Acoustic surveys in the full monte: simulating uncertainty
A method is presented to estimate and diagnose the sources of uncertainty in acoustic survey measures of fish density. The method involves simultaneous Monte Carlo simulation of density and uncertainty resulting from imprecision in all terms in acoustic analyses. Known bias can also be considered. R...
Published in: | Aquatic Living Resources |
---|---|
Main Authors: | , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
EDP Sciences
2000
|
Subjects: | |
Online Access: | https://research.library.mun.ca/1722/ https://research.library.mun.ca/1722/1/Acoustic_surveys_in_the_full_monte_simulating_uncertainty.pdf https://research.library.mun.ca/1722/3/Acoustic_surveys_in_the_full_monte_simulating_uncertainty.pdf https://doi.org/10.1016/S0990-7440(00)01074-3 |
Summary: | A method is presented to estimate and diagnose the sources of uncertainty in acoustic survey measures of fish density. The method involves simultaneous Monte Carlo simulation of density and uncertainty resulting from imprecision in all terms in acoustic analyses. Known bias can also be considered. Results are presented from theoretical simulations based on assumed distributions of fish density and acoustic parameters at sample sizes from 10–1000, and on survey data for Atlantic cod and redfish. Uncertainty can be reduced by either increasing the sample rate or by decreasing the error in input variables (system and ocean parameters, backscatter, target strength, species identification, detectability). Uncertainty in inshore cod surveys in Newfoundland waters can be attributed for the most part to heterogeneous fish distribution and detectability variance (total R2 = 0.64). Uncertainty in offshore redfish surveys is attributed to heterogeneous fish distribution, and variance in target strength and species identification (total R2 = 0.82). Uncertainty can be reduced by survey design, not only by the classical methods of achieving less diverse measures of backscatter, but also by increasing precision in the input parameters to the density estimate, in particular, target strength, detectability, and species identification. |
---|