IUCN / SSC Cat Specialist Group - Digital Cat Library |
|
||||||||||
|
||||||||||
|
|
| Blecha, K.A.; Alldredge, M.W. | |
| Improvements on GPS location cluster analysis techniques for predicting feeding activities | |
| 2014 Conference Proceeding | |
GPS location cluster analysis has become a staple technique in large predator feeding studies. We demonstrate three improvements to the method based on study design and technological advancements in a four year study of 4600 cluster ground-truthing visits to GPS location clusters from 53 cougar subjects: 1) Ground-truthing potential feeding sites have generally relied on either exhaustively visiting all clusters of an individual subject or by opportunistic non-random sampling. The former may be cost prohibitive and limit the sample size of animal subjects, while the latter may lead to biased estimates of kill-rate/prey composition. We demonstrate a stratified random sampling scheme for monthly visitation periods, along with a sample of "single point clusters" consisting of a single GPS location. 2) Feeding event prediction models using classic cluster analyses are likely biased toward larger prey items. We demonstrate a simple integration of accelerometer sensor data as an additional covariate to help improve prediction models. 3) It is suspected that the probability a technician successfully classifies a cluster while ground-truthing declines as the visitation lag (time delay between technician visitation and cougar feeding date) increases. Direct satellite communication capabilities now allow real-time downloading of GPS data, and thus near real-time ground truthing. We tested this on a sample of clusters (n= 202) with visitation by one technician 1-3 days post initiation by the cougar, and then by a second independent blind observer 1-60 days later. The double observers concurred in the presence/absence of feeding remains in 98.6% of the cases, while concurring on species identification in 93.6%. While classification success of prey species, sex, and age will likely decrease as prey remains degrade, researchers should be confident that the presence/absence classification component is robust for the logistically feasible visitation lags that were tested. |
|
 |
PDF files are only accessible to Friends of the Cat Group. Joining Friends of the Cat Group gives you unlimited access and downloads in the Cat SG Library for one year, and allows you to receive our newsletter Cat News (2 regular issues per year plus special issues). More information how to join here |
(c) IUCN/SSC Cat Specialist Group ( IUCN - The World Conservation Union) |