A multivariate habitat model for female bobcats: a GIS approach

摘要

We developed a Geographical Information System (GIS) based habitat model for female bobcats (Lynx rufus) and subjected the model to internal-validation, crossvalidation, and validation using independent data. The model predicted probability of an area being used by female bobcats increased (P <0.001) as slope and distance to mature pine stands increased. Probability of an area being used by female bobcats decreased (P <0.001) with increasing distance to sapling stands, mature hardwood stands, pavedroads, maintenance roads, and creeks. Forest type (non-forested, pine dominated, or hardwood dominated) also influenced (P <0.001) probability of use. Internal- and cross-validation indicated the model performed relatively well (75.5 percent and 73 percent correct classification, respectively). However, when the model was tested with an independent data set. predicted values were only slightly better than random (57.5 percent correct classification). Our validation results indicate habitat models should not be trusted in absence of thorough verification.