Maintenance of community function through compensation breaks down over time in a desert rodent community

Abstract Understanding the ecological processes that maintain community function in systems experiencing species loss, and how these processes change over time, is key to understanding the relationship between community structure and function and predicting how communities may respond to perturbations in the Anthropocene. Using a 30‐year experiment on desert rodents, we show that the impact of species loss on community‐level energy use has changed repeatedly and dramatically over time, due to (1) the addition of new species to the community, and (2) a reduction in functional redundancy among the same set of species. Although strong compensation, initially driven by the dispersal of functionally redundant species to the local community, occurred in this system from 1997 to 2010, since 2010, compensation has broken down due to decreasing functional overlap within the same set of species. Simultaneously, long‐term changes in sitewide community composition due to niche complementarity have decoupled the dynamics of compensation from the overall impact of species loss on community‐level energy use. Shifting, context‐dependent compensatory dynamics, such as those demonstrated here, highlight the importance of explicitly long‐term, metacommunity, and eco‐evolutionary perspectives on the link between species‐level fluctuations and community function in a changing world.


Explanation
In order to calculate energetic compensation and the total energy ratio, we require an estimate for the baseline values of total energy use, kangaroo rat energy use, and small granivore energy use on control plots. Estimating these baselines requires aggregating over between-plot variability among the control plots. For consistency, in the main analysis, we also aggregate across the exclosure plots and focus on treatment-level means throughout. Here, we explore the effect of between-plot variability on our analyses, to the extent possible. We used treatment-level means across control plots to calculate energetic compensation and the total energy ratio, but calculated these quantities separately for each exclosure plot, and conducted analyses including a random effect of plot. We also conducted analyses of Dipodomys and C. baileyi proportional energy use using plot-level data, again including plot as a random effect. Results were qualitatively the same as using treatment-level means.

Model specification and selection
We fit linear mixed-effects models (using the lme function in the R package nlme; Pinheiro et al. 2021) of the form compensation ~ time period with a random effect of plot and temporal autocorrelation structure to account for autocorrelation between monthly census periods within each time period. We compared these to models without the autocorrelation structure, without the random effect, and without the term for time period. The best-fitting model included terms for time period, random effect of plot, and autocorrelation.

Model specification and selection
As for compensation, we fit linear mixed-effects models fitting total_energy_ratio ~ time period with a random effect of plot and a temporal autocorrelation term to account for autocorrelation between monthly census periods within each timeperiod. We compared these to models without the autocorrelation term, without the random effect, and without the term for time period. The best-fitting model included terms for time period, random effect of plot, and autocorrelation.  1988-1997 -1997-2010 -0.4146370

Model specification and selection
To compare proportional energy use across time periods, we used binomial generalized linear mixed models (using the glmer function in the R package lme4; Bates et al. 2015), which allowed us to include a random effect of plot.

Model specification and selection
As for kangaroo rat proportional energy use, we used a binomial generalized linear mixed effects model to compare C. baileyi proportional energy use across time periods. Because C. baileyi occurs on both control and exclosure plots, we investigated whether the dynamics of C. baileyi's proportional energy use differed between treatment types. We compared models incorporating separate slopes, separate intercepts, or no terms for treatment modulating the change in C. baileyi proportional energy use across time periods, i.e. comparing the full set of models: • cbaileyi_proportional_energy_use ~ timeperiod + treatment + timeperiod:treatment We also tested a null (intercept-only) model of no change across time periods: • cbaileyi_proportional_energy_use ~ 1 We compared all of these models with and without a random effect of plot.
We found that the best-fitting model incorporated a random effect of plot, and fixed effects for time period and for treatment, but no interaction between them (cbaileyi_proportional_energy_use ~ timeperiod + treatment). We therefore proceeded with this model.