Climate warming detected in caves of the European Alps

Abstract

Cave air temperatures in four caves in the European Alps show statistically significant warming trends of about 0.2 °C per decade over the last two decades (2000–2020). These trends are about half as large as those observed outside and are characterized by a remarkable spatial and temporal consistency. The investigated caves represent different types in terms of their ventilation regime and one of them also hosts perennial ice. Key observation sites are located in cave sections where the temporal variability of air temperature is strongly attenuated compared to outside conditions and data from different cave sections show that the main results are valid for large parts of the investigated caves. Continued warming will lead to broad changes in alpine cave environments, including changes in strength and direction of air flow in caves, karst hydrology and subsurface ecosystems. The observed subsurface warming has a particular strong effect on the long-term preservation of perennial ice present in some of these caves. This is shown for an ice cave in the Austrian Alps, where enhanced melt of ice correlates with the observed warming. This cave (and similar ones) will not be able to hold perennial ice beyond the next decade.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-78658-y.

Introduction

Atmospheric warming is a global phenomenon of the Anthropocene. Global mean temperatures and temperatures averaged over land areas have warmed by 1.1 °C and 1.6 °C from 1850 to 1900 to 2011–2020, respectively¹. Mountain regions warmed at an even higher rate, which is due to reduced snow and ice feedbacks, reduced local cooling due to shrinking glacier extent, and local land use changes². The European Alps (hereafter Alps for short) have warmed twice as much as the global average between the late 19th and the early twenty-first century^3,4. In recent decades, the average warming rate in the Alps has been 0.3 °C per decade, outpacing the global rate of 0.2 °C per decade⁵.

The environmental impact of this strong atmospheric warming of the Alps is most prominently manifested by the loss of glacier extent and volume^6,7. Parallel to this reduction in surface ice masses, high-alpine regions have also witnessed a degradation of permafrost^8,9. Rising temperatures have also affected surface waters in the Alps¹⁰. A recent study of two catchments in the Tyrolian Alps identified mean warming trends of local streams of 0.24 and 0.45 °C per decade for the time period 1977 to 2020¹¹. This signal has also propagated into the subsurface and led to an increase in groundwater temperatures. Shallow groundwaters in Austria (average depth 7 ± 4 m) increased on average by 0.37 °C per decade between 1994 and 2013^12,13. A study in the northern Alpine foreland (Bavaria, Germany) reported a decrease of the warming amplitude with depth, from 0.28 °C per decade in 20 m to 0.09 °C per decade in 60 m depth between the early 1990s and 2019¹⁴. Rising temperatures have also been detected inside high alpine glaciers, e.g., in the Mont Blanc area, revealing a mean warming rate of 0.14 °C per decade between 1900 and 2004¹⁵. Perrier et al.¹⁶ provided evidence that long-term changes in surface climate have also propagated into urban underground.

Little is known about whether and to what extent this atmospheric warming signal has already “arrived” in the interior of cave systems. Caves are abundant in many part of the Alps where karstifiable rocks (chiefly limestone, dolomite and/or marble) are present. The Alps not only host some of the largest caves in Europe (e.g., Hölloch, Siebenhengste-Hohgant Höhlensystem, Schönberg-Höhlensystem), but also caves with ice deposits many of which showing strong signs of retreat, e.g., Dachstein Mammuthöhle¹⁷, Hundsalm Eis- und Tropfsteinhöhle¹⁸ and Obstanser Eishöhle¹⁹. Warming of underground cavities has been shown to have also significant consequences for subterranean biota and ecosystems^20,21.

Here, we document for the first time decadal-scale warming in the interior of caves across the Alps. Although local time series of temperature measurements in different caves (and parts thereof) exist since many years^22–24, this is the first study that systematically evaluates temperature measurements obtained in four caves in Switzerland and Austria. The aim of this study is to document and quantify the ongoing warming in the interior of these alpine caves and to encourage follow-up research, in order to take full advantage of the wealth of temperature data available for caves elsewhere around the globe.

The microclimate of caves

Caves are underground structures developed in the earth’s lithosphere which can exchange mass and energy with the atmosphere in a variety of ways. The most important processes are related to advection of air and water. Diurnal and seasonal temperature variations that characterize the Earth surface are attenuated in the cave environment. Ignoring latent heat and humidity effects, the attenuation of the diurnal and seasonal temperature amplitudes follows an exponential function from the entrance to the interior of caves²⁵. In the near-entrance part of caves – the heterothermic zone – diurnal and seasonal temperature cycles are well expressed. Deeper into the cave, the attenuation approaches an asymptotic value. This is referred to as the homothermic zone, where diurnal cycles are below detection and also the seasonal temperature amplitude typically does not exceed a few tenths of a degree. This stable temperature is commonly close to the mean annual surface temperature, except where local effects play a role. To exemplify the latter, cold air can be trapped in sag-type cave sections, while turbulent processes in passages with strong ventilation force mixing of air masses and thermal equilibrium with neighbouring regions. In the vicinity of substantial ice bodies the summer air temperature is limited to about 0 °C as long as the ice melts, which manifests in a correspondingly disturbed seasonal cycle and can induce anormal temperature gradients across small distances. Ice or snow can also plug cave sections and thus change local ventilation and thermal conditions.

The air temperature close to snow and ice is strongly controlled by processes exchanging mass and energy at the surface. For many ice caves, the subzero temperature regime is mainly supported by advection of cold air during periods of inflow during winter, when the turbulent fluxes withdraw energy from the surface and the overlying air. Summer conditions normally overcompensate the winter deficit, which is mainly driven by a reversal of the turbulent fluxes. Excess energy finally induces melt, which excerts a strong control on air temperature, since the corresponding energy is mostly supplied by air, whose temperature is kept at 0 °C²⁶. This results in asymmetric seasonal temperature cycles.

A compilation of data from 40 caves in Austria located between 403 and 2347 m a.s.l. revealed that most caves show interior air temperatures close to the multi-annual mean of the outside temperature²⁷. Anomalously low temperatures were measured in caves that are characterized by restricted ventilation and act as cold traps in winter. This study also reported preliminary results of long-term measurements in one alpine cave (Spannagelhöhle), where temperatures had increased since the year 2000 in a near-linear manner. Preliminary data of a similar monitoring work in caves of the Swiss Alps documented rising temperatures since the year 1990²⁸. In contrast to these well-ventilated caves, a study in a side passage of Postojna Cave (Slovenia) with a dead-end gallery with limited air advection (in contrast to the rest of the system) concluded that there the external warming signal was largely transferred into the cave by thermal conduction, resulting in a delay of 20–25 years²⁹. An extreme endmember is a monitoring site in the interior of Bärenschacht, a 60 km-long cave network spanning 946 m of vertical extent. The measurement site is 700 m below the surface with limited air exchange and showed a constant air temperature within 0.1 °C between 1999 and 2017 (W. Janz, pers. comm.).

Study sites

Four caves in different climatological regions across the Alps were selected showing a range in size and elevation (Table 1). Rasslsystem (RS), Spannagelhöhle (SP) and Hundsalm Eis- und Tropfsteinhöhle (HA) are located in different parts of the Austrian Alps, while Schrattenhöhle (SH) is located in the Swiss Alps (Fig. 1).

Table 1.

Key parameters of the studied caves.

	Spannagelhöhle	Schrattenhöhle	Hundsalm Eis- und Tropfsteinhöhle	Rasslsystem
Abbreviation	SP	SH	HA	RS
Cave register number	2515/1	OW 4/61 (22363)	1266/1	3925/9
Coordinates (lat./long.)	47.08°/11.67°	46.77°/8.27°	47.55°/12.03°	46.51°/14.54°
Length of cave (m)	12,000	19,718	264	1085
Vertical extension (m)	329	573	51	161
Classification of cave system	Dynamic	Dynamic	Static with ice	Dynamic
Number of entrances	3	8	1	1
Elevation of (upper) entrance (m a.s.l.)	2523	2090	1495	1085
Name of principle cave monitoring site	Porzellanladen	Windkluft	Jugendgang	Perlenhalle
Abbreviation	SP-PL	SH-WK	HA-JG	RS-PH
Elevation of monitoring site (m a.s.l.)	2408	1743	1480	1070
Rock overburden of monitoring site (m)	17	60	50	95
Distance of monitoring site to nearest entrance	840	250	85	270
Relevant outside weather stations	JE, PK, PL	PI, EN, BO	JE, PK	BA, SE, ZE

Fig. 1.

Locations of the studied alpine caves and the weather stations used for comparison. Abbreviations of cave names: Hundsalm Eis- und Tropfsteinhöhle (HA), Rasslsystem (RS), Schrattenhöhle (SH), Spannagelhöhle (SP). Abbreviations of weather stations: Bad Eisenkappel (BA), Bonistock (BO), Engelberg (EN), Innsbruck (IN), Jenbach (JE), Mayrhofen (MA), Patscherkofel (PA), Pilatus (PI), Plattkopf (PA), Seeberg (SE), Zell-Pfarre (ZE). Map was produced using GoogleEarth (https://www.google.com/intl/de/earth/about/) and Adobe Illustrator (v27.2, www.adobe.com).

Spannagelhöhle is a 12 km-long system in the western Zillertal Alps (Tyrol; Fig. 1) with three upper entrances and unknown lower entrance(s) extending over 329 m in the vertical dimension (Supplementary Fig. S1). The main cave parts are well ventilated and characterized by upward directed air flow in winter and a reversed flow in summer. The main monitoring site, Porzellanladen (SP-PL), is located in the interior of the cave very close to the main conduit where the air is flowing. Additionally, a site in a dead-end side gallery (Tropfsteingang, SP-TG) and one located in the main air flow but closer to the main upper entrance (Gaunhalle, SP-GH) were included in this study (Supplementary Fig. S1). We also present data from a small cave that opens 27 m beneath the main entrance of Spannagel Cave and contained an ice body when discovered in 1998. This so-called Spannagel-Eishöhle (SP-EH) shows a descending geometry and has lost its ice by the year 2002.

Hundsalm Eis- und Tropfsteinhöhle is located in the Northern Calcareous Alps of Tyrol (Fig. 1) and classifies as a sag-type cave given its subvertical geometry and the lack of a lower entrance (Supplementary Fig. S2). HA is also exceptional because the cave consists of two compartments, a higher one that also contains perennial ice, and a lower section, which is ice-free. The two cave parts were artificially connected in 1996 and then subsequently thermally re-separated using an airlock. Our study concerns primarily the lower section (location Jugendgang, HA-JG) whose air temperature is close to the mean air temperature outside (in contrast to the cold trap in the upper cave part). Temperature data obtained in the main shaft and in a short borehole in the upper part of the cave are also considered in this study (Supplementary Fig. S2).

Rasslsystem is one of several caves in the Obir Massif of the Northern Karawanks in SE Austria (Fig. 1). Its air flow regime was modified as a result of mining activities which came to an end at the beginning of the 20th century. An abandoned mining adit acts as the lower entrance of cave system (Supplementary Fig. S3) with strong inward and outward air flow during the cold and warm season, respectively, while the pathways of the air in the upper inaccessible parts of the cave are largely unknown²². The monitoring site is Perlenhalle (RS-PH), a high and elongated terminal chamber in the innermost part of Rasslsystem with no measurable air flow.

Schrattenhöhle is located in the Melchsee-Frutt region of Switzerland (Obwalden; Fig. 1) and comprises an extensive network of mostly well-ventilated galleries extending over 573 m and showing multiple entrances (Supplementary Fig. S4). The monitoring site, Windkluft (SH-WK), is located in the main air flow route 250 m from the next entrance, while the subsidiary site SH-WG (Wermutgang) is located close to the uppermost entrance of the Schrattenhöhle cave system.

Results

We examined daily averages of data measured in the four caves and in the free atmosphere of nearby meteorological stations. The data were subjected to quality control and restricted to the period providing the most consistent records (2000–2020). The main analyses concerned climatological features and relationships with external conditions, trends using different methods including seasonal aspects, spatial and temporal representativeness and the relationship with cave ice developments.

Climatological characteristics

The analyzed records are characterized by (a) different mean values, (b) regular annual cycles of comparable magnitude, and (c) continuously increasing temperatures at all observation sites (Fig. 2).

Fig. 2.

Time series of cave temperatures observed in the four caves during 2000–2020. The partly lower resolution of some records before about 2005 is due to the use of different instruments.

The average conditions in the caves as well as in their outside atmosphere are summarized in Table 2, which refers to the central range of the overall investigation period (2008–2010).

Table 2.

Mean and standard deviation of air temperature inside and outside of the investigated caves and respective differences (2008–2010). Bracketed numbers in the lowest panel denote Pearson correlation coefficients between air temperature in the caves and outside.

	SP-PL	SH-WK	HA-JG	RS-PH
Elevation of cave sites [m a.s.l.]	2408	1743	1480	1070
Cave temperature [°C]
Mean annual	+ 1.82 ± 0.066	+ 3.17 ± 0.064	+ 4.15 ± 0.054	+ 5.47 ± 0.022
Winter (open period)	+ 1.77 ± 0.044	+ 3.12 ± 0.046	+ 4.13 ± 0.055	+ 5.46 ± 0.023
Summer (closed period)	+ 1.86 ± 0.046	+ 3.21 ± 0.044	+ 4.17 ± 0.043	+ 5.47 ± 0.021
Outside temperature [°C] *
Mean annual	+ 0.40 ± 7.551	+ 3.75 ± 7.052	+ 4.18 ± 7.603	+ 4.94 ± 8.214
Winter (open period)	− 6.07 ± 5.18	− 1.42 ± 4.96	− 1.65 ± 4.99	− 1.39 ± 5.37
Summer (closed period)	+ 5.66 ± 5.09	+ 8.79 ± 4.75	+ 9.93 ± 4.83	+ 11.2 ± 5.20
Temperature difference in-out [°C] (correlation coefficient)
Mean annual	+ 1.42 (+ 0.72)	− 0.58 (+ 0.66)	− 0.03 (+ 0.46)	+ 0.53 (− 0.11)
Winter (open period)	+ 6.96 (+ 0.50)	+ 4.54 (+ 0.35)	+ 5.78 (+ 0.44)	+ 6.85 (− 0.06)
Summer (closed period)	− 4.04 (+ 0.31)	− 5.58 (+ 0.25)	− 5.76 (− 0.03)	− 5.73 (− 0.23)

*at the elevation of measurements in the cave.

¹calculated from AWS data JE and PA, bias from comparison with PA.

²calculated from AWS data EN and PI, bias from comparison with BO.

³calculated from AWS data JE and PA, bias from comparison with HA-entrance.

⁴calculated from AWS data BA and SE, bias from comparison with ZE.

The average cave temperatures correspond to the elevations of the cave sites. Lowest values are observed at SP-PL (+ 1.8 °C) being located at an elevation of 2408 m a.s.l. SH-WK (1743 m a.s.l.) is characterized by an average temperature of + 3.2 °C, HA-JG (1480 m a.s.l.) by + 4.2 °C and RS-PH by + 5.5 °C (1070 m a.s.l.). The mean temperatures are positive at all cave sites, as well as those in the atmosphere outside of the caves, which range from + 0.4 °C at the elevation of SP-PL to + 4.9 °C (for RS-PH), respectively. The variability of cave temperatures is almost two orders of magnitude smaller compared to that outside ( < ± 0.1 °C vs. ca. ±8 °C) and increases with elevation (from ± 0.02 to ± 0.07 °C). Thus, RS-PH stands out by a homothermic regime, i.e., quasi-constant temperatures despite of the somewhat larger outside variability. Figure 3 graphically summarizes these basic characteristics in terms of boxplots.

Fig. 3.

Boxplots of observed cave temperatures (left) and outside temperatures calculated for the same elevation (right). Red lines denote medians, blue boxes indicate the 25th and 75th percentiles, whiskers represent extremes and red crosses outliers.

Table 2 includes information about the average conditions during winter and summer, which in the context of cave microclimate are also referred to as the so-called open and closed periods¹⁸. The latter are both characterized by positive air temperatures and all cave sites experienced strongly attenuated seasonal temperature differences. Winters are just about 0.1 °C colder than summer while in the free atmosphere at the same elevation where winters are about 10 °C colder than summer and have negative mean temperatures. During winter the cave temperatures are about 5 °C higher than outside, while caves remain colder than outside during summer.

Table 2 also shows that the average cave temperatures observed in HA and SH are slightly lower than the outside temperatures at the same elevation, while SP and RS are warmer than outside. This is corroborated by observations at nearby automatic weather stations (AWS), although the latter represent a less reliable measure of local outside conditions due to being located at different elevations. The observed cave temperatures follow a quasi-vertical gradient of about − 0.4 °C 100 m ^− 1 (red dots in Fig. 4), which is somewhat smaller than the lapse rate of outside temperatures in the relevant elevation range, i.e., 1000–2500 m a.s.l. and the moist adiabatic lapse rate (blue triangles and dotted line), respectively. However, it should be noted that the cave sites are located in different climatological regimes characterizing the Swiss Alps (SH), the Northern Calcareous Alps (HA), the Central Alps (SP) and the southeastern part of the Alps (RS). Judged by AWS data, temperature gradients are different along valleys in the vicinity of the investigated caves (-0.57 °C 100 m^− 1 in the Central Alps compared to -0.43 and − 0.41 °C 100 m^− 1 in the western and eastern and western fringes of the Alps, respectively).

Fig. 4.

Quasi-vertical gradients of mean temperatures measured in the investigated caves (2008–2010; red dots) compared to outside conditions, i.e., locally relevant AWS data (open triangles) and values interpolated to the elevation of cave observations (filled triangles). The dashed line shows a moist-adiabatic lapse rate (-0.5 °C 100 m^− 1).

Figure 4 also suggests that cave temperatures are correlated with the outside temperatures at the same elevation, which can be considered as an indicator of a physically based response of cave temperatures to changes of the outside atmosphere. This is supported by Table 2 (lowest panel) which also shows that the correlation of daily cave temperatures with concurrent outside values is stronger at the high-elevation sites and during winter, RS being outstanding in this respect. These results are related to cave-specific ventilation patterns which themselves strongly depend on their morphological features (concerning entrances essentially) as will be discussed below.

Trend analysis

Long-term temperature trends were investigated by calculation of linear trends based on ordinary least squares regression and the Theil–Sen trend estimator as well as using Mann-Kendall trend tests. The methods are applied to quality-controlled cave temperatures covering two decades as well as to AWS data representing temperatures at elevations below and above the cave sites.

Visual inspection of the observed records suggests that cave temperatures have systematically increased at all observation sites (Fig. 2). This is confirmed by quantitative analyses for the period 2000–2020 (Table 3), which reveals positive trends for all investigated sites, i.e., in the caves themselves as well as in the outside atmosphere. Cave temperatures increased by about + 0.2 °C per decade, with RS and SP showing higher values than SH or HA.

Table 3.

Decadal temperature trends for the period 2000–2020 (linear and Theil-Sen, the latter being represented by the bold numbers). MK+ qualifies the calculated trends to be significant according to the Mann-Kendall test. Outside temperatures refer to the elevation of the cave sites.

	Spannagelhöhle	Schrattenhöhle	Hundsalm Eis- und Tropfsteinhöhle	Rasslsystem
Cave [°C per decade]	SP-PL	SH-WK	HA-JG	RS-PH
Total period	+ 0.24 | + 0.26 (MK+)	+ 0.14 | + 0.15 (MK+)	+ 0.20 | + 0.12 (MK+)	+ 0.24 | + 0.28 (MK+)
Open periods	+ 0.22 | + 0.25 (MK+)	+ 0.14 | + 0.15 (MK+)	+ 0.21 | + 0.21 (MK+)	+ 0.25 | + 0.28 (MK+)
Closed periods	+ 0.25 | + 0.29 (MK-)	+ 0.13 | + 0.14 (MK+)	+ 0.20 | + 0.19 (MK+)	+ 0.24 | + 0.26 (MK-)
Outside [°C per decade]	at elevation of SP-PL	at elevation of SH-WK	at elevation of HA-JG	at elevation of RS-PH
Total period	+ 0.54 | + 0.53 (MK+)	+ 0.66 | + 0.72 (MK+)	+ 0.59 | + 0.57 (MK+)	+ 0.58 | + 0.53 (MK+)
Open periods	+ 0.61 | + 0.59 (MK+)	+ 0.63 | + 0.94 (MK+)	+ 0.62 | + 0.62 (MK+)	+ 0.76 | + 0.74 (MK+)
Closed periods	+ 0.22 | + 0.23 (MK-)	+ 0.48 | + 0.53 (MK+)	+ 0.29 | + 0.28 (MK+)	+ 0.10 | + 0.03 (MK+)
AWS low [°C per decade]	JE	EN	JE	BA
Total period	+ 0.68 | + 0.63 (MK+)	+ 0.65 | + 0.61 (MK+)	+ 0.68 | + 0.63 (MK+)	+ 0.63 | + 0.59 (MK+)
Open periods	+ 0.69 | + 0.68 (MK+)	+ 0.68 | + 0.65 (MK+)	+ 0.69 | + 0.68 (MK+)	+ 0.64 | + 0.68 (MK+)
Closed periods	+ 0.36 | + 0.32 (MK+)	+ 0.36 | + 0.32 (MK+)	+ 0.36 | + 0.32 (MK+)	+ 0.30 | + 0.27 (MK+)
AWS high [°C per decade]	PA	PI	PA	SE
Total period	+ 0.54 | + 0.53 (MK+)	+ 0.70 | + 0.77 (MK+)	+ 0.55 | + 0.53 (MK+)	+ 1.11 | + 1.17 (MK+)
Open periods	+ 0.61 | + 0.60 (MK+)	+ 0.68 | + 0.76 (MK+)	+ 0.61 | + 0.60 (MK+)	+ 0.96 | + 0.99 (MK+)
Closed periods	+ 0.22 | + 0.23 (MK-)	+ 0.36 | + 0.31 (MK+)	+ 0.22 | + 0.23 (MK-)	+ 0.88 | + 0.93 (MK+)

The outside temperatures at the elevation of the cave sites increased at higher rates of about 0.5 °C per decade. This value is independent of the applied calculation method and there is no clear evidence regarding elevation dependencies or regional patterns. It is notable, however, that temperatures at low-lying weather stations are characterized by rather consistent trends in the order of + 0.6 °C per decade, while some distinction can be observed regarding the investigated high-elevation sites. Hence there is an indication that the eastern and western fringes of the Alps (i.e., the RS and SH regions) experienced enhanced warming compared to the Central (SP) and the Northern Calcareous Alps (HA), which however, is not reflected by the cave data.

We also examined seasonal trends representing open (Nov-May) and closed periods (Jun-Oct), which provide evidence that the outside temperatures increased stronger during winter compared to summer (Table 3, open vs. closed periods). Naturally, this also concerns the underlying AWS data and contrasts the cave conditions, where significant differences between winter and summer temperature trends can hardly be discerned.

Table 3 also shows the results of the Mann-Kendall tests to explore the temporal consistency of the observed cave temperature trends. There is evidence that all investigated cave records are characterized by statistically significant trends, which also holds true for the investigated sub-periods (open and closed periods) and the calculated outside temperatures. The significance of positive trends can also be stated for the weather stations being used for the calculation of the outside temperatures, with the exception of summer temperatures at Patscherkofel.

These results are based on observations representing the two decades providing the most consistent data (2000–2020). However, the underlying time series have different length and episodic gaps (Fig. 2), which in principle can affect the robustness of the so far presented analysis. Moreover, the spatial representativeness of the presented cave records may be influenced by local effects (e.g., ventilation conditions, episodic water influx or icing) and diverse observation problems (such as unrepresentative sites, different instruments and/or failure due to e.g. icing). A rigorous analysis of respective effects on our trend analysis is hampered by the scarceness of appropriate cave records over longer time spans. To at least qualitatively strengthen the respective validity of the presented results we considered some additional temperature records from other positions in these caves. Corresponding results are addressed in the following sections and we here just mention strong evidence that all investigated cave sites experienced a warming since at least 1990 and beyond 2020 until the recent days.

In this context, we also refer to temperature data across the so-called Greater Alpine Region (GAR), which represents a gridded data set across the Alps derived from long-term homogenized station data³ and allows putting the presented results in a broader temporal and spatial context. We here consider a limited ensemble of these data, i.e., back to 1970 as well as low/high-elevation, West and Southeast regions, respectively. Figure 5 demonstrates that these records are characterized by consistent and fairly linear trends in the order of + 0.5 °C per decade, with indication of enhanced warming in the lowlands compared to high-lying regions (+ 0.56 vs. +0.48 °C per decade). The analysis further shows that the observed cave records as well as the calculated outside temperatures fit these regional developments, which essentially confirms their spatial and temporal representativeness. Note that the Austrian caves are located in HISTALP-regions West and Southeast. It may be added that Jungfraujoch, i.e., a Swiss mountain station some 50 km south of SH, is included in the GAR-High-Level data, where a warming rate of + 0.45 °C per decade was observed during the period 1970–2020.

Fig. 5.

Mean annual cave temperatures (dots) compared to calculated temperatures outside of the caves (thick solid lines) and HISTALP data for selected Greater Alpine Regions (thin lines).

Case study SP

SP has several entrances at different elevations, which has implications for the ventilation regime of the cave. Figure 6 (red line) again shows the previously presented SP-PL data and outside temperatures which are both characterized by consistent warming (Table 3). Addressing the question of respective spatial representativeness, we explored whether trends can also be detected in two other parts of this cave. Site SP-GH is located closer to the (upper) entrance while site SP-TG is located deeper into the cave, in a dead-end side passage (Supplementary Fig. S1). SP-GH largely mirrors the SP-PL data concerning the overall level (+ 1.68 vs. +1.82 °C), the seasonal variability and the trend (+ 0.36 vs. +0.24 °C per decade). The slightly higher temperatures at SP-PL can be attributed to the larger distance from the main (upper) entrance (840 m compared to 470 m for SP-GH). SP-TG is characterized by a different thermal regime with higher temperatures (+ 2.16 °C on average) and a strikingly reduced seasonal variability. This can be related to its location in kind of chamber deeper down in the cave and off the main air flow route. It is remarkable that also this remote cave section experienced a warming (+ 0.27 °C per decade) which according to the Mann-Kendall test is statistically significant.

Fig. 6.

Time series of air temperatures in Spannagel cave. SP-PL and SP-outside denote the so far discussed data (red and grey lines) compared to SP-GH (green), SP-TG (yellow) and SP-EH (blue). The lower resolution of the measurements at the beginning of the SP-GH and SP-PL records is due to use of different instruments that were replaced at about 2005.

We further studied temperature records from a small separate cave close to the entrance of SP (Spannagel-Eishöhle, SP-EH) which contained perennial ice until the year 2002. Figure 6 (blue line) shows that also this cave experienced a warming during the last two decades. The overall lower air temperature as well as the tenuous trend at the beginning of the record may have been related to the presence of ice. The seasonal amplitude of air temperature is comparatively large due the small size and effective ventilation of this cave. Summarizing, there is strong evidence that Spannagel cave as a whole and nearby caves were affected by a consistent warming over the last two decades at least.

Case study HA

HA is special in the context of the four studied alpine caves in that its ventilation regime is much more restricted given its sag-type geometry and the presence of perennial ice (Supplementary, Fig. S2). This leads to a cold-trap setting in the upper part of the cave, which will be addressed in more detail in this chapter thereby also considering wall-rock temperature data.

First, we again note the progressive damping of outside temperature changes as they progress into the cave, which is a key feature of cave microclimate and strongly depends on ventilation characteristics (Fig. 7). In sag-type caves (as HA) the damping is most pronounced during summer when stable thermal stratification prohibits effective exchange of air between the cave interior and the outside. Conversely, cold air can relatively easy penetrate into the deeper parts of the cave. We observed a seasonal temperature variation of about 30 °C outside, while at a depth of 29 m below the main entrance (site HA-29) the air temperature varies by just about 5 °C. We note that the conditions in this cave section are not only preconditioned by effective cold air pooling but also by the existence of perennial ice (Supplementary Fig. S2). The lower part of the cave is thermally separated from the upper one (site HA-JG), contains no ice and is characterized by a homothermic regime (air temperature hardly varies on an annual time scale) at comparatively high temperatures (average + 4.3 compared to + 0.1 °C at site HA-29). Figure 7 also shows long-term observations of rock temperatures (HA-126) measured at a distance of 126 cm behind the cave wall next to site HA-29. The average rock temperature is + 0.4 °C, thus being slightly higher compared to the cave air at the same depth (+ 0.1 °C). During winter the rock does not cool as much as the air, while air and rock are thermally adjusted during summer. Hence there is a systematic difference in seasonal variability. Closer inspection also reveals that changes in rock temperature systematically lag behind those in the air¹⁸.

Fig. 7.

Time series of air and rock temperatures in Hundsalm Eis- and Tropfsteinhöhle. HA-JG and HA-outside denote the previously presented data (red and grey lines), while HA-29 (blue) and HA-126 (yellow) refer to additional data measured in air and within rock in the upper, ice-bearing part of the cave.

Figure 7 finally supports that each of the considered HA sites experienced a warming throughout the investigation period 2010–2020. This is corroborated by the calculated trends (0.24 / 0.79 / 0.99 °C per decade for HA-JG / HA-29 / HA-126) and the according Mann-Kendall trend tests (all attesting statistical significance). It can thus be concluded that the results, which refer to HA-JG, can serve as a reference for the long-term evolution of air temperature in this cave. Moreover, there is indication that trends are stronger in cave sections containing ice (HA-29) and rock temperatures (HA-126) compared to relatively remote cave sections without ice (HA-JG).

Surface height changes of ice in HA were monitored by annual stake readings close to site HA-29 (Supplementary Fig. S2). This effort was motivated by the expectation that long-term warming traces in enhanced melt of ice, as is generally known for snow and glaciers outside⁵. Indeed, Fig. 8 (blue line) shows a continuous loss of ice, which was accelerated during the second half of the investigation period (2008–2018). The values of air temperature averaged over time between the stake readings (black line) show a transition from initially negative values towards mostly positive values. This analysis confirms a strong physical relationship between long-term temperature evolution and the development of ice in caves, as is known from ice caves elsewhere^18,19,30,31 .

Fig. 8.

Mean air temperature observed at site HA-29 during consecutive stake measurements (red line) and concurrent surface height changes of ice located at the same location in HA (blue line).

Discussion

Ongoing climate change is leading to a regional and global rise in the temperature of the atmosphere. The evidence is essentially based on a sophisticated synthesis of observations from weather stations, radiosondes and satellites, as well as numerical models¹. Equally important evidence comes from indirect sources relating to corresponding changes in the ocean (e.g., rising sea-surface temperatures), in the cryosphere (e.g., vanishing glaciers and ice sheets and thawing permafrost) and in the biosphere (e.g., shifting vegetation belts). This raises the question of whether climate change has also “arrived” in caves, which is investigated in this study on the basis of long-term observations from caves in the Alps. The data from four caves in different parts of this mountain range provide ample evidence that this is indeed the case. Principal data were selected considering their quality as well as their spatial and temporal representativeness. The main analysis is based on data covering two decades (2000–2020) and linear trends were calculated and tested for statistical significance. Extended temporal representativeness of key results is shown back to the 1990s, as well as correspondence to developments in other cave sections and regional climate changes outside the caves.

Context to previous work

The analysis of the studied cave temperature data reveals a consistent picture in terms of linear trends ranging from + 0.1 to + 0.2 °C per decade (Table 3). These figures are backed by the application of two calculation methods (least squares and Theil-Sen) and the statistical significance is confirmed by the Mann-Kendall test. Additional temperature records from each cave show that the principal records are representative for major parts of the investigated caves and not just for individual sites. These results overall agree with previous findings which are, however, mostly based on shorter time series and less elaborated analyses. Spötl and Pavuza²⁷ reported temperature trends for Spannagel cave of + 0.5 and + 0.3 °C per decade (2000–2015) obtained from sites GH and TG, respectively. These sites are located above and below site SP-PL, which is considered in this work. Although these figures refer to different periods, they also indicate that the sections closer to the upper entrance have experienced greater warming.

Trüssel²⁸ analyzed temperature records from Schrattenhöhle and reported a warming of 0.7 °C for Windkluft (SH-WK in this study) for the period 1990–2018, which corresponds to a trend in order of + 0.3 °C per decade (slightly decreasing through time). Similar warming rates were observed for a shorter period of time (1990–2010) in cave sections above Windkluft (Fünferkluft and Wermutsgang). The author critically pointed out that the cave sections for climate studies must be carefully selected in order to avoid misinterpretation that can arise, for example, from an inappropriate selection of observation sites. Filipponi³² proposed criteria suggesting that SH-WK is representative of large parts of the cave.

Wind et al.¹⁸ addressed long-term temperature changes in Hundsalm Eis- und Tropfsteinhöhle. Significant warming is apparent in the ice-free part of the cave throughout the period 2008–2020 (+ 0.24 °C per decade at site HA-JG, which is a principal site in this study). This value agrees with the findings of this study. Ongoing warming is also observed in the ice-bearing section of the cave, where trends depend on the distance to the ice (+ 0.27 for T36 vs. +0.54 for T29; the latter being located closer to the ice).

Some related information is available from neighboring countries. Domínguez-Villar et al.²⁹ observed that temperatures in Postojnia cave (Slovenia) increased at a rate of 0.04 °C per decade (2009–2013). The application of a model indicates that the cave was prone to warming since the 1980’s at least and enhanced warming is predicted for the future (+ 0.15 °C per decade). Analyses of historical data since 1935 indeed show that in this cave the average temperature has increased up to 2 °C³³, corresponding to a trend of + 0.25 °C per decade. Significant warming was also reported for another cave in Slovenia (Predjama Cave) as well as for the average temperature across Slovenia (+ 0.34 °C per decade during period 1961 and 2011).

The calculated temperature trends for the free atmosphere at the elevation of our studied cave sites are about twice as large as those observed inside the caves (+ 0.5 to + 0.7 °C per decade). These figures are in agreement with site-specific observations for Patscherkofel²⁷, Bonistock²⁸ and Hahnenkamm¹⁸. Since a straightforward comparison is hampered by the different elevations and periods, we also placed our results in an extended temporal and regional context provided by regionally gridded climate data i.e., the HISTALP data set^3,34,35. These data show the generally strong warming of the alpine region, which is about twice as large as the Northern Hemispheric average and affected all subregions of the Alps. Enhanced and fairly homogeneous warming occurred since the 1980, with rates matching those from our study (ca. +0.5 °C per decade). Our analysis indicates that trends in the low-lying regions are relatively uniform, while mountain stations show some regional distinction, e.g. smaller trends in the Central Alps compared to the Swiss Alps. Referring to the vicinity of Monlesi Ice Cave, Luetscher et al.³⁶ note that the Jura Mountains of northern Switzerland experienced a warming of ca. 0.5 °C per decade (1980–2000). Based on a gridded data set across Switzerland, Ceppi et al.⁴ found an average warming rate of + 0.35 °C per decade (1959–2008). The Obir mountain area in South-Eastern Austria is characterized by outstanding warming rates (ca. +1 °C per decade) which, however, is not fully reflected in corresponding HISTALP data.

Elevation-dependent and thus also regional warming in the Alps is controversially discussed in literature. Recent reviews conclude that a general picture is difficult to draw due to the variety of influencing factors and processes being strongly related to small-scale topographic effects^2,35,37. It is therefore also difficult to interpret the result that trends are larger during the open periods (winter) compared to the closed periods (ca. +0.6 vs. +0.3 °C per decade). A thorough investigation of local characteristics (such as cold air pooling, mesoscale wind systems, snow cover, radiation, cloudiness) that influence outside temperature is beyond the scope of this work. Using data from more than 1000 mountain sites across the globe, Pepin and Lundquist³⁸ showed that high-elevation temperature trends are highest near the annual 0 °C isotherm due to snow-ice feedback, which was also observed in Switzerland⁴.

Context for cave-specific conditions and processes

A comparison of the average cave temperatures with those outside reveals a small difference and a clear dependence on elevation (Table 2; Fig. 4). Both suggest that the cave temperatures overall correspond to the external conditions, as is corroborated by an independent analysis considering more than 60 caves across the Austrian Alps²⁷. This result is also in agreement with theoretical considerations^25,39, which show that cave temperatures are characterized by a damped and delayed response to external forcings, both increasing with distance from the entrance(s). Observations from caves across the world support the related hypothesis that temperature in deeper cave zones asymptotically approaches an equilibrium rock temperature. Following Badino⁴⁰, the latter roughly corresponds to the long-term outside temperature. Short-term fluctuations are effectively filtered out due to the large thermal capacity of the rock in which the caves formed. Application of this conceptual model to karstic rock with a thickness of 100 m yields equilibration time scales (i.e., the time to attain thermal equilibrium) in the order of 5–50 decades, depending on the relative importance of energy exchange involving flow of air and water. Since the studied caves have a rock overburden of less than 100 m it can be argued that the analyzed data were obtained at sites where the seasonal variability is strongly damped, while the climatically relevant variability is still captured, albeit with a delay. These theoretical considerations support the observation that warming trends inside caves are smaller that outside (Table 3). It also helps to understand the nature of the time series mainly considered in this study, which essentially show long-term changes (trends) superimposed by weak seasonal variability (Fig. 2). Systematically delayed responses of cave air temperatures to external forcings cause thermal imbalances, which are reflected in the differences between mean temperatures inside and outside of the caves, as our results also show (Fig. 4). Cross-correlation of observed annual cave temperatures with long-term outside temperatures (HISTALP data) indicates that the during the period 2000–2020 observed interannual variability of cave temperatures strongly correlates with that of regionalized outside temperatures approximately 15 years back in time (Supplementary Fig. S5). This figure is comparable to a model-based estimate for a site in Postojna cave, which shows that cave temperatures respond with a delay of ca. 20 years to external changes²⁹.

The thermal cyclicity observed in cave records has often been used to classify caves and zones therein. A basic distinction concerns the heterothermic and homothermic zones⁴¹. Considering that the analyzed cave data are characterized by a seasonal variability of the order of a few tenths of a degree (compared to an external variability of about +/-15 °C), they cannot represent truly homothermic conditions, but remote sections of heterothermic cave sections. Sedaghatkish et al.⁴² used a numerical model to simulate heat transfer in a karst conduit due to air flow and thermal conduction, which suggests a more physically based distinction of a “diffusive region” (where external forcings mainly propagate through thermal conduction in the rock), a “convective region” (where heat is mainly transported by heat transfer in air) and a “deep region” (where temperatures are quasi-constant). This distinction basically matches the observationally based approach by Cropley⁴³ and suggests that three of the principal sites being considered in this work can be assigned to the second class (SP, SH and RS; i.e., heterothermic), while HA-JG is similar to the third one (which essentially conforms to a deep i.e., homothermal regime). Medina et al.²¹ analyzed records from twelve caves across the globe and found three patterns which represent caves sites retracing external variations with little modification except for reduced amplitudes, sites featuring low correlation to surface temperature and a slight thermal delay, and cave sections showing extreme delays or even no relationship to surface variations. Our cave data match the first category, which the authors expect to represent caves with multiple entrances, where air flow plays a dominant role in controlling the cave microclimate (supported by our observations).

Seasonally changing airflow patterns are known for three of the investigated caves, i.e. SP, SH and RS^27,28,32, which have two or more entrances at significantly different elevations (Table 1). This, as well as the seasonally changing signs of the thermal gradient (inside vs. outside) are strong indications that processes related to advection and/or convection are important for the exchange of matter and energy with the outside. The so-called chimney effect has been recognized as a key process in this context, which is driven by seasonally changing density differences between the subsurface air and the outside air^39,44 and can be significantly modified by synoptically induced pressure differences⁴⁵. The effect is known to be important in Spannagelhöhle, where it induces a positive deviation of the cave temperature in sections close to the upper entrances compared to outside (Fig. 2; Table 2)²⁷. This is most evident during winter when warm air is forced towards the upper entrance to which sites SP-PL and SP-GH are relatively close. The direction of the air flow switches at a threshold temperature of about 1.5 °C⁴⁶.

Trüssel²⁸ showed that air flow in Schrattenhöhle is also driven by the chimney effect, but it is interesting to see that the temperature at the investigated site (SH-WK) is lower than outside. This can be related to the position of the investigated site closer to lower entrances, which in the context of the chimney effect are prone to be colder than regions closer to the upper entrances. However, air flow in SH is generally more complex due to the many entrances, which favor differential air flow in different branches of the cave. We note that wind-driven pressure differences can additionally complicate cave ventilation⁴⁵, which could play a role in Schrattenhöhle, too.

The RS is characterized by relatively high temperatures compared to outside. Also this cave is known for air flow being driven by the chimney effect which, however, is relatively weak due to the remoteness of this site relative to the main pathway of the air (Supplementary Fig. S3). Still, the direction of air flow seasonally reverses at an outside temperature threshold of 5.5 °C²².

Hundsalm Eis- und Tropfsteinhöhle is exceptional in several respects. First, it is the smallest cave and its sag-type geometry effectively traps cold air during winter, which enters the cave by density-driven convection. During summer, the air pools without significant exchange with the outside due to a stable thermal stratification, which supports the existence of perennial ice in the upper cave Sect¹⁸. The cave is therefore classified as “static with ice”⁴⁷. These characteristics give rise to an asymmetric seasonal temperature cycle, which is not observed in the other caves. However, the mainly considered data (HA-JG) are much less affected by these processes, since they were collected in a relatively isolated chamber in the deepest section of the cave.

Having thus addressed some individual characteristics of the studied caves, it is even more remarkable that the magnitudes of the calculated temperature trends are comparable, i.e., ranging between + 0.14 to + 0.24 °C per decade for the principally investigated sites. It should also be remembered that these sites are located in different regions across the Alps and at different elevations. The rather uniform warming rates can be explained by the fact that records with comparable seasonal variability were considered, which in principle can lead to similar long-term external signals.

The microclimates of different cave sections, however, retain some spatial variability due to individual physical characteristics such as the relative distance to entrances which influences related processes (e.g., heat conduction, ventilation and flow of water). Our case studies show a correspondingly different seasonality, which is reflected by a progressive attenuation (Figs. 6 and 7), which was already discussed. As far as the trends in the different cave sections are concerned, the results indicate that the decadal warming was stronger near the entrance(s) than in more remote sections (SP-GH vs. SP-TG and HA-29 vs. HA-JG). This is in agreement with theoretical concepts, which predict that external variability is delayed with increasing distance from the entrances due to the large thermal capacity of the overburden rock⁴⁰.

Indirect evidence of warming in the investigated caves

Temperature strongly controls the formation and existence of ice, which is present in two of the investigated caves. Hundsalm Eis- und Tropfsteinhöhle hosts perennial firn and ice in its upper section, which currently reaches a maximum thickness of about 5 m. The ice formes by metamorphism of snow falling though one of the entrances, as well as by refreezing of seepage water^18,48.

Our results confirm a development towards increasingly negative mass balances during the recent decade (Fig. 8), in parallel with the concurrent atmospheric warming in the Alps. Wind et al.¹⁸ developed a model, which allows to calculate annual melt rates as a function of the external temperature. Given the current warming rates, the model predicts that within the next decade, this cave will no longer preserve perennial ice (only seasonal ice). Note, however, that the mass balance of an ice body not only responds to surface melt (ablation) but also changes in accumulation and basal melting, both of which also depend on temperature.

A small cave next to Spannagel cave (Spannagel Eishöhle, SP-EH) contained perennial ice until about 2002 and has been ice-free since then. This adds evidence that ice caves are sensitive indicators of regional climate change, similar to snow cover and glaciers outside. Significant reduction of ice in caves is a wide-spread phenomenon across Europe^{17,18,41,49–51} and there is no indication that this trend will not continue (if not accelerate).

Potential influences of anthropogenic cave manipulations

Certain sections in the investigated caves are open to guided tours, which necessitates some related discussion. Touristic use in Spannagelhöhle began in 1994, involving ca. 500 m of galleries behind the upper entrance (open all year round). HA was opened as a show cave in 1967, and the guided tours cover most of the upper, i.e., ice-bearing part (open from May until October). The lower section (in which site HA-JG is located) was discovered in 1984 and is not accessible to visitors (locked). Rasslsystem (RS) is part of the Obir Cave system. Guided tours are only offered for part of this system that are far away from RS (since 1991; open April to October). Schrattenhöhle can also be visited within small guided tours, but this is only possible since the year 2000, i.e., after the investigation period and in cave sections far off site SH-WK.

Touristic use can have severe impacts on the thermal regime of caves⁵². The main influencing factors are the heat emitted by people and the lighting. Evidence is provided by comparing historical temperature data with recent ones and by comparing cave sections or neighboring caves with and without touristic influence³³. Microclimatic studies usually show increased daily variability which is related to the number of visitors or concurrent changes in, e.g., humidity and CO₂ concentrations^53,54. In general, anthropogenic influences are difficult to disentangle from natural changes, as they are cave-specific and less pronounced in well-ventilated caves or sections.

Regarding the potential impact on the investigated caves, it should first be noted that in three caves the main study sites (SP-PL, HA-JG and RS-PH) are not part of guided tours, i.e. the sites are relatively far from the direct impact of visitors. In terms of timing, tourist activities are rather confined to summer and autumn, which means that excess heat in outer cave sections (where guided tours are offered) is likely to be pushed out through the entrances. This may be relevant for Spannagel and Rasslsystem, but not for Hundsalm Eis- und Tropfsteinhöhle. This sag-type cave is special because its microclimatic summer regime is characterized by stable air stratification. In principle, this has meant that the excess energy input from the tourists has slowly warmed the cave, which in turn will likely have been compensated for by the increased melting of snow and ice (Fig. 8). Wind et al.¹⁸ noticed small transient disturbances of the cave air temperature being related to guided tours, but these were equilibrated shortly after. Potential long-term effects have not been observed. Regarding the principally considered site in this cave (HA-JG in the lower, ice-free part), warming by tourists cannot play a role, since it is located in a thermally isolated chamber underneath the ice-bearing part (Supplementary Fig. S2). Finally, as far as the decadal trends are concerned, it should be noted that touristic use in all these caves began before the main period under consideration (2000–2020). It can therefore be assumed that the trends analyzed are not distorted by tourist activities. Data from Schrattenhöhle are not at all affected by such effects because guided tours started after the investigation period.

Consequences of warming caves

The demise of cave ice as a consequence of warming of cave environments was already been addressed in the context of the Hundsalm Eis- und Tropfsteinhöhle and the (former) Spannagel Eishöhle (Figs. 7 and 8). Related processes have been elucidated by e.g. Luetscher et al.⁴¹ or Obleitner and Spötl⁵⁵. These studies also showed the importance of refreezing seepage water, whose availability strongly depends on the permeability of the host rock, which at high-elevation alpine caves may also contain sporadic permafrost. Systematic warming is therefore expected to significantly change the availability (amount and timing) of water in these caves, which in turn will also affect the thermal environment⁴⁰. The degradation of sporadic permafrost can also lead to a destabilization of cave passages, which happened in Hundsalm cave in 2020 (rockfall). Warming can affect the ventilation regime of high-elevation caves by changing the upper entrances (which otherwise may be plugged by snow or ice). Such effects have already been observed, e.g., in Schrattenhöhle²⁸.

Finally, warming of subterranean environments will potentially have a large impact on biota. For example, Mammola et al.²⁰ argued that caves are largely unexplored habitats for species (bats, spiders, bacteria) that are likely to be sensitive to even small changes in the microclimate⁵⁶.

Implications for speleothem-based paleoclimate research

Many caves contain carbonate deposits, collectively known as speleothems, which are highly sought-after archives for paleoclimate studies. Speleothems can be accurately dated using U-series methods and a range of proxies are available to reconstruct e.g. air temperature (changes) over long time scales^57,58. Worldwide studies have shown that the air temperatures in the homothermic zone of most caves – with the exception of those with special geometries that serve as ‘traps’ for cold or warm air – are close to the mean annual air temperature outside the cave at this altitude^25,59. Therefore speleothem-based paleoclimate studies mostly focus on samples from the inner parts of caves. Less is known how quickly a given warming or cooling of the outside atmosphere propagates into caves. This clearly depends on the detailed geometry and dimensions of the cave, as well as on the interplay between cave ventilation and possible major water infiltration (e.g., through cave streams). The question how much the climate signal ultimately captured by speleothems is delayed relative to the external forcing is of less importance for studies that focus on long periods of time, e.g. orbital time scales. However, for high-resolution studies using fast-growing (sometimes annually laminated) speleothems, this question is relevant. Long-term studies such as our alpine network provide relevant empirical data to assess these lags. A first step in this direction was presented (Supplementary Fig. S5), but advanced modelling studies following e.g^29,42,44, are needed to gain quantitative insights for cave sites where no such monitoring information is available.

Conclusions

Statistically significant changes of air temperature were observed in four caves across the Alps. Our analysis covers the most recent two decades and different cave types (static vs. dynamic and also with ice) as well as different sections therein. Key results, interpretations and conclusions are summarized as follows:

• Mean air temperatures inside these caves are close to those outside and are characterized by much reduced seasonal variability. Caves are systematically warmer in winter (colder in summer) and quasi-vertical temperature gradients in caves are smaller than outside.

• Trends of cave temperature increase are about half of that outside (0.2 vs. 0.5 °C per decade).

• These warming rates are characterized by a remarkable linearity and a pronounced spatial and temporal consistency compared to external conditions.

• These figures represent cave sections characterized by an almost homothermic regime, although there are indications of increased warming rates near the entrances. This and other observations (reduced and lagged seasonal variability) are in line with basic theoretical expectations.

• The consistency of these trends is remarkable given the variable geometry of the investigated caves, and the results are considered representative of cave environments in the Alps. The synthesis of the data confirms that global warming has reached the interior of these cave systems.

• The currently observed interannual variability of cave temperatures is strongly correlated with outside conditions having occurred ca. 15 years back in time.

• There is no indication that the touristic use of some of these alpine caves has significantly influenced the results.

• Continued warming may cause feedbacks and broader changes in cave environments, e.g. changes in the ventilation regime and increased ice melt, which may also affect water resources and subsurface ecosystems.

• This study shall stimulate sustained long-term monitoring of temperature and other physical or chemical parameters, as well as related modelling, to better understand the current and future responses of caves in a warming world.

Methods

Cave air temperatures were logged using different instruments. Most data from the Austrian caves were obtained using HOBO TempPro v2 logger (Onset Computer Corp.; resolution 0.02 °C, accuracy ± 0.21 °C); the first few years of some of these time series were recorded by an earlier model, HOBO Optic StowAway Temp (Onset Computer Corp.; resolution 0.16 °C, accuracy ± 0.2 °C). In one cave, Hundsalm Eis- und Tropfsteinhöhle, we also logged the rock temperature in a borehole drilled horizontally into the cave wall (diameter 2.5 cm) using a HOBO TempPro v2 logger with an external sensor (Onset Computer Corp.; resolution 0.02 °C, accuracy ± 0.21 °C). The calibration of both these logger types was checked using an ice-water bath in a dewar prior to the installation in the cave and after the loggers were brought back to the laboratory. No drift was observed in any of the annually recovered time series. The final data were corrected for the offset at 0 °C, which for most HOBO TempPro v2 logger was less than 0.1 °C. The logging interval in the Austrian caves varied between 1 and 3 h.

Temperature data in the Schrattenhöhle were collected using a different logger model, i.e., Hamster and Hamster-A (Elpro; resolution 0.05 °C, accuracy ± 0.2 °C) and the calibration provided by the manufacturer was used. Occasional cross-checks using other logger types prove an overall accuracy of ± 0.1 °C. The logging interval in Schrattenhöhle was 1.5 h (1990–1999) and 1 h (since 2000).

These instruments proved reliable to recover cave temperatures in a methodically consistent manner and over long periods of time. Minor gaps in the records were mostly due maintenance (downloading the data or change of sensors) which in caves can need longer time. Major gaps evolved in the context of post-processing raw data focused on detection of physically unplausible data and comparison to parallel series as far as available. It is helpful in this context, that cave temperatures are less variable compared to outside temperatures. The reasons for physically unplausible data can be related to various environmental issues including, e.g. intermittent influences by water, rime or ice on sensors. About 15% of the original data from Hundsalm and Spannagel caves were finally excluded from further analysis, significantly more from SH and RS (62 and 42%). The postprocessed data were then merged to daily averages, which form the basis of this study.

The mostly complex geometrical structure of caves and related impacts on their microclimate questions the spatial representativeness of analysis referring to a certain cave section. Practically, the siting of a sensor is often determined by on-site compromises considering scientific arguments, accessibility and constraints to safely mount sensors, which usually can hardly be visited several times a year on the other hand. Related uncertainties are difficult to quantify and would require enhanced measurement networks or detailed numerical simulations of air flow along the investigated caves (both being beyond the scope of this work). Nevertheless, in order to address the topic of this investigation (temperature trends) we also analyzed observations in neighboring sections of the studied caves to study the spatial representativity of the best possible chosen reference time series.

Daily values of air temperature measured at automatic weather stations (AWS) in the vicinity of the four caves were used to calculate free-atmosphere temperatures at the elevation of the cave sites. The latter were derived by linear interpolation based on daily lapse rates calculated from station pairs located at the valley bottom and at crest height in representative valleys. Correspondingly, we used the station pair JE (530 m a.s.l.) and PA (2251 m a.s.l.) to derive the lapse rate for the SP and HA environment, while BA (623 m a.s.l.) and SE (1040 m a.s.l.) are used for RS and EN (1000 m a.s.l.) and PI (2100 m a.s.l.) for SR, respectively. The calculated temperatures were validated using additional AWS data (Plattkopf (2260 m a.s.l.), HA-entrance (1520 m a.s.l.), Bonistock (2160 m a.s.l.) and Zell-Pfarre (900 m a.s.l.; Fig. 1) and corresponding offsets were applied to refine the calculated temperatures at the elevation of the cave sites. The ice development in HA cave was monitored using stakes drilled into the ice yielding annual measures of surface height changes with an estimated accuracy of ± 1 cm¹⁸.

Characterization of the typical meteorological conditions in the studied caves and their outside environment is based on standard statistics for a reference period (2008–2010) with most consistent data. Shorter and longer periods are considered in the context of case studies addressing specific problems, e.g. the spatial representativity. Trend analysis employs the calculation of linear trends and Mann-Kendall tests^60,61. Since these methods can suffer from its sensitivity to outliers, we also employed the Theil–Sen trend estimator^62,63. This is a comparatively robust method to fit a line to sample points by considering the median of slopes determined by all pairs of successive data.

Calculations were performed using Matlab®⁶⁴ and were also applied to seasonal subsets covering the so-called open and closed periods. Following Wind et al.¹⁸, the latter comprise the months November to April and May until October, respectively.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

The idea of the manuscript was developed by C.S. and M.T., who also provided the cave data. Main analysis was provided by F.O., who also drafted the manuscript, tables and figures except of Fig. 1. Figures in the supplement were contributed by C.S., and M.T. The manuscript was reviewed by all authors, thereby also considering comments by Beat Niederberger. The authors declare no competing interests and underlying data can be provided on request.

Funding

Long-term monitoring in the Austrian caves was funded in part by the Austrian Science Fund (FWF) [Grant DOI: 10.55776/Y122, 10.55776/P31874, 10.55776/P20618 an 10.55776/P18637Y122]. For open access purposes, the authors have applied a CC BY public copyright license to any author-accepted manuscript version arising from this submission.

Data availability

All data being used in this study are available upon request. Please contact friedrich.obleitner@uibk.ac.at or christoph.spoetl@uibk.ac.at. Key data can be downloaded from https://zenodo.org/uploads/14050143.

Declarations

Competing interests

The authors declare no competing interests.