Estimating fish abundance from acoustic surveys: calculating variance due to acoustic backscatter and length distribution error

Juan Zwolinski, Paul G. Fernandes, Vitor Marques, Yorgos Stratoudakis

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

Estimation of fish abundance from acoustic surveys requires the estimation of total acoustic backscatter of the target species in the sampled region. Although the arithmetic mean of acoustic backscatter is an unbiased estimator of the mean backscatter for regular or random sampling designs, under the presence of spatial structure, its use leads to a loss of information and the estimation of its variance is not trivial. Here, we tackle these shortcomings by building a spatial model of acoustic backscatter using spline-based generalized additive models (GAMs). GAMs were used to provide local and global estimates of acoustic backscatter, and their precision was calculated by statistical simulations of the models' parameters. For a series of surveys performed off the western and Southern Iberian Peninsula, GAM estimates were unbiased and more precise than the arithmetic mean estimates. Simulations of the acoustic backscatter fields were combined with resampling of the trawls to provide confidence intervals for fish numbers and biomass. The relative standard errors of the estimates were within 13% and 46% (average 22%) for numbers and within 12% and 35% (average 19%) for biomass. Acoustic sampling error was the major contributor to the overall variance.

Original languageEnglish
Pages (from-to)2081-2095
Number of pages15
JournalCanadian Journal of Fisheries and Aquatic Sciences
Volume66
Issue number12
Early online date16 Nov 2009
DOIs
Publication statusPublished - Dec 2009

Keywords

  • Sardine Sardina-Pilchardus
  • trawl survey data
  • statistical-models
  • precision
  • sea
  • uncertainty
  • variability
  • simulation
  • Portugal
  • splines

Cite this

Estimating fish abundance from acoustic surveys : calculating variance due to acoustic backscatter and length distribution error. / Zwolinski, Juan; Fernandes, Paul G.; Marques, Vitor; Stratoudakis, Yorgos.

In: Canadian Journal of Fisheries and Aquatic Sciences, Vol. 66, No. 12, 12.2009, p. 2081-2095.

Research output: Contribution to journalArticle

@article{d7b39134ff314a9eabd3e5ce43f1574a,
title = "Estimating fish abundance from acoustic surveys: calculating variance due to acoustic backscatter and length distribution error",
abstract = "Estimation of fish abundance from acoustic surveys requires the estimation of total acoustic backscatter of the target species in the sampled region. Although the arithmetic mean of acoustic backscatter is an unbiased estimator of the mean backscatter for regular or random sampling designs, under the presence of spatial structure, its use leads to a loss of information and the estimation of its variance is not trivial. Here, we tackle these shortcomings by building a spatial model of acoustic backscatter using spline-based generalized additive models (GAMs). GAMs were used to provide local and global estimates of acoustic backscatter, and their precision was calculated by statistical simulations of the models' parameters. For a series of surveys performed off the western and Southern Iberian Peninsula, GAM estimates were unbiased and more precise than the arithmetic mean estimates. Simulations of the acoustic backscatter fields were combined with resampling of the trawls to provide confidence intervals for fish numbers and biomass. The relative standard errors of the estimates were within 13{\%} and 46{\%} (average 22{\%}) for numbers and within 12{\%} and 35{\%} (average 19{\%}) for biomass. Acoustic sampling error was the major contributor to the overall variance.",
keywords = "Sardine Sardina-Pilchardus, trawl survey data, statistical-models, precision, sea, uncertainty, variability, simulation, Portugal, splines",
author = "Juan Zwolinski and Fernandes, {Paul G.} and Vitor Marques and Yorgos Stratoudakis",
year = "2009",
month = "12",
doi = "10.1139/F09-138",
language = "English",
volume = "66",
pages = "2081--2095",
journal = "Canadian Journal of Fisheries and Aquatic Sciences",
issn = "0706-652X",
publisher = "CANADIAN SCIENCE PUBLISHING, NRC RESEARCH PRESS",
number = "12",

}

TY - JOUR

T1 - Estimating fish abundance from acoustic surveys

T2 - calculating variance due to acoustic backscatter and length distribution error

AU - Zwolinski, Juan

AU - Fernandes, Paul G.

AU - Marques, Vitor

AU - Stratoudakis, Yorgos

PY - 2009/12

Y1 - 2009/12

N2 - Estimation of fish abundance from acoustic surveys requires the estimation of total acoustic backscatter of the target species in the sampled region. Although the arithmetic mean of acoustic backscatter is an unbiased estimator of the mean backscatter for regular or random sampling designs, under the presence of spatial structure, its use leads to a loss of information and the estimation of its variance is not trivial. Here, we tackle these shortcomings by building a spatial model of acoustic backscatter using spline-based generalized additive models (GAMs). GAMs were used to provide local and global estimates of acoustic backscatter, and their precision was calculated by statistical simulations of the models' parameters. For a series of surveys performed off the western and Southern Iberian Peninsula, GAM estimates were unbiased and more precise than the arithmetic mean estimates. Simulations of the acoustic backscatter fields were combined with resampling of the trawls to provide confidence intervals for fish numbers and biomass. The relative standard errors of the estimates were within 13% and 46% (average 22%) for numbers and within 12% and 35% (average 19%) for biomass. Acoustic sampling error was the major contributor to the overall variance.

AB - Estimation of fish abundance from acoustic surveys requires the estimation of total acoustic backscatter of the target species in the sampled region. Although the arithmetic mean of acoustic backscatter is an unbiased estimator of the mean backscatter for regular or random sampling designs, under the presence of spatial structure, its use leads to a loss of information and the estimation of its variance is not trivial. Here, we tackle these shortcomings by building a spatial model of acoustic backscatter using spline-based generalized additive models (GAMs). GAMs were used to provide local and global estimates of acoustic backscatter, and their precision was calculated by statistical simulations of the models' parameters. For a series of surveys performed off the western and Southern Iberian Peninsula, GAM estimates were unbiased and more precise than the arithmetic mean estimates. Simulations of the acoustic backscatter fields were combined with resampling of the trawls to provide confidence intervals for fish numbers and biomass. The relative standard errors of the estimates were within 13% and 46% (average 22%) for numbers and within 12% and 35% (average 19%) for biomass. Acoustic sampling error was the major contributor to the overall variance.

KW - Sardine Sardina-Pilchardus

KW - trawl survey data

KW - statistical-models

KW - precision

KW - sea

KW - uncertainty

KW - variability

KW - simulation

KW - Portugal

KW - splines

U2 - 10.1139/F09-138

DO - 10.1139/F09-138

M3 - Article

VL - 66

SP - 2081

EP - 2095

JO - Canadian Journal of Fisheries and Aquatic Sciences

JF - Canadian Journal of Fisheries and Aquatic Sciences

SN - 0706-652X

IS - 12

ER -