Double-observer line transect surveys with Markov-modulated Poisson process models for animal availability

David Louis Borchers, Roland Langrock

Research output: Contribution to journalArticlepeer-review

11 Citations (Scopus)
5 Downloads (Pure)

Abstract

We develop maximum likelihood methods for line transect surveys in which animals go undetected at distance zero, either because they are stochastically unavailable while within view or because they are missed when they are available. These incorporate a Markov-modulated Poisson process model for animal availability, allowing more clustered availability events than is possible with Poisson availability models. They include a mark-recapture component arising from the independent-observer survey, leading to more accurate estimation of detection probability given availability. We develop models
for situations in which (a) multiple detections of the same individual are possible and (b) some or all of the availability process parameters are estimated from the line transect survey itself, rather than from independent data. We investigate
estimator performance by simulation, and compare the multiple-detection estimators with estimators that use only initial detections of individuals, and with a single-observer estimator. Simultaneous estimation of detection function parameters and availability model parameters is shown to be feasible from the line transect survey alone with multiple detections and double-observer
data but not with single-observer data. Recording multiple detections of individuals improves estimator precision substantially when estimating the availability model parameters from survey data, and we recommend that these data be gathered. We apply the methods to estimate detection probability from a double-observer survey of North Atlantic minke whales, and find that double-observer data greatly improve estimator precision here too.
Original languageEnglish
Pages (from-to)1060-1069
Number of pages10
JournalBiometrics
Volume71
Issue number4
Early online date1 Jul 2015
DOIs
Publication statusPublished - Dec 2015

Keywords

  • Abundance estimation
  • Availability bias
  • Cox point process
  • Mark-recapture
  • Maximum likelihood

Fingerprint

Dive into the research topics of 'Double-observer line transect surveys with Markov-modulated Poisson process models for animal availability'. Together they form a unique fingerprint.

Cite this