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 set3,34,35. 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 effects2,35,37. 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 considerations25,39, 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 RS27,28,32, 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 air39,44 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 water18,48.
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 Europe17,18,41,49,50,51 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 concentrations53,54. 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 scales57,58. 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 altitude25,59. 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.g29,42,44, are needed to gain quantitative insights for cave sites where no such monitoring information is available.