Quantitative comparison and selection of home range metrics for telemetry data
Article first published online: 19 APR 2012
© 2012 Blackwell Publishing Ltd
Diversity and Distributions
Volume 18, Issue 11, pages 1057–1065, November 2012
How to Cite
Cumming, G. S., Cornélis, D. (2012), Quantitative comparison and selection of home range metrics for telemetry data. Diversity and Distributions, 18: 1057–1065. doi: 10.1111/j.1472-4642.2012.00908.x
- Issue published online: 9 OCT 2012
- Article first published online: 19 APR 2012
- Oppenheimer Foundation
- University of Cape Town
- French National Research Agency
- the European Union
- French Foreign Ministry through the French Embassy in Zimbabwe. Grant Number: RP-PCP Grant 2008
- USAID through a subcontract from the Wildlife Conservation Society's Global Avian Influenza Network for Surveillance (GAINS) programme
- DST/NRF Centre of Excellence at the Percy FitzPatrick Institute
- home range;
- movement ecology;
- PTT ;
- ROC plot ;
- species occurrence model;
Home range (HR) metrics are widely used in ecology and conservation, but the quantitative basis for choosing and parameterizing metrics is weak. Home range estimates are ecological and statistical hypotheses that must balance type I and type II errors. Here, we present and test a new approach to fine-tuning and comparing HR estimates using the area under the curve (AUC) statistic.
Test data are taken from telemetry studies of 44 individual ducks in southern Africa and nine buffaloes in southern and western Africa.
We use a meta-analysis of AUC statistics to compare the performance of four standard HR metrics on data from 44 ducks (two species) and nine African buffaloes.
The AUC method emerges as a useful and accessible statistical tool. It captures clear differences between HR estimators as well as providing a way of fine-tuning parameters for an individual HR estimate. Code to run the HR AUC analyses in R is provided. As argued by others, we found that kernel density estimators offer the best combination of ecological and statistical validity, while estimators that use minimum convex polygons at any stage of the algorithm perform poorly and should be avoided.
The AUC statistic provides a readily implementable and straightforward approach to comparing different HR metrics and to selecting parameters for individual metrics. It thus offers a valuable tool for conservation efforts that seek to define HRs for species or populations. The use of the AUC in this new context further contributes to solidifying the interface between species occurrence models and HR estimators.