This study contains the first quantitative description of mountain permafrost in Iceland, obtained through analysis of four ground temperature time series from recently established shallow boreholes at high altitude (∼900 m above sea level) in central and eastern Iceland. Permafrost is present at three of the four boreholes, where temperatures range from −2° to 0°C. During the data-logging period, 2004–2006, the thickness of the active layer at the individual sites varied between less than 2 m and more than 5 m, mainly owing to differences in near-surface ice content. The ground temperature profiles are nonlinear, with near-surface temperature perturbations considered to be related to a recent warming of the ground surface. At the one nonpermafrost site the insulating influence of winter snow cover is more pronounced. A one-dimensional heat-conduction model was calibrated using ground temperature measurements and driven by surface temperature time series. The model results reproduce the measured ground temperature pattern well. This simplified model, simulating the results of possible future climate scenarios, indicates that permafrost degradation will occur at these sites within decades, depending on climate scenarios chosen and subsurface ice content.
 So far, only a few systematic studies concerning mountain permafrost in Iceland have been published. The literature is focused on sporadic permafrost connected to palsa formations [e.g., Friedman et al., 1971; Thórhallsdóttir, 1996] and geomorphological features indicating extensive periglacial activity [e.g., Friedman et al., 1971; Schunke, 1974; Thorarinsson, 1964; Van Vliet-Lanoë et al., 1998]. A range of borehole temperature measurements exist in Iceland, most of which are related to geothermal energy extraction [e.g., Arnórsson, 1995a], but to our knowledge there is no record of mountain permafrost being determined from boreholes. Recently, Guðmundsson  provided a rock glacier inventory, showing that active rock glaciers are widely distributed in northern and eastern Iceland. Detailed studies on rheology and morphodynamics of such rock glaciers exist [Martin and Whalley, 1987; Wangensteen et al., 2006; Whalley et al., 1995], and these features have recently been related to the existence of mountain permafrost [Farbrot et al., 2007]. Gridded data of mean annual air temperature (MAAT; 1961–1990) for Iceland [Gylfadóttir, 2003; Tveito et al., 2000] show considerable nonglaciated mountain areas with MAAT below −3°C, a value indicative of permafrost existence for areas with a thin or absent snow cover [Etzelmüller et al., 2007]. The overall altitude for this isotherm increases from north (c. 700 m above sea level, asl) to south (>1000 m asl). Thus we hypothesize that widespread mountain permafrost exists at higher altitudes in the mountains of Iceland.
 To address this hypothesis, four shallow boreholes situated at the inferred regional lower limit of mountain permafrost were drilled and equipped to continuously record the ground temperatures. This study presents two years (2004–2006) of temperature data from these boreholes. Here we characterize the ground thermal regime and provide the first estimate of permafrost sensitivity to past and future climate changes using a numerical transient 1-D heat flow model. Furthermore, we present direct current (DC) resistivity tomography measurements to characterize and delineate the extent of possible subsurface permafrost bodies.
2. General Setting and Borehole Locations
 Iceland is located at a point where the asthenospheric flow under the Mid-Atlantic Ridge interacts and mixes with a deep-seated mantle plume [Shen et al., 1998; Wolfe et al., 1997]. This implies generally high heat fluxes due to both heat conduction from the partially molten layer at approximately 10 km depth, as well as from mechanical energy released within the volcanic rift zone by crustal movements [Flóvenz and Sæmundsson, 1993]. Conditions are, however, complicated by convection of surface water, especially within the volcanic rift zone where near-surface heat flow is negligible because of a thick permeable lava formation at the surface [Arnórsson, 1995a, 1995b]. Furthermore, cooling and solidification of local intrusions add heat, altogether resulting in a highly heterogeneous geothermal pattern [cf. Flóvenz and Sæmundsson, 1993].
 Iceland is characterized by maritime conditions with cool summers and mild winters. In the lowland areas, the MAAT for the period 1961–1990 was 4°–5°C in the south, 3°–4°C in the east and west parts and 2°–3°C in northern, coastal parts of the country [Tveito et al., 2000]. The majority of the precipitation is brought with easterly and southerly prevailing winds [Einarsson, 1984]. Thus mean annual precipitation increases from about 400 mm in the central and northern parts of the country to more than 4000 mm in the southeast.
 The four boreholes are located in central and northeastern Iceland at ∼890–930 m asl (Figure 1). The Hágöngur borehole is situated within the volcanic rift zone, the Snæfell borehole within an intraplate volcanic belt, whereas Gagnheiði and Vopnafjörður boreholes are located to the east within Tertiary basalt formations (Figure 1; see Table 1 for details). All boreholes are relatively shallow (12–22 m deep), penetrating through a shallow sediment cover into basaltic bedrock. All localities are on elevated plateaus; significant slopes are more than 100 m away from the boreholes. The sites are not vegetated and are exposed to winds, and are therefore prone to considerable snow redistribution. At Gagnheiði, a meteorological station run by the Icelandic Meteorological Office is present, providing more or less continuous daily air temperatures since 1994.
Table 1. Borehole Locations and Key Site Information
The site lies at Sauðafell, a small, SW-NE-aligned hyaloclastite ridge to the north of Snæfell. This ridge is a product of subglacial, basaltic eruptions along the poorly developed fissure swarm [Hards et al., 2000]. Volcanic activity probably occurred latest about 0.7–0.8 Ma ago [Kristjansson et al., 1988].
The bedrock in the area is dense basaltic strata, with an Upper Tertiary age [Saemundsson, 1979].
The site lies in a massif, dissected by u-shaped valleys with individual tops reaching well above 1200 m asl. Active patterned ground is dominant on most of the plateau, indicating extensive periglacial activity. The bedrock consists of basaltic strata of Upper Tertiary age [Saemundsson, 1979].
sand, weathering material/volcanic ash, very young volcanic area
sand, weathering material/volcanic ash
thin sediment cover, ground moraine
fine-grained sediments, 2–2.5 m, patterned ground and polygons
 Initially, the boreholes were equipped with UTL-1 miniature temperature data loggers (MTDs) having an accuracy of ±0.27°C or better [cf. Hoelzle et al., 1999]. Measurements were recorded at intervals of 2 hours in the upper parts of the boreholes and 6 hours in the lower parts. In August 2005, the boreholes at Snæfell and Gagnheiði were equipped with thermistor chains consisting of 26 and 15 YSI S40006 thermistors, respectively. These thermistors have an absolute accuracy of ±0.05°C [Vonder Mühll and Holub, 1992], and measurements were performed at 6 hourly intervals. The depths of the installed MTDs and thermistors are summarized in Table 1. During a field campaign in August 2006, higher-accuracy temperature measurements were made in the Hágöngur and Vopnafjörður boreholes, using a multimeter connected to a thermistor cable, similar to those installed in the boreholes in 2005. Single MTDs, measuring at intervals of 2 hours, were installed near the ground surface (c. 5 cm depth) to record ground surface temperature (GST), and in radiation shields close to the boreholes to measure surface air temperatures (SAT). At Gagnheiði, the SATs from the meteorological station have been used.
 By correlating measured daily SAT values at the borehole sites with corresponding SAT values from nearby weather stations, a simple linear regression model was established. Using MAATs from the actual weather stations as input, MAAT at the borehole sites for the “normal” period 1961–1990 was estimated. Correspondingly, measured daily GST values were correlated with SAT values from nearby weather stations, and a linear regression procedure was used to estimate mean annual ground surface temperatures (MAGST) at the borehole sites. A 30-year period is considered long enough to calculate a representative average, smoothing out pronounced fluctuations and interannual variability [cf. Isaksen et al., 2005]. For estimates of MAGST to be reliable, however, snow conditions need to have been fairly similar during the 2004–2006 and the 1961–1990 periods. Since data on timing and depth of past snow covers are unavailable for the earlier period, we use only data from areas which at present remain more or less snow-free to estimate 1961–1990 MAGSTs. That the sites remain snow-free is indicated by the frequent temperature fluctuations during the winter. We assume that these locations were also predominantly snow-free during the 1961–1990 period.
3.2. Heat-Conduction Modeling
 To model the subsurface temperature distribution, a numerical 1-D model capable of modeling ground temperatures forced by annual, monthly or daily GST input values is used. The model solves the heat-conduction equations, following
[e.g., Williams and Smith, 1989]. As boundary conditions, we prescribe time series of GST and the geothermal heat flux Qgeo = −k · at depth. Values for Qgeo were obtained from Flóvenz and Sæmundsson , and must be regarded as regionally applicable values rather than local estimates. The thermal properties of the ground are described in terms of density ρ, thermal conductivity k and heat capacity c. Typical values for Icelandic basalt were derived from the literature [Flóvenz and Sæmundsson, 1993] and are listed in Table 3. The presence of water in the substrate has a twofold effect on the thermal properties. First, the thermal properties of water and ice are different to those of the matrix, and we consider effective values as a linear mixture of the substrate and its pore water content w. For example, the effective heat conductivity is expressed as
In case the material is frozen, we use the value for ice (e.g., kice) instead of that for water (kH2O). Second, the water content affects the thermal properties during the phase transition from liquid to solid or vice versa. The migration of the phase transition is also known as the “Stefan-problem” [e.g., Lunardini, 1996]: During freezing or thawing, the latent heat associated with this phase change is released or consumed, respectively. Thereby, the propagation of the freezing or thawing front is effectively retarded. Comparing two samples of the same soil at identical temperature and having identical thermal properties except for the water content, the one having the lower water content will freeze or thaw earlier, while the other will stay at 0°C until most of the water is frozen or thawed [e.g., Williams and Smith, 1989]. This effect is termed the “zero curtain” effect since, until the phase change is completed, the temperature remains constant at 0°C. In our model, we consider the change of latent heat L due to phase changes of the pore water by increasing the heat capacity within a small temperature interval of ±0.1°C around the freezing temperature [e.g., Wegmann et al., 1998]
Further, any effects of heat advection related to groundwater flow are neglected in this study. The heat-conduction equation (equation (1)) was discretized along the borehole depth using finite differences and subsequently solved by applying the method of lines [Schiesser, 1991]. The model was calibrated using measured ground temperatures, and the values of w and k were adjusted to match modeled and measured temperature distributions, annual amplitudes at a given depth and modeled and measured thicknesses of the active layer.
3.3. DC Resistivity Tomography
 The principle of the DC resistivity method relies on the fact that different materials have different abilities to conduct electricity [e.g., Reynolds, 1997]. The application of resistivity soundings in permafrost mapping is justified through the large contrast between the resistivity of water (∼101 to 102 Ωm) and permafrost with ground ice (∼103 to 106 Ωm). Therefore the method allows for the identification of ground ice, and hence permafrost in water-containing material [e.g., Hauck and Vonder Mühll, 2003]. For two-dimensional (2-D) resistivity tomography, an ABEM Lund (®ABEM Sweden) multielectrode, high-resolution 2-D resistivity system was used. This system comprises a linear array of 61 electrodes at constant spacing connected through a multicore cable. The DC current flows into and out of the ground through two current electrodes (C1 and C2) and the potential difference is measured between two potential electrodes (P1 and P2). The control software is designed to carry out a series of constant separation traverses with increasing electrode spacing at each traverse. A Wenner Alpha array (electrodes equally spaced C1-P1-C2-P2) was used owing to its favorable geometrical factor [cf. Hauck et al., 2003]. We used electrode spacing of 2 m and 5 m, giving maximum penetration depth of roughly 11 m and 26 m, respectively. The length of a section can be increased by moving individual cables from one end to the other, allowing flexible profile lengths. The system automatically measures the resistance for a given electrode combination and spacing, and calculates apparent resistivity. A 2-D-model interpretation was undertaken using the software package RES2DINV, which performs smoothness-constrained inversions using finite difference forward modeling [Loke and Barker, 1996]. In all cases topographic corrections were incorporated in the inversion algorithm.
4.1. Air and Ground Surface Temperatures
 At Hágöngur the SAT installation was broken in October 2004, so the main part of the 2004–2005 data record is missing. Furthermore, the MTD measuring SAT at Snæfell malfunctioned January to mid-March 2005. Since the SATs at the three eastern sites are highly correlated to each other, the missing SATs from Snæfell have been estimated through correlation with daily SATs from Gagnheiði and subsequent linear regression (R2 = 0.973, N = 519, p < 0.001).
 The site measurements during the period 2004–2006 show that during summer GSTs follow the SATs well, while during winter, differences occur owing to the influence of the snow cover (Figure 2). The influence of snow is especially pronounced for the Vopnafjörður borehole record where GSTs are fairly stable during longer periods of the winter. In late January and early February 2005 several melting periods occurred during which the snow at the Vopnafjörður site apparently disappeared as indicated by the returned coincidence of the GST and SAT records (Figure 2). Thus, in this case, a high SAT can be responsible for subsequent cooling of the ground in case the formerly present, insulating snow cover disappears.
 During the measurement period mean SATs were fairly similar at the three eastern borehole sites (−1.6° to −1.4°C in 2004–2005; −2.0° to −1.8°C in 2005–2006), whereas Hágöngur had somewhat higher temperatures (−1.3°C in 2005–2006) (Table 2). For the reference period (1961–1990), the estimated MAATs are ∼0.8°C lower at all sites. For the measurement period, the mean ground surface temperature (MGST) ranged from −1.2°C at Gagnheiði to 0.2°C at Vopnafjörður (Table 2). The estimated MAGSTs (1961–1990) were −1.9°C, −1.6°C and −0.8°C at Snæfell, Gagnheiði and Hágöngur, respectively. In these analyses, R2 ≥ 0.85 and ≥0.75 were achieved for the MAATs and MAGSTs, respectively (Figure 3; for all records, p < 0.001). For Vopnafjörður, MAGST was not estimated since the prevalent snow cover during winter inhibits any direct linkage between SAT and GST. The measured surface offsets between annual MGST and MAT were ∼2°C at Vopnafjörður and ≤1°C at the other sites, whereas the estimated long-term surface offset (1961–1990) is 0.7°C at Gagnheiði and 1.2°C at Snæfell and Hágöngur.
Table 2. Measured and Estimated SATs and GSTs at the Borehole Sitesa
Sauðafell (Near Snæfell)
Vopnafjörður Mountain Plateau
MAT, mean air temperature; MGST, mean ground surface temperature; FDD, freezing degree days; TDD, thawing degree days; MAAT, mean annual air temperature; MAGST, mean annual ground surface temperature; N, number of values in the statistical analyses; R2, coefficient of determination.
MGST – MAT 04–05
MGST – MAT 05–06
MGST – MAT 04–06
Estimated long-term offset between air and ground temperatures
4.2. Borehole Temperatures
 Continuous subzero temperatures over the two years period were found within the boreholes at Hágöngur, Snæfell and Gagnheiði (Figures 4 and 5) . Temperature fluctuations within the boreholes were small (<2.5°C in 3 m depth), and temperatures ranged from −2° to 0°C beneath the active layer. Thermistor chain measurements at Hágöngur showed unfrozen ground at 12.2 m depth (Figure 4). This is also indicated by the observation of moisture on the sensor when the cable was pulled up. Below 5 m depth in the Vopnafjörður borehole, the temperature ranged from 0.3° to 2.0°C (Figures 4 and 5). The annual amplitudes of ground temperatures are somewhat larger than those observed in the permafrost boreholes at corresponding depth, thereby indicating low water content. Active layer thicknesses of the permafrost boreholes were <2 m, ∼5 m and ∼6 m in the Snæfell, Gagnheiði and Hágöngur boreholes, respectively (Figure 4). At Gagnheiði and Hágöngur the active layer totally refroze between late December and early January, whereas the thinner active layer at Snæfell was already refrozen by late October. At the Vopnafjörður borehole seasonal freezing prevailed down to ∼2 m for both monitoring years (Figure 4).
4.3. DC Resistivity Validation
 At the Snæfell borehole, the resistivity measurements reveal horizontally homogenous conditions along the top plateau (Figures 6a and 6b). The vertical profile consists of a low-resistivity upper zone associated with the active layer (700–1500 Ωm), this is underlain by a high-resistivity zone (2000–10,000 Ωm), followed by an intermediate resistive zone at the base (1000–1500 Ωm). The upper layer is not visible in Figure 6a owing to lower resolution of the resistivity tomography profile. From calibration with the borehole, resistivities >1000 Ωm are considered as permafrost indicators in the Snæfell area. The resistivity profile down the northern slope of the hyaloclastite ridge reveals distinct high-resistive areas dissected by less resistive zones (Figure 6c). At lower elevations, a low-resistive area is present, indicating that permafrost seems to disappear at 850 m asl. The resistivity profile across the Vopnafjörður borehole reveals resistivity values mainly below 1000 Ωm, with some areas having values above 2000 Ωm. The profile across Gagnheiði borehole reveals fairly heterogeneous conditions showing generally lower resistivity values than at Snæfell; all values are <1000 Ωm. The site is probably relatively dry with temperatures close to 0°C, thus resistivity differences are presumably not related to permafrost.
4.4. Numerical Modeling
 Ground temperatures were modeled for the Gagnheiði and Snæfell sites for the season 2005–2006. The model performed least well for the last reading in August 2006 for both sites (see Figures 7 and 8) . This is probably because the model does not consider vertical water flow and the resulting variations in water content over the melting season. The subsurface substrate was very likely drier during summer than it was during spring and fall. On the basis of the modeling it is evident that the water content at the Gangheidi site is considerably lower than at Snæfell (Table 3). The model indicates a permafrost thickness of c. 35 m at Snæfell, and a somewhat lower value (<30 m) at Gagnheiði.
Table 3. Modeling Parameters
Sauðafell (Near Snæfell)
Number of layers
Depth of discontinuity, m
Depth increment, m
Geothermal heat flux, W m−2
Thermal conductivity, W m−2 K−1
Density, kg m−3
Heat capacity, J kg−1 K−1
Water content, volume %
Total RMS error, °C
Total R2 between measured and modeled temperature
 In order to investigate the dynamics of the permafrost at Snæfell and Gagnheiði, we first estimated the mean annual GSTs since 1955 from linear regression with known SATs from nearby Egilsstaðir meteorological station (Figure 9). These GSTs were set as constrains for the 1-D model, which used a steady state linear thermal gradient with a MAGST 0.5°C lower than the present MAGST as starting conditions (−1.5°C). This approach probably underestimates near-surface temperatures, but the deeper layers were probably influenced by the cool climate at the end of the “Little Ice Age” and lower temperatures are therefore more realistic [cf. Kirkbride, 2002].
 In order to investigate the sensitivity of the permafrost to climatic warming, we modeled the thermal response at Snæfell and Gagnheiði to linear increases in SATs of 0.01, 0.02 and 0.03°Ca−1. These estimates are in line with recent climate change scenarios for Iceland from the Intergovernmental Panel on Climate Change (IPCC) [Benestad, 2005]. Also the response of a step change of MAGST from the estimated values at the various sites to +0.5°C was modeled. The results of these sensitivity experiments indicate that for continued warming, the permafrost would disappear relatively rapidly at both sites (Table 4). The results further imply that the response is somewhat slower at Snæfell compared to Gagnheiði. In these scenarios we did not consider any changes of snow cover duration and thickness due to a future temperature increase.
Table 4. Estimated Numbers of Years Before Total Degradation of Permafrost Forced by Different Warming Rates of MAGSTa
Start ground surface temperature was −1°C, and the start ground temperature profile was the last modeled profile defined in Figure 7. One run defined a step function to a constant MAGST of +0.5°C.
5.1. Surface Offset Between Air and Ground Surface Temperatures
Smith and Riseborough  have shown that for arctic regions, the air-ground temperature difference is mainly related to the insulating effect of winter snow cover (nival offset) and especially in continental regions the insulating effect of vegetation. They used so-called n factors (depending on snow depth and MAAT) as transfer functions between MAAT and MAGST. Anisimov and Nelson  used the same concept for driving a regional permafrost model using predicted SAT for the Northern Hemisphere as input data. Our procedure for extrapolating MGSTs at the boreholes back in time has two important assumptions for the timescales in question (decades): (1) The difference between annual SAT at the actual weather station and at the borehole is fairly constant, and (2) the surface offset between annual GST and SAT is fairly constant. Hence the regression between GST and SAT at a weather station may not be valid if any of these assumptions are violated. Furthermore, the modeled impact of future climate scenarios may be unreliable if the second assumption is violated.
 We believe that the first assumption is reasonable since the weather stations used in the correlation procedures are situated close to the boreholes. However, this still requires that annual temperature lapse rates are fairly constant. The borehole sites are wind-exposed, and except for Vopnafjörður the sites seem to be more or less blown bare during winter. Hence we believe that there is a close coupling between SAT and GST at these sites. Furthermore, owing to low nival offset, the air-ground temperature differences are low (<1.5°C), both for the measuring period, with low variability between the two measuring years, and the estimated 1961–1990 period. Similar observations have been made at other permafrost boreholes in Scandinavia and in Svalbard [Isaksen et al., 2000, 2001, 2003, 2007]. It is shown by, for example, Putnam and Chapman  and Harris and Chapman  that SAT and GST are closely related on timescales significant for climate change (decadal), even under conditions of changing snow coverage. Hence we believe that our estimates of the thermal offset at bare-blown sites, based on our site measurements, are valid also for past and future conditions. This assumption will be further evaluated using future measurements of GSTs and SATs at the borehole sites.
 At regularly snow-covered sites, the snow-buffering effect is certainly highly variable since the snow cover in Iceland often is temperate, even at higher altitude, owing to the maritime climate involving frequent melting periods during winter [Etzelmüller et al., 2007]. On the other hand, when the melting periods are strong enough to melt the snow completely, a subsequent penetration of the freezing front is enhanced. Hence some of the warming effect may be counteracted. However, the surface offset between SAT and GST in areas having considerable winter snow cover requires further systematic investigation. Nevertheless, we have shown that mountain permafrost is present in Iceland at sites where MAAT is below −2°C. Thus permafrost is presumably widespread at high elevations in northern and eastern parts of Iceland [Etzelmüller et al., 2007].
5.2. Active Layer and Permafrost Thicknesses
 Differences in active layer thickness were found between Snæfell (<2 m) and Gagnheiði (∼5 m) even though the temperature conditions were fairly similar. This difference is attributed to higher water/ice content at the Snæfell site due to greater porosity of the bedrock. The model calibration indicates a volumetric average annual water content of >10% at Snæfell and <5% at Gagnheiði. Resistivity profiles also indicate larger ice content close to the surface, which in turn would reduce the thaw during summer. Moreover, the thermal conductivity of Icelandic nonfrozen basalts varies relatively little (1.6–1.9 Wm−1°C−1) [Flóvenz and Sæmundsson, 1993], so differences in thermal properties of the bedrock at the sites are unlikely to be the main reason for the differences in active layer thickness. The resistivity profile with 2 m electrode spacing (Figure 6b) indicates that the active layer thickness is fairly constant across the top plateau at Snæfell.
 At Hágöngur unfrozen ground is present at 12 m depth. With an active layer thickness of 5 to 6 m, the permafrost thickness is ∼5–6 m. Thus a slight increase in temperature or snow cover thickness would rapidly degrade permafrost completely at this site. Neither of the other two permafrost boreholes reaches the permafrost base, thereby inhibiting direct measurements of the permafrost thicknesses.
 The resistivity profile down the northern slope of the hyaloclastite ridge, close to the Snæfell borehole, indicates permafrost down to ∼850 m asl, but some areas where snow patches are present during winter are apparently unfrozen (Figure 6c). In the steeper parts where snow cover does not develop, permafrost seems to be present. Thus the presence of snow cover tends to increase ground temperatures by protecting the ground from heat loss during winter [e.g., Goodrich, 1982]. This effect is presumably additionally amplified by the assumed relatively high snow temperatures in Iceland.
 The lack of permafrost at the Vopnafjörður borehole may be explained by the greater influence of snow cover. In general, however, the borehole measurements at the other locations clearly indicate the presence of warm, shallow permafrost at more or less bare blown sites at the studied elevations, as the SAT were comparable for all of the three eastern sites. Thus the limited permafrost thickness is related to the general high geothermal heat fluxes in Iceland, while the occurrence is controlled by SAT and snow cover.
5.3. Climate Change Impact on Permafrost Temperatures
 At Gagnheiði and Snæfell, where high accuracy temperature time series are present, borehole-temperature profiles show deviation from steady state conditions in form of convex temperature-depth curves. Similar observations have been made in other permafrost boreholes in the Northern Hemisphere [e.g., Harris et al., 2003; Lachenbruch and Marshall, 1986], and the geothermal profiles have been interpreted as signs of recent climatic warming. Also, an increase in snow cover will raise the GSTs owing to its insulating effect during winter [e.g., Goodrich, 1982]. However, significant snow covers changes at the sites are unlikely to occur because of the sites exposure to wind. Furthermore, since the borehole depths are modest compared to distances to steep slopes, topographic effects are considered being minimal [cf. Gruber et al., 2004]. Provided that the porosity and the water content of the substratum remain constant, changes in the thermal properties are exclusively associated to phase changes of the pore water. Changing material properties may affect the shape of the temperature profiles in the vicinity of 0°C. However, these effects are considered in our model formulation. Also, major changes in the thermal conductivity of the bedrock are improbable. Our preliminary modeling reproduces temperature profiles as a response to the most recent warming. Thus the borehole-temperature profiles at Gagnheiði and Snæfell are considered to relate to a recent warming of permafrost temperatures in response to warming of SAT. Continued permafrost monitoring will acquire the database necessary to evaluate this suggestion.
 The modeled temperature evolution since 1955 indicates that the present-day permafrost thicknesses prevail from the cooling period in the late 1960s/early 1970s. Furthermore, the modeling indicates that the inverted temperature gradients represent a response to warmer annual temperatures, which since the 1990s, have occurred more frequently. Concerning possible future climate warming the longer response time at Snæfell with somewhat warmer GSTs is due to the higher ice content of the subsurface which reduces the rate of thaw [cf. Kukkonen and Šafanda, 2001]. The fast decay of permafrost, on the order of decades as indicated by our modeling, is due to the modest vertical extent of the permafrost and the large influence of the high geothermal heat flux. In comparison, the anticipated geothermal heat flux at Snæfell (∼170 mW m−2) is roughly five times larger than the corresponding values at the permafrost boreholes at Tarfalaryggen and Juvvasshøe, Scandinavia [Isaksen et al., 2001]. At Hágöngur, the geothermal heat flux is probably even larger [cf. Flóvenz and Sæmundsson, 1993]. In summary, the results indicate that the mountain permafrost in Iceland is highly temporal dynamical, especially within the volcanic rift zone.
 From this study the following conclusions are drawn:
 1. Permafrost is present within the Hágöngur, Snæfell and Gagnheiði boreholes, the former two sites situated within active volcanic areas. This confirms the hypothesis of widespread occurrences of mountain permafrost at elevations above ∼900 m asl in northern and eastern Iceland, even in areas of high geothermal heat flux. At the Vopnafjörður borehole the warmer, nonpermafrost conditions experienced are probably due to greater influence of snow cover than the other sites.
 2. Permafrost temperatures are between −2° and 0°C at the measurement sites and permafrost thicknesses are ∼6 m at Hágöngur and in the range of a few tens of meters at Snæfell and Gagnheiði. The active layer thicknesses range from less than 2 m up to more than 6 m. The considerably thinner active layer thickness at Snæfell compared to Gagnheiði is mainly due to the presumably higher ice content of the uppermost layers at Snæfell.
 3. The resistivity tomography measurements indicate that permafrost is present down to elevations of 850 m asl in a northerly exposed slope close to the Snæfell borehole, but unfrozen ground is apparently present at higher altitudes due to the insulating effects of long-lasting snow coverage.
 4. Results of 1-D heat conductivity modeling suggest that the deviation from linear gradients toward higher temperature observed at shallow depths in the Snæfell and Gagnheiði boreholes is likely to result from the rise in SAT since the early 1990s. Continued permafrost monitoring will acquire the database necessary to evaluate this suggestion.
 5. The modeling further suggests a high sensitivity of Icelandic permafrost to changes in surface temperature. This behavior is associated with the modest thickness of the permafrost and the large influence of the geothermal heat flux in the region. Hence permafrost in Iceland may react relatively quickly to future climate warming and may completely disappear within a few decades at elevations comparable to the sites at Gagnheiði and Snæfell. The somewhat longer response time at Snæfell, even at slightly higher ground surface temperature conditions, is caused by the presumably larger ice content of the substratum.
 Meteorological data were kindly provided by the Icelandic Meteorological Office. Bjørn Wangensteen, Geir Moholdt, and Andreas Kellerer-Pirklbauer gave valuable field assistance. The study was financed by the Norwegian Research Council (project 157837/V30), Jarðfræðistofan Geological Services, Reykjavik, Iceland, and the Department of Geosciences, University of Oslo, Norway. Two anonymous reviewers made valuable comments on an earlier version of this manuscript. Kirsty Ann Langley helped in smoothing the text. We want to thank all mentioned individuals and institutions.