 222Rn movement to the cave atmosphere as well; therefore, concentration levels of both gases undergo a remarkable increment.

The end of the annual cycle (December 2004, segment A′) is characterised by a situation of relative stability with a limited surface-atmosphere exchange of air, similar to that described for segment A. This leads to high concentrations of both gases with short-period variations clearly influenced by rapid barometric oscillations in cave atmosphere.

In short, the monitored and simulated series show a high degree of correlation for both trace gases with key factors, validating the optimum results obtained from segmentation of the functional relationship between the gas levels and the fluctuations of factors. The temporal pattern of CO2 and 222Rn throughout an annual cycle seems to be governed by hydrogeological, meteorological and microclimatic processes within the subterranean environment. Rainfalls and external humidity determine the water content of the porous and fissures system from host rock and soil. The low porosity of the host rock (shales) determines that small increases in humidity or rainfall can generate an intense process of water filling of the porous system and host rock fissures, up to reach a fixed limit that restricts the gas exchange between the inner cave and the atmosphere.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

The application of entropy of curves analyses has provided temporal details regarding the trace gases variations in Castañar cave throughout an annual cycle, revealing a complicated functional relationship between the trace gases (CO2 and 222Rn) and each of the key factors. Trace gases within the subterranean atmosphere of Castañar cave do not follow the stair-step patterns observed in many others caves, with strong seasonal shifts in function of the thermal exterior-cave relationship and with high summer and low winter levels (i.e. Bourges et al., 2001; Perrier et al., 2007). The partial water filling of the porous system and fissures from the membranes covering the cave (host rock and soil) is determined by the external relative humidity (controlled by the external air temperature) as well as by rainfalls, which play a key role in confining the cave atmosphere. Under confined conditions, barometric pressure variations play an important role in controlling subsurface radon concentrations, as shown by the direct relationship between both parameters. This functional relationship is contrary to the standard model described in other caves for 222Rn (Hakl et al., 1996, 1997), or similarly for CO2 (Denis et al., 2005). The monitored and simulated series show a high degree of correlation for both trace gases with environmental factors, as shown by the optimum results obtained from the segmentation of the functional relationship between the gas levels and the fluctuations of factors. Thus, in the course of an annual cycle, different stages and processes can be distinguished:

  • (1)
    Isolated cave atmosphere. During the winter and much of the spring (December to May) the outside temperature is constantly below the cave air temperature, the external relative humidity is high (60–90%) and rainfall represents the 41% of the annual total. Under these environmental conditions the membranes covering the cave (host rock and soil) tend to a full state of water saturation, so a limited air exchange prevails between the cave and exterior atmosphere through the network of fissures and pores. During this period both trace gases reach the maximum mean levels: 3700–4500 ppm of CO2 and 35 000–45 000 Bq/m3 for 222Rn. When the confined conditions are reached (especially during the rainy periods), the microfracture networks, which are penetrative structures (< mm), carry diphasic infiltration (water plus air) delaying the air gaseous transfer between the cave atmosphere and the exterior, thus favouring the barometric control on the air exchange through the cave entrance and in function of barometric falls of external atmosphere. Since the achievement of confined conditions is gradual, during the semiconfined phases barometric increments affect air trapped in the fissure network, favouring gas mobilisation to the cave. During this stage the cave behaves as a CO2 reservoir.
  • (2)
    Underground air renewal. During late May and throughout the summer (from June to October) the exterior air temperature is constantly above the cave air temperature, the external relative humidity falls below 60% and registered rainfall represents 33.4% of the annual total, but corresponding to a long period of 130 days. These factors provoke the partial opening of the porous system of upper soil and the network of host rock fissures (isolating membranes). Thus, the air exchange between the cave and the outer atmosphere is favoured, with an intense degasification process taking place such that the levels of trace gases drop up to moderate concentrations of 3200–3400 ppm for CO2 and 22 000–25 000 Bq/m3 for 222Rn. Degassing processes involve a net emission of CO2 to the atmosphere.
  • (3)
    Gas recharge phase. surface-atmosphere exchange is maintained throughout the summer until October, when the outside temperature drops below the cave temperature, external relative humidity increases and the first events of intense rain are registered (142.6 l/m2 in only 18 days, 25.6% of the annual rainfall). Then, the degree of water content existing on the rock-soil upper membranes increases continuously, restricting again the cave-exterior air exchange and thereby resulting in recharge of CO2 and 222Rn levels.

Similar stages and processes have recently been observed in other caves in Spain (Hoyos et al., 1998; Sanchez-Moral et al., 1999; Kowalski et al., 2008). The results, conclusions, and designed methodology of the present study may be valuable for future research regarding cave air dynamics, which requires focus on several aspects: (1) consideration of the intra-annual CO2 fluctuations in terms of recent and current climate variability (Perrier et al., 2005) and paleoclimatic interpretations from speleothem records (e.g. in line with studies such as Baker et al., 2002; Smith et al., 2006), (2) the key role of karst geochemical processes in the local carbon cycle on short timescales, and (3) application of the knowledge regarding mass and energy fluxes involved in the subterranean environments, in order to ensure the constant climate required for the conservation of caves (De Freitas and Schmekal, 2003; Fernandez-Cortes et al., 2006a), as well as for designing, operating, and maintaining underground facilities (Salve et al., 2008).

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References

This research was supported by the Spanish Ministry of Education and Science, project CGL2006-11561/BTE, in agreement with the Scientific Cooperation between the Museo Nacional de Ciencias Naturales (CSIC) and the Department of Agriculture and Environment of the Extremadura Autonomous Region and through FEOGA-ORIENTACION-FEDER funds. Thanks are due to the latter for authorising the publication of the results. Fernandez-Cortes benefits from the Spanish Ministry of Education and Science Research Programme ‘Juan de la Cierva’ and JAE-Doc Program (CSIC). We are sincerely grateful for all the help given by the cave guides Ana Blázquez and Antonio Baltasar of Castañar cave, and for their valuable collaboration throughout the entire investigation. Special thanks to Dr Denis (Centre de Développement des Géosciences Appliquées, Université Bordeaux, France) for advice on the technique of entropy of curves, and Dr Kowalski (Department of Applied Physics, University of Granada, Spain) for his useful comments and language support.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgements
  9. References
  • Alonso-Zarza AM, Gil-Peña I, Martínez-Flores E, Muñoz-Barco P. 2005. La cueva de Castañar. In Patrimonio Geológico de Extremadura: Geodiversidad y lugares de interés geológico, Muñoz BarcoP, Martínez FloresE (eds). Junta de Extremadura; 99111.
  • Alonso-Zarza AM, Martin-Perez A 2008. Dolomite in caves: Recent dolomite formation in oxic, non-sulfate environments. Castañar Cave, Spain. Sedimentary Geology 205(3-4): 160164, DOI:10. 1016/j.sedgeo.2008.02.006.
  • Baker A, Brunsdon C 2003. Non-linearities in drip water hydrology: an example from Stump Cross Caverns, Yorkshire. Journal of Hydrology 277(3-4): 151163, DOI:10.1016/S0022-1694 (03) 00063-5.
  • Baker A, Proctor CJ, Barnes WL 2002. Stalagmite lamina doublets: a 1000 year proxy record of severe winters in northwest Scotland? International Journal of Climatology 22(11): 13391345, DOI: 10.1002/joc.800.
  • Baldini JUL, Baldini LM, McDermott F, Clipson N. 2006. Carbon dioxide sources, sinks, and spatial variability in shallow temperate zone caves: Evidence from Ballynamintra Cave, Ireland. Journal of Cave and Karst Studies 68(1): 411.
  • Baldini JUL, McDermott F, Hoffmann DL, Richards DA, Clipson N 2008. Very high-frequency and seasonal cave atmosphere P-CO2 variability: implications for stalagmite growth and oxygen isotope-based paleoclimate records. Earth and Planetary Science Letters 272(1-2): 118129, DOI:10.1016/j.epsl.2008.04.031.
  • Batiot-Guilhe C, Seidel JL, Jourde H, Hébrard O, Bailly-Comte V. 2007. Seasonal variations of CO2 and 222Rn in a Mediterranean sinkhole spring (Causse d'Aumelas, SE France). International Journal of Speleology 36(1): 5156.
  • Benavente D, Cañaveras JC, Cuezva S, Laiz L, Sanchez-Moral S 2009. Experimental definition of microclimatic conditions based on water transfer and porous media properties for consevation of prehistoric constructions: Cueva Pintada at Galdar, Gran Canaria, Spain. Environmental Geology 56: 14951504, DOI 10.1007/s00254-008-1246-8.
  • Bourges F, Genthon P, Mangin A, d'Hulst D 2006. Microclimates of l'Aven d'Orgnac and other French limestone caves (Chauvet, Esparros, Marsoulas). International Journal of Climatology 26(12): 16511670, DOI: 10.1002/joc.1327.
  • Bourges F, Mangin A, d'Hulst D. 2001. Carbon dioxide in karst cavity atmosphere dynamics: the example of the Aven d'Orgnac (Ardeche). Comptes Rendus de l Academie des Sciences Serie Ii Fascicule A-Sciences de la Terre et des Planetes 333(11): 685692, DOI:10.1016/S1251-8050 (01) 01682-2.
  • Cigna AA. 2005. Radon in caves. International Journal of Speleology 34(1-2): 118.
  • De Freitas CR, Littlejohn RN 1987. Cave climate: assessment of heat and moisture exchange. International Journal of Climatology 7: 553569, DOI: 10.1002/joc.3370070604.
  • De Freitas CR, Littlejohn RN, Clarkson TS, Kristament IS. 1982. Cave climate: assessment of airflow and ventilation. International Journal of Climatology 2: 383397, DOI: 10.1002/joc.3370020408.
  • De Freitas CR, Schmekal A 2003. Condensation as a microclimate process: Measurement, numerical simulation and prediction in the Glowworm Cave, New Zealand. International Journal of Climatology 23(5): 557575, DOI: 10.1002/joc.898.
  • Denis A, Crémoux F. 2002. Using the entropy of curves to segment a time or spatial series. Mathematical Geology 34(8): 899914.
  • Denis A, Lastennet R, Huneau F, Malaurent P. 2005. Identification of functional relationships between atmospheric pressure and CO2 in the cave of Lascaux using the concept of entropy of curves. Geophysical Research Letters 32: L05810, DOI: 10.1029/2004GL022226.
  • Dinh-Chau N, Chrusciel E, Prokólski, K. 2005. Factor controlling measurement of radon mass exhalation rate. Journal of Environmental Radioactivity 82: 363369, DOI:10.1016/j.jenvrad.2005.02.006.
  • Duenas C, Fernandez MC, Canete S, Carretero J, Liger E. 1999. Rn-222 concentrations, natural flow rate and the radiation exposure levels in the Nerja Cave. Atmospheric Environment 33(3): 501510, DOI:10.1016/S1352-2310 (98) 00267-2.
  • Ek C, Gewelt M 1985. Carbon-dioxide in cave atmospheres—new results in belgium and comparison with some other countries. Earth Surface Processes and Landforms 10(2): 173187, DOI: 10.1002/esp.3290100209.
  • Faimon J, Stelcl J, Sas D 2006. Anthropogenic CO2-flux into cave atmosphere and its environmental impact: a case study in the Cisarska Cave (Moravian Karst, Czech Republic). Science of the Total Environment 369: 231245, DOI: 10.1016/j.scitotenv.2006.04.006.
  • Fernandez-Cortes A, Calaforra JM, Sanchez-Martos F, Gisbert J. 2006a. Microclimate processes characterization of the giant Geode of Pulpi (Almeria, Spain): technical criteria for conservation. International Journal of Climatology 26(5): 691706, DOI: 10.1002/joc.1269.
  • Fernandez-Cortes A, Calaforra JM, Sanchez-Martos F. 2006b. Spatiotemporal analysis of air conditions as a tool for the environmental management of a show cave (Cueva del Agua, Spain). Atmospheric Environment 40(38): 73787394, DOI: 10.1016/j.atmosenv.2006.06.045.
  • Fernandez-Cortes A, Calaforra JM, Sanchez-Martos F, Gisbert J. 2007. Stalactite drip rate variations controlled by air pressure changes: an example of non-linear infiltration processes in the ‘Cueva del Agua’ (Spain). Hydrological Processes 21(7): 920930, DOI: 10.1002/hyp.6283.
  • Frantziskonis G, Denis A. 2003. Complementary entropy and wavelet analysis of drilling-ability data. Mathematical Geology 35(1): 89103.
  • Genty D, Deflandre G 1998. Drip flow variations under a stalactite of the Pere Noel cave (Belgium): Evidence of seasonal variations and air pressure constraints. Journal of Hydrology 211(1-4): 208232, DOI:10.1016/S0022-1694 (98) 00235-2.
  • Hakl J, Csige I, Hunyadi I, Varhegyi A, Geczy G. 1996. Radon transport in fractured porous media—experimental study in caves. Environment International 22: S433S437.
  • Hakl J, Hunyadi I, Csige I, Geczy G, Lenart L, Varhegyi A 1997. Radon transport phenomena studied in karst caves—international experiences on radon levels and exposures. Radiation Measurements 28(1-6): 675684, DOI:10.1016/S1350-4487 (97) 00163-7.
  • Heaton T. 1986. Caves: a tremendous range in energy environments in earth. National Speleological Society News 44(8): 301304.
  • Helsel DR, Frans LM 2006. Regional Kendall test for trend. Environmental Science and Technology 40(13): 40664073, DOI: 10.1021/es051650b.
  • Helsel DR, Hirsch RM 2002. Statistical Methods in Water Resources. U.S. Geological Survey Techniques of Water-Resources Investigations, A3: 524 pp. Available online at: http://water.usgs.gov/pubs/twri/twri4a3/.
  • Helsel DR, Mueller DK, Slack JR 2006. Computer Program for the Kendall Family of Trend Test. Report: 2005-5275. U.S.Geological Survey Scientific, 4 pp.
  • Hirsch RM, Slack JR. 1984. A nonparametric trend test for seasonal data with serial dependence. Water Resources Research 20: 727732.
  • Hirsch RM, Slack JR, Smith RA. 1982. Techniques of trend analysis for monthly water quality data. Water Resources Research 18: 107121.
  • Hoyos M, Soler V, Canaveras JC, Sanchez-Moral S, Sanz-Rubio E 1998. Microclimatic characterization of a karstic cave: human impact on microenvironmental parameters of a prehistoric rock art cave (Candamo Cave, northern Spain). Environmental Geology 33(4): 231242, DOI: 10.1007/s002540050242.
  • Kaufmann G 2003. Stalagmite growth and palaeo-climate: the numerical perspective. Earth and Planetary Science Letters 214(1-2): 251266, DOI:10.1016/S0012-821X (03) 00369-8.
  • Kowalski AS, Serrano-Ortiz P, Janssens IA, Sanchez-Moraic S, Cuezva S, Domingo F, Were A, Alados-Arboledas L. 2008. Can flux tower research neglect geochemical CO2 exchange? Agricultural and Forest Meteorology 148(6-7): 10451054, DOI:10.1016/j.agrformet.2008.02.004.
  • Lario J, Sanchez-Moral S, Cuezva S, Taborda M, Soler V. 2006. High Rn-222 levels in a show cave (Castanar de Ibor, Spain): proposal and application of management measures to minimize the effects on guides and visitors. Atmospheric Environment 40: 73957400, DOI:10.1016/j.atmosenv.2006.06.046.
  • Mateu J, Montes F, Fuentes M. 2003. Recent advances in space-time statistics with applications to environmental data: an overview. Journal of Geophysical Research-Atmospheres 108(D24): 8774, DOI: 10.1029/2003JD003819.
  • Perrier F, Le Mouel JL, Poirier JP, Shnirman MG 2005. Long-term climate change and surface versus underground temperature measurements in Paris. International Journal of Climatology 25(12): 16191631, DOI: 10.1002/joc.1211.
  • Perrier F, Richon P, Crouzeix C, Morat P, Le Mouel JL. 2004. Radon-222 signatures of natural ventilation regimes in an underground quarry. Journal of Environmental Radioactivity 71(1): 1732, DOI:10.1016/S0265-931X (03) 00117-6.
  • Perrier F, Richon P, Gautam U, Tiwari DR, Shrestha P, Sapkota SN. 2007. Seasonal variations of natural ventilation and radon-222 exhalation in a slightly rising dead-end tunnel. Journal of Environmental Radioactivity 97: 220235, DOI:10.1016/j.jenvrad.2007.06.003.
  • Ritzi RW, Sorooshian S, Hsieh PA. 1991. The estimation of fluid-flow properties from the response of water levels in wells to the combined atmospheric and earth tide forces. Water Resources Research 27(5): 883893.
  • Salve R, Krakauer NY, Kowalsky MB, Finsterle S 2008. A qualitative assessment of microclimatic perturbations in a tunnel. International Journal of Climatology 28(15): 208120U3, DOI: 10.1002/joc.1697.
  • Sanchez-Moral S, Cuezva S, Lario J, Taborda-Duarte M. 2006. Hydrochemistry of karstic waters in a low-energy cave (Castañar de Ibor, Spain). In Karst, cambio climático y aguas subterráneas, DuránJJ AndreoB, CarreteroJ (eds). Hidrogeología y Aguas Subterráneas n° 18. IGME: Madrid, Serie; 339347.
  • Sanchez-Moral S, Soler V, Canaveras JC, Sanz-Rubio E, Van Grieken R, Gysels K. 1999. Inorganic deterioration affecting the Altamira Cave, N Spain: quantitative approach to wall-corrosion (solutional etching) processes induced by visitors. Science of Total Environment 243-244: 6784.
  • Smith CL, Baker A, Fairchild IJ, Frisia S, Borsato A. 2006. Reconstructing hemispheric-scale climates from multiple stalagmite records. International Journal of Climatology 26(10): 14171424, DOI: 10.1002/joc.1329.
  • Spötl C, Fairchild IJ, Tooth AF. 2005. Cave air control on dripwater geochemistry, Obir Caves (Austria): Implications for speleothem deposition in dynamically ventilated caves. Geochimica et Cosmochimica Acta 69(10): 24512468, DOI:10.1016/j.gca.2004. 12.009.
  • Vanmarcke EH. 1983. Random Field: Analysis and Synthesis. Massachusetts Institute of Technology Press: Cambridge, MA.