Quantifying spatial variation in the size and structure of ecologically stratified communities

Abstract

Summary Including ecological specialization (e.g. functional guild) in analyses performed across regions can help to study how size and structure of communities vary across environmental gradients. Multi‐species occupancy models, and their extension to a multi‐region framework, represent useful tools for such gradient analysis based on functional traits. However, in these models species richness is only a derived parameter and therefore explicit relationships cannot be inferred. We provide a novel hierarchical multi‐region community model that allows for direct modelling of trait‐based patterns of species richness along environmental gradients by partitioning communities into ecologically relevant strata (e.g. guilds). We illustrate the flexibility of the model by simulations, where we estimate guild richness and community composition using both categorical and continuous species traits. In addition, we apply our model to data from eight avian communities to explore how guild richness varies across elevational gradients. Simulations indicate that the model is able to estimate relationships between guild richness and environmental gradients even when the number of regions is low. Applying the model to the bird community data set, we show how the number of species in four feeding guilds varies in different ways across an elevational gradient. Understanding large scale variation in species richness and how it relates to environmental gradients requires an explicit treatment of community structure. Our modelling approach allows the testing of hypotheses directly and simultaneously on strata‐ and community‐specific species richness while accounting for the issues of rarity and detectability, and can be used to investigate the mechanisms which shape large scale spatial variation in species richness.

Poor management in protected areas is associated with lowered tropical mammal diversity

“…Variation across regions in guild-specific richness is assumed to be a Poisson process where N g,r~P oiss(k g,r ) and k g,r is the expected guild-and region-specific richness. Following the model formulation of Tenan et al (2017), we marginalized over a binomial prior distribution (assuming M trials each having probability Ω r of occurring) to estimate the total number of species in each community, N r . We specified Ω r = (Σ g k g,r )/M and assumed a guild indicator variable g i,r~C at(p r ) that allowed the model to estimate guild membership for the augmented species, with p r = (p 1,r,.., p G,r ) and p g,r = k g,r /Σ g k g,r .…”

“…For example, we assume that a1 i,r~N (l a1,r , r a1,r ), where l a1,r is the mean community intercept (across species) and r a1,r is the standard deviation (among species). To account for the fact that rare species may be more difficult to detect, we included a correlation structure between occupancy and detection probability with region-specific correlation parameters q r (Dorazio & Royle, 2005; see Appendix S1 in Tenan et al, 2017).…”

“…We analyzed the data using a multi‐region community occupancy model (Tenan et al ., ). We structured our model to compare species richness and community composition of the sampled mammal communities between the two regions (Mwanihana and Uzungwa Scarp, r = 2) based on four trophic guilds ( g = 4), with the underlying model assumption that each species i = 1,2, …, n r (where n r is the number of observed species in each region) can be assigned to only one guild.…”

Landbird trends in protected areas using time‐to‐event occupancy models

“…We ran single species, multi-year occupancy models rather than hierarchical multi-species models (Sutherland et al 2016, Tenan et al 2017 because model selection and computation time were challenging with the multi-species models. Our model included a random effect for sampling site to account for repeated measures in occupancy over time.…”

“…Long-term monitoring programs across large spatial scales are required to understand how the direct and indirect effects of human activity affect wildlife populations. Occupancy models are frequently used to monitor changes in distribution (K ery et al 2013), relative abundance (Steenweg et al 2019), and species diversity (Sutherland et al 2016, Tenan et al 2017. They are also used to understand how factors such as climate change (Mor an-Ord oñez et al 2017), inter-specific competition (Wisz et al 2013, Staniczenko et al 2017, and industrial development affect wildlife populations (Johnson et al 2015, Rich et al 2017.…”