Three deep boreholes (≥100 m) in mountain permafrost were recently drilled in Svalbard (Janssonhaugen) and in Scandinavia (Tarfalaryggen and Juvvasshøe) for long-term permafrost monitoring. These holes form part of a latitudinal transect of boreholes in permafrost through Europe, established by the Permafrost and Climate in Europe (PACE) project. Six-year thermal time series data collected from the three boreholes are presented. These data provide the first opportunity for temporal trends in permafrost temperatures in Svalbard and Scandinavia to be analyzed. Results show that the permafrost has warmed considerably at all three sites. Significant warming is detectable down to at least 60 m depth, and present decadal warming rates at the permafrost surface are on the order of 0.04°–0.07°C yr−1, with greatest warming in Svalbard and in northern Scandinavia. The present regional trend shows accelerated warming during the last decade.
 This paper is based on temperature data from three mountain permafrost boreholes that have recently been installed in Svalbard and in Scandinavia [cf. Sollid et al., 2000]. The boreholes were drilled and instrumented between May 1998 and March 2000 for long-term permafrost temperature monitoring and form the northern part of a latitudinal transect established by the European Fourth Framework project, PACE (Permafrost and Climate in Europe [Harris et al., 2001b]). At the permafrost borehole in Svalbard (Janssonhaugen) a ground surface temperature reconstruction based on a heat conduction inversion model of the ground temperature profile indicated a near surface warming of 1°–2°C over the last 6–8 decades [Isaksen et al., 2000]. At the two permafrost boreholes in Scandinavia (Tarfalaryggen and Juvvasshøe) similar data showed a near surface warming of 0.5°–1.0°C over the last 3–4 decades [Isaksen et al., 2001]. The present study gives the first systematic documentation of recent temperature changes in permafrost in Svalbard and in Scandinavia.
2. Research Context
 Permafrost is linked to the atmosphere by the intervening active layer, vegetation, and snow cover, which vary strongly with time and location [Romanovsky and Osterkamp, 1995]. Variable snow cover, especially in early winter, exerts an important influence on ground temperatures [e.g., Smith, 1975; Goodrich, 1982; Vonder Mühll et al., 1998; Imhof et al., 2000; Zhang, 2005]. Circulation of water and air within coarse blocks in typical steep Alpine slopes may result in highly variable and sometimes extreme thermal offsets [e.g., Hoelzle et al., 2001]. Thus permafrost ground temperatures may integrate effects of several processes involved in the heat transfer regime of the air-ground interface. This integration is highly sensitive and is reflected in the thermal signal conducted downward into the permafrost below. It may result in ground temperature histories that are different to air temperature histories. A detailed knowledge of the energy exchange processes involved in the air-ground interface is therefore of fundamental importance for understanding the ground thermal regime response to climate change [e.g., Mittaz et al., 2000; Isaksen et al., 2003; Gruber et al., 2004a; Hanson and Hoelzle, 2004]. Scientific research concerning such processes in mountain permafrost is still in its infancy but now experiences rapid development [cf. Hoelzle et al., 2001].
 In mountain permafrost, thermal gradients may be influenced by four major factors; past changes in ground surface temperatures, ground surface topography, regional geothermal heat flux and variation in lithology and ice content with depth [Harris et al., 2003]. At lower elevations local hydrology could also have major effects. Therefore interpretation of measured temperature profiles should be undertaken carefully, especially in view of the strong 3-D effects within rugged alpine topography [Gruber et al., 2004c]. Temperature measurements in permafrost boreholes are particularly well suited to reconstructing past surface temperature changes, mainly because heat advection by groundwater or air circulation is often negligible. The geothermal profile is therefore primarily a function of heat conduction. Heat flow from the Earth's interior toward its surface and the heat flux from the energy exchanges at the ground surface determine the near-surface geothermal profile. Temperature perturbations at the surface are propagated downward and attenuated through time. The annual thermal cycle, with typical amplitude of 20°–30°C generally penetrates to a depth of 15–20 m, but larger perturbations in surface temperature of longer periodicity penetrate much deeper. Thus changes in the subsurface thermal gradient provide a record of recent ground surface temperature history.
 Global climate change simulations show a larger increase in annual mean air temperature over the Arctic and sub-Arctic than anywhere else in the world [Intergovernmental Panel on Climate Change, 2001]. Although climate predictions suggest strong warming at high latitudes, the air temperature records in this region show pronounced fluctuations and large interannual variability, making identification of longer-term trends more difficult. Recorded ground temperature changes at 40–50 m depth may provide direct evidence of thermal trends at the ground surface during recent decades [e.g., Cermak et al., 2000; Osterkamp and Romanovsky, 1999]. Permafrost temperatures represent a systematic running mean that filters the higher-frequency signal of the atmosphere and preserves only the low-frequency, long-term signals [cf. Lachenbruch and Marshall, 1986]. Thus analyses of permafrost ground temperatures obtained at carefully selected drill sites may constitute a key research tool in climate studies.
 Locations of the drilling sites in Svalbard and in Scandinavia are shown in Figure 1. The northernmost borehole is located at Janssonhaugen (depth 102 m), western Svalbard (78°10′N, 16°28′E, 270 m asl). It was drilled in May 1998. In Scandinavia, boreholes were drilled at Tarfalaryggen (depth 100 m), northern Scandinavia (67°55′N, 18°38′E, 1550 m asl) in March 2000 and at Juvvasshøe (depth 129 m), southern Scandinavia (61°40′N, 08°22′E, 1894 m asl) in August 1999. The drill sites were carefully selected by J. L. Sollid, in discussion with O. Liestøl, in order to avoid geothermal disturbance from undesirable and nonclimate sources [cf. Isaksen et al., 2000].
 At Janssonhaugen, estimated mean annual air temperature (MAAT for the standard normal period 1961–1990) is approximately −8°C. Only the months of June, July and August have average air temperatures above zero. The annual precipitation at Janssonhaugen is estimated to be 300–500 mm. The snow thickness at the drill site is very low and for most of the time snow is completely absent, due to redistribution by wind. Janssonhaugen is a west-east oriented hill with a top plateau and gentle slopes down to the valley bottom. Bedrock consists of fine-grained porous sandstone with a high content of silt and some thin interbeds of shale [Dypvik et al., 1991]. Vegetation cover is sparse and superficial material, sorted into polygons, is thin (0.2–0.5 m) and dominated by in situ weathered bedrock and patchy till.
 At Tarfalaryggen, the estimated MAAT for the drill site is approximately −6°C. The annual precipitation is approximately 500 mm [National Committees for the International Hydrological Decade in Denmark, Finland, Iceland, Norway and Sweden, 1976]. During a normal winter a thin layer of snow covers the area. Tarfalaryggen is a north-south oriented mountain ridge. The ground surface is blocky and is sorted into polygons and the vegetation cover is sparse. The thickness of the unconsolidated regolith is approximately 4 m. This loose material is believed to be largely in situ weathered bedrock. The underlying bedrock is massive amphibolite in the Seve-Köli nappe [Andreasson and Gee, 1989].
 MAAT is estimated to be approximately −4.5°C and the average precipitation 800–1000 mm a –1. Snow thickness is very low due to the exposed setting and associated redistribution of snow by wind. The vegetation cover is sparse and the surface is bouldery (block field), consisting mainly of in situ weathered material of thickness of 3–4 m. The borehole is located at the rim of a 4 km2 mountain plateau. The bedrock is light to medium gray quartz monzonorite, rich in plagioclase and quartz [Battey, 1960].
 None of the three boreholes reaches the permafrost base. The permafrost depth at Janssonhaugen is estimated to be 220 m [Isaksen et al., 2000]. In contrast, despite surface mean annual temperatures some 3°–4°C higher, low geothermal gradients on Tarfalaryggen and Juvvasshøe indicate deep mountain permafrost, estimated to exceed 300 m at both sites [Isaksen et al., 2001].
 At each of the three drill sites a 15–20 m deep control borehole was drilled between 5 and 20 m away from the main borehole. These shallow boreholes were installed to detect the thermal influence of the protection structure located at the top of the main boreholes and to ensure a good resolution and quality control (e.g., to control long-term stability of thermistors) of the annual ground thermal variations. Borehole temperatures were measured with negative-temperature-coefficient (NTC) thermistors; Yellow Spring Instruments YSI 44006 with a resistance of about 2.95 × 104 Ω at 0°C, with a temperature coefficient of about 5% per °C. The calibration was undertaken at VAW (Versuchanstalt für Wasserbau, Hydrologie und Glaziologie), ETH Zürich. The absolute accuracy is estimated as ±0.05°C and the relative accuracy ±0.02°C [Vonder Mühll and Holub, 1992]. The thermistors were coupled to a Colorflex CY thermistor chain. A weight was placed at the lower end of the chain to ensure that the thermistor positions remained constant and that the chain reached the bottom of the hole. Depths of thermistors followed the general instructions for the PACE boreholes [Harris et al., 2001b]; levels were: 0.2 m, 0.4 m, 0.8 m, 1.2 m, 1.6 m, 2.0 m, 2.5 m, 3.0 m, 5.0 m, 7.0 m, 9.0 m, 10.0 m, 13.0, 15.0 m, 20.0 m, 25.0 m, 30.0 m and 10 m spacing to 80.0 m, and then denser again to 100.0 m. On Juvvasshøe, an additional chain of 5 thermistors from 106.5 to 129.0 m was installed below the standard PACE depth. The measurement interval of the thermistors in the upper 15 m of the control borehole is every 6 hours. Temperatures of the thermistors below 5 m in the main borehole are taken once every 24 hours. Periodic recalibration of the installed thermistors is possible and the holes remain accessible for other probes in future. Borehole casing, sensors and data logging equipment were assembled according to guidelines provided by the PACE project in order to standardize procedures and ensure comparability between sites [Harris et al., 2001b]. This also ensured reliability and serviceability. For more details about borehole instrumentation, see Isaksen et al. .
 Meteorological stations were established at all drill sites. At Janssonhaugen air temperatures were measured at three levels (2.0 m, 1.0 m and 0.1 m above ground level) and wind speed and direction at 2 m. At Tarfalaryggen air temperature were measured at 2 m. At Juvvasshøe an automatic micrometeorological station was established, measuring air temperature, humidity, radiation and wind speed and direction. The station has a 2 m anemometer height. An analysis of the near surface energy fluxes on Juvvasshøe is presented by Isaksen et al. .
 Data series from Janssonhaugen are complete since instrumentation of the borehole in May 1998. Because of instrument failure at Tarfalaryggen there are some major data gaps in 2001, 2002 and 2003. At Juvvasshøe there is a major data gap in 2001. Drilling of the boreholes lasted 3–5 days. To reduce any thermal disturbance from the drilling [e.g., Stulc, 1995] analyses of the borehole data time series starts approximately 15 months after the boreholes were established. Thus, in the present analyses Janssonhaugen time series starts in August 1999, Tarfalaryggen in June 2001 and Juvvasshøe November 2000.
 The main purpose of thermal monitoring at these three boreholes is to collect a long time series of data through several decades. Long-term monitoring will contribute data to the Global Terrestrial Network for Permafrost (GTN-P) of the Global Climate Observing System (GCOS). The Norwegian Meteorological Institute is responsible for the data from the Janssonhaugen and the Juvvasshøe boreholes. The Tarfala Research Station (Stockholm University) is responsible for the data from the Tarfalryggen borehole. On Juvvasshøe the Norwegian Meteorological Institute intends to install a permanent weather station associated with the deep borehole.
5.1. Seasonal Ground Temperatures and Relation With Climate
 Correlation between monthly ground temperature at 0.2 m depth and monthly air temperatures at standard screen height (2 m) is shown in Figure 2. The results indicate a strong correlation with air temperatures at all sites. The coefficient of determination (R2) at Janssonhaugen is 0.98, at Tarfalaryggen 0.91 and at Juvvasshøe 0.96. Mean temperature values are found in Table 1.
Table 1. Present Mean Temperatures at Study Sitesa
MAT, mean air temperature; MGT, mean ground temperature.
Data from Janssonhaugen are complete, except for some missing days of air temperature in March 2003, which were interpolated.
Data period is between April 2003 and March 2005.
Missing ground temperature data in 2001 are interpolated.
 The differences between these three sites are mainly explained by differences in thickness of snow cover and type of ground surface. In general all sites are exposed to strong winds, leading to an absence of snow or only a thin snow cover. The high correlation at Janssonhaugen can be explained by fine-grained in situ weathered bedrock extending virtually to the ground surface, and the absence of a significant boundary layer of snow, vegetation and organic material. At Tarfalaryggen, field observations reveal that during a normal winter 0.2–0.3 m of snow covers the site. At Juvvasshøe strong winds result in little snow cover, with a maximum of 0.1–0.2 m in April. In contrast to Janssonhaugen, the blocky surface materials at Tarfalaryggen and Juvvasshøe collect some snow and have more complex thermal properties.
Figure 3 presents a temperature-time series from selected depths in the zone of annual temperature variations at Janssonhaugen. In winter absence, or only very thin snow cover gives no insulating effect, resulting in high-frequency temperature variations at the ground surface. During freezing, the ground temperature within the active layer falls rapidly below 0°C, indicating low water content and a dry surface. Only during freezing in the year 2000 was a significant zero curtain effect observed. This was due to exceptional rainfall in September and October, with monthly values 2–3 times above normal, combined with the warmest October in Svalbard since records began in 1912. Below the permafrost table, high-frequency temperature variations diminish rapidly and closely follow a sinusoidal curve at 5.0 m depth. At 10 m depth, the annual wave is delayed by approximately 220 days and the annual amplitude is diminished to 0.8°–1.0°C.
 At Juvvasshøe high turbulent fluxes give good ventilation in the upper surface layers [Isaksen et al., 2003]. In addition weakly developed zero curtain effects suggest low water content within the active layer. Analyses of ground temperatures at Tarfalaryggen indicate similar near-surface ground conditions to those at Juvvasshøe, but slightly higher ice content near permafrost table. In general there is a close relation between air, ground surface, and ground temperatures, especially at Janssonhaugen and Juvvasshøe. This results in a preserved climate signal that penetrates into the permafrost, with no large perturbations caused by near-surface and surface conditions, such as significant variations in snow cover and latent heat related to phase changes.
5.2. Active Layer Thickness
 The term “active layer” refers to the layer of ground between the surface and permafrost that undergoes seasonal freezing and thawing [e.g., Burn, 1998a]. Figure 4 shows the maximum annual development of the active layer depth at each site. On Janssonhaugen active layer depth was greatest in 2004, with a maximum depth of 1.7 m. The minimum active layer depth of 1.4 m was recorded in 1999. On Juvvasshøe, depth of the active layer shows a significant response to warm summers. The summers of 2002 and 2003 were among the warmest on record in the region. Active layer depths were 20% greater in these summers than in previous years. On Tarfalaryggen, active layer thickness is between 1.5 and 1.6 m. As the snow cover at the boreholes is usually thin or absent, surfaces are normally dry and water content in the ground is low, so that active layer thickness is well correlated with local summer air temperatures on an interannual basis.
5.3. Permafrost Temperature Profiles
 The temperature profile from Janssonhaugen recorded one year after drilling was analyzed by Isaksen et al. . A ground surface temperature reconstruction based on a heat conduction inversion model indicated a warming of the permafrost surface of 1.5°C ± 0.5°C over the last 6–8 decades. A brief description of the thermal profiles from all three permafrost boreholes was presented by Isaksen et al.  and was later compared with all the boreholes in the European PACE permafrost network [Harris et al., 2003].
Figure 5 shows updated ground temperature profiles from Janssonhaugen, Taralaryggen and Juvvasshøe, recorded in April 2005. The smooth profiles support a low geothermal disturbance from undesirable elements and nonclimate sources [cf. Isaksen et al., 2000]. The upper 15–20 m of the temperature profiles is influenced by seasonal temperature variations and is less ideal for analysis of interannual to decadal variations in ground temperatures. On Janssonhaugen, the gradient changes gradually downward from 0.010°C m –1 at 25 m to 0.037°–0.038°C m –1 at 95 m. At both Tarfalaryggen and Juvvasshøe, the gradients are smaller and negative in the upper 40–45 m. On Tarfalaryggen the temperature gradient changes from −0.015°C m –1 at 25 m to 0.010°–0.011°C m –1 at 95 m. Similarly, the Juvvasshøe thermal gradient is −0.011°C m –1 at 25 m and 0.010°–0.011°C m –1 at 126.5 m. Large-scale topographic influence may explain the low geothermal gradients at Tarfalaryggen and Juvvasshøe since adjacent valleys are ∼400 and ∼1100 m deeper, respectively, than the location of the boreholes. To penetrate into an undisturbed temperature field and obtain undisturbed heat flow values, the boreholes should probably be at least 5 to 10 times deeper. A careful evaluation of the thermal gradients on Tarfalaryggen and Juvvasshøe should be performed by comparison with geological conditions and three-dimensional transient models, which include topography and effects of differences in local surface temperatures [e.g., Gruber et al., 2004c; Kohl, 1999; Wegmann et al., 1998]. This could lead to a more detailed interpretation on the thermal gradients in the two boreholes.
 Although derivation of climate signals from temperature profiles in mountain permafrost is complicated by 3-D effects [e.g., Gruber et al., 2004c], the three present boreholes are situated on plateaus or ridges with minor topographic relief within a radius of 100–200 m, with smooth ground surfaces and quite uniform snow conditions. This ensures a low disturbance from small-scale 3-D thermal effects in the upper parts of the boreholes. In Figure 6 data were obtained by subtracting temperatures for steady state conditions from measured temperatures below the ZAA. The steady state temperatures were estimated by extrapolating the thermal gradient measured in the lowermost part of the borehole, which is assumed to be unaffected by recent warming trends. All boreholes show a significant warmside deviation in their thermal profiles to 70 m depth, which is most likely associated with surface warming during the last decades, with the greatest change occurring in the northernmost borehole in Svalbard [cf. Harris et al., 2003; Isaksen et al., 2001]. Upward extrapolation to the surface of the temperature gradient between 30–20 m depth indicates surface temperature changes with a magnitude of ∼1.4°C, ∼1.1°C and ∼1.0°C for Janssonhaugen, Tarfalaryggen and Juvvasshøe respectively (Figure 6). No evidence exist so far that the observed anomalies in the upper part of the thermal profiles are caused by other factors than past changes in ground surface temperatures. In addition, the similarity of the two thermal profiles from Tarfalaryggen and Juvvasshøe suggest a general common effect, that is, a warming of the upper permafrost surface. The lower part of the temperature profiles represents fine-scale fluctuations around the quasi-steady state field and probably minor variations due to thermal conductivity contrasts.
5.4. Time Series at Different Depths
 In all boreholes the annual temperature signal below 20 m depth is free of any response to annual or shorter-term temperature variations. At these depths any recorded systematic temperature time variations must correspond to a longer period of several years [e.g., Cermak et al., 2000; Osterkamp and Romanovsky, 1999]. Figure 7 shows results from the continuous ground temperature monitoring at selected depths below ZAA. These time series suggest that permafrost is warming at a significant rate. Results from four to six years monitoring on Janssonhaugen, Tarfalaryggen and Juvvasshøe show that the ground temperature has increased by 0.26°C, 0.18°C, and 0.17°C, respectively, at 25 m depth and increased by 0.19°C, 0.15°C, and 0.14°C, respectively, at 30 m depth. At 40 m depth ground temperatures have increased by 0.07°–0.09°C in all boreholes (Figure 7). Observed warming is statistically significant to 60 m depth at all sites. This result strongly supports the previous interpretation that most of the anomalies observed in the temperature depth profiles (see Figure 6) are associated with surface warming.
 Because temperature has been monitored continuously over a period of several years, it is possible to calculate the actual rate of temperature change as a function of depth. Figure 8 shows calculated linear trends of time series between 20 and 60 m depth. At greater depths longer time series are needed to draw conclusions on the rate of warming. Present warming rates at for instance 30 m depth are in the order of 0.025°–0.035°C yr−1, with greatest warming rates on Tarfalaryggen and Janssonhaugen.
 Determination of the thermal diffusivity is important in transient calculations, as it governs the rate of propagation of a temperature disturbance through the ground [e.g., Gold and Lachenbruch, 1973]. Higher diffusivity accelerates the temperature signals. Given the parameters for the mean annual surface temperature T0 (°C), the annual amplitude of the surface temperature A0 (°C) and the depth in the ground z (m), the solution for the annual cycle based on thermal diffusivity is given by:
were κ is the thermal diffusivity (m2 s−1) and P is the period of the annual signal (s).
 The thermal diffusivity κ in the zone of seasonal temperature variations may be calculated in terms of the amplitude decrease with depth using the term A0 · e−z(π/κP)1/2:
where ω is the frequency of the annual signal (ω = 2π/P), h is the distance between the two temperature sensors (m), A1 is the amplitude of the upper temperature sensor (°C) and A2 is the amplitude of the lower temperature sensor (°C).
 When plotting ln(A(z)) against z the values of κ fall in a narrow range below 4 m in all the boreholes. The mean thermal diffusivity κ can be determined from the slope m of a linear fit to the natural logarithm of the maximum amplitude A(z), plotted versus depth, assuming that the period p is one year [e.g., Vonder Mühll and Haeberli, 1990]:
which gives typical diffusivity values for bedrock of 1.0 ± 0.2 × 10−6 m2 s−1 on Janssonhaugen and 1.2 ± 0.2 × 10−6 m2 s−1 on Juvvasshøe [cf. Isaksen et al., 2000, 2003]. Because of greater data gaps in time series on Tarfalaryggen, missing data were calculated using regression and thermal diffusivity was estimated to 1.4 ± 0.3 × 10−6 m2 s−1. Generally the drilled rock strata are relatively homogenous without any pronounced conductivity contrasts [Isaksen et al., 2000, 2001], thus the diffusivity values obtained in the zone of annual variations were also used for calculations at greater depths.
 The amplitude A(z) of the subsurface response to the variation of the surface conditions (A0) is a function of thermal diffusivity κ [e.g., Cermak et al., 2000]:
 Recorded temperature trends at 40–50 m are used to calculate warming rates of the permafrost surface, representative for the last decades [cf. Cermak et al., 2000]. Present decadal warming rates at the permafrost table at the three study sites are in the order of 0.04–0.07°C yr−1, with greatest warming on Janssonhaugen and Tarfalaryggen. The present trend is for accelerated warming during the last decade at all sites.
6.1. Recent Warming of the Permafrost and Relation to Air Temperature Records
 During the instrumental record of air temperature in the 20th century there have been substantial decadal and multidecadal temperature variations in the regions of the three study sites. A rather cold period around 1900 was followed by “the early 20th century warming”, which culminated in the 1930s. A period of cooling followed, before the recent period of warming, which has dominated Svalbard and most of Scandinavia since the 1960s [Hanssen-Bauer and Førland, 2000; Hanssen-Bauer, 2002]. Recent trends in annual mean air temperature from meteorological stations adjacent to the boreholes are presented in Figure 9. During the period 1965–2004, positive trends at the 5% significance level (Mann-Kendall) were detected at all stations. For the 35 year series, the linear trend at the northernmost station in Svalbard is 0.05°C yr−1, at Tarfala in northern Scandinavia it is 0.04°C yr−1 and at the southernmost location, adjacent to Juvvasshøe, it is 0.03°C yr−1. For the three selected stations, regression analyses indicate high correlation with the local air temperature observations made at the study sites. On a monthly basis the coefficient of determination (R2) at the two northernmost locations is 0.99 and at the southernmost it is 0.97. However, the pronounced 20th century air temperature fluctuations and large interannual variability complicate the analyses of long-term trends [e.g., Hanssen-Bauer and Førland, 2000]. For the weather station in Svalbard, the difference between the highest and lowest monthly mean in December–March is about 25°C. In one or two years the annual air temperature can differ by 5°–6°C, which is a large interannual fluctuation. An eventual trend signal has therefore to be extracted from the noisy air temperature record and significant trends from the air temperature record may be difficult to establish.
Smith and Riseborough  conclude that permafrost monitoring at exposed bedrock sites produces the most direct signal of climate change on the ground thermal regime. The close permafrost-climate relationship reported in this paper suggests that the geothermal records from the study sites may be powerful indicators of climate change in Svalbard and in Scandinavia. All three boreholes show a significant warmside deviation in their thermal profiles, which are most likely associated with surface warming during the last decades, with the greatest change occurring in the northernmost borehole in Svalbard. Present decadal warming rates at permafrost surface are in order of 0.04°–0.07°C yr-1 and statistically significant positive trends are found to 60 m depth at all sites.
6.2. Possible Consequences and Impacts of Warming Permafrost in Svalbard and Scandinavia
 In the discontinuous permafrost in the Arctic and sub-Artic, thaw settlement related to permafrost degradation is presently responsible for damage to houses, roads, airports, military installations, pipelines, and other facilities founded on ice-rich permafrost [Osterkamp et al., 1997]. Permafrost in these areas is already at temperatures close to thawing, and further temperature increases are very likely to result in extremely serious impacts on infrastructure. In the continuous permafrost in Svalbard, projected climate change [e.g., Hanssen-Bauer, 2002] is not likely to pose an immediate threat to infrastructure if the correct permafrost engineering design procedures have been followed [Instanes, 2003]. However, downscaled temperature scenarios for Svalbard indicate a warming of about 0.03°C yr−1 in summer to year 2050 [Hanssen-Bauer, 2002] leading to a warming of the cold continuous permafrost during the thaw period, which is likely to increase the thickness of the active layer, leading to possible thaw settlement and slope instability. Long-term monitoring of such changes is co-ordinated under the Circumpolar Active Layer Monitoring program (CALM) of the International Permafrost Association (IPA) [Nelson and Anisimov, 2002].
 In the high mountains of Scandinavia, warming permafrost may lead to rapid deterioration of the load bearing capacity of frozen soil and rock. On slopes this will affect both permafrost creep rates and landslide-related processes [Harris et al., 2000, 2001a, 2001c; Davies et al., 2001], leading for instance to increased rockfall [cf. Gruber et al., 2004b]. In Scandinavia permafrost underlies a significant proportion of the mountain areas. If recent trends continue, it will take several centuries for permafrost to disappear completely. However, negative consequences of this degradation will be pronounced from the very beginning because the highest ice content in permafrost usually is found in the upper few tens of meters.
 With expanding tourism, communication networks, hydropower production etc. in thermally sensitive Scandinavian mountain permafrost terrain, knowledge and understanding of potential future changes in permafrost temperatures is essential [cf. Haeberli, 1992]. Because of the sparse network of meteorological stations in high altitudes in Scandinavia and in Svalbard, permafrost monitoring of the three reported boreholes should be continued for several decades, along with the measurements from the associated weather stations. The three boreholes and the remaining PACE permafrost-monitoring network will, in future, provide fundamental data, with a high temporal resolution, for the study of permafrost temperatures under future climate development and probable accelerated climate warming in the mountains and northern latitudes of Europe.
 The PACE mountain permafrost borehole network offers the prospect of long-term monitoring as part of the Global Terrestrial Network for Permafrost (GTN-P). The three boreholes in Svalbard and Scandinavia reported in this study show a significant warmside deviation in their thermal profiles, suggesting surface warming through the 20th century, with the greatest change occurring in the northernmost borehole in Svalbard. The results from four to six years of ground temperature monitoring in permafrost at the three sites indicate that the permafrost has warmed considerably during the last decades. Significant warming is detectable down to at least 60 m depth. Present decadal warming rates at the permafrost table are in the order of 0.04°–0.07°C yr−1, with greatest warming in Svalbard and in northern Scandinavia. The present trend seems to be an accelerated warming during the last decade.
 This research was mainly funded by the European Community through the 4th Framework Permafrost and Climate in Europe (PACE) Project (Contract ENV4-CT97-0492, period 1997–2001). Support was also received from the University of Oslo, the Norwegian Meteorological Institute, the Tarfala Research station (the Stockholm University), the Norwegian Geotechnical Institute, the University Courses in Svalbard, and the Gjøvik University College. At the end of the PACE project in 2001 the Norwegian Meteorological Institute and the Tarfala Research station have been the main financial supporters. Olav Liestøl assisted in the selection of the Janssonhaugen and Juvvasshøe drill sites. Trond Eiken and Rune Strand Ødegård gave valuable field assistance during data collection and maintenance of the permafrost stations at Janssonhaugen and Juvvasshøe. Ole Humlum was responsible for the meteorological station at Janssonhaugen. Øyvind Nordli assisted with the Mann-Kendall tests and gave useful comments to the manuscript. The reviewers, Tom Osterkamp and Wilfried Haeberli, and the Associate Editor, Tingjun Zhang, gave important improvements to the manuscript. The contribution of all persons and institutions mentioned is gratefully acknowledged by the authors.