Study Areas and Geochemical Features

The three study areas are located in the middle alpine belt (here from 800 to 1205 m a.s.l.) of the North Norwegian Caledonides, which are the remnants of the Caledonian mountain range (Fig. 1). The sites are equally located along the coast/inland climatic gradient (Werner Johannesen, 1974; Aune, 1981), in the transition between the oceanic and continental vegetation sections (Moen, 1998).

Figure 1

Maps of Northern Europe and Troms County, with the location of the three municipalities where the study took place.

On the basis of mineralogical composition and chemical weathering rate, the bedrock of the study areas was classified into five categories: (1) Carbonate rocks (Carb), which include limestone and dolomite, usually metamorphosed into marble. Prevailing minerals are calcite and dolomite. (2) Calcareous mica schist (CMS). Calcite and/or dolomite are important minerals in this group as well, but mica schists are characterized by the micas muscovite and/or biotite in addition to quartz. Feldspars and chlorite occur in varying amounts. (3) Noncalcareous mica schist (N-CMS). This is similar to the preceding group, but lacking calcite. Less soluble calcium silicates may be present. (4) Mafic rocks (Mafic). This generally denotes rocks with a high content of magnesium and iron. Amphibolite and gabbro are widespread in the study area, often intruding the surrounding metasedimentary rocks. The main minerals are plagioclase feldspars, pyroxenes, and amphiboles. (5) Felsic rocks (Felsic). This is a common name for light colored rocks enriched with quartz. Feldspars (both plagioclase and potassium feldspar) and micas occur in varying amounts. Granite, gneiss, quartzite, and arkose are examples of felsic rocks, which often are hard and resistant to weathering and erosion.

Dry, exposed ridges were selected as study sites. A ridge is here defined as a well-drained convex landform of bedrock above the tree line, which accumulates a minimal amount of snow during winter. Dead plant debris is blown away from such sites, leaving a sparse and discontinuous mineral soil derived from the local bedrock as the substrate for vegetation. For the study sites to be as uniform as possible, they were selected on the following criteria: moderate cryoturbation, pronounced wind deflation, minimal admixture of allochthonous sediments such as ablation moraine, minimal thickness or nonexistence of a humic layer, no stagnant surface water, and horizontal or subhorizontal inclination. The bedrock categories and elevations of the selected sites were not correlated (Fig. 2).

Figure 2

Geological map of the study area in Storfjord, SW of Skibotndalen valley. The 14 study sites in this area are shown as black quadrates. The western study site at Mount Luhčavárri is situated on a lens of hornblende schist, too small to appear on the map. The equidistance is 20 m. For further information of adjacent areas and lithostratigraphy, see Zwaan (1988). Carb = carbonate rocks, CMS = calcareous mica schist, N-CMS = noncalcareous mica schist, Mafic = mafic rocks, and Felsic = felsic rocks.

Field Sampling and Laboratory Procedures

Vegetation and soil sampling was performed at 17 sites in three study areas: (1) Two sites on the mountain Gråhøgda (1200–1267 m a.s.l.), Målselv, with predominantly carbonates and calcareous mica schist. (2) Fourteen sites in the mountains between the valley of Kitdalen and the valley of Skibotndalen (830–1205 m a.s.l.), Storfjord. The local distribution of bedrock in Storfjord is drafted in Figure 2, supplemented by Zwaan (personal communication, 1996), Padget (1955), and our own field observations. (3) One site on the mountain Pihkahistama/Javroaivit (909 m a.s.l.), Nordreisa, with calcareous mica schist.

For floristic analysis, 85 sampling units were analyzed, 5 at each site. The units consisted of five 50 × 50 cm quadrates: a central one and one placed at each corner of the central one. In each of the quadrates, species frequency was recorded using the grid-frequency method, with a sampling frame subdivided into 16 squares. Saxicolous and crustose lichens were not recorded. Three samples of the bedrock-derived substrate were collected at each study site. The texture of the weathered material is partly caused by the physical properties of the parent bedrock. Marbles disintegrate into gravel formed by the original crystals of calcite or dolomite, while mica schists, especially the calcareous kind, often produce a more fine-grained material of mica and calcite. The material weathered from felsic and mafic rocks is often a product of frost shattering, containing a variety of grain sizes. More important, however, subsequent cryoturbation produces a considerable amount of fine material in the bedrock-derived soils, independent of bedrock category. The soil occurring at our study sites can be classified as leptosols according to the FAO soil classification system (FAO, 1998), and the thickness varied between 0 and 10 cm. It was sampled across its entire depth, where available. Pebbles were removed, along with loose plant litter. The samples were then put to dry at 30°C, stored in paper bags.

In addition to pH and LOI, the amounts of the elements calcium, magnesium, potassium, phosphorus, and nitrogen were measured. Nitrogen analysis was carried out using the Kjeldahl method (Farstad et al., 1985), which gives the total nitrogen content. The amount of the elements P, K, Mg, and Ca available to plants was estimated using the so called Acetate-Lactate (AL) method (based on the quantity of elements extractable with a solution of acetic acid [0.4 M] and ammonium lactate [0.1 M]). The method is described in Egner et al. (1960) and Houba et al. (1997). Two particle size fractions were separated using sieving, namely the fraction of fine particles, i.e. clay and silt (≤0.06 mm), and the fraction of coarser particles, i.e. sand and gravel (>0.06 mm). These particle size limits follow the standard of the U.S. Department of Agriculture (USDA) presented in Gee and Bauder (1986). From the two fractions a ratio was calculated reflecting the dominance of fine particles in the soil. This soil texture ratio (STR) is likely to reflect the moisture retention capability of the soil (Arya and Paris, 1981). For each site and soil variable, the averages of the values measured in the three soil samples were used in the analysis.

Numerical Analysis

In order to relate species diversity to the selected bedrock categories, the Shannon diversity index (H) was calculated for vascular plants, bryophytes, lichens, and for the entire data set, for each of the five bedrock categories. This commonly used index accounts for both abundance and evenness of the species present to express species diversity in a community:

1

To visualize the differences and similarities in the floristic composition of the different sites, we used nonmetric multidimensional scaling (NMDS; Kruskal, 1964) in combination with the Bray-Curtis dissimilarity index to quantify the ecological distance between sites (Bray and Curtis, 1957). NMDS is a robust form of unconstrained ordination which has the potential of revealing gradient structures in the data if the calculated similarity between plots reflects their position along the gradients. The Bray-Curtis dissimilarity index, standardized to unit maximum for each species and to unit sample total, has shown to be an index with this property (Faith et al., 1987). NMDS requires that the dimensionality of the data is specified in advance. As the aim of the NMDS here is to visualize the variation in the floristic data, we used NMDS in two dimensions. Namely, it can be presented as a plot where the pairwise distances between the sites are ranked similarly to the distances calculated from the floristic data. In other words, NMDS maximizes the correlation between rank order of the distances between sites in ordination spaces and the rank order of the dissimilarities calculated for the sites from the species data.

As a statistic of goodness of fit, the discrepancy between the distances in the ordination space and the dissimilarities calculated from the species data is quantified, and termed stress. Algorithms for NMDS rely on iterative processes, and therefore we repeated the procedure with multiple random start configurations aiming to find a constellation of the sites that produces the lowest stress. This constellation does not depend on the environmental data. To facilitate the interpretation of the NMDS, environmental variables were plotted as arrows on top of the NMDS plot. The direction of an arrow indicates the direction of the gradient and its length indicates the correlation with the NMDS plot. The significance of these correlations was quantified using permutations tests. These tests verify that the correlations are not found when the soil variables are randomly permuted.

To test the importance of the relationships between the measured bedrock-related soil variables and the floristic composition of the sites, we constrained the ordination using the soil data in a Canonical Correspondence Analysis (CCA). While the NMDS showed the major gradients in the floristic data, the CCA optimizes the fit of the sites relative to the environmental variables (ter Braak and Prentice, 1988). Rather than showing all the variation in the data, as with the NMDS, the CCA quantifies how much of the variation in the species data can be accounted for by the measured environmental gradients. The contributions of the different soil variables to the CCA model were inspected separately. This was done using permutation tests that quantified if the effect of a variable was still significant after the variation caused by the other variables had been removed (Legendre and Legendre, 1998).

Results from one outlier site on mafic rocks were omitted from the statistical analyses, as soil conditions appeared to be locally influenced by an adjacent marble outcrop. The software package R with the MASS (Venables and Ripley, 2002) and vegan (Oksanen, 2005) packages, were used for the statistical analysis and modeling.

Soil Conditions

pH was higher in soil derived from carbonates (mean pH 7.4) and calcareous mica schist (mean pH 6.6) than in soil from the noncalcareous bedrocks (Table 1). Felsic soil had the lowest pH with a mean of 4.9.

Table 1

Descriptive statistics for soil parameters in the weathered material of different bedrock types. Carb = carbonate rocks, CMS = calcareous mica schist, N-CMS = noncalcareous mica schist, Mafic = mafic rocks, and Felsic = felsic rocks, LOI = loss on ignition, STR = soil texture ration (% clay and silt) / (% sand and gravel).

Calcium and magnesium are the main cations of carbonate minerals. They showed very high values in soils from the carbonate-containing bedrock types. In soil from carbonate rocks, the mean value of Ca was 4.3 × 10⁴ parts per million (ppm), and that of Mg was 948 ppm (one of the marbles was dolomitic). The average Ca and Mg values in soil derived from calcareous mica schist were 7448 and 354 ppm, respectively. In comparison, average Ca and Mg values of felsic soils were 51 ppm and 2 ppm, respectively. For the statistical analysis, the concentrations of Ca and Mg were transformed using the natural logarithm.

The phosphorus content was generally low. In soil from carbonate rock, all mica schists, and felsic rocks, none of the mean values exceeded 3.5 ppm. However, the values in soil derived from mafic rocks were higher, with a mean of 6.9 ppm and maximum, 11.3 ppm. Mean potassium values were highest in soil from calcareous and noncalcareous mica schists; values being 35 and 31 ppm, respectively. Both soil from carbonates, mafic and felsic rocks had low K values with means between 11 and 16 ppm. The organic admixture in the soil samples, as quantified by nitrogen content and LOI, hardly differed between soil samples of different bedrock origin. Ranges of mean total nitrogen content and LOI were 0.1–0.2% and 2.3–6.8%, respectively (Table 1).

Plant Diversity

Dry, exposed ridges in the middle-alpine belt have a low cover of vascular plants (5–30%). The overall species richness of vascular plant, bryophyte, and lichen species is low to moderate at most of the analyzed sites, varying from 16 to 52 species.

One general pattern of carbonate ridges is that the diversity of vascular plants, bryophytes, and lichens is low (Fig. 3, Table 2). Nevertheless, on felsic and mafic substrates the diversity of vascular plants is even lower than on carbonate rocks. The diversity of vascular plants is considerably higher on calcareous mica schist, and somewhat higher on noncalcareous mica schist, than on the other substrates. The diversity of bryophytes shows little variation except being slightly lower on carbonates than on all other bedrock categories. Lichen diversity is higher than the diversity of vascular plants and bryophytes on all bedrock categories, and on mafic rocks in particular. The exception is on calcareous mica schist, where the diversity of vascular plants is highest.

Figure 3

Shannon diversity indices showing the diversity of vascular plants, bryophytes, and lichens as well as the total diversity on the five bedrock categories. The whiskers indicate minimum and maximum values.

Table 2

Mean frequency of the recorded species on the different bedrock types. The species could have a frequency score ranging between 0 and 16. Carb = carbonate rocks, CMS = calcareous mica schist, N-CMS = noncalcareous mica schist, Mafic = mafic rocks, and Felsic = felsic rocks.

Soil–plant Relations

In most cases, the vegetation of plots on the same bedrock types tended to be more similar than on different bedrock types, as indicated by the NMDS in two dimensions (stress = 5.7%; Fig. 4). pH, P, ln(Ca), and ln(Mg) showed a significant correlation with the NMDS ordination configuration (p_pH < 0.001, p_P = 0.033, and p_ln(Ca) < 0.001, p_ln(Mg) = 0.004). There was weak statistical evidence that N and K were correlated with the NMDS configuration (p_N = 0.079 and p_K = 0.075), while LOI, altitude, and soil texture ratio did not appear correlated with it (p_LOI = 0.12, p_alt = 0.81, and p_STR = 0.259, respectively). Since ln(Ca) and ln(Mg) were highly correlated with pH (R² = 0.95 and R² = 0.84, respectively), however, only pH was included in the CCA.

Figure 4

Nonmetric multidimensional scaling of the sites based on the floristic data and the Bray Curtis dissimilarity to quantify the ecological distance between the sites. Environmental variables were overlaid on the result as arrows. The length of each arrow indicates the strength of the correlation between the variable and the nonmetric multidimensional scaling (NMDS), and its direction indicates the orientation of the gradient. Variables that were significantly correlated with the NMDS constellation are underlined (confidence level = 0.05). Carb = carbonate rocks, CMS = calcareous mica schist, N-CMS = noncalcareous mica schist, Mafic = mafic rocks, and Felsic = felsic rocks, LOI = loss on ignition, STR = soil texture ration (% clay and silt / % sand and gravel).

Three environmental variables quantifying bedrock related soil chemistry were included in the CCA model, namely phosphorus and potassium concentrations as well as pH (Fig. 5). The first CCA axis was strongly related to pH, while the second and third axes were most correlated with phosphorus and potassium, respectively (Table 3). It was clear that pH and P were the variables that contributed most to the model (pseudo-F_pH = 3.4, p_pH < 0.005, and pseudo-F_P = 2.3, p_P = 0.005, respectively), while the significance of the contribution of K was less clear (pseudo-F = 1.9, p = 0.04). We tested conditioning the model using the environmental variables altitude, N, and soil texture ratio, which are not or indirectly related to the chemical composition of the bedrock. Including the conditions did not reduce the constrained inertia by more than 7% and permutation tests showed that they did not improve the model significantly (p < 0.005 in the three cases). This indicates that the effects attributed to pH and phosphorus are highly unlikely to have been due to variation in N, altitude, or soil texture.

Figure 5

Canonical correspondence analysis plot, showing the sites and the four variables related to soil chemistry. Carb = carbonate rocks, CMS = calcareous mica schist, N-CMS = noncalcareous mica schist, Mafic = mafic rocks, and Felsic = felsic rocks.

Table 3

Canonical correspondence analysis of the species data using selected edaphic variables. Scores of the different environmental variables along the first three axes of the analysis, pseudo-F values and associated p-values produced by a permutation test comparing the complete model with a model excluding the variable in question.

Phosphorus Availability

The availability of phosphate may be a limiting factor in arctic and alpine communities due to slow decomposition rates (Wielgolaski, 1975; Wielgolaski et al., 1981; Crawford, 1989; Shaver and Kummerow, 1992; Hobbie et al., 2005). Ultimately the element phosphorus is derived from the mineral apatite, but in most systems the phosphorus available to plants is recycled from decomposing litter. At exposed mountain ridges this may not be the case, as most dead litter is removed. This suggests that the local availability of phosphorus could be lower than elsewhere in alpine systems. Our data confirm this assumption; the sites were unusually poor in phosphorus with a mean value of 2.7 ppm. In comparison, Haselwandter et al. (1983) found a mean plant-available phosphorus concentration in soils in the nival zone of the Alps of 14.6 ppm, and Walker (1985) found a mean value of 8.0 ppm in different Alaskan tundra habitats. It needs to be noted however, that some of this difference could be due to different extraction methods (Otabbong et al., 2004). Furthermore, the variation in the small amounts of available phosphorus was correlated neither with nitrogen content (r = −0.16) nor with LOI (r = −0.17), confirming that decomposing organic matter was not the source of phosphate compounds at the study sites.

The only other available source of phosphorus is the bedrock. The cycle of phosphorus is slow and can be considered as unidirectional from the weathering of rocks to sediments and sea deposits (Holtan et al., 1988). The phosphate mineral apatite occurs in igneous and metamorphic rocks (Walker and Syers, 1976; Holtan et al., 1988). Thus, it is plausible that mafic rocks and certain other silicate rocks leach more phosphate than other local bedrocks. As the availability of phosphorus is inhibited by high soil pH (Holtan et al., 1988; Tyler, 1992), this topic can be more readily observed at the noncalcareous sites. Among the sites on mafic rocks, felsic rocks, and noncalcareous mica schist there seemed to be a varying availability of phosphorus that was not correlated with soil pH (Fig. 5). The mean values of P in soil derived of mafic rocks were higher than elsewhere. Both mafic sites and one felsic site had high scores on the phosphorus axis. The remaining noncalcareous sites with comparable pH levels showed lower scores on the phosphorus axis in the CCA (Fig. 5). This suggests that bedrock derived phosphorus in certain noncalcareous soils could be significant to vegetation when there is limited availability of phosphorus from organic matter. What seems to characterize the wind-exposed ridges, where most phosphorus originates from bedrock, is a higher diversity and abundance of lichens (Fig. 3), as well as a higher abundance of certain vascular plant species (Table 2). Further studies, however, should address to what extent the vegetation itself, and the lichens in particular, facilitate the weathering of the minerals and subsequent release of phosphorus (Wilson, 1995). This facilitation could occur, for example, through the release of organic acids.

The Role of Ph And Calcium

pH was the measured soil parameter proving to be most important to the floristic composition. Calcium, which was abundant in carbonate-containing substrates, was highly correlated with pH, which is a commonplace observation. Meanwhile, there seemed to be a difference between the gravelly marble soils producing basic soil conditions, and the more composite soil derived from calcareous mica schist, producing a circumneutral pH. The soils of carbonate rocks had extremely high values of calcium (Table 1), sometimes in combination with high values of magnesium, due to dolomitic admixture. Available calcium and magnesium levels in soil weathered from calcareous mica schist were also considerable, but with few exceptions lower than those in soil weathered from carbonates.

Phosphorus and potassium concentrations in the latter soils were generally low. The high amount of carbonate/bicarbonate ions produces a high pH, which may be of importance to the root absorption of nutrients (Tyler, 1992). Especially, the absorption of phosphate is inhibited by basic soil conditions (Holtan et al., 1988). On the other hand, several alpine species growing on carbonate rocks are referred to as “lime-demanding” (Gjærevoll, 1990). This is probably an inadequate expression, as no plant species has the extreme calcium demand corresponding to the amount of Ca ions available in carbonate soils (Table 1). It would be more suitable to refer to these species as carbonate specialists, since they appear adapted to the actual nutrient stress at such sites.

The impression in the field was that middle-alpine marble ridges had a low species diversity and supported plant individuals with a poor vegetative development. Ridges of calcareous mica schist had a higher diversity with well-developed specimens, e.g. Astragalus alpinus and Rhododendron lapponicum, which is typically absent on ridges of carbonate rocks. This was confirmed by the analyses of floristic data. The Shannon diversity index was low at carbonate sites and higher at calcareous mica schist sites, especially for vascular plants. In the CCA plot, sites on carbonate rocks and one site on calcareous mica schist occurred in the upper left corner (Fig. 5). These are species-poor sites occupied by carbonate specialists. The species-rich sites on calcareous mica schist occurred as a distinct group lower and more centrally in the CCA plot. In known classifications of alpine and arctic areas, vegetation of exposed ridges on calcareous substrate is treated as one vegetation unit (often denoted Caricion nardinae or Kobresieto–Dryadion, in the classic phytosociological system). Our findings suggest, however, that the vegetation on ridges of carbonate rocks with basic soil pH is different from the vegetation on ridges of calcareous mica schists. Vegetation growing directly in soil originating from weathering of carbonate rocks is not rare in middle- and high-alpine environments or in the High Arctic (Scholander, 1934, Rønning, 1965, Edlund and Alt, 1989), and should probably be treated as a separate vegetation unit.

Potassium and Magnesium

The CCA indicated a relationship between floristic composition and the availability of potassium, although this relationship was less clear than that with pH and phosphorous. It is possible that a survey on a larger scale including more sites can throw more light on the relationship between flora and potassium availability.

Soil weathered from mafic rocks, which is the most magnesium rich bedrock category included in this study, had low magnesium values, indicating that the large amount of magnesium bonded in the dark minerals (pyroxenes and amphiboles) is not subject to extensive chemical weathering. Another interesting topic, which was not treated here, is the difference between dolomitic and calcitic carbonates. Both minerals are easily weathered, but calcite is less resistant than dolomite. On the other hand, dolomite carries, per definition, a larger proportion of magnesium carbonate in its mineral structure. This accounts for a great input of Mg to the soil. To assess the importance of Mg in different carbonate soils, it would be necessary to perform a comparative study, including several sites on dolomitic as well as calcitic carbonates.