Comparison of two least-cost path modeling methods for predicting external home range movements of white-tailed deer in a fragmented agricultural landscape

Background/Question/Methods

Least-cost path (LCP) modeling provides a framework to assign habitat variables a resistance level; this information can be used to predict potential movement paths of animals across a landscape. Several methods to rank resistance of habitat variables have been used to predict movement across space but few comparisons and validations of these methods have occurred. We compared 2 popular methods of ranking resistance to actual external home range movements by white-tailed deer (Odocoileus virginianus). During 2011-2013, we GPS collared (GPS location/hour) 52 white-tailed deer in agriculturally dominated central Illinois to monitor external home range movements. We collected 47 external home range paths (dispersal and exploratory movements) to use for analysis. We ranked 11 habitat variables for resistance using expert opinion (EO) surveys and a resource selection function (RSF), and created resistance maps for each within a GIS. We then predicted movement paths of white-tailed deer using the starting and ending points of the collected movement paths. We compared the 2 resistance models and a null model using 3 repeated measures ANOVAs on 3 path metrics, path deviation index (PDI), path cost, and path sinuosity.

Results/Conclusions

We found that neither the EO or RSF models outperformed the null model at predicting deer movement paths with no difference observed between PDI values (P >0.60). We found no difference in path resistance between the LCP models, actual deer paths, or the null model (P >0.30). We found significant differences in path sinuosity between the actual deer paths and both LCP models and the null model (P <0.01), with actual deer paths being more sinuous than both LCP models and the null model. Our results show that these 2 methods of LCP modeling did not accurately predict white-tailed deer movement paths better than the null model. Our results may have been influenced by frequent sampling of movement paths which incorporated more variation in behaviors during external home range movements. The use of LCP methods should be investigated further to determine their efficiency at predicting actual movement paths for other species, especially since LCP modeling is popular in conservation corridor planning.