Marine Mammal Survey and Assessment Methods

J.L Laake, D.G. Robertson, Steven C. Amstrup, B.F.J Manly

• Publisher: CRC Press
• Year: 2021
• eBook ISBN: 9781000448368
• Topic: Biological Sciences
• Subtopic: Biology
• Index: Biological Sciences

This volume comprises the proceedings of a symposium on marine mammal survey assessment methods, which took place in Seattle, Washington, USA.

1 Survey design and application

Adaptive line transect survey for harbor porpoises

D. Palka

National Marine Fisheries Service, Woods Hole, Mass., USA

J. Pollard

School of Mathematical and Computational Sciences, University of St. Andrews, North Haugh, UK

ABSTRACT: When surveying for animals that are rare and spatially clumped, adaptive sampling has been shown to be theoretically more precise than non-adaptive methods because the increased number of sightings enables model parameters to be estimated more precisely. Adaptive line transect sampling, developed by Pollard and Buckland (1997), is a technique that permits additional survey effort in areas of high animal density. Computer simulation studies indicate this method is unbiased and more precise than traditional line transect methods. To field test this method, shipboard surveys for harbor porpoises (Phocoenaphocoena) were conducted in the Gulf of Maine/Bay of Fundy during August 1996. Both adaptive and traditional line transect surveys were conducted on the same day over the same track lines. This field experiment demonstrated that adaptive line transect sampling was easy to implement and resulted in more precise and lower harbor porpoise density estimates as compared to traditional line transect sampling.

Keywords: adaptive sampling, harbor porpoise, line transect sampling, Phocoena phocoena, sampling heterogeneity, statistical efficiency.

1 INTRODUCTION

A problem encountered during some line transect abundance surveys of marine mammals is low precision in estimating density due to low detection rates and spatial heterogeneity. Adaptive cluster sampling has been suggested as a method that leads to more detections, which could result in a more precise density estimate (Thompson 1992, Thompson & Seber 1996, Pollard & Buckland 1997). In adaptive cluster sampling designs, neighboring units are added to the sample whenever the value of some trigger variable satisfies a chosen criterion. However, these methods can lead to insufficient coverage of the entire survey area, may be difficult to implement, may be computationally complex to analyze, or may require unrealistic amounts of prior knowledge about the population.

Pollard & Buckland (1997) suggested a strategy for line transect surveys that is adaptive, produces unbiased density estimates, and, for highly aggregated populations, results in more precise density estimates than that from traditional line transect methods. To implement this strategy, two factors are determined a priori: the total amount of survey effort and the minimum amount of straight-line survey effort necessary to cover the entire survey region, termed the nominal effort. In addition to the nominal effort, survey effort is increased when high density areas are encountered. The amount of increased effort is a function of the difference between the total effort still available and the nominal effort remaining. The increased effort starts when the number of observations exceeds a pre-determined criterion and is achieved by zig-zagging back and forth across the nominal straight-line transect. When the sighting rate on the zig-zagging track line decreases to below the pre-determined criterion, the survey track line goes back to the nominal straight-line tracks (Figure 1).

Figure 1. Actual effort is increased by zig-zagging. Nominal effort refers to the equivalent straight-line track.

When using this sampling strategy, effort is biased towards high density areas and so the density estimate, if not corrected, is biased upwards. The corrected estimate is adjusted by weighting the data by the nominal (straight-line) length. Thus, each leg of effort is weighted in proportion to the length of inverse of the effort factor, which is the ratio of the length of transect line actually traveled to the nominal effort through that section. The increased effort legs can vary by altering the zig-zags, either in length, angle, number or a combination of all of these. Thus, in this strategy the adaptive component can be modified as the survey progresses to ensure the nominal length is covered and the entire study area is surveyed.

Simulations show that little bias is introduced by this strategy, and especially for highly clustered populations, the density estimate variance was less than the variance of traditional line transect estimates (Pollard & Buckland 1997). These simulations demonstrated a mean efficiency increase of 7% for highly clustered populations and, as expected, a loss in efficiency (4%) for spatially random populations.

Our objective in this study was to test Pollard and Buckland’s adaptive line transect strategy in the field to determine the practicality of implementing the survey design, and the analytical efficiencies when applied to actual sightings of harbor porpoises (Phocoena phocoena) from the Gulf of Maine / Bay of Fundy region. An experimental survey was conducted where both adaptive line transect and traditional line transect sampling methods were used in the same area during the same day.

2 BACKGROUND: POLLARD & BUCKLAND’S DENSITY ESTIMATION METHOD

Nominal values refer to the values expected if a traditional straight-line transect was followed. Nominal effort is signified by a prime, such as L^/, whereas the corresponding actual effort is denoted by L. A detection may consist of one or more animals.

Each transect was divided into a number of sub-transects or legs, where the start and end of each leg occurred at a change in direction. The straight sections within each zig-zag all had (approximately) the same angle, and hence effort factor, so the complete zig-zag section was consider...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• In Memoriam: Dr Gerald W.Garner, 1944–1998
• Preface
• 1 Survey design and application
• 2 Visibility bias and missed observations
• 3 Modeling
• Keyword index
• Author index