Navigation – Plan du site

AccueilNumérosvol. 21 – n° 4A first estimate of permafrost di...

A first estimate of permafrost distribution from BTS measurements in the Romanian Carpathians (Retezat Mountains)

Première évaluation de la distribution du pergélisol dans les Carpates roumaines (massif de Retezat) par la méthode BTS
Adrian C. Ardelean, Alexandru L. Onaca, Petru Urdea, Raul D. Șerban et Flavius Sîrbu
p. 297-312

Résumés

Cet article fournit une première évaluation de la distribution du pergélisol dans la partie centrale du massif de Retezat (Carpates du Sud, Roumanie), utilisant des valeurs BTS (température à la base de la neige) comme indicateur de pergélisol dans un modèle statistique. Un total de 170 points BTS a été mesuré pendant la saison hivernale 2013-2014. La distribution du pergélisol est modélisée à l’aide d’une régression linéaire multiple entre les valeurs BTS et cinq variables indépendantes : l’altitude, la radiation solaire, la végétation, la pente et la courbure du profil. L’analyse statistique montre que la pente et la courbure du profil ne sont pas significatives. Par suite, les trois autres variables ont été incluses dans le modèle final, indiquant une précision de 0,48. Ce modèle montre que le pergélisol couvre une surface de 31 km2 (52% de la zone étudiée) dont 14 km2 représentent du pergélisol probable (PRP) et 17 km2 du pergélisol possible (PP). Les résultats mettent l’accent sur l’importance de la radiation solaire, l’altitude et la végétation dans l’occurrence du pergélisol, dans la partie centrale du massif de Retezat.

Haut de page

Notes de la rédaction

Article soumis le 28 octobre 2014, reçu sous sa forme révisée le 10 mars 2015, définitivement accepté le 28 octobre 2015.

Texte intégral

The authors would like to thank two anonymous reviewers for their valuable feedback on an earlier version of this manuscript. This work has been supported from the strategic grant POSDRU/159/1.5/S/133391, Project “Doctoral and Post-doctoral programs of excellence for highly qualified human resources training for research in the field of Life sciences, Environment and Earth Science” cofinanced by the European Social Fund within the Sectorial Operational Program Human Resources Development 2007-2013”.

1. Introduction

1The degradation of mountain permafrost due to rising global temperature leads to significant changes in the local ecosystems and subsequently may accelerate the instability of mountain slopes (Haeberli and Gruber, 2009). Therefore, a better understanding of the permafrost characteristics and distribution within complex steep mountain terrain is important for the assessment of future geomorphological and ecological changes in alpine environments (Schrott et al., 2012).

2Recent studies focusing on investigating permafrost occurrence within the Southern Carpathians have revealed that patches of permafrost may occur in the Southern Carpathians within block slopes composed of large blocky materials above 2000 m (Vespremeanu-Stroe et al., 2012; Onaca, 2013; Onaca et al., 2013a; Onaca et al., 2013b; Popescu et al., 2015). The occurrence of isolated patches of permafrost in the Southern Carpathians is strongly controlled by the presence or absence of coarse blocks, incoming solar radiation, elevation and snow cover characteristics (north-facing rock glaciers, block fields, and debris cones composed of unconsolidated materials situated above 1950 m; Otto et al., 2012). Despite the recently increased interest in permafrost occurrence within the Southern Carpathians, the problem of modelling its spatial distribution has raised limited scientific interest (Török-Oance, 2004; Szepesi, 2007). In this context, the current study offers new information regarding permafrost occurrence and distribution in the eastern mountains of Europe. Despite some recent findings, the mountain permafrost characteristics and spatial distribution in the Retezat Mountains are poorly known. Early permafrost observations in the Retezat Mountains were made by Urdea (1993) who measured bottom temperature of snow cover (BTS) values of -3°C or lower in several rock glaciers, and low (below +2°C) summer spring temperatures. Based on these data, he concluded that sporadic permafrost could occur in the Retezat Mountains. These first results, dealing with the existence of permafrost in the central part of the Retezat Mountains, were confirmed by recent studies involving indirect measurements (ground temperature monitoring, DC electrical resistivity soundings and ground penetrating radar (GPR) investigations; Vespremeanu-Stroe et al., 2012; Onaca, 2013; Onaca et al., 2013b).

3Since the 1990s, several models were proposed to estimate the spatial distribution of permafrost in different Alpine mountain ranges (Boeckli et al., 2012), either empirical-statistical or physically based models (Etzelmüller et al., 2001; Riseborough et al., 2008). BTS values were used for model calibration in several papers (Hoelzle, 1992; Keller, 1992; Gruber and Hoelzle, 2001; Lewkowicz and Ednie, 2004; Julián and Chueca, 2007), while the empirical-statistical approach highlighted the role of different topo-climatic parameters (air temperature, solar radiation, slope, etc.) on the probability of permafrost distribution (Etzelmüller et al., 2006).

4This paper aims to realize the first permafrost distribution model for the central part of the Retezat Mountains using independent BTS measurements as input data for a GIS-based empirical-statistical analysis. Because the ecological conditions show extreme spatial variability in mountain areas (Haeberli and Gruber, 2009), this paper aims to derive an empirical-statistical permafrost distribution model based on the relationships between topographical and climatic factors, vegetation cover and permafrost occurrence (Hoelzle et al., 2001). A GIS-based modelling approach derived from BTS measurements and digital elevation model parameters has been carried out, while previous geophysical measurements (DC electrical resistivity soundings and ground penetrating radar investigations) and temperature data were used to validate the model (Onaca, 2013; Onaca et al., 2013b).

2. Study Area

5The Retezat Mountains are located in the Western part of the Southern Carpathians, having two peaks above 2500 m (Păpușa 2504 m and Peleaga 2509 m). The study area, covering 59 km2 (fig.1), is situated in the central part of the Retezat Mountains where a high density of landforms indicative of creeping permafrost, such as rock glaciers (Haeberli, 1985), is observed.

Fig. 1 – Location of the study area.
Fig. 1 – Distribution des glaciers rocheux et des mesures BTS dans la zone étudié.

Fig. 1 – Location of the study area. Fig. 1 – Distribution des glaciers rocheux et des mesures BTS dans la zone étudié.

1: BTS measurements; 2: rock glaciers.
1 : mesures BTS ; 2 : glaciers rocheux.

6Within the study area, elevations range from 2509 m to 1505 m while the tree line lies at 1650-1750 m. The main structure of these mountains is the granitoid Retezat Unit, striking ENE-WSW and entirely covering the central part of the Retezat Mountains (Urdea, 2000). In the Pleistocene, the central part of the Retezat Mountains was glaciated by valley glaciers reaching 1050 m in the extensive Lolaia advance (Reuther et al., 2004). The present-day landscape of the alpine belt of the Retezat Mountains is dominated by periglacial (talus slopes, rock glaciers, block fields, scree slopes, block streams, debris cones, etc.) and glacial landforms (glacial cirques and valleys, moraines, ice-smoothed rocks, etc.). During the Holocene cold periods, the rates of debris production were higher than today, leading to an enhanced formation of unconsolidated slope deposits such as talus cones, rock glaciers, and block streams. These rock glaciers are considered to be the most important morphological expressions of high mountain permafrost (Barsch, 1996). According to Urdea (1998), 64 rock glaciers (covering 4 km2) are situated within the study area, of which 20 are of the intact type according to their morphology and previous geophysical and thermal data, while 44 are considered relict.

7The study area shares the general climatic conditions of the Southern Carpathians, with the 0°C isotherm situated at approximately 2050 m and regional recorded mean annual air temperature (MAAT) values of -2.4°C at 2505 m (Omu meteorological station) and -0.6°C at 2180 m (Țarcu meteorological station; Onaca et al., 2013b). The annual average precipitation varies between 1100-1400 mm above 2000 m, and the monthly maximum precipitation is 160 mm in July at the Țarcu meteorological station. The study area is situated within Retezat National Park.

3. Methods

3.1. BTS data collection

8The BTS method was first described by Haeberli (1973) to detect mountain permafrost in the Swiss Alps, but since then, it became widely used in the Alps (Hoelzle, 1992; King et al., 1992; Hoelzle et al., 1993; Hoelzle, 1996; Gardaz, 1997; King and Kalisch, 1998; Imhof et al., 2000; Gruber and Hoelzle, 2001; Bodin, 2005) and in other mountain ranges of the world (Dobinski, 1998; Ødegård et al., 1999; Kneisel et al., 2000). It is considered one of the most efficient methods for mapping permafrost distribution (Vonder Mühll et al., 2002) in non-arid mountains (Harris and Pedersen, 1998) and is the most used method for subsequent permafrost modelling(Lewkowicz and Ednie, 2004). Despite strong variations in BTS values induced by winter conditions (depth and duration of snow cover and air temperature) and summer surface energy exchange (influenced especially by the incoming solar radiation), there were several studies showing successful correlations between BTS values and the permafrost predictor variables (elevation, incoming solar radiation, vegetation cover, etc.) (Hoelzle, 1992; Lewkowicz and Ednie, 2004; Etzelmüller et al., 2006; Bonnaventure et al., 2012).

9In Romania, the BTS method was used to map permafrost occurrence in the Southern (Retezat and Parâng Mountains) and Western Carpathians(Apuseni Mountains; Urdea, 1993; Urdea, 2000; Urdea and Vuia, 2000; Vespremeanu-Stroe et al., 2012; Onaca et al., 2013b; Popescu et al., 2015). The scope of BTS measurements is to assess the winter equilibrium temperature (WEqT), typically in March or even in the beginning of April, relying on the assumption that the winter snow depth, as long as it remains thicker than 80 cm, is an excellent insulator that protects the ground surface from external air temperature fluctuations (Haeberli, 1973). However, as it is strictly dependent on the history of the snow pack at the measurement location, WEqT are not reached every year or at every location (Bodin, 2005). WEqT requires that a continuous snow cover is present since the beginning of the winter to provide sufficient insulation during the cold days of November-December, which could lead to a drastic cooling of the soil, resulting in a deep seasonal frost and implicitly low negative values of WEqT even in nonpermafrost areas. Thus, if optimum snow conditions are met, the temperature at the snow-ground interface will remain relatively stable at the end of the winter or beginning of spring (Ishikawa, 2003) when, according to the ”rules of thumb”, values lower than -3°C indicate that permafrost occurrence is probable, values of -2 to -3°C suggest the possible presence of permafrost while values higher than -2°C indicate the absence of permafrost (Haeberli, 1973; Hoelzle, 1992). Nevertheless, Hoelzle et al. (1993) noted that probable does not imply 100% permafrost occurrence, nor that improbable indicates 0% likelihood. Recent studies have suggested that these thresholds could be reliable only within those areas where the relationship between BTS values and permafrost occurrence were initially observed (Swiss Alps), with different local conditions leading to different BTS thresholds (Ishikawa, 2003; Lewkowicz and Ednie, 2004; Lambiel and Pieracci, 2008). Despite this, the conditions from the study site, characterized by rugged topography and a pronounced shadow effect from the ridges, lead to a continuous snow cover during the winter that can easily surpass 2 m, resembling local conditions met in the Alps. Therefore, based on previous works (Vespremeanu-Stroe et al., 2012; Onaca, 2013; Onaca et al., 2013b), we regard the categorization of permafrost classes used in the current paper as a good proxy for indicating permafrost occurrence within the investigated area. Two homemade, 2.5 m long carbon fibre BTS probes, equipped with PT 100 (0.1°C accuracy) digital thermometers were used in March 2013 to collect a total of 170 BTS measurements. Due to a combination of elements (bad weather, rough terrain and thick snow layer), it was impossible to obtain a more complete set of BTS measurements. Each thermometer was calibrated and tested, showing no notable differences in results. Temperatures were registered for sites having a snow depth of at least 80 cm, while the location of each individual measurement was obtained using a differential GPS (Trimble GeoExplorer 6000) with an overall sub metric accuracy. The sampling strategy was elaborated based on previous results which indicate permafrost occurrence within several rock glaciers from the Southern Carpathians(Onaca et al., 2013b). A mesh of equally distributed individual BTS locations was generated for each investigated rock glacier using ArcGIS 10 (Data Management Tools). In the case of the Pietrele rock glacier, the complex local topography composed of a series of furrows and ridges, along with a significant change in block size and snow cover thickness (due to snow input from neighbouring avalanche tracks), led to an uneven distribution of BTS measurements. To gain a better coverage of elevation, often considered one of the main controlling variables of permafrost distribution (Gruber and Hoelzle, 2001; Isaksen et al., 2002), several BTS measurements were executed at lower elevations outside of the investigated landform. Snow cover thickness was recorded for each individual measurement to ensure that it was at least 80 cm. To provide enough time for the probe temperature to equilibrate with the snow cover, temperature readings were made after 10 minutes (Brenning et al., 2005; Julián and Chueca, 2007). As suggested by Brenning et al. (2005), analyses for detecting autocorrelation of the data were undertaken using Moran’s I (ArcGIS 9.3). The analyses confirmed that spatial autocorrelation was not significant after the BTS measurements situated at a distance of < 25 m were grouped together and averaged.

3.2. Derivation of predictor variables

10A 15-m digital elevation model (DEM) generated from a point cloud with 15 m spacing, obtained by stereo restitution of aerial images and covering the investigated area, was used to generate and extract the topographic parameters with the highest influence on permafrost distribution in alpine regions. To generate the DEM, the Kriging interpolation method (implemented in the GIS-software ArcMap 10) was applied to the original set of elevation points. A lowpass filter was applied to the resulting DEM to reduce the significance of anomalous cells.

11Based on the 15-m DEM, several parameters were calculated for the evaluation of permafrost distribution within the study area. One of the key factors controlling permafrost occurrence is the local elevation because the altitude largely controls the mean annual air temperature and thus may provide a rough estimation of permafrost distribution (Etzelmüller et al., 2001). Global solar radiation is considered to be one of the leading factors in determining permafrost distribution in mountainous areas (Etzelmüller et al., 1998; Harris and Pedersen, 1998; Kneisel et al., 2000; Julián and Chueca, 2007). The amount of the received solar radiation was computed using the Solar Analyst extension implemented in the software ArcMap 10 only for the summer season (July-September) because the albedo of snow does not influence the solar radiation balance in these months (Funk and Hoelzle, 1992). Another determining factor is the slope, according to the assumption that steep slopes generally correspond to rockwalls where unconsolidated debris deposits with high porosity are missing and thus the cooling air circulation is absent (Harris and Pedersen, 1998; Delaloye and Lambiel, 2005). The profile curvature was computed using Landserf 2.3 and was considered for this analysis based on the assumption that concave surfaces are thought to favour snow cover persistence for a longer period of time, and thus prevent permafrost melting during the summer (Frauenfelder et al., 1998). The last parameter considered was the vegetation cover, which was included as a variable that influences permafrost occurrence based on the assumption that the existence of Alpine permafrost is strongly related to areas free of permanent vegetation (Hoelzle et al., 1993). The vegetation cover was quantified using the normalized difference vegetation index (NDVI), derived from a LandsatTM image from August 2011 using the image processing module implemented in the software IdrisiAndes.

12Individual point-parameters were extracted from the five predictor variables mentioned above to evaluate the strength of their linear correlations with the BTS data values. The resulting variables were initially tested for normality to a 95% confidence level using the Shapiro-Wilk normality test (Shapiro and Francia, 1972). A multiple linear regression analysis was performed using the statistical package SPSS to model the relationships between the BTS data values, used as the dependent variable, and the predictor variables (elevation, solar radiation, slope, profile curvature and NDVI), used as independent variables. A Forward Stepwise regression method was applied to perform the analysis, thus only the independent variables, which significantly improved the regression model, were included in the resulting model.

3.3. Model validation

13To validate the resulting model of permafrost distribution, the following analyses were performed: (i) Spatially overlapping and quantifying the differences between the resulting model of permafrost distribution with the rock glacier distribution in the central part of the Retezat Mountains (Urdea, 2000). Intact and relict rock glaciers are commonly used to confirm permafrost presence within a given area (Barsch, 1996; Humlum, 2000; Etzelmüller et al., 2001; Etzelmüller and Hagen, 2005; Haeberli et al., 2006; Etzelmüller et al., 2007) as well as to investigate the lower limits of mountain permafrost (Frauenfelder et al., 2001; Janke, 2005; Ribolini and Fabre, 2006). Therefore, 20 intact rock glaciers (fig. 2A) are used to confirm permafrost presence, while 44 relict rock glaciers (fig. 2B) are used as indicators for permafrost absence; (ii) Comparing the resulting model of permafrost distribution with the existing geophysical investigations (ERT and GPR profiles; Onaca, 2013; Onaca et al., 2013b); (iii) Comparing the resulting model of permafrost distribution with previously undertaken BTS campaigns and ground surface temperature data (Urdea, 1992; Vespremeanu-Stroe et al., 2012; Onaca, 2013).

3.4. Ground surface temperature (GST) regime

14One miniature thermistor (iButton Digital Thermometers DS 1922L; fig. 2) was installed in August 2012 at Judele rock glacier to automatically monitor the ground surface temperature (GST) evolution during one season. The thermistor registered the GST every two hours, with a resolution of 0.065° C and a ±0.5°C accuracy.

Fig. 2 – Data logger placement and classification examples of rock glaciers within the study area.
Fig. 2 –Placement des capteurs autonomes de température et exemples de classification des glaciers rocheux dans la zone étudié. 

Fig. 2 – Data logger placement and classification examples of rock glaciers within the study area. Fig. 2 –Placement des capteurs autonomes de température et exemples de classification des glaciers rocheux dans la zone étudié. 

A. 1: intact rock glacier; 2: data logger; B. 3: relict rock glacier.
A. 1 : glaciers rocheux intact ; 2 : capteurs autonomes de température ; B. 3 : glaciers rocheux relictes.

15The temperature data were calibrated at 0°C using zero-curtain intervals because snowmelt isothermal conditions occur at the ground surface. The thermistor was installed in a trench, free of vegetation, soils or other fine-grained sediments. We covered the thermistor with 5-10 cm of pebbles to avoid direct exposure to solar radiation. Based on the data recorded by the thermistor, we determined the winter equilibrium temperature (WEqT) when sufficiently thick snow cover provides a complete isolation of the ground and the GST remains relatively constant.

4. Results

4.1. Temperature measurements (BTS) and distribution

16A total of 170 BTS values ranging from -0,8°C to -7.3°C were measured on the Pietrele (47), Pietrele II (12), Pietricelele (22), Valea Rea (15), Valea Rea III (5) Judele (41) and Ana (8) rock glaciers or nearby (20) in March 2013. Of these, 86 (51%) BTS values lower than -3°C indicate that permafrost occurrence is probable, 36 (21%) suggest the possible presence of permafrost (-3 < BTS < -2°C), while 48 (28%) indicate the absence of permafrost (BTS > -2°C). Despite its eastern orientation, the Judele rock glacier shows a very low mean BTS of –4,1°C, with most of the values (32) suggesting probable permafrost occurrence, while Pietrele has a mean BTS of -2.5°C, due to a higher proportion of BTS measurements indicating the absence of permafrost (13; tab. 1).

Tab. 1 – Permafrost occurrence at the investigated rock glaciers, as indicated by BTS measurements.
Tab. 1 – Occurrence du pergélisol dans les glaciers rocheux étudiés, d'après les mesures BTS.

Rock glacier

No. of BTS

points

Mean BTS (˚C)

Temperature range (˚C)

PPR points

PP points

NP

points

Pietrele

47

-2.5

-4.3 - -0.8

17

17

13

Pietrele II

12

-3.7

-4.8 - -2.3

9

2

-

Pietricelele

22

-3.3

-5.3 - -1.5

14

5

3

Valea Rea

15

-2.9

-5.1 - -1

6

5

4

Valea Rea III

5

-4.1

-5.7 - -3.1

5

-

-

Judele

41

-4.1

-7.3 - -1.1

32

6

3

Ana

8

-2.3

-3.2 - -1.5

2

3

3

17The large variability of BTS values measured on Pietrele is probably due to the greater roughness of the surface of this rock glacier due to small scale variations in the microtopography and changes in boulder size. Additionally, the characteristics of the active layer (thickness and porosity) could also induce large differences in the energy fluxes at the ground surface.

18Permafrost occurrence at the investigated rock glaciers, as indicated by BTS measurements, is present at altitudes above 2000 m, with a maximum concentration above 2150 m and restricted to areas with solar radiation values generally lower than 550 KWH/m2. Permafrost occurrence is strictly related to vegetation free surfaces (as indicated in table 3), where most of the BTS measurements were performed. A synthesis of the spatial distribution of BTS data in relation to each predictor variable is given in figure 3.

Fig. 3 – Spatial distribution of BTS data according to predictor variables.
Fig. 3 – Distribution spatiale des valeurs BTS en accord avec les variables prédictives.

Fig. 3 – Spatial distribution of BTS data according to predictor variables. Fig. 3 – Distribution spatiale des valeurs BTS en accord avec les variables prédictives.

A: BTS values (oC); B: snow thickness (cm); C: solar radiation (WH/m2); D: elevation range (m); E: NDVI values; F: slope angle (degrees); G: profile curvature values.
A : valeurs BTS (oC) ; B : épaisseur de la neige (cm) ; C : radiation solaire (WH/m2) ; D : valeurs d’altitude (m) ; E : valeurs de NDVI ; F : angle de pente (degrés) ; G : valeurs de la courbure du profil.

4.2. Spatial modelling of the BTS values

19The statistical relations between the BTS values and the predictor variables (elevation, solar radiation, slope, profile curvature and NDVI) were tested using a Linear Correlation Analysis and are reported in figure 4. These analyses reveal that BTS data are strongly associated with solar radiation, elevation and NDVI (p < 0.001), while slope and profile curvature (p > 0.001) reveal a very weak correlation to the BTS values.

Fig. 4 – Boxplots showing the distribution and correlation coefficients between BTS and predictor variable data.
Fig. 4 – Boîtes à moustaches indiquant la distribution et le coefficient de corrélation entre les valeurs BTS et les variables prédictives.

Fig. 4 – Boxplots showing the distribution and correlation coefficients between BTS and predictor variable data. Fig. 4 – Boîtes à moustaches indiquant la distribution et le coefficient de corrélation entre les valeurs BTS et les variables prédictives.

A: BTS values (oC); B: solar radiation (WH/m2); C: elevation range (m); D: NDVI values; E: slope angle (degrees); F: profile curvature values.
A : valeurs BTS (oC) ; B : radiation solaire (WH/m2) ; C : valeurs d’altitude (m) ; D : valeurs de NDVI ; E : angle de pente (degrés) ; F : valeurs de la courbure du profil.

20The Pearson correlation coefficient indicates significant differences between BTS data and the variables, thus the most significant correlation coefficient was reported between BTS values and solar radiation, P = 0.548 (fig. 4), indicating that solar radiation, which is strongly influenced by slope, aspect and topographic shading, is the primary factor controlling permafrost distribution at the study site. The existence of permafrost seems to have an acceptable negative correlation with elevation (P = -0.478) and a positive correlation with NDVI (P = 0.391). The slope and profile curvature are not significantly correlated with permafrost occurrence. Tables 2, 3 and 4 show the results of the multiple linear regression analysis. Because the Forward Stepwise regression algorithm in the resulting model only includes the independent variables, which significantly improves the strength of the prediction, only the solar radiation, vegetation cover and elevation were considered to control the permafrost occurrence at the study site. As indicated previously by the linear correlation analysis, slope and profile curvature cannot be considered when modelling permafrost occurrence at the investigated site, due to a low correlation coefficient.

Tab. 2 – Summary of the generated regression model.
Tab. 2 – Statistiques du modèle de régression.

Model

R

R2

Standard error of estimate

Durbin-Watson

Retezat

0.69

0.48

0.97

1.967

Tab. 3 – ANOVA of the models (p value of F-statistic < 0.001).
Tab. 3 – Résultats de l’analyse de variance ANOVA (valeur-p de la statistique F < 0,001)

Model

Sum of squares

Degrees of freedom

Mean square

F

Retezat

Regression

129.52

3

43.176

45.279

Residual

139.21

167

0.954

Total

172.177

170

Tab. 4 – Coefficients of the resulted model.
Tab. 4 – Coefficients du modèle résultant.

Model

Unstandardized Coefficients

Standardized Coefficients

Beta

t

P value

B

Std. Error

Constant

1.478

3.313

0.446

0.036

Solar Radiation

0.000014

0.000

0.440

7.045

0.000

NDVI

4.0373

0.867

0.286

4.657

0.000

Elevation

-0.00571

0.00135

-0.271

-4.207

0.000

21The summary of the generated model (tab. 2) exhibits an R2 of 0.48, being within the range of the values obtained for R2 in similar approaches, where the highest R2 = 0.63 was achieved by Ødegård et al. (1999) for the Jotunheimen in southern Norway and the lowest value (R2 = 0.20) by Tanarro et al. (2001) in the Spanish Sierra Nevada. Similar values of R2 (R2 = 0.40) were reported for the Alps, by Gruber and Hoelzle (2001) and for the Yukon Territory by Lewkowicz and Ednie (2004). The value of the standard error of the estimate indicates that permafrost distribution temperatures could be over or underestimated by up to 0.97°C. The independence of the variables is confirmed by the value of the Durbin-Watson parameter (1.967). According to the ANOVA table (tab. 3), the F statistics of the model has a significance value of 0.001, thus rejecting the null hypothesis and confirming the potential of the variables included in the model to predict permafrost distribution.

22Following the literature (Julián and Chueca, 2007) several statistical tests were conducted to check the validity of the model: a) linear correlation (Pearson correlation value); b) independence of the variables (Durbin-Watson parameter); c) normality distribution of the residuals (visual inspection of P-P plots of regression standardized residuals and a Shapiro-Wilk normality test); d) constant variance (homoscedasticity; visual inspection of scatterplots of standardized residuals versus standardized predicted values); and e) colinearity between variables (VIF). Each of these statistical tests confirmed the reliability of the multiple linear regression model. Based on the coefficients of the model (tab. 4), we noticed that all of the factors included in the model (p < 0.05) have predictive ability for determining the BTS distribution within the investigated area. According to the standardized beta coefficients, which enable us to observe the importance of each factor in controlling permafrost distribution, it is possible to assume that solar radiation plays the most important role in controlling permafrost occurrence in the investigated area, followed by vegetation cover (NDVI) and elevation. Elevation has an inverse relation with the BTS data, thus confirming the role of air temperature as a restrictive factor for alpine permafrost occurrence.

23The equation for the Permafrost distribution model was generated using the unstandardized coefficient values (tab. 4) and is as follows:
BTS= 1.477798 + 0.000014 * Solar Radiation + 4.037341 * NDVI - 0.005711 * Elevation

24The above equation was used in ArcMap 10 to generate a modelled BTS surface for the entire test site (fig. 5).

Fig. 5 – Predicted basal temperature of snow (BTS) cover (oC).
Fig. 5 – Distribution des valeurs de température à la base de la neige (BTS), prédite par le modèle.

Fig. 5 – Predicted basal temperature of snow (BTS) cover (oC). Fig. 5 – Distribution des valeurs de température à la base de la neige (BTS), prédite par le modèle.

A: intact rock glaciers; B: relict rock glaciers.
A : glaciers rocheux intacts ; B : glaciers rocheux relictes.

25According to the modelled BTS surface, very low BTS values (below -3°C) are expected to be found above 2000 meters, where the incoming solar radiation is very small, especially on Northern exposed valleys. The permafrost probability model was generated by reclassifying the modelled BTS surface according to the predicted permafrost thresholds. Thus, the permafrost occurrence at the study site in regard to the permafrost distribution assigned by the `rules of thumb` (Haeberli, 1973) is as follows (fig. 6):

  • Probable Permafrost (PPR) areas occupy 17 km2 (28% of the investigated area), extending mainly at elevations above 2000 m. PPR areas are strictly related to areas free of vegetation and are characterized by a very small influx of solar radiation (less than 550000 WH/m2 in summer). PPR areas have a greater extent in north-facing valleys, where a pronounced shadow effect from the ridges exists and blocky slopes prevail.

  • Possible Permafrost (PP) areas occupy 14 km2, covering 24% of the total investigated area, a slightly smaller area than that of PPR. The elevation range of PP is situated between 1900 and 2000 m, especially in the north-facing valleys, while the specific amount of incoming solar radiation ranges between 55000 and 450000 WH/m2 during the summer months.

  • Non Permafrost (NP) areas cover 28 km2 (48% of the total investigated area) below 1900 m elevation, where most of the surfaces are vegetated and the solar radiation influx is over 550000 WH/m2 during the summer months.

Fig. 6 – Predicted permafrost probability.
Fig. 6 – Distribution du pergélisol dans la zone d'étude telle que prédite par le modèle.

Fig. 6 – Predicted permafrost probability. Fig. 6 – Distribution du pergélisol dans la zone d'étude telle que prédite par le modèle.

1: probable permafrost; 2: possible permafrost; 3: no permafrost; 4: intact rock glaciers; 5: relict rock glaciers.
1 : pergélisol probable ; 2 : pergélisol possible ; 3 : absence de pergélisol ; 4 : glaciers rocheux intacts ; 5 : glaciers rocheux relictes.

4.3. Model validation

26The validation of the permafrost probability model was performed using independent data sets confirming the presence or absence of permafrost. For that purpose, we used all of the measurements and observations available from previous studies (Urdea, 1992; Urdea, 2000; Vespremeanu-Stroe et al., 2012). Because long-term creep deformation of rock glaciers requires permafrost conditions, rock-glaciers are often used to model the past or the present distribution of discontinuous permafrost (Frauenfelder et al., 2001; Etzelmüller et al., 2007). The contours of all of the rock glaciers from the centre part of the Retezat Mountains (Urdea, 2000), as well as their intact/relict character, were taken into account as indicators of permafrost absence or presence (fig. 6). According to our permafrost probability model, all 20 landforms that are considered intact rock glaciers and are situated above 2000 m near the Retezat (2482 m), Bucura (2433 m), Judele (2398 m) or Peleaga (2509 m) peaks fall entirely in PPR areas. Of these, three of the largest rock glaciers within the study area (Valea Rea, 0.4 km2; Pietrele, 0.2 km2 and Ana, 0.15 km2) also exhibit PP areas toward their frontal part. In regard to predicting nonpermafrost areas according to our model, from the total number of 44 relict rock glaciers, 41% fall into NP areas (situated below 1900 m), 6 landforms (14%) have less than half of their surface areas classified as PP areas (situated between 1900 and 2000 m), and the remaining 20 landforms (45%) have more than half of their surfaces classified as either PPR or PP areas (situated at over 2000 m).

27The comparison between the permafrost probability map and previous geophysical and thermal investigations (Urdea, 1998; Vespremeanu-Stroe et al., 2012; Onaca, 2013), carried out on 8 intact rock glaciers (Pietrele, Pietricelele, Judele, Lower Ana, Upper Ana, Valea Rea 1, Valea Rea 2 and Știrbu; tab. 5), showed that the correspondence between these is also reliable, but local mismatches may also occur. In the case of the Judele rock glacier, there is a good correspondence between previous investigations and the current permafrost probability model (fig. 7).

Tab. 5 – Previous studies concerning permafrost occurrence within the rock glaciers from the central part of the Retezat Mountains.
Tab. 5 – Études antérieures concernant la présence du pergélisol dans les glaciers rocheux de la partie centrale du massif de Retezat.

Investigated Rock glacier

Permafrost occurrence

Research method

Years of investigation

References

Geophysical investigations

Thermal investigations

Pietrele

Yes

-

BTS

1998; 2010; 2012

Urdea, 1992; Vespremeanu et al., 2012; Onaca, 2013.

ERT

-

2007

Urdea et al., 2008 .

GPR

-

2012

Onaca, 2013.

-

GST regime

2008-2010; 2012-2013

Vespremeanu et al., 2012; Onaca, 2013.

Pietricelele

Yes

-

GST regime

2008-2009

Vespremeanu et al., 2012.

Judele

Yes

-

GST regime

2008-2010; 2012-2013

Vespremeanu et al., 2012; Onaca, 2013.

GPR

-

2013

Onaca, 2013.

Lower Ana

Yes

ERT

-

2007

Urdea,et al. 2008.

-

GST regime

2008-2009; 2010-2011

Vespremeanu et al., 2012; Onaca, 2013.

Upper Ana

Yes

-

GST regime

2008-2009

Vespremeanu et al., 2012.

Valea Rea 1

Yes

-

GST regime

2008-2009; 2012-2013

Vespremeanu et al., 2012; Onaca, 2013.

Valea Rea 2

Yes

-

GST regime

2012-2013

Onaca, 2013.

Știrbu

Yes

-

GST regime

2008-2010

Vespremeanu et al., 2012.

ERT: electrical resistivity tomography; GPR: ground penetrating radar; BTS: bottom temperature of winter snow; GST: ground surface temperature.
ERT : tomographie de la résistivité électrique ; GPR : radar à pénétration de sol ; BTS : mesures de température à la base de la neige ; GST : température de surface du sol.

Fig. 7 – Synthesis of permafrost occurrence within the investigated area revealed by previous geophysical and thermal investigations.
Fig. 7 – Synthèse d’occurrence du pergélisol dans la zone étudiée, révélée par des investigations géophysiques and thermiques antérieures.

Fig. 7 – Synthesis of permafrost occurrence within the investigated area revealed by previous geophysical and thermal investigations. Fig. 7 – Synthèse d’occurrence du pergélisol dans la zone étudiée, révélée par des investigations géophysiques and thermiques antérieures.

A: probable permafrost; B: possible permafrost; C: no permafrost; D: permafrost occurrence according to previous investigations; E: intact rock glaciers; F: relict rock glaciers. Previously investigated rock glaciers: 1: Pietrele; 2: Pietricelele; 3: Valea Rea; 4: Valea Rea III; 5: Ana; 6: Upper Ana; 7: Judele; 8: Știrbu.
A : pergélisol probable ; B : pergélisol possible ; C : absence de pergélisol ; D : occurrence du pergélisol en accord des investigations précédentes ; E : glaciers rocheux intacts ; F : glaciers rocheux relictes. Glaciers rocheux investigués précédemment : Pietrele ; 2 : Pietricelele ; 3 : Valea Rea ; 4 : Valea Rea III ; 5 : Ana ; 6 : Ana supérieur ; 7 : Judele ; 8 : Știrbu.

28A total of 6 thermistors measured -3°C or lower WEqT values and negative mean annual ground surface temperatures, while only one datalogger located in the south-eastern part recorded higher values indicating a non-permafrost area (Onaca, 2013). Additionally, a transversal GPR profile, made in 2013 in the upper part of the rock glacier, showed a continuous and consistent reflector that was interpreted as the top of the permafrost layer, at 4 m depth (Onaca, 2013). According to our model, the Judele rock glacier is almost completely within the PPR class, except for some small patches close to its northern limit, despite the fact that one thermistor showed that non-permafrost surfaces occur in its southeastern part. In the case of the Lower Ana rock glacier, a high resistivity anomaly, interpreted as a permafrost body, was identified in the upper part of the rock glacier by an ERT profile performed in 2007, while the resistivity values were lower in the lower part of the profile, suggesting an absence of permafrost (Vespremeanu-Stroe et al., 2012). Our permafrost probability model showed that PPR and PP areas occur in the upper and central part of the rock glacier, while NP is present in the lower part. The data from the thermistors are contradictory because BTS values lower than -2°C suggested probable or possible permafrost existence in three cases, while one thermistor corresponding to a PP area suggested the absence of permafrost (Onaca, 2013). Regarding the Pietrele rock glacier, both the model and the validation data (BTS from March 2012, thermistors, ERT and GPR profiles) showed the existence of permafrost in the upper and western part of the rock glacier (fig. 7). In the frontal part, where our model estimated PP areas, most of the BTS data suggest permafrost absence. In the case of the Pietricelele rock glacier, a GPR transversal profile performed in 2013 showed several strong reflectors at a depth of 3 m, interpreted as permafrost lenses. Despite local mismatches regarding permafrost prediction at the location of 20 relict rock glaciers situated mainly above the 2000 m mark, verification with the rock glacier inventory and with both previous geophysical and thermal investigations showed a good fit with the predicted permafrost distribution. Thus, it can be assumed that the model offers a realistic overview of today’s permafrost distribution within the central part of the Retezat Mountains.

4.4. GST regime

29The GST evolution at the Judele site is shown in figure 8. During the snow free interval, short-term fluctuations in GST values occurred at the Judele site. In the second part of the autumn, a significant ground cooling was observed before the installation of the insulating snow cover. After the height of the snow cover increased sufficiently to ensure a thermal insulation to the ground, the daily fluctuations disappeared and a slight increase in the negative GST regime occurred. The insulating snow layer occurred on 14 December, whereas the winter equilibrium temperature interval occurred between 21 February and 26April. The mean temperature for this period (-3.9°C), together with the mean annual ground surface temperature (-0.1°C) suggests the likely occurrence of permafrost at this site. The snow cover disappeared completely on 25 June, after a zero-curtain interval of 61 days.

Fig. 8 – GST evolution in the 2012-2013 season at Judele site.
Fig. 8 – Évolution de GST dans le site Judele, saison 2012-2013.

Fig. 8 – GST evolution in the 2012-2013 season at Judele site. Fig. 8 – Évolution de GST dans le site Judele, saison 2012-2013.

A: time of first freezing; B: BTS field measurements; C: BTS interval; D: zero curtain.
A : temps de première congélation ; B : mesures BTS sur le terrain ; C : intervalle BTS ; D : Phase zéro.

5. Discussions

5.1. Permafrost distribution in the central part of the Retezat Mountains

30The distribution of mountain permafrost for the central part of the Retezat Mountains was successfully estimated based on statistical modelling of the relations between the measured BTS values and the permafrost predictor variables within the study area (global solar radiation, elevation and vegetation cover). The resulting map provides valuable input data for permafrost distribution scenarios, thus contributing to a better understanding of our geosystem. Permafrost occurrence, as predicted by the model, showed a good correspondence with both the geophysical and thermal data (BTS and GST regime; tab. 5) and the considered empirical data, especially in the case of the rock glaciers found between 2000 and 2300 m. Permafrost occurrence was successfully predicted at the location of several pre investigated rock glaciers: Pietrele, Pietricelele, Lower and Upper Ana, Judele, Valea Rea and Valea Rea II. The accuracy of the model decreases at higher altitudes where most of the northern and some of the eastern and western oriented slopes were predicted as PRP areas. In this respect, recent studies studies (Onaca et al., 2013b; Popescu et al., 2015) have shown that permafrost is likely to occur in the Southern Carpathians on north-facing rock walls above 2300 m where the recorded long-term mean annual surface temperatures proved to be negative. Below 2000 m, the permafrost existence is scattered, but we expect isolated patches of permafrost to occur within a few rock glaciers where favourable conditions are present. Regardless, several predictions may be incorrect, especially in the case of the PRP indicated below 1900 m outside the surfaces of the rock glaciers. A few examples can be found in the Pietrele and Stânişoara valleys, below 1900 m and having a western exposition, where our field observations revealed that permafrost is unlikely to occur.

31The main source of uncertainty in the proposed permafrost distribution model is related to the BTS traditional thresholds, which in some cases can have a limited validity due to strong dependence on small scale variations of the surface topography (occurrence of large boulders) and to local snowpack (Otto et al., 2012). Surface roughness has a great impact on the local solar radiation influx, contributing to the shading effect in mountain regions, which is well known to significantly impact permafrost occurrence within a given area (Schrott, 1994, 1996). As stated before, snowpack history plays a crucial role in determining WEqT through BTS measurement, requiring a continuous snowpack of at least 80 cm since the beginning of winter to provide sufficient insulation during the cold days of winter. Additionally, subsurface ventilation exhibits a substantial influence on permafrost occurrence, locally reducing ground temperatures and causing permafrost existence within the lower parts of talus slopes and warmer sites within their upper parts (Delaloye and Lambiel, 2005). Therefore, because the winter snow cover showed a sufficient thickness at the location of the BTS measurements (fig. 3) and that the conditions for WEqT were met (fig. 8), it can be assumed that any mismatches between permafrost distribution as depicted by the model and actual permafrost occurrence within the investigated area are due to specific local subsurface air circulation and local variations in surface roughness.

32A visual comparison between our local (15-m resolution) permafrost distribution model of the central part of the Retezat Mountains with the corresponding model extracted from the high-resolution estimate of global permafrost zonation (PZI; Gruber, 2012) revealed strong similarities. The PZI model has a lower resolution (30 arc-seconds/<1 km) and uses the proportion of a given area that is underlain by permafrost (PE) as a basis for obtaining a global permafrost zonation index (PZI). The approach is based first on modelling PE at a global scale as a function of MAAT alone and as a combined result with an aggregate effect of several stochastically parameterized phenomena with a fine scale variability (snow cover, exposition to solar radiation, vegetation and subsurface characteristics; Gruber, 2012). In figure 9, both models are presented, depicting the central part of the Retezat Mountains.

Fig. 9 – Comparison of permafrost distribution models for the central part of the Retezat Mountains.
Fig. 9 – Comparaison des modèles de répartition du pergélisol dans la partie centrale du massif de Retezat.

Fig. 9 – Comparison of permafrost distribution models for the central part of the Retezat Mountains. Fig. 9 – Comparaison des modèles de répartition du pergélisol dans la partie centrale du massif de Retezat.

A: Permafrost distribution model based on the approach of Gruber (2012); B: Proposed permafrost distribution model; 1: fringe of uncertainty; 2: permafrost only in favourable conditions; 3: probable permafrost; 4: possible permafrost; 5: no permafrost.
A : modèle de répartition du pergélisol basée sur la stratégie proposée par Gruber (2012) ; B : modèle de répartition du pergélisol proposé ;  1 : frange d’incertitude ; 2 : pergélisol à conditions le plus favorable ; 3 : pergélisol probable ; 4 : pergélisol possible ; 5 : absence de pergélisol.

33It is clear that our small scale model (fig. 9A) offers a more realistic picture of the local permafrost pattern and isolated patches, which have a greater extent in north facing valleys. In contrast, the world wide permafrost zonation map (fig. 9B), explicitly made for determining permafrost distributions at larger scales, overestimates the permafrost distribution within the study site and indicates large surfaces of permafrost occurrence. Thus, according to the PZI model, the area near Bucura Lake (which has a southern exposition) is depicted as a permafrost occupied location, when permafrost occurrence is highly improbable according to both our model and specific local conditions (areas occupied by alpine meadows). Nevertheless, despite its limitation over small areas, we consider the PZI model a very useful approach that practically confirms the possibility of permafrost occurrence within our study area and also enables the comparison amongst mountain ranges of similar scale.

5.2. Controlling factors

34In the central part of the Retezat Mountains, the distribution of mountain permafrost is clearly linked to the specific topography of the study site, which plays an extremely important role in controlling solar radiation inputs. According to our model, the probability of permafrost occurrence is highly dependent on the solar radiation influx. These results are confirmed by similar studies, suggesting that low-altitude mountain permafrost is strongly controlled by the solar radiation input (Schrott, 1994; Lewkowicz and Ednie, 2004). Incoming solar radiation is very small, especially in the case of N, NE and NW facing slopes, due to the pronounced shadow effect of the ridges which could also have significant influence on local permafrost preservation (Schrott, 1996).

35A remarkable correlation was found between BTS values and the vegetation cover (NDVI), thus confirming the assumption that alpine permafrost is strongly related to those areas free of permanent vegetation (Hoelzle et al., 1993). Previous studies (Onaca et al., 2013b; Popescu et al., 2015) have shown that permafrost existence in the Southern Carpathians is restricted to debris-covered slopes composed of large boulders that allow a colder ground temperature regime than elsewhere (Harris and Pedersen, 1998; Scapozza et al., 2011). This thermal anomaly (particularly through the Balch effect) frequently occurs in high porosity unconsolidated debris deposits because cold air is trapped between large boulders, thus enhancing permafrost preservation (Humlum, 1997). In many situations, due to large air-filled voids within the coarse debris mantle, subsurface ventilation (chimney effect) can locally occur, causing a reduction of ground temperature in the lower part of the talus slopes (Delaloye and Lambiel, 2005).

36Despite the fact that the elevation is the main factor controlling air temperature, the statistical correlation between BTS values and elevation is not as good as in the European Alps (Boeckli et al., 2012), but it was considered acceptable. A possible explanation for this could be the limited altitudinal range of BTS measurements (1900-2200 m). According to previous studies, the lower boundary of permafrost in the Retezat Mountains is approximately 2000 m as in our model. Isolated patches of permafrost may occur even lower, but only in sites where favourable conditions for permafrost preservation occur. In addition, the cold-air drainage (chimney effect) within debris covered surfaces modifies the relationship between BTS values and elevation, and some mismatches may occur. In the case of the Lower Ana rock glacier, this circulation was not identified in 2010-2011, but it was highlighted in other cases in the Southern Carpathians (Popescu et al., 2015). The two other predictor variables, slope and curvature, were not considered in the multiple linear regression model due to low correlations recorded between BTS values and these parameters.

6. Conclusions

37For the first time, a high-resolution permafrost probability model is available for the central part of the Retezat Mountains. According to this model, permafrost is especially likely to occur above 2000 m on north-facing valleys, while at lower altitudes, permafrost occurrence is scattered and may be present at only a few sites where favourable conditions occur.

38The empirical-statistical approach has proven to be successful for modelling the permafrost distribution in the central part of the Retezat Mountains. With the exception of profile curvature and slope, the other predictor variables (solar radiation, elevation and vegetation cover) were successfully correlated with the BTS data. As noted by similar studies (Tanarro et al., 2001; Julián and Chueca, 2007; Schrott et al., 2012), solar radiation was found to be the leading factor in controlling mountain permafrost occurrence at the study site. The resulting model showed a good correspondence with previous field investigations regarding permafrost occurrence at the investigated area, having a good overall accuracy (R2 = 0.48). The permafrost probability model was validated with previous geophysical and thermal data, showing a reasonable correspondence.

39The current study offers valuable information regarding permafrost occurrence in the Retezat Mountains, indicating that from a total of 56 km2 investigated, 31 km2 (52%) seems to exhibit favourable conditions for permafrost occurrence, amongst which 14 km2 was indicated as probable permafrost (PRP), while the remaining 17 km2 was indicated as possible permafrost (PP).

40In the future, an important challenge would be to generate a regional model of permafrost probability for the Southern Carpathians, but high resolution measurements of the predictor variables are required for the entire area, as well as a larger amount of well distributed calibration data.

Haut de page

Bibliographie

Barsch D. (1996) – Rockglaciers. Series: Springer Series in Physical Environment, ISBN: 978-3-642-80095-5. Springer Berlin Heidelberg (Berlin, Heidelberg), Dietrich Barsch (Ed.), vol. 16, 16.

Bodin X. (2005) – L'état thermique du glacier rocheux de Laurichard en 2003-2004: analyse des températures de surface, spatialisation du régime thermique et implications géodynamiques. Environnements périglaciaires, 19-38.

Boeckli L., Brenning A., Gruber S., Noetzli J. (2012) – A statistical approach to modelling permafrost distribution in the European Alps or similar mountain ranges. The Cryosphere 6, 125-140.

Bonnaventure P.P., Lewkowicz A.G., Kremer M., Sawada M.C. (2012) – A permafrost probability model for the southern Yukon and northern British Columbia, Canada. Permafrost and Periglacial Processes 23, 52-68.

Brenning A., Gruber S., Hoelzle M. (2005) – Sampling and statistical analyses of BTS measurements. Permafrost and Periglacial Processes 16, 383-393.

Delaloye R., Lambiel C. (2005) – Evidence of winter ascending air circulation throughout talus slopes and rock glaciers situated in the lower belt of alpine discontinuous permafrost (Swiss Alps). Norsk Geografisk Tidsskrift 59, 194-203.

Dobinski W. (1998) – Permafrost occurrences in the alpine zone of the Tatra Mountains, Poland. Proceedings, Seventh International Conference on Permafrost, Yellowknife, 231-237.

Etzelmüller B., Berthling I., Sollid J.L. (1998) – The distribution of permafrost in Southern Norway; a GIS approach. Seventh International Conference on Permafrost, Proceedings. Collection Nordicana. Centre d'Etudes Nordiques, Universite Laval, Quebec, PQ, Canada, 251-257.

Etzelmüller B., Farbrot H., Guðmundsson Á., Humlum O., Tveito O. E., Björnsson H. (2007) – The regional distribution of mountain permafrost in Iceland. Permafrost and Periglacial Processes 18, 185-199.

Etzelmüller B., Hagen J. O. (2005) – Glacier-permafrost interaction in Arctic and alpine mountain environments with examples from southern Norway and Svalbard. Geological Society, London, Special Publications, 242, 11-27.

Etzelmüller B., Heggem E.S., Sharkhuu N., Frauenfelder R., Kääb A., Goulden C. (2006) – Mountain permafrost distribution modelling using a multi-criteria approach in the Hövsgöl area, northern Mongolia. Permafrost and Periglacial Processes 17, 91-104.

Etzelmüller B., Hoelzle M., Flo Heggem E.S., Isaksen K., Mittaz C., Mühll D.V., Ødegård R.S., Haeberli W., Sollid J.L. (2001) – Mapping and modelling the occurrence and distribution of mountain permafrost. Norsk Geografisk Tidsskrift 55, 186-194.

Frauenfelder R., Allgower B., Haeberli W., Hoelzle M. (1998) – Permafrost investigations with GIS – a case study in the Fletschhorn area, Wallis, Swiss Alps. Proceedings of the 7th International Conference on Permafrost, Yellowknife, 551-556.

Frauenfelder R., Haeberli W., Hoelzle M., Maisch M. (2001) – Using relict rockglaciers in GIS-based modelling to reconstruct Younger Dryas permafrost distribution patterns in the Err-Julier area, Swiss Alp. Norsk Geografisk Tidsskrift 55, 195-202.

Funk M., Hoelzle M. (1992) – A model of potential direct solar radiation for investigating occurrences of mountain permafrost. Permafrost and Periglacial Processes 3, 139-142.

Gardaz J.M. (1997) – Distribution of mountain permafrost, Fontanesses Basin, Valaisian Alps, Switzerland. Permafrost and Periglacial Processes 8, 101-105.

Gruber S. (2012) – Derivation and analysis of a high-resolution estimate of global permafrost zonation. The Cryosphere 6, 221-233.

Gruber S., Hoelzle M. (2001) – Statistical modelling of mountain permafrost distribution: local calibration and incorporation of remotely sensed data. Permafrost and Periglacial Processes 12, 69-77.

Haeberli W. (1973) – Die Basis-Temperatur der winterlichen Schneedecke als moglicher Indikator fur die Verbreitung von Permafrost in den Alpen. Z. Gletscherk. Glazialgeol. 9, 221-227.

Haeberli W. (1985) – Creep of mountain permafrost: internal structure and flow of alpine rock glaciers. Mitteilungen der Versuchsanstalt für Wasserbau, Hydrologie und Glaziologie 77, p 142.

Haeberli W., Gruber S. (2009) – Global warming and mountain permafrost. Permafrost soils. Springer Berlin Heidelberg, 205-218.

Haeberli W., Hallet B., Arenson L., Elconin R., Humlum O., Kääb A., Kaufmann V., Ladanyi B., Matsuoka N., Springman S., Mühll D.V. (2006) – Permafrost creep and rock glacier dynamics. Permafrost and Periglacial Processes 17, 189-214.

Harris S.A., Pedersen D.E. (1998) – Thermal regimes beneath coarse blocky materials. Permafrost and Periglacial Processes 9, 107-120.

Hoelzle M. (1992) – Permafrost occurrence from BTS measurements and climatic parameters in the Eastern Swiss Alps. Permafrost and Periglacial Processes 3, 143-147.

Hoelzle M. (1996) – Mapping and modelling of mountain permafrost distribution in the Alps. Norsk Geografisk Tidsskrift 50, 11-15.

Hoelzle M., Haeberli W., Keller F. (1993) – Application of BTS-measurements for modelling mountain permafrost distribution. Proceedings of the sixth international conference on permafrost, Beijing, 272-277.

Hoelzle M., Mittaz C.B. E., Haeberli W. (2001) – Surface energy fluxes and distribution models of permafrost in European Mountain areas: an overview of current developments. Permafrost and Periglacial Processes 12, 53-68.

Humlum O. (1997) – Active layer thermal regime at three rock glaciers in Greenland. Permafrost and Periglacial Processes 8, 383-408.

Humlum O. (2000) – The geomorphic significance of rock glaciers: estimates of rock glacier debris volumes and headwall recession rates in West Greenland. Geomorphology 35, 41-67.

Imhof M., Pierrehumbert G., Haeberli W., Kienholz H. (2000) – Permafrost investigation in the Schilthorn Massif, Bernese Alps, Switzerland. Permafrost and Periglacial Processes 11, 189-206.

Isaksen K., Hauck C., Gudevang E., Ødegård R.S., Sollid J.L. (2002) – Mountain permafrost distribution in Dovrefjell and Jotunheimen, southern Norway, based on BTS and DC resistivity tomography data. Norsk Geografisk Tidsskrift - Norwegian Journal of Geography 56, 122-136.

Ishikawa M. (2003) – Thermal regime at the snow–ground interface and their implications for permafrost investigation. Geomorphology 52, 105-120.

Janke J.R. (2005) – The occurrence of alpine permafrost in the Front Range of Colorado. Geomorphology 67, 375-389.

Julián A., Chueca J. (2007) – Permafrost distribution from BTS measurements (Sierra de Telera, Central Pyrenees, Spain): assessing the importance of solar radiation in a mid-elevation shaded mountainous area. Permafrost and Periglacial Processes 18, 137-149.

Keller F. (1992) – Automated Mapping of Mountain Permafrost Using the Program PERMAKART within the Geographical Information System ARC/INFO. Permafrost and Periglacial Processes 3, 133-138.

King L., Gorbunov A.P., Evin M. (1992) – Prospecting and mapping of mountain permafrost and associated phenomena. Permafrost and Periglacial Processes 3, 73-81.

King L., Kalisch A. (1998) – Published. Permafrost distribution and implications for construction in the Zermatt area, Swiss Alps. 7th International Conference on Permafrost. Proceedings. Collection Nordicana. Centre d'Etudes Nordiques, Université Laval, Yellowknife, Canada, 569-574.

Kneisel C., Hauck C., Mühll D.V. (2000) – Permafrost below the timberline confirmed and characterized by geoelectrical resistivity measurements, Bever Valley, eastern Swiss Alps. Permafrost and Periglacial Processes 11, 295-304.

Lambiel C., Pieracci K. (2008) – Permafrost distribution in talus slopes located within the alpine periglacial belt, Swiss Alps. Permafrost and Periglacial Processes 19, 293-304.

Lewkowicz A.G., Ednie M. (2004) – Probability mapping of mountain permafrost using the BTS method, Wolf Creek, Yukon Territory, Canada. Permafrost and Periglacial Processes 15, 67-80.

Ødegård R.S., Isaksen K., Mastervik M., Billdal L., Engler M., Sollid J.L. (1999) – Comparison of BTS and Landsat TM data from Jotunheimen, southern Norway. Norsk Geografisk Tidsskrift 53, 226-233.

Onaca A. (2013) – Periglaciar processes and landforms from the Southern Carpathians. A geomorphological and geophisical approach. Phd. Thesis, p. 237.

Onaca A., Urdea P., Ardelean A., Serban R. (2013a) – Assesment of internal structure of periglaciar landform from Southern Carpathians (Romania) using DC resistivity tomography. Carpathian Jurnal of Earth and Environmental Sciences 8, 113-122.

Onaca A.L., Urdea P., Ardelean A.C. (2013b) – Internal Structure and Permafrost Characteristics of the Rock Glaciers of Southern Carpathians (Romania) Assessed by Geoelectrical Soundings and Thermal Monitoring. Geografiska Annaler: Series A, Physical Geography 95, 249-266.

Otto J.C., Keuschnig M., Götz J., Marbach M., Schrott L. (2012) – Detection of mountain permafrost by combining highresolution surfaceandsubsurface information - an example from the Glatzbach catchment, Austrian Alps. Geografiska Annaler: Series A, Physical Geography 94, 43-57.

Popescu R., Vespremeanu-Stroe A., Onaca A., Cruceru N. (2015) – Permafrost research in the granitic massifs of Southern Carpathians (Parâng Mountains). Zeitschrift für Geomorphologie 59, 1-20.

Reuther A., Geiger C., Urdea P., Niller H.P., Heine K. (2004) – Determining the glacial equilibrium line altitude (ELA) for the Northern Retezat Mountains, Southern Carpathians and resulting paleoclimatic implications for the last glacial cycle. Analele Universităţii de Vest din Timişoara Seria geografie 14, 11-34.

Ribolini A., Fabre D. (2006) – Permafrost existence in rock glaciers of the Argentera Massif, Maritime Alps, Italy. Permafrost and Periglacial Processes 17, 49-63.

Riseborough D., Shiklomanov N., Etzelmüller B., Gruber S., Marchenko S. (2008) – Recent advances in permafrost modelling. Permafrost and Periglacial Processes 19, 137-156.

Scapozza C., Lambiel C., Gex P., Reynard E. (2011) – Prospection géophysique multi-méthodes du pergélisol alpin dans le sud des Alpes suisses. Géomorphologie, Relief, Processus, Enrivonnement, 1, 15-32.

Schrott L. (1994) – Die Solarstrahlung als steuernder Faktor im Geosystem der subtropischen semiariden Hochanden (Agua Negra, San Juan, Argentinien). Heidelberger Geographische Arbeiten 94, p. 199.

Schrott L. (1996) – High mountain permafrost and its relation to landform and solar radiation. A case study in the High Andes of San Juan, Argentina. The Erasmus 94-95 Programme in Geomorphology: Intensive Course Tyrol (Austria) and Student Mobility. Dipartimento di Scienze della Terra, Università degli Studi di Modena, M. Panizza, M. Soldati, D. Barani & M. Bertacchini (Eds.), 77-87.

Schrott L., Otto J.C., Keller F. (2012) – Moddeling alpine permafrost distribution in the Hohe Tauern region, Austria. Austrian Journal of Earth Sciences 105, 169-183.

Shapiro S.S., Francia R. (1972) – An approximate analysis of variance test for normality. Journal of the American Statistical Association 67, 215-216.

Szepesi A. (2007) – Masivul Iezer. Elemente de geografie fizică. Edit. Universitară, Bucureşti, p. 208.

Tanarro L.M., Hoelzle M., García A., Ramos M., Gruber S., Gómez A., Piquer M., Palacios D. (2001) – Permafrost distribution modelling in the mountains of the Mediterranean: Corral del Veleta, Sierra Nevada, Spain. Norsk Geografisk Tidsskrift 55, 253-260.

Török-Oance M. (2004) – Geographic Information Systems as a tool for permafrost investigation, a case study in the west part of the Southern Carpathians (Romania). Geomorphologia Slovaca  4, 68-71.

Urdea P. (1992) – Rock glaciers and periglacial phenomena in the Southern Carpathians. Permafrost and Periglacial Processes 3, 267-273.

Urdea P. (1993) – Permafrost and periglacial forms in the Romanian Carpathian. Proceedings of Sixth International Conference on Permafrost, Beijing, July 5-9, I, South China University of Technology Press, 631-637.

Urdea P. (1998) – Rock glaciers and permafrost reconstruction in the Southern Carpathians Mountains, Romania Permafrost - Seventh International Conference Proceedings, Yellowjnife, Canada, Collection Nordicana 57, Univ. Laval, 1063-1069.

Urdea P. (2000) – Munţii Retezat, Studiu geomorfologic. Edit. Academiei, Bucureşti.

Urdea P., Vuia, F. (2000) – Aspects of the periglacial relief in the Parâng Mountains. Revista de Geomorfologie 2, 35-39.

Vespremeanu-Stroe A., Urdea P., Popescu R., Vasile M. (2012) – Rock Glacier Activity in the Retezat Mountains, Southern Carpathians, Romania. Permafrost and Periglacial Processes 23, 127-137.

Vonder Mühll D., Hauck C., Gubler H. (2002) – Mapping of mountain permafrost using geophysical methods. Progress in Physical Geography 26, 643-660.

Haut de page

Annexe

Version française abrégée

Du fait des changements climatiques globaux, la dégradation accélérée du pergélisol a un effet significatif sur les écosystèmes alpins, pouvant notamment déstabiliser les versants de montagne (Haeberli and Gruber, 2009). Une meilleure connaissance de la répartition du pergélisol et de ses caractéristiques est importante pour évaluer les changements morphologiques et écologiques dans l’environnent alpin (Schrott et al., 2012). Depuis les années 1990, plusieurs modèles empirico-statistiques ou physiques (Etzelmüller et al., 2001 ; Riseborough et al., 2008) ont été proposés pour l’évaluation de la distribution spatiale du pergélisol alpin (Boeckli et al., 2012). Les valeurs BTS sont utilisées dans plusieurs recherches pour calibrer le modèle (Hoelzle, 1992 ; Keller, 1992 ; Gruber and Hoelzle, 2001 ; Lewkowicz and Ednie, 2004 ; Julián and Chueca, 2007), tandis que l’approche empirico-statistique souligne le rôle des différents facteurs topo-climatiques : la température de l’air, la radiation solaire, la pente, etc. (Etzelmüller et al., 2006). Depuis la première utilisation par Haeberli (1973) des mesures BTS pour la détection du pergélisol alpin dans les Alpes Suisses, la méthode est devenu largement utilisée dans les Alpes (Hoelzle, 1992 ; King et al., 1992 ; Hoelzle et al., 1993 ; Hoelzle, 1996 ; Gardaz, 1997 ; King and Kalisch, 1998 ; Imhof et al., 2000 ; Gruber and Hoelzle, 2001 ; Bodin, 2005) et d’autres chaines montagneuses (Dobinski, 1998 ; Ødegård et al., 1999 ; Kneisel et al., 2000). En Roumanie, la méthode BTS a été utilisée pour cartographier la distribution spatiale du pergélisol dans les Carpates du Sud (les monts Retezat et Parâng) et de l’Ouest (les monts Apuseni ; Urdea, 1993 ; Urdea, 2000 ; Urdea and Vuia, 2000 ; Vespremeanu‐Stroe et al., 2012 ; Onaca et al., 2013b ; Popescu et al., 2015).

Cette étude présente le premier modèle de répartition du pergélisol dans la partie centrale du massif de Retezat à l’aide de la méthode BTS et de la prospection géophysique : sondages et profils géoélectriques (ERT) et géoradar (GPR). La prospection géophysique a été utilisée pour valider le modèle final. La zone d’étude couvre une surface de 59 km2. Cette zone est caractérisée par une forte densité de formes du relief spécifiques du déplacement lent du pergélisol, comme les glaciers rocheux (Haeberli, 1985). L’altitude est comprise entre 1 505 m et 2 509 m, tandis que la limite forestière se trouve entre 1 650 et 1 750 m. La méthodologie proposée pour la modélisation de la distribution du pergélisol présente quatre étapes principales: a) récolter les mesures BTS, b) dériver les variables prédictives, c) modéliser les valeurs BTS et d) valider le modèle spatial. Un total de 170 valeurs BTS a été récolté pendant la saison d’hiver 2013-2014, basé sur la stratégie proposée par Onaca et al. (2013b). Un modèle numérique d'altitude (d'une résolution de 15 m) a été utilisé pour dériver les variables prédictives contrôlant l’occurrence du pergélisol dans les régions alpines : l’altitude, la radiation solaire, la végétation, la pente et la courbure du profil. La relation entre ces variables et les valeurs BTS a été analysée à l’aide d’une régression linéaire multiple. Puis, le modèle final a été validé par les analyses suivantes : (i) recouvrement entre le modèle de répartition du pergélisol et la distribution des glaciers rocheux dans la partie centrale du massif de Retezat (Urdea, 2000) ; (ii) comparaison entre le modèle de la distribution du pergélisol et les investigations géophysiques (profils ERT et GPR ; Onaca, 2013 ; Onaca et al., 2013b) ; (iii) comparaison entre le modèle de distribution du pergélisol et les valeurs BTS mesurées lors des campagnes précédentes et les températures de surface du sol (Urdea, 1992 ; Vespremeanu-Stroe et al., 2012 ; Onaca, 2013). Les mesures BTS ont été réalisées sur les glaciers rocheux Pietrele (47), Pietrele II (12), Pietricele (22), Valea Rea (15), Valea Rea III (5) et Ana (8) ou dans leurs environs (20). Les valeurs BTS mesurées varient entre -0,8°C et -7,3°C. En accord avec les classifications proposées par Haeberli (1973) et Hoelzle (1992), les valeurs inférieures à -3°C indiquent que la présence du pergélisol est probable, les valeurs entre -2 et -3°C indiquent que la présence du pergélisol est possible, tandis que les valeurs supérieurs à -2°C indiquent l’absence du pergélisol. Ainsi, dans la zone d’étude, 86 valeurs BTS (51 %) indiquent la présence probable du pergélisol, 36 valeurs (21 %) révèlent que la présence du pergélisol est possible et 48 valeurs (28 %) indiquent l’absence du pergélisol. Les relations statistiques entre les valeurs BTS et les variables prédictives ont été évaluées par une corrélation linéaire, avec un coefficient de corrélation de Pearson de 0,545 pour la radiation solaire, -0,478 pour l’altitude et 0,391 pour la végétation. La pente (-0,201) et la courbure du profil (0,036) ne présentent pas une corrélation significative avec la présence du pergélisol. La valeur de la régression linéaire multiple indique un R2= 0,48 qui représente la précision du modèle final. Cette valeur est considérée significative en raison d’autres recherches similaires (Ødegård et al., 1999 ; Tanarro et al., 2001 ; Gruber and Hoelzle, 2001 ; Lewkowicz and Ednie, 2004 ). L’équation utilisée est :
BTS = 1,477798 + 0,000014 * Radiation Solaire + 4,037341 * NDVI - 0,005711 * Altitude.

L’occurrence du pergélisol dans la zone étudiée est conforme avec la classification du pergélisol proposée par Haeberli, 1973 : le pergélisol probable (PPR), le pergélisol possible (PP) et l’absence de pergélisol (NP). Ainsi, le PPR couvre une surface de 17 km2, le PP une surface de 14 km2 et le NP une surface de 28 km². Les résultats confirment le recouvrement entre le modèle théorétique et la distribution des glaciers rocheux Pietrele, Pietricelele, Judele, Lower Ana, Upper Ana, Valea Rea 1, Valea Rea 2 et Știrbu. Ces résultats représentent le premier modèle à haute résolution spatiale de répartition du pergélisol dans la partie centrale du massif de Retezat.

Haut de page

Table des illustrations

Titre Fig. 1 – Location of the study area. Fig. 1 – Distribution des glaciers rocheux et des mesures BTS dans la zone étudié.
Légende 1: BTS measurements; 2: rock glaciers. 1 : mesures BTS ; 2 : glaciers rocheux.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-1.jpg
Fichier image/jpeg, 3,4M
Titre Fig. 2 – Data logger placement and classification examples of rock glaciers within the study area. Fig. 2 –Placement des capteurs autonomes de température et exemples de classification des glaciers rocheux dans la zone étudié. 
Légende A. 1: intact rock glacier; 2: data logger; B. 3: relict rock glacier. A. 1 : glaciers rocheux intact ; 2 : capteurs autonomes de température ; B. 3 : glaciers rocheux relictes.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-2.jpg
Fichier image/jpeg, 2,2M
Titre Fig. 3 – Spatial distribution of BTS data according to predictor variables. Fig. 3 – Distribution spatiale des valeurs BTS en accord avec les variables prédictives.
Légende A: BTS values (oC); B: snow thickness (cm); C: solar radiation (WH/m2); D: elevation range (m); E: NDVI values; F: slope angle (degrees); G: profile curvature values. A : valeurs BTS (oC) ; B : épaisseur de la neige (cm) ; C : radiation solaire (WH/m2) ; D : valeurs d’altitude (m) ; E : valeurs de NDVI ; F : angle de pente (degrés) ; G : valeurs de la courbure du profil.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-3.jpg
Fichier image/jpeg, 1,6M
Titre Fig. 4 – Boxplots showing the distribution and correlation coefficients between BTS and predictor variable data. Fig. 4 – Boîtes à moustaches indiquant la distribution et le coefficient de corrélation entre les valeurs BTS et les variables prédictives.
Légende A: BTS values (oC); B: solar radiation (WH/m2); C: elevation range (m); D: NDVI values; E: slope angle (degrees); F: profile curvature values. A : valeurs BTS (oC) ; B : radiation solaire (WH/m2) ; C : valeurs d’altitude (m) ; D : valeurs de NDVI ; E : angle de pente (degrés) ; F : valeurs de la courbure du profil.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-4.jpg
Fichier image/jpeg, 572k
Titre Fig. 5 – Predicted basal temperature of snow (BTS) cover (oC). Fig. 5 – Distribution des valeurs de température à la base de la neige (BTS), prédite par le modèle.
Légende A: intact rock glaciers; B: relict rock glaciers. A : glaciers rocheux intacts ; B : glaciers rocheux relictes.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-5.jpg
Fichier image/jpeg, 3,1M
Titre Fig. 6 – Predicted permafrost probability. Fig. 6 – Distribution du pergélisol dans la zone d'étude telle que prédite par le modèle.
Légende 1: probable permafrost; 2: possible permafrost; 3: no permafrost; 4: intact rock glaciers; 5: relict rock glaciers. 1 : pergélisol probable ; 2 : pergélisol possible ; 3 : absence de pergélisol ; 4 : glaciers rocheux intacts ; 5 : glaciers rocheux relictes.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-6.jpg
Fichier image/jpeg, 3,1M
Titre Fig. 7 – Synthesis of permafrost occurrence within the investigated area revealed by previous geophysical and thermal investigations. Fig. 7 – Synthèse d’occurrence du pergélisol dans la zone étudiée, révélée par des investigations géophysiques and thermiques antérieures.
Légende A: probable permafrost; B: possible permafrost; C: no permafrost; D: permafrost occurrence according to previous investigations; E: intact rock glaciers; F: relict rock glaciers. Previously investigated rock glaciers: 1: Pietrele; 2: Pietricelele; 3: Valea Rea; 4: Valea Rea III; 5: Ana; 6: Upper Ana; 7: Judele; 8: Știrbu. A : pergélisol probable ; B : pergélisol possible ; C : absence de pergélisol ; D : occurrence du pergélisol en accord des investigations précédentes ; E : glaciers rocheux intacts ; F : glaciers rocheux relictes. Glaciers rocheux investigués précédemment : Pietrele ; 2 : Pietricelele ; 3 : Valea Rea ; 4 : Valea Rea III ; 5 : Ana ; 6 : Ana supérieur ; 7 : Judele ; 8 : Știrbu.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-7.jpg
Fichier image/jpeg, 3,1M
Titre Fig. 8 – GST evolution in the 2012-2013 season at Judele site. Fig. 8 – Évolution de GST dans le site Judele, saison 2012-2013.
Légende A: time of first freezing; B: BTS field measurements; C: BTS interval; D: zero curtain. A : temps de première congélation ; B : mesures BTS sur le terrain ; C : intervalle BTS ; D : Phase zéro.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-8.png
Fichier image/png, 2,0M
Titre Fig. 9 – Comparison of permafrost distribution models for the central part of the Retezat Mountains. Fig. 9 – Comparaison des modèles de répartition du pergélisol dans la partie centrale du massif de Retezat.
Légende A: Permafrost distribution model based on the approach of Gruber (2012); B: Proposed permafrost distribution model; 1: fringe of uncertainty; 2: permafrost only in favourable conditions; 3: probable permafrost; 4: possible permafrost; 5: no permafrost. A : modèle de répartition du pergélisol basée sur la stratégie proposée par Gruber (2012) ; B : modèle de répartition du pergélisol proposé ;  1 : frange d’incertitude ; 2 : pergélisol à conditions le plus favorable ; 3 : pergélisol probable ; 4 : pergélisol possible ; 5 : absence de pergélisol.
URL http://journals.openedition.org/geomorphologie/docannexe/image/11131/img-9.jpg
Fichier image/jpeg, 2,5M
Haut de page

Pour citer cet article

Référence papier

Adrian C. Ardelean, Alexandru L. Onaca, Petru Urdea, Raul D. Șerban et Flavius Sîrbu, « A first estimate of permafrost distribution from BTS measurements in the Romanian Carpathians (Retezat Mountains) »Géomorphologie : relief, processus, environnement, vol. 21 – n° 4 | 2015, 297-312.

Référence électronique

Adrian C. Ardelean, Alexandru L. Onaca, Petru Urdea, Raul D. Șerban et Flavius Sîrbu, « A first estimate of permafrost distribution from BTS measurements in the Romanian Carpathians (Retezat Mountains) »Géomorphologie : relief, processus, environnement [En ligne], vol. 21 – n° 4 | 2015, mis en ligne le 23 novembre 2015, consulté le 28 mars 2024. URL : http://journals.openedition.org/geomorphologie/11131 ; DOI : https://doi.org/10.4000/geomorphologie.11131

Haut de page

Auteurs

Adrian C. Ardelean

West University of Timișoara, Department of Geography – Blvd. V. Pârvan, no. 4 – Timișoara, 300223 – Romania (adrian.ardelean86@e-uvt.ro).

Alexandru L. Onaca

West University of Timișoara, Department of Geography – Blvd. V. Pârvan, no. 4 – Timișoara, 300223 – Romania (alexandru.onaca@e-uvt.ro).

Petru Urdea

West University of Timișoara, Department of Geography – Blvd. V. Pârvan, no. 4 – Timișoara, 300223 – Romania (pentru.urdea@e-uvt.ro).

Raul D. Șerban

West University of Timișoara, Department of Geography – Blvd. V. Pârvan, no. 4 – Timișoara, 300223 – Romania (raul.serban88@e-uvt.ro).

Flavius Sîrbu

West University of Timișoara, Department of Geography – Blvd. V. Pârvan, no. 4 – Timișoara, 300223 – Romania (Flavius.sirbu@gmail.com).

Haut de page

Droits d’auteur

Le texte et les autres éléments (illustrations, fichiers annexes importés), sont « Tous droits réservés », sauf mention contraire.

Haut de page
Search OpenEdition Search

You will be redirected to OpenEdition Search