Habitat use of wild Reindeer (Rangifer t.tarandus) in Hardangervidda, Norway

Abstract

As a migratory species living in the (sub-)arctic reindeer are highly sensitive to the effects of habitat fragmentation, cumulated human disturbance and climate change. In a rapidly changing environment a better understanding of their habitat use is therefore crucial for the longterm conservation of wild reindeer and in order to improve management plans. In this thesis I present research on the habitat selection of wild reindeer (Rangifer t. tarandus) in Hardangervidda, Norway, focusing on (A) how seasonal differences in distribution and movement patterns of wild reindeer can be described and quantified (chapter 2), (B) how the land cover of the alpine study area can be classified accurately using selected parametric and nonparametric classifiers on satellite imagery with and without ancillary data (chapter 3), (C) how the current status of lichen pastures can be mapped and quantified (chapter 4) and (D) how reindeer habitat selection can be modeled at different temporal and spatial scales (chapter 5). To answer question (A), 28 wild, female reindeer in Hardangervidda were equipped with global positioning system collars (GPS). Based on the recorded tracking data, a method is presented to detect and define annual seasons empirically (floating cut-off points within correlated random walks (CRW) models). Reindeer are shown to perform seasonal migrations within the study area and the underlying movement patterns are described and quantified using multiple parameters (directionality and regularity of movement, activity level and size/stability of seasonal home range). Significant seasonal differences in movement patterns are detected and related to (potential) underlying factors (climate, differences in spatial distribution of resources between seasons, human disturbance and social behaviour of reindeer). Answering question (B), a method for land cover classification within a mountainous area is proposed based on Landsat imagery and ancillary data (DEM and derivates, images from different seasons). Relevant land cover classes for resource selection of wild reindeer are defined and the study area is classified accordingly. The kNN classifier demonstrates superior results as compared to maximum likelihood and decision tree classifiers. It is shown, that the use of ancillary data can significantly improve the overall classification accuracies for all tested classifiers and the class accuracies of the vast majority of classes.