Spatial and temporal variability of ecosystems has received a lot of attention in the last decades in the context of biodiversity loss, climate change and their impact on ecosystem functioning. Ecosystem functioning is known to vary i) among ecosystems due to the environmental context of each ecosystem, especially edaphic and climatic conditions ; ii) within ecosystems because individual communities respond differently to local conditions at the ecosystem scale [4, 42, 46, 50]; iii) inter-annually due to temporal variation in edaphic and climatic conditions [27, 50]; and iv) within communities because individual species respond differently to local conditions at the community scale [8, 17, 21, 26, 55]. In temperate and arctic regions, factors such as climate constrain the onset and offset of ecosystem processes . During the growing season, competition for resources induces species specific responses to environmental conditions which have been hypothesized to be driving the observed variation among communities within the ecosystems and dampening the effect of edaphic and climatic conditions [46, 52]. The relative importance of these different sources of variation on ecosystem functioning has yet to be quantified in natural ecosystems and the dampening effect of the diversity of plant assemblages remains to be investigated.
Plant phenology is a key functional trait of plant that links growth and reproduction events of plants to the functioning of ecosystems . Over large geographic extents, plant phenology is driven by the effect of climatic and edaphic factors on plant growth and stress tolerance [6, 18, 22]. In turn, plant phenology determines several ecological functions, such as pollination , herbivory  and carbon uptake . One main advantage of studying plant phenology is that the timing of biological events can be monitored at high spatial and temporal resolution through satellite or time-lapse imagery [10, 54].
Although climatic and edaphic factors were identified as important drivers of plant phenology at both large and small observational scales, recent studies also emphasized the importance of species richness and composition [9, 28, 32]. Species subjected to similar climatic and edaphic conditions tend to show large interspecific differences in their phenology [11, 25, 49, 51, 54]. For example, Wilsey et al.  compared grassland communities in northern latitudes and found that their growing season length differed by nearly 40 days. A study by Meng et al.  reported large inter-annual variations in the flowering sequence (i.e., ranking order) of 15 co-occurring plant species. Thus, the biodiversity of plant assemblages could be an important driver of plant phenology by introducing spatial variability and temporal asynchrony within ecosystems.
Land managers and conservation biologists use estimates of plant species richness to characterize temporal changes in the ecological dynamics of ecosystems. Yet, the role of plant species richness on the regulation of plant phenology was investigated in a few cases only. A lengthening of the growing season with increasing plant species richness was observed across six biogeographic regions of central Europe, independently of altitude and land-use descriptors . Rheault et al.  monitored 28 wetland plant communities and showed that the growing season length was, on average, 30 days longer in species-rich communities. However, the latter authors noted that the relationship between plant species richness and growing season length was contingent on the climatic conditions . Studies of plant phenology have yet to disentangle the relative importance of species richness, community asynchrony, and community temporal variance on the dynamics of ecosystems.
The coefficient of variation of an ecological function (e.g., aerial biomass, growing season length) measured on several occasions is a standard measure of temporal variation; i.e., the reciprocal of stability. Using this metric, it can be shown that three key variables determine the temporal variability of a plant community : i) species asynchrony, ii) species temporal variance and iii) species average functioning. Species asynchrony is a measure of how temporally de-correlated the functioning of each species is relative to the others in the community. The variability of a plant community will be low (i.e., stability will be high) if species asynchrony is high and if species temporal variance is low . The above principles can be scaled-up to the ecosystem level, such that the variability of an ecosystem is this time determined by: i) community asynchrony, ii) community temporal variance and iii) community average functioning. For the diversity of plant assemblages to stabilize the functioning of ecosystems, the expectation is that community asynchrony is an important determinant, while community temporal variance is comparatively less important. High spatial variability in the average functioning of communities is also stabilizing because it buffers differences among ecosystems.
The objective of this study was to evaluate whether the growing season phenology of plant communities within wetland ecosystems is mostly determined by climatic and edaphic factors (i.e., ecosystem identity), by the diversity of species within plant communities, or by the diversity of plant assemblages. The approach that we developed in this paper consists of partitioning the growing season phenology of plant communities (green-up and green-down dates, and growing season length) into five components using linear models: i) Ecosystem identity, ii) Community temporal variance, iii) Community average functioning, iv) Species richness and v) Community asynchrony. A schematic representation of the partitioning procedure for growing season length is illustrated in Fig. 1. Disentangling between these alternative scenarios is critical because they involve different management scales and policies. We also provide a direct test of the diversity-stability relationship using the species richness of individual communities as a measure of plant diversity and the coefficient of variation of plant phenology (e.g., green-up date) as a standard measure of temporal variation.