Negative biotic interactions drive predictions of distributions for species from a grassland community.

Phillip P. A. Staniczenko,K. Blake Suttle,Richard G. Pearson

doi:10.1098/rsbl.2018.0426

Abstract

Understanding the factors that determine species' geographical distributions is important for addressing a wide range of biological questions, including where species will be able to maintain populations following environmental change. New methods for modelling species distributions include the effects of biotic interactions alongside more commonly used abiotic variables such as temperature and precipitation; however, it is not clear which types of interspecific relationship contribute to shaping species distributions and should therefore be prioritized in models. Even if some interactions are known to be influential at local spatial scales, there is no guarantee they will have similar impacts at macroecological scales. Here we apply a novel method based on information theory to determine which types of interspecific relationship drive species distributions. Our results show that negative biotic interactions such as competition have the greatest effect on model predictions for species from a California grassland community. This knowledge will help focus data collection and improve model predictions for identifying at-risk species. Furthermore, our methodological approach is applicable to any kind of species distribution model that can be specified with and without interspecific relationships.

Highlights

Species’ distributions are commonly estimated using only abiotic environmental variables, but recent studies have shown that modelling biotic interactions can improve range predictions [1]
When using the maxSens threshold to convert habitat suitability values to binary ranges, only two of the eight models resulted in positive changes in data compression (i.e. DM . 0): the model representing nine negative biotic interactions led to a percentage change in total length of D%BI2 1⁄4 9.7%; and the model representing all 15 positive and negative biotic interactions led to D%BI 1⁄4 3.7%
The all interspecific relationships (ALL) model led to the largest absolute change, and the rank order of models followed the number of interspecific relationships represented in compression models; the single negative change resulted from the model representing six positive biotic interactions

Summary

Introduction

Species’ distributions are commonly estimated using only abiotic environmental variables, but recent studies have shown that modelling biotic interactions can improve range predictions [1]. For the posterior community distribution matrix, we used the same Maxent parameter estimates for species’ responses to bioclimate variables as above and modelled the effects of 52 interspecific relationships (classified from experiments [3] and long-term monitoring studies [4]; see electronic supplementary material for results for two alternative sets of interspecific relationships) on 14 focal species using a method that has been shown to improve range predictions for these species [1]. We began by calculating total lengths for the two matrices and a BN model representing no interspecific relationships: TLE,prior and TLE,posterior Such ‘Empty BN’ models have no conditional dependencies among species and provide a baseline measurement of the amount of information complexity inherent in a community distribution matrix. This means percentage changes are useful for investigating which types of interspecific relationships provide the most compression relative to their preponderance

Results

Discussion

Methods