Ungulate population monitoring in an open tundra landscape: distance sampling versus total counts

Mathilde Le Moullec,Brage Bremset Hansen,Åshild Ønvik Pedersen,Ronny Aanes,Nigel Gilles Yoccoz,Jarle Tufto

doi:10.2981/wlb.00299

Abstract

Researchers and wildlife managers strive for low bias and high precision (i.e. high accuracy) when estimating animal population sizes. Distance sampling is currently one of the most widely used monitoring methods. However, it relies on strict sampling designs and modeling assumptions that can be difficult to meet in the field. Here, we use data from two sub‐populations of non‐migratory wild Svalbard reindeer Rangifer tarandus platyrhynchus inhabiting flat, open and isolated coastal tundra plains, to demonstrate some challenges related to the distance sampling methodology. To achieve this, we compared distance sampling line transect estimates with repeated total population counts and combined available software tools (R packages unmarked, Distance and dsm) to fulfill the analytical requirements of small study sites in which large areas are surveyed relative to the study area size. Based on low variation among repeated total counts (CV = 0.02 ‐ 0.06) and the virtual absence of false negatives and positives of marked animals, the total counts could be used as reference population sizes. Distance sampling estimates were not statistically different from the total count estimates. Our relatively large sample size of 143 observations enabled precise distance sampling abundance estimates (CV = 0.16 ‐ 0.26) compared with other studies in the wild. However, capturing the processes shaping population dynamics would likely require even higher sampling effort or other, more resource demanding monitoring tools, such as total counts or mark‐recapture. In this type of ecosystem, distance sampling nevertheless represents a cost‐effective tool suitable for ‘population state’ assessment and studies of large‐scale spatial distribution patterns. Our study stresses the importance of choosing the appropriate analytical tools and estimating the accuracy of the monitoring methods that are used to achieve specific scientific, management or conservation goals.

Highlights

Stochasticity) and observational error (Clark and Bjørnstad 2004, Buckland et al 2007, Sæther et al 2007)
distance sampling (DS) relies on four key assumptions related to study design and statistical analysis of the data (Buckland et al 2015): 1) animals are distributed independently of the transects; 2) objects on or close to the transects are always detected; 3) distances are measured without error; and 4) objects are detected at their original position
To estimate reindeer abundance from Total Count (TC) data and its uncertainties we considered two types of errors related to the observers; 1) an animal could be counted twice with a probability p or 2) an animal could be undetected with a probability q

Summary

Introduction

Stochasticity) and observational error (Clark and Bjørnstad 2004, Buckland et al 2007, Sæther et al 2007). The most common method to estimate population abundance of wild animals is distance sampling (hereafter referred to as DS) (Buckland et al 2004). Recent spatial modeling developments have been incorporated into the DISTANCE interface (Thomas et al 2010), whereby a two-stage approach analyzes detection and density separately This is suitable for large-scale study areas. The distinctive landscape characteristics of Svalbard are highly suitable to evaluate the precision and sources of error in two methods of estimating animal abundance; DS and TC. In this open tundra landscape, the detection of reindeer should in principle only vary with distance from the transect line as visibility is good.

Methods

Results

Conclusion