Elevation gradient of successful plant traits for colonizing alpine summits under climate change

A total of 120 summits in south-eastern Switzerland (figure 1) were revisited to record changes in floristic composition over the past century based on historical species lists (on average 102 years old). The summits are distributed across bedrock types (46 on limestone and 74 on siliceous rock) and across elevations between 2449 and 3418 m asl.

Zoom In Zoom Out Reset image size

The historical data were judged to be highly reliable according to Stöckli et al (2011). They had been collected by qualified botanists to assemble complete species lists in a spatially restricted area, and thus can be regarded as presence/absence data. As in the earlier surveys, during the summers 2010 and 2011, all vascular plants within the uppermost 10 m were recorded without previous knowledge of the old data (Stöckli et al 2011; Wipf et al). 48 of the summits were simultaneously visited by two botanists. They worked completely independently and the longest species list was retained, considering that historical inventories were realized by highly skilled botanists. To avoid overestimating species turnover, doubtful identifications of difficult taxa (e.g. Festuca sp.) were excluded unless confirmed by the previous record, and pairs of similar species that differed between surveys were merged. The species nomenclature complies with Aeschimann et al (2004).

A total of twenty plant traits in three trait categories (species ecology, morphology and dispersal; table 1) were selected for their potential influence on whether a species colonized, established or became extinct. Traits related to ecology consisted of eight ecological indicator values (Landolt et al 2010), reflecting a species' requirements for climate (temperature, continentality, light) and soil conditions (moisture, pH, nutrients, humus content, aeration), as well as its sociological group (table 1). Morphological traits were the species life form, type of clonal growth, maximum height and plant strategy (competitive—stress tolerant—ruderal). Traits related to the dispersal ability of the species were reproduction type, pollen vector, fruit type, diaspore morphology, flowering duration, dispersal duration and diaspore mass. Trait values were gathered from the literature and databases (see table 1 for references). Where possible, missing values were extrapolated from closely related species. We measured the average diaspore mass of 36 plant species not present in the databases by weighing for each species a minimum of five viable, dry diaspores from three individuals in three populations (table S1, supplementary data available at stacks.iop.org/ERL/8/024043/mmedia). The final dataset contained <11% missing values per trait (NA), which we considered low enough for valid analyses.

For each species on each of the 120 summits, we assessed whether it had colonized (i.e. been historically absent but recently present) or become locally extinct (i.e. been historically present but recently absent) on the summit. We analysed whether a species trait affected the likelihood of colonizing or becoming extinct on a summit with generalized linear mixed models (GLMMs) with binomial error distribution using the package 'lme4' (Bates et al 2012) and the function 'glmer' in R 2.15.1 (R Core Team 2012). Any collinearity between explanatory variables had been previously identified by the 'corvif' function (package 'AED', Zuur et al 2009). As suggested by Booth et al (1994), pairs of traits with correlation coefficients >0.5 were considered collinear and the ecologically less meaningful variable was removed. Thus, the indicator values for light, continentality and soil humus content were discarded because they were collinear with temperature, soil moisture and soil aeration, respectively. The three categories of plant traits were analysed in separate GLMMs, with traits as fixed factors, and summit and species identity as random factors. The best models were selected through a stepwise backward selection. The interactions between summit elevation and traits were similarly tested with GLMM, but only traits significant in the first analyses were then retained.

Differences between traits were investigated by sum contrasts. The final models were visually validated on the basis of logistic regression quantile–quantile plots (calculated with 500 model runs and response residuals; Zuur et al 2009), and plots of working residuals against the significant explanatory variables.

For clarity of presentation in figures, we considered the proportion of each trait state on the total number of plant occurrences (excluding the NAs) in the historical and recent records, and among colonizers. Bar plots were produced to illustrate the changes in trait proportions. This involved dividing summits into three elevation categories of equal size (N = 40), corresponding approximately to the belt division of Gottfried et al (2011). The resulting belts were: lower alpine (2449–2802 m), upper alpine (2804–3006 m) and nival (3020–3418 m). Since continuous values for elevation were, however, used in the analyses, these limits are only indicative.

Summit shifts in a functional trait space was presented with a PCA based on mean trait values for quantitative traits and the proportion of trait states for categorical traits. For these latter traits, the less frequent state was skipped before analyses because of data redundancy (sum of state proportions is 100%). The significance of temporal shifts in functional trait space along the first three axes of PCA was tested by multivariate analysis of variance (MANOVA) by differences of axis scores among inventory dates against the intercept.

Traits indicating a species' ecological preference in terms of temperature, soil nutrient content, soil aeration, and their respective interactions with elevation had a significant effect on the colonization probability of a species (table 2; figure 3(a)). At lower elevations, the proportion of species originally from the subalpine and montane belt (T2 and T3) increased, while at higher elevations, lower alpine species (T1.5) increased most, and some montane species (T3) were also recorded. Alpine and nival species (T1) increased in frequency at all elevations (not shown), although proportionally they decreased. Most of the colonizing species were typical of infertile soils with a moderate to good aeration. However, at all elevations, the number of species indicating fertile soils (especially in the nival belt) and poor soil aeration slightly increased (figures 3(b) and (c)). Overall, the distribution of ecological indicator values on nival summits in recent times became more similar to that of lower alpine summits in historical times.

Zoom In Zoom Out Reset image size

Life form was the only morphological trait with a significant effect on the probability of species colonization, both alone and in interaction with elevation (table 2; figure 3(d)). Over the past century, hemicryptophytes and fruticose chamaephytes (ChF, i.e. woody dwarf shrub) have become more abundant at the expense of suffruticose (ChS, i.e. woody only at stem base) and cushion (ChC, i.e. cushion plants) chamaephytes. This trend was most pronounced in the nival belt, where the proportions of stalked hemicryptophytes (HeS, i.e. herbaceous plants with leaves along the stem) were similar to those in the upper alpine belt, while the proportion of rosette hemicryptophytes (HeR, i.e. herbaceous rosette plants) slightly decreased. The proportion of therophytes (Th, i.e. annual species) increased on the upper alpine summits.

Diaspore morphology and fruit type were the only dispersal traits that had a significant effect on the probability of a species colonizing a summit, but only the interaction between diaspore morphology and elevation was significant (table 2; figures 3(e) and 4). Compared to historical records, the proportion of diaspores with pappus significantly increased, while the proportion of species without appendages decreased. This was most apparent in the nival belt, where an increase in diaspores with awns was also detected. While in the historical records there were more capsules than achenes, in the recent records the situation was reversed. The proportion of species without fruit (mainly ferns) increased significantly, although it was still less than 2%.

Zoom In Zoom Out Reset image size

The variance due to random effects was similar for all models, ranging between 0.61 and 1.13 for species and between 0.93 and 1.10 for summits (table 2). Hence, the among-site variation was considered to have a limited influence on the results.

As in previous floristic studies, we also found that relatively thermophilous species were the ones that successfully colonized high summits (Camenisch 2002, Hofer 1992, Vittoz et al 2008, Wipf et al 2013), as was also the case in a recent pan-European study (Gottfried et al 2012). Higher temperatures and a longer growing season (Jolly et al 2005), which strongly influence high-elevation plant communities (Körner 2003), have alleviated growth restrictions for an ever increasing number of species. Thus, each elevation belt received new species that are typically considered as growing at lower sites (Landolt et al 2010): montane-subalpine species (T2 and T3, table 1) have colonized new summits of the lower alpine belt, and alpine species (T1 and T1.5) summits of the upper alpine and nival belts. Interestingly, alpine-nival species (T1) likewise colonized new summits of the lower alpine belt, indicating that these species still find suitable habitats at these elevations and that competition from subalpine species is not yet strong enough to limit their establishment.

Alpine-nival species (T1) require mostly oligotrophic soil conditions, whereas montane-subalpine species are more frequently nitrophilous (Güsewell et al 2012, Landolt et al 2010). Hence, the meso- to nitrophilous species (N3 and N4) that have recently colonized the summits may have been absent historically due to low temperatures rather than to poor soil conditions. However, increased nutrient input from the atmosphere (Tørseth and Semb 1997), from herbivores (Stöckli et al 2011) or from the faster humus mineralization under warmer conditions cannot be excluded as potential explanations.

Plant traits related to soil pH and moisture (potentially related to bedrock type) had no significant effects on species colonization; hence, we conclude that the observed colonization patterns are independent of the bedrock type. Nevertheless, Holzinger et al (2008) found a higher increase in species numbers on calcareous than on siliceous mountains, but the calcareous mountains were on average lower. Also our data set was not perfectly balanced given the availability of historical data (highest siliceous summit with historical data = 3418 m, highest calcareous = 3267 m). Hence, potential confounding effects of elevation and bedrock type on species colonization deserve attention in future studies.

The general increase in hemicryptophytes we observed is due to species colonization from the subalpine belt, where this growth form is the most common, whereas they are more in balance with chamaephytes in the nival belt (Aeschimann et al 2012, Vittoz et al 2010). The change is particularly evident for stalked hemicryptophytes in the upper alpine and nival belts, so that they today show similar proportions at all elevations. The simultaneous decrease in rosette hemicryptophytes (HeR) at all elevations was caused by a higher-than-average extinction rate, although this was not significant (not shown).

Cushion chamaephytes are well adapted to harsh high-mountain climate and rarely occur below 2500 m asl in the Alps (Pellissier et al 2010, Körner 2003). Their general proportional decrease reflects the low potential for new colonization as the species pool on many summits may have been saturated (most of the 14 cushion species were already frequent in historical times). Conversely, the increase in fruticose chamaephytes in the nival belt corresponds to the global increase in (dwarf-) shrubs in alpine grasslands (Vittoz et al 2008) and in alpine and arctic tundra (see Myers-Smith et al 2011; Epstein et al 2013 for reviews).

Although we could expect a colonization of pioneer species (ruderal strategy and sociological group of rocks and scree) like on glacier forelands (Burga 1999, Cannone et al 2008), sociology and plant strategy appeared to have no significant effect on the colonization success. This indicates that the floristic changes were not related to a succession driven by soil development. In contrast to the raw soils of recently deglaciated moraines, summits possess a wide range of soil types and a high heterogeneity in microtopography, providing suitable habitats for a large variety of species.

The composition of dispersal traits in the historical records reflects the elevational pattern described by Pellissier et al (2010), who found the proportion of species with pappus seeds rises from subalpine to alpine belts, but decreases in the nival belt, where species that have seeds with wings are more prevalent. During the past century, however, species with pappus seeds colonized sites intensively over the whole elevation range. As a result, they occur in similar or higher proportions than species that produce seeds with wings. Very light seeds without appendages are still well dispersed by wind (Greene and Calogeropoulos 2002), especially by the turbulent winds prevailing in mountainous regions (Tackenberg and Stöcklin 2008). Seeds with pappus, however, are efficiently dispersed by both wind and animals (Couvreur et al 2004), and allow the dispersal of heavier diaspores, richer in reserve compounds. This is advantageous for seedling establishment on rocky soils (Moles and Westoby 2006). Seeds with awns, mainly graminoids, are typically most frequent at lower elevations (subalpine belt; Pellissier et al 2010). Their proportion has slightly increased with the upwards shift of these species including, for instance, that of the most successful colonizer, Poa alpina (Wipf et al).

The historical distribution of fruit types again reflects the pattern observed by Pellissier et al (2010), who recorded a shift between achenes and capsules as dominant traits at around 2500 m. The high colonization rate of species with achenes is in accordance with the recent arrival of subalpine and lower alpine species on summits. This also corresponds with the pattern of diaspore morphologies described above, as many achenes are extended by pappus or awns, while capsules contain mainly seeds without appendages. The increase in species without fruit corresponds to the establishment of ferns, which used to be rare or even absent on most summits, but have since been some of the most successful colonizers (Wipf et al). Their tiny spores are dispersed over very large distances, which suggests their historical absence on summits is not due to dispersal, but rather to temperature limitation. Their recent establishment on the re-surveyed summits thus indicates that warmer conditions are very important for colonization processes (Vittoz et al 2008).