Figure legends
Figure 1 Geographic distribution of MRC occurrence and background-weighted points. Background-weighted points were allocated based on a probability distribution of a density raster of training occurrence points.
Figure 2 Habitat suitability map of the tuned-up models derived from random (Rndm) and background weighting (BkWt) schemes.
Figure 3 Variation of sensitivity (Sens); the proportion of correctly predicted presence data, and specificity (Spec); the proportion of correctly predicted pseudo-absence data, across a gradient of suitability thresholds. Dashed lines indicates test data. Rndm: random and BkWt: background weighting schemes. The black diamonds indicate the corresponding threshold of the 10 percentile suitability of the training dataset.
Figure 4 Response curve of the explanatory variables derived from models obtained based on the random (top) and background weighting (bottem) schemes. Models were fitted in biomod2 package.