University of Veterinary Medicine Vienna - Research portal

Diagrammed Link to Homepage University of Veterinary Medicine, Vienna

Selected Publication:

Type of publication: Diploma Thesis
Type of document:

Year: 2010

Authors: Huber, Nikolaus

Title: Calibrating behavioural information from GPS/SOB units through simultanious direct observation of a collared Przewalski's horse.

Other title: Beobachtung einer besenderten Wildpferdgruppe mit neuer Sendertechnologie und R├╝ckschluss auf Verhalten

Source: Diplomarbeit, Vet. Med. Univ. Wien, pp. 27.


Advisor(s):

Kaczensky Petra
Walzer Christian

Reviewer(s):
Arnold Walter

Vetmed Research Units:
Research Institute of Wildlife Ecology


Graduation date: 12.07.10


Project(s): Landscape level research for the conservation of Asiatic wild ass in Mongolia


Abstract:
Behavioural observations of free ranging animals can provide important insight into many aspects of their biology but are time consuming and can be invasive. The use of Very High Frequency (VHF) radio tracking systems in the early 1960s, later used in combination with motion sensors, and the application of global positioning system (GPS) technology since the mid 1990s, made it possible to monitor fine-scale movements of large and medium-sized mammals. The recent development of GPS technology allows the remote collection of high precision location data at fixed intervals. In this study we tested whether it is possible to classify the behaviour of a Przewalski's horse in the Mongolian Gobi based on the distance between successive GPS fixes by comparing GPS data with direct observations. In summer 2008 we collared a one-year-old Przewalski's horse with a GPS unit that collected locations at 15 minute intervals. The unit included a motion sensor but data of this sensitive technical tool was unfortunately corrupted. To determine the distance range and get a first idea of suitable cut points to differentiate behaviour based on distances travelled, we plotted the % of correctly classified behaviour against distance. We further modelled the best cut-points by using Fisher's discriminant analysis and a Regression tree (CRT) analysis. The three methods yielded similar cut points (plot: resting/grazing 35m, grazing/movement 570m; Fisher's discriminant: resting/grazing 31 ,2m, grazing/movement 510m; CRT: resting/grazing 21 , 5m, grazing/movement 677m), which suggests that the separation criteria is rather robust. Although behavioural categories lasting for 15 minutes could be fairly reliably separated based on the distances covered between successive fixes, almost half the dataset consisted of mixed intervals. Thus, fifteen minute intervals are too long to register just one behavioural category, which makes classification based on GPS fixes alone problematic. Although our present approach was not particularly successful, we believe that using GPS data in combination with activity sensors holds great potential for inferring main behavioural categories.


© University of Veterinary Medicine ViennaHelp and Downloads