Skip to main content

Niche differentiation in the bank vole: Maxent input and output data

Cite this dataset

Escalante, Marco; Horníková, Michaela; Marková, Silvia; Kotlik, Petr (2022). Niche differentiation in the bank vole: Maxent input and output data [Dataset]. Dryad.


Species-level environmental niche modelling has been crucial in efforts to understand how species respond to climate variation and change. However, species often exhibit local adaptation and intraspecific niche differences that may be important to consider in predicting responses to climate. Here, we explore if phylogeographic lineages of the bank vole originating from different glacial refugia (Carpathian, Western, Eastern and Southern) show niche differentiation, which would suggest a role for local adaptation in biogeography of this widespread Eurasian small mammal. We first model the environmental requirements for the bank vole using species-wide occurrences (210 filtered records) and then model each lineage separately to examine niche overlap and test for niche differentiation in geographical and environmental space. We then use the models to estimate past [Last Glacial Maximum (LGM) and mid-Holocene] habitat suitability to compare to previously hypothesized glacial refugia for this species. Environmental niches are statistically significantly different from each other for all pairs of lineages in geographical as well as environmental space and these differences cannot be explained by habitat availability within their respective ranges. Together with the inability of most of the lineages to correctly predict the distributions of other lineages, these result support intraspecific ecological differentiation in the bank vole. Model projections of habitat suitability during the LGM support glacial survival of the bank vole in the Mediterranean region as well as in central and western Europe. Niche differences between lineages and the resulting spatial segregation of habitat suitability suggest ecological differentiation has played a role in determining the present phylogeographic patterns in the bank vole. Our study illustrates that models pooling lineages within a species may obscure the potential for different response to climate change among populations.


Environmental data
Raster layers representing 19 bioclimatic variables were downloaded from the WorldClim ( v.1.4 dataset (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005), at 30 arc-second resolution, and were clipped to the boundaries of the study area.  To account for climate modelling uncertainties (Schorr, Holstein, Pearman, Guisan, & Kadereit, 2012), two General Circulation Models (GCM) of past climate, CCSM4 (Gent et al., 2011) and MIROC-ESM (Watanabe et al., 2011), were used for the LGM (about 22 kyr ago) and mid-Holocene (about 6 kyr ago).

Environmental Niche Modelling
The ENM was performed using the maximum entropy approach implemented in MaxEnt, v.3.4.1 (Phillips, Dudík, & Schapire, 2004).
Two different sets of predictor variables were used to ascertain the robustness of the results (Araújo et al., 2019). Both sets were selected based on a reiterative jackknife procedure of model construction and stepwise removal of the least contributing variables (Zeng, Low, & Yeo, 2016), but each used a different data set and metric (i.e. training gain or test gain) to measure the contribution of the variables to the model. After removing the uninformative variables by the jackknife procedure, the final set of variables was produced in each case by removing one variable from each pair of correlated variables, based on a correlation matrix of the climate layers (cut-off of r ˂ 0.8; Merow, Smith, Silander, Merow, & Silander, 2013). The first set of predictors (Set 1) was selected based on the training gain using the species-level occurrences as the training data set and it contained Mean Diurnal Range (BIO 2), Temperature Annual Range (BIO 7), Mean Temperature of Wettest Quarter (BIO 8), Mean Temperature of Driest Quarter (BIO 9), Precipitation Seasonality (BIO 15), Precipitation of Wettest Quarter (BIO 16), Precipitation of Driest Quarter (BIO 17) and Precipitation of Warmest Quarter (BIO 18). The second set (Set 2) was selected using Western Siberia as an independent test area. The variables included in Set 2 based on the test gain were Isothermality (BIO 3), Mean Temperature of Warmest Quarter (BIO 10), Precipitation of Wettest Quarter (BIO 16) and Precipitation of Warmest Quarter (BIO 18).
Niche models were built independently with Set 1 and Set 2. A total of 50 replicates of each model were generated by the subsampling method in MaxEnt, which randomly selected 25% of the occurrence points reserved as test data (Phillips et al., 2004). Subsequently, the estimates are based on the overall mean of the replicates. To minimize the possible effect of inadequate representation of the environmental background (Guevara, Gerstner, Kass, & Anderson, 2018), 1,000,000 background points were used. Default values were used for the other parameters, as recommended when comparing models at different evolutionary levels (i.e. species and intraspecific lineages) and with different sampling efforts (Merow et al., 2013; Phillips & Dudík, 2008).
Our training area encompasses the bank vole distribution range and surrounding areas, covering Europe, western Siberia, the Anatolian Peninsula and the Caucasus. 
The models were projected to the CCSM4 and MIROC-ESM bioclimatic layers for the LGM and mid-Holocene. The logistic format and ascii type were used to generate the raster output. 


Czech Science Foundation, Award: 16-03248S

Czech Science Foundation, Award: 20-11058S

Ministry of Education, Youth and Sports of the Czech Republic, Award: EXCELLENCE CZ.02.1.01/0.0/0.0/15_003/0000460 OP RDE

Ministry of Education, Youth and Sports of the Czech Republic, Award: EXCELLENCE CZ.02.1.01/0.0/0.0/15_003/0000460 OP RDE