Limits to convergence of vegetation during early primary succession del Moral, Roger
|
|  Yüklə 2.39 Mb.
|
- Preliminary draft Limits to convergence of vegetation during early primary succession del Moral, Roger Department of Biology, University of Washington, Box 351800, Seattle, Washington 98195-1800 USA moral@u.washington.edu Abstract Questions: Primary succession, measured by changes in species composition, is slow, usually forcing a chronosequence approach. I use a unique data set to explore spatial and temporal changes in vegetation structure after a 1980 volcanic eruption. Using data from a transect of 20 permanent plots with an altitudinal range of 250 m sampled through 2005, I asked: Do changes along the transect recapitulate succession? Do plots converge to similar composition over time? Location: A ridge between 1218 and 1468 m on Mount St. Helens, Washington, USA. Methods: Repeat sampling of plots for species cover along a 1 km transect. Floristic changes were characterized by techniques including DCA, clustering and similarity. Results: Species richness and cover increased with time at rates that decreased with increasing elevation. The establishment of Lupinus lepidus accelerated the rate of succession and may control its trajectory. Diversity (H') at first increased with richness, then declined as dominance hierarchies developed. Primary succession was characterized by overlapping phases of species assembly (richness), vegetation maturation (diversity peaks, cover expands) and inhibition (diversity declines). Each plot passed through several community classes, but by 2005, only four classes persisted. Succession trajectories (measured by DCA) became shorter with elevation. Similarity between groups of plots defined by their classification in 2005 did not increase with time. Similarity within plot groups converged slightly at the lower elevations. Despite similarities between temporal and spatial trends in composition, trajectories of higher plots do not recapitulate those of lower plots, apparently because Lupinus was not an early colonist. Any vegetation convergence has been limited to plots that are in close proximity. Keywords: long-term permanent plots, Mount St. Helens, primary succession, succession trajectories; vegetation structure Abbreviations: AG = amalgamated group; CC = community class; DCA = detrended correspondence analysis; MRPP = multiple response permutation procedure; PS = percent similarity; SD = standard deviation Nomenclature: A list of species used in this study is at http://protist.biology.washington.edu/delmoral/. Nomenclature follows the Integrated Taxonomic Information System: http://www.itis.usda.gov/advanced_search.html. Introduction The cataclysmic 18 May 1980 lateral eruption of Mount St. Helens, Washington provided a unique chance to explore primary succession (del Moral et al. 2005). In 1984, I established a transect of permanent plots up a devastated ridge to initiate one of the longest continuous records of early volcanic primary succession (cf. Magnússon & Ólafsson 2003; Weber et al. 2006). Permanent plot studies have a long history and are an elegant alternative to chronosequence studies because they eliminate assumptions based on space-for-time substitutions (Austin 1981). The Buell secondary succession study (Bartha et al. 2003) clarified aspects of community dynamics and restoration (Pickett et al. 2001). Dutch ecologists have used permanent plots for decades to show cyclical and directional vegetation dynamics (Leendertse et al. 1997; Smits et al. 2002). Permanent plots are useful in studies of climate change (Petriccione 2005), and in field experiments (Brys et al. 2005) and restoration (Bekker et al. 1999; Grootjans et al. 2001; Bernhart & Koch 2003; Kiehl & Wagner 2006). Despite their advantages, permanent plots are rarely used to study primary succession because vegetation changes slowly (del Moral & Grishin 1999), but the 25-year period of this study was sufficiently long to reveal patterns. Measuring sequential percent cover and species composition in permanent plots allows the trajectory of vegetation change to be quantified by multivariate methods and similarity measures (Philippi et al. 1998). If two sites become floristically more similar over time, their trajectories are convergent (Baer et al. 2005). Whether or not vegetation converges to a single community is a critical theme in studies of primary succession (Fukami et al. 2005). The tendency for vegetation samples to approach a “target” or to become more similar to each other has been noted where biotic interactions are strong Weiher & Keddy (1999; del Moral et al. 2007). However, failure to converge can result from several factors (Capers et al. 2005; McEwan & Muller 2006). Early in primary succession, biotic interactions are weak (Callaway & Walker 1997), so landscape (Belyea & Lancaster 1999) or priority effects can alter establishment success (Drake 1991; Inouye and Tilman 1995; Temperton & Zirr 2004) and preclude convergence. Differing environmental conditions among plots also may reduce convergence. Species assembly and vegetation maturation dominate early succession as populations increase in response to abiotic factors (Walker & del Moral 2003). After most available species have established and expanded their dominance, inhibition can exclude rare species, leading to turnover and lower richness (del Moral & Ellis 2004). In this paper, I describe vegetation structure (i.e. richness, cover, diversity) changes on a volcanic ridge between 1984 and 2005 and with increasing elevation to compare rates of succession. I use floristic trajectories and transitions among “community classes” (CCs, a term used to emphasize that groups are not formal associations) to describe succession and to explore mechanisms. Trajectories of high elevation plots could be seen as a recapitulation of those followed by lower plots delayed by environmental stress (see del Moral & Ellis 2004). However, this view assumes that surfaces are similar, that landscape factors are irrelevant and that priority effects are unimportant. Higher plots are slightly steeper, with less soil, and Lupinus lepidus has not yet formed dense populations. I will investigate if vegetation samples converge over time and discuss factors that may preclude convergence. Methods I used several complementary approaches to address questions about succession rates and convergence. I used regression analyses of time as a predictor of descriptors of plant structure to document temporal and spatial changes. Community classes were developed to enhance visualization of overall changes in vegetation in time and space. I used detrended correspondence analysis (DCA) of these data and percent similarity (PS) to compare rates of floristic change and to explore trajectory patterns. Location Studebaker Ridge is on the northwest flank of Mount St. Helens, Washington, USA. During the eruption, the direct lateral blast removed all vegetation and most soil (Dale et al. 2005). I established a transect between 1218 m and 1480 m, centered at 46° 13' 52" N, 122° 11' 39" W. Plot SR-1 is 3.9 km from the focus of the 1980 eruption (i.e. the first dome), while SR-20 is 2.9 km distant. I presented structural data of these plots through 2002 (del Moral et al. 2005), but floristic patterns have not been presented. Sampling I established 250-m2 circular plots up the ridge at 50 m intervals. The first ten were established in 1984; in 1989, I added ten more plots uphill. I marked the center of each plot and the ends of four 9-m radii. Slopes range between 9 and 15°. More soil survived in SR-1 to SR-8 (except at SR-5) than higher plots. I measured species percent cover annually in 6-¼ m2 quadrats arrayed at 1 m intervals along each radius (n=24). Species within the plot but not in any quadrat were given cover of 0.1%. My reconnaissance in 1983 located no plants above 1200 m. A few plants may have established by 1988 on plots first sampled in 1989. Other than Salix sp., all taxa were identified to species. I posted the data on the H. T. Andrews LTER web site http://www.fsl.orst.edu/lter/data/abstract.cfm?dbcode=TV070&topnav=97. Pattern descriptions For each plot, I calculated the number of species (richness), percent cover and the Shannon-Weiner index (H'= [-(Pi lnPi)], where Pi is the proportion of total plot cover of a species. Graphs were produced using AXUM 7 (Mathsoft 2001). I used percent cover data to classify each year of data from each plot into community classes using agglomerative sorting with absolute Euclidean distance and flexible sorting (β = -0.25), and determined provisional classes from the dendrogram. The validity of the ten CCs formed was assessed by multi-response permutation procedures (MRPP; McCune & Mefford 2002) that use Sørensen’s distance to compare the within- with between-group distances. Trajectories β diversity is compositional change in space or time (Carey et al. 2007). I used DCA with Hill’s scaling (McCune & Mefford 2002) to describe composition changes (species with < 30 occurrences removed). DCA axes are expressed in standard deviation units (SD), so that a typical species enters and leaves along the axis in about four SD. So measured, distance in DCA ordination-space estimates β diversity such that about 1.3 SD is one half-change (two samples have 50% compositional similarity). I used DCA, despite its well-known disadvantages (Minchin 1987), because it is robust with low β diversity and it is suited to estimating gradient lengths (Legendre & Anderson 1999). Nonmetric multidimensional scaling (McCune & Mefford 2002) showed that only one axis was significant, so I used only DCA-1. I conducted four analyses using data from 1989 to 2005: A. Individual plots—cover percentage; DCA axis reflects both changes in cover and shifts in species composition. B. Individual plots—cover values transformed to proportion of species maximum in the data set; DCA axis reflects changes in species composition. C. Amalgamated groups (AG)—cover percentage; formed by assignment of plots to one of four groups based on its community class in 2005. AG-J consisted of plots SR-1 to 4; AG-I was divided into lower (AG-IL, SR-6 to 10) and upper (AG-IU, SR-11 to 15) plot groups; and AG-H consisted of SR-16 to 20. The anomalous SR-5 was omitted. D. AG as in (C)—but cover values were transformed as in (B). I conducted a two-way indicator species analysis (McCune & Mefford 2002) to determine a reasonable order for species in Table 3. Convergence To assess convergence, I evaluated changes in the SD of DCA scores of plots that were in the same CC at the last (2005) census (based on absolute and relative cover). I compared changes in DCA scores for each of the four plot groups using absolute and relative cover. A complementary approach uses PS for pair-wise comparisons: Statistics Statistical analyses and summaries were conducted with Statistix 8 (Analytical Software 2003). The simultaneous analysis of temporal changes in plots without replication and along environmental gradients is problematic. Lack of independence and pseudoreplication preclude using these data to test hypotheses, but I did use statistical procedures to illustrate patterns, seek trends and generate hypotheses. Changes in structure were assessed by multiple regressions with elevation and time as independent variables. The number of years it took DCA scores of a plot to reach a benchmark DCA score was compared using Spearman’s rank order correlations (Sr). DCA scores generally declined with age, so the benchmarks were 0.25, 0.5, 1.0 and 1.5 SD. The logic was that lower plots would reach each value sooner than higher plots, so the regression was between plot number (1 to 20) and number of years for each plot to reach each benchmark. I used linear regression of DCA scores vs. time to assess trends. I assessed changes in similarity within and between groups over time with one-way ANOVA. Comparisons among groups were based on the conservative Bonferroni statistic. Results Vegetation structure Each measure of structure changed significantly over time in each plot and between plots Table 1). Lupinus lepidus was often the first species to arrive, followed closely by Anaphalis margaritacea. The remaining early colonizers were wind-dispersed species (e.g., Chamerion angustifolium, Hieracium albiflorum and Hypochaeris radicata; Fuller and del Moral 2003). Although cover of Lupinus increased progressively up the ridge, it was not among the pioneers at the higher plots. I noted 48 species, many of them uncommon, or, like ferns, confined to gullies. Initially, richness increased in each plot (Fig. 1), then after about 2001 it declined. Mean richness increased from 1.5 species per plot in 1984 to 19 species per plot in 2001 (P < 0.0001 by ANOVA). Only time (P < 0.0001; r2 = 0.68) predicted species richness in a regression of richness vs. time and elevation. Percent cover increased with time, but it decreased with elevation in any particular year. Using cover derived from the 24 individual quadrats per plot permitted a statistical test of differences in cover between years and between plots within each year (Fig. 2). Cover increases of each plot over time were significant (P < 0.0001). Total percent cover increased progressively with age from negligible values in each plot (P < 0.0001; Table 1). Changes in cover over elevation and time were visualized in a 3-dimensional plot (Fig. 3). Cover increased with time and decreased with elevation. The rate of cover increase was decreased with elevation. Cover of SR-1 increased most quickly, with rates decreasing sequentially for higher plots. Mean cover over all years decreased from 48.1% in SR-1 to < 3% in higher plots (ANOVA, P < 0.0001). At lower elevations, peak values were due to a pulse of Lupinus, followed by its decline and increases in other species. At higher elevations, grasses and mosses became more important. Coincidentally, time and elevation contributes nearly equally to cover predictions (r2 = 0.53; tplot = -16.2; tyear = 14.7; P < 0.0001). This relationship will change. The regression lines of cover vs. time of the lower 10 plots declined with elevation from 4.03 to 0.96 (P < 0.0001); the slopes of this regression for the upper 10 plots decreased from 1.3 to 0.18 (P < 0.0001). Plot diversity (H') variation in time and elevation was complex (Table 1). Both time (P < 0.0001, tyear = 11.8) and elevation (P < 0.0002, tplot = 3.81) were significant predictors (combined r2 = 0.25), but nonlinear responses were evident. Adding quadratic functions improved the fit such that for time, r2 = 0.42, while elevation had little power (r2 = 0.04). Mean H' was least at lower elevations due to the dominance of Lupinus. High elevation plots had similar richness, but little dominance, producing an Table 1. Summary of structural characteristics. Superscripts indicate group membership determined after ANOVA by Bonferroni comparisons (P < 0.05). Left side of table: Np is number of plots in a year, values are means for all plots in the year; right side of table: Ny is number of sample years for each plot, values are means of the plot across all sampled years. Richness = mean number of species per plot; Cover = mean percent cover in plots; H' is mean Shannon-Wiener diversity, excluding plots with no species. N.B. Cover in 1998 is skewed upward because the sparsely vegetated plots SR-11 to SR-20 were not sampled. Groups defined by Bonferroni comparisons are for illustrative purposes only.
|
, where i and j are two samples, there are n species, xik and xjk are