Fig 1.
South Converse (Wyoming, USA) study area displaying GPS points from 122 collared mule deer.
Gray shading indicates elevation, with darker gray indicating higher elevation. The towns of Casper and Douglas are indicated with black circles. Data for elevation, road, and river were obtained from publicly available data sources [38–40]. Map was created using R software [41] including packages ‘ggplot2,’ ‘ggspatial,’ ‘maps,’ ‘cowplot,’ and ‘terra’ [42–46].
Table 1.
Environmental variables used to model the risk of mule deer newly testing positive for chronic wasting disease (CWD) based on habitats visited by the deer. Variables fall within the categories: agriculture, human development, water sources, topography, soil, and vegetation. Variables included in risk models exploring hypothesis 3 were expected to affect risk by controlling habitat suitability (either attracting or deterring mule deer from using certain habitats), and variables included in risk models exploring hypothesis 4 were expected to affect risk by increasing or decreasing prion persistence in the environment. Spatial resolution of all data sets is 30m unless otherwise stated.
Table 2.
Numbers of GPS-collared mule deer used in models predicting relative risk of a deer newly testing positive for chronic wasting disease based on habitat visited by the deer. Deer individuals were identified by genotype and were either homozygous for serine at codon 225 (“SS” genotype), or were heterozygous/homozygous for phenylalanine at codon 225 (“*F” genotype). All individuals included in models did not test positive for chronic wasting disease (CWD) on first capture. The total number of individuals in each genotype group is further broken down into individuals that never tested positive for chronic wasting disease for the duration of the study (“Remained CWD-“) and those that had tested positive by the end of the study (“Became CWD+”). All individuals in this analysis are female.
Table 3.
Comparison of models predicting probability a mule deer newly tested positive for chronic wasting disease (CWD) in a one-year period. Models listed are the top models representing each of five hypotheses: 1) null hypothesis (non-environmental covariates), 2) deer density hypothesis (containing kernel density estimate as a covariate), 3) habitat suitability hypothesis (containing environmental covariates related to habitat suitability), 4) prion persistence hypothesis (containing environmental covariates related to prion persistence), and 5) combination hypothesis (containing environmental covariates related to both habitat suitability and prion persistence). K indicates the number of parameters in each model, and AICc refers to Akaike’s Information Criterion for small sample sizes.
Fig 2.
Standardized coefficient estimates with standard errors from top model predicting chronic wasting disease transmission risk for the South Converse Mule Deer Herd based on spatial properties of habitat used by deer.
Covariates included PRNP genotype (shows the effect of a deer being genotype SS compared to reference type *F), compound topographic index (CTI) during summer, CTI during winter, distance to cropland during winter, distance to perennial water source during summer, and distance to secondary road.
Fig 3.
Marginal effects of environmental covariates from top model predicting chronic wasting disease transmission risk for the South Converse Mule Deer Herd based on spatial properties of habitat used by deer.
Spatial covariates included A) distance to perennial water source during summer, B) compound topographic index (CTI) during summer, C) CTI during winter, D) distance to secondary road, and E) distance to cropland during winter. Color indicates differences in effects by genotype (*F or SS).
Fig 4.
Map showing relative risk of chronic wasting disease (CWD) transmission for mule deer in the South Converse Mule Deer Herd as predicted from top risk model.
Spatial predictors in top risk model included compound topographic index (CTI), distance to perennial water source, distance to cropland, and distance to secondary road. The black solid line bisecting the study area is Interstate 25. Since predictors are specific to season and deer genotype was also a predictor in the top model, maps are specific to genotype and season: A) genotype SS, winter, B) genotype *F, winter, C) genotype SS, summer, D) genotype *F, summer. Because one of the predictors, distance to cropland, varies by year, these maps are specific to data from 2014.