ENVIREM: an expanded set of bioclimatic and topographic variables increases flexibility and improves performance of ecological niche modeling

Abstract

Species distribution modeling is a valuable tool with many applications across ecology and evolutionary biology. The selection of biologically meaningful environmental variables that determine relative habitat suitability is a crucial aspect of the modeling pipeline. The 19 bioclimatic variables from WorldClim are frequently employed, primarily because they are easily accessible and available globally for past, present and future climate scenarios. Yet, the availability of relatively few other comparable environmental datasets potentially limits our ability to select appropriate variables that will most successfully characterize a species’ distribution. We identified a set of 16 climatic and two topographic variables in the literature, which we call the ENVIREM dataset, many of which are likely to have direct relevance to ecological or physiological processes determining species distributions. We generated this set of variables at the same resolutions as WorldClim, for the present, mid‐Holocene, and Last Glacial Maximum (LGM). For 20 North American vertebrate species, we then assessed whether including the ENVIREM variables led to improved species distribution models compared to models using only the existing WorldClim variables. We found that including the ENVIREM dataset in the pool of variables to select from led to substantial improvements in niche modeling performance in 13 out of 20 species. We also show that, when comparing models constructed with different environmental variables, differences in projected distributions were often greater in the LGM than in the present. These variables are worth consideration in species distribution modeling applications, especially as many of the variables have direct links to processes important for species ecology. We provide these variables for download at multiple resolutions and for several time periods at envirem.github.io. Furthermore, we have written the ‘envirem’ R package to facilitate the generation of these variables from other input datasets.

Genomic evidence of a widespread southern distribution during the Last Glacial Maximum for two eastern North American hickory species

“…Briefly, occurrence records were obtained from the US Forest Service Forest Inventory Analysis Database (O'Connell et al., ), whereas environmental variables were obtained at 2.5‐arcminute resolution from the WorldClim 1.4. (Hijmans, Cameron, Parra, Jones, & Jarvis, ) and envirem (Title & Bemmels, ) databases. SDMs were constructed using Maxent 3.4.1 (Phillips, Anderson, Dudík, Schapire, & Blair, ; Phillips, Anderson, & Schapire, ; Phillips, Dudík, & Schapire, ) in the R package “dismo” (Hijmans, Phillips, Leathwick, & Elith, ), with models optimized according to best practices, following Title and Bemmels ().…”

“…(Hijmans, Cameron, Parra, Jones, & Jarvis, ) and envirem (Title & Bemmels, ) databases. SDMs were constructed using Maxent 3.4.1 (Phillips, Anderson, Dudík, Schapire, & Blair, ; Phillips, Anderson, & Schapire, ; Phillips, Dudík, & Schapire, ) in the R package “dismo” (Hijmans, Phillips, Leathwick, & Elith, ), with models optimized according to best practices, following Title and Bemmels (). Habitat suitability was projected for the LGM according to each of the CCSM4, MIROC‐ESM and MPI‐ESM‐P general circulation models (GCMs), but since projections were similar for all three GCMs, results were averaged into a single map.…”

Incorporating microclimate into species distribution models

“…Overlooking such microrefugia likely results in overestimations of future species' range shifts (Lenoir et al 2013). CHELSA (Karger et al 2017), WorldClim (Fick and Hijmans 2017), TerraClimate (Abatzoglou et al 2018) or ENVIREM (Title and Bemmels 2017)). CHELSA (Karger et al 2017), WorldClim (Fick and Hijmans 2017), TerraClimate (Abatzoglou et al 2018) or ENVIREM (Title and Bemmels 2017)).…”

Conventional methods for enhancing connectivity in conservation planning do not always maintain gene flow

“…We obtained data on five predictor variables to make spatial predictions of habitat suitability and landscape resistance to gene flow (Supporting Information Figure ). These variables included soil alkalinity (5 km resolution; Land Resources Management Unit, Institute for Environment and Sustainability, European Commission, Joint Research Centre, ), available growing conditions (i.e., number of days in the year with temperatures over 5 ° C; 2.5′ resolution; Title & Bemmels, ), annual precipitation (BIO12), and precipitation seasonality (BIO15; 30 ′′ resolution; http://worldclim.org; Hijmans, Cameron, Parra, Jones, & Jarvis, ) to account for potential habitat requirements (Alvarez et al., ; Meineri, Skarpaas, & Vandvik, ). Additionally, we obtained land cover data (500 m resolution corresponding to 2015; NASA LP DAAC, ) and reclassified it into three categorical classes representing: forest (1–5), human modified land (12–14), and other cover classes (6–11, 15–16; Figure b).…”