Fig 1.
AGS cases and land use in North Carolina, South Carolina, and Virginia (2019). (Base map: ESRI https://www.arcgis.com/home/item.html?id=6e850093c837475e8c23d905ac43b7d0, Land Cover: MRLC mrlc.gov/data, State Boundaries: ESRI https://www.arcgis.com/home/item.html?id=540003aa59b047d7a1f465f7b1df1950).
Table 1.
Variable names, adjusted relative risks, 95% confidence intervals, and p-values of GLM.
Table 2.
Variable importance for GLM, MaxEnt, and BRT. Based on AUC measure.
Fig 2.
Predicted Individual-level AGS risk for the mid-Atlantic region from GLM, MaxEnt, and BRT. GLM and BRT predict highest risk within the mountainous eastern sides of the region, with lower risk occurring along the eastern coastline. However, BRT predicts a steeper change in risk between regions than the GLM. MaxEnt predicts patchy risk across most of the region. (County Boundaries: US Census Bureau https://services2.arcgis.com/FiaPA4ga0iQKduv3/arcgis/rest/services/TIGERweb_Counties_v1/FeatureServer).
Fig 3.
Predicted Population-level AGS risk for the mid-Atlantic region from GLM, MaxEnt, and BRT. All models predict areas around population centers to have the highest risk of AGS cases, likely due to the large disparity in population density and ease of access to specialists (e.g., allergists) between urban and rural areas. (County Boundaries: US Census Bureau https://services2.arcgis.com/FiaPA4ga0iQKduv3/arcgis/rest/services/TIGERweb_Counties_v1/FeatureServer).
Fig 4.
Receiver Operating Characteristics (ROC) curves for GLM, MaxEnt, and BRT.