Combining phylogeny and co‐occurrence to improve single species distribution models

Abstract

Aim We present a novel quantitative framework that combines information on phylogeny and the spatial distributions of related species to enhance the single‐species distributional models commonly used in ecology. Innovation While species distribution models (SDMs) are becoming increasingly sophisticated, they rarely take into consideration the shared evolutionary histories of species. Species are not independent entities, and phylogenies may capture how species have configured their spatial distributions as a response to their ecological similarities and interactions in space and through evolutionary time. Our framework provides a flexible approach to include phylogenies as a surrogate for missing trait data within SDMs, and may be particularly valuable for disentangling current and historical drivers of species distributions and modelling data‐poor species when their close relatives are better sampled. Main conclusions Using both simulations and empirical examples, we demonstrate how the inclusion of phylogenetic information can significantly improve the fit of species distribution models by up to 30% for certain species. We show that the potential of phylogenies to improve model fit directly relates to the phylogenetic structure of species distributions and suggest that our framework has the potential to reconcile the apparent conflict between current and historical drivers of biodiversity patterns.

Estimating the population size of lemurs based on their mutualistic food trees

“…We have taken a straight‐forward approach to incorporate species interactions: using the model of one type of interactor in addition to the environment to model distribution and abundance of the other interactor. New methods that model the distributions of multiple species jointly and incorporate phylogenetic relatedness allow the potentially negative or positive effects of species on each other to inform the model (Morales‐Castilla, Davies, Pearse, & Peres‐Neto, ). We were interested in the one‐way effects of lemur food tree distributions on lemur abundance, however, and thus to explore these interactions directly we investigated each lemur species separately.…”

Combining Phylogenetic and Occurrence Information for Risk Assessment of Pest and Pathogen Interactions with Host Plants

“…We suggest that this step is critical in anticipating potential interactions; from our analysis, we can conclude that the potential geographic distribution depends on, or is a consequence of, the number and type of interactions (i.e., according to host range and level of phylogenetic constraint; also see [23,24]). An extension of this approach could be to improve single species distribution models [25] by including both the customary environmental information and species interactions. Moreover, since this index may be interpreted as a suitability index for predicting ecological interactions, it could also be interpreted as summarizing the parts of the realized ecological niche of the species related to the Eltonian ecological niche [26,27].…”

Predicting potential current distribution of Lycorma delicatula (Hemiptera: Fulgoridae) using MaxEnt model in South Korea

“…In addition, MaxEnt creates a distribution model, using regularization multiplier (default value = 1), which mitigates model complexity or overfitting, to make general interpretation [36]. Nevertheless, MaxEnt does not always create the best model by a given default parameter setting [38]. Therefore, to select the best model we adjusted the parameters setting and developed 24 candidate models with different feature combinations and regularization multipliers by using four feature combinations (LQ, LQP, LQH, LQPH) and six regularization parameters (1, 2, 5, 10, 15, 20).…”