Varying dataset resolution alters predictive accuracy of spatially explicit ensemble models for avian species distribution.

Claire M Curry,Eli S Bridge,Jeremy D Ross,Andrea J Contina

doi:10.1002/ece3.4725

Abstract

Species distribution models can be made more accurate by use of new “Spatiotemporal Exploratory Models” (STEMs), a type of spatially explicit ensemble model (SEEM) developed at the continental scale that averages regional models pixel by pixel. Although SEEMs can generate more accurate predictions of species distributions, they are computationally expensive. We compared the accuracies of each model for 11 grassland bird species and examined whether they improve accuracy at a statewide scale for fine and coarse predictor resolutions. We used a combination of survey data and citizen science data for 11 grassland bird species in Oklahoma to test a spatially explicit ensemble model at a smaller scale for its effects on accuracy of current models. We found that only four species performed best with either a statewide model or SEEM; the most accurate model for the remaining seven species varied with data resolution and performance measure. Policy implications: Determination of nonheterogeneity may depend on the spatial resolution of the examined dataset. Managers should be cautious if any regional differences are expected when developing policy from range‐wide results that show a single model or timeframe. We recommend use of standard species distribution models or other types of nonspatially explicit ensemble models for local species prediction models. Further study is necessary to understand at what point SEEMs become necessary with varying dataset resolutions.

Highlights

Species distribution modeling (SDM) is a tool that uses envi‐ ronmental and geographic variables to predict what areas are suitable for a species and to better understand what factors con‐ strain species’ ranges (Elith & Leathwick, 2009)
We focused on species of grassland birds found during our general surveys [Northern Bobwhite (Colinus virginianus); Upland Sandpiper (Bartramia longicauda); Horned Lark (Eremophila alpestris); Cassin’s Sparrow (Peucaea cassinii); Field Sparrow (Spizella pusilla); Lark Sparrow (Chondestes grammacus); Grasshopper Sparrow (Ammodramus savannarum); Dickcissel (Spiza americana); Eastern Meadowlark (Sturnella magna); and Western Meadowlark (Sturnella neglecta)], plus the obligate brood parasite Brown‐headed Cowbirds (Molothrus ater) for which presence often depends on land use factors (Benson, Chiavacci, & Ward, 2013), for a total of species
spatially explicit ensemble model (SEEM) increase model accuracy over continental scales (Fink et al, 2013, 2010), our study found their performance differed by species and predictor resolution even in a state with variable cli‐ mate and diverse ecoregions

Summary

| INTRODUCTION

Species distribution modeling (SDM) is a tool that uses envi‐ ronmental and geographic variables to predict what areas are suitable for a species and to better understand what factors con‐ strain species’ ranges (Elith & Leathwick, 2009). Newer regression and machine learning techniques incorporated into SDM continue to increase prediction accuracy (Cutler et al, 2007; Elith, Leathwick, & Hastie, 2008; Elith et al, 2006; Lorena et al, 2011; Phillips, Dudík, & Schapire, 2004) One such method, Spatiotemporal Exploratory Modeling (STEM), has recently been introduced as a means of coping with variation in regional drivers. Forest species, which likewise occupy a single habitat type, show spatial and temporal variation in predictor importance (Zuckerberg et al, 2016) Such a technique has been used on shorebirds in habitats with structural similarity to grasslands at a statewide scale (Johnston et al, 2015). Our results will allow others to make decisions on whether increased accuracy in modeling is worth the additional computational effort required by newer modeling techniques and provide guidance for future work into where given modeling ap‐ plications are useful

| METHODS

Findings

| DISCUSSION