Integrating intraspecific differentiation in species distribution models: Consequences on projections of current and future climatically suitable areas of species

Thomas Lecocq,Oliver Schweiger,Pierre Rasmont,Alexander Harpke,Janine Bolliger

doi:10.1111/ddi.12916

Abstract

AbstractAimConventional species distribution models (SDMs) usually focus on the species level but disregard intraspecific variability. Phylogeographic structure and evolutionary significant units (ESU) have been proposed as pragmatic proxies to incorporate intraspecific differentiation in SDMs. Nevertheless, the efficiency of using these proxies in SDMs has been poorly investigated. We analysed how the projections of current and future climatically suitable areas can be affected when using ESU‐based or lineage‐based models compared to a species‐level model.LocationWest‐Palaearctic region.MethodsAs examples, we used three bumblebee species (133,787 observations). We assessed potential climatic niche differentiation between species level, lineages and ESUs, by determining the niche position and niche breadth for each classification level by outlying mean index analyses. Subsequently, we developed SDMs for each species and classification level separately using boosted regression trees prior to a comparison of their performances. Finally, we used the alternative models to project the extent of climatically suitable areas in 2070.ResultsWe found that in spite of highly similar overall model accuracy, integrating intraspecific variability significantly increases model sensitivity (i.e., better predicting presences) while decreased model specificity (i.e., over‐predicting the range). Consequently, both predictions of current and projections of future suitable conditions differed among the three approaches.Main conclusionsWe showed that although integrating lineage or ESU information did not improve the accuracy of conventional species‐level SDMs, it led to considerably different conclusions. As SDM‐based climatic risk assessments are increasingly used to help and improve conservation plans, divergences and limitations of each modelling approach should be taken into account for developing efficient biodiversity management strategies. Lineage and ESU‐based SDMs offer the advantage to draw attention to species in which allopatric populations could display physiologically different responses to climate change when they lead to different results than species‐based models.

Highlights

Species distribution modelling (SDM) is among the most widely used methods in macro‐ecology and conservation ecology (Mainali et al, 2015; Zurell et al, 2016)
We developed SDMs for each species and classification level separately using boosted regression trees (BRTs) (Elith, Leathwick, & Hastie, 2008) in r (R‐package gbm version 2.1.; Ridgeway, 2013) assuming a binomial error structure and using a learning rate of 0.001, a tree complexity of three, and a bag fraction of 0.75
We argue that these datasets can be considered reliable for SDM analyses

Summary

| INTRODUCTION

Species distribution modelling (SDM) is among the most widely used methods in macro‐ecology and conservation ecology (Mainali et al, 2015; Zurell et al, 2016). The definition of units below the species level based on solely neutral marker remains controversial because (a) DNA sequences analysed are often chosen arbitrarily (Cruaud et al, 2014) and (b) ecologically differentiated populations are not always characterized by the accumulation of many genetic differences (Ferguson, 2002; Salvato et al, 2002) These limitations can be overcome by combining different lines of evidence such as phylogeographic structure based on multi‐markers along with phenotypic traits and/or ecological features within the context of the integrative taxonomy (Schlick‐Steiner et al, 2010). This approach allows detecting conspicuous allopatric differentiated populations (Lecocq, Brasero, Meulemeester et al, 2015b) that are closely related to evolutionarily significant units (ESUs). We use three widespread bumblebee species (Bombus lapidarius, B. pascuorum and B. terrestris) with comparable distributions across Europe, for which phylogeography and ESU delimitations are available (Lecocq, Brasero, Martinet, Valterovà, & Rasmont, 2015; Lecocq, Coppée et al, 2016; Lecocq, Dellicour et al, 2015)

| METHODS

| DISCUSSION

Findings

| Limitations and relevance of the approach