Fig 1.
Spatial extent and occurrence record locations of common vampire bats (Desmodus rotundus) in North America.
The spatial extent is indicated by grey background. Black dots represent 1,029 occupied pixels using 7,094 individual occurrence records for vampire bats in México. The white stars indicate fossil locations of vampire bats (genus Desmodus) known from the United States and Cuba.
Table 1.
Final set of five predictor variables used in modeling potential habitat suitability and distribution of common vampire bats (Desmodus rotundus) in North America under current climate and future climate projections.
Table 2.
List of the CMIP5 climate models used in this study.
Table 3.
Comparison of area under the curve (AUC), sensitivity, specificity, and average of sensitivity and specificity ((sensitivity + specificity)/2) statistics using test data and the cross-validation mean for each of 5 species distribution model (SDM) approaches using the final set of predictor variables in modeling potential habitat suitability and distribution of common vampire bat (Desmodus rotundus) in North America.
Table 4.
Variable importance using increase in area under the curve (AUC) statistic when each predictor variable is permuted using 5 species distribution model approaches to model patterns of habitat suitability and distribution of common vampire bats (Desmodus rotundus) in North America.
Fig 2.
Comparison of response curves for the highest-scoring predictor variable, minimum temperature of the coldest month (Bio6; °C) for each of the 5 species distribution modeling approaches used in this analysis of common vampire bats (Desmodus rotundus) in North America.
GLM = Generalized Linear Models; MARS = Multivariate Adaptive Regression Splines; BRT = Boosted Regression Trees; RF = Random Forest; and Maxent = Maximum entropy. The x-axis is minimum temperature of the coldest month (Bio6) and the y-axis is the estimated probability that a given value for Bio6 will result in suitable habitat for vampire bats, with all other variables in the model held at their mean values. The “Max” value indicated is the temperature at which the probability is maximized using a given SDM approach.
Fig 3.
Ensemble map for common vampire bats (Desmodus rotundus) in North America using binary prediction from 5 species distribution modeling approaches.
Occurrence records are indicated by red dots. The grey-scale indicates the proportion of the 5 model approaches that predict a given cell on the template as suitable habitat. Black indicates that all of the models predict the area to be suitable habitat, while white indicates that all of the models predict the area to be unsuitable habitat.
Fig 4.
Future climate ensemble map for common vampire bats (Desmodus rotundus) in North America using binary predictions from 5 species distribution model approaches projected into 17 downscaled IPPC5 (CMIP5) future climate projections (85 future models; see Table 4) using the 8.5 representative concentration pathway for 2070 (average for 2061–2080).
The grey-scale indicates the proportion of the 85 model predictions that considered a given cell on the map as suitable habitat in year 2070. Black indicates that all of the models predict the area will be suitable habitat, while white indicates that all of the models predict the area will be unsuitable habitat. Three possible routes of vampire bat and vampire-borne rabies dispersal into the ecoregions of the United States are also indicated: (1) via the Western Gulf Coastal Plain (GP) and Southern Texas Plains (TP) ecoregions in the México-Texas borderlands; (2) via Cuba and the Southern Florida Coastal Plain (FP) and Southern Coastal Plain (SCP) ecoregions in south Florida; and (3) via the Mandrean Archipelago (MA) ecoregion in New Mexico and Arizona (this route is not likely through year 2070, given the results of this analysis). Outlines of North American ecoregions are indicated, but only ecoregions associated with possible routes of dispersal are named using a 2 or 3 letter acronym.