Surface and cave climatology: annual cycles and thermal diffusivity

The mean annual precipitation recorded at Altamura in the last 20 years was 622 ± 98 mm. Rainfall is concentrated in the autumn and winter months (mean precipitation 63 mm/month), while the summer months (JJA) are the driest (32 mm/month). Mean annual temperature (MAT) at Altamura for the last 60 years (mean 14.85 ± 0.87°C) showed a steady increase of more than 2°C from 1980 to 2015, while a decreasing trend was observed from 2015 to 2020 (Fig. 3). High temperatures coupled with low precipitation during the warm season result in a negative budget between rainfall and potential evapotranspiration with almost no infiltration May–September, while most of the total infiltration (94% of the total 250 mm/yr) occurs during the late autumn and early spring, from November to March (Fig. 3). The hydrology of Lamalunga Cave reflects the surface observations: a stalactite drip rate monitored in the northern end of the cave between 2017 and 2020 revealed that active infiltration starts in early November and typically lasts until May. In winter and spring, the infiltration is triggered by rainfall events >15–20 mm, while during the dry season the rainfall events triggering infiltration peaks in the stalactite record usually exceed 40 mm.

Figure 3. Left panel: mean monthly rainfall, infiltration, and temperature at Altamura for the period 1960–2020. Right panel: mean annual temperatures at Altamura 1960–2020. The blue line represents the 3-year running mean, while the red and the black horizontal lines indicate the mean annual temperature between August 2017 and August 2020 in Lamalunga Cave towards the South End (red) and in the Abside chamber (black).

The surface MAT during the period 2017–2020 (mean: 15.53 ± 7.70°C) follows nearly sinusoidal cycles with maximum values around 24 July and minimal values around 23 January each year, with a thermal amplitude Ta = ± 4.95°C (Fig. 4). The temperature cycles of the 3-year dataset were modeled with the sinusoidal function:

(1)$$T_x = {\rm MAT} + {\rm A}_{\rm z} \bullet \;\sin \left({\displaystyle{{2\pi t} \over {365.242}}} \right)\;-\varphi _z + \varepsilon $$

Figure 4. Three-year record of air temperature at five locations in Lamalunga Cave compared to the surface air temperature record (Altamura station). The dashed lines indicate the modeled temperatures calculated with sinusoidal functions (see text). The record for the South End (S-End) is limited to the first year due to data logger malfunction. The triangles mark the visits inside the cave that lasted about 6 hours/day for periods between one and four days. The sudden cave temperature anomalies in August 2017, April 2018, October 2018, and April 2019 are related to visits to the cave.

where Tx is the temperature at any time t (in days), MAT is the mean annual temperature, A_z is the amplitude of the annual temperature excursion, 365.242 is the period representing the duration in days of the tropical year, φ_z is the is the phase shift and ɛ is the initial phase shift of the cycle expressed in radians. The optimal parameters of the three variables MAT, A_z, and φ_z were calculated interactively with the least-squares method by minimizing the sum of squared residuals between the modeled and the measured temperatures during the 3-year observational period.

The temperature measurements in the five locations inside the cave exhibit a much clearer sinusoidal cyclicity, with different mean annual values, phase shift, and thermal attenuation with respect to the external temperature signal (Fig. 4). By using the same approach as for the external temperature, we modeled the sinusoidal temperature cycles by fitting the time-series with the least-squares method.

(2)$${\rm T}_{\rm x} = {\rm T}_{\rm z} + A_z\;\bullet \sin \left({\displaystyle{{2\pi t} \over {365.242}}} \right)\;- \varphi _z + \varepsilon $$

Where Tx is the temperature at any time t (in days), Tz is the mean annual temperature at depth z, and A_z is the amplitude of the annual temperature excursion. The calculated parameters document a decrease in the thermal amplitude A_z and increase in the phase shift (ɛ) as a function of increasing rock cover above the ceiling of the cave (Table 2).

Near-sinusoidal annual cycles in cave air temperature have been documented in several studies, and are particularly clear in shallow caves (Buecher, Reference Buecher1999; Mattey et al., Reference Mattey, Fairchild, Atkinson, Latin, Ainsworth, Durell and Roger-Son2010; Domínguez-Villar et al., Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013, Reference Domínguez-Villar, Krklec and Sierro2023; Breitenbach et al., Reference Breitenbach, Lechleitner, Meyer, Diengdoh, Mattey and Marwan2015; Guerrier et al., Reference Guerrier, Doumenc, Roux, Mergui and Jeannin2019; Sainz et al., Reference Sainz, Fábrega, Rábago, Celaya, Fernandez, Fuente, Fernandez, Quindos, Arteche and Quindos2022; Perrier et al., Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023). Domínguez-Villar et al. (Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013) studied the attenuation of the annual thermal amplitude in a shallow barometric cave in central Spain and highlighted an inverse-linear relationship between the thickness of rock/soil and the natural logarithm of the thermal amplitude at each site (Fig. 5b). Similarly, heat transmission in conduction-dominated soil and rock profiles exhibits a linear correlation between the phase shift and rock/soil thickness related to the exponential amplitude attenuation with depth (Smerdon et al., Reference Smerdon, Pollack, Enz and Lewis2003; Fig. 5a). In fact, the propagation of a temperature signal as a function of time (t) and depth (z) is described by the function:

(3)$${\rm T}( {{\rm z}, \;{\rm t}} ) = {\rm T}_{\rm a}{\rm e}^{- kz}\cos ( {\omega {\rm t\ } + \varepsilon - kz} ) $$

Figure 5. Correlation between rock cover and phase shifts (a) and between rock cover and thermal amplitude (b) for the cave temperature time-series with the corresponding k values. The slope of the regression line is compared with the slopes calculated from a shallow cave in central Spain (gray dash–dot line: Domínguez-Villar et al., Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013), the entrance narrow gallery of Los Pilones Cave, Spain (gray dashed line: Domínguez-Villar et al., Reference Domínguez-Villar, Krklec and Sierro2023), and in a soil profile in North Dakota, USA (gray dotted line: Smerdon et al., Reference Smerdon, Pollack, Enz and Lewis2003) (see text).

where T_a and ω are the amplitude and angular frequency of the harmonic surface temperature, respectively, k is the wave vector, and ɛ is the initial phase of the signal (Carslaw and Jaeger, Reference Carslaw and Jaeger1959). Following the approach of Smerdon et al. (Reference Smerdon, Pollack, Enz and Lewis2003) and Domínguez-Villar et al. (Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013, Reference Domínguez-Villar, Krklec and Sierro2023), we calculated the wave vector k from the slopes of the linear regressions between the rock thickness (depth z), the natural logarithm of the thermal amplitude ω, and the phase shift ɛ of the annual signal (Fig. 5a, b):

(4)$${\rm Phase\ shift\ }( {{\rm rad}} ) = 0.176{\it z} + 0.507\;\to \;( {{\rm R}^2 = 0.994} ) $$

(5)$${\rm Ln\ }( {{\rm T}_{\rm a}} ) ( {^\circ {\rm C}} ) = 0.183z + 1.673\;\to \;( {R^2 = 0.995} ) $$

The solution of equation (3) represents a thermal wave with wave vector k = (π/Pκ)^1/2, where P and κ are the period of oscillation and the thermal diffusivity of the material, respectively. This formula allows the calculation of thermal diffusivity κ by using the regression slope of both datasets and the corresponding wave velocity v = (4πκ/P)^1/2 (Table 2).

The calculated mean wave vector k = 0.178 ± 0.01/m, the thermal diffusivity κ = 30.8 ± 2.3×10⁻⁷ m²/s and the wave velocity v = 34.2 ± 1.3 m/yr are much higher that the values calculated in soil profiles where the heat transmission is dominated by conduction and characterized by k values around 0.5 (Fig. 5). Less-steep slopes (k = 0.4–0.3) can indicate air-filled cavities and passages in the host rock and/or the contribution of thermal convection, while the slope approaches 0 for tunnel caves and highly ventilated passages dominated by convective airflow (Smerdon et al., Reference Smerdon, Pollack, Enz and Lewis2003; Badino, Reference Badino2010; Domínguez-Villar Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013). For sites influenced only by thermal conduction the wave vector k is mostly related to the thermal diffusion coefficient of the rock/soil cover above the ceiling of the cave and, therefore, is typical for each bedrock/soil cluster as a function of their lithology, porosity, and water content. On the other hand, where thermal convection is also operating, the wave vector k as well as the thermal diffusivity integrate also the contribution of the air flow.

In shallow limestone caves with negligible ventilation, dominated by thermal conduction, thermal diffusivity has been quantified as 7.6 × 10⁻⁷ m²/s (Domínguez-Villar et al., Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013), 8.0 × 10⁻⁷ m²/s (Guerrier et al., Reference Guerrier, Doumenc, Roux, Mergui and Jeannin2019) and 5.1 × 10⁻⁷ m²/s (Domínguez-Villar et al., Reference Domínguez-Villar, Krklec and Sierro2023). These values fall within the range of carbonate rocks ranging from 3.1 × 7.6 × 10⁻⁷ m²/s to 15.3 × 7.6 × 10⁻⁷ m² /s (Cermak et al., Reference Cermak, Huckenholz, Ryback, Schmid, Schopper, Schuch, Stoffler and Wohlenberg1982). In a recent study, Domínguez-Villar et al. (Reference Domínguez-Villar, Krklec and Sierro2023) explained the low thermal diffusivity at Los Pilones Cave, Spain (5.1 × 7.6 × 10⁻⁷ m²/s) as a consequence of the high water content of the host rock related to its enhance porosity (thermal diffusivity of water = 1.4 × 7.6 × 10⁻⁷ m²/s), while higher thermal diffusivity would be associated with air-filled porosity (dry climate settings) given that the κ value of air at standard temperature and pressure is ~186 × 10⁻⁷ m²/s (Massman, Reference Massman1999).

The thermal diffusivity calculated for Lamalunga Cave (κ = 30.8 ± 2.3 10⁻⁷ m²/s) is well above the documented range for caves dominated by thermal conduction as well as for carbonate rocks. The reason for the anomalously high thermal diffusivity may be associated with cavity and air-filled porosity in the limestone above the ceiling, as well as derived by the minor contribution of thermal advection in the cave galleries. In this case, the thermal diffusivity should not be referred to the encasing limestone but to the whole rock–air ensemble.

However, the calculation of thermal diffusivity assumes a semi-infinite space as a boundary condition, and if this condition is not satisfied, the thermal diffusivity cannot be calculated by this method. For example, Smerdon and Stieglitz (Reference Smerdon and Stieglitz2006) explored the heat transport of harmonic temperature signals in Earth's shallow subsurface, which highlighted that the temperature oscillations are much less attenuated with respect to the infinite half space scenario in the presence of a lower boundary. In the case of Lamalunga Cave, the lower boundary can be related to the presence of unexplored large deeper galleries connected through fissures to the cave, although no evidence for this currently exists. On the other hand, the influence of a phreatic level close to the cave can be ruled out because the phreatic level in this part of the High Murge plateau is more than 200 m below the surface (Parise, Reference Parise2011).

Whatever the reason, Lamalunga Cave exhibits a reduced attenuation of the seasonal signal with respect to other published cave sites, which allows detection of annual temperature cycles as deep as 25 m below the surface.

The influence of water infiltration

When comparing MAT between cave sites and surface, it is noteworthy that all cave MATs are significantly higher than the surface MAT. This is particularly true for the South Gallery, where MAT is 0.7°C higher with respect to the surface MAT recorded during the same period. This result should be expected because the hill slope under which the cave developed lacks arboreal vegetation cover and is facing SE, thus receiving a significantly higher amount of solar radiation (Domínguez-Villar et al., Reference Domínguez-Villar, Fairchild, Baker, Carrasco and Pedraza2013). On the other hand, the Abside chamber displays systematically lower (−0.4°C) MAT with respect to the South Gallery. The lower MAT in the Abside, with respect to any other measured point on the cave, likely is related to the fact that most infiltration is concentrated in the Abside chamber and, to a minor extent, in the northern galleries, whereas infiltration in the South Gallery is very limited. Given that the infiltration occurs almost exclusively during the winter months (November–March), it produces a “forced cooling” in the galleries where it occurs. Differential heating (or cooling) in the cave MAT with respect to the surface MAT of up to 3°C has been documented in areas where the infiltration is concentrated during the summer or winter months (Badino, Reference Badino2010). This is confirmed by Lamalunga Cave dripwater temperature measurements (average 14.84 ± 0.41°C), which are systematically lower with respect to the average air temperature (16.21 ± 0.48°C) along the cave main gallery (Table 1).

Table 1. Statistics of Lamalunga Cave temperature measurements and corresponding modeling parameters of the sinusoidal functions for the period 2017–2020.

¹ Rock cover includes bedrock and soil (where present).

² Distance from entrance includes the vertical distance along the Entrance pit.

³ Standard deviations (SD) of temperature residuals are calculated with respect to the running 24-h average values.

⁴ Phase shifts are calculated with respect to the date of the maximum temperature at the surface.

⁵ Rock cover above the measuring point along the entrance pit was arbitrarily set at 0.05 m as representative of the thermal characteristics of the double-steel trapdoor.

⁶ Temperature measurements in the South End lasted from August 2017 to July 2018.

⁷ Average value of Sala della Jena, South Gallery, and South End series.

Cave ventilation, relative humidity, and CO₂ concentration

A few series of spot measurements were performed at 11 points along the main axis of the cave in order to verify the temperature-logger readings and to characterize the seasonal variability of microclimate parameters (temperature, relative humidity, and air pCO₂). By considering a temperature phase shift of around three months in the northern part of the cave (which exhibits greater seasonal temperature fluctuations), the expected warmer period should occur in October–November, while the cooler period should be in April–May (Fig. 4, Table 2). These shifts are reflected in the temperature profiles. In fact, by excluding the clear perturbation in the area near the base of the entrance pit, a nearly exponential decrease from the homothermic South Gallery towards the Abside chamber in April can be observed, and a corresponding nearly exponential increase occurs in October (Fig. 6). By contrast, quasi-homothermic conditions characterize the entire cave in late January and late July (Figs. 4, 6). The relative humidity profiles follow the surface seasons because they are influenced by rainfall and evapotranspiration, with minimum values in August and near-saturated conditions in winter and spring.

Table 2. Wave vector, thermal diffusivity, and wave velocity calculated from the linear regression of the natural logarithm of the thermal amplitude and the from phase shift as a function of the thickness of rock/soil cover.

Figure 6. (a) Temperature, (b) CO₂, and (c) relative humidity profiles in different months along the main axis of the cave (see Fig. 1). The distances are calculated from the Entrance Hall towards the Abside chamber at the NE end of the cave. The dashed horizontal lines on the graphs represent the mean annual values for the inner parts of the cave. In (b) the black line represents the concentration profile on 15 April 2019 measured approximately three hours after the opening of the entrance trapdoor in correspondence with a rapid rise in the surface air pressure (see text).

Only four series of air pCO₂ could be performed along the axial part of the cave (Fig. 6b). The CO₂ concentrations are relatively stable in the South Gallery with values around 8000 ppmv in April 2018 and 4000 ppmv in August 2017. CO₂ concentrations tend to gradually increase/decrease in the northern part of the cave.

Cave entrance opening and transient phenomena

The opening of the trapdoor at the cave entrance affects all the microclimate variables and is particularly evident when the atmospheric pressure at the surface is high or increasing rapidly. In order to assess its effect, and the microclimate perturbations related to the presence of visitors inside the cave, we monitored the temperature fluctuations with respect to cave CO₂ concentration, relative humidity in the entrance pit, and the atmospheric surface air pressure (SAP).

During the visits on 23–26 October 2018 the surface temperature was slightly lower than the cave temperature, which caused short-lived negative temperature peaks along the South Gallery during the inflow events of 23 October 2018 and 25 October 2018 (Fig. 7) when the SAP was high or rising. On the other hand, on 24 October 2018 and 26 October 2018 the atmospheric pressure was low and/or lowering, and the air current was reversed with flux estimated around 1 m³/s in the key-hole passage at the base of the entrance pit exiting the cave, which caused a modest negative peak in the entrance pit temperature. During four October visits, two persons were present most of the time in the Abside chamber, which caused transient temperature rises of up to 2°C, although the air temperature inside the Abside chamber switched back to pre-event values in 24 hours.

Figure 7. Temperature fluctuations during the four visits to the cave 24–27 October 2018 compared to the external air pressure and mean daily temperature from the Altamura station. During each visit, the duration of which is indicated by gray bars at the top of the graph, the trapdoor at the entrance remained open, thus enhancing the cave ventilation. The pale blue vertical bars and the blue arrows indicate the sucking of cold air inside the cave.

In the period 7–9 August 2017 the rising atmospheric pressure during the first two visits (7 and 8 August 2017) triggered the sucking of warm air inside the cave (Fig. 8), with a steady air inflow detected in the key-hole passage at the base of the entrance pit with estimated fluxes around 1 m³/s. This caused a slight air temperature rise in the areas near the base of the entrance pit (Entrance Hall and Sala Jena), whereas no appreciable temperature change was detected in the deeper part of the cave (South End). The 07 August 2017 event lasted only 1 hour and resulted in a short negative peak in the CO₂ concentration, whereas on 08 August 2017 the sudden atmospheric pressure rise triggered an inflow event lasting around 3 hours that resulted in a dramatic fall in the CO₂ concentration from 4500 to less than 1000 ppmv in the Entrance Hall. On the other hand, on 09 August 2017 the atmospheric pressure was high but lowering, which caused only a short-lived inflow event with a negligible negative peak in the CO₂ concentration that was swiftly canceled by the reversing air flow (Fig. 8). In all the three visits, the cave CO₂ concentration and temperature switched back to the pre-event values in 4–6 hours as soon as the trapdoor was closed.

Figure 8. Microclimate regime during the visits to the cave in August 2017. (a) Temperature and relative humidity (reverse scale) fluctuations in the entrance pit; (b) CO₂ concentration fluctuations in the Entrance Hall; (c) variations in atmospheric pressure at Altamura station (hourly values); (d) temperature fluctuations in the Entrance Hall and Sala della Jena. During each visit, the duration of which is indicated by three gray bars at the top of the graph, the trapdoor at the entrance remained open. The pink vertical bars and the red arrows indicate the sucking of warm and low-CO₂ air inside the cave, whereas the blue arrows indicate the outflow of cold air, as detected in the key-hole passage at the base of the entrance pit. Note that the temperature peaks in Sala Jena were also enhanced by the presence of visitors.

A similar situation occurred on the visit of 15 April 2019, which was caused by a steady rise in the SAP (0.5 mbar/hour) that triggered a fast, cold air inflow from the entrance pit with estimated flux around 1 m³/s that lasted about two hours. This resulted in a rapid decrease in the CO₂ concentration in the entire South Gallery from 5040 ± 390 ppmv to 740 ± 150 ppmv and from 5300 to 1900 ppmv in Sala della Jena (Fig. 6).

Ventilation is usually assessed in caves by using Radon-222 (Kowalczk and Froelich, Reference Kowalczk and Froelich2010), as well as by studying transient CO₂ episodes (Faimon et al., Reference Faimon, Troppová, Baldík and Novotný2012; Kukuljan et al., Reference Kukuljan, Gabrovšek, Covington and Johnston2021; Sainz et al., Reference Sainz, Fábrega, Rábago, Celaya, Fernandez, Fuente, Fernandez, Quindos, Arteche and Quindos2022). If the seasonal CO₂ budget has to take into account the CO₂ outgassing from dripwater entering the cave, as well as other potential in-cave sources (Buecher, Reference Buecher1999; Mattey et al., Reference Mattey, Atkinson, Barker, Fisher, Latin, Durrell and Ainsworth2016), these latter are not relevant during rapid ventilation episodes in which surface air with atmospheric CO₂ concentration is forced inside the cave. Of particular interest is the study of d'Aven d'Orgnac (Southern France), a barometric-descending cave with a vertical development similar to Lamalunga Cave, where continuous monitoring of temperature, relative humidity, Radon-222, and CO₂ concentration allowed the detection of “emptying events” during autumn and winter (Bourges et al., Reference Bourges, Mangin and d'Hulst2001, 2006). These events are triggered by significant SAP variations and cause the almost instantaneous (less than few hours) decrease in cave air temperature, relative humidity, Radon-222, and CO₂ concentration. The subsequent filling periods, during which the pre-event conditions are restored, last about one week. Similar emptying–recharging events in Radon-222 and CO₂ concentration have been described in the Polychrome Room of Altamira Cave (northern Spain), although in this case the events lasting one to more than seven days are symmetrical and triggered by temperature gradients between the surface and the cave air (Sainz et al., Reference Sainz, Fábrega, Rábago, Celaya, Fernandez, Fuente, Fernandez, Quindos, Arteche and Quindos2022).

By applying a simple mass-balance model for the 15 April 2019 event, we can observe that the estimated air inflow for 2 hours (~7000 m³) exceeded the mapped volume of the entire cave. The fast and drastic drop in CO₂ concentration, especially in the South Gallery (volume estimate 4400 m³), which accounts for most of the mapped cave volume, is not compatible with a mixing model. Therefore, other air passages acting as air outflows have to exist, albeit the presence of any other sizable entrances to the cave can be excluded because no entrances are visible in the the barren karst landscape above the cave (Fig. 1b). However, the sporadic discovery in the northern galleries of coprolites of small mammals seems to suggest the existence of cm-wide passages connecting the surface, while the asymmetric lowering of the CO₂ concentration between the South and the northern galleries suggests that forced outflow from the cave preferentially occurs in the South Gallery. This situation is similar to that of Obir Cave (Austria) and Aven d'Orgnac where ventilation occurs through karst micro- and macrofissures and a network of inaccessible (or unexplored) underground passages (Spötl et al., Reference Spötl, Fairchild and Tooth2005; Bourges et al., Reference Bourges, Genthon, Mangin and D'Hulst2006).

Daily to sub-daily temperature cycles

Perturbations and daily to sub-daily cycles in the cave microclimate parameters are well known in the literature, and can affect air temperature, relative humidity, dripwater flow rate, radon-222, CO₂ concentration, and barometric air pressure (Genty and Deflandre, Reference Genty and Deflandre1998; Perrier et al., Reference Perrier, Morat and Le Mouël2001, Reference Perrier, Richon, Crouzeix, Morat and Le Mouël2004, Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023; Sondag et al., Reference Sondag, van Ruymbeke, Soubiès, Santos, Somerhausen, Seidel and Boggiani2003; Bourges et al., Reference Bourges, Genthon, Mangin and D'Hulst2006; Richon et al., Reference Richon, Perrier, Pili and Sabroux2009; Drăgușin et al., Reference Drăgușin, Tîrlă, Cadicheanu, Ersek and Mirea2018; Sekhon et al., Reference Sekhon, Novello, Cruz, Wortham, Ribeiro and Breecker2021; Gomell and Pflitsch, Reference Gomell and Pflitsch2022). In underground environments and, particularly, in barometric caves, daily to sub-daily cycles are induced by atmospheric pressure variations (APV), which are characterized by strong 24-hour (S1 = 1.0 cpd) and 12-hour (S2 = 2.0 cpd) solar harmonic frequencies (Perrier et al., Reference Perrier, Morat and Le Mouël2001; Sondag et al., Reference Sondag, van Ruymbeke, Soubiès, Santos, Somerhausen, Seidel and Boggiani2003; Richon et al., Reference Richon, Perrier, Pili and Sabroux2009; Mentes, Reference Mentes2018). Although APV can cause thermodynamic adiabatic thermal changes in the air that exceed 0.1°C/hPa, the thermal amplitude is damped by heat inertia of the rock walls and can be significantly modified by water phase changes in near-saturated conditions (Bourges et al., Reference Bourges, Genthon, Mangin and D'Hulst2006; Perrier et al., Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023). APV can induce barometric tides S1 and S2 not only in the cave atmospheric pressure and air temperature, but even in cave rock temperatures, although with amplitudes 10 times smaller than for cave air temperature (Perrier at al., Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023). Pressure-induced temperature (PIT) variations are particularly relevant near the cave entrances, producing temperature variations up to 1°C/mBar, while in more-confined galleries the PIT transfer function decreases to few 0.001°C/hPa and increases with the distance to the closest cave wall (Perrier et al., Reference Perrier, Morat and Le Mouël2001, 2010, Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023).

In their recent study of four barometric caves in France, Perrier et al. (Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023) calculated that the modulus, at barometric tide S2, varies from 0.002–0.014°C/hPa, with a median value of 0.0048°C/hPa. In particular, along the Red Corridor of Pech Merle Cave, a dead-end horizontal gallery with a topographic trend similar to that of Lamalunga Cave South Gallery, the PIT modulus at barometric tide S2 is relatively constant throughout the year (~0.0036 ± 0.0007°C/hPa) (Perrier et al., Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023). These small semi-diurnal temperature cycles have been explained by PIT changes that diminished in less than one hour as a result of thermal exchanges with the surrounding rock walls (Bourges et al., Reference Bourges, Genthon, Mangin and D'Hulst2006). Propagation of the APV underground can be delayed and dampened as a function of the distance from the entrance, the morphology of the entrance, and the topographic trend of the galleries. In their study of Wind Cave and Jewel Cave in South Dakota, Gomell et al. (Reference Gomell and Pflitsch2022) calculated delays between 38 minutes and 2 hours 22 minutes for distances between 0.4 km and 1.5 km from the entrance.

The time-series of Lamalunga temperature residuals (calculated as differences between the hourly measurements and the 24-hour means) reveals the occurrence of PIT variations at all the monitored sites (Fig. 9). Apart from the Entrance pit, where large daily temperature variations (± 0.523°C) are mainly due to the barometric thermal advection of surface air, the PIT variations in the inner cave sites have similar absolute amplitudes with the highest values in the South Gallery (± 0.0054°C), intermediate values in the Sala della Jena (± 0.0035°C) and South End (± 0.0037°C), and the lowest values in the Abside chamber (± 0.0021°C). Although these values are only 1.5–4 times the reported instrumental resolution (0.0014°C), the systematic coherence of the different data loggers and the reproducibility of the signals in three consecutive years corroborate the findings. Moreover, the absolute amplitudes of the PIT variations are remarkably similar to the values observed in other barometric caves in France (Bourges et al., Reference Bourges, Genthon, Mangin and D'Hulst2006; Perrier et al., Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023). Each site, however, is characterized by a different pattern and seasonality: the Abside chamber and the South Gallery display the highest daily thermal fluctuations in the winter season (December–May) when most of the water infiltration occurs, while Sala della Jena displays higher amplitude in the summer season (June–September), which is similarly to the pattern at the Entrance pit (Fig. 9). Conversely, the South End (not shown) exhibits almost constant daily thermal fluctuations throughout the year.

Figure 9. Time-series of surface air temperature and pressure at Altamura station (24-h moving window average) compared to Lamalunga infiltration record, cave temperatures, and their corresponding temperature residuals (dT = differences from the 24-h averages). The pale blue shading indicates winter seasons when the surface temperature was significantly cooler that the cave temperature, while the pink shading indicates summer seasons when the surface temperature was significantly warmer that the cave temperature.

Spectral analysis

In barometric caves the spectral analysis of temperature time-series can be compared with the corresponding SAP time-series and gives precious information about the thermal confinement of the galleries (Perrier at al., 2023). In order to account for different seasonal behavior, we performed fast Fourier transform (FFT) analysis on series constructed with the data of two consecutive winter (DJF) and summer (JJA) seasons (Fig. 10).

Figure 10. Winter and summer fast Fourier transform for: (a) surface air pressure at Altamura, (b) Lamalunga entrance temperature, (c) Sala della Jena temperature, and (d) South End temperature. Each series is constructed with the data of two consecutive winter (DJF) and summer (JJA) seasons. The black line in each series represents the 90% confidence level.

As expected, the results for the Entrance pit are very similar to the SAP, with a clear barometric S1 (24-h) signal, a most prominent S2 (12-h) signal followed by weaker S3 (8-h) and S4 (6-h) harmonic peaks, as already observed in many French caves (Perrier et al., Reference Perrier, Bourges, Girault, Le Mouël, Genty, Lartiges, Losno and Bonnet2023). The presence of a clear S1 peak confirms that the Entrance pit is thermally influenced by daily temperature cycles, while the lack of S1 peaks in all the cave sites confirms that most parts of the main gallery are thermally confined. Other than the S1 peak, the spectra of the South End are very similar to the spectra of Entrance pit, with S3 and S4 peaks most prominent in the winter season. In the South Gallery (not shown) the FFT spectra are similar to those of the South End, but with only a very weak S2 signal. In contrast, Abside chamber (not shown) and Sala della Jena (Fig. 10c) display a high variance range and very weak S2 signal. This difference possibly is related to the position of the room at the intersection of two galleries, and/or to the periodic water infiltration, with subsequent water films flowing on the rock surface, and to evaporation–condensation episodes.

Seasonal variations, phase shifts, and delays

In order to evaluate seasonal behavior, phase shift, and delays in the daily temperature cycles we used a graphical approach, which is better suited to illustrate the complexity of the mechanisms acting simultaneously in different parts of the cave and during different seasons. For this, we analyzed three time slices: (1) January, when the surface temperature is lower and the cave is almost in homothermic condition; (2) June, when the surface temperature is high but the northern part of the cave (Sala della Jena and Abside) has the lowest temperatures; and (3) September, when the external temperature is still high and temperatures in the northern part of the cave are close to their maximum values (Fig. 4).

Comparison between the surface air pressure residuals (SAPr) and the Entrance pit temperature residuals (TEPr) revealed that the temperature cycles are synchronous (within the range of the acquisition frequency of 1 hour) and mimic, in most details, the SAP variations (Fig. 11a, e, i). The SAP-to-air temperature transfer function varies from 0.5–1.0°C/mBar in summer and from −0.2°C/mBar to −0.32°C/mBar in winter, which confirms that thermal advection is actively operating along the entrance pit. Some details, however, can differ as a result of the distance from the cave of the Altamura meteorological station (6 km). For this reason, we used the Entrance pit TEPr as a proxy for the SAPr at the cave site in order to compare the temperature residuals in the inner part of the cave. This hypothesis was corroborated by the fact that temperature residuals in the internal cave sites systematically have a higher correlation coefficient with TEPr than with SAPr.

Figure 11. Time-series of surface air pressure (Altamura station) and cave temperature residuals during three significant periods representing (a–d) the winter season (January 2018), (e–h) early summer (June 2018), and (i–l) late summer (September 2017) (see text). Note the inverted scale on the January secondary axes (a–d).

Sala della Jena temperature residuals (TJEr) exhibited larger daily fluctuations in summer, especially in June and July (Fig. 9), while for the rest of the year the daily fluctuations were more muted (Fig. 11b, f, j), which is similar to the signal in the TEPr. The SAP-to-air temperature transfer function varies from 0.0013–0.003°C/mBar in summer and from 0.0012–0.0015°C/mBar in winter, suggesting a minor influence of barometric thermal advection especially in summer. The South Gallery temperature residuals (TSGr), on the other hand, exhibit a more complex behavior with the best correlation to TEPr in winter and late summer but no clear correlation in the intermediate seasons (Fig. 11 c, g, k). The South End temperature residuals (TSEr) are correlated more consistently to TEPr throughout the year, regardless of the thermal conditions at the surface (Fig. 11d, h, l), with SAP-to-air temperature transfer function varying from 0.0048–0.0033°C/mBar in summer and from 0.0018–0.0024°C/mBar in winter. Only in the Abside chamber did temperature residuals (not shown here) have no clear cyclicity without a consistent correlation with both SAPr and TEPr.

The information gathered from quantification of the thermal diffusivity, the analyses of the annual temperature cycles, the FFT spectra, and the study of the temperature residuals allow detailed description of barometric air flow inside Lamalunga Cave and the possible contribution of barometric and thermal advection. We summarized the results in two diagrams that represent the winter and summer situations (Fig. 12). In summer, the overpressure created by rising SAP gently pumps warm air inside the cave with strong thermal advection in the Entrance pit and a limited influence up to Sala della Jena, while in the other parts of the cave no measurable thermal advection is detectable, and the daily temperature fluctuations are controlled mainly by PIT variations. The air flux and thermal advection are reversed during under-pressure created by lowering SAP. In January, most of the cave is at near-homothermic condition and the thermal advection is mostly confined to the Entrance pit. However, as towards both cave ends (Sala della Jena and South End) where the SAP-to-air temperature transfer function is systematically higher in summer than in winter, a minimal influence of thermal advection has to operate. Because the relative humidity of surface air is generally much lower than inside the cave, the faint air flow can play a crucial role in the growth and morphology of delicate speleothem formations such as coralloids which grew from hydro-aerosol, evaporation, and capillary flow (Vanghi et al., Reference Vanghi, Frisia and Borsato2017, Reference Vanghi, Borsato, Frisia, Howard, Gloy, Hellstrom and Bajo2019).

Figure 12. Simplified cross section of Lamalunga Cave (vertical exaggeration = 3) with air-flow direction in summer (August) and winter (January) as a function of surface air pressure rise (↑ = overpressure) and fall (↓ = underpressure). Ep = Entrance pit; Se = South End; SG = South Gallery; Eh = Entrance Hall; Je = Sala della Jena; Ab = Abside. The number near each station is the mean daily temperature recorded in each month. The arrows indicate the direction of air flow inside the cave, with the colors indicating the influence of thermal advection: red = warm; blue = cold; black = no detectable thermal advection (see text).

The absence of thermal perturbation during winter, when cold and dense external air can potentially sink inside the cave, testifies to the limited extent of this phenomenon and the mitigation mechanism produced by the Entrance pit whose narrow diameter hinders temperature transmission inside the main gallery of the cave. In this regard, the partial closure of fissures along the trapdoor effectively hinders rapid and dynamic exchange with the surface, although the constriction does not cut off smooth and continuous equilibration with the surface air pressure, thus conveying a tiny convective temperature signal even in the deeper parts of the cave. It has to be noted, however, that the contribution of thermal convection in most parts of the cave is minimal, affecting the thermal regime in the order of magnitude of hundredths of a degree Celsius without adding any significant noise to the temperature records, and this allows the detection of barometric tides throughout the year in all the cave galleries.

Speleothem formation in Lamalunga Cave

Speleothem formation in ventilated galleries depends on a number of variables, which can vary on a seasonal to daily basis and include drip rate, temperature, air CO₂, calcite supersaturation, relative humidity, and the presence of trace elements, impurities, and organic matter (Frisia et al., Reference Frisia, Borsato, Fairchild and McDermott2000). High drip rates and low to moderate calcite supersaturation (SI_C ) are associated with flowstone and dome-shaped or cone-shaped stalagmites, whereas low drip rates and higher calcite supersaturation favor the formation of candle-shape stalagmites (Miorandi et al., Reference Miorandi, Borsato, Frisia, Fairchild and Richter2010). Conversely, in the absence of direct dripping, minute calcite coralloids can grow from hydro-aerosol generated by the fragmentation of drops (popcorn coralloids; Vanghi et al., Reference Vanghi, Frisia and Borsato2017) or from evaporation and through intercrystalline and intercoralloid capillary flow (branching coralloids; Caddeo et al., Reference Caddeo, Railsback, De Waele and Frau2015). Both popcorn and branching coralloids are present in Lamalunga Cave and their distribution is related to the hydrological and microclimate conditions of the different parts of the cave.

Lamalunga Cave water, which was analyzed during 2017 and 2018, revealed substantial spatial and seasonal homogeneity, with low Mg/Ca molar ratio (0.12 ± 0.03) and moderate calcite supersaturation in both drips (SI_C = 0.45 ± 0.05) and pools (SI_C = 0.44 ± 0.08) (Vanghi et al., Reference Vanghi, Borsato, Frisia, Howard, Gloy, Hellstrom and Bajo2019). This suggests a limited modification of dripwater by prior calcite precipitation, and the potential of calcite precipitation on speleothem surfaces. In these situations, ventilation controls the actual SI_C by lowering the air CO₂ during SAP rises, particularly during the fast emptying events.

Given that infiltration is preferentially concentrated in the northern galleries, a higher density of large popcorn coralloids (Vanghi et al., Reference Vanghi, Frisia and Borsato2017) is expected in this part of the cave (Fig. 2a, c). The fabric of these coralloids varies from columnar-elongated to fiber-like in their apical parts, alternating with micritic and microsparitic layers that correspond with growth interruptions and condensed intervals (Vanghi et al., Reference Vanghi, Frisia and Borsato2017). These coralloids are particularly enriched in minor and trace elements and impurities, including Mg, Sr, and Si, organized in cyclic lenticular layers (Vanghi et al, Reference Vanghi, Borsato, Frisia, Howard, Gloy, Hellstrom and Bajo2019), which testify to the more dynamic microclimate and hydrological regime in the Abside chamber. It also has been noted that coralloids in this part of the cave do not show a preferential orientation of growth, which should have been expected in the presence of strong air currents.

On the other hand, reduced infiltration along the South Gallery explains the absence of large popcorn coralloids and the high density of delicate branching coralloids, which grew from evaporation and through intercrystalline and inter-coralloid capillary flow (Caddeo et al., Reference Caddeo, Railsback, De Waele and Frau2015; Vanghi et al., Reference Vanghi, Frisia and Borsato2017) (Fig. 2b, d). These coralloids, made by honey-yellow translucent compact columnar calcite with few impurities and low trace element content, are concentrated preferentially along the edges of limestone debris on the floors of the galleries, often aligned in delicate rows that are spaced several mm apart (Fig. 2b, d). Both globular and dendritic branching morphologies are present, the latter preferentially concentrated on the debris covering the cave floor.

The delicate morphology, low trace-element content, and compact translucent fabric all indicate a very stable environment without significant changes in temperature, relative humidity, and calcite supersaturation (Frisia and Borsato., Reference Frisia, Borsato, Alonso-Zarza and Tanner2010). This, again, confirms the confined state of the South Gallery where the exceptionally stable temperature, the near-saturation relative humidity favored by steady water infiltration during the winter combined with high air CO₂ concentration prevent any fast coralloid growth, and allow the development of delicate and branching morphologies. Given that most of the coralloid growth occurred in the Holocene and 64–36 ka (Vanghi et al., Reference Vanghi, Borsato, Frisia, Howard, Gloy, Hellstrom and Bajo2019), their ubiquitous presence in the South Gallery suggests that stable microclimate conditions persisted for several tens of thousands of years inside the cave.