Figure 4 Breeding habitats of the Eastern population of the Lesser White-fronted Goose based on the minimum training presence threshold. Projection: Asia North Albers Equal Area Conic. Background: World Imagery from ESRI (http://services.arcgisonline.com/arcgis/rest/services).

Effects of environmental factor on the summering range of A. erythropus

Of the nine environmental variables included in model building, elevation was the most important, strongly contributing to the scaling of the Maxent model (59.7% based on the model gain and 50.1% based on re-evaluation of the random permutation of training presence and background data, Table 2). Other highly influential variables (with more than 5% permutation contribution) include precipitation of the warmest quarter, distance to streams, and mean temperature of the warmest quarter (Table 2).
Although highly correlated environmental predictors were excluded from model fitting, there are still collinearities in the remaining variables. For example, the Pearson r between Bio10 (precipitation of the warmest quarter) and Bio18 (mean temperature of the warmest quarter) is relatively high (-0.82) in the study area. Thus, the variable contributions in Table 2 should be interpreted with caution (Phillips 2005).
Table 2 Relative contributions of the environmental variables to the breeding habitat distribution of A. erythropus ranked by permutation importance.