RESEARCH PAPER
A comparison of four different methods to estimate population size of Alpine marmot (Marmota marmota)
 
More details
Hide details
1
University of Natural Resources and Life Sciences Vienna
2
Stelvio National Park
Online publish date: 2017-03-15
Publish date: 2017-03-15
 
Hystrix It. J. Mamm. 2017;28(1):61–67
KEYWORDS:
ABSTRACT:

Obtaining reliable information on animal abundance in mountainous landscapes is challenging. Highly heterogeneous habitats tend to reduce detection probabilities, and the three-dimensional, rugged nature of the terrain poses severe limits to the fulfilment of a number of assumptions underlying several statistical methods. In this study, we aimed to compare the performance of 4 different methods to estimate population size of Alpine marmot (Marmota marmota), a highly social semifossorial rodent widely distributed on the European Alps. Between May and August 2015, in a study area within the Stelvio National Park (Italy) we conducted 8 sessions of capture-mark-recapture, 6 sessions of mark-resight from vantage points, 8 sessions of line distance sampling along 4 transects, and 2 sessions using double-observer methods from vantage points. The minimum number of animals alive, obtained during the mark-resight surveys, was n=54 individuals. Capture-mark-recapture models estimated a population size of n=56 individuals [95% CI (45,87)]; similar, but more precise estimates were obtained with the mark-resight approach {Bowden’s estimator: n=62 [95% CI (54,71)]; Poisson log-normal estimator: n=62 [95% CI (55,69)]}. Line distance sampling and double-observer methods were severely biased low {Line distance sampling: n=24 individuals [95% CI (19,31)]; Independent double-observer: n=24 [95% CI (22, 35)]; Dependent double-observer: n=15 [95% CI (15,20)]}. Our results suggest that the probabilistic approach based on marked individuals yielded fairly robust estimates of population size. The underestimates obtained using distance sampling and double-observer methods were likely due to the violation of some underlying assumptions. While the topography of the mountainous landscape makes it difficult to randomize the sampling scheme, the semifossorial behaviour of the target species is likely to lower the detection probabilities and violate the assumption of perfect detection on the transect. 

eISSN:1825-5272
ISSN:0394-1914