Phylogenetic occupancy models integrate imperfect detection and phylogenetic signal to analyze community structure

Abstract

Biological communities are structured phylogenetically-closely related species are typically more likely to be found at the same sites. This may be, in part, because they respond similarly to environmental gradients. Accurately surveying biological communities is, however, made difficult by the fact that detection of species is not perfect. In recent years, numerous statistical methods have been developed that aim to overcome deficiencies in the species detection process. However, these methods do not allow investigators to assess phylogenetic community structure. Here, we introduce the phylogenetic occupancy model (POM), which accounts for imperfect species detection while assessing phylogenetic patterns in community structure. Using simulated data sets we show that the POM grants less biased estimates of phylogenetic structure than models without imperfect detection, and can correctly ascertain the effects of species traits on community composition while accounting for evolutionary non-independence of taxa. Integrating phylogenetic methods into widely used occupancy models will help clarify how evolutionary history influences modern day communities.

Phylogeny and species traits predict bird detectability

“…Phylogeny and traits might be used to define groups of species for which the combined detection probabilities can be estimated jointly (Alldredge et al 2007), or estimates available for closely related species can serve as a prior distribution in a Bayesian or empirical Bayes analytical framework, alleviating sample size limitations. A multi-species model could also utilize available traits and phylogenetic correlation, the statistical 'shrinkage' effect thus enabling more accurate inference for rare and cryptic species (Frishkoff et al 2017). Phylogeny based correlations with the tempering effect of λ should provide valuable information to multi-species occupancy and abundance models in general because phylogenetic information is available for many vertebrate species (Jetz et al 2012, Søren andSvenning 2015).…”

“…Understanding how variation in singing behaviour and signal transmission leads to imperfect detection is important for single species conservation and management (Kellner and Swihart 2014, Guillera-Arroita et al 2015, Alexander et al 2017, Guillera-Arroita 2017. There is growing evidence that species traits can be important predictors of detectability (Seoane et al 2005, Garrard et al 2013, Denis et al 2017 and that phylogenetic relationships can be utilized in joint community models that consider imperfect detection (Frishkoff et al 2017). Because avian singing behaviour and acoustic characteristics of songs are under directional selection (Thomas 1999), we hypothesized that species' detectability in birds might be predictable by species traits, phylogenetic relatedness, or both.…”

Integrating over uncertainty in spatial scale of response within multispecies occupancy models yields more accurate assessments of community composition

“…For example, it could be directly integrated into abundance models, or combined with flexible hierarchical modeling approaches to query population dynamics through time, or the effects of phylogenetic relatedness or interspecific competition on community structure (Yackulic et al 2014, Frishkoff et al 2017). Here, we take a similar approach to integrate scale selection into a series of models of greater complexity, using a parameter (or group of parameters) that explicitly estimates the spatial scales of response to environmental covariates.…”

“…While the approach here is demonstrated with a simple single-season occupancy model for multiple species, the internalized scale selection is broadly applicable. For example, it could be directly integrated into abundance models, or combined with flexible hierarchical modeling approaches to query population dynamics through time, or the effects of phylogenetic relatedness or interspecific competition on community structure (Yackulic et al 2014, Frishkoff et al 2017).…”

The importance of accounting for imperfect detection when estimating functional and phylogenetic community structure

“…Thus, the functional and phylogenetic distinctiveness of species may have opposing effects on the detectability of diversity (Jarzyna and Jetz 2016). Imperfect detection of species can also bias estimates of community structure (i.e., overdispersion, or clustering, relative to the structure of random communities generated from a regional species pool) when the process of detecting species is non-random (Frishkoff et al 2017). Specifically, we hypothesized that the structure of observed communities could appear more clustered than those of the actual composition of communities if undetected species are more likely to be functionally or phylogenetically distinct.…”