Here we present the results of the inversion of a multi-annual temperature
profile (2013, 2014, 2015) of the deepest borehole (235
The thermal regime of the uppermost ground is determined by the geothermal
heat flow and by the fluctuations of temperature at the surface. If rock was
homogeneous and no temperature change were to occur at the surface, the
temperature would increase linearly with depth, unless spontaneous heat
production is present in the vicinity of the well. The gradient of this
temperature increase would be governed solely by the magnitude of the
terrestrial heat flow and by the thermal conductivity of the rock. However,
variations in ground surface temperature (GST) propagate downwards into the
rock as attenuating thermal waves, superimposed on the aforementioned linear
temperature profile. The depth to which disturbances can be recorded is
determined mainly by the amplitude and duration of the temperature change at
the surface. Generally, propagation of climate signals is slow and it can
take more than 1000 years to reach the depth of 500
Lachenbruch and Marshall (1986) were among the first to demonstrate that
thermal profiles obtained from boreholes drilled in permafrost can be used
to reconstruct GST changes. These do not require
calibration because the heat conduction equation is directly used to infer
temperature changes at the ground surface. Today, the majority of permafrost
boreholes used to reconstruct GSTs are located in the
polar regions of North America and Eurasia where the boreholes can be
drilled on flat terrain, with negligible topographical effects, and with
permafrost thicknesses typically exceeding 100
The Share Stelvio Borehole (SSB) in the Italian Alps is the deepest drilled
within permafrost in the mid-latitude mountains of Europe. Because the
permafrost thickness exceeds 200
Recent atmospheric warming (over the last century) in the European Alps has
been roughly twice the global average (Böhm et al., 2001; Auer et al.,
2007). Despite its high sensitivity, no GST reconstruction based on borehole
thermal profiles is available for this part of the world. Instead,
reconstructions of summer air temperatures have been based on either
tree rings (e.g. Büntgen et al., 2006; Corona et al., 2010) or lake
sediments (e.g. Larocque-Tobler et al., 2010; Trachsel et al., 2010) for the
last 500–1000 years, or both (Trachsel et al., 2012). With rare exceptions
(e.g. ice cores; Barbante et al., 2004), the other proxy data are from sites
at elevations that rarely exceed 2000
The SSB data provide GST history from a high elevation site (3000
This paper reconstructs the GSTs inferred from this borehole and compares the results with existing multi-proxy reconstructions for the European Alps and elsewhere.
Study area:
Topography of the SSB site:
The Stelvio–Livrio area is a summer ski location, located between the
Stelvio Pass (2758
The SSB was drilled in early July 2010, using refrigerated
compressed-air-flush drilling technology. The stratigraphy was obtained using
analyses of the cuttings (sampled every 10
The thermal properties of the three main facies observed in the stratigraphy
were measured in the laboratory at three different temperatures
(0,
Example of a GST history parametrized by Eq. (2).
The temperature anomaly in the borehole at time
In order to exploit the abovementioned explicit solution, it is customary to
approximate the GST with a piecewise constant function (see Fig. 3):
Synthetic data for the present time. It is remarkable that by also
varying the
Once the sequence
A common choice for
In order to validate our GST inversion method we have generated a synthetic
data set as follows. An ideal GST was chosen (dashed curve in Fig. 4) and
Eq. (1) was solved using a finite difference method with a spatial grid
spacing of 1
In the first experiment we fed our inversion algorithms only with the
synthetic data for the present time. The value of
Next we also fed the inversion algorithm with the synthetic data for the
past years. First, the inversion is expected to be more accurate since the
algorithm can average not only the temperature at a given depth but also the variation in the temperature in the last years at that depth.
Moreover, the algorithm should also be more robust since it relies on a
larger data set. Both these effects can be appreciated in Fig. 5, where it
can be seen that the inversion in the last 50 years is more accurate than
the inversion of Fig. 4 and that a wider variation in the value of
Synthetic data for three past years (2013, 2014 and 2015). It can be seen that the inversion in the last 50 years is more accurate than the inversion of Fig. 4. Legend: 0.15: green line; 0.2: orange line; 0.25: red line.
Share Stelvio Borehole (SSB) stratigraphy. Legend:
SSB mean annual ground temperature profiles in 2013, 2014 and 2015.
Thermal gradients (
Thermal properties of the three different facies occurring in SSB measured in the laboratory at three different steps of temperature
(0;
Effects of different thermal diffusivity used in the model. The temperature profiles a posteriori of 2015 obtained in the case of a constant thermal diffusivity value of the more widespread facies a (red dots) and in the case with multi-layer thermal diffusivities following the different facies according to Fig. 6 (blue dots). The bars indicate the variations in the measured temperature in the same year.
The SSB stratigraphy is characterized by four different facies of dolostone
(Fig. 6): a massive dolostone (from grey to pinky grey) comprises more
than 90 % of the profile; three other facies (white dolostone, black
stratified limestone, brownish dolostone) are thin intercalations (maximum
3.5
The mean annual thermal profiles of the last three years (2013, 2014 and 2015) show a
negative gradient between 20
Table 2 shows the thermal properties of the three main stratigraphic facies
encountered in the borehole. Facies a and c show similar density and thermal
properties while facies b has higher density and higher conductivity. All
facies have heat capacity values that increase with a decrease in
temperature. In facies a, this behaviour also occurs for thermal conductivity
and diffusivity values. In contrast, facies b and c show a reversed bell
shape behaviour, with the minimum value recorded at
According to the model proposed in Sect. 3, we found the best fitting
with the thermal profiles (Fig. 7) using a heat flow of
70
The linear system (4) was assembled including the detrended data measured at
SSB in 2015 (
Example of different GST histories with different
Comparison among the anomaly of the mean annual GST
reconstructed by the SSB (black thick line), its uncertainty range (red
shaded) and the MAAT anomaly reconstructed for the European Alps by Christiansen
and Ljungqvist (2011) (grey line with dots; data available online at
Trend of monthly mean of GST (red line) and air temperature (blue line) at SSB since 1998.
Effect of the snow cover at SSB. The winter 2010/11 is
representative of the average conditions of the snow cover at SSB while the
following season 2011/12 was the snowiest of the whole monitoring period.
The difference between the daily mean GST and air temperature (
In permafrost environments, snow cover can influence GST variability in both
space and time (e.g. Zhang, 2005; Schmidt et al., 2009; Morse et al.,
2012; Rodder and Kneisel, 2012; Schmid et al., 2012; Guglielmin et al.,
2014). This is especially the case for alpine areas where topography
influences both the re-distribution of the snow by wind drift and actual
snow cover evolution (e.g. melting date and duration).
Nevertheless, GST and air temperature are well correlated (
Our reconstruction after the cold GST anomaly, between AD 1906 and 1941,
shows a slightly positive peak (ca. 0.1
On a regional scale, the Stelvio data can be compared with the MAAT obtained for the Alps by Christiansen and Ljungqvist (2011) (Fig. 10) and Trachsel et al. (2010). The maximum of the slight temperature increase during the first half of the 20th century in the Stelvio data (1930) falls exactly in the middle of the relative warming period between 1925 and 1935 in the Alps found by Trachsel et al. (2010) and is in good agreement with the date (1928) indicated by Christiansen and Ljungqvist (2011). Later, the sharp GST anomaly decrease was delayed in the Stelvio data (1989) with respect to the 1950–1965 period found by Trachsel et al. (2010) and 1965–1975 period found by Christiansen and Ljungqvist (2011). Finally, the most recent increase in temperature culminated in the Alps in 1994 (Christiansen and Ljungqvist, 2011), while in the Stelvio data it culminated in 2011.
The Stelvio reconstruction shows a long period of negative anomaly between
AD 1560 and 1860 with colder conditions (
The LIA period has also been characterized by a widespread worldwide glacier advance, although the comparison between glacial evidences and temperature fluctuations is problematic because glaciers respond with different timescales (mainly depending on their size) and also reflect the precipitation regime, which is even more variable in space and time. According to Holzhauser et al. (2005), the LIA advance of the main Swiss glaciers has three peaks around AD 1350, 1640 and 1820–1850 respectively with the two later phases almost synchronous, also in the Eastern Alps (Nicolussi and Patzelt, 2000).
Close to the location of the SSB, the maximum LIA advance was diachronous. Nearby glaciers show a maximum LIA advance in AD 1580 (Trafoi Valley glacier; Cardassi, 1995), around AD 1770 (Solda glacier; Arzuffi and Pelfini, 2001) and in AD 1600 (La Mare glacier; Carturan et al., 2014).
The borehole area was presumably over-capped by the Vedretta Piana glacier until 1868. Due to the geomorphological position (on a watershed divide) the possible glacier should have been very thin and possibly cold based, as already stressed by Guglielmin et al. (2001). However, considering Fig. 10, the glacier should have been present in the borehole area with a buffering effect only between AD 1711 and 1834, with a peak at 1760, when the difference between the GST anomaly and the MAAT anomaly was maximum. This peak is pretty similar to the peak of the LIA in the Solda glacier (AD 1770) but not to the peak in the Trafoi glacier (AD 1580); this could be related to Vedretta Piana having a more similar glacier size and aspect (NE-N) to the Solda glacier than to the Trafoi glacier, although this latter is the closest to the Vedretta Piana.
Main climatic events enhanced by anomalies of MAAT through different proxies in all of Europe: A (modified from Luterbacher et al., 2004); central Europe: B (re-elaborated from Dobrovolný et al., 2010); Alps: C (modified from the same data of Fig. 5; Christiansen and Ljungqvist, 2011) and SSB: D (this paper).
Several deep Alaskan boreholes have been used to demonstrate the 20th century
warming (e.g. Lachenbruch and Marshall, 1986; Lachenbruch et al., 1988) but
only a few studies in Europe illustrate GST reconstructions that span a time
period greater than 100–150 years (e.g. Isaksen et al., 2001; Guglielmin,
2004). In North America, only Chouinard et al. (2007) show a GST pattern of
the last 300 years in the context of the permafrost of northern Québec.
There, after the LIA (AD 1500–1800), an almost constant and
marked warming of ca. 1.4
There is a some similarity between the Stelvio reconstruction and the
pattern of Canadian permafrost GST reported by Chouinard et al. (2013)
after the LIA. Indeed, at our site there was also an almost simultaneous but
greater cooling (0.9
The general climatic pattern of the last 500 years recorded by this mountain
permafrost borehole is similar to the majority of other studies in the
European Alps and central Europe. The main difference concerns post LIA
events. In fact, the different multi-disciplinary proxies considered (see
Fig. 13) do not indicate cooling between 1940 and 1989, with the
exceptions of the shorter and less severe cooling found for the Alps. It is
also relevant to stress that the rapid and abrupt GST warming (more than
0.8
The Stelvio borehole GST reconstruction also allows one to estimate changes in the Vedretta Piana glacier. This glacier presumably buried the site of the Stelvio borehole with an ice thickness sufficient to exert a significant buffering effect upon the ground thermal regime between AD 1711 and 1834. This was a time when the difference between the Stelvio GST anomaly and the MAAT anomaly was greatest.
After publication the borehole temperature data cited here
will be uploaded to the GTn-P database at
The temperature anomaly in the borehole at time
Given the detrended measures
In this paper, we use as regularizer a standard discretization of the
Laplacian
In order to ascertain the effect of the terrain geometry we conducted a number of forward simulations with the model (1) using the synthetic GST shown in Fig. 4 (dashed line) and already employed for the sensitivity analysis as boundary data.
First we computed the solution of the one-dimensional model Eq. (1). Next we
computed the solution of the corresponding three-dimensional model in a
computational domain of
Comparison of predictions of the forward model for the same GST and different geometrical setups. Legend: 1-D: red dots; 2-D flat terrain: blue line; 2-D N–S: green line; 2-D SSW–NNE: orange line; 3-D: dashed black line (see the Appendix for the details).
Figure B1 compares the temperature anomalies that each of the models would
predict at SSB at present time. The red dots are the predicted well
anomalies at the depth of the thermometers at SSB. One can see that the
predictions of the two-dimensional model with flat terrain (blue line)
almost coincide with those of the one-dimensional one. Furthermore, the
two-dimensional model applied to the section with the steeper sides (the
SSW–NNE one, orange line) gives rise to predictions that are within the
instrumental error (
Finally, let us remark that for the forward model, a numerical 3-D simulation takes hours to complete on 16 computing nodes of our HPC cluster. Using a numerical multi-dimensional simulator in the inverse problem would of course require computing the forward model several times and would thus take a lot longer than the few seconds in which our proposed method can compute the reconstructed GST depicted in Fig. 10.
The authors declare that they have no conflict of interest.
The SSB was drilled and equipped thanks to the project “Share Stelvio” managed by EvK2-CNR and funded by Regione Lombardia. The research was also funded through the PRIN 2008 project “Permafrost e piccoli ghiacciai alpini come elementi chiave della gestione delle risorse idriche in relazione al Cambiamento Climatico” led by Claudio Smiraglia. Special thanks to the Stelvio National Park, SIFAS and Umberto Capitani for the permission and the logistical support. We also want to thank Hugh M. French for revision and English editing of a previous version of the paper. Edited by: Volker Rath Reviewed by: Volker Rath and one anonymous referee