|Home | About | Journals | Submit | Contact Us | Français|
Understanding processes and mechanisms governing changes in plant species along primary successions has been of major importance in ecology. However, to date hardly any studies have focused on the complete life cycle of species along a successional gradient, comparing pioneer, early and late-successional species. In this study it is hypothesized that pioneer species should initially have a population growth rate, λ, greater than one with high fecundity rates, and declining growth rates when they are replaced by late-successional species. Populations of late-successional species should also start, at the mid-successional stage (when pioneer species are declining), with growth rates greater than one and arrive at rates equal to one at the late successional stage, mainly due to higher survival rates that allow these species to persist for a long time.
The demography of pioneer- (Saxifraga aizoides), early (Artemisia genipi) and late-successional species (Anthyllis vulneraria ssp. alpicola) was investigated together with that of a ubiquitous species (Poa alpina) along the Rotmoos glacier foreland (2300–2400 m a.s.l., Central Alps, Austria) over 3 years. A matrix modelling approach was used to compare the main demographic parameters. Elasticity values were plotted in a demographic triangle using fecundity, individual growth and survival as vital rates contributing to the population growth rates.
The results largely confirmed the predictions for population growth rates during succession. However, high survival rates of larger adults characterized all species, regardless of where they were growing along the succession. At the pioneer site, high mortality rates of seedlings, plantlets and young individuals were recorded. Fecundity was found to be of minor relevance everywhere, but it was nevertheless sufficient to increase or maintain the population sizes.
Demographically, all the species over all sites behaved like late-successional or climax species in secondary successions, mainly relying on survival of adult individuals. Survival serves as a buffer against temporal variation right from the beginning of the primary succession, indicating a major difference between primary and secondary succession.
Glacier forelands offer an excellent opportunity to study processes of primary succession (Matthews, 1992; Chapin et al., 1994; Walker and del Moral, 2003). Immediately after glacial retreat, succession is initiated by colonization processes (Matthews, 1992; Raffl and Erschbamer, 2004). The first plants arrive at barren moraines via seeds, vegetative propagules and/or by clonal propagation (Chapin et al., 1994; Stöcklin and Bäumler, 1996; Niederfriniger Schlag and Erschbamer, 2000). Although infrequent germination and low seedling survival have often been recorded in alpine habitats (Bliss, 1958; Urbanska and Schütz, 1986; Scherff et al., 1994; Chambers, 1995a; Forbis, 2003), genetic studies have shown variation in alpine plants similar to that of lowland species (Bingham and Ranker, 2000; Till-Bottraud and Gaudeul, 2002; Raffl et al., 2006a). Thus, recruitment from seeds should occur and species with high fecundity, high ability for seed dispersal and short generation time should be predominant among the first colonizers (Chapin et al., 1994). But stressful alpine habitats may also select for long-lived perennials with life histories showing low fecundities, few offspring and high adult survival rates from the very beginning of the succession onwards (Billings and Mooney, 1968; Bell and Bliss, 1980). Therefore a combination of high colonization abilities (Niederfriniger Schlag and Erschbamer, 2000; Erschbamer et al., 2001) and large seed production (S. Marcante, University of Innsbruck, Austria, unpubl. res.) with extreme longevity (Morris and Doak, 1998; Forbis and Doak, 2004), clonal growth (Stöcklin and Bäumler, 1996), or a combination of sexual and asexual reproduction (Weppler et al., 2006) is to be expected for alpine pioneer species. Despite such contrasting life-history traits, life cycles as a whole – the consequences of all life-history phases for recruitment, establishment and replacement during primary succession under alpine conditions – have, to date, barely been studied.
In the course of succession, as in other ecological processes, demography captures the fate of species during succession, reflecting the effect of interactions and of environmental impacts. Demographic studies usually cover a time period of only a few years. But primary succession in alpine ecosystems is a slow process, proceeding over decades or even centuries. A space-for-time substitution in a chronosequence approach, using moraines as fixed landmarks with known age in the glacier foreland, gives an appropriate framework for combining the results of demographic short-term studies. To quantify demographic factors from such field studies and to evaluate long-term consequences, matrix modelling offers a very useful approach. Population-dynamic models allow investigation of the current population structure and dynamics, and projection of future population development (Caswell, 2001). Such modelling has been applied to a variety of plant species differing in life history and environment, but to only a few species from arctic–alpine habitats, most of them showing clonal growth (Callaghan, 1976; Erschbamer, 1994; Erschbamer et al., 1998; Morris and Doak, 1998; Tolvanen et al., 2001; Dinnétz and Nilsson, 2002; Erschbamer and Winkler, 2005; Weppler et al., 2006). None of these studies have compared different species with each other, attempted to relate their dynamics to certain stages of primary succession, or drawn general conclusions on life-history strategies.
An important attempt to arrange demographic strategies of species along a successional gradient was the dichotomy between r- and K-selected strategies (Pianka, 1970). Grime (1979) proposed competitive, stress-tolerant and ruderal species' characteristics (C-S-R), originally for the study of secondary succession, but applied also for primary succession (Caccianiga et al., 2006). In a demographic matrix-model framework, the strategy of a species can be described by a scheme introduced by Silvertown et al. (1993). From matrix models they calculated elasticities and plotted these elasticities – grouped under the terms fecundity, growth and survival – in ‘demographic triangles’. For species occurring in the course of secondary succession they demonstrated that strategies can be arranged on a parabolic trajectory within this demographic triangle. Species with high fecundity (pioneer species; mainly short-lived herbs) are replaced by species with high growth rates of individuals (early successional species; herbs of open habitats) and these are finally followed by species with high survival rates (late-successional species, climax species; forest herbs, shrubs, trees). However, this ‘arc model’, offering an appropriate background for a comparative approach, has not been tested along either a primary succession or in the alpine environment. The longevity of alpine species, considered to be one of the major factors shaping successional changes from the early stages of the beginning of succession (Chapin et al., 1994), may represent a challenge for this model (Walker and del Moral, 2003). We therefore need to ask how primary succession under the harsh environmental conditions of glacier forelands will fit into the ‘arc model’ (Silvertown et al., 1993), considering that this model was based on results from species occurring in secondary successions.
In the present paper we have analysed, in a comparative manner, the demography and the population dynamics of four key species located along a primary successional gradient in a central Alpine glacier foreland. We selected one pioneer species, one early and one late-successional species, and one ubiquitous species. Permanent plots were established at three moraines with well-defined ages from which field records were taken over 3 years, and these data gave the basis for parameterizing and analysing matrix models for the four species in the study at different successional stages. From this combination of field work and modelling we asked the following questions: (1) Do population size, population structures and rates of population increase vary with moraine age (successional stage) in a characteristic manner? (2) Are there changes in fecundity, individual growth and survival rates with successional stage? (3) Do the demographic parameters of the species follow a general strategy concept?
The research area lies in the Central Alps of Austria on the glacier foreland of the Rotmoos Valley (Obergurgl, Oetztal, Tyrol, 46 °49′N, 11 °02′E) at 2300–2400 m a.s.l. The U-shaped valley is almost level and only slightly ascending near the glacier tongue. Over the last 150 years the glacier has retreated by more than 2 km. The largely well-preserved chronosequence exhibits a series of glacier moraines (e.g. 1971, 1923, 1870), delimitated by a terminal moraine ridge dated 1858 (G. Patzelt, University of Innsbruck, Austria, unpubl. res., 1995). Being a part of the northern climatic region with the character of a continental climate, the study site exhibits comparably low mean annual precipitation of approx. 1460 mm, and the average monthly air temperatures in summer range from 5·9 to 8·4 °C (Kaufmann, 2001). Permanent snow cover lasts from October to late May or beginning of June, with only small differences between the moraines.
The study sites were arranged along a primary successional gradient comprising the pioneer stage, and the early and the late successional stages (Raffl and Erschbamer, 2004; Raffl et al., 2006b). The pioneer stage (= moraine 1971) is characterized by pioneer assemblages mainly composed of Saxifraga aizoides, S. oppositifolia, Linaria alpina and Artemisia genipi (Raffl and Erschbamer, 2004; Raffl et al., 2006b). On the second site (= moraine 1923) the pioneers are replaced by early successional species such as Trifolium pallescens and Poa alpina. Additionally, Festuca spp., Silene acaulis and Salix spp. occur. On the 1858 moraine, late-successional species prevail, such as Agrostis alpina, Anthyllis vulneraria ssp. alpicola, Kobresia myosuroides, Leontodon hispidus ssp. alpinus and T. pratense ssp. nivale (Jochimsen, 1975; Raffl and Erschbamer, 2004).
Soil development along the glacier foreland is rather slow, from a Syrozem on the youngest moraines to Pararendzinas of 3–4 cm depth on the oldest parts (Erschbamer et al., 1999).
Four key species of the successional gradient were selected. They are all long-lived perennial herbs. The seeds are usually dispersed by wind and they form long-term, persistent seed banks except for Artemisia genipi, which has only a transient seed bank (Marcante et al., 2009; Schwienbacher et al., 2010). Nomenclature follows Fischer et al. (2005).
Saxifraga aizoides (Saxifragaceae) has small, semi-succulent leaves forming multi-ramet mats or cushions (Kaplan, 1995). Each raceme bears, at a maximum, ten yellow or orange flowers (Webb and Gornall, 1989). Each capsule releases between 100 and 500 extremely small seeds (seed mass: 0·04 ± 0·01 mg, S. Marcante, University of Innsbruck, Austria, unpubl. res.). In Europe, this arctic–alpine species is widely distributed within an altitudinal range between 800 and 3000 m a.s.l. It occurs in a broad variety of moist and calcareous habitats (glacier forelands, scree slopes, snow beds, alluvial material, river banks), and on the study site it is the most characteristic pioneer species immediately after deglaciation (Niederfriniger Schlag and Erschbamer, 2000; Raffl et al., 2006b).
Artemisia genipi (Asteraceae) is endemic to the Alps and is distributed in the alpine and subnival belt. It produces small fruits with a seed mass of 0·26 ± 0·03 mg (Schwienbacher and Erschbamer, 2002). Seeds are produced at an early life stage by small mother plants (i.e. plants with two rosettes). Artemisia genipi is an early successional species characteristically occurring on young moraines deglaciated from 35 to 83 years ago.
Anthyllis vulneraria ssp. alpicola (Fabaceae) is distributed in nutrient-poor grasslands or snow–heather pine forests. Along the successional gradient of the Rotmoos glacier foreland, it is a late-successional species especially abundant on the oldest moraine (Niederfriniger Schlag and Erschbamer, 2000; Raffl et al., 2006b).
Poa alpina (Poaceae) is an evergreen, pseudoviviparous grass, reproducing sexually by seeds as well as asexually by plantlets (Wilhalm, 1996). The plantlets are capable of photosynthesis (Lee and Harmer, 1980; Pierce, 1998), produce adventitious roots and rapidly become established after wind or water dispersal. The speed of establishment gives them an advantage over seeds in short arctic/alpine growing seasons (Lee and Harmer, 1980). Poa alpina is a ubiquitous species along the Rotmoos glacier foreland, occurring from the pioneer stage to the late successional one.
The four species were studied in permanent plots at two or three successional stages (moraines 1971, 1923, 1858) according to the frequency of their occurrence (Raffl et al., 2006b; Table 1). In July 2004, for each species separately, three permanent plots of 1 m2 were established at each stage in question, distributed randomly on approximately homogeneous areas and fenced off against grazers (sheep, horses) and tourists. Population data were collected over a period of 3 years, from 2004 to 2006. In each plot, one main census per year was conducted during the growing season (July–September). In order to record exactly the same individuals in every census, each 1 m2 plot was fixed by metal nails and subdivided into dm2. Each individual (adults and seedlings) was marked with a coloured wire. For each adult individual of S. aizoides, P. alpina and A. alpicola, the numbers of vegetative shoots and flower heads were counted, whereas for A. genipi the numbers of vegetative and generative rosettes were determined.
For the study species, as for many other perennial plants, age estimation based on above-ground morphological traits was not possible. Following field observations and previous studies (Schwienbacher, 2004), we therefore distinguished life-cycle stages on a size basis (for details see Table 2).
Seedlings: with cotyledons (or coleoptile for P. alpina) and generally one pair of leaves. Seeds germinate soon after snowmelt in the early summer following their production in the previous year or, most likely, from a permanent seed bank (Marcante et al., 2009). Only newly recruited individuals were considered as seedlings; small individuals with one shoot were placed in the first adult category.
This classification of individuals into stage categories followed a biological approach combining reproductive and survival criteria with size criteria (Horvitz and Schemske, 1995). For adult plants the number of shoots was assumed to be the best predictor of plant size in the case of S. aizoides, A. alpicola and P. alpina; whereas for A. genipi the number of rosettes was best suited for this purpose, following a previous study (Schwienbacher, 2004). Thus changes in plant size were due to changes in rosettes or shoot numbers giving individual growth and transition rates. Effects of herbivory were not explicitly addressed in the study.
As individuals were classified according to their size, we used Lefkovitch matrices to perform matrix analyses of population demography (Caswell, 2001). Each matrix comprised a time interval of 1 year and five or six stages: one or two non-reproductive stages (seedlings and, for P. alpina, plantlets), and four or five adult stages (Table 2). For determination of the number of individuals per size class and for calculation of transitions, we pooled vegetative and generative shoots. Transition probabilities from one life-cycle stage to another were calculated for each stage as the proportion of individuals remaining in that stage or having changed to another stage after an interval of 1 year (Caswell, 2001). Seedling mortality was calculated as the proportion of dead seedlings per year, whereas the adult mortality, mi, in size class i was calculated by
where aji are the elements of the ith column (except for fecundities).
Fecundities were calculated under the assumption of anonymous reproduction. The contribution of each adult stage to the production of seedlings or plantlets in the following years was determined by a partition on these stages of all generative shoots of the population. Hence, the fecundity, fi, of an individual of class i was given by:
where S represents the total number of newly emerged seedlings (or plantlets), gi is the proportion of generative shoots belonging to the individuals of class i, and xi is the number of these individuals. All generative shoots were considered to contribute equally to reproduction, irrespective of the size class of individuals to which they belonged.
Due to the anonymous character of reproduction, we had to assume that the number of seeds emigrating from the plot approximately equalled the number of immigrating seeds. An explicit ‘seed-bank stage’ was omitted in the model as we had no data on dynamics of the permanent seed banks. Therefore, our analysis depended on the assumption that there was a stable relationship between seed bank and above-ground population and also an equilibrium between study plots and their environment.
Transition matrices were constructed separately for each species plot per site and for each transition (2004–2005 and 2005–2006). The three plot matrices per site and species were then pooled in a non-weighted manner, thus averaging spatial environmental variation. Following the distribution of species' plots over the sites (Table 1) and the two transition steps, altogether 18 matrices were the basis for further evaluation. Presentations of population structure, fecundity and mortality rates were additionally averaged over the two time steps.
For each of the 18 transition matrices, population growth rates, λ, were calculated using the software package RAMAS EcoLab 2·0 (Sinauer Associates, Inc.). The stable stage distribution was computed from each matrix and averaged over the two transitions.
The elasticities (proportional sensitivities) of λ to changes in elements of the transition matrix were calculated separately for each year and population, giving 18 elasticity matrices. Following the approach of Silvertown et al. (1993), these elasticity matrices were divided into regions describing different parts of the life cycle along the scheme
where the scheme for P. alpina was extended by a second row and column for plantlet fecundity and survival. Within each region the elasticity sums were calculated. These regional G-S-F elasticities were used to characterize the relative contribution of growth (G), survival (S) and fecundity (F) to the finite rate of increase (λ) and were plotted in G-S-F triangle graphs (Silvertown et al., 1993; Silvertown and Franco, 1993).
Stochastic simulations projecting population development considered demographic and environmental stochasticity. Demographic stochasticity was incorporated into an individual-based model by treating transitions as multinomial processes and reproduction as Poisson processes. To account for environmental stochasticity in each simulation step, one of the two transition matrices in question was drawn at random with equal probability (Münzbergová, 2005). Environmental stochasticity was assumed to be uncorrelated over time. Population development, starting with the numbers of individuals per stage from the field study in 2004, was followed over 20 years (100 runs per case). Stochastic growth rate, λS, was approximated by averaging, over n runs, the quotient
The values xt and xt−1 were given by the final pair of population sizes, per run, with x the total number of individuals and t the time of simulation stop point (here, t = 20). From the results at t = 20, asymptotic stable stage distributions were determined, and the averages over all runs were compared with the 3-years' mean of observed distributions of individuals on stages, using Keyfitz's Δ. This measure quantifies the distance between observed and stable distributions, with values ranging from 0 to 1, equating to maximum similarity and maximum difference, respectively (Caswell, 2001).
For each species Table 3 gives the observed numbers of adults, seedlings, plantlets and the total number of individuals during the study period (2004–2006), summed over all plots per successional stage. Differences in population sizes between successional stages were clearly larger than temporal changes within the 3 years. Total population sizes as well as seedling numbers of S. aizoides and A. genipi decreased with the progress of succession (Table 3). Poa alpina, having two reproductive modes, showed the highest number of seedlings in the early successional stage, whereas the number of plantlets decreased steadily along the successional gradient, and in the late successional stage no plantlets at all were detected (Table 3). Total population sizes of P. alpina and A. alpicola were highest in the early successional stage.
Figure 1 presents the population structure of the four species from the field data, averaged over the three census years and the plots. The distribution of adult individuals over the life-cycle stages listed in Table 2 is included in these graphs. The population structure of the species investigated changed markedly between successional stages. The proportion of seedlings (and plantlets) decreased with ongoing succession for all species, with the exception of A. alpicola where this proportion did not change. A high proportion of small individuals (one rosette/shoot) characterized the populations of A. genipi and P. alpina all along the successional gradient. For S. aizoides and P. alpina already occurring at the pioneer stage, the proportion of large individuals increased with older successional stages. In contrast, large individuals (more than five rosettes) of A. genipi were scarcely detected at both successional stages, and for A. alpicola the proportion of large individuals even decreased, in favour of medium sized ones, with the progress of succession.
Rates of population growth (λ values) and elasticities aggregated over the fecundity, growth and survival regions (eqn. 3) are given in Table 4 for all species and successional stages, differentiated for the two transition steps, while the complete transition and elasticity matrices are presented in the Appendix. Overall, the populations studied showed population growth rates close to 1 in both years of study (Table 4). Remarkable exceptions were seen for the P. alpina population at the pioneer stage, where the highest growth rate (λ = 1·15) was found, while the late-successional population of A. alpicola gave the lowest one (λ = 0·82; Table 4).
Along the successional gradient, the populations investigated at the pioneer stage (A. genipi, S. aizoides and P. alpina) were progressive (λ-values above 1) with the exception of S. aizoides in the second year. At the early successional stage the populations of these species were regressive (λ -values below 1; Table 4). At this successional stage only A. alpicola increased in population size. At the late successional stage, the population growth rates of P. alpina were close to 1 (λ = 1 and λ = 0·96, respectively, in the first and second year), whereas the values for A. alpicola were lower than 1 in the first year (0·82) and higher (1·11) in the second year, indicating high fluctuations over the years.
Mortality rates per species are illustrated in Fig. 2. Mortality affected all the life-cycle stages of the populations investigated. Mortality could not be quantified in the study period only for very large individuals of S. aizoides (>50 shoots) at both the pioneer and the early successional stages and for the large Artemisia individuals at the early successional stage. In general, mortality decreased with increasing size of individuals: whereas small individuals (one shoot/rosette) showed mortalities up to 0·5, all larger individuals had mortalities of at most 0·2, and even less than 0·1 for large individuals (stages 4 and 5). With the progress of succession, mortality increased for A. genipi, with the exception of very large individuals. In contrast, A. alpicola, S. aizoides and P. alpina showed decreasing mortality in all life-history stages (including plantlets and seedlings) along the three successional stages.
Fecundity rates, i.e. the contribution of individuals to seedling formation as dependent upon life-history stage, are given in Fig. 3 and values were averaged over two annual steps. In most cases only the individuals of the largest stage class contributed to reproduction in a significant manner, but seed production could not be fully excluded for individuals of smaller size classes, not even for the smallest individuals for A. alpicola. The same held for plantlet formation of P. alpina individuals. In most cases there was a clear decrease in offspring production with ongoing succession, with the exception of A. alpicola, where seedling formation slightly increased at the late successional stage. For A. genipi the decrease in net fecundity with succession was enhanced by the increase in seedling mortality (Fig. 2), whereas for P. alpina the strong decrease in fecundity rate was weakened by a reduction in seedling and plantlet mortality. The very high fecundity rate of S. aizoides in the pioneer stage was accompanied by exceptionally high seedling mortality.
The elasticity values were grouped and plotted according to the scheme of eqn (3), and together with the lambda values the results are given in Table 4. Figure 4 shows that there were large variations in the relative contribution of each region of the elasticity matrix between years and sites. Nevertheless, these regional elasticity values were clearly positioned along a line in the diagram from the upper-left to the lower-right: from a region with low fecundity, medium growth and high survival, to a region with very low fecundity, lower growth but very high survival. Associated with this shift in matrix position to the bottom right of the diagram was a decrease in λ (Table 4). The species in the study ‘moved’ along this line (describing the ‘movement’ of the species along the successional gradient) in a different manner. Artemisia genipi and P. alpina both moved towards the bottom right corner of the plot, but the values for S. aizoides and A. alpicola did not shift. The values for S. aizoides remained in the bottom right corner, indicating that populations were already fully established at the pioneer stage, whereas the values for A. alpicola stayed at the upper-left end of the region covered by the plot (Fig. 4), in agreement with the increasing population size of this species in the late succession stage.
In general, an increase of population growth rate, λ, was accompanied by an increase in the importance of fecundity and individual growth and a decrease in that of survival (Table 4). None of these correlations was significant (results not shown), indicating a high influence of species-specific and site-specific aspects. However, high values of survival (S) were characteristic for all the four species (Table 4). Growth (G) and fecundity (F) elasticity values were lowest in S. aizoides; higher fecundity values were calculated for the other three species (Table 4). The elasticity analyses of the pioneer and the early successional P. alpina populations showed that fecundity, given by plantlets (mainly in the pioneer stage) or seedlings (early successional stage), contributed considerably to the population growth rate, whereas at the late successional stage this contribution was rather irrelevant. At this stage, in addition to the survival of small and large individuals, the growth of small individuals to medium ones was critical as well (Appendix). The most critical process with respect to changes in λ for S. aizoides was the survival of large individuals at the pioneer as well as at the early successional stage (Appendix). Small and medium individuals were the most sensitive stages for A. genipi. For A. alpicola large individuals, but also seedlings and younger individuals, proved to be sensitive stages at the early and late successional stages.
Results of stochastic simulations for the populations studied over 20 years and the corresponding stochastic λ-values per successional stage are given in Fig. 5. These stochastic λ-values were in close agreement with the averaged deterministic λ-values in Table 4.
Projections of population development and their dependence upon successional stages were in agreement with the stochastic λ -values (Fig. 5), with increasing populations at λ > 1 and decreasing ones at λ < 1. The standard deviations of the projections show that even regressive populations would not disappear from the study plots after 20 years, assuming an ongoing validity of the actual conditions of the study. On the other hand, increasing populations in the pioneer stage (A. genipi and P. alpina, but also S. aizoides) and in the early successional stage (A. alpicola) would reach (under constant conditions) extremely high individual numbers per plot after 20 years.
Along the successional gradient, the time-averaged observed frequencies of life-cycle stages were similar to the averages of the stable distributions derived from the transition matrices (results not shown) as well as to the stable stage distributions predicted from the stochastic simulations (Fig. 1). Keyfitz Δ-values comparing simulated and observed values were about 0·1 in all cases (Fig. 1). For A. alpicola, the agreement between the distributions was even better, both at the early and late successional stages (Δ-values: 0·051 and 0·076, respectively).
Different processes and mechanistic models have been proposed to explain colonization and succession in glacier forelands (reviewed in Matthews, 1992; Whittaker, 1993; Chapin et al., 1994; Walker and del Moral, 2003; Caccianiga et al., 2006), but none of these have considered the complete life cycle of key species dominating the course of succession. This paper is the first to compare demography and population development of several species along a primary successional gradient.
In this study we observed, for the most part, populations of stable structure, indicated by a high similarity between observed and predicted stage distributions. This good agreement shows that in the study plots all species (except S. aizoides in the pioneer stage) had reached a population structure that may only change in the long term, relative to the study period of 3 years and even to the simulation period of 20 years. Such a stability of populations was also found by Tolvanen et al. (2001) and Weppler et al. (2006), and it supports our assumption of a stationary relationship between adult plants and a permanent seed bank. The result implies that processes giving rise to succession manifest themselves only over time periods much larger than the study period.
Demographic processes and resulting growth rates, λ, varied with moraine age. At the pioneer stage, population growth rates of the pioneer S. aizoides, the early successional A. genipi, and the ubiquitous P. alpina were progressive (λ > 1) or close to equilibrium. In these stages enough free space was available for newly arriving individuals, and hence a lack of competition prevailed. High fecundity was counterbalanced by high seedling/plantlet mortality and by low seedling/plantlet establishment, but nevertheless it was sufficient to produce an increase in population sizes. Although the pioneer site is characterized by a great abundance of small- and medium-sized stones and cushion plants, offering safe sites for seedling/plantlet recruitment and establishment (Frenot et al., 1998; Jumpponen et al., 1998, 1999; Niederfriniger-Schlag and Erschbamer, 2000; Erschbamer et al., 2001; Acuna-Rodriguez et al., 2006; Cavieres et al., 2007), seedlings/plantlets are subjected to an inconsistent water supply, low nutrient availability and other harsh conditions (frost heave; Chambers, 1995b). The risk of drought stress is probably the major factor for seedling/plantlet mortality (Matthews, 1992; Chapin and Bliss, 1989; Giménez-Benavides et al., 2007), due to the poor water retention capacity of the raw soils and the patchy vegetation (Chapin and Bliss, 1989). At the pioneer stage, the recruitment of P. alpina by plantlets was more important than the recruitment by seeds. The advantage of reproduction by plantlets has already been highlighted by different authors (Matthews, 1992; Pierce, 1998; Pierce et al., 2000), but in our study both seedling and plantlet mortality were rather high at the pioneer stage.
In contrast to the pioneer successional stage, population growth rates decreased considerably at the early successional stage with the exception of A. alpicola, which started population growth in this stage. Seedling density of S. aizoides was much lower than in the pioneer stage, but mortality remained at the same level. We suggest that S. aizoides formed remnant populations at the early successional stage, which resisted local extinction for a long time despite a negative population growth rate (Eriksson, 1996, 2000). According to Kaufmann (2001) and Raffl et al. (2006b), diversity increases exponentially up to this successional stage. Soil organic content (Erschbamer et al., 1999), moisture and nutrient availability (Matthews, 1992; Chapin et al., 1994) and microbial activity (Tscherko et al., 2003) are clearly higher compared to the pioneer stage. These factors enhance the ability of different late-successional species (e.g. A. alpicola) to become established and to expand in this area. Thus, the pioneer and early successional species experienced an increase in competition. The higher competition and the lower availability of ‘free’ areas increased the mortality of individuals from pioneer and early successional species, regardless of size (except P. alpina). Recruitment and establishment of seedlings and plantlets of P. alpina, A. genipi and S. aizoides were almost prevented, and thus the population growth rates were regressive. On the other hand, A. alpicola – as a late-successional competitor – behaved like a pioneer, i.e. invading this successional stage by having progressive populations.
At the late successional stage the populations of A. alpicola and P. alpina exhibited a slightly decreasing trend. Because of higher vegetation, mean cover and higher beta-diversity (70 % and ~3, respectively; Raffl et al., 2006b), competition with other species is assumed to be a determining factor. Competition with Kobresia myosuroides – the dominant graminoid species on this site (Raffl et al., 2006b) – may be suggested, as found in a study on Carex subspecies (Erschbamer and Winkler, 2005). Plantlets of P. alpina were hardly present in the early and late successional stage, suggesting that the switch from asexual to sexual reproduction may depend upon the amelioration of the environmental conditions.
For a demographic matrix model approach, the calculation and plotting of grouped elasticities was an adequate method for a general interpretation. Silvertown et al. (1996) and Franco and Silvertown (2004) showed that the elasticity of demographic rates may explain the relevance of life history traits and population parameters. Following their method, our species were placed very close together within the possible range of positions in the plot. All four species were characterized by high survival and extremely low net fecundity contributions. The survival of adult individuals was found to be essential for the performance of the glacier foreland populations; however, populations cannot rely upon it exclusively, and hence a low fecundity, at least, was indispensable for the populations. Life-history characteristics, such as very high survival rates of reproductive plants with sporadic mortality events and a population growth rate limited by low values of recruitment, have already been detected for other long-lived perennials in stable habitats (Nault and Gagnon, 1993; Ehrlén, 1995; Damman and Cain, 1998; Nantel and Gagnon, 1999; Forbis and Doak, 2004). In all successional stages, the life history of our species resembled that of forest herbs and trees, i.e. late-successional or climax species in the classification of Silvertown et al. (1993). This result was rather surprising, as the four key species represented pioneer, early, and late-successional species, appearing along the successional gradient under different environmental conditions and among different neighbours. Nevertheless, the main trait of the four species is longevity, giving rise to the observed distribution in the G-S-F triangle. It remains open whether at the very beginning of succession, on areas much younger than our pioneer plots, populations will rely to a higher degree on the arrival and establishment of new individuals. But due to the harsher conditions in such areas, higher fecundity can hardly be assumed. Here, the source of seeds or plant fragments must be outside these areas, giving rise to a net import of propagules.
The projections of population development revealed two types of population. Those at the pioneer stage, S. aizoides and P. alpina, increased considerably in size. The same held for A. alpicola at the early successional stage. The other three species were already decreasing at this stage, and so were the populations at the late-successional stage. But for all declining populations (with the exception of S. aizoides in the early successional stage), the risk of disappearance was low, often non-existent, over several decades due to the low mortality rate of large individuals.
Projection by matrix models depends upon the assumption that environmental conditions will not change within the projection period. In general, this is not true and therefore matrix models are known to reflect rather the present state of conditions, instead of providing a reliable picture of future developments (Caswell, 2001). In our system there are at least four sources for environmental change: (1) increase in intraspecific competition due to an increase in density, (2) changes in interspecific interactions during the progress of succession, (3) impact of stochastic environmental effects and of catastrophes, and (4) drastic changes in alpine ecosystems as a consequence of ongoing climatic change (Erschbamer, 2007; Pauli et al., 2007). The impact of environmental stochasticity will clearly be incompletely covered by a 3-year study with three plots per case; nevertheless, the present study will mark a reference point for an ongoing evaluation of succession processes by demographic analyses.
Alpine species over their complete life cycle do not fit into the demographic succession model proposed by Silvertown et al. (1993). Whereas this model, mainly developed for secondary successions, assumes that the course of succession starts with short-lived species characterized by high fecundity, our results regarding primary succession indicate that, under harsh conditions, survival of adult individuals right from the beginning is crucial in the life cycle of the species studied. From a demographic point of view, all our species behave like late-successional or climax species. The survival of adult individuals serves as a buffer against temporal variation, and is thus a condition of particular relevance in disturbed habitats such as alpine scree or glacier forelands. Fecundity was found to be of secondary importance, but it was sufficient to ensure an increase in population sizes under low-competition conditions. Our results emphasize stability of population structure in the comparably slow progress of primary succession. The dynamic counterparts, initial colonization as well as reactions to disturbances and catastrophic events, remain to be studied and classified in order to provide a complete picture of the main demographic processes in primary successions on glacier forelands.
We warmly acknowledge Erich Schwienbacher for helpful discussions and for field assistance, and Barbara Beikircher and Fabian Nagl for help in the field. Many thanks are due to Meinhard Strobl for support during the field work. We are grateful to Paul Ronning for linguistic help. We also are indebted to the editor and two anonymous reviewers for their constructive and helpful comments. This work was supported by FWF Austrian Science Fund (FWF P16615-B06).
Stage-based transition matrices and elasticity matrices for the four species studied for the successional stages according to Table 1, and for the two annual steps 2004–2005 and 2005–2006. Life-cycle stages are described in Table 2 (S, seedlings; P, plantlets).
|Transition matrices||Elasticity matrices|
|(A) Saxifraga aizoides (Sa)|
|Pioneer stage, 2004–2005||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5|
|Pioneer stage, 2005–2006||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5|
|Early successional stage, 2004–2005||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5|
|Early successional stage, 2005–2006||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5||S||Stage 1||Stage 2||Stage 3||Stage 4||Stage 5|
|(B) Artemisia genipi (Ag)|
|Transition matrices||Elasticity matrices|
|Pioneer stage, 2004–2005||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|Pioneer stage, 2005–2006||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|Early successional stage, 2004–2005||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|Early successional stage, 2005–2006||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|(C) Anthyllis alpicola (Aa)|
|Transition matrices||Elasticity matrices|
|Early successional stage, 2004–2005||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|Early successional stage, 2005–2006||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|Late successional stage, 2004–2005||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|Late successional stage, 2005–2006||S||Stage 1||Stage 2||Stage 3||Stage 4||S||Stage 1||Stage 2||Stage 3||Stage 4|
|(D) Poa alpina (Pa)|
|Transition matrices||Elasticity matrices|
|Pioneer stage, 2004–2005||S||P||Stage 1||Stage 2||Stage 3||Stage 4||S||P||Stage 1||Stage 2||Stage 3||Stage 4|
|Pioneer stage, 2005–2006||S||P||Stage 1||Stage 2||Stage 3||Stage 4||S||P||Stage 1||Stage 2||Stage 3||Stage 4|
|Early successional stage, 2004–2005||S||P||Stage 1||Stage 2||Stage 3||Stage 4||S||P||Stage 1||Stage 2||Stage 3||Stage 4|
|Early successional stage, 2005–2006||S||P||Stage 1||Stage 2||Stage 3||Stage 4||S||P||Stage 1||Stage 2||Stage 3||Stage 4|
|Late successional stage, 2004–2005||S||P||Stage 1||Stage 2||Stage 3||Stage 4||S||P||Stage 1||Stage 2||Stage 3||Stage 4|
|Late successional stage, 2005–2006||S||P||Stage 1||Stage 2||Stage 3||Stage 4||S||P||Stage 1||Stage 2||Stage 3||Stage 4|