Discussion
We utilised a niche-based approach to investigate five hypotheses relating to factors influencing community stability. Our first hypothesis supported was that species richness will be higher if the site was located closer to the niche centre of the species pool. This was expected as niche-distribution modelling is dependent on the relationship between species niche-space and the probability of occurrence (Elith & Leathwick 2009). We add here that a simple aggregate score was informative about the species richness of our communities. This, however, leaves room for improvement when predicting niche effects on richness, such as utilising individual niche models to sum occurrence at each site, or alternatively, applying joint approaches that utilise species covariances to improve predictive performance (Poggiato et al. 2021). A spatial correlation structure also accounted for differences in species richness not accounted for by climatic niche characteristics. Other factors are, therefore, important for predicting richness such as latitudinal gradients in species richness (Hillebrand 2004), lower species richness of island faunas (MacArthur & Wilson 1967), and with smaller scale effects such as changing land cover.
That our second hypothesis was supported, lends the first large scale support for the abundant centre hypothesis (Andrewartha & Birch, 1954; Brown, 1984; Lawton, 1993) operating in butterfly communities. The effect size here was modest, however, we applied a simple aggregate metric (mean niche mismatch) and more variation in abundance might be explained for each species by niche mismatch. Similarly, increasing the dimensionality of our niche constructions, or accounting for other factors such as land use may better isolate the effect of niche position on abundance. We also found increased niche volume was associated with increased mean abundance. This suggests that in communities where species had larger niches, i.e. more generalist in terms of climate association, the populations were larger. This could be a reflection of the wide-scale declines that have been noted for specialist species (Clavel et al. 2011). However, studies of generalist-specialist trends in butterflies typically utilise species traits, including host plant association, and focus on single countries (Stefanescu et al. 2011; Dapporto & Dennis 2013; Eskildsen et al. 2015) and so linking these results must be cautious. Nevertheless, our results suggest that evaluating species trends in relation to climatic niches could be valuable, particularly given the recent, and projected, increase in extreme climatic events (Donohue et al. 2016; Ummenhofer & Meehl 2017). Finally, a path not predicted a priori, suggested communities with higher niche overlap had populations that were on average more abundant. It is possible, that after accounting for niche mismatch, a community with greater overlap means more species situated in optimal conditions producing a positive association between high overlap and abundance. However, we suggest further work is needed to resolve interactions between niche overlap, volume, and mismatch.
Results from our third hypothesis showed a significant negative association between species richness and synchrony, thereby supporting a key diversity-stability mechanism for butterfly communities (Thibaut & Connolly, 2013). That we found no effect of climatic niche-overlap on synchrony is surprising as butterflies are strongly influenced by weather (McDermott Long et al. 2017). This suggests that other factors, such as species traits and local adaptation, may also be needed to predict species responses to weather variation. Alternatively, the niches constructed here using yearly average weather apply space-for-time substitutions that might be too coarse to account for the weather events causing changes in population size (White & Kerr 2006). Our results suggest that other factors must generate asynchrony. One factor could be intra-specific density dependence of population growth rates (Roy et al. 2001) as density varies between species, asynchrony could occur even between species with similar climatic niches. We also noted additional associations, as niche overlap decreased and niche mismatch increased with species richness. The relationship with niche-overlap is perhaps expected as, so long as the niche shapes vary between species, then adding species will reduce overlap. The result with niche mismatch is harder to interpret, but could be due to sites with higher species richness include more species near the edges of the niche positions, or that the UK with a moderate climate and lower species richness may induce a niche-mismatch and species richness association.
Results from our fourth hypothesis were partly as predicted. Average population stability decreasing with niche mismatch is an extension of the abundant centre hypothesis, and natural populations at range edges have previously been shown to be more variable (Oliver et al.2012; Mills et al. 2017). However, we expected mean niche volume to increase average population stability as species with broader niches should tolerate a wider range of conditions leading to higher stability. It is possible that species with broader niches tolerate more marginal conditions leading to lower stability at a site while still being able to persist. Species with broad niches may also be more affected by competition, reducing the stability of the population. By comparison, specialists with narrower niches may only occupy favourable areas where they are generally more stable and relatively freer from inter-specific competition. Finally, we found no effect of mean abundance on average stability which is surprising due to expected mean-variance relationships (Taylor 1961; Kilpatrick & Ives 2003). However, mean abundances take no account of species weighting, whereas species stability was abundance weighted so that the evenness of the community might obscuring the effect.
Finally, we hypothesised synchrony and average population stability will explain differences in community stability (Doak et al. 1998; Tilman et al. 1998; Thibaut & Connolly 2013; Wang & Loreau 2014) was supported. The standardised effect size (Table 1) and the model fits (Figure 3e,f) suggest that asynchrony was, overall, having a larger impact on community stability as expected from theoretical considerations (Thibaut & Connolly 2013; Wang & Loreau 2014). In butterfly communities asynchrony may be particularly important as being r-selected – with high reproductive and interannual population growth rates – populations are characterised by high levels of population variability (Pianka 1970; May 1974), and community stability may be more influenced by how the dynamics of populations combine than differences in population stability. Mean abundance also increased community stability. This is expected from mean-variance scaling relationships, such as Taylor’s power law (Taylor 1961; Kilpatrick & Ives 2003) and we show this affects the stability of butterfly communities in the aggregate even if it was not detected at the individual level. In addition to asynchrony, we also noted that species richness increased stability through mechanisms not directly accounted for. This could be related to how species richness reflects the evenness of species. For example, if species-poor communities are dominated by one species in terms of absolute abundance, then even with high levels of asynchrony, the aggregate community abundance will largely reflect the population variability of the dominant species (Grime 1998).
Our approach demonstrates that consideration of hyper-dimensional niches and the derived metrics are useful for understanding population and community dynamics (Barros et al. 2016). To date, they have been applied more in plant communities (Enrique et al. 2018; Papugaet al. 2018), though their uses for predicting biogeography of populations has been recently demonstrated for birds (Osorio‐Olveraet al. 2020). However, there are some possible weaknesses in our approach. First, several metrics can be generated from niche-hyper-volumes (Mammola 2019) and for some hypotheses different metrics may have been more effective. For example, niche mismatch could consider minimum distances from niche volumes to site locations, and Mahalanobis distances could provide better measures than Euclidean distances from the centroids. Similarly, niche-overlap could be quantified using a range of metrics. However, we justify our efforts here as applying a straightforward approach that uses a minimum number of intuitive metrics to test key hypotheses regarding butterfly community stability. A second limitation of the niche approach is that the strength of any one event on any single niche dimension is not measured as directly. Consequently, identifying the threats of any class of extreme event such as drought (Oliver et al. 2013; De Palmaet al. 2017) may still require a single variable approach. Combining the approaches, for example tolerance to drought would be predicted to have lower scores on an axis related to precipitation-aridity, could provide a way to test and generate new hypotheses around the factors affecting the population dynamics of species.
In conclusion, we find support for the mechanisms purported to influence community stability operating in butterfly communities. We utilised metrics derived by niche hyper-volumes to provide a unique overview of the environmental drivers behind these mechanisms. Thus, our method provides a novel test of factors affecting the stability of terrestrial animal communities and demonstrates how considering niches can allow consideration of mechanisms, operating at a range of scales, that ultimately influence community stability.