Home ABOUT Latest OFP News Identifying FAD-fishing
Identifying FAD-fishing
Tuesday, 25 August 2015 08:54

stevenh2015_08_25-sp_brailleTuna purse-seine catches associated with Fish Aggregation Devices (FADs) are typically more diverse than catches set on “free school” tuna, including both smaller target tuna as well a variety of bycatch species that are often discarded.  In response, an annual FAD-fishing moratorium, of several months’ duration and Pacific-wide, has been implemented since 2009.  Despite the moratorium, there remains concern about adherence to FAD-free fishing requirements.  Additionally, seafood producers and consumers seek certification regarding the source of purse seine captured tuna.

Given this background, we became interested in developing a FAD-association verification test.  We have developed a simple technique, based on observer catch sampling, to determine whether a purse seine set is likely to have been associated with a FAD.  Our methodology, which can be easily implemented “in the field”, has a prediction accuracy rate of up to 86%.  Details of the methodology have been published, in Open Access format, in the journal Fisheries Research.




• We predict school association in purse seine sets using low intensity observer sampling data.
• Simple classification tree model prediction accuracy ranged from 77 to 83%.
• Bycatch data substantially reduced misclassification of free schools as FAD sets.
• Use of auxiliary data and more complex resampling models did not improve set association prediction accuracy.
• Purse seine sets from FAD-closure periods were classified equally well as sets from the non-FAD closure periods.


We investigate the potential of verifying whether individual purse seine sets were made in association with a fish aggregation device (FAD) or on an unassociated (FAD-free) tuna school, on the basis of low intensity catch sampling by onboard observers. The target tuna catch and length compositions and bycatch amounts were analyzed from more than 50,000 purse seine sets sampled by onboard observers who had, in addition to collecting sampling data on species and size composition of target tunas in the catch, and set-level estimates of total bycatch, also identified the sets as either “associated” or “unassociated”. Classification tree (CT) models were developed based on 2007–2011 observer data and tested for misclassification error rates on 2012 data. Two types of model misclassification errors (MCE) are possible: unassociated sets misclassified as associated (termed false positive or Type I) and associated sets misclassified as unassociated sets (false negative or Type II error). A third error measure, overall MCE, is a weighted average of Type I and Type II errors. The classification rules developed on the basis of observer catch sampling tended to be nearly presence/absence, e.g. greater than 99% skipjack composition or presence of 0.5 kg rainbow runner, likely keyed by the modest observer sample sizes. Overall MCE rates were 21.8% for the initial tuna-only CT model and 14.4% for the bycatch-included model. The improvement in overall classification for the bycatch models derived principally from a reduction in Type I errors. The addition of auxiliary non-sampling variables (e.g., longitude, month) and use of more complex resampling extensions to CT modelling led to little to no improvement in MCE rates. We employed our methodology to analyze a particular subset of the purse seine data, i.e., sets from the FAD-closure periods of 2009–12. The intent was to determine if MCE rates of these particular sets were greater than the MCE rates found in the more general analysis. Reassuringly, the MCE rates of sets during the FAD closure period were found to be equal, or even a bit lower than MCE rates in the broader analyses based on our best performing model.

The paper has been published as Open Access and is freely available on ScienceDirect or in the icon SPC Digital Library.


Nouméa, New Caledonia, August 25th 2015,
Photo courtesy of Tim Lawson, SPC.

Copyright © 2024 Oceanic Fisheries Programme. All Rights Reserved.
Joomla! is Free Software released under the GNU/GPL License.
   SPC Homepage | About SPC | Copyright © SPC 2010