The effects of temperature and dispersal on species diversity in natural microbial metacommunities

Dispersal is key for maintaining biodiversity at local- and regional scales in metacommunities. However, little is known about the combined effects of dispersal and climate change on biodiversity. Theory predicts that alpha-diversity is maximized at intermediate dispersal rates, resulting in a hump-shaped diversity-dispersal relationship. This relationship is predicted to flatten when competition increases. We anticipate that this same flattening will occur with increased temperature because, in the rising part of the temperature performance curve, interspecific competition is predicted to increase. We explored this question using aquatic communities of Sarracenia purpurea from early- and late-successional stages, in which we simulated four levels of dispersal and four temperature scenarios. With increased dispersal, the hump shape was observed consistently in late successional communities, but only in higher temperature treatments in early succession. Increased temperature did not flatten the hump-shape relationship, but decreased the level of alpha- and gamma-diversity. Interestingly, higher temperatures negatively impacted small-bodied species. These metacommunity-level extinctions likely relaxed interspecific competition, which could explain the absence of flattening of the diversity-dispersal relationship. Our findings suggest that climate change will cause extinctions both at local- and global- scales and emphasize the importance of intermediate levels of dispersal as an insurance for local diversity.


Fig. S3
Average per tube ("alpha diversity"; black circles) and per treatment ("gamma diversity"; grey circles) and total species richness (white circles) as a function of sampling week in early-(left panel) and late-successional (right panel) communities.
Error bars indicate ± one standard error. Total species richness is the cumulative number of species over all treatments in a given week. In both early-and late-successional stages, alpha-diversity decreased markedly with time, which can be captured by a hyperbolic relationship (fitted solid lines). Gamma-diversity (total number of species in the 5 tubes forming a metacommunity) also showed a negative trend, but with weaker decrease during the first sampling weeks (dashed lines describe a linear and a quadratic relationship for early-and late-successional stages, respectively).
There was no evidence of a decrease in total species richness with time in both successional stages; the variability was attributable to rare species being undetected.  Given is the best model after model selection based on AIC, starting with a model containing all interactions between the treatment variables (succession, temperature, and the linear and quadratic terms of dispersal). Analyses were performed at the tube level. Sampling week and total density were included as covariates. Sampling week was considered as a continuous variable and, to account for the sharp decrease in alpha diversity with sampling weeks (see Fig. S3), was reciprocally transformed (week -1 ). To account for repeated measures, we modelled correlation between observations with an AR1 approach. Given is the best model after model selection based on AIC, starting with a model with all interactions between temperature and the linear and quadratic terms of dispersal. Both response variables were log transformed to satisfy normality assumption; this assumption was not reached for total density in the late succession and the p-values must be interpreted with caution. Sampling week (reciprocally transformed for evenness) and total density were included as covariates.
Repeated measures were modelled with an AR1 correlation approach between observations. Both diversity indices were measured at the treatment level (4 temperature x 4 dispersal) at each sampling week. Beta-diversity was Box-Cox transformed. Sampling week and total density were included as covariates. Sampling week was considered a continuous variable, with a quadratic term necessary for beta-and gamma-diversity in the late-successional stage (see Fig. S3). To account for repeated measures, we modelled correlation between observations with an AR1 approach.

Fig. S4
Evenness, total density, gamma-diversity, and beta-diversity as a function of dispersal rate for the four temperature levels and in the two successional stages.  Taxonomic status (see Table S1) was used as a random factor (the Bdelloid rotifer is excluded from the analysis). Size was entered as an ordered variable (ordered by increasing body size: small to medium to large). The significant linear term indicates that small morphospecies experienced higher extinction rates. A logistic regression performed exclusively for the Flagellata morphospecies confirms this result (for this analysis, size was considered a quantitative variable with category small=1, medium=2, and large=3). Fig. S5 represents the proportions of extinction in the three size categories.  variable was the number of small-vs. the number of medium-plus large-sized morphospecies in each tube at each sampling week. Medium-and large-sized species were pooled as their response to the treatments were qualitatively similar (see Fig. 2 in the main text). To account for repeated measures, we used tube identity as a random factor. The results for the change in proportions with sampling weeks as a function of temperature and dispersal in early-and late-successions are shown in Fig. S6 and S7. Legend: 0 = no dispersal; 1 = low dispersal; 2 = medium dispersal; 3 = high dispersal. Proportion of species is expressed as percentages within each tube. Boxes as in Fig. S5. dispersal for the late-successional stage.