Predicting ecosystem responses to disturbance events and environmental changes requires understanding the mechanisms that govern community assembly during primary succession and, thus, modulate biodiversity. According to our results, compositional changes during successions after the retreat of glaciers are shaped by both the addition and the replacement of taxa. Within 400 years after deglaciation, both mechanisms showed a similar contribution to compositional differences in plant communities (Fig. 2). Nevertheless, their contribution to total β-diversity varied over time. Immediately after glacier retreat, richness differences contributed more to β-diversity than replacement, as expected in harsh environments23. This suggests an overall predominant role of taxa addition in plant primary succession, but replacement becomes dominant after more than 50 years following glacier retreat.
The more the communities differed in age, the more dissimilar they were in terms of composition. Furthermore, the dissimilarity between communities with strong age differences was mostly driven by β-richness rather than β-replacement (Fig. 3b and d, Table S3). This matches the observed pattern of plant taxonomic accumulation from recently deglaciated terrains to late-successional stages13,35,36. Our results question studies advocating that severe environments are characterized by a constant initial floristic composition37 without changes in species composition over time resulting from the lack of the establishment of additional species (autosuccession)12,14. In fact, only 16% of the taxa detected with eDNA and 35% of taxa detected with traditional sampling persisted after 50 years of succession (Fig. S3). Such apparent incongruity could be explained by the ambiguity of the concept of “severe” or harsh” environment. These terms apply to limiting conditions both linked to regional climate (high altitude and/or latitude, and related climatic conditions) and occurring specifically on recently-deglaciated terrains, characterized by nutrient-poor soils and frequent geomorphological disturbances38. While the first conditions were reported to reduce species replacement ultimately leading to autosuccession12,14, the second ones can favor fast-growing species (ruderals28), which often include ubiquitous species39,40, but are subsequently replaced by species exhibiting more competitive strategies20,38,41. “Severe” environmental conditions thus include a wide array of ecological factors and relative responses with multifaceted effects along succession.
Moreover, total community dissimilarity is not only affected by the differences in ages between communities, but also by their mean age. The total β-diversity between communities in early successional stages was generally larger than between communities having similar age differences but being in late successional stages. Thus, our results point out that dissimilarity among sites decreases over time during succession. Considering that stochastically structured communities should exhibit divergent taxonomic compositions, our results suggest that community composition in early stages is strongly affected by initial conditions and/or stochastic processes (e.g., priority effects, probabilistic dispersal, and local extinction42), while deterministic processes (e.g., habitat filtering, competitive interactions) may drive more convergent community structures in late successional stages, converging with temporal observations from studies using permanent plots9,18,43.
When we compared communities in early successional stages (having on average less than ~ 50 years), richness differences contribute much more than replacement in determining the dissimilarity between communities (the credible intervals of the importance of the two components did not overlap for any method, Fig. 3a-d), while replacement becomes the dominant pattern for late-successional stages. This supports the hypothesis that the mechanisms driving succession after glacial retreat can change through time18,35,44,45. Immediately after glacial retreat, soils are generally nutrient-poor and affected by physical disturbances but early colonizers do not inhibit the establishment of new colonizing taxa1. This may be explained either by neutral interactions (due to the predominant role of the environment46 or of stochastic processes9,10) in these species-poor early stages or by facilitative interactions1, where the beneficiary species are not constraining the already established species47. However, the importance of taxa addition quickly decreased over the succession sequence, suggesting an increasing competition, as expected when environmental harshness decreases and species richness and cover increases23. In late successional stages, the stabilization of resources and terrains can reduce the strength of environmental filtering, and the most competitive taxa may replace the least competitive ones. Such substitutions are likely driven by biotic interactions, either because early arrivers modify the environment making the conditions suitable for other colonizers but less suitable for them1 or because later successional species exploit similar resources and outcompete the already established early species47.
We found a consistent temporal trend of beta-diversity and its components across 46 chronosequences around the world, suggesting common features in the changes in the assembly rules underlying plant successions, despite existing variability across forelands. Local and regional environmental conditions are known to strongly influence the trajectory and rate of plant successions12,14,35,46,48. Forthcoming studies using a global dataset covering a wide geographical scale, such as the one used here, will allow assessing the local and regional drivers that shape successional trajectories. The eDNA metabarcoding method enables rapid, cost-efficient, and standardized biodiversity assessment over broad geographical and taxonomic scales, yielding data that would have been challenging to assemble with traditional methods. Like all sampling approaches, eDNA has some limitations such as missing the rarest taxa, not providing estimates of absolute biomass, and limited taxonomic resolution49. Nevertheless, eDNA yielded temporal patterns of β-diversity strongly consistent with traditional methods, suggesting that our results are robust to these methodological limitations.
Threshold dynamics have been proposed during the biotic colonization of glacier forelands, with a fast increase in community alpha richness during the first 60 years followed by a plateau and a decline in β-diversity18,41. However, in our study, the global trends of β-diversity and its components during succession were characterized by linear relationships over time without significant breakpoints (Table S4), in agreement with Clements's views of successions as progressive, linear changes. It should be noted that our plant communities did not reach a stable point in the considered time frame. Even in our late successional communities, the total β-diversity remained substantial and β-richness remained well above zero. Our sampling was focused on terrains deglaciated since the Little Ice Age (LIA, after ca. 1850 for the majority of forelands), whose ages may remain too young for observing a stabilization of the composition of plant communities. Longer time series (thousand years) would be required for a complete understanding of β-diversity changes and their drivers20,50, especially to identify whether and when β-diversity changes stop.
The robustness of conclusions about successions obtained from the analysis of chronosequences has been questioned, given that sites placed at different distances from the glacier margin can differ from each other due to microhabitat conditions or the identity of first colonizers (priority effects) and might thus potentially follow different trajectories30,51. Nevertheless, the analysis of temporal data from permanent plots yielded temporal patterns of β-richness and β-replacement highly consistent with our results (Supplementary Note1, Fig. S4, Table S5), confirming the robustness of conclusions from the chronosequence approach. Importantly, the combination of the chronosequence approach and eDNA sampling allowed readdressing old ecological questions left behind due to the lack of properly standardized and comprehensive datasets. Even though we find common successional trends among different regions of the world, the temporal patterns of total β-diversity and its components highlight the complexity of plant successions after glacial retreat, where the drivers of community assembly changed over time. The debate on processes shaping succession continues since the onset of community ecology, with both stochastic6 and deterministic4 processes pinpointed as key drivers during the last century5. Our broad-scale study suggests that both processes have a fundamental role in community composition changes, with neutral and/or positive interactions dominating compositional variations in early successional communities, while competition becomes more important in late successional communities.
Today, glaciers are retreating at an unprecedented rate, and plant communities have an important role in ecosystems developing after deglaciation13. The temporal changes of compositional drivers are expected to go beyond plant communities, thus affecting taxa interacting with plants through pollination, mutualism, herbivory, or parasitism13,52. Understanding how β-diversity measures co-vary across different components of communities will be a key challenge in predicting the long-term consequences of climate change on ecosystems53.