Discussion
Based on a globally coordinated experiment in grasslands, our analysis demonstrates that (1) nutrient addition and herbivore exclusion mainly had additive effects, with nutrient addition consistently reducing stability at the local and larger spatial scales, while herbivore exclusion weakly reduced stability at both scales; (2) nutrient addition reduced stability primarily by increasing temporal community dissimilarity and decreasing species richness and evenness. In contrast, herbivore exclusion reduced gamma stability mainly by reducing spatial asynchrony, but also weakly by decreasing local species richness. Temporal and spatial community dissimilarity was mainly attributed to balanced variation (i.e. change in relative abundance among species but not total abundance in communities across time or space), pointing at the importance of turnover driven by species replacement in determining grassland stability.
In contrast to our hypothesis 1, our analysis provides weak support for interactive effects of nutrients and herbivores on stability across spatial scales and other plant community properties (additive for plant diversity and stability but a synergistic effect for spatial asynchrony). Previous analyses using different subsets of NetNut data looking at the joint effects of nutrient and herbivores on species richness and aboveground biomass also found weak interactive effects (Borer et al. 2014b, 2020). Several non-exclusive processes may explain the lack of the interactive effects of nutrients and herbivores found here. First, relatively low grazing intensity at many sites. Indeed, we found that the effects of herbivores on spatial asynchrony and gamma stability tended to be more apparent at sites with high grazing intensity under the ambient conditions (Fig. S5). Second, duration of the current experiment may not be long enough to capture changes in plant communities (Chen et al. 2019, 2020). Boreret al . (2020) found that the interactive effects of nutrient addition and herbivore exclusion on aboveground biomass became stronger using 8-10 years post-treatment data compared with those using 2-4 years post-treatment data. Third, the effects of nutrients and herbivores on plant communities may act at different spatial scales, where nutrient addition influences local communities and herbivores modulate spatial heterogeneities (see next paragraph). Consistent with previous analyses (Zhang et al. 2019; Hautier et al. 2020), our results show that the negative effects of nutrient addition alone cascaded to larger spatial scales. Herbivore exclusion alone had weak negative effects on stability at the two scales investigated. Again, this may be due to low grazing intensity at many sites (Table S1).
Confirming our hypothesis 2, nutrients and herbivores impacted gamma stability via different pathways. Nutrient addition reduced alpha and gamma stability probably because nutrient addition intensified interspecific competition within local communities, causing a decline in alpha diversity, a decrease in evenness, and an increase in temporal community dissimilarity. Moreover, increased temporal community dissimilarity contributed to decreased alpha and gamma stability as much as the combined effects of decreasing alpha diversity and evenness. This confirms previous results showing a stronger contribution of temporal community dissimilarity to alpha stability compared to alpha diversity (Koerner et al. 2016). In contrast, herbivore exclusion weakly decreased gamma stability primarily by decreasing spatial asynchrony, and weakly by decreasing alpha diversity. It suggests that the effects of herbivores may be more apparent at the larger spatial scale probably by impacting vegetation heterogeneity, particularly in aboveground biomass (Glenn et al. 1992; Howison et al. 2017). Our results suggest that maintaining stability from nutrient addition and herbivore exclusion in grasslands needs to focus on different processes across spatial scales.
Additionally, nutrient addition decreased alpha and gamma stability via increasing temporal balanced variation, while its negative effect on temporal abundance gradients does not translate into changes in stability. In contrast, the effects of herbivore exclusion on spatial asynchrony and gamma stability were not related to spatial community dissimilarity, which is consistent with previous analyses (Wilcoxet al. 2017; Zhang et al. 2019). Community dissimilarity across time is commonly used as an index of compositional stability (Hillebrand et al. 2018; Hillebrand & Kunze 2020; White et al. 2020) and higher compositional stability has been suggested to lead to higher biomass stability (Allan et al. 2011). Thus, the relationship between compositional stability and biomass stability may depend on the spatial scale considered. This necessitates looking at multidimensional stability (Donohue et al. 2013). As cover is usually easier to measure in the field and less destructive compared with biomass harvesting, many researchers evaluate stability based on total cover (i.e. mean of total cover through time divided by its standard deviation; e.g. Post 2013; Beck et al. 2015; Blüthgenet al. 2016; Wilcox et al. 2017). Our results suggest that using cover data to calculate stability may fail to capture changes induced by balanced variation. For instance, two communities (in different years or places) can differ markedly in biomass due to species replacement even when their total cover remains the same (Fig. S2E). As a result, a cover-based metric of stability may overestimate ecosystem stability relative to its biomass-based counterpart (Fig. S7). That said, community dissimilarity (and its partitioning) serves as a useful index to predict biomass stability, but spatial scales need to be considered.
In this study, we regarded aggregated local communities within treatments across blocks as “larger spatial scale” following previous analyses (Chalcraft et al. 2008; Wilcox et al. 2017; Zhanget al. 2019; Hautier et al. 2020). However, subplots belonging to different treatments within blocks are closer to each other compared with subplots belonging to the same treatments across blocks (i.e. larger spatial scale), thus dispersal may be stronger for subplots within blocks than the larger spatial scale used here and dispersal may bias the results. We argue that dispersal may not influence our results for two reasons. First, blocks typically spread within a relatively small area at each site (around 1000 m2), thus all subplots within sites may be connected by dispersal (Zhang et al.2019). Second, we found that subplots belonging to different treatments within blocks had relatively higher spatial community dissimilarity (0.61 average across sites) than those belonging to the same treatments across blocks (0.55, 0.58, 0.53, 0.57 for control, fence, NPK, and NPK+fence). This suggests that treatments are the dominant filter for plant community assembly.
Our results—based on 34 grasslands across four continents—advance our knowledge in that (1) nutrients and herbivores mainly have additive effects on stability in grasslands; (2) nutrients and herbivores impact stability across multiple spatial scales through different pathways, wherein turnover driven by species replacement is more important than species richness in determining grassland stability. Our results point to the need to reduce nitrogen deposition while preserving or reintroducing herbivores to ensure the stable provisioning of grassland biomass. More importantly, our results highlight that maintaining grassland stability in the face of increasing nutrient addition and herbivore extirpation requires a multi-scale framework to disentangle the influences of processes operating at different scales to guide conservation and management practices.