STUDY SITE

Our investigation refers to the late-melting snowpatch called the Map of the Republic, situated in the Giant Mountains (Krkonose in Czech, Karkonosze in Polish), the topmost range of the Sudetes, a chain of middle mountains along the border between the Czech Republic and Poland (Fig. 1). In spite of their moderate altitude (the highest peak being Mount Snezka, 1603 m a.s.l.), the Giant Mountains display two regions with a continuous alpine timberline, large arctic-alpine ecosystems, numerous avalanche tracks, and occasional landslides (Jenik, 1961; Soukupova et al., 1995; Stursa, 1998). On the summit plateaus, average annual temperatures range between 1 and 2°C, and average annual precipitation reaches 1450 mm. Continuous snow cover persists from November to the beginning of May, with a maximum snow depth observed in March or at the beginning of April. Mean snowpack thickness reaches values of about 1.8 m (Vrbatova Bouda, meteorological station at 1400 m a.s.l., period 1963–1975).

The studied snowpatch, Map of the Republic, is situated in a nivation niche on a south-facing slope beneath the Modré Saddle (1499 m a.s.l.) between Mount Studnicni (1554 m a.s.l.) and Mount Lucni (1555 m a.s.l.). The prevailing bedrock is mica schist, and quartzite debris at the eastern edge of the area; soils are a mosaic of shallow leptosols and lithosols. The inclination of the slope surface ranges between 20° at the upper level and 37° at the bottom level of the area typically covered by snow in summer. The first written records of this snowbed appeared in the 18th century; occasional observations have been published ever since (Jenik, 1958). In most years the last snow disappears at the end of July or during the first weeks of August. The snowbed is situated mainly in the lee of the northern and northwestern winds affecting the snow deposition (a) during particular snowfall events and (b) during long-term wind loading by snow drifted from the neighboring summit etchplain. Exceptional amounts of snow and their duration on the leeward site of the Map of the Republic were described by several authors (Jenik, 1958, 1961; Vrba, 1964; Spusta et al., 2003a, 2003b).

Regular records of the snow depths and data of disappearance of the continuous snow cover are available due to measurements near Lucni Bouda chalet in the center of the etchplain; these data provide a suitable comparison with observations in the studied snowbed. In the area around the Map of the Republic, snow starts melting out usually at the end of April, and the snow persists up to the end of July or the first week of August; old observations suggest the ultimate melting date to fall in September (Partsch in Jenik, 1958); Vrba (1964) and Hejcman and Dvorak (2001) recorded the snow disappearance at the end of July. The typical area of the Map of the Republic, usually appearing in June, covers an area of about 10 ha. Jenik (1958) described the Map of the Republic as a starting zone of periodic snow avalanches which induces local soil disturbance and creates conspicuous insertion of dwarf pine (Pinus mugo) and distinct herbaceous vegetation into surrounding spruce forests. Vrba (1964) recorded considerable snow creeping, which was probably responsible for local landslides and soil surface disturbances in the locality.

SNOW COVER MEASUREMENT

The kinematic carrier phase–based GPS measurements with TRIMBLE Pathfinder ProXR and Pathfinder Power receivers were conducted in periods of the highest snow accumulation, i.e., in April 2000, May and June 2000, May 2001, February and May 2002, March and April 2003, and April 2004. For a greater part of the snowbed with potential high snow accumulation, relevant data were collected by parallel line measurements 2 m apart (collection of line data by walking or slow horizontal skiing or by a rope-driven sledge with a GPS receiver across the snowbed), combined with a “stop and go method” (walking followed by a stop for measurement, collection of point data; Fig. 2) applied in April 2000. Complete data for the whole snowpatch were collected in a regular network by the “stop and go method” in a later period of spring when the snow cover was safe enough for surveying. Subsequent construction of the snow surface model was performed for data sets belonging to each year. Data for the construction of the snow-free ground surface model were collected after the snow melting in mid-August 2000. We used the line measurements in a plot 400 m × 250 m in size, which covered the entire snowbed area. For the creation of digital elevation models (DEMs), the TOPOGRID command in ArcGis 8.3 (ArcInfo) was used, which generated a grid of elevations from 3-D point, line, and polygon data. This interpolation method, specifically designed for the creation of a hydrologically correct DEM, was based on the ANUDEM program developed by Hutchinson (1988, 1989) for hydrological research. All parameters in this tool were used in default. These two digital models (snow surface model and snow-free ground surface model) were both overlaid and after a simple arithmetical operation in the GRID extension of ArcInfo, the snow depths for each relevé were obtained and snow maps constructed accordingly.

MELTING RATE

The snow melting rate was recorded every year at weekly intervals in two ways. (1) In four transects, situated downward across the whole snowbed, we measured the distance of the snow-free area between the upper and bottom edge of the snowbed; wooden poles were used to indicate the margin of the snow cover each week; the GPS was used for the poles' position after complete snow recession. (2) Alteration of the outline of the melting snowbed was recorded with the GPS while walking around the snowbed. Both methods were applied every spring, starting in April 2000 and finishing after the snow disappearance in 2003.

PLANT SPECIES COMPOSITION

Phytosociological relevés were recorded along the same four transects as where the snow melt gradients were measured. Plant cover was estimated in percentages in 2 m × 2 m square plots in September 2000. Altogether, 111 relevés were recorded in the investigated area. Nomenclature of vascular plant species follows Kubat et al. (2002).

SOIL CHARACTERISTICS

The depths of the soil profile and the litter horizon were recorded using a soil probe in all 111 plots. The soil skeleton ratio (gravel and stones) was estimated from excavated material in the soil probes according to weight share. Part of the investigated area becoming snow-free relatively early was covered by Quaternary quartzite debris. The rest of the study site with leptosols and lithosols was strongly affected by processes of nivation (Fig. 3).

DATA ANALYSIS

To evaluate multivariate vegetation data and their relation to explanatory (environmental in Canoco terminology) variables, we used Canonical Correspondence Analysis (CCA) followed by the Monte Carlo permutation test. We used the unimodal method because of the great length of the environmental gradient; all environmental data, with the exception of the relevé position only, were of continuous character. Vegetation data variability explained by a single environmental variable was used as its explanatory power measure. We used restricted permutations and a cyclic shift of samples because the data were collected in transects. Altogether 499 permutations were performed and the results of these analyses were illustrated in the form of ordination diagrams in the Canodraw program. All multivariate analyses were performed in the Canoco for Windows (ver. 4.5) program (ter Braak and Šmilauer, 1998). Spearman correlation and regression analyses were performed in the STATISTICA (ver. 6.0) program.

SNOW COVER MEASUREMENTS

In the Map of the Republic snowbed, depths of the snowpack and date of snow melting were highly positively correlated. The obtained regression coefficient was 0.97 and the relationship was significant (P > 0.001; Fig. 4). In the center of the snowbed, the snow depth was 15.7, 6.1, 13.4, 7.6, and 14.2 m in 2000, 2001, 2002, 2003, and 2004, respectively. The depth of the snowpack depended on the winter weather, as seen in the comparison of maximum snow depths in the Map of the Republic and respective maxima at Lucni Bouda chalet (Table 1). At both localities the deepest snowpack was recorded in 2000, while the minimum depth occurred in 2001 and 2003. Despite the high yearly variability of the snow depth, the pattern of snow accumulation remained relatively stabilized, as shown in Figure 5. Maximum snow depth was recorded in the first week of April 2000, but the snow cover model in Figure 5 is from May 2000, when the snowpack was relatively shallow. Obviously, a part of the variability of the snowpack depth within individual years was caused by the date of actual snow measurements, but unfavorable weather prevented us from eliminating this factor. The depth of accumulated snow depended not only on the total amount of snow precipitation in particular winters, but also on wind direction and wind speed during and after the snowfalls. Northern and northwestern winds were favorable for snow accumulation at this locality.

The effects of snow cover on the cover of vascular plants was significantly negative (P = 0.014; R² = −0.296; Fig. 4). Mean vegetation cover in the early melting sites was 85%, with a high variability caused by the substrate, while only 17% cover was found in the center of the snowbed loaded with the maximum amount of snow. The low degree of vegetation cover in some early melting sites was caused particularly by the presence of quartzite debris fields of Quaternary origin.

VEGETATION STRUCTURE

Twenty-seven vascular plant species were recorded in our 111 phytosociological relevés covering the area of the Map of the Republic. According to the dominant species, we determined several vegetation units. The Calluna vulgaris unit (1) was present in the marginal zone of early melting edges situated on stony debris of Quaternary origin. The Molinia caerulea unit (2) was represented by grassland patches distributed below the zone of quartzite debris; this grass species was present only in a zone with a thick layer of rough humus emerging relatively early from the snowpack; Polygonum bistorta grew jointly with M. caerulea; this unit was predominately present above the snowbed in more favorable moisture conditions caused by water drained from the debris fields. The Nardus stricta unit (3) was a type of grassland tolerating a relatively high snow accumulation and a short snow-free period; a small patch dominated by this grass was developed above the central part of the snowbed, in a position lacking dense vegetation. This unit evidently remains unaffected by disturbances of creeping snow and landslides (Fig. 3); Avenella flexuosa, Anthoxanthum alpinum, Carex bigelowii, and Solidago virgaurea subsp. alpestris with a low cover degree were associated with Nardus stricta. The Avenella flexuosa–Deschampsia cespitosa unit (4) included grasses growing on the edges of the scarcely vegetated central part of the nivation niche; optimum growth conditions of Deschampsia cespitosa occurred at the bottom of the snowbed, i.e., in a zone of more favorable moisture conditions caused by water running from the melting snow. The Gnaphalium supinum unit (5) was represented by this single forb species and occurred in the central part of the snowbed or in the area where the upper soil horizons were destroyed by landslides; the soil skeleton amounted to more than 90%, and fine earth was missing. The Vaccinium vitis-idaea–V. uliginosum unit (6) occurred on the debris emerging relatively early from the snow above the upper margin of the snowbed where a fermented litter horizon was only found; Vaccinium myrtillus formed a belt of only a few meters broad, at the bottom margin of the snowbed where less than 9–10 m of snow was recorded.

In CCA, all investigated environmental variables explained together 23.1% of the species data variability (Table 2; An. 1). Results of this analysis are illustrated in the form of an ordination diagram (Fig. 6). Depth of the litter horizons explains 10.5%, and soil skeleton present in the soil profile reflects 10.2% of the species data variability.

Snow depth explained 3.8% and the snow melting rate 3.7% of the species data variability only (Table 2; An. 5 and 6). In the ordination diagram, the relationship of plant species to environmental variables is visible: Vaccinium vitis-idaea, V. uliginosum, Calluna vulgaris, Trientalis europaea, and Hieracium alpinum agg. were species more common above the central part of the snowbed, where the amount of soil skeleton was high and where an early date of melting, shallow snowpack, and shallow soil profile were recorded. Calamagrostis villosa differs from this group by preferring deeper soils with a smaller share of soil skeleton. In Figure 6, the overlap of the arrow for depth of litter horizons (H) and triangle for Molinia caerulea indicates occurrence of this grass in places with favorable moisture and geomorphological conditions for litter accumulation. Galium saxatile and Gentiana asclepiadea were species more common below the snowbed (Fig. 6), where relatively deep soils developed without obvious disturbances of ground surface. Juncus filiformis grew below snow accumulation only where waterlogging during snow melting was present. Viola palustris occurred on the banks of small periodical water flows below the snowbed where fine soil particles were washed out.

SNOW COVER MEASUREMENTS

Detailed measurements of the snow cover in the area of the Map of the Republic had never been performed, and comparable data are missing even in other regions. Earlier attempts were not successful, e.g., those undertaken with wire probes or a fixed steel pole by Messner (in Vrba, 1964). According to the experienced team of the Mountain Rescue Service in the Giant Mountains, collapsible avalanche probes serve as a reliable tool for snow measurements down to 3 m depth only. As witnessed by Billings and Bliss (1959), Stursa et al. (1973), or Klimesova (1993), wire probes can be successful in measurements of a relatively shallow snowpack. In a deep snowpack, a rapid increase of the estimation error must be expected due to the probe's flexibility and deviation from the investigated direction. For that reason we verified our GPS measurements by comparison with a probe down to 3 m, a snow depth found at the edge of the studied snowbed only. We consider the depths obtained from our model as “fairly precise” because differences between measurements performed by a wire probe and those recorded with our new method were less than 10 cm. Verification of the snow depths in central parts of the snowbed was impossible, but realistically no decrease of precision with increasing snow depths could be expected. Indirect support for the feasibility of our measurements is the repeated yearly coincidence of highest snow thickness and place of latest melting. The human factor is evidently more decisive for the model's reliability than the centimeter accuracy level of the kinematic carrier phase–based GPS measurements; we consider the insufficient density of collected data in the dissected ground surface and changes in antenna height due to movement difficulties in the mountain landscape to be human-induced factors (Fig. 2). The advantages of our new method are (1) computation of the snow depth without decrease of precision in thick snow accumulations, (2) recording of snow depth for each point of an area, and their subsequent attribution to each vegetation relevé, and (3) quantification of snow volume and therefore water storage in the investigated snowbed.

Accessibility of satellite signals was the only limitation in our procedure. If this method is applied in a narrow mountain valley surrounded by steep slopes, data collection must in many cases be limited to a time of day when a sufficient number of satellites is available for high precision measurements. According to our experience, GPS data collection can in many cases be performed without walking or skiing on dangerous slopes; in several cases we fixed the GPS receiver on a sledge driven by a rope from a safe site above the snow-patch where no avalanche hazard exists.

The majority of snowbed studies were based only on the melting rate or length of the snow-free period, irrespective of actual snow depth measurements (Stanton et al., 1994; Kudo and Ito, 1992; Tomaselli et al., 2003). A high correlation of melting rate and snow depth, found in our study, indicates that both these environmental variables gave similar results if used in an analysis of vegetation structure as explanatory variables.

A surprising feature of the Map of the Republic is the relatively high durability and stability of snow accumulation on the south-facing slope. In spite of occasional slab avalanches with a highly destructive effect recorded by Jenik (1958), these events are not frequent or regular. Spusta et al. (2003a, 2003b) recorded one avalanche moving on the ground and nine slab avalanches there within a period of 42 years from winter season 1961/1962 to 2002/2003. Most avalanches were caused by a new snow fracture during or following several days after heavy snowfall attended by strong northern or northwestern winds. Accumulation of snow from east and west is prevented by the presence of neighboring mountains ridges. Wind ripples and rills directed north–south in the snow were a direct evidence of accumulation due to snow transport from the northern etchplain. These ripples and rills, often more than 0.5 m deep, were present above the investigated snowbed. It is obvious that the thickness of snow accumulation is a result of both regular snowfall and successive wind loading. The last big avalanche in 1985 resulted from a snow storm accompanied by a NNW wind blowing at a speed of 140 km h⁻¹ on the day preceding the event; the crown fracture of the avalanche was 180 m long and 4 m high; snow density was 440 kg m⁻³ (Spusta and Kocianova, 1998).

VEGETATION STRUCTURE

We have determined a negative relationship between the cover of vascular plants and snowpack thickness. The negative relationship is in accordance with findings of Galen and Stanton (1999). The low regression coefficient indicates a high variability in the data and relative weakness of the relationship in the Map of the Republic. In places with a shallow snow cover, total plant cover was lower than in places with average snow depths (Fig. 4). This was predominately caused by the substrate, not by a low protective effect of the snowpack. Under a shallow snowpack, the recorded high variability of total vegetation cover degree was an undisputable proof of the variability of the substrate. Low plant cover degree was revealed on the edge of Quaternary debris fields. On the other hand, a high total cover of plants was recorded in places with a deep soil profile. Also, the depth of snow cover on the neighboring etchplain is around 2 m, and a dense plant cover is developed there. The threshold limit for the decrease of herb layer cover are snowpack depths of 1 m, providing a low protective effect and enabling plant communities of wind-exposed stands to occur there (Jenik, 1961).

The main factors causing the variability in the cover degree of vascular plants was a high soil diversity across the snowpatch, caused mainly by nivation processes and snow creeping generated by thick snow layers. Landslides disturbed the soil surface in a range of areas differing in snow depth, namely in the upper part of the snowbed. Gnaphalium supinum was strictly restricted to these sites, with a low participation of the other species. Surprisingly, Salix herbacea, a typical snowbed willow in the Central European mountains, does not grow in the study area despite its scattered occurrence in the surrounding area; unlike Gnaphalium supinum, the smallest willow evidently prefers snowbeds with a relatively stable soil surface. The preference of Salix herbacea for “stable nivation niches” is in agreement with a study performed in the Apennines by Tomaselli (1991).

The share of plant species variability explained by melting rate or by snowpack depth gradient was surprisingly small, in contrast to the high variability explained by soil parameters. Evidently, one explanation could be a high variability of the soil, and a high degree of soil disturbance created by solifluction, snow creeping, and especially by snow and firn erosion. The low explanatory power of the melting rate is in sharp discrepancy with findings of Stanton et al. (1994). The snowpack thickness gradient is obviously decisive for vegetation development in snowbeds with a quiet, stabilized ground surface, which is not the case in the Map of the Republic with high dynamic nivation processes on a sloping ground surface. We should not forget, however, that the majority of soil variability in the snowbed was caused by nivation; to divide strictly the effect of snow and soil parameters is simply impossible, and all environmental parameters determining plant growth must be evaluated jointly. Another factor probably responsible for the weakness of the snow gradient as a predictor of plant responses was the absence of specialized snowbed species in our study, in comparison with other studies performed in extensive tundra or alpine environments (see Hirano and Kudo, 2004; Komarkova, 1993; Sandvik and Heegaard, 2003; Sandvik et al., 2004; Vanderpuye et al., 2002). In our locality, changes in plant composition in relation to the snowpack gradient were thus in many cases just quantitative, not qualitative.

Presence of Agrostis rupestris was recorded in the central part of the snowpatch Map of the Republic, yet this species was excluded from our analyses because of its low abundance in the transects. A stand dominated by Nardus stricta was restricted to the area of medium melting date, on the ground surface obviously lacking any mechanical disturbance. Nardus stricta is evidently able to cope with a deep snowpack and short growing season, but it is unable to regenerate under frequent disturbances, in contrast to the grasses Avenella flexuosa and Deschampsia cespitosa. In similar conditions, Nardus stricta is a frequent snowbed species also in other European mountains, from the Mediterranean to Scandinavia (Jenik, 1961; Krahulec 1985; Palacios et al., 2003; Thebaud and Etlicher, 1997; Tomaselli, 1991).

A similar case of Nardus-dominated patches and the distribution of other vegetation patches including Molinia caerulea, Gnaphalium supinum, Deschampsia cespitosa, and Avenella flexuosa in relation to snow cover were first described by Jenik (1958). He used a different method in data collection and therefore his chosen statistical approach was inappropriate to compare with ours. In spite of this, it is evident that after nearly half a century the vegetation structure of the Map of the Republic snowbed has remained stabilized. The vegetation pattern and soil disturbance are evidently long-term features as confirmed by similar vegetation and substrate patterns recorded in the area in 1954 and 2003.