Fig 1.
Study area and land cover classes.
Fig 2.
BU positive and BU negative community locations.
Fig 3.
Examples of land cover configurations resulting in high and low values for the three landscape metric calculations.
Fig 4.
New locations gridded at 5 km intervals for BU rate predictions at unobserved locations.
Table 1.
Deviance Information Criterion for non-spatial vs. spatial models.
Lower DIC values indicate a better model fit. Variable abbreviations correspond to metrics as follows: SHAPE_MN (shape index mean), PLADJ (percent land cover adjacency), LSI (landscape shape index).
Table 2.
Variable significance comparison of non-spatial versus spatial models.
The “+” and “-”symbols indicate significant coefficients in the direction indicated by the symbol, and the “X” symbol indicates non-significant coefficients.
Table 3.
Root mean square predictive error of actual BU rates versus non-spatial and spatial GLM predicted rates per 1,000 people.
Table 4.
BU actual rates vs. predicted rates per 1,000 people at 18 testing locations.
Table 5.
Credible intervals for 1200m wetland best-fitting spatial binomial GLM.
Variables that maintain positive or negative signs across their corresponding interval region may be interpreted as having a high probability of the true value of the variable lying within this credible region.
Fig 5.
Observed (A) vs. fitted (B) BU rates.
Fig 6.
Map of spatial random effects. Red, orange, and yellow colors indicate areas where unknown, spatially structured variables that were not included in the model make a higher contribution to BU rates than in other areas of the study region.
Fig 7.
Interpolated mean of the posterior predictive distribution.
Predicted rates per 100,000 individuals. Red, orange, and yellow colors indicate areas with higher BU rates predicted, while blue and green colors represent areas predicted to have low BU rates.
Fig 8.
A) Areas with aggregated and complexly shaped wetland and forest patches surrounding locations with high BU prevalence.
B) Uniformly shaped forest patches surrounding locations with no BU prevalence.