Fig 1.
Spatial distribution of presence and absence and background samples.
The zoomed area presents an example of the raw samples’ distribution before data cleaning.
Fig 2.
Risk maps generated from the four diseases distribution models and their means: Presence-background MaxEnt (MPB), presence-background Random Forest (RFPB), weighted mean of both presence-background models based on AUCc (MeanPB), corrected presence-absence MaxEnt (MPAc), corrected presence-absence Random Forest (RFPAc), weighted mean of both presence-absence models based on AUCc (MeanPAc).
The green dots represent the centroids of the 2014–2015 HPAI outbreaks. The color gradient of each pixel represents the AI presence probability from clear red shading (low presence probability) to bright red shading (high presence probability). High resolution versions of the maps are available in Disease BioPortal (http://bioportal.ucdavis.edu).
Table 1.
AUC and AUCc values for the four DDMs: Maxent Presence Bakground (MPB), Maxent Presence-Absence (MPA), Random Forest Presence Background (RFPB) and Random Forest Presence-Absence (RFPA).
Table 2.
Spearman’s correlation coefficient between the spatial predictions of the for the four DDMs, Maxent Presence Background (MPB), Corrected Maxent Presence-Absence (MPAc), Random Forest Presence Background (RFPB), Corrected Random Forest Presence-Absence (RFPAc) and, the model averages for Presence-Absence (WPAc) and Presence-Background (WPB) of Avian Influence occurrence in California, USA.
Predictions were made for 500 × 500 m grid cells in California (n = 3,301,320).
Table 3.
Variable importance for the final models of the four DDMs: Maxent Presence Background (MPB), Maxent Presence-Absence (MPA), Random Forest Presence Background (RFPB) and Random Forest Presence-Absence (RFPA) of Avian Influence occurrence in California, USA.
Variable importance was determined with percent contribution for Maxent and mean decrease in accuracy for Random Forest.