Abstract

We describe an approach to estimation of the spatial distribution of reindeer (Rangifer tarandus). Spatial auto-correlation, inherent to the data describing the distribution of wildlife species, contains information that can be utilized to improve the effciency of field inventories. Our data included reindeer fecal pellet counts, satellite imagery and a digital terrain model. We applied ordinary logistic regression, autologistic regression, and the Gibbs sampler to predict spatial distribution of reindeer based on the combined data. A training set was used to compare the outcome for different field sampling designs for each method. Results suggested the possibility to reduce the number of plots by up to 75% with a 15% reduction in prediction accuracy (quality). We also showed that the Gibbs sampler outperformed, in terms of accuracy, the logistic regression. The outcome however, was dependent on the spectral homogeneity of the area and on the relative position of the sampling design to the elevation curves. Our results justify the incorporation of spatial information when modeling the distribution of reindeer at finer scales (< 1 km)

Published in

AMBIO: A Journal of the Human Environment
2003, Volume: 32, number: 8, pages: 568-572
Publisher: ROYAL SWEDISH ACAD SCIENCES

SLU Authors

• Teterukovsky, Alex

  • Department of Forest Economics, Swedish University of Agricultural Sciences
    

• Edenius, Lars

  • Department of Animal Ecology, Swedish University of Agricultural Sciences
    

UKÄ Subject classification

Environmental Sciences related to Agriculture and Land-use


Publication identifier

DOI: https://doi.org/10.1639/0044-7447(2003)032[0568:EFSFPT]2.0.CO;2

Permanent link to this page (URI)

https://res.slu.se/id/publ/1361