Limits to convergence of vegetation during early primary succession del Moral, Roger

Each plot changed through several CCs (Table 2) reflecting accumulation of species and increases in cover. Plots reached their 2005 class by 2001. Tran­sitions through CCs reflected increasing dominance by persistent species (Table 3). Cover increased pro­gressively in SR-1 to SR-4, which stabilized as CC-J. Although initially dominated by Lupinus, they now have substantial concentrations of Penstemon, Agrostis and mosses. The anomalous SR-5 is exposed, barren and rockier than other plots and was arrested in CC-G and lacks dominance by Lupinus or mosses. Plots SR-6 to SR-15 stabilized as CC-I, although at different times and by different trajecto­ries. CC-I had less cover than CC-J and was domi­nated by Agrostis pallens and mosses, with some Lupinus. I divided plots in CC-I into lower and upper sub-classes for studies described below. Plots SR-16 to SR-20 were sparsely populated and diverse. They stabilized at CC-H, characterized by Agrostis pallens, Cistanthe and low cover by mosses and Lupinus.

The relative importance of Lupinus lepidus declined over time in many plots. I calculated rela­tive Lupinus cover for each plot over time. When I corre­lated relative Lupinus cover with year, there were significant negative correlations in SR-1 to SR-3, SR-13 and SR-14 and SR-19. Lupinus was domi­nant in these plots. In contrast, SR-5 and SR-20 had positive relationships.
Trajectories
I conducted DCA to compare vegetation change over space and time. DCA scores over time for indi­vidual plots form succession vectors (i.e. trajectories) because they are directional and species composition at the end of the study differs systematically from the first samples.

Mean DCA scores (absolute cover, plots) changed through time (r²=0.27; P < 0.0001); the first axis extended for 2.97 SD, with 25% of the total variance. DCA scores of representative plots over time demon­strate this pattern (Fig. 4A). The change of DCA scores was large in low elevation plots and small in higher elevation plots. High DCA scores occurred in early years and in high elevation plots dominated by Chamerion, Cistanthe, Anaphalis and Luetkea. Low scores reflected dominance by Lupinus, Penstemon, Agrostis pallens, Salix and mosses.

DCA scores of plots based on relative cover changed little with time (r² = 0.03, P < 0.004). The first axis was 2.49 SD long and had 17% of the vari­ance (Fig. 4B). Annual changes slowed after initial changes. DCA scores of lower sites diverged from those of higher ones between 1989 and 1995, reflecting disproportionate increases in Lupinus, Pen­stemon and mosses.

The mean DCA scores (absolute cover) of each amalgamated group (AG; each yearly sample of those plot in community class in 2005) declined with time (Fig. 5A). AG-J changed 1.5 SD (P < 0.0001, slope = -9.3); AG-IL shifted over 1.5 SD (P < 0.0002, slope = -11.4); AG-IU declined 1.1 SD (P < 0.0001, slope =-9.5); and AG-H changed 0.3 SD (P < 0.0001, slope =-4.2). Each regression was significantly different from the other, although those of the AG-I subgroups were barely distinct. When the mean DCA scores (relative cover) of the AGs were regressed with time, patterns were subtler than with absolute cover (Fig. 5B). AG-J increased 1.57 SD (P < 0.0001, slope = 5.4); AG-IL increased 0.6 SD (P < 0.0001, slope = 3.0); AG-IU increased 0.7 SD (P < 0.0001, slope = 2.9); and AG-H decreased 0.6 SD (P < 0.007, slope =-1.9). The slopes of the AG-I subgroups did not differ.
Convergence
Reduced variation in DCA scores (SD) of plots within each amalgamated group over time would suggest convergence. This was rarely the case. This regression showed a significant decline only in AG-J (t = -2.53, P < 0.03, r² = 0.31), and SD in 2005 remained large (0.4 SD). With DCA scores based on relative cover, the SD of AG-J declined (t=-2.26, P < 0.04, r² = 0.27), while the initially heterogeneous AG-H declined more steeply (t=-4.24, P < 0.001, r² = 0.57). In 2005, SD in AG-J was 0.96, compared to 0.23 in AG-IL, 0.19 in AG-IU and 0.14 in AG-H.

I calculated linear regressions of the mean PS within an amalgamated group vs. time (Table 4). The plots in AG-J converged significantly (P < 0.0001) as the mean similarity increased from 20 to 70%. There were no linear trends for plots within AG-IL. The similarities were always high (60 to 70%). A quad­ratic regression fit the trend of increasing similarities as species assembled, followed by decreasing simi­larities as dominance hierarchies formed at different rates (r² = 0.66; P < 0.002; t = 4.22). There were no trends in AG-IU, where similarities ranged from 51 to 61%. The trend among plots of AG-H was weakly positive (P < 0.0003). However, after 1991, when all plots had plants, there was no trend, with PS varying around 65%.

I used linear regressions of between-amalga­mated group similarity to determine if there was con­ver­gence at larger scales. There were six annual pair-wise comparisons. The increase in similarity between AG-J and both AG-I sub-classes was sig­nificant (Table 4), although final values were lower than PS between the two AG-I sub-classes. There were no significant changes between AG-J and AG-H, because PS remained very low. The similarity between each AG-I set and AG-H decreased over time, indicating divergence.
Discussion
With this long-term data set, I addressed ques­tions about patterns of structure, rates of vegetation development, trajectory convergence and transitions from assembly to inhibition phases during primary succession. I found consistent changes in structure (richness, diversity, percent cover) among the plots, but rates of change depended on elevation. Despite these structural changes, I found little evidence for floristic convergence.
Structure
After 26 growing seasons, the vegetation on Studebaker Ridge retains steep compositional gradi­ents over elevation, but community composition is changing only slowly as each plot has remained in its CC for several years. Richness stabilized by 2001 in all plots, similar to other Mount St. Helens sites (del Moral 2000). Although many plots were still quite barren, richness subsequently declined. Species that disappeared were short-lived species (e.g. Poly­gonum) or present only as seedlings (e.g. Abies and Cistanthe). Most species on this ridge are wind-dis­persed, yet Lupinus lepidus, dispersed by ants, plays a major role in many plots (Bishop, et al. 2005). Lupinus increases soil nitrogen and appears to have promoted grasses and mosses. Plots yet to experi­ence dense Lupinus populations may take much longer to reach levels of vegetation cover found in community class J and may undergo distinct trajecto­ries. The early establishment of Lupinus is the most important example of priority effect on this transect. Based on the pattern of development at lower eleva­tion (cf. del Moral & Rozzell 2005), the development of dense populations of Lupinus will almost surely govern both subsequent rates and trajectories.

Cover had stabilized only in the lowest plots where conditions are less severe (cf. del Moral & Lacher 2005), while limited soil, lower nutrients and greater exposure seem to inhibit cover accumulation at higher elevations. This occurred after species rich­ness stabilized (see also Wiegleb & Felinks 2001). Eventually, seedlings of conifers, which now occur sporadically, will establish and permit shade-tolerant species to establish. Though cover remained low at higher elevations, it too is increasing while richness has declined. Most common species present on the transect occur near all plots, most are wind-dispersed and all species found can occur above 1500 m (del Moral & Eckert 2005), so the pool of species on the

transect is similar throughout. However, increasing environmental stress is likely to restrict cover below values obtained at lower elevations on the transect. Diversity (H') is likely to continue the decline noted for each plot because dominance hier­archies will intensity and uncommon species will be eliminated. Grasses, Penstemon, Salix, mosses and perhaps Lupinus should become dominant if patterns reflect those of lower elevation (del Moral & Lacher 2005). If conifers become dominants, diversity should decline further (cf. del Moral & Ellis 2004) because, in this flora, fewer species thrive beneath conifers compared to meadows (del Moral 2000).

Richness, cover and diversity combine to describe overlapping assembly, maturation and inhi­bition phases in early primary succession. Assembly con­tinues until all probable colonists have established (cf. del Moral & Jones 2002). Maturation occurs as vegetation cover expands in the plot. During this phase, facilitation may permit the establishment of additional species (del Moral & Wood 1993). Diver­sity peaks at the end of assembly, then declines with differential cover expansion. While vegetation con­tinues maturing, facilitation of some species may continue, but the inhibition phase commences when restricted seedling establishment causes further declines of diversity, while cover remains stable (cf. Walker & del Moral 2003).
Rates
Rates of succession are critical measures of the community, reflecting species turnover and changes in the dominance hierarchy (Brown et al. 2006). Here, I used changes in richness, cover percentages and ordination scores over time to describe rates of succession. Cover alone can reflect mere recovery from disturbance, but in this case, it is an interesting measure of successional rates since the system started on barren surfaces. Ordination scores use species composition to provide a more comprehensive esti­mate of successional rates.

The slopes of richness vs. time for each plot were similar, indicating that elevation had a small effect in slowing establishment. In contrast, the accumulation of cover was substantially retarded at higher eleva­tions evidenced by sharply declining slopes of cover vs. time with increasing elevation. Within an eleva­tion range of only 260 m, the rate of succession dif­fers due to a shorter growing season and increased environmental stress. In the 25 years encompassed by this study, the effects of increasing elevation were qualitatively similar to effects of reduced succes­sional time-span. Although this rela­tionship will change substantially, it confirms that succession was delayed progressively up the slope.

DCA scores of lower plots changed more than higher ones. Mean annual DCA shifts in plots in AG-J were large until 2000. In contrast, AG-I con­tinued directional change to 2004, while AG-H changed little. The rate at which DCA scores of individual plots reached specific benchmarks de­clined with ele­vation, additional evidence that the development of high elevation vegetation was retarded at least by the shorter growing seasons (del Moral & Ellis 2004).

The classification produced clear patterns. Each plot developed sequentially, but at different rates. All initial community classes transitioned often and assembly and vegetation maturation were consistent across plots (Table 2). By 2005, there was a regular progression of classes (excluding CC-G) with the most mature (CC-J) found in the lowest plots, and the immature CC-H found in the highest plots. Lower plots had developed through more community classes than had upper ones.

These alternative ways to assess succession rates confirm that rates declined with elevation. At least some individuals of most species can establish as soon as there are suitable safe-sites, so richness pat­terns are similar throughout the transect. In con­trast, vegetation development is slowed by limited soil and less favorable conditions at higher eleva­tions. Thus, negative species interactions, now intense at lower elevations, have scarcely com­menced in the higher plots.
Convergence
The pattern of species establishment was similar at each plot, except that Lupinus was the first colonist at lower elevations, but was sparse in higher eleva­tion plots. Trajectories differed depending on early establishment of Lupinus. As in other habitats on Mount St. Helens (del Moral et al. 2005), mosses were not pioneers. Mosses require facilitation, after which they may form dense mats that preclude seed­ling establishment (del Moral & Rozzell 2005). In SR-8 to SR-13, a weak Lupinus invasion occurred in the mid-1990s, but dense vegetation has yet to form. Thus, the degree to which the relatively poorly dis­persed Lupinus lepidus has established is a determi­nant of the degree to which vegetation will converge.

Trajectories described by DCA exhibited several patterns. The likelihood of convergence decreased as distance between samples increased. When amalga­mated groups were analyzed using absolute cover, there was some evidence for conver­gence between AG-J and AG-IL, but AG-IU diverged from both, and AG-H, once similar to the others, also diverged. When relative cover was used to emphasize floristics, convergence was not found. The lack of Lupinus at higher elevation could be permitting alternative tra­jectories to form. Thus, given the rapid decline of similarity with distance, any convergence on this transect should occur only within short distances (del Moral 1998; del Moral & Ellis 2004), and divergence is possible.

Plots within a community class were spatially proximate and there was weak evidence for conver­gence within a CC. The SD of DCA scores for plots within amalgamated groups declined in AG-J and AG-H, but remained very high in the lower plots, indicating that these plots retained large floristic dif­ferences. The PS within AGs increased in AG-J, and slightly in AG-H, but they remained relatively low. There was no directional change in either AG-I sub­class, which suggests that trajectories of individual plots were changing in parallel, not converging. The PS between AGs weakly suggested incipient conver­gence, but the similarities between them were low. AG-J has very low similarity to the distant AG-H, while AG-IL is diverging from AG-H. Spatial effects alone will limit convergence between these AGs, and even convergence among plots within AGs will be partial, at most.

None of the approaches that I used in this study demonstrated either that vegetation within classes will become more homogeneous or that the classes will tend to merge as they mature. Limited conver­gence may occur within closely spaced samples (< 250 m), but even if such processes as priority effects were overcome, the environmental gradient and local variations would appear to preclude tight conver­gence.

Permanent plot studies such as this can be used to explore for assembly rules, but thus far, the time is too short to test specific predictions based on rules. However, this study does offer clues. Each plot has passed through a consistent sequence that demon­strated a network of trajectories with elements of convergence and divergence. Three initial CCs existed at the first sampling. A network of transitions followed to form different CCs that were sorted by elevation. Establishment started sooner and vegeta­tion grew more quickly at lower elevations than at higher elevations. If higher plots were to recapitulate the trajectories of lower ones, then CC-H would eventually develop into CC-J. However, dense populations of Lupinus have directed trajectories at lower elevation, so it is equally likely that trajectories will differ because Lupinus may never become com­mon at high elevation on this ridge. Local environ­mental conditions and priority effects also may pre­vent the occurrence of similar vegetation along this transect.

This study suggests that forces that may pro­duce vegetation convergence early in primary succes­sion are weak, while those that promote divergence (e.g. priority effects and intrinsic environmental dif­fer­ences) are strong. Trajectories of plots at higher ele­vations do not closely recapitulate those of lower ones, emphasizing that environmental severity affects trajectories. Weak evidence for convergence exists, but only among plots located in close proximity. Common space-for-time substitution methods used in toposequence studies may be even more suspect than commonly acknowledged.