1. INTRODUCTION
Knowledge about the global response of the >200 000 glaciers on Earth to climate change has considerably improved in recent years (Vaughan and others, Reference Vaughan and Stocker2013; Pfeffer and others, Reference Pfeffer2014), especially through large-scale modelling (e.g. Gardner and others, Reference Gardner2013; Radić and others, Reference Radić2014; Huss and Hock, Reference Huss and Hock2015). On a local scale, however, large uncertainties remain for individual glaciers because processes and glacial dynamics are implemented in a simplified manner in the models. Indeed, each glacier shows specific behaviour to some extent, influenced by local topoclimatic factors, glacier hypsometry or surface characteristics (Carturan and others, Reference Carturan2013a; Fischer and others, Reference Fischer, Huss and Hoelzle2015). These particularities may generate non-linearities that complicate the response of glaciers to climate change (Paul and others, Reference Paul, Kääb and Haeberli2007; Scherler and others, Reference Scherler, Bookhagen and Strecker2011; Carrivick and others, Reference Carrivick2015). Understanding of this local complexity and its integration in global approaches remain a major task (Huss, Reference Huss2012; Huss and Fischer, Reference Huss and Fischer2016; Pellicciotti and others, Reference Pellicciotti2015).
In high-relief environments, three characteristics can strongly influence the individual response of glaciers to climatic variations: (1) small glacier size, (2) debris cover on the ice and (3) permafrost conditions. These characteristics affect an increasing number of mountain glaciers with the current rise in the ELA (Benn and others, Reference Benn, Kirkbride, Owen, Brazier and Evans2003; Vaughan and others, Reference Vaughan and Stocker2013). Indeed, shrinking glaciers are becoming smaller and thinner. Glacier melt is especially rapid in temperate environments and glaciers tend to be confined within the coldest areas of the Earth's surface (e.g. Huss, Reference Huss2012; Huss and Hock, Reference Huss and Hock2015). Moreover, because ice deformation is related to ice thickness, the dynamics of shrinking glaciers is decreasing (Paul and others, Reference Paul, Kääb and Haeberli2007) and their capacity to evacuate their sediment load is thus reducing. At the same time, deglaciation and permafrost degradation in the surrounding relief increase the debris supply. As a result, supraglacial debris covers are expanding on mountain glaciers (Carturan and others, Reference Carturan2013a; Kirkbride and Deline, Reference Kirkbride and Deline2013).
Very small glaciers (<0.5 km2) have a reduced motion (Haeberli, Reference Haeberli1985; Benn and others, Reference Benn, Kirkbride, Owen, Brazier and Evans2003; Paasche, Reference Paasche, Singh, Singh and Haritashya2012). Usually too thin to allow noticeable ice deformation, they are also weakly affected by basal processes. The production of meltwater and the shear stress acting on the glacier bed are both limited. In general, these glaciers react more rapidly to climatic variations (Jóhannesson and others, Reference Jóhannesson, Raymond and Waddington1989; Kuhn, Reference Kuhn1995; Colucci and Guglielmin, Reference Colucci and Guglielmin2014). Due to the small altitudinal range, variations in the ELA of these glaciers rapidly change the proportion of accumulation and ablation areas, and strongly impact on glacier mass balance. On the other hand, the influence of the surrounding topography on small cirque glaciers can significantly attenuate their climate sensitivity (DeBeer and Sharp, Reference DeBeer and Sharp2009; Carturan and others, Reference Carturan2013b; Scotti and others, Reference Scotti, Brardinoni and Crosta2014). Accumulation and ablation anomalies occur in relation to the snow redistribution by avalanches, shadow effects or the presence of a debris cover. They can decouple the climatic control on mass-balance variations. Debris covers exceeding several decimetres in thickness are especially efficient thermal insulator layers and significantly reduce the ablation rate (Benn and others, Reference Benn, Kirkbride, Owen, Brazier and Evans2003; Deline and others, Reference Deline, Gardent, Magnin and Ravanel2012). Finally, permafrost conditions can strongly influence glacier dynamics. The thermal regime of the glacier can be cold or polythermal because of the surface and basal energy balance (Etzelmüller and Hagen, Reference Etzelmüller, Hagen, Harris and Murton2005; Cuffey and Paterson, Reference Cuffey and Paterson2010). Otherwise, repeated freeze/thaw cycles increase the sediment production in the surrounding relief (Deline and others, Reference Deline, Shroder, Haeberli and Whiteman2015) and the amount of debris deposited on the glacier surface. Ablation rate is then significantly reduced in these environments when the thickness of the debris cover is close to that of the permafrost active layer (i.e. seasonally thawing layer; Lilleøren and others, Reference Lilleøren2013). At the glacier margins, the deposition of the large amount of debris may induce the build-up of moraine dams (Benn and others, Reference Benn, Kirkbride, Owen, Brazier and Evans2003; Iturrizaga, Reference Iturrizaga2013). Complex glacier/permafrost interactions can also occur around the glacier (Ackert, Reference Ackert1998; Berthling and others, Reference Berthling, Schomacker, Benediktsson, Shroder, Giardino and Harbor2013). Thereby, the presence of cold ice, a moraine dam and/or a rock glacier commonly restrain the glacier's downward flow under permafrost conditions (Kirkbride and Deline, Reference Kirkbride and Deline2013). Consequently, very small heavily debris-covered glaciers located in permafrost environments often react with thickness variations rather than length changes.
Studies focusing on this type of glacier are relatively few and are led by geomorphological rather than glaciological questions. Mainly focused on glacier/permafrost interactions, research has characterised the internal structure (e.g. Reynard and others, Reference Reynard, Phillips, Springman and Arenson2003; Ribolini and others, Reference Ribolini2010) and its relation to glacier dynamics or hydrology (e.g. Kneisel and Kääb, Reference Kneisel and Kääb2007; Pourrier and others, Reference Pourrier, Jourde, Kinnard, Gascoin and Monnier2014; Seppi and others, Reference Seppi2015; Bosson and Lambiel, Reference Bosson and Lambiel2016). There have been fewer studies on the decadal behaviour of these glaciers and, at this timescale, their past and future evolution still remains largely unknown. This paper combines remote sensing and geophysical surveys to determine the response of a very small mountain glacier in the Swiss Alps to climate variations over the past five decades. Changes in surface elevation and ice dynamics were assessed for successive observation periods by comparing high-precision DEMs with aerial images. Geodetic glacier mass balance was computed for each period. In addition, a GPR survey was carried out to determine the current ice thickness distribution and glacier volume. Finally, we compare the decadal evolution of the Tsarmine Glacier with other local glaciers to uncover potential particularities in the behaviour of very small and heavily debris-covered cirque glaciers located in permafrost environments.
2. STUDY SITE AND CLIMATIC CONTEXT
2.1. The Tsarmine Glacier system
The Tsarmine Glacier system is located in the Arolla valley in Switzerland (46°03′N, 7°31′E; 0.49 km2; 2600–3070 m a.s.l.; Fig. 1). Detailed geomorphological analysis and mapping of this complex assemblage of interacting landforms can be found in Lambiel and others (Reference Lambiel, Reynard, Cheseaux and Lugon2004), Micheletti and others (Reference Micheletti, Lambiel and Lane2015a) and Bosson and Lambiel (Reference Bosson and Lambiel2016). This very small cirque glacier (0.25 km2 in 2015) occupies the foot of the 400–600 m high rock walls of the Grande Dent de Veisivi and Blanche de Perroc. Until the 1980s, a north-facing ice apron covered the latter (Delaloye, Reference Delaloye and Rothenbühler2008). Nowadays, the rock walls are affected by strong rockfall activity. For instance, a major rockfall occurred near the top of the west face of the Grande Dent de Veisivi in September 2015, following an extraordinarily warm summer. The presence of water in the scarp coming from ice melt indicated the existence of permafrost conditions in the area.
Geographic setting and main components of the Tsarmine Glacier system: (a) location of the Arolla valley. (b) Location of the Tsarmine Glacier system and of the three larger valley glaciers referred to in this study (aerial image: Landsat 25 1990–94; Swisstopo). (c) Main components of the Tsarmine Glacier system (aerial image: orthophoto 2012).
Situated at the shade-dominated base of Blanche de Perroc between 3070 and 2920 m a.s.l., the uppermost debris-free part of the glacier receives little solar radiation (zone 1; Fig. 1c). Avalanching enhances snow accumulation, leading to a cone-like surface morphology. Below, down to ~2830 m a.s.l., a discontinuous and relatively thin (millimetres to decimetres) layer of debris covers the glacier today (zone 2a; Fig. 1c). This zone showed continuous variations in surface conditions (debris-free vs. debris-covered) in recent decades, but overall the debris cover progressively extended upward through time. Downslope, a thick (decimetres to few metres) and continuous debris mantle has been present since at least 1967 (zone 2b; Fig. 1c). The supraglacial layer is made up of coarse-grained angular deposits of rockfalls, indicating their passive transfer on the glacier. Ice outcrops become very rare and are related to thermokarstic erosion. Arcuate round ridges are visible on the surface (Fig. 1c). The debris layer, which thickens downslope, masks the lower limits of the glacier. It was detected with electrical resistivity tomography (ERT) near 2715 m a.s.l. (Bosson and Lambiel, Reference Bosson and Lambiel2016). These results also allowed distinguishing debris-covered dead ice (zone 3; Fig. 1c) from ice-free debris located around a proglacial lake (zone 5 and 6). Downslope, a large gullied moraine dam delimits the Tsarmine system. Accumulated during the Holocene glacial pulses (Ivy-Ochs and others, Reference Ivy-Ochs2009), it illustrates that this fluctuating glacier probably always contained large amounts of debris, produced by the erosion of the surrounding rock walls composed of fractured gneiss. The presence of an active rock glacier at the northern margin of the Tsarmine system points to the local occurrence of permafrost conditions (zone 4; Fig. 1c). The presence of push moraines indicates a deformation by the readvance of Tsarmine Glacier during the Little Ice Age (LIA; Lambiel and others, Reference Lambiel, Reynard, Cheseaux and Lugon2004). Like other similar systems (Bosson and others, Reference Bosson2015; Micheletti and others, Reference Micheletti, Lambiel and Lane2015a), Tsarmine is disconnected from downslope sediment transfer systems because of the presence of the moraine dam and the rock glacier, the relatively weak slope angle (17° of mean slope for the longitudinal profile between the top of the glacier and the top of the moraine dam), the reduced magnitude of sediment transfer processes and the absence of a proglacial meltwater stream.
Bosson and Lambiel (Reference Bosson and Lambiel2016) quantified the surface kinematics of the Tsarmine system between 2011 and 2014 with repeated dGPS measurements. The results highlighted the existence of different dynamic behaviours at annual and seasonal timescales between zones 2b, 3, 4 and 5. The mean horizontal surface displacement was 60 cm a−1 and the mean vertical change rate −17 cm a−1 in the debris-covered glacier zone (zone 2b). Surface dynamics were weaker elsewhere: ~20 cm a−1 of downslope creep in the marginal rock glacier (zone 4), ~15 cm a−1 of surface lowering in the proglacial dead ice body (zone 3) and discontinuous, very slow movements in ice-free debris (zone 5).
2.2. Climatic context
The Tsarmine Glacier system is located in a dry intra-Alpine valley. Topography acts as a barrier for wet air masses and reduces annual precipitation rates considerably. Combining meteorological data collected by Hydro-exploitation SA between 2001 and 2013 at 2000 m a.s.l. on the valley floor with monthly lapse rates estimated locally (between 0.54 and 0.63°C (100 m)−1 with an annual mean of 0.59°C (100 m)−1 at these elevations, Bouët, Reference Bouët1985), the mean annual air temperature at 2800 m a.s.l. is currently ~−0.8°C. Mean annual precipitation measured on the valley floor between 2001 and 2013 is ~900 mm.
Figure 2 shows the evolution of the regional climate (south-western Valais) between 1930 and 2015. Homogenised monthly data for Sion (482 m a.s.l.; 24 km distance from Tsarmine, Fig. 1a), and Col du Grand St Bernard (GSB) (2472 m a.s.l.; 34 km distance from Tsarmine, Fig. 1a) weather stations were used (MeteoSwiss). To highlight the variations of the most important climatic parameters on the glacier, the mean deviation of annual and summer (July–September in this high-mountain environment) air temperatures and annual and winter precipitation (October–May) relating to the 1961–1990 normal climate period were calculated for both stations. Between the 1950s and the mid-1980s, annual temperatures fluctuated around or slightly below the climatic norm. Subsequently, a clear warming trend set in: temperatures exceeded the norm by at least 1°C and continued to progressively increase until 2012. Summer temperatures behaved similarly, but the rapid increase had already started some years before, around 1980. Annual and winter precipitation showed larger fluctuations. Both were generally below the climatic norm before 1980 and after 2000 while the 20 years in between were wetter.
Evolution of annual (solid curve) and summer (dotted curve; means of monthly values from July to September) air temperature and annual (solid curve) and winter (dotted curve; sums of monthly values from October to May) precipitation in the south-western Valais between 1950 and 2014. Based on homogenised monthly series of MeteoSwiss, the mean deviations of data from Sion and Grand St Bernard (GSB) stations (location in Fig. 1a) from the 1961 to 1990 climatic normal period have been computed. Only the 5-year running means are represented.
3. METHODS
The methodology of this study was based upon: (1) the creation of DEMs using archival digital photogrammetry; (2) use of these DEMs to derive surface elevation changes and orthorectified imagery; (3) calculation of mass balance from the surface elevation changes; (4) analysis of surface displacement rates through image correlation applied to the orthorectified images; (5) application of GPR to quantify the spatial ice thickness distribution and volume.
3.1. Creation of DEMs using archival digital photogrammetry
DEMs were created by applying archival digital photogrammetry (Walstra and others, Reference Walstra, Chandler, Dixon, Dijkstra, Lacerda, Erlich, Fontoura and Sayao2004; Chandler and others, Reference Chandler, Lane, Walstra, Fryer, Mitchell and Chandler2007) to historical aerial photographs held by the Swiss Federal Office of Topography (Swisstopo) for the period 1967–2005. Additionally, we used aerial images acquired in 2012 by the commercial company Flotron AG (Table 1). DEM creation from aerial orthoimagery is described in detail by Micheletti and others (Reference Micheletti, Lane and Chandler2015b), and therefore only summarised here. Digital aerial photogrammetry requires a number of Ground Control Points (GCPs) for the photogrammetric restitution. A network of GCPs was established using a Leica System 500 dGPS during a field campaign in 2012. These points must be stable through the whole observation period, well distributed over the area of interest and easy to identify on the images. Given varying image quality and contrast, it was not always possible to retrace all GCPs locations with high accuracy and confidence. It was crucial to obtain as many GCPs as possible, to always have adequate observations well distributed over the (varying) extent covered by the photographs. Consequently, GCPs employed (and unused) varied between pairs of images. Survey data were post-processed and corrected using data from an automated GNSS Network for Switzerland (AGNES) to a vertical and horizontal precision better than ±0.05 m. Image processing for DEM creation was performed using ERDAS IMAGINE Leica Photogrammetry Suite. It involved three main steps: (1) definition of the internal orientation of the camera using camera calibration certificates available for each image pair; (2) after manually identifying the GCPs, estimating the exterior orientation using a conventional bundle adjustment; (3) application of automatic stereomatching algorithms to identify homologous points on image pairs and so to compute ground coordinates. The resulting point clouds were used to compute 1 m resolution DEMs for each period by applying a kriging interpolation in ArcGIS 10. DEMs were derived for eight dates between 1967 and 2012 (Table 1).
Characteristics of the aerial imagery, DEM accuracy and precision estimates using dGPS survey data (Micheletti and others, Reference Micheletti, Lane and Chandler2015b), mean snowline altitude and accumulation area ratio at each date
BW, black and white; RGB, red, green, blue (colour); NIR, near infrared.
The two values in brackets have to be considered carefully because fresh snow hampered the identification of the real snowline in 1995.
3.2. Calculation of surface elevation changes
To obtain surface elevation changes, DEMs of different years were differenced to obtain DEMs of Difference (DoDs). At this stage, areas of ineffective photogrammetric restitution became visible; these corresponded to zones where, because of low image texture, shadow effects, excessive steepness or occlusion, stereomatching algorithms failed and only a few low quality points were used to represent the topography. As a consequence, unrealistic elevation differences were found, especially in the upper part of the glacier. To avoid unreliable estimates, these areas were identified and removed from the analysis.
To distinguish real topographical changes from noise associated with random error in the derived surface elevation changes, a standard error propagation procedure was employed (Lane and others, Reference Lane, Westaway and Hicks2003). First, the quality of each DEM was assessed with dGPS data from unused GCPs by computing the difference in elevation with DEM values in the respective locations (Table 1). The precision of each DEM, estimated as the standard deviation of these errors, was used to quantify limits of detection for each DoD, following Lane and others (Reference Lane, Westaway and Hicks2003). The significance level was set to 90% in order to obtain a greater confidence that values in the DoD actually corresponded to real topographical changes. The limits of detection for each pair of DEMs are given in Table 2. Only differences larger than these values were employed to compute surface elevation and volumetric changes (ΔV) for individual observation periods.
Limits of detection of elevation changes for each pair of images (both absolute values and normalized per year) with a confidence level of 90% computed following the approach by Lane and others (Reference Lane, Westaway and Hicks2003)
See Micheletti and others (Reference Micheletti, Lane and Chandler2015b) for details.
3.3. Calculation of average mass balance
Area-averaged specific geodetic mass-balance rates
$\dot B$
(m w.e a−1) were computed for each observation period with:
$$\dot B = \displaystyle{{\Delta V\;. f\Delta v} \over {{\!A}\; \cdot \Delta t}},$$
where A is the glacier area (m2), and Δt the length of each observation period. Equation (1) includes a conversion factor fΔv, set to 0.85, which corresponds to a density of volume change of 850 ± 60 kg m−3. Such a value is commonly used (e.g. Fischer and others, Reference Fischer, Huss and Hoelzle2015) and is recommended for periods >5 years, for a study site composed by a firn area and for volume changes significantly different from zero (Huss, Reference Huss2013). ERT (Bosson and Lambiel, Reference Bosson and Lambiel2016) and GPR results allow us to quantify the current glacier area to 252 000 m2 (zones 1, 2a, 2b; Fig. 1c). This value was always used to calculate the glacier's average mass balance because the presence of extensive debris cover has hampered a clear determination of downslope glacier boundaries in recent decades. Weak variations of debris-covered glacier areas observed elsewhere at a decadal timescale support this approach (e.g. Deline, Reference Deline2005; Schmidt and Nüsser, Reference Schmidt and Nüsser2009; Benn and others, Reference Benn2012). The evolution of the specific geodetic mass balance rates between 1967 and 2012 were compared with the mean glaciological mass balances from Huss and others (Reference Huss, Dhulst and Bauder2015) for two nearby glaciers (Giétro and Allalin, 5.5 and 9.7 km2, at 12 and 30 km distance from Tsarmine; Fig. 1a) and to the decadal extrapolated mass balance means of glaciers in the European Alps (Huss, Reference Huss2012). In addition, we compared the surface elevation and mass changes of the Tsarmine Glacier since the 1980s with those of the neighbouring Tsijiore Nouve Glacier, Mont Collon Glacier (also known as Bas Glacier d'Arolla) and Haut Glacier d'Arolla (Fig. 1b) using the data provided by Fischer and others (Reference Fischer, Huss and Hoelzle2015).
3.4. Quantification of horizontal surface displacement rates and aerial imagery analysis
Archival aerial photogrammetry also allowed orthorectification of the aerial photographs, which were then used to quantify horizontal surface displacement rates within the observation periods. For this purpose we used the 7D software (Vacher and others, Reference Vacher, Dumoulin, Morestin and Mguil-Touchal1999). Based on image autocorrelation, this software identifies the same features on images of two different dates based upon correlation of image grey-scale values. The comparison of images generates vectors illustrating the value of the pixel displacement, which, by using image resolution, can be expressed in m a−1. The quality and the quantity of identified vectors depended on quality and characteristics of the source images themselves and was impaired considerably by the presence of shadow or fresh snow, and by poor image contrast. Thus, it was not always possible to quantify horizontal surface displacements for some areas (especially the upper half of the glacier) or during some periods. We only present results obtained within the Tsarmine Glacier system. Horizontal surface displacement rates detected in neighbouring stable areas were below the limit of detection (<0.1 m a−1). Otherwise, movements of some large boulders were manually tracked on the photographs. The horizontal velocity values obtained with image autocorrelation and feature tracking for the most recent period are close to those measured between 2011 and 2014 by dGPS (Bosson and Lambiel, Reference Bosson and Lambiel2016).
Mean snowline altitude and accumulation area ratio (AAR) were calculated for each period based on the orthorectified aerial photographs to investigate the evolution of the accumulation area of the Tsarmine Glacier over the past few decades (Table 1). These values provide interesting information on the recent evolution of the Tsarmine Glacier. Nevertheless, their interpretation requires care because the identification of annual accumulation of snow and ice is challenging on aerial images. Moreover, snow accumulation on very small glaciers can be very patchy and variable from year to year (Huss, Reference Huss2010). Aerial photographs were taken a bit too late (7 October) in 1995 and a fresh snow cover hinders the identification of the end of ablation season snowline.
3.5. Measuring glacier thickness and volume
GPR data for Tsarmine Glacier were acquired on 10 March 2015 with a Malå Geoscience ProEx System consisting of a control unit mounted in a backpack, a portable monitor and a 100 MHz rough terrain antenna with a constant offset of 2.2 m. Taking four measurements s−1, the antenna was pulled across the glacier at walking speed. For each measurement, 728 samples were recorded within a time window of 550 ns. The measurement (trace) positions were acquired with a conventional GPS receiver connected to the monitor. Horizontal accuracy in the position of the GPR-derived glacier thickness data is ±5 m. Individual radar profiles were collected along a total length of ~3500 m and covered the entire debris-free and debris-covered parts of the glacier (1, 2a, 2b; Fig. 1c), at high-spatial resolution.
A series of standard processing steps (e.g. Sold and others, Reference Sold2013) were applied to the GPR raw data using Reflexw 2D data analysis (Sandmeier Scientific Software). These included a spatial interpolation to a constant 1 m spacing of traces to account for variations in the walking speed, a frequency bandpass filtering to reduce noise, a background removal, a gain function to compensate for radar wave power loss with depth and a fk (Stolt) migration to account for omnidirectional reflections of the signal and correct geometrical irregularities. Only unambiguous reflection horizons of the glacier bed were picked and stored as x-, y-, thickness data. For time/depth conversion, a constant radar propagation velocity in ice of 0.167 m ns−1 was assumed (Hubbard and Glasser, Reference Hubbard and Glasser2005). Following Fischer (Reference Fischer2009), we expect the accuracy of measured glacier thicknesses to be in the order of 5–10% of the measured values. Uncertainty may be caused by the measurement system itself, but is primarily attributed to changes of the signal velocity below the glacier surface, e.g. within the debris-cover layer. The measured glacier thickness data were imported in ArcGIS to derive the spatial distribution of ice thickness and glacier volume. To avoid and/or reduce linear interpolation artefacts, the measured point density was reduced to 1% of the originally recorded data by only considering every hundredth measurement (dots of measured glacier thickness in Fig. 7). Zero glacier thickness was assigned to the glacier outline prior to the spatial interpolation of point thickness data. Because different algorithms have been shown to produce significantly varying results (Fischer and Kuhn, Reference Fischer and Kuhn2013), topo to raster, kriging and inverse distance weighting (IDW) techniques were applied to create 10 m × 10 m grids of the 2015 glacier thickness, and their results were subsequently compared with each other.
4. RESULTS
4.1. Movement and elevation changes in the proglacial zones
The marginal zones of the Tsarmine system (3, 4, 5, 6; Fig. 1c) are not analysed in detail here. Their comprehensive analyses at both decadal and annual/seasonal timescales were carried out by Micheletti and others (Reference Micheletti, Lambiel and Lane2015a) and Bosson and Lambiel (Reference Bosson and Lambiel2016). Elevation changes of the proglacial zones between 1967 and 2012 revealed a surface lowering of up to 20 m that can be attributed to three main processes: (1) ice melt in the debris-covered dead ice body and the rooting zone of the rock glacier (zones 3 and 4; Fig. 3h), (2) thermokarstic erosion of the ice by lake formation near the glacier front (zone 6; Fig. 3h) and (3) gullying of the outer side of the moraine dam (zone 5; Fig. 3h). Except for the development of a proglacial lake between 1983 and 1999 (Figs 3c–e), the magnitude of these processes was generally too weak to be systematically observable over the short term. Horizontal surface displacement rates (Fig. 4) showed a recent acceleration of the rock glacier, where velocities reached 70 cm a−1 in the rooting zone between 1999 and 2005 (zone 4). Discontinuous, slow movements were reconstructed for the proglacial area (zones 3 and 5; Fig. 4). They show that the delineation of glacier boundaries in debris-covered environments can be challenging. For instance, the dead ice zone was almost completely stagnant between the 1967–1988 and 1995–1999 periods and was flowing during the 1999–2005 period. Still, the distinction of this zone from the upslope glacier (zone 1, 2a, 2b) relies on their different dynamic behaviour on both decadal (Fig. 4) and seasonal timescales (Bosson and Lambiel, Reference Bosson and Lambiel2016). The few movements reconstructed at these timescales for the area surrounding the lake indicate that most of the remnant ice present in this zone melted during recent decades (Fig. 3).
Surface elevation, volume and average geodetic mass balance changes of Tsarmine Glacier for seven different periods between 1967 and 2012, derived from archival aerial photogrammetry. The longitudinal transect in the synthesis map (h) is referred to in Fig. 5, 10. The limit of detection (LoD) values are listed in Table 2.
Horizontal surface displacement rates for different parts of the Tsarmine Glacier system between 1967 and 2005, obtained with 7D software.
4.2. Elevation changes and geodetic mass balances of the Tsarmine Glacier
There were clear differences in the elevation changes between the upper part of the glacier, composed of the debris-free zone, the middle part covered by a thin debris mantle (zones 1, 2a; Fig. 3), and the downslope heavily debris-covered one (zone 2b). From 1967 to 1977, a lowering of ~0.18–0.4 m a−1 was found in the heavily debris-covered part and contrasted with elevation gain up-glacier (Fig. 3a). Between 1977 and 1983, changes were more homogeneous over the whole glacier and a significant positive elevation change up to 1 m a−1 affected its central part (Fig. 3a). Between 1983 and 1995 (Figs 3c, d), surface lowering was observed in the upper half, but a discontinuous elevation increase affected the heavily debris-covered zone and migrated progressively downslope. This is clearly noticeable between 1983 and 1988 for instance (Fig. 3c), when a significant volume loss of 820 000 m3 in the upper part contrasts with a positive volume change of 140 000 m3 in the lower part. Since 1995, elevation loss has been dominant, but the lowering rate decreased after 1999 in zone 1 (Figs 3e–g).
For the entire period (1967–2012) (Fig. 3h), the volume change was negative, with a loss of 2 × 106 m3 over 45 years. Whereas the volume losses mainly affected the upslope part (with up to 30 m of elevation loss in zone 2a) and secondarily the glacier front, the middle part, covered by a thick layer of debris, actually gained in volume (~260 000 m3). This corresponds to a maximum positive elevation change of 15 m. These clear heterogeneous spatial patterns are also observable on a central longitudinal flowline transect (Fig. 5a) and have changed the glacier surface topography: whereas in 1967 the surface was convex for the debris-free zone and slightly concave for the heavily debris-covered part, the surface of the glacier was generally convex in the debris-covered portion and slightly concave in the debris-free zone by 2012 (Fig. 5b).
(a) Surface elevation changes between 1967 and 1983 (blue), 1983 –1995 (green), 1995–2012 (red) and 1967–2012 (bold black) along a longitudinal transect of Tsarmine Glacier, visible in Figure 3h. The graphs were computed with the ‘interpolate line’ tool in ArcGIS 10. (b) Longitudinal elevation profiles in 1967 (blue) and 2012 (red). The dashed black curve shows the glacier bed derived from ground penetrating radar data. The snowline altitude along the longitudinal transect, detected from aerial imagery, is reported at each date. On the right-hand side, the horizontal displacement rates along the longitudinal transect are derived from the image autocorrelation (Fig. 4).
Between 1967 and 2012, two stages of resulting mass changes were recorded for the Tsarmine Glacier. It went from a positive mass-balance context, with rates of 0.20 m w.e. a−1 for the period 1967–1977 (Fig. 3a), and 0.30 m w.e. a−1 between 1977 and 1983 (Fig. 3b), to a negative mass-balance context with most negative values between 1995 and 1999 (−0.62 m w.e. a−1; Fig. 3e). However, mass balances for the 2000s were less negative again (−0.26 m w.e. a−1 between 1999 and 2005 and −0.30 m w.e. a−1 between 2005 and 2012, respectively; Figs 3f, g). Over the entire period 1967–2012, there was a negative average mass balance of −0.15 m w.e. a−1.
4.3. Horizontal surface displacement rates of the Tsarmine Glacier and aerial imagery analysis
The horizontal surface displacements shown in Figures 4, 5b illustrate significant spatial and temporal variations. The period 1967–1977 revealed horizontal surface displacements of ~1–2 m a−1, which decreased down-glacier to ~0.1–0.2 m a−1 (Fig. 4a). For the following decade, values between 0.1 and 0.5 m a−1 were obtained over the debris-covered glacier snout (Fig. 4b). In the late 1990s, horizontal surface displacements increased to up to 1 m a−1 for this area (Fig. 4c). Meanwhile, displacement rates exceeded 3 m a−1 in the central up-glacier zone and decreased towards the margins. Except for the glacier terminus, a velocity decrease affected all the debris-covered part in the early 2000s (Figs 4d, 5b).
Two main outcomes were obtained from tracking large boulders (Fig. 6): (1) horizontal surface displacements decreased down-glacier and (2) the periods with highest displacement rates varied spatially. Glacier surface motion in the central upslope zone reached its maximum between 1977 and 1983, with values >7 m a−1 (block No. 1). At the glacier snout (block No. 7) the displacements were much lower and highest velocities were reached between 1999 and 2005 (0.7 m a−1). Thus, from upslope to downslope, there was an apparent attenuation of the maximum surface displacement rates and a temporal delay of the maximum values over time. This shows the progressive translation down-glacier of a wave of maximum superficial displacement rates. Between 2005 and 2012, the motion of Tsarmine Glacier decreased, as shown by the displacement rates of all boulders.
(a) Horizontal surface displacement rates obtained by tracking some large boulders on the debris-covered glacier and (b) their successive position on the glacier.
Reconstructions of estimated mean snowline altitudes and AARs from aerial photographs showed that the relative proportion of the accumulation area of the Tsarmine Glacier generally decreased in recent decades (Table 1; Fig. 5b). More than the upper third of the glacier was snow covered in 1967, 1977 and 1983. Without considering the unrepresentative values of 1995, the accumulation area was above 2900 m a.s.l. and corresponded to <30% of the glacier surface in 1988, 1999, 2005 and 2012.
4.4. Current ice thickness and volume of Tsarmine Glacier
The 2015 GPR survey (Figs 5b, 7) revealed a maximum ice thickness of 36 m, recorded in the lower heavily debris-covered part. There, cross-section profiles showed a U-shaped glacier bed with the steeper thickness gradients towards the left LIA moraine (Fig. 7a). Upslope, in the middle part covered by a thinner debris layer, the glacier thickness decreased and was particularly shallower towards the lateral margins. The glacier bed here has a gentle V-shaped profile along a cross section, with thickest measured values of slightly >20 m for a very narrow central part (Fig. 7b). Up-glacier a trough-like bed was found, with glacier thicknesses >15 m in its centre.
Measured ice thickness and interpolated ice thickness contours obtained from March 2015 GPR data, as well as two examples of radargrams showing the glacier bed with black lines (aerial image: Swiss image level 2 2013; Swisstopo).
The mean interpolated glacier thickness was ~15 m (Table 3). Due to the comparatively high relative spatial density and the complete coverage of measured GPR profiles, the different interpolation algorithms applied to the measured point data returned very similar results, both for the spatial glacier thickness distribution and mean glacier thickness. The present (2015) bed topography and volume of Tsarmine Glacier, which amounted to 4.05 ± 0.4 × 106 m3, can be derived if the interpolated glacier thickness grids are subtracted from the latest high-resolution DEM of 2012 and the 2012–2015 surface elevation changes are considered in the calculation. The volume change exceeded 2 × 106 m3 between 1967 and 2012. The glacier thereby lost 1/3 of its initial volume in the past 45 years.
Mean (h mean) and maximum (h max) glacier thickness derived from GPR point measurements and different interpolation algorithms
5. DISCUSSION
5.1. Main controls on decadal elevation changes of Tsarmine Glacier
From the spatial distribution of the surface elevation changes (Fig. 3), it is possible to infer that the most dynamic part of the glacier is the upper half, including the debris-free zone (1) and the zone covered by a relatively thin and discontinuous debris layer (2a). Volume changes were marked here and related to snow and ice accumulation and ablation, ice fluxes and, to a lesser extent, debris deposition. They were likely to be strongly controlled by air temperature, especially during summer, as reported in other studies (Arendt and others, Reference Arendt, Walsh and Harrison2009; Diolaiuti and others, Reference Diolaiuti, Bocchiola, Vigiasindi, D'Agata and Smiraglia2012; Thibert and others, Reference Thibert, Eckert and Vincent2013). The periods of mass gain correspond to the coldest years (Figs 2, 3a, b). Significantly higher summer temperatures resulted in rapid surface lowering (Figs 3c–g). Thus, the temperature-driven summer mass balance is a primary control on the annual mass balance in debris-free zones as observed for other glaciers (e.g. Huss and others, Reference Huss and Hock2015). Because of the consistency of air temperature on the regional scale (Begert and others, Reference Begert, Schlegel and Kirchhofer2005), the decadal evolution of surface elevation changes in the upper part of Tsarmine is very representative for the behaviour of Alpine glaciers during this period (Deline and others, Reference Deline, Gardent, Magnin and Ravanel2012; Huss, Reference Huss2012; Vaughan and others, Reference Vaughan and Stocker2013; Huss, and others, Reference Huss and Hock2015). However, winter snow can modify and enhance these effects. The greatest mass gain (Fig. 3b) occurred during a period when winter precipitation generally exceeded the long-term average (Fig. 2). Since the 2000s, lower winter precipitation may have reduced snow cover at the start of summer, leading to an earlier transition to ice exposure, lower albedo and enhanced melt (e.g. DeBeer and Sharp, Reference DeBeer and Sharp2009). The most rapid and spatially homogeneous volume loss occurred between 1983 and 1999 (>650 000 m3; Figs 3c, d). Three main reasons, visible on aerial photographs, can explain the deceleration of lowering rates since 1999 while air temperature continued to increase: (1) the snow and ice accumulated from the 1960s to the mid-1980s has largely and rapidly melted in subsequent warmer years, as observed elsewhere in the Alps (Paul and others, Reference Paul, Kääb and Haeberli2007); (2) the increasing emergence and deposition of debris in this zone has reduced the ablation rate; (3), the concomitant confinement of the accumulation area at the shaded and avalanche-fed backwalls' foot has limited the sensitivity to climate (Table 1; Fig. 5b).
The influence of climatic variations on the elevation changes of the debris-covered glacier tongue (zone 2b) is not so obvious. Elevation changes were slightly negative on the tongue between 1967 and 1977, while they were positive upslope (Fig. 3a). Conversely, they were positive or close to zero between 1983 and 1999 in zone 2b, when elevation in the upper zone decreased rapidly (Figs 3c–e). From 1967 to 2012, the centre of the down-glacier parts presented positive surface elevation changes, whereas the frontal zone was lowered by up to 10 m (Fig. 3h). Two processes may interact to explain this effect. First, the increased thickness of the debris layer down-glacier will slow down melt rates because ablation rate and the superficial debris thickness are negatively correlated (e.g. Benn and others, Reference Benn2012). It induces an inversion of the ablation pattern down-glacier, as revealed by the surface elevation changes along the longitudinal profile since 1967 (Fig. 5a). The maximum melt rate then occurs at the transition between debris-covered and debris-free, just below the ELA (Kellerer-Pirklbauer and others, Reference Kellerer-Pirklbauer, Lieb, Avian and Gsurning2008; Shea and others, Reference Shea, Immerzeel, Wagnon, Vincent and Bajracharya2015), where ice loss generally exceeded 20 m between 1967 and 2012 (zone 2a; Fig. 3h). Second, there was probably a progressive down-glacier migration of the mass accumulated upslope between the 1960s and the mid-1980s (Figs 3a–e; e.g. Carrivick and others, Reference Carrivick2015). The downward propagation of a wave of maximum but decreasing surface displacement over time supports this assumption (Fig. 6a). Similar observations were reported by Thomson and others (Reference Thomson, Kirkbride, Brock, Nakawo, Raymond and Fountain2000) for a larger debris-covered glacier (Miage Glacier; area of 11 km2) over the 20th century. They illustrate the delayed and attenuated effects of climatic variations on mass changes of heavily debris-covered tongues: upslope surface elevation changes are dynamically transferred but attenuated towards the glacier front. Because of their reduced ablation rates, heavily debris-covered glacier tongues are less affected than debris-free glaciers by negative mass-balance perturbations, such as downwasting or frontal retreat (Thomson and others, Reference Thomson, Kirkbride, Brock, Nakawo, Raymond and Fountain2000; Scherler and others, Reference Scherler, Bookhagen and Strecker2011; Haeberli and others, Reference Haeberli, Huggel, Paul, Zemp, Shroder, James, Harend and Clague2013). Conversely and as observed for Tsarmine, significant positive mass-balance perturbations can lead to an increase in surface elevation of the debris-covered glacier snouts.
5.2. Surface displacement rates and associated dynamics
The analysis of surface displacement rates (Figs 4, 6) highlights four main characteristics of the decadal dynamics of Tsarmine Glacier. (1) Due to the propagation of the snow and ice accumulated from the 1960s to the mid-1980s, a kinematic wave of maximum displacement rates with progressively decreasing intensity migrated down-glacier (Fig. 6; Thomson and others, Reference Thomson, Kirkbride, Brock, Nakawo, Raymond and Fountain2000; Cuffey and Paterson, Reference Cuffey and Paterson2010). (2) Surface displacement rates decrease significantly towards the glacier terminus (Figs 5b, 6), which can be explained by the reduced driving stress in this area (Cuffey and Paterson, Reference Cuffey and Paterson2010). This effect may be enhanced by the downstream accumulation of sediments. Increasing the compressive flow, it often contributes to restraining the glacier motion (Kirkbride and Deline, Reference Kirkbride and Deline2013). Hence the uplift of the central part of the glacier is probably both due to an advective mass transfer and associated compression (Gudmundsson and others, Reference Gudmundsson2000; Kellerer-Pirklbauer and others, Reference Kellerer-Pirklbauer, Lieb, Avian and Gsurning2008). The formation of arcuate superficial ridges around the uplift zone in recent decades is very likely related to this compressive stress increase (Fig. 3h). (3) The glacier motion generally tends to decelerate over time, as observed elsewhere in the Alps since the 1980s in the context of negative mass balances (Paul and others, Reference Paul, Kääb and Haeberli2007). There are multiple causes of this reduction in dynamics. Several metres of ice have been lost in the upper half and frontal glacier zone of Tsarmine since 1967 and, except for the central zone of the heavily debris-covered part, it is now mostly thinner than 25 m (Fig. 7). Moreover, the glacier surface is flattening, as revealed by its current linear or even concave profile (Fig. 5b). Both processes lead to a decrease in driving stress (Benn and others, Reference Benn2012; Carrivick and others, Reference Carrivick2015). Finally, snow and ice accumulation on Tsarmine Glacier have been limited and discontinuous since 1983, reducing the mass turnover and thus glacier dynamics (Paul and others, Reference Paul, Kääb and Haeberli2007). The overall decrease of the glacier motion contrasts with the acceleration of the rock glacier and dead ice zones between 1999 and 2005 (Fig. 4). Accelerations of similar landforms were also observed in the Alps (e.g. Delaloye and other, Reference Delaloye, Kane and Hinkel2008). They were especially related to the increase of the permafrost temperature, the melt of ground ice and the increased presence of ground water induced by the 2003 heatwave. (4) The spatially rather homogeneous motion between 1995 and 2005 (Figs 4c, d) points to the domination of an en masse flow. Even if ice and may be bed deformation can also be active, it indicates the occurrence of basal sliding (Mayer and others, Reference Mayer, Lambrecht, Belò, Smiraglia and Diolaiuti2006; Copland and others, Reference Copland2009). Hence, even though this glacier may be polythermal because of the surrounding permafrost conditions, a large portion of the basal ice likely has a temperate thermal regime, as indicated by its basal slip (Bingham and others, Reference Bingham, Hubbard, Nienow and Sharp2008). Similar results and interpretations were proposed from dGPS monitoring data (Bosson and Lambiel, Reference Bosson and Lambiel2016).
5.3. A particular behaviour?
Before 2000, the average geodetic mass balance of Tsarmine Glacier was consistent with the average glaciological mass balance of Giétro and Allalin glaciers (Huss and others, Reference Huss and Hock2015) and with the extrapolated decadal mean of glaciers in the European Alps (Huss, Reference Huss2012; Fig. 8). Except in comparison with the negative 1967–1977 anomaly for Giétro Glacier, the mass balance of Tsarmine Glacier shows both magnitude and temporal variations comparable with the other larger glaciers. It illustrates the homogenous control of regional climate on glaciers during this period. The computed mass balance of Tsarmine between 1988 and 1999 has to be considered with care. Indeed, fresh autumnal snow is observable in the upper zone on the aerial image of 1995 and integrated in the calculation. Thus, the mass balances between 1988–1995 (−0.32 m w.e. a−1) and 1995–1999 (−0.63 m w.e. a−1) are likely under- and overestimated and the real values are probably closer to those obtained for Giétro and Allalin glaciers. The period 2003–2012 appears to be the one with strongest mass losses for the Swiss glaciers for at least the last century (Huss, and others, Reference Huss and Hock2015). However, this remarkably negative trend does not affect the Tsarmine mass balance, which stabilised at ~−0.3 m w.e. a−1 after 1999 (Fig. 8). This particular behaviour is not observable on other (very) small debris-free Alpine glaciers. For instance, the mass balances of Pizol (Huss, Reference Huss2010) and Sarennes (Thibert and others, Reference Thibert, Eckert and Vincent2013) glaciers were three times more negative (<−1 m w.e. a−1) in recent years compared with Tsarmine. On a larger timescale, while Tsarmine lost 1/3 of its initial volume between 1967 and 2012; this value exceeded 68% at Pizol between 1973 and 2006 (Huss, Reference Huss2010). The progressive shrinking of the accumulation area in a north-facing shaded and avalanche-fed niche and the concomitant extension of the debris cover explain this peculiar variation in the Tsarmine Glacier mass balance since 1999. It highlights the recent decrease of the control of regional climate on this glacier once most of the snow/ice accumulated during the 1960s to mid-1980s had melted. This consequence of the increasing influence of topography and debris cover supports results found in other studies that analysed the behaviour of some small cirque glaciers in the final stage of glaciers vanishing (e.g. López-Moreno and others, Reference López-Moreno, Nogués-Bravo, Chueca.Cía and Julián-Andrés2006; DeBeer and Sharp, Reference DeBeer and Sharp2009; Carturan and others, Reference Carturan2013b; Carrivick and others, Reference Carrivick2015). It contrasts with the dynamics recently undergone by climatically-driven (very) small bare-ice mountain glaciers (e.g. Huss, Reference Huss2010; Thibert and others, Reference Thibert, Eckert and Vincent2013; López-Moreno and others, Reference López-Moreno2015).
Average geodetic mass balance of Tsarmine Glacier, average glaciological mass balance of Giétro and Allalin glaciers (location on Fig. 1a, Huss and others, Reference Huss, Dhulst and Bauder2015) and extrapolated decadal mean of the glaciers of the European Alps (Huss, Reference Huss2012).
Comparing glacier surface elevation changes of Tsarmine since the 1980s with those of the nearby larger valley glaciers also reveals a particular spatial pattern for Tsarmine (Fig. 9). While the three large glaciers showed strongest lowering towards their terminus, Tsarmine Glacier lost most within its upper third. This highlights the importance of air temperature and thus altitudinal gradients for large glaciers. As very small cirque glaciers have a limited altitudinal gradient, other processes can superimpose on the effects of altitudinal temperature gradients. The decadal elevation changes of Tsarmine Glacier rather illustrate the influence of the delayed and attenuated mass transfer from the upper debris-free part to the debris-covered part and the debris cover thickening down-glacier. Interestingly, the snout of Tsijiore Nouve Glacier was also covered by a several decimetre thick debris layer in recent decades (e.g. Small, Reference Small1983) and is significantly less exposed to solar radiation than the Tsarmine snout (2.5 times smaller mean solar radiation; Fig. 9). However, up to 50 m of surface lowering occurred at the Tsijiore Nouve tongue since 1988. There, the presence of a thick debris cover and a weak exposure to solar radiation are thereby not sufficient to prevent rapid ice melt. The Tsarmine Glacier snout is located 300 m above the one of Tsijiore Nouve and within the discontinuous permafrost belt. The neighbouring rock glaciers show that ground ice can be preserved from noticeable melt under a several metre thick superficial debris layer (Lambiel and others, Reference Lambiel, Reynard, Cheseaux and Lugon2004; Bosson and Lambiel, Reference Bosson and Lambiel2016). Therefore, (1) the sheltering of a large portion of the glacier under a heavy debris mantle and (2) its confinement in a zone where air and ground temperatures are lower than for the other glaciers explain why the average mass balance since the 1980s is up to two times less negative for Tsarmine Glacier (−0.5 m w.e. a−1) compared with the three larger systems (−0.58 to −1.15 m w.e. a−1; Fig. 9). On a larger scale, it is also less negative than the average of glaciers in the European Alps from 1980 to 2000 (−0.69 m w.e. a−1; computed from decadal means of Huss, Reference Huss2012; Fig. 8).
Spatial distribution of elevation changes within the digital glacier outlines from the 1973 (dark blue) and the 2010 (light blue) Swiss Glacier Inventories (Müller and others, Reference Müller, Caflish and Müller1976; Fischer and others, Reference Fischer, Huss, Barboux and Hoelzle2014), derived by Fischer and others (Reference Fischer, Huss and Hoelzle2015) through a comparison of the DHM25 Level 1 DEMs from the 1980s with the 2010 swissALTI3D DEMs for (a) Tsijiore Nouve Glacier, (b) Tsarmine Glacier (location on Fig. 1a), (c) Mont Collon Glacier and (d) Haut Glacier d'Arolla. Mean area solar radiation was computed with ArcGis Spatial Analyst tool on the 2010 swissALTI3D DEMs and according to the 2010 glacier extents (Fischer and others, Reference Fischer, Huss, Barboux and Hoelzle2014).
Projections of climate change for the next few decades suggest continued increase of air temperature (Collins and others, Reference Collins and Stocker2013). The observed varying patterns of elevation changes since 1983 are expected to sustain and govern the future evolution of Tsarmine Glacier. Negative mass balance stabilised between 1999 and 2012 and the mean lowering rates of 0.5, 0.7 and 0.2 m a−1 were computed for the zones 1, 2a and 2b. In Figure 10, we extrapolate these rates, taking into account the known ice thickness. In the upper half of the glacier, the lowering exceeded 10 m in the past 30 years, roughly corresponding to the remaining glacier thickness in 2015 (Figs 5b, 7). This rapid thinning particularly affects the transition zone between debris-free and heavily debris-covered zones and could cause the disintegration of the Tsarmine Glacier into two entities, as evoked for other heavily debris-covered systems (Benn and others, Reference Benn2012; Shea and others, Reference Shea, Immerzeel, Wagnon, Vincent and Bajracharya2015). In contrast, the uppermost debris-free part appears a bit less sensitive to the air temperature increase because of the influence of topography on snow accumulation and solar radiation. The control of air temperature tends to decrease in this shaded avalanche-fed area, as observed since 1999 (Fig. 3). Conversely, the control of winter precipitation can increase here in the next few decades (Carturan and others, Reference Carturan2013b; Scotti and others, Reference Scotti, Brardinoni and Crosta2014). Sheltered in its proximal niche, this thinning glacier zone may progressively become an inactive and debris-covered ice patch (Serrano and others, Reference Serrano, González-Trueba, Sanjosé and Del Rìo2011), as the input of debris on the glacier surface is likely to increase in the next few decades (Deline and others, Reference Deline, Shroder, Haeberli and Whiteman2015). Finally, this zone could evolve toward a talus slope (Gomez and others, Reference Gomez2003). In the lower half, the glacier thickness exceeds 30 m (Fig. 7). Here, the ice melt could take more than 100 years if the mean lowering rates of 0.2 m a−1 observed over the past decade remain constant in the future. Several studies mention the recent transformation of similar heavily debris-covered glacier tongues into rock glaciers under permafrost conditions (e.g. Shroder and others, Reference Shroder, Bishop, Copland and Sloan2000; Monnier and Kinnard, Reference Monnier and Kinnard2015; Seppi and others, Reference Seppi2015). Bosson and Lambiel (Reference Bosson and Lambiel2016) discuss this possible evolution, which seems unlikely, given the probable domination of temperate thermal conditions in the glacier (Kneisel, Reference Kneisel2003). However, the integration of glacier ice in the rock glacier during or after the LIA is probable on the northern margin of the Tsarmine Glacier system (Bosson and Lambiel, Reference Bosson and Lambiel2016).
Evolution model of Tsarmine Glacier for the next few decades. The potential future surface elevation changes have been simply generated using the mean zonal lowering rates between 1999 and 2012.
6. CONCLUSIONS
Archival aerial photogrammetry, autocorrelation of orthophotos and GPR have been used to study the decadal behaviour of Tsarmine Glacier, a very small heavily debris-covered glacier located in an Alpine permafrost environment.
Spatial and temporal heterogeneity has characterised the glacier dynamics in recent decades. The upper half of the glacier presents the largest elevation changes, which are mainly related to (summer) air temperature variations. The accumulation of snow/ice in the period 1967–1983 and its rapid, almost complete, melt between 1983 and 1999 have induced related significant volume gains and losses. Volume loss has particularly affected the upper central part, where the surface lowering exceeded 30 m between 1967 and 2012. The consequences of climatic variations have been attenuated on the lower half of the glacier because of its covering under a thick debris layer. While the frontal zone showed several metres of lowering between 1967 and 2012, the lower central zone surface rose by up to 10 m. This was related to the downward transfer of the ice accumulated before 1983 and associated compressive stress. The surface displacement rates have slowed down in recent decades in relation with the glacier thinning. In 2015, the glacier volume was 4.05 ± 0.4 × 106 m3, corresponding to 2/3 of the volume in 1967. The mean interpolated 2015 glacier thickness was 15 m. In the next few decades, the differential ablation observed at the glacier scale could lead to its separation in two entities: an uppermost shrinking glacier and a lowermost heavily debris-covered dead ice body.
The decadal behaviour of Tsarmine Glacier exhibits particularities in time and space compared with the other local larger glaciers. Until the late 1990s, average geodetic mass balances ranged between 0.30 and −0.63 m w.e. a−1. The temporal variation and magnitude of these values were consistent with those measured for other local glaciers, illustrating the dominant control of regional climate. However, Tsarmine has had less negative mass balances than neighbouring glaciers (stabilised above −0.30 m w.e. a−1) since 1999. The spatial pattern of elevation changes also contrasts with neighbouring glaciers. Whereas Tsarmine was mainly affected by lowering in its upper half, the larger valley glaciers lost most of their mass at their tongues. Both temporal and spatial particularities highlight the limitation of the regional climate control on Tsarmine Glacier. The recent occurrence of this negative feedback is mostly related to the retreat of the debris-free area into a shaded and avalanche-fed north-facing niche, whereas the thick debris cover and cold air/ground conditions limit the ablation rate in this permafrost environment.
These results show that the evolution model of larger valley glaciers cannot be transferred 1:1 to very small debris-covered glaciers. Glacier system characteristics may complicate glacier responses to the climate forcing in time and space, and local additional controls generate heterogeneities and non-linearities in dynamics. In that way, studies focusing on complex glacial behaviours have to be thorough and backed up by field surveys. This study also highlights the potential of archival aerial photogrammetry to improve the knowledge on complex glacial dynamics on a decadal timescale.
ACKNOWLEDGEMENTS
The first two authors contributed equally to this work. We acknowledge those people who helped during field measurements (especially D. Bonneaux). Hydro-exploitation SA, Alpiq SA and Grande Dixence SA kindly provided the Arolla climate data. This manuscript was improved thanks to the valuable advice and comments of J-M Fallot and N. Deluigi (Uni. Lausanne) and M. Huss (Uni. Fribourg & ETH Zürich). B. Scott is acknowledged for proofreading the English text.
Open access
