![The Convention Area of the Western and Central Pacific Fisheries Commission (Source: https://www.wcpfc.int/ [Accessed 15 May 2022]).](https://www.researchgate.net/profile/Johanna-Pierre/publication/365368646/figure/fig4/AS:11431281097156558@1668456909810/The-Convention-Area-of-the-Western-and-Central-Pacific-Fisheries-Commission-Source_Q320.jpg)
Content uploaded by Johanna P Pierre
Author content
All content in this area was uploaded by Johanna P Pierre on Nov 14, 2022
Content may be subject to copyright.
How much is enough?
Review optimization methods to deliver best value from
electronic monitoring of commercial fisheries
Pew Project: 2021-IF-02
Final report: 30 October 2022
Johanna P. Pierre
Alistair Dunn▪
Abby Snedeker
Morgan Wealti
Corresponding author: johanna@jpec.co.nz; Johanna Pierre Environmental Consulting Ltd, Lower Hutt, New Zealand
▪Ocean Environmental Ltd, Wellington, New Zealand
Saltwater Inc., Anchorage, USA
Present address: Anchorage, USA
1
Glossary
ADP Annual Deployment Plan (set out plans for fishery monitoring)
AFG Alaska Fixed Gear fishery
AGAC Association of Large Tuna Freezers (an industry body)
AI Artificial Intelligence
AR Activity recognition software
BFT Atlantic bluefin tuna (Thunnus thynnus)
CCSBT Commission for the Conservation of Southern Bluefin Tuna
CGP Code of Good Practice
CV Coefficient of Variation
dFAD Drifting Fish Aggregating Device
DOS Digital Observer Services (an EM review service provider)
EEZ Exclusive Economic Zone
EFL Electronic Fishing Log
EM Electronic Monitoring
ERandEM-IWG WCPFC Intersessional Working Group on Electronic Monitoring and
Reporting
GNSS Global Navigation Satellite System
GPS Global Positioning System
HMS Highly Migratory Species
IATTC Inter-American Tropical Tuna Commission
IBQ Individual Bluefin Quota Program
ICCAT International Commission for the Conservation of Atlantic Tunas
IFOP Fisheries Development Institute (Chile)
IOTC Indian Ocean Tuna Commission
KPI Key Performance Indicator
MCS Monitoring, Control and Surveillance
MSY Maximum Sustainable Yield
NOAA National Oceanic and Atmospheric Administration (USA)
NPFC North Pacific Fisheries Commission
OLE Office of Law Enforcement (NOAA Fisheries)
2
OPAGAC Organization of Associated Producers of Large Tuna Freezers (an
industry body)
PFAs Principles, functions and actions (specified in RFMO convention texts)
RFMO Regional Fisheries Management Organization
SERNAPESCA National Fisheries and Aquaculture Service (Chile)
SUBPESCA Undersecretariat for Fisheries and Aquaculture (Chile)
TNC The Nature Conservancy
UNCLOS United Nations Convention on the Law of the Sea
VMP Vessel Monitoring Plan
VMS Vessel Monitoring System
WCPFC Western and Central Pacific Fisheries Commission
WGEMS IOTC Ad-hoc Working Group on the Development of Electronic
Monitoring Programme Standards
3
Executive summary
Electronic monitoring (EM) using on-vessel cameras can effectively collect a broad range of data
to support fisheries management. Key advantages of EM include its flexibility, scalability,
verification capability, and the avoidance of health, safety and logistical challenges that human
observer deployments can involve. EM can also offer cost efficiencies relative to other
monitoring methods. In this report, we consider the use of EM to meet a range of fishery
monitoring objectives, present case studies from EM programs in real-world fisheries, and
evaluate the level of review needed to extract EM-collected information to support management
objectives. Our goal is to show how the efficiency of EM review can be maximized to support
management, within budgetary requirements. We focus on Regional Fisheries Management
Organizations (RFMOs) managing tuna fisheries, and also set out the broader application of
findings across other management entities and fishing methods.
RFMOs require extensive datasets to meet their management objectives. Supporting such data
requirements, EM has the capacity to collect comprehensive data on fishery catch (retained and
discarded), catch handling, fishing gear, and operational characteristics of fisheries (e.g. date,
time and location of sets and hauls). Opportunistic or partial data collection supported by EM
includes discarded gear and other marine pollution events. Most RFMOs have taken significant
steps towards progressing EM, while adoption is at different stages.
Case studies spanning the Pacific, Indian, and Atlantic Oceans show the efficacy of EM in
collecting fisheries data in the real world. Monitoring objectives to be met by EM and
approaches to review of EM imagery and associated information vary among these and other
EM programs. Using EM to capture 100% of fishing activity is recognized as best practice, while
EM review may be undertaken as a census (all imagery reviewed) or with samples of imagery
collected. Auditing EM-derived data against other sources, typically logbook information, offers
additional options for review.
EM review efficiency, in terms of time and cost, can be increased by considering review
requirements during the EM program design (e.g. development of EM-appropriate data
definitions) and on-vessel data capture phases (e.g. lens cleaning to improve image clarity).
Efficiency of the review phase itself can also be increased, for example by reviewing at speeds
faster than real time and supporting review with computer vision tools.
EM review costs as a proportion of program costs vary from 2.5 – 60% (noting that what is
incorporated in review process costs differs among programs). Review costs do not scale
linearly with review rates, and service providers emphasize that collaboration among
themselves, clients and vessel operators is important for maximizing review cost efficiencies.
Identifying the minimum level of review necessary to provide the data required for
management is also recommended, to maximize cost efficiency. To investigate minimum EM
review rates, we prototyped a simulation tool based in R, EMoptim, that uses stratified random
sampling to address one or more fishery monitoring objectives. EMoptim also incorporates a
cost function, developed based on pricing estimates for analysis of EM imagery and associated
information. Using EMoptim, fishery-specific information can be used to fine-tune review rates,
within specified limits including cost, and across a suite of fishery monitoring objectives.
We applied EMoptim using publicly available information from longline and purse seine
fisheries operating in the western and central Pacific Ocean, and the scientific literature. Results
confirmed that minimum effective review rates increase as catch frequency decreases, and as
the required coefficient of variation decreases. Stratified sampling approaches were effective in
reducing the level of review required for more commonly caught taxa. However, stratification
4
had little effect on review rates for rare capture events that were geographically widespread. As
a result, significantly higher levels of EM review are required to estimate numbers of rare
events effectively.
EM programs often include multiple monitoring objectives, and we used EMoptim to explore
optimized rates of EM review required to estimate target and bycatch catch, to achieve specified
coefficients of variation. Outputs highlight that optimizing review regimes for different
monitoring objectives is most effective among more commonly caught species. The required EM
review rate increases dramatically when rarely caught species are considered, such that
“optimizing” at a lower review rate is not effective for monitoring these taxa. Outside strata with
higher review rates set using EMoptim, we recommend that a minimum baseline level of
random review should be maintained to enable detection of fishery changes.
EM has great potential to collect data cost-effectively at scale to support fisheries management.
Information requirements that can be met by EM are broadly consistent across RFMOs and
other management bodies. Furthermore, service providers operate across jurisdictional
boundaries. Therefore, there is significant potential and opportunity to accelerate the
development and adoption of methods to optimize EM review, both in the immediate future and
longer term.
5
Contents
Glossary ................................................................................................................................................................................ 1
Executive summary ......................................................................................................................................................... 3
1. Introduction .............................................................................................................................................................. 6
2. RFMO fishery information requirements ...................................................................................................... 9
3. EM adoption by RFMOs ..................................................................................................................................... 13
4. Case studies ............................................................................................................................................................ 14
4.1. EM adoption by industry to demonstrate responsible fishing practices:
Association of Large Tuna Freezers (AGAC) – Pacific, Atlantic and Indian Oceans ...................... 14
4.2. EM to verify fisher reported catch and discarding of a quota-limited species: USA ...... 17
4.3. EM to support management of fishery discards and incidental bycatch: Chile ............... 19
4.4. EM to provide catch composition information for fishery management:
Alaska Fixed Gear fishery: USA ........................................................................................................................... 22
4.5. Regional EM initiatives: Pacific Ocean tuna fisheries .................................................................. 24
5. Optimizing EM review ........................................................................................................................................ 28
5.1. Approaches to EM review ....................................................................................................................... 28
5.2. Increasing EM review efficiency .......................................................................................................... 29
5.3. Costs of EM review .................................................................................................................................... 32
6. EMoptim: a prototype tool to evaluate EM review rates...................................................................... 37
6.1. Exploring EM review rates ..................................................................................................................... 37
6.2. Optimizing EM review rates .................................................................................................................. 43
7. Best-value approaches to EM review: present and future .................................................................. 43
8. Acknowledgements ............................................................................................................................................. 47
9. References ............................................................................................................................................................... 48
Appendix 1. Data requirements that support fisheries management by selected Regional
Fisheries Management Organizations. ................................................................................................................. 55
Appendix 2. EM review rate evaluation and optimization by EMoptim ................................................. 59
6
1. Introduction
Electronic monitoring to support fisheries management
Monitoring commercial fisheries is essential for their effective management. Human observers
have been a mainstay of on-vessel fisheries monitoring, used alongside methods such as
position monitoring through satellites, at-sea patrols and aerial surveillance28,44. While
monitoring by human observers can work well, challenges such as occupational safety,
representativeness of data collected, and cost, have catalyzed the development of
complementary monitoring methods.
Electronic monitoring (EM) using on-vessel cameras is a fishery monitoring tool that has
developed since the late 1990s. In that time, EM has been trialed in more than 100 fisheries and
operationalized in some, to address a range of fishery monitoring objectives17,71,89. In addition to
the cameras that record fishing activity, typical functions and components of EM systems
include GPS tracking, a control unit that monitors the operation of the system and records data,
satellite reporting of system status, and sensors that indicate fishing activity (Figure 1).
From a fishery management perspective, key benefits of EM compared to other monitoring tools
include17,25,49,84,89:
• the capability to collect high quality, comprehensive and detailed information on fishing
activities
• flexibility of the monitoring method which enables scaling across fleets and in
accordance with risk and evolving management priorities,
• ability to support incentive-based management (including fishery access and market-
based incentives),
• verification capability; and,
• relative cost efficiency.
Significant broader benefits include the avoidance of health, safety and logistical challenges that
human observer deployments can involve25,49,51.
While the benefits of EM are well-recognized, there are perceived barriers impeding its
adoption. Barriers are largely human-focused rather than technological, e.g. culture change
required to accept working in a monitored environment, and the need to encompass a new
monitoring tool in existing regulatory and management frameworks (e.g. where regulations
specify that human observers must be used to meet at-sea monitoring requirements)49,89.
Another key barrier is cost50. The perception of cost impacts is heightened by costs being
immediately calculable and incurred in the short-term (and on an ongoing basis). In contrast,
benefits may be variable and accrue in a longer timeframe49,51. Further, EM programs generally
cost more per unit of monitoring effort and information in their initial stages (pilot or trial
programs), becoming cheaper when scaled up and implemented as operational programs49,68. If
the operational stage is never reached, the cost to benefit ratio of EM cannot be optimized,
recognized or realistically compared with other monitoring tools.
The costs of EM can be partitioned into fixed and variable components (Figure 2). Fixed costs
include the EM system hardware, installation onboard vessels and some maintenance elements.
Variable costs can include administration of the EM program, software, and review of the EM
imagery and associated information. Data storage may be a fixed or variable cost. Fixed and
variable costs and the ratio of these cost types vary with program, vessel and fishery-specific
factors, e.g. monitoring objectives, fishery scale, geographic location, level of engagement and
support from industry operators and management bodies, and program standards (which may
7
include specifying imagery review rates)84. Beyond monetized costs, so-called “soft costs” are
the on-vessel changes in operational practice that can support or improve the efficacy of EM
(e.g. catch handling protocols)65.
Information needs of Regional Fisheries Management Organizations
Among fishery management entities, Regional Fisheries Management Organizations (RFMOs)
are multilateral bodies that hold critical fishery management responsibilities across most of the
world’s oceans. RFMOs comprise countries (represented by their governments) which may be
termed members, parties and contracting parties. Countries that are not full RFMO members
may hold other status, e.g. as cooperating non-members or cooperating bodies. The
management roles of RFMOs are typically defined in relation to fished species within a
particular geographic area. RFMOs are focused on sustainable management of focal species,
which also involves managing the impacts of fishing activity on non-target species and the
marine environment (Appendix 1).
To support fishery management, RFMOs set requirements for information collection and
monitoring, control and surveillance (MCS) within their areas of competence. Human observers
are a common component of on-vessel MCS and minimum levels of observer coverage are often
specified (e.g. by fishing day, trips, vessels, or hooks). However, these minima do not necessarily
reflect data requirements for robust fisheries management (e.g. to effectively characterize catch
composition4,7,42,90). Furthermore, coverage achieved by some RFMO members falls below levels
required on an ongoing basis39,70,95.
EM for RFMOs and this report
The emergence of EM as a fishery monitoring tool has led some RFMOs, and their members, to
evaluate the possibilities for EM-based data collection. This has included considering data
requirements that may be met using EM, and standards for data collection. Members of RFMOs
managing tuna fisheries have been particularly active in this regard19,55,73,76,93. Furthermore,
opportunities for accelerating EM adoption have been identified49,50,51. However, foundations for
structuring EM review have seldom been investigated analytically, including trade-offs of data
quantity, quality and cost. This is despite review processes being highlighted repeatedly as a
vital consideration for EM program design and cost management17,65,84.
In this report, we focus on the review of EM imagery and associated data that can be used to
support RFMO fishery management objectives. We:
- identify the RFMO fishery management objectives that can be supported by information
collected using EM
- present case studies of EM implementation, to show how fishery management and
monitoring objectives are being met using EM in the real world
- demonstrate a prototype simulation tool, EMoptim, to explore EM imagery review rates
that provide information supporting fishery management, and associated costs; and,
- illustrate how monitoring costs incurred at the review stage of EM programs can be
reduced, while optimizing the suite of data collected.
We focus on RFMOs managing tuna fisheries, while reflecting broader application of findings
across other management entities and fishing methods.
We consider EM review as the process of extracting and processing data collected by EM
systems into a form ready for consideration by end-users. We do not consider broader review-
related elements of an EM program (e.g. training, data management, data storage).
8
Figure 1. Generalized schematic showing electronic monitoring system components on a fishing vessel.
Figure 2. Overview of EM program costs. Source: Michelin et al. 2018.
9
2. RFMO fishery information requirements
RFMO information requirements are defined by the objectives or purposes of these
management bodies. Key themes among RFMO objectives are sustainable use and conservation
in the long-term. Both fished species and non-target species are in-scope for management.
Among those considered in detail here, one RFMO explicitly includes ecosystem protection in its
overarching objective (Appendix 1).
Principles, functions and actions (PFAs) specified in RFMO convention texts provide insights on
how the objective or purpose of RFMOs is defined and may be addressed (i.e., what
conservation or sustainable use means in practice). PFAs can be grouped into three categories:
biological, environmental and operational (Figure 3). Key biological PFAs include maximum
sustainable yield (MSY) for focal or target species that are fished, and ensuring non-target
species affected by fishing activities are maintained above levels at which reproduction may be
threatened. Taking account of biological uncertainties may also be highlighted (Appendix 1).
Broader environmental PFAs include addressing pollution originating from vessels, lost gear,
and ecosystem impacts. Operational PFAs cover implementation and compliance, e.g.,
determination of total catch and fishing effort, adopting evidence-based management measures,
and ensuring compliance with binding measures. Some conventions also include a specific
requirement for a precautionary approach (Appendix 1). While each invokes specific data
needs, there is significant overlap such that some data support multiple PFAs.
Figure 3. Schematic diagram of linkages between fishery management objectives, principles, functions and actions, and
information and data needs. The two-way flow indicates that each layer informs the other on an ongoing basis.
Information needs
Principles, functions, actions
RFMO Objective
Biological OperationalEnvironmental
Data fields
Environmental
impact
Stock/
population
size
Implementation Compliance
Catch Effort Fishing
gear
Bycatch
mitigation
Operational
data
10
A significant amount of the data needed to support RFMO management can, and must, be
captured onboard fishing vessels during fishing operations. As an on-vessel monitoring tool, the
data that EM can collect in support of fisheries management traverses all PFAs. For example,
fishing catch and effort data are vital for assessing the fishery impacts on target stocks and non-
target species, supporting the development of management measures, evaluating compliance
with management measures, and assessing the impacts of fishing on the environment. In recent
years, some RFMOs (and their members) have investigated whether their data requirements
can be met using EM. Such evaluations have typically included a comparison of EM capabilities
with data recorded by fisheries observers55,93.
An overview of data requirements that can be effectively met using EM is set out below, for five
fishing methods used in the RFMOs considered in Appendix 1. However, given ongoing
technological and practical developments in EM systems and applications, considering how EM
can meet management and monitoring objectives is recommended as any monitoring program
is conceived. What has been achieved to date provides a baseline but does not limit future
possibilities.
Catch and discard information
More than 75 EM projects or programs have been conducted worldwide with the objective of
monitoring catch. Focal catch components have included target species, fish bycatch, and
bycatch of endangered, threatened and protected (ETP) species and other megafauna. Data
recorded have included retained and discarded catch species, size, and life status17,71,89.
Information capture using EM is most straightforward when catch items come aboard serially
(e.g. piece by piece on a longline) or in smaller clusters (e.g. gillnets), compared to when catch is
landed on deck or into storage holds in bulk (e.g. purse seine and trawl methods). For all gears,
catch handling protocols may facilitate enumeration, identification, size and life status
assessments of catch items. For bulk fishing gears, catch handling protocols are essential to
support quantitative data capture from larger catches using EM46.
Discarded catch items may be landed on deck (for enumeration and identification with the rest
of the catch prior to discarding) or removed from gear without being brought aboard. For catch
discarded after being brought aboard, EM and landed catch reconciliations (e.g. conducted by
dockside monitors) may be viable monitoring methods46. When catch items are removed,
released in the water or dropped from gear before being brought aboard, EM-supported
enumeration is achievable (with appropriate camera placement) though the view may not
enable identification to the same level of granularity as when catch items are brought aboard
(e.g. to family or genus level, rather than species). Similarly, determining life status and size is
less achievable when catch items are removed, released or dropped directly into the water, and
not brought aboard vessels31.
Catch and discard data inform RFMO information needs relating to stock/population status of
species caught, implementation of fishing operations, and compliance with management
requirements (Figure 3).
Fishing effort
Among more than 100 trial and operational EM programs worldwide, monitoring fishing effort
has been the most common objective. The efficacy of monitoring fishing effort is demonstrated
across the longline, purse seine, trawl, gillnet and pot/trap methods17,89. The duration of fishing
activity may also be used to define and quantify fishing effort (i.e. hours fished), and for purse
seine fishing, effort characteristics include searching and setting time and whether sets are
made on fish schools associated with floating objects, or unassociated schools.
11
Fishing effort data are relevant across the four categories of RFMO information needs (Figure
3).
Fishing gear
EM can be effective in capturing imagery of some fishing gear characteristics, e.g., presence of
floats and weights on longlines, presence of shark lines, and characteristics of floating objects
used in purse seine fishing19,29.
The presence of some bycatch mitigation devices is also discernible from EM imagery20,71. For
example, sorting grids used to reduce ETP bycatch in trawl fisheries can be detected as gear is
deployed. Wire traces (associated with increased shark bycatch, and prohibited in some
fisheries), and tori lines (also known as streamer lines) used to reduce seabird captures in
longline and trawl fisheries are detectable in EM imagery (though tori line dimensions are not
currently well captured). Pingers deployed on gillnets to reduce cetacean impacts are
detectable. Seal exclusion devices and some operational practices (e.g. backdowns to release
marine mammals from purse seines) to reduce ETP captures are also expected to be detectable1,
73.
Fishing gear characterization is relevant to RFMO information needs including catch per unit
effort, stock/population status of species caught, implementation of fishing operations, and
compliance with management requirements (Figure 3). Broader environmental impacts of
fishing gear may include accounting for lost gear (e.g. reconciling gear hauled against gear set).
Bycatch handling
Handling practices used to remove bycatch from the gear and release it into the water affect
post-release survival96. Some fisheries management entities including RFMOs have adopted
mandatory provisions for carrying release equipment (e.g. dehookers and line-cutters16) and
best practice handling guidelines to promote post-release survival of live-captured animals (e.g.
supplements to the Western and Central Pacific Fisheries Commission’s (WCPFC) Conservation
and Management Measure (CMM) 2018-03 for the safe handling and release of seabirds, and
CMM 2019-04 for some sharks, mantas and mobulids. EM can be used to collect information on
bycatch handling, as well as to identify opportunities to improve handling (e.g. developing
guidance materials and training17). This information is relevant to RFMO fishery impacts on
populations of species caught, implementation of fishing operations, and compliance with
management requirements (Figure 3).
Operational data
A range of general operational data characterizing fishing activities is readily collectible using
EM, e.g., the date, time and location of various fishing activities including (but not limited to) the
start and end of sets and hauls73,89,93.
Operational fishery data is critical for addressing all categories of RFMO information needs
(Figure 3). While not in-scope for this report, the potential for EM to contribute to monitoring of
labor and human rights onboard fishing vessels has also been identified25,49.
12
Table 1. Data required for fisheries management, that can be obtained in whole (
) or in part (*) using electronic monitoring systems in commercial fisheries. Catch fate includes whether
catch is released alive or dead, and injury status. In general, less detailed information is expected to be obtained for discarded catch items because at least some animals are released before
being landed on the vessel. For example, animals may be identified to genus level rather than species level, if released while still in the water and not subject to detailed examination. Life
status may also be difficult to estimate effectively for catch released without being brought aboard. The presence of line weights can be detected by EM, while the weight of weights may not
be. FAD = Fish Aggregating Device, used in purse seine fisheries.
RFMO
principles,
functions or
actions
Catch and discard information
Gear and operational information
Catch
species
/ stock
Landed
catch
size
Discarded
catch
species /
stock
Discarded
catch size
Discarded
catch life
status
Hooks
set,
hauled
Floats
Set / haul
time /
location
FAD use,
type
Gear not
retrieved;
discarded
Biological
Target species
*
*
*
*
Non-target
species
*
*
*
*
Environmental
Environmental
impacts
*
Operational
Implementation
and compliance
*
*
*
*
Objectives
Bycatch mitigation usage information
General
Tori
lines
Line
weights
Hook-
shielding
devices
Dyed bait
Bird
curtain
Fish
waste
discharge
Wire
traces
Dehooker /
linecutter
use
Bycatch /
unwanted
catch
handling
Marine
pollution
Biological
Target species
*
*
Non-target
species
*
*
Environmental
Environmental
impacts
*
*
Operational
Implementation
and compliance
*
*
13
3. EM adoption by RFMOs
RFMOs are at different stages in the progression of EM. For example, the Inter-American
Tropical Tuna Commission (IATTC) held its first workshop on the implementation of electronic
monitoring in 2021. This followed the 2019 Commission resolution (C-19-08) that the IATTC
Scientific Staff would prepare a draft proposal for the development of minimum standards for
EM implementation on longline vessels for consideration by the Scientific Advisory Committee
in 2020. Subsequent work undertaken has included the development of agreed definitions for
EM-related terminology, a proposed framework for EM implementation for longline and purse
seine vessels (including draft minimum standards, data collection and reporting requirements,
institutional structure supporting an EM program, and data management, among other
content)35, and a workplan for the introduction of EM34. The workplan identified 1 January 2025
as the date at which the Electronic Monitoring System should be operative on longline and
purse seine tuna fishing vessels, subject to Commission agreement.
The International Commission for the Conservation of Atlantic Tunas (ICCAT) recommended
the adoption of minimum standards for purse seine vessels on which EM was voluntarily
implemented in 2016 and 201780. In 2021, the ICCAT Subgroup on Electronic Monitoring
Systems was established to consider EM, with a focus on billfish and longline fishing, noting that
other methods would require attention in due course (e.g. gillnet). Group recommendations
included that focal species should be expanded, to include sharks, albacore tuna, and other
species. The first meeting of the ICCAT Working Group on Electronic Monitoring Systems (WG-
EMS) took place in early 202294.
The Indian Ocean Tuna Commission (IOTC) adopted preliminary minimum standards in 2017
for purse seiners voluntarily using EM to augment observer coverage. The development of
minimum standards for all IOTC fisheries was recommended in 2018 by the Scientific
Committee. In 2020, EM data capture capabilities were documented for purse seine, longline,
gillnet and pole and line vessels greater than 24 m in overall length, and vessels under that
length using the same or other methods when operating in the high seas 55. Areas of
consideration for EM program and data standards were also summarized and minimum
requirements and the definitions of key terms stated. The IOTC Ad-hoc Working Group on the
Development of Electronic Monitoring Programme Standards (WGEMS) held its first meeting in
late 202138. The group adopted a workplan which identified the facilitation of pilot EM projects
and development of minimum data standards as the highest priority work areas for 2022/23.
WCPFC has held five meetings of its intersessional working group on electronic reporting and
monitoring to date (ERandEM-IWG). Proceedings of the 2020 meeting included consideration of
draft minimum standards for that RFMO’s electronic monitoring program22. These draft
standards include program standards (e.g. the independence and impartiality of EM programs),
technical standards (e.g. requirements for camera capabilities, tamper-evident systems,
malfunction alerts), logistical standards (e.g. operational procedures to ensure the secure
collection and distribution of data storage devices) and data analysis standards (e.g. analyst
training, data entry checks, sub-sampling considerations for audit-based review). WCPFC has
also drafted a consultative proposal for a future CMM for a regional EM program23. At its 2022
meeting, the ERandEM-IWG updated its workplan including the consideration of integrating EM
with other elements of the management framework (e.g. its regional observer program) and
progressing the drafting of the EM CMM24.
Members of the Commission for the Conservation of Southern Bluefin Tuna (CCSBT) have
contributed papers on EM to meetings of this RFMO and its subsidiary bodies over time, and the
2021 meeting of the Compliance Committee discussed the monitoring method. That meeting
14
recommended that EM systems be the main item of discussion for the Technical Compliance
Working Group in 202212. Information sharing on electronic monitoring is also a workplan (and
agenda) item for CCSBT’s Ecologically Related Species Working Group.
The North Pacific Fisheries Commission (NPFC) is in the preliminary stages of considering the
potential for EM as a monitoring tool in the fisheries in its area of competence77. At its 2021
meeting, the NPFC Scientific Committee tasked subsidiary bodies with reporting to its next
meeting on the potential use of EM (and other data collection methods) to address data needs
and gaps for NPFC priority species and non-target species78.
4. Case studies
The following five case studies exemplify the diverse application of EM in real-world fisheries to
date. Case studies demonstrate EM adoption by fishing industry and government bodies
operating within Exclusive Economic Zones, in areas beyond national jurisdictions, and within
RFMO areas of competence. Monitoring objectives include assessing conformance with good
practice measures, monitoring catch limits of quota-managed species and broader
characterization of catch composition, and detecting incidental bycatch and discarding. Case
studies were compiled using published sources and additional information provided by
program participants listed below.
4.1. EM adoption by industry to demonstrate responsible fishing practices:
Association of Large Tuna Freezers (AGAC) – Pacific, Atlantic and Indian Oceans
Information contributors and reference sources: M.A. Herrera, J. Morón, I. Moniz, J. López
(OPAGAC-AGAC), G. Legorburu (Digital Observer Services (DOS)), I. Canive (Datafish Technology
Solutions), J. Ruiz (AZTI); 43, 52, 53, 72, 74, 75, 76
Purpose of EM
AGAC is an industry body that represents the interests of vessels registered in nine countries
(Spain, Belize, Curacao, Ecuador, El Salvador, Guatemala, Panama, Tanzania and the Seychelles).
The AGAC fleet comprises 48 purse seiners and 10 support vessels.
AGAC adopted a Code of Good Practice (CGP) in 2012. The main objective of the CGP68 is to
reduce the environmental impact of the AGAC fleet’s fishing activities. Vessel crew are
responsible for implementing the Code. EM is one of the methods used to monitor conformance
with the Code, primarily:
• safe release of sensitive bycatch species (e.g. sharks, turtles and marine mammals); and,
• the use of non-entangling FADs.
EM is also used in some cases to monitor retained non-target catch and collect other scientific
monitoring information.
Context for EM implementation
The AGAC fleet operates in the Pacific, Atlantic and Indian Oceans, in the high seas and various
EEZs. The fleet targets tropical tunas, mostly in association with drifting Fish Aggregating
Devices (dFADs). Full documentation of AGAC purse seine vessel activities has been compulsory
since 2015. This was extended to support vessels from 2017, and is achieved through a
combination of human observers and/or EM. At present, EM systems are installed on 28 purse
seiners and all support vessels.
15
The EM program
The first EM pilot program was conducted in the AGAC fleet in 2011 – 2012. The pilot program
involved comparing data collection from EM and by human observers on three purse seine
vessels to investigate the efficacy of EM. Additional trials followed using EM systems from a
range of EM service providers. This work underpinned the development of the minimum EM
standards for tropical tuna purse seine fisheries, published by the International Seafood
Sustainability Foundation in 2014.
EM-based verification of AGAC’s CGP was first investigated on support vessels in 2015. Two
years later and coincident with the adoption of preliminary EM standards by ICCAT and IOTC,
science provider AZTI incorporated data obtained through EM in its verification of the AGAC
fleet’s conformance with the CGP.
The AGAC program is designed to capture 100% of fishing trips. Therefore, EM systems record
every day for 24 h/day. Recording frame rates can be configured based on sensor and/or GPS
data. Support vessels carry 2 – 3 cameras each, to document FAD-related activity. Purse seiners
are fitted with 4 – 8 cameras per vessel, to enable monitoring of all fishing-related activities.
Crew routinely clean camera covers, but otherwise have no need to engage with the EM system
at sea. Hard drives are used to capture EM imagery and associated information. Vessel captains
are responsible for shipping drives to the EM review service provider.
AGAC has adopted EM as a core component of its monitoring approach for several reasons:
• Economic: monitoring costs were generally lower for EM than other methods, and EM
has also proven more time efficient.
• Logistical: EM circumvents the logistical difficulties associated with boarding
observers, which would otherwise be necessary to meet flag state and some coastal
state requirements.
• Comprehensive data collection: EM can be used to monitor all activities on the vessel,
including when these occur concurrently in different locations (e.g., brailing on the
upper deck and loading of catch on the lower deck), and in areas unsafe for people.
Such activities would not be possible for a human observer, therefore EM data are
more complete for some tasks.
• Information validation: EM enables the objective validation of divergent reports of
vessel activities, e.g., when observer and skipper reports have differed, EM information
has been used for conflict resolution.
Other benefits of EM adoption recognized by AGAC include high acceptance overall among
vessel owners and most crew (noting that some vessel owners retain a preference for human
observers, due to crew preference for not being monitored by cameras), independence of
monitoring, tamper resistance of systems, the ability to review imagery multiple times as
required, and that high levels of monitoring can still be achieved when the health and safety of
human observers may be compromised (including where there are piracy threats and during
the COVID-19 pandemic). Space available onboard vessels is also not a limitation for EM,
whereas this is an important constraint for human observer deployments on support vessels.
Challenges for EM operation were technological and operational, including equipment failure
and maintenance needs in remote locations, inability to collect biological samples, difficulties
with some species identifications, and the current time delay between the collection and
extraction of EM information (e.g. due to hard drives being shipped on return to port after
lengthy trips). In the broader operational environment, AGAC considers that the adoption of EM
16
minimum standards by management bodies (e.g. RFMOs) will foster acceptance and
implementation of EM among flag and coastal states.
A census approach is taken to EM review for monitoring conformance with CGP bycatch
handling and FAD-related requirements. Algorithms in the analytical software are used to
identify different vessel activities based on characteristics such as vessel speed and course, and
georeferencing is in place with position, date and time.
Data routinely captured during EM review include:
• For sensitive bycatch species: species identification (and sex, where possible), size,
origin (encircled, entangled by purse seine net/on FAD), location of release (net, brailed
to upper deck/lower deck), handling (including the tools used, e.g. hopper, stretcher,
etc.), time spent from capture to release, and condition at release (to estimate fate).
• For FADs: identification of the device (using the tracking buoy); materials used on the
surface and for the underwater structure of FADs, both for new deployments and
visits/encounters of dFADs already in the water; any modifications made to FAD
structure.
EM review service providers have put standards in place to reduce bias and improve
consistency among EM analysts (termed dry observers).
Key Performance Indicators (KPIs) used to monitor the time it takes to extract EM data
collected include:
• Time between the end of the last trip recorded and receipt of the hard drive by the
review company (generally less than 2 months, but a broad range of 15 days to 15
months depending on landing location)
• Time spent between receipt of the hard drive and completion of data extraction from EM
onto data forms (35 days on average, ranging from 7 – 75 days)
• Observer working hours, calculated as an Analysis Ratio of Sea Days/Office Days. Office
days are considered as the sum of hours dedicated to one fishing trip divided by the
office day of eight hours. (Ratios show that it takes 4 – 8 times more working hours for a
human observer at sea to complete purse seine trip records compared to EM analysis by
a dry observer. For support vessels, this figure is 10 – 15 times).
Program development
The performance of the EM program is reviewed annually by AZTI and the EM service
providers. This review includes an assessment of any changes that may be needed to improve
monitoring or to address new requirements. For example, the CGP is updated regularly to
incorporate new provisions or amend the existing ones based on the results of research
activities. In addition, UNE 195007 was adopted in Spain in 2021, requiring some updates for
AGAC vessels. (UNE 195007:2021 is the first European standard developed to harmonize
requirements for the use of on-board cameras among industry and data users88). The
development of AI and machine learning is expected to support improvement in EM analysis
KPIs in future.
Overall, AGAC reports that for monitoring fleet compliance with the CGP, data collected using
EM are as good or better than data collected by human observers.
17
4.2. EM to verify fisher reported catch and discarding of a quota-limited species:
USA
Information contributors and reference sources: B. McHale, I. Miller (Office of Sustainable
Fisheries, NOAA), M. Wealti (Saltwater Inc.); 57
Purpose of EM
In the Atlantic Highly Migratory Species (HMS) fishery, catches of Atlantic bluefin tuna (Thunnus
thynnus) are quota-managed. It is required that both landings and dead discards are accounted
for within the US national quota (established through binding recommendations of ICCAT). This
operational EM program was initially designed as a tool to audit logbook information on bluefin
tuna being retained and discarded in the pelagic longline fishery. More recently, the program
has been broadened to include monitoring catch, retention and discarding of shortfin mako
sharks (Isurus oxyrinchus).
EM is intended to provide an incentive for accurate logbook reporting; it gives the National
Marine Fisheries Service the ability to verify vessel owner/operator catch and discard records
for these two species, and their life status at the haul.
Context for EM implementation
In 2006, conservation and management measures for the Atlantic bluefin tuna (BFT) were
updated, through Amendment 7 of the 2006 Consolidated Highly Migratory Species Fishery
Management Plan. This included the development of new management measures for the pelagic
longline fishery. Targeting of Atlantic bluefin tuna (BFT) by pelagic longline fishers is not
permitted, though the species is caught in the course of fishing for swordfish (Xiphias gladius),
yellowfin tuna (Thunnus albacares), and bigeye tuna (Thunnus obesus). As part of Amendment 7,
the Individual Bluefin Quota Program (IBQ) was developed. For several years prior to the
development of this program, catches (landings plus dead discards) of BFT by pelagic longline
vessels had regularly exceeded the longline quota. The IBQ was developed to incentivize pelagic
longline fishers to minimize interactions with this species, and to support individual
accountability for BFT catch. While landing limits had been in place previously, management
changes included new discard limits. EM was introduced to provide an independent verification
measure for logbook reporting of landings and discards.
Amendment 11 to the 2006 Consolidated HMS Fishery Management Plan addresses Atlantic
shortfin mako sharks64. The North Atlantic shortfin mako shark stock is overfished, and subject
to overfishing. ICCAT Recommendation 17-08 sets out the multilateral management
requirements for this species, which Amendment 11 is designed to respond to. The measures in
Recommendation 17-08 are expected, by ICCAT, to prevent further deterioration in the status of
the shortfin mako stock, stop overfishing and enable the stock to start to rebuild. Up to July
2022 in the commercial pelagic longline sector of the Atlantic HMS fishery, all live shortfin mako
sharks were required to be released alive with a minimum of harm (with due consideration of
crew safety). Only shortfin makos that were already dead on haulback could legally be retained.
Reported disposition of shortfin makos at the haul was verified using EM. Since 5 July 2022,
release of all shortfin makos has been required, regardless of life status67.
EM was selected for implementation in the pelagic longline component of the Atlantic HMS
fishery because pilot programs elsewhere had proven the capabilities of the monitoring method.
Implementation could also be supported by the budget available for fishery management. The
audit design envisaged by managers would be difficult to achieve with human observers; EM
system presence on all vessels and 100% capture of fishing activities enables monitoring bias to
be eliminated.
18
The EM program
The EM program was initiated on 1 June 2015. Around 112 vessels have EM systems installed,
and 80 of these vessels are currently fishing. Most vessels carry two cameras. Cameras must be
installed to (i) record close-up images of the deck near the hauling bay or processing area,
where gear is retrieved and catch is removed from the hook (this view is intended to collect
information on species identification and length estimation), and (ii) record fish that are caught
and discarded without being brought aboard, as well as whether those fish are alive or dead
when released. Some vessels require a third camera if these views are obstructed (e.g. by
structures on the vessel). Overall, camera views must capture fish from when they appear at the
surface, through to being brought aboard and ultimately discarded, or processed and stored.
Two sensors are fitted as part of the EM setup: a hydraulic pressure sensor and a rotation
sensor on the longline reel. These show when the fishing gear is engaged and trigger the
cameras to record only when the sensors indicate the gear is being hauled back. This haul-only
recording was in direct response to a request made by the fishing industry when Amendment 7
was first announced.
NMFS selects specific sets for review at the end of each quarter. The sample design is stratified
based on historical data on where and when BFT have been caught. Hauls are sampled for
review within this stratification, aiming to capture trips with BFT interactions. The goal is to
review 10% of hauls and at least one longline haul from each active vessel each year. The extent
of review is subject to budget allocations.
The focus of review is compliance with the landing and discard requirements for BFT and
shortfin mako sharks. Date, time, location and disposition of the two focal species are the key
data fields recorded from EM imagery and associated information. A key challenge with the
audit model has been alignment of data fields from EM and VMS-reported data from vessels (e.g.
local time versus UTC time stamps). Also, hauls selected for review based on the stratification
approach and vessel reporting information have not always been available for EM review, e.g.
due to hard drive loss.
The clearly defined scope of review has contributed to the success of the program. Currently,
the data collected are used for compliance purposes, with the EM program acting as an incentive
for accurate reporting by those on vessels. The level of review makes the detection of
underreporting trends difficult. However, the IBQ limits are not being met or exceeded
currently, diminishing the imperative to increase review rates.
Beyond providing assurance of the quality of vessel reporting, EM has delivered other
advantages. These include the ability to review imagery and associated information more than
once. EM also provides a basis for the continued evolution of management. A third unforeseen
advantage of EM has been vessels utilizing EM-collected information for legal reasons.
Program development
Activity recognition (AR) software has been in development to improve the efficiency of review.
AR is intended to facilitate reviewer detection of fish on the line, by recognizing human
activities associated with fish captures (e.g., increased movement of crew, presence of fish
shapes). Critical to the success of AR will be testing to demonstrate that the accuracy of review
is maintained (or increased). Initial testing showed comparable results between human and AR-
facilitated review for retained and total catch. However, discard detection was unacceptably low
(less than 70% of discards were detected). Development of the software continues.
19
A new rule has been proposed to require the installation of booms to improve camera views
over vessel rails, and a measuring grid to enable recording of lengths of retained catch and
smaller catch items. Future developments could include real time transmission, while this is not
feasible within the program budget currently.
4.3. EM to support management of fishery discards and incidental bycatch: Chile
Information contributors and reference sources: L. Cocas (SUBPESCA), R. Toro (SERNAPESCA);
14, 15, 37
Purpose of EM
EM is being implemented in Chilean fisheries to improve fisheries sustainability and facilitate
high-end market access. EM implementation to date has focused on monitoring compliance with
regulations applying to discarded catch and incidental bycatch (seabirds, marine mammals,
turtles, sharks and rays), as well as fishery regulations on fishing locations and gear.
Context for EM implementation
Two government agencies hold responsibility for fisheries management in Chile - the
Undersecretariat for Fisheries and Aquaculture (SUBPESCA) sets fisheries and aquaculture
policies, regulations and management measures. The National Fisheries and Aquaculture
Service (SERNAPESCA) conducts monitoring, compliance and enforcement (e.g. conducting EM
review and applying sanctions). Additionally, the private Fisheries Development Institute
(IFOP) is in charge of research in fisheries and scientific observation programs. IFOP provides
the information used to make management decisions.
In 2001, a broad prohibition on discarding was introduced in Chilean fisheries. Sanctions were
introduced for violating the prohibition, without at-sea monitoring in place. The substantial
penalties in place impeded the acquisition of information about discarding, and the extent of
discarding remained unknown.
In 2012, the Chilean government reviewed fisheries legislation such that the main objective
became the conservation and sustainable use of marine resources, incorporating ecosystem and
precautionary approaches. The revised legislation identified discards and incidental catch
(including seabirds, marine mammals and sea turtles), and specified control mechanisms for
these. Additionally, the legislation provided for vessel-specific exemptions to the discard
prohibition, conditional on at least two years of fishery-based research or monitoring. The
purpose of the two-year exemption period was to enable an unbiased quantification of
discarding and incidental bycatch, to understand these events, and to develop (with the sector)
proposals for how to address both issues. Proposals were later translated into mandatory plans
for each fishery, which contain management measures and technological means to reduce both
discards and bycatch, handling protocols, codes of good fishing practice, a continuous scientific
monitoring and compliance program, training and dissemination programs, etc. At-sea
observers and fisher logbooks were key sources of information throughout this process. New
management categories were implemented: species for which (i) discarding is prohibited (all
species with quota and species for human consumption), (ii) discarding is authorized (damaged
specimens, species with no current commercial value) and (iii) return to the sea is mandatory
(all bycatch, chondrichthyans, prohibited species and species not subject to exploitation).
Discards have been considered in total allowable catches since 2018. Now, procedures are in
development for the explicit incorporation of discards in fishing permits and licenses.
The difficulty of controlling discarding at sea led to the incorporation of EM in the revision of
the fisheries and aquaculture law in 2012 (Law No. 20.625, 2012). This enabled the use of EM to
support the management of discards and incidental bycatch, by detecting and recording any
20
discarding actions occurring (thereby enabling the monitoring of compliance with the reduction
plans). To implement this law, Supreme Decree No. 76 (2015) sets out the requirements for EM
systems on both industrial and smaller-scale vessels. Regulatory provisions include, for
example, EM system design and technical specifications, number and location of cameras by
fishery, details on image collection, processing and confidentiality, obligations of vessel owners,
and the role of SERNAPESCA. There is also a complementary Resolution that sets out the
technical standard for an EM system.
By 2022, 11 discard and incidental bycatch reduction plans were established, covering 17
fisheries (both artisanal and industrial), and other plans are still in the research phase.
Additionally, the list of species subject to such plans for each fishery, and the associated
management classification (i.e. prohibited discard, authorized discard, mandatory return) is
updated annually by Resolution. This Resolution also set out requirements for incidental
bycatch, including the use of mitigation devices and handling practices. All the industrial
fisheries subject to reduction plans are being monitored by EM.
The EM program
The use of EM by Chile’s industrial fleet (> 18 m in length) has been mandated since January
2020. The fleet comprises vessels using demersal and midwater trawl, purse seine and longline
fishing methods. EM roll-out in this fleet was preceded by extensive research, hundreds of
meetings with industry participants, and other preparatory work. The EM program was
immediately implemented as an operational one (no pilots occurred). In 2020 and 2021, 109
and 92 vessels, respectively, were operating in the fleet and all were covered by EM. (Vessel
numbers vary year to year, e.g., due to vessels being sold, repaired, or moving into a different
sector of industry). SUBPESCA and SERNAPESCA collaborated extensively throughout the
development and implementation of regulations relating to bycatch and discarding, to ensure
that the required measures could be efficiently monitored using EM. Collaboration is ongoing to
address issues arising, e.g., requiring additional cameras in some fisheries.
EM captures fishing trips from when vessels depart through to when they return to port. (The
regulatory framework also enables the use of sensor-based systems). Vessels carry 2 – 8
cameras each, depending on the vessel size, fishery and fishing operation, and extent of catch
processing that is undertaken onboard. In general, hard drive storage capacity determines the
schedule of collection and a range of days is specified by law. However, operators can be
compelled to provide hard drives at any time on demand from various authorities, and for
administrative or compliance purposes.
Vessel owners are legally responsible for the costs of EM equipment, installation and
maintenance. These costs are set by the chosen EM supplier.
The law enables the collection and processing of EM imagery to be carried out by SERNAPESCA
or outsourced. SERNAPESCA currently conducts the review of EM imagery and associated
information, recognizing that the experience gained will inform any future competitive
outsourcing process undertaken. Review costs could also be on-charged to vessel owners in
future.
After collection of hard drives, EM imagery and associated information is checked for integrity
and completeness. The number of trips and hauls is identified. Selection of the review sample
follows a stratified random process. Review is prioritized where there is considered to be a
higher probability of discarding or bycatch occurring. On each hard drive, a sample of 10% of
the hauls is randomly selected (without replacement). Review is more time consuming on
factory vessels conducting onboard processing and vessels with catch storage tanks/pounds,
21
because the catch is followed throughout processing in multiple areas of the vessel. When
instances of regulatory non-compliance are detected, additional samples are reviewed.
The Electronic Fishing Log (EFL) maintained by the captain of the vessel, and mandatory for
industrial fleets since 2020, provides the basis for comparison with the data extracted from EM.
In EFLs, captains must record all fishing events, and for each set, required records include
estimated catches by species or species group, geographical position, date and time of each
set/haul, quantities and species discarded, and the bycatch of seabirds, marine mammals, sea
turtles and chondrichthyans. At the end of a fishing trip, EFLs are submitted via cellular network
or Wi-Fi, to SERNAPESCA. When inconsistencies between the data extracted from EM and the
EFL are identified, the vessel owner is notified and sanctions may be applied. At this point, the
vessel owners, vessels and captains most likely to use fishing practices associated with
unacceptable discarding have been identified in each fishery across the industrial fleet.
Currently there are three full-time EM analysts conducting review, each working 44 hours per
week. Analysts conduct review in accordance with a documented standard. Analysts review EM
imagery, prepare reports and communicate findings to management. They also validate findings
as required to support sanctioning processes.
Program development
Implementing the EM program during the COVID-19 pandemic was challenging. Nonetheless,
the program has catalyzed change in the industrial fishery. EM information has contributed to
the management agencies’ understanding of behavior patterns and entities and individuals
associated with non-compliance, and supported a significant improvement in undesirable
practices at sea in a way that was not previously possible. Remaining challenges include species
identification in some fishery operating conditions, and catch (and discard) identification and
quantification using EM. In the cases when discards cannot be fully quantified using EM, other
information sources may be considered (e.g. average catches per trip).
Building on the knowledge acquired during the first two years of the program in the industrial
fleet, new approaches to sampling imagery for review are being explored, such as the
development of fleet-specific criteria. The program will continue to cover 100% of vessels and
fishing activity. New review technologies (using machine learning and artificial intelligence) will
be trialed in two pilots starting in 2022 in the artisanal fleet, supported by The Nature
Conservancy and Future of Fish.
Work underway also includes integrating various electronic monitoring and reporting tools to
provide more streamlined and efficient systems (such as EFLs, EM cameras, VMS, catch
certification, and weighing systems). Future steps include transitioning from hard drive storage
to wireless transmission over 5G networks and cloud storage, implementing pre-review within
the EM system onboard vessels, and improving image quality to support a broader range of
monitoring objectives. Exploration of the use of EM for scientific purposes and complementarity
with other observation systems is also underway.
The rapidly changing characteristics of the fishery and its environment are driving the need for
fishery data with higher spatial and temporal resolution, to account for growing uncertainty and
enable adaptive management. The implementation of EFLs and EM has enabled the
modernization and updating of fisheries data systems, and significant expansion of the
collection and analysis of information for both management and research. This has created an
opportunity to coordinate and enhance the work of SUBPESCA, SERNAPESCA and IFOP. The
collection of high-resolution data, as well as faster data processing, analysis and preparation of
more detailed reports enables management responses in times closer to real time.
22
The roll-out of EM across Chilean fisheries is ongoing, and implementation on artisanal vessels
15-18 m in length is regulated from January 2024 (Law No. 21.259, 2020).
4.4. EM to provide catch composition information for fishery management
Alaska Fixed Gear fishery: USA
Information contributors and reference sources: C. Paiva (Pacific States Marine Fisheries
Commission), J. Keaton (Alaska Regional Office, NOAA), J. Ferdinand, J. Calahan, G. Campbell
(Alaska Fisheries Science Center, NOAA), N. Munro, A. Snedeker (Saltwater Inc.), E. Torgerson
(Chordata LLC); 5, 6, 22, 53, 54, 56
Purpose of EM
EM is used to collect catch composition information, including characterizing retained and
discarded catch in the Alaska Fixed Gear (AFG) fishery. The AFG includes hook and line, and
pot/trap fishing gear.
EM-derived data are used in stock assessments, bycatch species risk assessments, to better
understand marine mammal depredation of catch, and for compliance purposes.
Context for EM implementation
EM had been of interest among Alaskan fixed gear fisheries for some years when an observer
program review conducted in 2013 triggered the progression of pre-implementation EM. The
review and associated restructure resulted in increases to the daily cost of observers, and the
addition of smaller vessels (<60 ft) to the North Pacific Observer Program’s partial coverage
category66. (In the partial coverage category, a sample of trips is monitored. By contrast, in a full
coverage program, all trips would be monitored). Fishery managers sought more data and more
randomized sampling from the AFG fleet targeting Pacific halibut (Hippoglossus stenolepis),
which included these smaller vessels. EM was seen by industry and the North Pacific Fishery
Management Council as a potential alternative to human observers to meet the increased
demand for fishery monitoring.
A pilot program was undertaken on volunteer vessels from 2014 – 2017. Review protocols for
the AFG program were initially sourced from the US West Coast Region fixed gear EM program
and updated as required for the AFG fishery. Learnings from the pilot program were used to
develop AFG program standards and specifications. For example, information from the pilot led
to the focus on collecting catch and fate data from EM, rather than hook counts for longline
fishing.
The EM program
The operational EM program has been implemented through regulation in the AFG fishery since
2018. The approach to delivering the program is set out in annual deployment plans (ADPs).
ADPs also describe requirements for monitoring by human observers. The EM selection pool
has comprised around 170 approved vessels from 2020 – 2022.
Vessels operate under Vessel Monitoring Plans (VMPs) that include requirements for
monitoring EM system functions, camera maintenance, and catch handling. Submission and
approval of a VMP are pre-requisites for participating in the AFG EM program on an annual
basis once accepted into the EM pool. Failure to comply with the VMP may result in exclusion
from the EM pool the following year. On average, 3 – 4 cameras are in place on vessels while
there is no regulated number and the focus is on what is necessary to achieve monitoring goals
on each vessel. The EM system is able to trigger camera recordings conditionally through the
use of various sensors, such as those installed on the vessel’s gear to monitor fishing activity, or
detect movement using GNSS/GPS. Sensor information facilitates review.
23
EM is used as a standalone data source for reviewed trips (i.e. audit of logbook information is
not conducted). In 2017, 50% of hauls were reviewed. This has since been reduced to one third
of hauls due to review time and cost considerations. All hauls from a selected trip are reviewed
if only two or one have occurred during that trip. Trips are randomly pre-selected for EM
review through the Observer Declare and Deploy System. This baseline approach to review may
change subject to a prioritization request from the Observer Program or the Office of Law
Enforcement (OLE). Observer Program prioritization requests may result from stock
assessment matters or other areas of concern. OLE requests may arise from unusual VMS
information and/or previous behaviors. Once selected for review, the full review process is
conducted as documented in the review protocol (i.e. there is no subsequent prioritization
among monitoring objectives or tasks).
In 2020, review ratios (that is, the ratio of EM review time to real time) ranged from 0.4 (1 hour
of catch handling took less than half an hour to review) to 2.2 (1 hour of catch handling took
more than 2 hours to review) (Table 2). For cod, the relatively higher review ratios were due to
the diversity of catch species and stern hauling (resulting in poor lighting at night and a side
view rather than a clearer top-down view). Industry buy-in is vital for optimizing catch handling
to support review, and providing prompt feedback when issues are detected has been
appreciated by vessel operators. This also supports prompt improvements in data quality.
Data collected from the AFG EM program have been used for management since 2018. Beyond
data provision for stock assessments, risk assessments, etc., EM also provides OLE with
evidence of illegal and egregious at-sea practices such as illegal discarding and shooting
seabirds. This can be used in support of legal proceedings.
Program development
The EM program is subject to ongoing development and improvement. This includes
considering ways to improve data quality, such as alternative sampling methods that would
alleviate catch handling requirements, the direct incorporation of effort/logbook data with
review, and a baseline level of review to achieve specified data quality objectives (e.g. for
discarded catch and species of interest).
AI tools to facilitate review are in development for this program. In the future, it is anticipated
that these tools will increase review speeds (enabling an increase in the amount of review that
can be completed). For example, if AI is effective in quantifying and identifying retained catch to
species level, reviewers would be able to focus their time on characterizing the discarded catch.
Table 2. Average ratio of review time to catch handling time, for the Alaska Fixed Gear electronic monitoring program in
2020. Review includes the characterization of both retained and discarded catch. Target species are Pacific halibut
(Hippoglossus stenolepis), Pacific cod (Gadus macrocephalus) and sablefish (Anoplopoma fimbria). (Source: Alaska Fisheries
Science Center and Alaska Regional Office 2021).
Gear
Target species
Review time: Real time
Fixed hook longline
Pacific halibut
0.66
Sablefish
0.68
Snap longline
Pacific halibut
0.61
Sablefish
0.42
Pacific cod
2.22
Single pot
Pacific cod
1.01
String pot
Sablefish
0.66
24
4.5. Regional EM initiatives: Pacific Ocean tuna fisheries
EM development in the region
Numerous EM programs have been undertaken in fisheries targeting tunas and billfish in the
Pacific Ocean (Table 389). To date, two have become operational, in Australia’s Eastern Tuna and
Billfish (longline) Fishery and the AGAC purse seine fishery program (described in detail in Case
Study 4.1). The objectives of Pacific-based programs have ranged from compliance to catch
accounting. There has been a focus on the comparability of data derived from EM and data
collected by human observers deployed on fishing vessels, and also logbook reporting in some
cases (Table 3). EM implementation is increasing in the region. For example, Thai Union and
The Nature Conservancy (TNC) have recently partnered to deploy EM and/or human observers
on all vessels in Thai Union’s tuna supply chain by 202585,86. Among government-led initiatives,
EM is scheduled for implementation on New Zealand’s pelagic longline fishing fleet from
November 202327.
Below, we consider two EM programs conducted in Pacific pelagic longline fisheries in more
detail. These are the Hawai’i longline fishery (Pacific Islands Region) and the TNC-Pacific Islands
Cooperative Longline EM Project10.
EM as a monitoring tool for the Hawaii longline fishery (Pacific Islands Region)
Information contributors and reference sources: K. Bigelow, J. Stahl, J. Tucker (Pacific Islands
Fisheries Science Center, NOAA); 11, 30, 48, 62, 79
EM has been explored as a monitoring tool for the Hawaii longline fishery since 2009, when an
initial trial was conducted. The objective of the pre-implementation program that commenced
in 2017 is to evaluate the efficacy of EM as a monitoring tool. Monitoring using EM is focused on
the detection of all catch events, including fish (marketable and nonmarketable species) and
protected species (sea turtles, seabirds, marine mammals, and sharks). For protected species,
monitoring objectives include the collection of mortality and serious injury information.
Currently, sea turtles and marine mammals are the focal taxa for assessing injury and mortality
using EM imagery. Other areas of current research include a catch-handling study to determine
whether detection of shark bycatch can be improved with fishers bringing sharks in closer to
the vessel, and whether catch items brought on deck and released in the water (including fish
and bycatch species) can be detected using machine learning. Routine analysis of EM imagery in
this fishery covers the longline haul, however EM has previously been used to address specific
research objectives, e.g., the efficacy of tori lines during longline setting.
A key motivator for the pre-implementation program was the relatively high cost of human
observers and opportunities to realize cost savings with EM. The work program is designed to
identify how EM and human observer coverage can be used together to best deliver the
required fishery monitoring information. There is currently no target coverage for EM
deployments. EM and human observers monitor 20% of fishing effort in the deep-set pelagic
longline fishery and 100% in the shallow-set fishery. Coverage levels are under evaluation
(bearing in mind human observer costs).
EM systems are carried voluntarily by 20 vessels. Sensors initiate camera recording and
facilitate the detection of catch events by EM analysts. Two cameras are fitted, to provide a rail
view and a deck view. These views record catch items in the water before hauling or release, as
well as catch items that come aboard. A Vessel Monitoring Plan is in place on each vessel, setting
out equipment maintenance protocols. Fishers keep cameras clean. There are catch handling
protocols in place for protected species, which may facilitate their detection at review. Some
25
fishers value the ability that EM provides, of viewing catch coming aboard without them having
to leave the wheelhouse.
The efficacy of various reviews speeds has been analyzed during this program (see section 5.2).
Other learnings include the importance of daily limits on review time to ensure EM analysts
retain their focus, and the need to avoid interpreting an event without watching all imagery of
that event (e.g. what initially appears to be a gear tangle may actually be a protected species
capture event).
Overall, EM has proven very effective at monitoring catch brought aboard vessels. Detecting fish
bycatch released in the water has been the most difficult at review. However, for commonly
caught species (e.g. longnose lancetfish (Alepisaurus ferox) or snake mackerel (Gempylus
serpens)), total catch across the fleet can be extrapolated from information collected by human
observers. Regulatory changes that require fishers to use monofilament leaders and bring
protected oceanic whitetip sharks into the camera view are expected to result in improved EM-
based detections of sharks.
It is expected that the overall coverage levels of 20% and 100% for deep- and shallow-set
fisheries will remain in place in the future, and that EM will continue to be used to supplement
human observer coverage. The integration of EM and human observer monitoring is an active
area of research. To facilitate the use of data collected from EM and by human observers, a data
integration group is being formed to pull data from both sources into one database. A regulatory
framework for the program is in development. Depending on the outcomes of current research,
AI could be used to detect catch events in future, while human review focuses on species
identification and assessment of mortality and serious injury.
EM-based catch characterization across multiple Pacific jurisdictions
Reference source: 10
The western and central Pacific is well suited to a multi-jurisdictional approach to EM
deployment for several reasons. The regional and subregional bodies in place (e.g. WCPFC,
Pacific Islands Forum Fisheries Agency, Parties to the Nauru Agreement) comprise a
management framework that is implemented across national jurisdictions, noting that
individual countries also develop their own legislative requirements. Other monitoring tools
(e.g. VMS) are established at the regional level92. There are many vessels active in tuna fisheries
in the western and central Pacific Ocean (WCPO) that fish across multiple EEZs, and both inside
EEZs and on the high seas.
The Nature Conservancy’s Pacific Islands Cooperative Longline EM Project is an example of a
multi-jurisdictional EM initiative. For this project, EM systems were installed on 15 longline
fishing vessels operating in the EEZs of the Republic of the Marshall Islands, Federated States of
Micronesia, and the Republic of Palau (and adjacent high seas). These included vessels
chartered into the EEZs of Pacific nations while flagged to Japan and Taiwan. While some
jurisdictional differences were detected, project findings demonstrate an overall pressing need
to improve the quality of logbook reporting of target, retained and discarded species from these
longline vessels. For example, catches of significantly more yellowfin and albacore tuna were
documented from EM-derived data compared to logbook reporting. Discards of tunas, billfish
and marine turtles detected by EM and human observers were almost never reported in
logbooks. Logbook reports included fewer species and species groups than EM-derived data
(typically five species in logbook records compared to an average of 8 – 10 for EM). Inaccurate
and under-reporting has clear implications for species and fishery management. EM was
26
identified as a key solution to the low level of monitoring in place in western and central Pacific
Ocean tuna longline fisheries.
Some vessels participating in the project were included in Units of Certification of Marine
Stewardship Council-certified fisheries. This was considered to provide a potential incentive for
better quality logbook reporting. From an EM review perspective, logbook reporting would
need to improve significantly before an audit model could be used to effectively document catch
composition.
27
Table 3. Results of selected electronic monitoring (EM) programs undertaken in Pacific Ocean longline and purse seine fisheries targeting tunas, 2014 - 2022.
Region/nation
Scope
Scale/Stage
Main objectives
Sources
Australia
Eastern Tuna and Billfish
(longline) Fishery
All vessels conducting 30 or
more longline sets per season
Operational
Seabird captures at hauling
Tori line deployment at setting
Composition of fish catch (logbook audit)
1, 17, 19, 20,
21
Hawaii (USA)
Hawaii EEZ and high seas
pelagic longline fishery
18 vessels
(20 vessels in 2022)
Pilot
EM efficacy as a monitoring tool
Catch accounting comparing human observer and EM-
derived data
11, 79
Case Study
4.5
Fiji
EEZ pelagic longline fishery
51 vessels
Pilot
Compliance monitoring, fishery information to support
market access (e.g. Marine Stewardship Council
certifications), and to improve onboard operations (e.g.
safety)
83
Solomon Islands
EEZ pelagic longline fishery
2 vessels
Pilot
Comparing human observer and EM data collection
capabilities
32
Federated States of
Micronesia, Republic of
the Marshall Islands,
Palau
EEZ pelagic longline fisheries
and adjacent high seas
15 vessels
Pilot
Comparison of catch rates between human observer, EM-
derived and logbook data
10
Pacific Ocean
Purse seine fishery
2 vessels
Pilot
Comparing fishery data collected by human observers and
EM
56
Pacific Ocean
Purse seine fishery
28 purse seiners
12 support vessels
Operational
Conformance with Code of Good Practice
Case Study
4.1
28
5. Optimizing EM review
5.1. Approaches to EM review
Fishery management objectives and EM program objectives should underpin the approach to
review used in an EM program. Ensuring appropriate levels of review are conducted to
efficiently meet management and monitoring objectives has been identified as possibly the
“number one near-term cost-reduction opportunity for EM programs”49. Census and sample-based
methods can be used to review EM imagery and associated information.
Census review
Census review involves the review of all imagery and associated information collected by EM
systems. This approach provides the most comprehensive dataset and it is often used in pilot or
trial programs, as well as to meet ETP monitoring objectives in operational programs17,51. In
pilot programs, census review has value beyond the data collected as it also provides a basis for
developing review processes and standards for scaling up to operational EM programs49,51.
Sample-based review
A sample-based approach to EM review enables the scaling of review to fit budget (and other)
constraints. Data derived from sampling can be used as a standalone information source. An
alternative approach is to use sampled data derived from EM to audit fisher logbook reporting.
Audit-based review with logbook data
Taking an audit approach, data collected from reviewing a sample of EM imagery are compared
to fisher reports and the deviations between the two datasets are scrutinized. If audited fisher-
reported data meet pre-defined accuracy thresholds, logbook data are accepted as the source of
fishery data at the fleet scale, and additional EM review is not pursued. Sampled data therefore
are not scaled up, and logbook reporting becomes the fleet-level record. Ideally samples used
for an audit approach would be randomly selected.
Where differences between EM and logbook datasets are significant at audit, further
investigation is required (e.g., additional EM review and evaluation of logbook data to identify
issues for improvement). Recovering additional review costs directly from individuals filing low
quality records is one approach to encouraging improvements in logbook data quality17,82.
Where logbook data are of low quality across a fleet, the audit approach will not work well and
use of EM data as a standalone data source (sample or census) is appropriate until logbook data
quality improves.
Sample data as a standalone source
EM review rates required to support different fishery monitoring objectives have not been
widely explored empirically, though the value of identifying these rates is well recognized84. In
one example, simulation modelling undertaken for the US Northeast Multispecies (groundfish)
Fishery found that bias in logbook reporting of 12 species of discards could be corrected with
EM review rates below 50%. (In this fishery, strong covariance was evident between logbook
reporting and EM-derived discard data). In an example year, 35% was the lowest EM review
rate that achieved a coefficient of variation of 30% for all species considered45. In a second case
in British Columbia, Canada, simulation modelling was not undertaken but 10% EM review
proved effective in meeting the requirements of the fishery, when used to audit logbook
reporting and in conjunction with other monitoring tools (such as dockside monitoring)81.
Clear objectives are essential to determine appropriate EM review rates and a higher level of
accuracy necessitates higher review rates. For example, when monitoring catch of a quota
limited species for compliance purposes is the objective, higher review rates will be needed
29
than when EM-based catch characterization is used for stock assessment purposes. The
characteristics of the event of interest will also affect EM review rates required to provide
information for management. For example, estimating catch levels of a commonly caught
species will require lower review levels than for a rarely caught species.
Such concepts are also evident in literature on human observer monitoring rates7,8,18,36,42.
However, a critical difference between human observer monitoring and EM is the ability to
sample and resample EM imagery and associated information after it is collected. Provided EM
captures all fishing activity, the review rate and sample selection process can both be set in
advance and adjusted retrospectively, in accordance with management priorities, risk,
resourcing, and any other relevant factors. This allows a much more agile and adaptive
approach to monitoring than is achievable with on-vessel human observer deployments.
Within a sampled unit of fishing effort, subsampling may be sufficient to meet data needs (while
also reducing time required for review). For example, subsampling catch from a portion of the
hooks from each longline haul may enable more hauls to be reviewed than if entire longline
hauls were sampled. Depending on the nature of the fishery, the subsampling approach may
result in more representative catch characterization (e.g. if a fishery operates across a large
geographic area with considerable variation in catch species distributions).
Combining census, sample and subsampling approaches in one EM program to meet a set of
monitoring objectives may also be appropriate (e.g. a census approach reviewing 100% of hauls
to identify ETP bycatch, 20% of hauls sampled to record fish bycatch, and hook counts
conducted on 5 baskets from each haul sampled for fish bycatch). Typically, the addition of
monitoring objectives adds complexity and therefore time, to review processes. Where
resources are finite, prioritizing monitoring objectives is another effective approach to
managing review time and cost.
Regardless of the review approach used, 100% capture of fishing operations (i.e. all vessels,
with all fishing activity recorded) is recognized as best practice, enabling avoidance of the
“observer effect” (when fishing operators change practice because they are being monitored,
resulting in data collected from observed trips not being representative of normal fishing
operations)17,49.
5.2. Increasing EM review efficiency
Practical steps to maximize the efficiency of EM review can be taken at the design, on-vessel
data capture and review stages (Table 4).
Suitability of data fields for EM
In many cases, EM programs are being developed and implemented in fisheries in which human
observers have operated. It is essential to consider, especially when transitioning from an at sea
observer program to an EM program, that not all data fields and definitions can be identically
transferred. Each data collection method differs, and EM analysts do not handle the organisms
they see. For example, Alaska’s at-sea observer and EM programs collect data on Pacific halibut
viability, injury, and gear release methods. This information is provided to the International
Pacific Halibut Commission and informs halibut mortality rates. The current condition codes are
defined based on at-sea observer fish-in-hand assessment. EM condition definitions have not
been adapted and remain the same as the observer definitions. This is problematic for EM
analysts, as they are often unable to view (and therefore assess condition of) both sides of the
halibut. Adjustments to data definitions that can be met using EM have been recommended13.
30
Considering alternative methods for collecting data from EM can also improve efficiency of
review. For example, hook counts are a typical measure of fishing effort. Depending on factors
such as camera views, image quality, and gear configuration, hook counts can be challenging
and time consuming for analysts. Alternative approaches explored to collect this data element
include census counts, computer vision (the use of artificial intelligence that enables computers
to derive meaningful information from visual inputs), and subsampling. Census counts were
found to be costly, computer vision was not complimentary to all programs, but subsampling
was considered to warrant further exploration.
Hook subsampling methods have included using time as a proxy, using gear segments/sections,
and a combination of time and gear segments. Determining which hook subsampling approach
would best complement a program depends on the type of gear used, the identifiable presence
of gear markers, and/or the reliability of gear documentation (logbooks) if available. Hook
count subsampling was found to be most representative for fixed longline gear with easy to
identify segment markers that were positioned at semi-regular intervals and an expected
number of hooks per segment. The amount of time to collect this information depends on the
proportion of gear subsampled across the trip and the total amount of gear. Preliminary results
showed a small percentage of added time (less than 10%) to complete subsampling, which is
less cost prohibitive than conventional hook enumeration approaches13. Having the ability to
efficiently collect hook count information could support options for sampling including audit
approaches.
Operational changes to facilitate data capture
Fishers can influence review costs by operating in ways that facilitate effective image capture.
When developing and operating in EM programs, a key question is how fishers can alter their
operations to support successful collection of data for management. For example, if the goal of
the program is catch accounting, the catch needs to be handled in a manner that allows
reviewers to identify and count catch to meet the minimum data need.
When establishing catch handling requirements for EM programs, it may not be necessary for
fishers to significantly alter their regular operations or not to the extent that some programs
currently require. Important considerations for the development of handling requirements
include gear configuration, hauling operations, and catch composition and volume. In parallel,
an awareness is required of the potential for handling requirements to lead to compliance
issues, slowed fish production, negatively impacted data, and/or increased review time due to
fishers’ difficulties in meeting these expectations. A collaborative approach that involves the EM
review service provider, fishers and the agency identifying data needs is recommended to
optimize the specification of any handling requirements. Prompt feedback to vessel crew is also
important for addressing on-vessel issues affecting image capture as quickly as possible. Where
review costs are on-charged to vessel operators, there is the opportunity to incentivize
facilitative operational changes such as catch handling practices through the commensurate
reduction in review time (and therefore cost).
Within the Alaska Fixed Gear (AFG) program, EM service provider Saltwater Inc.’s review team
actively researched current data needs and uses, reached out to fishers to gain a better
understanding of their operations, and assessed program protocols to look at other review
approaches and ways to reduce impacts on fishing operations. The AFG program currently
requires vessels using single pots to clear all catch from each pot prior to hauling and
processing the following pot. This is to ensure reviewers are able to identify and count all catch
items associated with that pot, the defined sampling unit. However, when catch volume and
species diversity are higher, the ability to sort, process, and clear the table prior to the next pot
arriving can become challenging if not impossible. This can lead to catch being mixed from
multiple pots and/or discarding species by the armfuls preventing reviewers from obtaining
catch composition information, and making the pots unsampleable.
31
The review team tested an altered sampling unit, defined as a string or cluster of pots, with an
allowance for clearing catch by the end of the string or cluster. Promisingly, results found that
increased flexibility in catch handling requirements paired with alternative sampling units led
to improved catch monitoring throughout fishing events. Furthermore, the amount of data per
dollar increased due to a decrease in unsampleable data caused by catch handling issues13.
Varying playback speed
EM imagery can be replayed at normal speeds (i.e., speeds equivalent to real time), sped up, or
slowed down. The appropriate playback speed will depend on monitoring objectives and data to
be extracted, as well as human constraints such as limits of analyst concentration and fatigue.
For example, for large and highly visible cetaceans, imagery review at 10 – 12x normal speed
has been effective41. In the Hawaii longline fishery, reviewer accuracy in detecting catch events
was tested at three playback speeds faster than real time (4x, 8x and 16x normal speed). EM
reviewers detected retained catch with similar accuracy at all three playback speeds. For
discarded catch, on average, detection accuracy was highest at a playback speed of 8x. At 4x
normal speed, reviewers did not detect some protected species, possibly due to waning focus as
the haul review progressed. At 16x normal speed, reviewers detected all protected species
caught except one albatross. The potential to miss protected species events at such rapid
playback speeds was noted, e.g., drop-offs or cut-offs would take place in an instant on-screen.
Above 16x normal speed, the EM video skipped and catch events may not have appeared on
screen at all79.
Ergonomic tools
Ergonomic efficiency can also save analysts’ time at review. EM analysts work by transitioning
back and forth between their keyboard and mouse to conduct review. While each movement is
short, cumulatively these transitions can account for a significant amount of time. Hotkeys
(project customizable key-bindings) assist reviewers in minimizing transitional movements,
navigating efficiently across the keyboard during review, and reducing the steps involved in
creating annotations at review. Hotkeys can be programmed to easily allow reviewers to
interact with playback speed, advance or reverse video, and create fishing and species
annotations within the data. This leads to an overall decrease in review time, and a potential
increase in data quality as the hotkeys reduce the number of fields and/or forms a reviewer
needs to complete. To maximize the benefit of hotkeys, collaborating with review teams is
recommended to establish hotkey utility and which key bindings would be best to implement.
The reviewers who directly interact with the data on a regular basis are best placed to identify
limitations and workable improvements for staff who are setting up these tools.
Computer vision and artificial intelligence
Computer vision tools can perform or augment the process of marking fishing events,
establishing sampling frames, monitoring for compliance, detecting catch and identifying
catch17,87. Compared to manual review and marking of events, the application of computer
vision/machine learning reduces review times and cost, and increases data confidence (as long
as algorithms are trained appropriately). One use case example involves detecting humans
interacting with gear or present on deck to direct EM reviewers to areas of interest - with an
accuracy of 95%. This dramatically reduced the amount of non-useful video that needed to be
scanned by the review team (Saltwater Inc. unpubl.).
Another computer vision tool utilized the automatic detection and event marking of fishing
gear, which can be a time-intensive component of analysis. This tool can achieve close to 100%
accuracy, reducing the amount of analyst time needed for this task considerably (e.g. reductions
of hours in the time needed for marking a high effort pot fishing trip). The detector finds and
pre-identifies selected pots for sampling in a fishery where gear deployment may be upwards of
a thousand pots per trip. Sampling rates can be set for gear detection tools to meet project
32
requirements. Other computer vision applications include discard compliance monitoring,
species identification, and low-level analysis of EM system performance13.
Many factors affect which computer vision tools will be effective in a monitoring program
including camera views, image quality, catch handling, general fishing operations, and
management objectives. Having a strong understanding of both the management objectives and
the EM data will help determine which computer vision tools could be beneficial. It is also
important to consider the start-up costs associated with computer vision tool development and
implementation. Additional time and resources are needed to create, train, assess functionality
and success, and incorporate the tool into the overall review workflow. Labelling and saving
data and metadata during EM review may provide longer term value that is unquantifiable in
the short term, by facilitating development of review processes that incorporate machine
learning when computer vision tools are developed.
5.3. Costs of EM review
EM review costs range from 2.5% to more than 60% of the total costs of EM programs (Table 5;
noting that the approach to defining the review component of EM programs affects the
comparability of cost information between programs). Review cost profiles for pilot and
operational programs are expected to differ significantly, because review methods and
standards will be developing in the former case while in place at scale for the latter. Also, for
operational programs, equipment may already be aboard vessels if a pilot program has been
conducted. This will affect the perception of review costs as a proportion of operational
program costs. Where review occurs may also affect costs, assuming EM analyst remuneration
reflects local labor costs.
How costs scale with review rates is not linear54,81. Regardless of the review rate, initial
screening of the EM imagery and associated information and determination of the sampling
frame (e.g. number of trips, sets/hauls) are required. This comprises a baseline minimum cost.
From there, irrespective of review rate, costs increase with some relationship to the complexity
of review tasks. The process of EM review enables significantly more granular analyses and
management of cost-per-datum compared to human observers. (For human deployments, the
length of a trip does not change with respect to the number of tasks undertaken during that trip,
and moving between vessels at sea has significant logistical implications). Generalized figures
from one EM review service provider illustrate the scale of variation in review costs across
different fisheries and monitoring objectives (Table 6). Service providers emphasize that
determining the best approach to review involves collaboration among providers, clients and
vessel operators, to ensure monitoring objectives are met with maximum cost efficiency (DOS
and Saltwater Inc., pers. comm.). Growth in the scope of an EM program can make costs more
difficult to predict, and also makes EM cost less comparable to pre-existing monitoring
programs with the same initial objectives as the evolving EM program.
33
Table 4. How EM review efficiency can be increased for the census and sample review methods.
When applicable
Approach
Review method
Census
Sample
Program design phase
Focused monitoring objectives
Information collection priorities set
Sample selection specified (random, stratified, risk-based)
Subsampling units identified
EM-appropriate data definitions developed
Use of the audit approach
On-vessel data capture
Catch handling protocols in place
Lens cleaning undertaken
High quality logbook reporting
Only for audit method
Incentives for operational practices that facilitate review
Feedback provided to crew rapidly to enable prompt on-vessel changes
At review
Review instructions that accurately reflect program design, objectives, data needs
EM-appropriate data collection units identified
Review speeds faster than real time
Hotkeys used by analysts
Review supported with Computer Vision, Artificial Intelligence
34
Table 5. Summary of published information on the cost of electronic monitoring review. (*excludes project management and some staff-related and training costs; +Conditions of the Marine
Stewardship Council certification of the fishery related to management of target stocks and some non-target species2. $Costs are estimates, predicted prior to the implementation of an
operational program. +EM review plus a 10% audit of review for quality assurance. @Calculation of this estimate excludes investments already made in the pilot trial for 50 vessels, therefore,
the percentage of total costs allocated to review will be less than 7.8%. ^Data review, processing analysis costs (2020). Avg. = average, EFP = Exempted Fishing Permit).
Monitoring objective /
data collected
Fishery
Target
Location
Program type
Review approach
Review cost
as % of
program cost
Source
Verify reported catch of Atlantic
bluefin tuna
Pelagic
longline
Highly
migratory
species
East Coast USA,
Caribbean and
international
waters
Operational
(67 active
vessels; 110
carrying EM)
10% baseline
16%
60
Monitor compliance, document
fishing practices, monitor setting
and hauling including safety
conditions, address MSC
conditions+
Pelagic
longline
Tuna
Fiji
Pilot
(50 vessels)
Objective of census
review.
Review of 44% of trips
achieved.
4.5%
83
Operational monitoring of
longline fishing
Pelagic
longline
Tuna
Fiji
Operational
(50 vessels)
Sample (~20% of
fishing days)
7.4%$
33
Operational monitoring of
longline fishing
Pelagic
longline
Tuna
Fiji
Operational
(90 vessels)
Risk-based (Low risk:
sample of 5%; High
risk: census of 100%
review; Avg. 22%
review overall)
<7.8%$@
33
Operational monitoring of
longline fishing
Pelagic
longline
Tuna
USA (Hawaii)
Operational
(160 vessels)
Sample (25% of sets)
31%$
+3%+
62
Operational monitoring of fixed
gear fishery:
• catch composition
(retained and discarded,
by species)
• discarded halibut
condition
• presence of streamer
lines and discarded
halibut release method
with longline gear
Hook and
line,
pots/traps
Sablefish
(Anoplopoma
fimbria)
Pacific cod
(Gadus
macrocephalus)
Pacific halibut
(Hippoglossus
stenolepis)
USA
(Alaska)
Operational
(169 vessels
approved for
EM)
Sample (From vessels
with EM, 100% of
string pot gear hauls;
33% of hauls with
other gear)
17.5%^
5
35
Collect data on fishing
operations
• Catch by species
• Discards
• Fishing effort (including FAD
activity)
Monitor compliance with
national and regional
management measures
Purse seine
Ghana
Pilot
Census
2.5%*
83
Discard accounting
Trawl
Groundfish
USA
(Pacific Coast)
EFP (84
vessels)
transitioning to
regulatory
program
Census
Review,
reporting,
storage
39% (Avg.
2015 – 21)
18% (2022)
63
Cetacean bycatch
Inshore
gillnet
Europe
Operational
Census
36%$
17
Complete sensor record of trip
Verify logbooks and audit catch
records (retained and discarded
catch)
Confirm fishing locations
Hook and
line
British Columbia,
Canada
Operational
(~200 active
vessels)
Audit (10%)
34%
81
36
Table 6. Relative costs per sea day for analysis of electronic monitoring imagery and associated information, for the purse seine, longline and trawl fishing methods. (Source: G. Legorburu,
DOS, pers. comm.). Figures are not necessarily applicable to any specific fishery and are provided as indicative. Fishing effort screening analysis includes trip and set start/end date, time and
locations; number of sets, detection and description of encounters with other vessels/vehicles. Detection and description of Fishing Aggregation Devices is also included for purse seine. For
longline, bait, hook and baskets are also characterized, and seabird mitigation measures checked. Species of special interest include sharks, rays and turtles, and this analysis includes
detection, handling, condition and fate analyses. Standard catch characterization for purse seine includes data collection for species of special interest, estimating total catch per set, per brail
and per well, species composition estimation, detection and estimation of tuna discards, detection and identification of fish bycatch and associated condition, fate and release information. In
addition to these data, full characterization for the purse seine method includes digital counting and sizing of catch per set and total catch. Longline catch characterization includes catch
composition by species, condition and fate, size and sex determination, hooks per basket, interactions with species of special interest; discard characterization includes date, time and location
of discard events, discard species composition and size, estimated reason for discarding. Trawl catch and discard characterization includes estimated total catch per haul discard estimation,
estimation of species composition, counting and sizing of samples of retained and discarded catch and estimates of their condition and fate.
Fishing effort screening
analysis
Catch analysis
Fishing methods
Trip and set description
Species of special interest
only
Standard catch
characterization
Full catch and discard
characterization
Purse seine
3x
4x
10x
15x
Longline
x
19x – 24x
Trawl
3x
20x
37
6. EMoptim: a prototype tool to evaluate EM review rates
6.1. Exploring EM review rates
EMoptim, a simulation tool
To investigate minimum EM review rates, we developed a prototype simulation tool based in R,
EMoptim, that uses stratified random sampling to address one or more monitoring objectives.
We used this tool to evaluate EM review rates when EM is implemented as a standalone
monitoring method (noting that other data collection tools that may complement EM will often
be in use and these should be considered when developing fishery-specific monitoring
programs, e.g. logbook data collection). A worked example showing implementation of EMoptim
using publicly available fishery data from WCPFC is provided at Appendix 2.
Our approach involves setting monitoring objectives to be met by EM (single or multiple
objectives can be set in EMoptim), and identifying accuracy/confidence requirements (e.g.
coefficient of variation, which can differ between objectives), cost limits, or other constraints.
We assume that 100% of fishing activity is captured on all vessels in the focal fishery. Existing
fishery knowledge is used to identify strata within which sampling effort is allocated for review.
Strata may be defined using statistical reporting areas, gear type, fisheries sector, time periods,
risks, species characteristics (e.g. distributions of age/size cohorts) or any other factor.
Information sources such as risk assessments can be used to estimate the distribution of taxa of
interest and interaction rates (e.g. if fishery-dependent information is inadequate).
Simulation modelling is conducted to identify the required review rate within the limits set. In
general, review to meet compliance monitoring objectives would require much greater certainty
(smaller coefficient of variation) than the collection of target catch information for stock
management purposes, for example, and such differences are accommodated when limits are
set.
Evolution of any EM program is expected based on lessons learned within the program and new
knowledge from external sources. That can be accommodated in EMoptim through changing to
monitoring objectives (including confidence limits), iterative updates to strata and repeating
review rate calculations. Outside of strata with higher review rates identified using EMoptim, we
recommend that a minimum baseline of 5% random review is maintained to enable detection of
significant changes in the fishery and previously unknown fishery issues (e.g. changes in fishing
location, bycatch hotspots, etc.) (Figure 4).
Bias is addressed by several facets of this approach. First, all fishing activity is monitored (there
is no observer effect) and is therefore available to be sampled. Second, a minimum of 5% of EM
imagery is randomly sampled for review in the fishery of interest, over and above any more
intensive sampling within strata (noting that small sample size considerations are relevant
here7). More broadly, bias introduced at the review stage by analysts can be addressed through
quality assurance processes implemented at review (e.g. a second independent reviewer
auditing 10% of EM imagery, after which the data extracted from the two reviews is compared
to assess accuracy)71.
Using EMoptim, we evaluated EM review rates appropriate to monitor target and non-target
catch to achieve specified coefficients of variation (Table 7). We emphasize the following
caveats on these review rates:
i. Review rates are estimated using aggregated data (WCPFC data at 5o x 5o resolution).
Set-level data were not available for use, and therefore we sourced estimates of the
38
number zero-catch sets for each fishing method and taxa from the literature (see
Appendix 2).
ii. At the aggregate level, set by set variation is no longer apparent. Therefore, for fishery-
specific determinations of review rates, the use of set-level data is strongly
recommended, as these data provide significantly more information about the statistical
characteristics of events of interest.
iii. In the absence of set-level data, we have based assumptions about the statistical
characteristics of events of interest assumed on published literature. These assumptions
strongly influence the estimation of review rates. For example, rare bycatch events are
characterized by zero-inflation and overdispersion and may be modelled using different
distributions depending on dataset characteristics8,9.
iv. We present EMoptim outputs generated from 1,000 simulations. The nature of the
approach using genetic algorithms means that consecutive iterations at the same
number of runs are likely to have slightly different outputs (i.e. review rates), but these
will be in proximity. Simulations should be increased in number until emergent outputs
show an acceptable level of stability in review rates. (We also used 10,000 runs of
EMoptim to generate review rates in Table 7 and explore optimized review rates in
Table 8. Many values were the same in the outputs from both sets of runs. With 10,000
runs, only seven values differed by 5% or more, and all except one of these differences
was for very rarely caught ETP species characterized by highly overdispersed and zero-
inflated capture rate distributions).
EM review rates to monitor catch
Simulations using EMoptim demonstrate that stratification can increase the efficiency with
which monitoring objectives are met for common catch species (e.g. the target species of
yellowfin tuna in this example). For example, to estimate (with CV = 0.1) the number of
yellowfin tuna caught in WCPFC longline fisheries, 26% review is required without
stratification. When stratified sampling at the 25o x 30o level is introduced, the required EM
review rate decreases to 4.4%. If a CV of 0.3 is required, the review rates become 7.8% and ~1%
without and with stratification, respectively (Table 7).
Whether or not sampling is stratified, the amount of review required increases as catch
frequency decreases. For example, where porbeagle sharks (Lamna nasus) are captured on 20%
of longline sets, 90% EM review would be required to estimate catch numbers with CV = 0.1 in
the absence of stratification. With stratification at the 25o x 30o level, the required review rate
decreases to 27% (Table 7).
Endangered, threatened and protected species bycatch events are generally characterized as
rare with zero-inflated distributions. As a result, significantly higher levels of EM review are
required to estimate numbers of these events effectively. Without stratification, estimating
seabird, turtle, and marine mammal bycatch events with a CV of 0.1 would require very high
levels of review (effectively a census review in most cases). Considering ETP species groups,
stratifying sampling, and requiring a CV of 0.3 can reduce required review rates (Table 7).
However, stratification should be expected to have little effect on review rates when rare events
are widespread geographically. In such cases, very high review rates will still be required.
Ensuring a baseline level of monitoring and considering fishery-independent data across the
fishery of interest are critical in this regard to ensure that areas in which bycatch occurs are not
overlooked (Figure 4).
39
Table 7. EM review rates calculated using EMoptim for a range of tuna fishery catch elements. Publicly available fishery data from the Western and Central Pacific Fisheries Commission were
used in EMoptim to derive review rates. p0 = the proportion of zero-catch sets, derived from published sources (for sources and a description of EMoptim, see Appendix 2). ETP = Endangered,
threatened and protected species.
Catch
element
Example
species/group
Statistical characteristics of
capture events
Target CV
Longline fishery review %
Purse seine fishery review %
No
stratification
25ox30o
stratification
No
stratification
25ox30o
stratification
Target
species
Yellowfin tuna
Thunnus albacares
Lognormal
p0 = 0
0.3
7.8
~1.0
3.8
~1.0
0.1
25.8
4.4
10.8
2.1
Other
retained
species
Porbeagle
Lamna nasus
Zif Poisson
p0 = 0.40 – 0.80
0.3
9.4 - 11.7
3.2 - 4.2
0.1
37.9 - 90.1
10.8 - 26.9
ETP species
Oceanic whitetip shark
Carcharhinus longimanus
Zif Poisson
p0 = 0.75 – 0.90
0.3
11.1 – 47.4
3.8 – 18.3
Zif Poisson
p0 = 0.75 – 0.90
0.1
12.3 – 73.0
4.8 – 44.6
Zif Poisson
p0 = 0.99
0.3
0.1
~99.0
~99.0
Silky shark
C. falciformis
Zif Poisson
p0 = 0.99
0.3
34.2
18.7
0.1
95.1
32.4
Black-footed albatross
Phoebastria nigripes
Zif Poisson
p0 = 0.99
0.3
~99.0
91.2
0.1
~99.0
95.1
Whale shark
Rhincodon typus
Zif Poisson
p0 = 0.99
0.3
~99.0
95.1
0.1
~99.0
~99.0
ETP species
groups
Seabirds
Zif Poisson
p0 = 0.95
0.3
~99.0
18.4
0.1
~99.0
~99.0
40
Turtles
Zif Poisson
p0 = 0.90 – 0.95
0.3
76.4 - ~99.0
9.3 - 95.1
95.1 - ~99.0
8.4 - 87.2
0.1
95.1 - ~99.0
84.1 - ~99.0
~99.0
80.4 - 91.2
Marine mammals
Zif Poisson
p0 = 0.99
0.3
92.1
87.2
87.2
51.3
0.1
~99.0
91.2
~99.0
~99.0
Table 8. Examples of optimized EM review rates estimated by the EMoptim simulation tool, as required to monitor the number of yellowfin tuna (Thunnus albacares) and two shark species
(porbeagle, Lamna nasus, and oceanic whitetip shark, Carcharhinus longimanus) caught in longline and purse seine fisheries. Optimization was conducted using publicly available catch
information from the Western and Central Pacific Fisheries Commission. CV = Coefficient of variation. p0 = the proportion of zero-catch sets, derived from published sources (for sources and a
description of EMoptim, see Appendix 2).
No stratification
Optimized stratification
No stratification
Optimized stratification
Species
Target CV
% review
% review
Achieved CV
% review
% review
Longline
Yellowfin p0 = 0
Porbeagle p0 = 0.4
0.1
0.3
25.8
9.5
~1.0
~2.0
0.05
0.22
25.8
~2.0
Purse seine
Yellowfin p0 = 0
Oceanic whitetip shark p0 = 0.99
0.1
0.3
9.7
~99
~1.1
~99
0.09
1.07
~99.0
~99.0
41
Table 9. EM review rates calculated using EMoptim with 10,000 runs, showing differences of >5% (italicized) among in outputs among example species/groups compared to 1,000 runs {as
shown in Table 7}.
Catch
element
Example
species/group
Statistical
characteristics of
capture events
No. runs
conducted in
EMoptim
Target CV
Longline fishery review %
Purse seine fishery review %
No
stratification
25ox30o
stratification
No
stratification
25ox30o
stratification
Other
retained
species
Porbeagle
Lamna nasus
Zif Poisson
p0 = 0.40 – 0.80
1,000
0.1
37.9 - 90.1
10.8 - 26.9
10,000
37.1 – 84.7
10.5 – 26.5
ETP species
groups
Turtles
Zif Poisson
p0 = 0.90 – 0.95
1,000
0.3
76.4 - ~99.0
9.3 - 95.1
95.1 - ~99.0
8.4 - 87.2
10,000
73.9 - ~99.0
8.7 – 82.1
95.1 - ~99.0
9.0 – 87.2
1,000
0.1
95.1 - ~99.0
84.1 - ~99.0
~99.0
80.4 - 91.2
10,000
~99.0
82.1 - ~99.0
~99.0
83.8 - ~99.0
Marine
mammals
Zif Poisson
p0 = 0.99
1,000
0.3
92.1
87.2
87.2
51.3
10,000
~99.0
~99.0
~99.0
30.4
42
Figure 4. Approach to setting review rates for fishery information collected through electronic monitoring. See Appendix 2 for a description of EMoptim, a prototype simulation tool based in R
that was developed to identify review rates required to meet monitoring objectives within identified limits.
100% of fishing activity captured by EM
No prior knowledge
Low level of prior
knowledge (e.g. 1%
observer coverage)
Review 5% of EM
imagery, selected
randomly
Identify strata to
structure
targeted EM
review
Define limits
(e.g. CV,
review budget)
Identify review rate
that meets limit
requirements
Conduct EM
review at
calculated rate
Consider newly available
information
(e.g. findings of EM
review, risk assessments)
Review baseline 5% of EM
imagery randomly selected
from areas not already
sampled
Higher level of fishery
knowledge (e.g. based
on 10% observer
coverage, good quality
logbook data)
Define
monitoring
objectives EM
will meet
Run
EMoptim
43
6.2. Optimizing EM review rates
EM programs often have multiple objectives. Broadly, two approaches can be taken to
optimizing EM review rates among management and monitoring objectives. These are (i) to
identify the highest priority objective and design review to address that objective. Data relevant
to other objectives are then collected to the extent possible within the regime designed for the
priority objective. Associated uncertainties can be determined analytically and considered
alongside review findings. Alternatively, approach (ii) is when review is designed to meet
multiple monitoring objectives, such that more review may be required than the minimum level
needed for any single objective. This could result from differences in the spatial distribution and
statistical characteristics of events of interest (e.g. common, ubiquitous events compared to
rare, clustered events).
Using EMoptim, we explored optimized rates of EM review needed to estimate target and
bycatch catch, to achieve specified coefficients of variation. Review rates necessary to optimize
any combination of objectives could be examined using this simulation tool. Outputs highlighted
that practically, optimizing sampling regimes for different monitoring objectives is most
effective among more commonly caught species. As soon as rarely caught species are introduced
to the optimization process, the required EM review rate increases dramatically (Table 8). For
optimization scenarios, increasing the number of simulations conducted by EMoptim from 1,000
to 10,000 had little impact on required review rates in most cases (Table 9).
The cost of review is usually a critical factor affecting EM review rate selection, and the key
trade-off becomes data needs versus available resources (budgetary and/or human, in the form
of analyst hours). To enable exploration of costs against monitoring objectives, EMoptim
incorporates a cost function, developed based on pricing estimates for analysis of EM imagery
and associated information (Table 6). Costs are estimated as a baseline for characterizing the
sampling frame, with an additional amount commensurate with the monitoring objectives. Cost
estimates would be refined over time for monitored fisheries, e.g., as fishery-specific
information becomes available from EM programs, analyst efficiency increases with experience,
etc., and this information could then be incorporated in EMoptim to refine review strategies.
7. Best-value approaches to EM review: present and future
Robust EM program design is critical for supporting efficient and cost-effective EM review, and
fishery-specific information is an essential input to program design. Indicative baseline rates
can be identified for appropriate EM review (Table 10), while fishery-specific review rates
should be determined using set by set data. Stratifying samples of EM imagery reviewed can
significantly increase review cost efficiency, notably for taxa that are commonly caught and/or
have restricted distributions. Nonetheless, a baseline cost is required to identify the sampling
frame for EM review, beyond which costs increase with review complexity.
Opportunities for increasing cost efficiency of EM review, both now and into the future, vary
among monitoring objectives and data types. Such opportunities include improving the quality
of logbook data to support audit-based EM review and investigating subsampling approaches
for data elements not expected to change dramatically within a fishing trip. Contributing to the
continued development and implementation of automated review tools supported by computer
vision and artificial intelligence is also strongly encouraged, e.g. through contributing imagery
to open access datasets used for computer vision work (Table 10).
EM has great potential to collect data at scale to meet the needs of commercial fisheries
management. Information requirements that can be met by EM are shared across RFMOs and
among other fishery management bodies. EM service providers operate across geographic and
44
jurisdictional boundaries. Existing knowledge provides a strong foundation for progress
without reinvention. Evidence requirements of seafood sustainability schemes47 and consumer-
driven demands for supply chain transparency90 provide additional stimuli for knowledge-
sharing, optimizing EM implementation including at review, and improving EM cost efficiencies.
As a result, there is a nascent and significant opportunity to accelerate EM adoption and accrual
of its benefits, to realize best value from this monitoring method in the immediate future and for
the longer term.
45
Table 10. EM review rate guidelines for the collection of selected EM-derived data elements to support longline and purse seine fishery management. Figures are indicative not definitive, do
not consider stratified review sampling options, and will vary with respect to management and monitoring objectives and fishery characteristics. Partial documentation by EM implies detection
in the absence of a dedicated set-up that would not routinely be part of an EM system (e.g. to monitor fish waste discharge, camera coverage of all points of waste exit from the vessel would
be required). Options to optimize the use of EM could be supported and implemented by a range of actors, including EM practitioners, fishery scientists and management bodies. (*EM is
expected to more accurate than logbook information, and in use in the absence of human observers).
Data types
Data elements
Review rate
(no stratification)
Considerations
Options to optimize the use of EM
Operational
Trip and set start and end
date, time, location
100%
Baseline review element required to
determine sampling frame.
Mainstream automated detection of fishing
events in EM review software.
Marine pollution events
Opportunistic
Partial documentation by EM.
Collect data from EM review as objectives
require and resources allow.
Abandoned/lost gear
Opportunistic
Partial documentation by EM.
Fishing gear,
effort
Floats (longline)
100%
Indicates number of baskets; proxy
for number of hooks per set.
Normalize float counts as a proxy for
extensive hook counts at EM review (see
below).
Hooks set/hauled
(longline)
Subsample
Float counts and hooks per basket
provide a proxy for total hooks;
subsampling approach assumes
consistent number of hooks/basket.
Investigate variation in hook numbers per
basket in focal fisheries, to determine
optimal subsampling rates in lieu of
extensive hook counts.
FAD use, type (purse
seine)
100%
Relevant to fishery characterization
and compliance.
Search time (purse seine)
100%
Relevant to fishery characterization
and compliance.
Catch
characterization
Target catch species
(assuming catch on all
sets)
~5 – 10%
(CV = 0.3)
Most likely catch component to be
recorded accurately in logbooks.
Verification of retained catch also
possible at landing.
Use set by set data from at least 5%
monitoring coverage (Figure 4) to stratify
sampling for review, estimate fishery-
specific EM review rates to meet
monitoring objectives.
EM imagery also provides a source of
length information (sampling process
determined by monitoring objectives;
sampling can also be stratified using
EMoptim).
Improve quality of logbook data to support
an audit approach to review.
Commonly caught
retained species (e.g.
caught on 20 - 40% of
sets)
~10 – 15%
(CV = 0.3)
May be relatively accurately
recorded in logbooks. Verification
also possible at landing.
46
Contribute imagery to open access datasets
that support the development of computer
vision tools for catch identification40.
Bycaught/discarded
species (e.g. caught on 10 -
25% of sets)
~10 – 50%
(CV = 0.3)
EM likely to be the most accurate
data source*.
Use set by set data from at least 5%
monitoring coverage (Figure 4) to stratify
sampling for review; estimate fishery-
specific EM review rates to meet
monitoring objectives.
Contribute imagery to open access datasets
that support the development of computer
vision tools for catch identification40.
Rarely bycaught species
groups (caught in low
numbers on 1 - 10% of
sets)
~75 – 100%
(CV = 0.3)
Explore review speeds faster than normal
time, to identify optimal speeds for
detection accuracy and efficiency.
Contribute imagery to open access datasets
that support the development of computer
vision tools for catch identification40.
Rarely bycaught species
(caught in low numbers on
~1% of sets)
~100%
(CV = 0.3)
Bycatch
mitigation
Tori line deployment
Recommended
minimum: all
sets/hauls for which
catch sampling is
undertaken.
Sampling level up to
100%.
Can quickly be assessed at the start
of each longline set/haul.
Relevant to fishery characterization
(seabird bycatch risk) and
compliance.
Bird curtain
Dehooker and line-cutter
use
Sampling rate same as
for focal taxa; up to
100%.
Relevant to fishery characterization
(impacts of bycatch on non-target
species) and compliance.
Bycatch handling
Wire traces
Subsample
Combine with hook subsampling.
Relevant to fishery characterization
(bycatch risk) and compliance.
Investigate consistency in usage per basket
and attrition/maintenance through trips, to
determine optimal subsampling rates in
lieu of requiring comprehensive counts.
Line weights
Subsample
Hook-shielding devices
Subsample
Dyed bait
Subsample
Fish waste discharge
Opportunistic
Partial documentation by EM.
Relevant to fishery characterization
(bycatch risk).
Collect data from EM review as objectives
require and resources allow.
47
8. Acknowledgements
The authors thank the many case study contributors, as well as C. Heberer, G. Hurry, G.
Legorburu, M. Michelin, and JT Mudge, for sharing their knowledge, experience and insights.
(See individual case studies for a full list of contributors). Thanks also to Pew’s International
Fisheries and Conservation Science teams and peer reviewers H. Walton and A. Barney for their
contributions, which enhanced the products of this project.
This work was completed with the support of The Pew Charitable Trusts.
The views expressed herein are those of the authors and contributors, and do not necessarily
reflect the views of The Pew Charitable Trusts.
48
9. References
1. AFMA. 2020. Australian Fisheries Management Authority Electronic Monitoring Program:
Program overview June 2020. Australian Fisheries Management Authority, Canberra.
Available at:
https://www.afma.gov.au/sites/default/files/australian_fisheries_management_authority_
electronic_monitoring_program_june_2020.pdf [Accessed 15 June 2022]
2. AFMA. 2020. Electronic monitoring trial. Feasibility and effectiveness of electronic
monitoring in the Commonwealth Trawl Sector of the Southern and Eastern Scalefish and
Shark Fishery. Available at:
https://www.afma.gov.au/sites/default/files/electronic_monitoring_trawl_trial_report_pu
blic_doc.pdf [Accessed 15 June 2022]
3. Akroyd, J. and McLoughlin, K. 2020. Fiji albacore, yellowfin and bigeye tuna longline. Marine
Stewardship Council Public Certification Report. Lloyd’s Register, Edinburgh. Available at:
https://fisheries.msc.org/en/fisheries/fiji-albacore-yellowfin-and-bigeye-tuna-
longline/@@assessments [Accessed 15 June 2022]
4. Amandè, M.J., Chassot, E., Chavance, P., Murua, H., de Molina, A.D. and Bez, N. 2012.
Precision in bycatch estimates: the case of tuna purse-seine fisheries in the Indian Ocean.
ICES Journal of Marine Science 69: 1501–1510. doi:10.1093/icesjms/fss106
5. Alaska Fisheries Science Center and Alaska Regional Office. 2021. North Pacific Observer
Program 2020 Annual Report. AFSC Processed Report 2021-03, Alaska Fisheries Science
Center, NOAA, National Marine Fisheries Service, Seattle.
6. Alaska Regional Office. 2021. Alaska Region Electronic Technologies Implementation Plan.
Alaska Fisheries Science Center, NOAA, National Marine Fisheries Service, Seattle.
7. Babcock, E., Pikitch, E.K. and Hudson, C.G. 2003. How much observer coverage is enough to
adequately estimate bycatch? Available at:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.201.3575&rep=rep1&type=pdf
[Accessed 15 June 2022]
8. Barlow, P.F. and Berkson, J. 2012. Evaluating methods for estimating rare events with zero-
heavy data: a simulation model estimating sea turtle bycatch in the pelagic longline fishery.
Fisheries Bulletin 110: 344–360.
9. Brodziak, J. and Walsh, W.A. 2013. Model selection and multimodel inference for
standardizing catch rates of bycatch species: a case study of oceanic whitetip sharks in the
Hawaii-based longline fishery. Canadian Journal of Fisheries and Aquatic Sciences 70:
1723-1740. https://doi.org/10.1139/cjfas-2013-0111
10. Brown, C.J., Desbiens, A., Campbell, M.D., Game, E.T., Gilman, E., Hamilton, R.J., Heberer, C.,
Itano, D. and Pollock, K. 2021. Electronic monitoring for improved accountability in western
Pacific tuna longline fisheries. Marine Policy 132: 104664.
https://doi.org/10.1016/j.marpol.2021.104664
11. Carnes, M.J., Stahl, J.P., Bigelow, K.A. 2019. Evaluation of electronic monitoring pre-
implementation in the Hawaiʻi-based longline fisheries. NOAA Technical Memorandum
NMFS-PIFSC-90. doi: 10.25923/82gg-jq77
12. CCSBT. 2021. Report of the Sixteenth Meeting of the Compliance Committee. Online. 5 – 7
2021. 5 – 7 October 2021. Commission for the Conservation of Southern Bluefin Tuna.
13. Chordata and Saltwater Inc. 2022. Alaska fixed gear data quality report. Unpublished
report.
14. Cocas, L. 2018. Chilean approach to understand and mitigate bycatch in fisheries. Pp. 14 –
17 in: Kennelly, S.J. and Borges, L. (eds.). 2018. Proceedings of the 9th International
Fisheries Observer and Monitoring Conference, Vigo.
49
15. Cocas, L. 2019. Experience from the implementation of camera and electronic surveillance
to monitor discards and bycatch in Chile. Presentation to the special session of the Icelandic
Seafood Conference: Remote Electronic Monitoring in Fisheries. Available at:
https://matis.is/media/fyrirlestrar/5.-Experience%20_from_EM_in_Chile.pdf [Accessed 15
June 2022]
16. Common Oceans (ABNJ) Tuna Project. 2017. Joint analysis of sea turtle mitigation
effectiveness. Final Report. Available at: https://www.fao.org/3/bq849e/bq849e.pdf
[Accessed 15 June 2022]
17. Course, G.P., Pierre, J., and Howell, B.K., 2020. What’s in the Net? Using camera technology
to monitor, and support mitigation of, wildlife bycatch in fisheries. Published by WWF.
Available at: https://www.wwf.org.uk/sites/default/files/2020-11/whatsinthenetfinal.pdf
[Accessed 15 June 2022]
18. Curtis, K.A. and Carretta, J.V. 2020. ObsCvgTools: Assessing observer coverage needed to
document and estimate rare event bycatch. Fisheries Research 225: 105493.
19. Emery, T. and Nicol, S. 2017. An initial examination of CCSBT observer program standard
data fields and their ability to be collected using electronic monitoring (EM) technologies.
Australian Bureau of Agricultural and Resource Economics and Sciences. Canberra.
Working Paper CCSBT-ESC/1708/10 prepared for the CCSBT Extended Scientific
Committee for the 22nd Meeting of the Scientific Committee. 28 August–2 September 2017.
Yogyakarta, Indonesia. Available at:
https://www.ccsbt.org/en/system/files/ESC22_18_AU_EM%20data%20standards.pdf
[Accessed 15 June 2022]
20. Emery, T., Noriega, R., Williams, A.J. and Larcombe, J. 2019. Changes in logbook reporting by
commercial fishers following the implementation of electronic monitoring in Australian
Commonwealth fisheries. Marine Policy 104: 135-145.
21. Emery, T., Noriega, R., Williams, A.J., Larcombe, J., Nicol, S., Williams, P., Smith, N., Pilling, G.,
Hosken, M., Brouwer, S., Tremblay-Boyer, L. and Peatman, T. 2018. The use of electronic
monitoring within tuna longline fisheries: implications for international data collection,
analysis and reporting. Reviews in Fish Biology and Fisheries 28. doi: 10.1007/s11160-
018-9533-2
22. ERandEMWG Chair. 2020. Draft consultative proposal for minimum standards for WCPFC’s
e-monitoring programme (REMP). WCPFC-ERandEMWG4-2020-03.
https://meetings.wcpfc.int/node/11906 [Accessed 1 June 2022]
23. ERandEMWG Chair. 2020. Draft consultative proposal for a CMM for a regional e-
monitoring programme (REMP). WCPFC-ERandEMWG4-2020-02.
https://meetings.wcpfc.int/node/11905 [Accessed 1 June 2022]
24. ERandEMWG Chair. 2022. Chair’s summary report. 5th E-reporting and E-monitoring
Working Group Meeting (ERandEMWG5). Electronic meeting.
https://meetings.wcpfc.int/node/16418 [Accessed 1 June 2022]
25. Ewell, C., Hocevar, J., Mitchell, E., Snowden, S. and Jacquet, J. 2020. An evaluation of Regional
Fisheries Management Organization at-sea compliance monitoring and observer programs.
Marine Policy 115. https://doi.org/10.1016/j.marpol.2020.103842
26. Faunce, C. 2013. The restructured North Pacific groundfish and halibut observer program.
Alaska Fisheries Science Center Quarterly Report. January-March. Available at:
https://apps-afsc.fisheries.noaa.gov/Quarterly/jfm2013/jfm13.htm [Accessed 13 May
2022]
27. Fisheries New Zealand. 2022. Digital Monitoring Update - 25 May 2022. Available at:
https://mpi.govt.nz/dmsdocument/51571/direct [Accessed 13 May 2022]
50
28. Flewwelling, P., Cullinan, C., Balton, D., Sautter, R.P. and Reynolds, J.E. 2002. Recent trends
in monitoring, control and surveillance systems for capture fisheries. FAO Fisheries
Technical Paper. No. 415. FAO, Rome.
29. Gilman, E. and Zimring, M. 2018. Meeting the objectives of fisheries observer programs
through electronic monitoring. The Nature Conservancy, Honolulu.
30. Gilman, E., Chaloupka, M., Ishizaki, A., Carnes, M., Naholowaa, H., Brady, C., Ellgen, S. and
Kingma, E. 2021. Tori lines mitigation seabird bycatch in a pelagic longline fishery. Reviews
in Fish Biology and Fisheries. https://doi.org/10.1007/s11160-021-09659-7
31. Gilman, E., Legorburu, G., Fedoruk, A., Heberer, C., Zimring, M. and Barkai, A. 2019.
Increasing the functionalities and accuracy of fisheries electronic monitoring systems.
Aquatic Conservation: Marine and Freshwater Ecosystems 29: 901 – 926.
https://doi.org/10.1002/aqc.3086
32. Hosken, M., Vilia, H., Agi, J., Williams, P., Mckechnie, S., Mallet, D., Honiwala, E., Walton, H.,
Owens, M., Wickham, C., Zaborovskiy, E. and Cheung, B. 2016. Report on the 2014 Solomon
Islands longline e-monitoring project.
https://oceanfish.spc.int/en/publications/doc_download/1482-hosken-2016-si-ereport
[Accessed 13 May 2022]
33. Hurry, G. 2019. Building a business case for electronic monitoring (EM) for the Fiji long line
(LL) fishing industry. MRAG AsiaPacific.
34. IATTC Staff. 2021a. A proposed workplan for the implementation of an electronic
monitoring system for the tuna fisheries in the eastern Pacific Ocean. Document EMS-01-02
Revised. Workshop on the Implementation of an Electronic Monitoring System (EMS). 1st
Meeting (by videoconference). 22 – 23 April 2021. Inter-American Tropical Tuna
Commission.
35. IATTC Staff. 2021b. Staff recommendations for the implementation of an electronic
monitoring system for the tuna fisheries in the eastern Pacific Ocean. Document EMS-01-
01. Workshop on the Implementation of an Electronic Monitoring System (EMS). 1st
Meeting (by videoconference). 22 – 23 April 2021. Inter-American Tropical Tuna
Commission.
36. ICCAT. 2018. Report of the Standing Committee on Research and Statistics (SCRS). Madrid,
Spain. 1 – 5 October 2018. International Commission for the Conservation of Atlantic Tunas.
37. ICES. 2021. Annex 4: Case studies of ET monitoring programme solutions. Working Group
on Technology Integration for Fishery-Dependent Data (WGTIFD; outputs from 2020
meeting). ICES Scientific Reports. 3:03. https://doi.org/10.17895/ices.pub.7684
38. IOTC-WGEMS01. 2021. Report of the 1st Session of the IOTC Ad-hoc Working Group on the
Development of Electronic Monitoring Programme Standards. IOTC–2021–WGEMS–R[E].
Online. 15-17 November 2021. Indian Ocean Tuna Commission.
39. IOTC Secretariat. 2021. Update on the implementation of the IOTC regional observer
scheme. IOTC-2021-WGEMS01-09. 17th Working Party on Data Collection and Statistics.
Online. 29 November – 3 December 2021. Indian Ocean Tuna Commission.
40. Kay, J. and Merrifield, M. 2021. The Fishnet Open Images Database: a dataset for fish
detection and fine-grained categorization in fisheries. Available at:
https://arxiv.org/pdf/2106.09178.pdf [Accessed 15 June 2022]
41. Kindt-Larsen, L., Dalskov, J., Stage, B. and Larsen, F. 2012. Observing incidental harbour
porpoise Phocoena phocoena bycatch by remote electronic monitoring. Endangered Species
Research 19: 75 – 83.
42. Lawson, T. 2006. Scientific aspects of observer programmes for tuna fisheries in the
western and central Pacific Ocean. SC2-2006/ST WP-1. 2nd Regular Session of the Scientific
Committee, Manila, Philippines. 7 – 18 August 2006. Western and Central Pacific Fisheries
Commission.
51
43. Legorburu G., Lekube X., Canive I., Ferré J.G., Delgado H., Moreno G., Restrepo V. 2018.
Efficiency of electronic monitoring on FAD related activities by supply vessels in the Indian
Ocean. ISSF Technical Report 2018-03. International Seafood Sustainability Foundation,
Washington, D.C.
44. Lenel, S. 2020. Monitoring, control, and surveillance of deep-sea fisheries in areas beyond
national jurisdiction. FAO, Rome. https://doi.org/10.4060/ca7320en
45. Linden, D. 2021. A predictive model of discarded catch that leverages self-reporting and
electronic monitoring on commercial fishing vessels. Available at:
https://media.fisheries.noaa.gov/2022-03/LindenEMdeltamodelpaperCIE-GARFO.pdf
[Accessed 15 June 2022]
46. Lowman, D.M., Fisher, R., Holliday, M.C., McTee, S.A. and Stebbins, S. 2013. Fisheries
Monitoring Roadmap. Environmental Defense Fund. Available at:
https://www.edf.org/sites/default/files/FisheryMonitoringRoadmap_FINAL.pdf [Accessed
15 June 2022]
47. Marine Stewardship Council. 2022. Consultation on review of the Fisheries Standard.
https://www.msc.org/standards-and-certification/developing-our-standards/the-
fisheries-standard-review/consultation [Accessed 22 June 2022]
48. McElderry, H., Pria, M.J., Dyas, M. and McVeigh, R. 2010. A Pilot Study Using EM in the
Hawaiian Longline Fishery. Archipelago Marine Research Ltd, Victoria. Available at:
www.wpcouncil.org/library/docs/Archipelago_EM_Pilot_Study_Final.pdf [Accessed 20 May
2022].
49. Michelin, M. and Zimring, M. 2020. Catalyzing the growth of electronic monitoring in
fisheries. Progress update. August 2020. CEA Consulting and The Nature Conservancy.
50. Michelin, M., Elliott, M. Bucher, M. Zimring, M. and Sweeney, M. 2018. Catalyzing the growth
of electronic monitoring in fisheries: Building greater transparency and accountability at
sea. Opportunities, barriers, and recommendations for scaling the technology. California
Environmental Associates and The Nature Conservancy.
51. Michelin, M., Sarto, N.M. and Gillett, R. 2020. Roadmap for electronic monitoring in RFMOs.
CEA Consulting. Available at: https://www.ceaconsulting.com/wp-
content/uploads/CEA.Roadmap-EM-Report-4.23.20.pdf [Accessed 13 May 2022]
52. Monteagudo J.P., Legorburu, G., Justel-Rubio, A. and Restrepo, V. 2015. Preliminary study
about the suitability of an electronic monitoring system to record scientific and other
information from the tropical tuna purse seine fishery. SCRS 2014/132. Collected Volumes
of Scientific Papers ICCAT: 71: 440-459.
53. Morón, J. and Herrera, M. 2020. Electronic Monitoring Systems: The OPAGAC Experience
Presentation available at: http://www.transparentfisheries.org/wp-
content/uploads/2020/06/Julio-Moro%CC%81n_REM.pdf [Accessed 13 May 2022]
54. MRAG. 2017. Building the business case for EMS in the Ghanaian tuna purse seine fleet.
WWF US, US2324. Final Report. MRAG, London.
55. Murua, H., Fiorellato, F., Ruiz, J., Chassot, E. and Restrepo, V. 2020a. Minimum standards for
designing and implementing electronic monitoring systems in Indian Ocean tuna fisheries.
IOTC–2020–SC23–12[E] rev2. 23rd session of the Scientific Committee. Online. 7 – 11
December 2020. Indian Ocean Tuna Commission.
56. Murua, H., Herrera, M., Morón, J., Abascal, F., Legorburu, G., Hosken, M., Roman, M., Panizza,
A., Wichman, M., Moreno, G. and Restrepo, V. 2020b. Comparing Electronic Monitoring and
human observer collected fishery data in the tropical tuna purse seine operating in the
western and central Pacific Ocean. WCPFC-SC16-2020/ST-IP-09. 16th Regular Session of
the Scientific Committee, Electronic Meeting. 11 – 20 August 2020. Western and Central
Pacific Fisheries Commission.
52
57. NFMS. 2019. Three-year review of the individual bluefin quota program. September 2019.
National Marine Fisheries Service, NOAA Fisheries.
58. NFMS. 2020. 2021 Annual Deployment Plan for Observers and Electronic Monitoring in the
Groundfish and Halibut Fisheries off Alaska. National Oceanic and Atmospheric
Administration, Juneau.
59. NFMS. 2021a. 2022 Annual Deployment Plan for Observers and Electronic Monitoring in
the Groundfish and Halibut Fisheries off Alaska. National Oceanic and Atmospheric
Administration, Juneau.
60. NFMS. 2021b. Atlantic highly migratory species electronic technologies implementation
plan. National Marine Fisheries Service, NOAA Fisheries.
61. NFMS. 2021c. Alaska region electronic technologies implementation plan. Alaska Regional
Office, Juneau and Fisheries Monitoring and Analysis Division and Alaska Fisheries Science
Center, Seattle. National Marine Fisheries Service, NOAA Fisheries.
62. NFMS. 2021d. Pacific Islands region electronic technologies implementation plan. National
Marine Fisheries Service, NOAA Fisheries.
63. NFMS. 2021e. West Coast region electronic technologies implementation plan. National
Marine Fisheries Service, NOAA Fisheries.
64. NOAA. 2019. Atlantic Highly Migratory Species; Shortfin Mako Shark Management
Measures; Final Amendment 11.
https://www.federalregister.gov/documents/2019/02/21/2019-02946/atlantic-highly-
migratory-species-shortfin-mako-shark-management-measures-final-amendment-11
[Accessed 15 June 2022]
65. NOAA. 2020. The National Electronic Monitoring Workshop Report 2019 | 2020. Office of
Science and Technology, NOAA Fisheries.
66. NOAA. 2022. North Pacific Observer Program.
https://www.fisheries.noaa.gov/alaska/fisheries-observers/north-pacific-observer-
program [Accessed 15 June 2022]
67. NOAA. 2022. Zero Atlantic shortfin mako shark retention limit. Final Rule.
https://www.fisheries.noaa.gov/action/zero-atlantic-shortfin-mako-shark-retention-limit
[Accessed 15 June 2022]
68. OPAGAC. 2020. Good practices for responsible tuna purse-seining. https://opagac.org/wp-
content/uploads/2022/02/Buenas-Pra%CC%81cticas-OPAGAC-ANABAC-2020_english.pdf
[Accessed 1 May 2022]
69. Paiva, C. 2019. EM Program costs: West Coast. Presentation to the US National Electronic
Monitoring Workshop – East Coast, New Castle.
70. Panizza, A., Williams, P., Falasi, C., Loganimoce, E. and Schneiter, E. 2021. Status of observer
data management. WCPFC-SC17-2021/ST-IP-02. 17th Regular Session of the Scientific
Committee, Online. 11 – 19 August 2021. Western and Central Pacific Fisheries
Commission.
71. Pierre, J. P. 2018. Using electronic monitoring imagery to characterise protected species
interactions with commercial fisheries: A primer and review. Final Report prepared for the
Conservation Services Programme, Department of Conservation. Available at:
https://dcon01mstr0c21wprod.azurewebsites.net/globalassets/documents/conservation/
marine-and-coastal/marine-conservation-services/reports/int2017-02-final-report-
em.pdf [Accessed 1 May 2022]
72. Restrepo, V., Ariz, J., Ruiz, J., Justel-Rubio, A. and Chavance P. 2014. Updated guidance on
Electronic Monitoring Systems for tropical tuna purse seine fisheries. ISSF Technical Report
2014 - 08 International Seafood Sustainability Foundation, Washington, D.C..
73. Román, M., Lopez, J., Lennert-Cody, C., Ureña, E., Aires-da-Silva, A. 2020. An electronic
monitoring system for the tuna fisheries in the eastern Pacific Ocean: Objectives and
53
standards. Document SAC-11-10. Scientific Advisory Committee 11th Meeting. 11 – 15 May
2020. La Jolla, California (USA). Inter-American Tropical Tuna Commission.
74. Ruiz, J., Batty, A., Chavance, P., McElderry, H., Restrepo, V., Sharples, P., Santos, J. and
Urtizberea, A. 2014. Electronic monitoring trials on in the tropical tuna purse-seine fishery.
ICES Journal of Marine Science 72: 1201–1213. https://doi.org/10.1093/icesjms/fsu224
75. Ruiz, J., de Lagos, E.M., Canive, I., Grande, M., Krug, I. and Santos, M. 2021. Electronic
monitoring programs conducted by AZTI and DATAFISH in the Spanish tuna fisheries.
IOTC-2021-WGEMS01-04_Rev1.
76. Ruiz, J., Krug, I., Justel-Rubio, A., Restrepo, V., Hammann, G., Gonzalez, O., Legorburu, G.,
Alayon, P.J.P., Bach, P., Bannerman, P. and Galán, T. 2017. Minimum standards for the
implementation of electronic monitoring systems for the tropical tuna purse seine fleet.
SCRS/2016/180. Collective Volumes of Scientific Papers ICCAT 73(2): 818-828.
77. Scientific Committee. 2020. 5th Scientific Committee Meeting Report. NPFC-2020-SC05-
Final Report. North Pacific Fisheries Commission. Available at:
https://www.npfc.int/sites/default/files/2021-01/SC05%20Report.pdf [Accessed 13 May
2022]
78. Scientific Committee. 2021. 6th Meeting Report. NPFC-2021-SC06-Final Report. North
Pacific Fisheries Commission. Available at: https://www.npfc.int/sites/default/files/2022-
02/SC06%20Report.pdf [Accessed 13 October 2022]
79. Stahl, J. and Carnes, M. 2020. Detection accuracy in the Hawaiʻi longline electronic
monitoring program with comparisons between three video review speeds. PIFSC Data
Report DR-20-012. https://doi.org/10.25923/n1gq-m468
80. Standing Committee on Research and Statistics (SCRS). 2021. Report of the Standing
Committee on Research and Statistics (SCRS). (Online, 27 September to 2 October 2021).
International Commission for the Conservation of Atlantic Tunas.
81. Stanley, R.D., McElderry, H., Mawani, T. and Koolman, J. 2011. The advantages of an audit
over a census approach to the review of video imagery in fishery monitoring. ICES Journal
of Marine Science 68: 1621–1627. doi:10.1093/icesjms/fsr058
82. Stanley, R.D., Karim, T., Koolman, J. and McElderry, H. 2015. Design and implementation of
electronic monitoring in the British Columbia groundfish hook and line fishery: a
retrospective view of the ingredients of success. ICES Journal of Marine Science 72: 1230–
1236. doi: 10.1093/icesjms/fsu212
83. Stobberup, K., Anganuzzi, A., Arthur-Dadzie, M., Baidoo-Tsibu, G., Hosken, M., Kebe, P.,
Kuruc, M., Loganimoce, E., Million, J., Scott, G., Spurrier, L., and Tavaga, N. 2021. Electronic
monitoring in tuna fisheries: strengthening monitoring and compliance in the context of
two developing states. FAO Fisheries and Aquaculture Technical Paper No. 664. FAO, Rome.
https://doi.org/10.4060/cb2862en
84. Sylvia, G., Harte, M. and Cusack, C. 2016. Challenges, opportunities and costs of electronic
fisheries monitoring. Prepared for: The Environmental Defense Fund, San Francisco.
Available at:
https://www.edf.org/sites/default/files/electronic_monitoring_for_fisheries_report_-
_september_2016.pdf [Accessed 13 May 2022]
85. The Nature Conservancy. 2021. Sea change: The Nature Conservancy and Thai Union
partner around game-changing transparency pledge. https://www.nature.org/en-
us/newsroom/nature-conservancy-thai-union-partner-game-changing-transparency-
pledge/ [Accessed 15 June 2022]
86. The Nature Conservancy. 2021. Shedding light on the sea to save it.
https://www.nature.org/en-us/what-we-do/our-insights/perspectives/shedding-light-
sea-save/ [Accessed 15 June 2022]
54
87. Tseng, C.-H. and Kuo, Y.-F. 2020. Detecting and counting harvested fish and identifying fish
types in electronic monitoring system videos using deep convolutional neural networks.
ICES Journal of Marine Science 77: 1367–1378. doi:10.1093/icesjms/fsaa076
88. UNE Standard 195007:2021. Observación electrónica en buques pesqueros. Requisitos.
https://www.en-standard.eu/une-195007-2021-observacion-electronica-en-buques-
pesqueros-requisitos/ [Accessed 15 June 2022]
89. van Helmond, A.T.M., Mortensen, L.O., Plet-Hansen, K.S., Ulrich, C., Needle, C.L., Oesterwind,
D., Kindt-Larsen, L., Catchpole, T., Mangi, S., Zimmerman, C., Olesen, H.J., Bailey, N.,
Bergsson, H., Dalskov, J., Elson, J., Hosken, M., Peterson, L., McElderry, H., Ruiz, J., Pierre, J.P.,
Dykstra, C. and Poos, J.J. 2020. Electronic monitoring in fisheries: Lessons from global
experiences and future opportunities. Fish and Fisheries 21: 162-189. doi:
10.1111/faf.12425.
90. Virdin, J., Vegh, T., Ratcliff, B., Havice, E., Daly, J. and Stuart, J. 2022. Combatting illegal
fishing through transparency initiatives: lessons learned from comparative analysis of
transparency initiatives in seafood, apparel, extractive, and timber supply chains. Marine
Policy 138: 104984. https://doi.org/10.1016/j.marpol.2022.104984
91. Wakefield, C.B., Hesp, S.A., Blight, S., Molony, B.W., Newman, S.J. and Hall, N.G. 2018.
Uncertainty associated with total bycatch estimates for rarely-encountered species varies
substantially with observer coverage levels: Informing minimum requirements for
statutory logbook validation. Marine Policy 95: 273 – 282.
92. WCPFC. 2022. Vessel Monitoring System. https://www.wcpfc.int/vessel-monitoring-
system [Accessed 15 June 2022]
93. WCPFC Secretariat. 2020. Outcomes of the review of the Commission’s data needs and
collection programmes (SC Project 93). WCPFC-ERandEMWG4-2020-04. 4th E-Reporting
and E-Monitoring Working Group Meeting (ERandEMWG4). Virtual meeting. 14 October
2020. Western and Central Pacific Fisheries Commission.
94. WG-EMS. 2022. Report of the meeting of the working group on electronic monitoring
systems (WG-EMS). Available at:
https://www.iccat.int/Documents/Meetings/Docs/2022/REPORTS/2022_EMS_WG_ENG.p
df [Accessed 15 June 2022]
95. Williams, P., Tuiloma, I. and Panizza, A. 2018. Status of observer data management. WCPFC-
TCC14-2018-IP04. 14th Regular Session of the Technical and Compliance Committee,
Majuro, Marshall Islands. 26 September – 2 October 2018. Western and Central Pacific
Fisheries Commission.
96. Zollett, E.A. and Swimmer, Y. 2019. Safe handling practices to increase post-capture
survival of cetaceans, sea turtles, seabirds, sharks, and billfish in tuna fisheries. Endangered
Species Research 38: 115 – 125.
55
Appendix 1. Data requirements that support fisheries management by selected Regional Fisheries
Management Organizations.
Regional Fisheries Management Organizations (RFMOs) are IATTC: Inter-American Tropical Tuna Commission, ICCAT: International Commission for the Conservation of Atlantic Tunas, IOTC:
Indian Ocean Tuna Commission, WCPFC: Western and Central Pacific Fisheries Commission, CCSBT: Commission for the Conservation of Southern Bluefin Tuna, NPFC: North Pacific Fisheries
Commission. Material relating to ICCAT is derived from the Protocol to amend the International Convention for the Conservation of Atlantic Tunas, as adopted by the Contracting Parties to
ICCAT on 18 November 2019. The Protocol has not entered into force as yet, while Contracting Parties deposit their instruments of approval, ratification, or acceptance. ‘Functions / actions’
are the obligations or actions of the Commissions that are relevant to the information that can be collected by electronic monitoring. Those in bold have specific quantitative or analytical
meanings. ‘Focal fish stocks’ are those in-scope for the Convention, that are not associated or dependent species, or otherwise identified as non-target species. MSY = Maximum Sustainable
Yield. FAO = Food and Agriculture Organization of the United Nations. *1982 Convention = the United Nations Convention on the Law of the Sea of 10 December 1982; Agreement = Agreement
for the Implementation of the Provisions of the United Nations Convention on the Law of the Sea of 10 December 1982 relating to the Conservation and Management of Straddling Fish Stocks
and Highly Migratory Fish Stocks. Italic table text indicates verbatim wording.
IATTC
https://www.iattc.org/
ICCAT
https://www.iccat.int/en/
IOTC
https://www.iotc.org/
RFMO Objectives
to ensure the long-term conservation and
sustainable use of the fish stocks covered by
this Convention, in accordance with the
relevant rules of international law
to cooperate in maintaining the populations of
tuna and tuna-like fishes and elasmobranchs
that are oceanic, pelagic, highly migratory and
found in the Atlantic Ocean, at levels that will
permit their long-term conservation and
sustainable use for food and other purposes
promote cooperation among Members with a
view to ensuring, through appropriate
management, the conservation and optimum
utilization of stocks covered by this Agreement
and encouraging sustainable development of
fisheries based on such stocks
Focal fish stocks
stocks of tunas and tuna-like species and
other species of fish taken by vessels fishing
for tunas and tuna-like species in the
Convention Area
populations of tuna and tuna-like fishes and
elasmobranchs that are oceanic, pelagic, and
highly migratory found in the Atlantic Ocean
populations of 16 tuna and tuna-like species
occurring in the Convention Area or
migrating into or out of that Area
Precautionary
approach explicit
Yes
Yes
No
Principles /
Functions /
Actions
research on the abundance, biology and
biometry of focal fish stocks, and
associated or dependent species as
necessary
protect biodiversity in the marine
environment
to gather scientific information, catch and
effort statistics and other data relevant to the
conservation and management of the stocks
and to fisheries based on the stocks
adopt evidence-based measures such
that harvested stock status supports
MSY
study of the populations of tuna and tuna-like
fishes and elasmobranchs that are oceanic,
pelagic, and highly migratory species, species
belonging to the same ecosystem, and
associated and dependent species
to adopt evidence-based conservation
and management measures to ensure the
conservation and optimum utilization of the
focal stocks
56
assess focal stock status, with specific
reference to whether stocks are fully
fished or overfished
collecting and analyzing statistical
information relating to the current conditions
and trends of focal species
ensure implementation and enforcement of
binding conservation and management
measures [Members]
adopt measures to ensure associated
and dependent species populations are
above levels at which reproduction
may be seriously threatened
studying and appraising information
relevant to ensuring focal species
populations are at or above levels
commensurate with MSY, and exploitation
is consistent with MSY
adopt appropriate measures to address
waste, discards, non-target catch, catch by
lost or discarded gear, and impacts on
associated and dependent species
make evidence-based recommendations
designed to ensure focal species are at or
above status commensurate with MSY
ensure fishing effort does not exceed a
level commensurate with sustainable
use of focal species
promote the conservation of associated
and dependent species such that
populations are above levels at which
reproduction may become seriously
threatened
ensure implementation of, and compliance
with, adopted conservation and
management measures [Parties]
take all actions necessary to enforce the
Convention [Members]
WCPFC
https://www.wcpfc.int/
CCSBT
https://www.ccsbt.org/
NPFC
https://www.npfc.int/
RFMO Objective
to ensure, through effective management,
the long-term conservation and sustainable
use of highly migratory fish stocks in the
western and central Pacific Ocean in
accordance with the 1982 Convention and
the Agreement*
to ensure, through appropriate management,
the conservation and optimum utilisation of
southern bluefin tuna
to ensure the long-term conservation and
sustainable use of the fisheries resources in the
Convention Area while protecting the marine
ecosystems of the North Pacific Ocean in which
these resources occur
Focal fish stocks
all fish stocks listed in Annex 1 of the 1982
Convention that occur in the Convention
Area (except sauries), and other fish
species as determined by the Commission
Southern bluefin tuna
Fish, mollusks, and other marine species
caught by fishing vessels, with specific
exclusions (sedentary species, indicator
species of Vulnerable Marine Ecosystems,
catadromous species, marine mammals,
reptiles and birds, other species covered by
pre-existing management instruments)
Precautionary
approach explicit
Yes
No
Yes
57
Principles /
Functions /
Actions
determine the total allowable catch or
total level of fishing effort and adopt
measures to ensure long-term
sustainability of highly migratory fish
stocks
consider regulatory measures for
conservation, management and optimum
utilization of southern bluefin tuna including
catch limits, and any other measures
promote the optimum utilization and ensure
the long-term sustainability of fisheries
resources
adopt evidence-based measures such
that stock status supports MSY
assess and analyze the status and trends of
the population of southern bluefin tuna
[Scientific Committee]
adopt evidence-based measures such that
stock status supports MSY and long-term
sustainability of fisheries resources is
ensured
assess the impacts of fishing on target
stocks, non-target species, and species
belonging to the same ecosystem or
dependent upon or associated with the
target stocks
report views on the stock status of southern
bluefin tuna and, as appropriate, ecologically
related species [Scientific Committee]
adopt measures in accordance with the
precautionary approach and an ecosystem
approach to fisheries
adopt measures to ensure non-target,
associated and dependent species
populations are above levels at which
reproduction may be seriously
threatened
provide scientific information, fishing catch
and effort statistics and other data relevant to
the conservation of southern bluefin tuna
and, as appropriate, ecologically related
species [Parties, to the Commission]
assess impacts of fishing on associated and
dependent species and those belonging to the
same ecosystem as target stocks
adopt appropriate measures to address
waste, discards, non-target catch, catch by
lost or discarded gear, and impacts on
associated and dependent species
ensure enforcement of the Convention and
compliance with binding measures [Parties]
adopt measures to ensure non-target,
associated and dependent species
populations are above levels at which
reproduction may be seriously threatened
collect complete and accurate data
concerning fishing activities, including
vessel position, catch of target and non-
target species and fishing effort
protect marine biodiversity including by
preventing significant adverse impacts on
Vulnerable Marine Ecosystems
implement and enforce conservation and
management measures through effective
monitoring, control and surveillance
ensure that complete and accurate data
concerning fishing activities are collected,
including target and non-target species
take into account, among other things,
uncertainties relating to the size and
productivity of the stocks, reference
points, stock condition in relation to such
reference points, levels and distributions
of fishing mortality and the impact of
minimize pollution and waste originating
from fishing vessels, discards, catch by lost or
abandoned gear, and impacts on other
species and marine ecosystems
58
fishing activities on non-target and
associated or dependent species
develop data collection and research
programs to assess the impact of fishing
on non-target and associated or
dependent species and their environment
ensure compliance with conservation and
management measures
subject stocks and species of concern to
enhanced monitoring to review their
status and the efficacy of conservation and
management measures (and update
measures in light of new information)
59
Appendix 2. EM review rate evaluation and optimization by EMoptim
Introduction
EMoptim is a prototype simulation tool based in R4, that optimizes a sampling stratification with
stratified random sampling (SRS) for multiple objectives assuming distributions and sampling
based on electronic monitoring of fisheries activity. SRS is a random sampling technique in
which the total population is divided into strata, with samples taken from each stratum and
combined to give a population estimate. When groups with similar properties are combined into
a stratum, greater efficiency can be obtained which reduces the total number of samples
required to achieve a given level of uncertainty in the resulting population estimate.
The EMoptim R software provides a simple interface into functions that (i) provide an expected
sampling coefficient of variation (CV) for a given stratification, based on assumed or actual data
from a fishery; and (ii) provide an estimate of optimal strata definitions and sampling allocation
that can meet multiple objectives with different underlying statistical and spatial distributions
to improve the overall sampling efficiency.
We assume that electronic recording of all fishing events is available, and that the expected
rates of capture are available for a fishery and are ordered in some spatially aggregated manner.
The simulation tool (EMoptim) takes data from an external file that defines the fishery, species
distributions, encounter rates expected (with the assumed statistical distributions), and
definitions of sampling objectives. This is read into R as (and object called) EMobject. Statistical
distributions for the encounter rates that are implemented are the binomial distribution
(parameterized by a proportion p), the Lognormal distribution (parameterized by µ and CV), the
negative binomial distribution parameterized by µ and θ), the normal distribution
(parameterized by µ and σ), the Poisson distribution (parameterized by λ), and the zero-inflated
Poisson (parameterized by λ and pzero, the probability of zero).
Here we use the Western and Central Pacific Fisheries Commission’s (WCPFC) publicly available
data for the longline fishery. Using these data, we provide an example of how the package can
work and be used to identify optimal strata and resulting coverage rates to meet different
objectives across different species of interest and different assumed statistical distributions.
In the example below based on the WCPFC longline data, the assumed data is held in EMobject
which can be created from a simple text file with a specific command and subcommand
structure.
Specification of the input configuration file
Defining the EMoptim fleets, species, encounter rates, and objectives
The EMoptim input configuration file is a plain text file and is made up of a number of
commands (each with subcommands) which specify various options for each of these
components. Commands always begin with an @ character, with several commands also
requiring a label.
Subcommands follow the command, with each subcommand having an argument.
Subcommands have a number of arguments that must be specified. Arguments can be strings,
numbers, or vectors of strings or numbers. The type of argument is always specific to the
subcommand. The order of subcommands or commands in a file does not matter, except that
the subcommands for each command must always follow the associated command and occur
before the next command.
60
For example, to specify the model structure in EMoptim, use the command @model to specify
the size of the grid (rows and columns), and the names of the strata, fleet, and species
definitions. For example, the Convention Area of WCPFC covers a region 60° S to 55° N and 100°
E to 141° W in the Pacific Ocean (Figure 1). This can be represented as a matrix of 29×25 cells of
5°×5° aggregated data.
The fisheries that operated in the region can be classified into ‘fleets’, however for this example
we will assume that there were two fleets, longline and purse seine as these two fleets are the
fisheries that were specified in the publicly available data. In practice, fleets could be defined
using vessel, nation, or other operating characteristics to identify and improve estimates of
incidence rates that may be more appropriate for management advice.
Species of interest will depend on the management requirements. In this example, we use the
publicly available data to identify a subset of species/species groups for the purposes of this
example: yellowfin tuna catch, shark catch, seabird captures, and marine mammal captures.
The WCPFC CMM 2018-03 (Conservation and Management Measure to mitigate the impact of
fishing for highly migratory fish stocks on seabirds) defines seabird management areas for
mitigation requirements: south of 30° S, at least two of the three defined mitigation measures;
25° S to 30° S at least one of the three defined mitigation measures; 25° S to 23° N, no mitigation
measures required; and north of 23° N, to use at least two of the mitigation measures in Table 1
of CMM 2018-03.
Hence, we can specify a model using a map size of 29×25 cells with the pre-defined strata for
the seabird mitigation management (labelled ManagementUnits in this example); the longline
(LL) and purse seine (PS) fleets; longline capture rates for yellowfin tuna (LL_yellowfin), sharks
(LL_shark), seabirds (LL_bird), and marine mammals (LL_mammal); purse seine capture rates
for yellowfin (PS_yellowfin), marine mammals (PS_mammal), and sharks (PS_shark).
This model structure is defined in the EMoptim input configuration file as:
@model
map_rows 29
map_cols 25
strata_definitions ManagementUnits
fleet_definitions LL PS
species_definitions LL_yellowfin LL_porbeagle LL_bird LL_shark LL_mammal PS_yellowfin PS_mammal PS_shark
Historically available public data from the WCPFC has data at 5°×5° aggregated cells for the
longline fishery. We can identify those cells where no sampling should be undertaken (i.e., as no
effort is recorded there, a cell that is on land, or to select subregions of the area of interest, etc.)
with the definition of a base map. This defines those cells that are available for EMoptim to use
in simulations or as valid cells for optimization of strata. Cells with a base map value of zero are
ignored. The base map used for the WCPFC example is given as Figure 2.
The definition of the base map in the input configuration file uses the table and end_table
subcommands to define the map of areas available, i.e.,
@base_map
table data
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
… [more rows]
end_table
61
Similarly, any pre-defined strata have a similar format but use the command @strata [label],
where [label] represents the label for that stratification. Multiple strata can be input using a
separate @strata command for each with unique labels for each one.
@strata ManagementUnits
table data
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
…[more rows]
5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
end_table
We use the 2019 data from the WCPFC in this example (the most recent year for which all data
were available and a year prior to any potential effect of COVID-19 on fishing patterns), but we
note that data could be averaged across several years or different years’ data could be trialed to
assess the effect of temporal variation in fishing patterns on the EMoptim model outputs.
Temporally specific fleet maps and the sampling could also be specified in the optimization to
consider fleet and event distributions for different seasons or temporal periods.
The @fleet definition command and subcommands are similar to @strata, but with the value in
each cell of the matrix indicating the amount of “effort”. The values in the ‘fleet’ represent the
available sampling units.
In this case the publicly available longline data are available in numbers of hooks, but the
sampling unit for any electronic monitoring sampling is most likely to be sets. We assume that
each set represents 3,500 hooks, broadly characteristic of a large-scale pelagic longline fishery3,
and hence assume the effort in each cell is the number of hooks reported in that cell and divided
by 3,500. For the purse seine data, the effort is given in days, and we assume that each day
represents one purse seine haul5. We note that these are approximations, and that set by set
data would ideally be used (while such data are generally publicly unavailable due to
confidentiality requirements).
Associated with each unit of effort is a cost. Here, we assume that identifying each unit of effort
has a fixed cost and is directly proportional to the amount of effort recorded in each fishery (i.e.,
fleet). In this case, we assume that the cost of identifying each unit of effort and characterizing
this using electronic review is: €5 for each day of longline fishing (corresponding to a set, with
approximately one set per day) with an additional €90 per day for analysis where catch
composition is relatively simpler (including identifying bycatch events); and €15 for each day of
purse seine fishing (corresponding to a set, with approximately one set per day) with an
additional €30 per day for standard analysis (including identifying bycatch events) but
excluding length sampling (G. Legorburu, DOS, pers. comm.).
In the case of the WCPFC, the longline fishery (Figure 3) is represented in the input definition
file as,
@fleet LL
table data
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
…[more rows]
0 0 0 0 0 0 0 0 0 25 29 6 0 0 4 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 13 14 33 17 0 0 25 0 0 0 19 11 0 0 0 0 0
…[more rows]
end_table
cost 5
62
For the species used in the WCPFC example, logbook (yellowfin, porbeagle, and oceanic
whitetip) and observer (seabirds and marine mammals) capture rates were available. Capture
rates are given in two @commands. @species[label] gives the expected spatial distribution of
capture rates (scaled to have a maximum of one). The scaling multiplier to scale the
distributions to reported capture rates is given in @encounter, along with the assumed
statistical distribution associated with the capture rate (i.e., Poisson, negative binomial,
lognormal, etc).
@species LL_yellowfin
table data
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 …[more columns]
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 …[more columns]
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 …[more columns]
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 …[more columns]
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0006 0.0015 0.0022 0.0000 0.0000 …[more columns]
…[more rows]
end_table
cost 90
With encounters of yellowfin in the longline fleet described by an assumed lognormal
distribution with µ = 684.5 and CV = 0.3 (i.e., the maximum capture number of yellowfin per
3,500 hooks in 2019 in the WCPFC 5°×5° aggregated data (Figure 4)),
@encounter LL_yellowfin
fleet LL
species LL_yellowfin
type lognormal
mu 684.4986
cv 0.3
Similar distribution maps and encounter assumptions are required for each species or species
group (e.g., seabirds, see Figure 5) and fleet combination that will be evaluated using EMoptim.
Note that in this example we have used the observed capture rates estimated from observer
coverage and available from the WCPFC public aggregated logbook data. Alternatively, species
distribution maps could be used or some other estimates of species’ distributions or capture
rates as the basis for this process.
Sampling objectives are defined using the command @objective[label]. For example, to define
an objective for the longline fleet for sampling of yellowfin, the @objective command specifies
the encounter (LL_yellowfin) along with the target CV. Given that yellowfin abundance is an
important quality to verify, we assume that the target coefficient of variation in the abundance
estimate for yellowfin was CV=0.10.
@objective LL_yellowfin
encounter LL_yellowfin
cv 0.1
Multiple objectives can be supplied (e.g., sampling for sharks, seabirds, marine mammals, etc),
and for each objective there must be ‘fleet’ and ‘species’ maps that correspond to the objective
subcommands fleet and species respectively. (Note that EMoptim will only optimize across
multiple objectives if multiple objectives are defined. See later for optimizing over multiple
objectives).
Simulation ranges and values are defined with the @simulation command. This specifies the
number of simulations (in this example, n = 1000) to use when evaluating an objective; the
63
default sampling_rate (for evaluating current sampling rates for example and set to 3.8%, based
on the level of coverage achieved by observers in the WCPFC in 2019), and the range of
sampling rates to value (min_sampling_rate to max_sampling_rate with the number of steps
equal to steps) to determine the sampling rates required for achieving sampling targets.
@simulations
N_simulations 1000
sampling_rate 0.038
max_sampling_rate 0.99
min_sampling_rate 0.01
steps 26
Note that more simulations improve accuracy but may take some time to undertake. Similarly, a
greater number of steps also may take some to evaluate and complete the simulations. The
speed of EMoptim will also strongly depend on the specifications of the computer used to
undertake the simulations.
Figure 1. The Convention Area of the Western and Central Pacific Fisheries Commission (Source: https://www.wcpfc.int/
[Accessed 15 May 2022]).
64
Figure 2. 5°×5° cells (orange shading) that have been historically fished with longlines in the Convention Area of the
Western and Central Pacific Fisheries Commission between the years 1950 and 2019. Axes show latitude (y axis) and
longitude (x axis) in 25o increments.
65
Figure 3. Density of the longline fleet by 5°×5° cells in the Convention Area of the Western and Central Pacific Fisheries
Commission in the year 2019 (yellow to red indicates increasing density and grey cells indicate zero density). Axes show
latitude (y axis) and longitude (x axis) in 25o increments.
66
Figure 4. Density of logbook yellowfin tuna capture rates in the longline fleet by 5°×5° cells in the Convention Area of the
Western and Central Pacific Fisheries Commission in the year 2019 (yellow to red indicates increasing density and grey
cells indicate zero density). Axes show latitude (y axis) and longitude (x axis) in 25o increments.
67
Figure 5. Density of seabird capture rates reported by observers in the longline fleet by 5×5° cells in the Convention Area
of the Western and Central Pacific Fisheries Commission in the year 2019 (yellow to red indicates increasing density and
grey cells indicate zero density). Axes show latitude (y axis) and longitude (x axis) in 25o increments.
Running EMoptim
EMoptim is run in R (R Core Team 2021), and is available as an R package. The R command
input.config.file() is used to read an EMoptim configuration file into R. This creates an
EMobject that is used by the rest of the EMoptim functions.
> library(EMoptim)
# Input config file
> EM <- input.config.file("WCPFC.def")
Simple lists of the objects available in the input file can be made with the utility function
getObject(). This function takes an EMobject, and character string of the object names (i.e., fleet,
species, encounter, or objective) and its label (or a list of labels if the label is not supplied).
> getObject(EM, "encounter", "LL_shark")
$encounter
[1] "LL_shark"
$cv
[1] 0.3
$min_value
[1] 0
Plotting of the supplied maps of data can be undertaken with the plotEMmap() function, i.e.,
plotEMmap(EM, type="base", xlab="Longitude (5 degree cells)", ylab="Latitude (5 degree
cells)")
plotEMmap(EM, type="fleet", label="LL", xlab="Longitude (5 degree cells)", ylab="Latitude (5
degree cells)")
68
Evaluating a pre-defined stratification
A single objective with a sampling stratification and a chosen sampling rate can be evaluated
using EMsample(). For example, to estimate the sampling fractions for the ManagementUnits
stratification defined above and using the default sampling rate,
> ans1 <- EMsample(EM, objective.label = "LL_bird", strata.label = "ManagementUnits")
This returns an object with the number of samples allocated using Neyman allocation to each of
the stratum in the defined stratification, the expected CV in each stratum, and the overall
expected CV for the sample design.
In this example, the overall expected number of samples would be n = 650, corresponding to a
sampling rate of 3.8% (the supplied default and equal to the achieved overall rate of observer
sampling in the WCPFC longline fishery in 2019), and giving an expected CV = 0.99 for the mean
seabird captures.
The overall CV is given in the cv element, total number of samples for a rate of 3.8% in the N
element, and the parameters for the sampling are given in the parameters element of the object
returned by EMsample(), i.e.,
> set.seed(0)
> ans1 <- EMsample(EM, objective.label = "LL_bird", strata.label = "ManagementUnits")
> ans1$cv
[1] 0.9869387
> ans1$N
[1] 649.572
> ans1$parameters
strata N.population fraction mu sd N
1 1 1342 0.01191153 0.017713459 0.002441512 8
2 2 14770 0.09872754 0.003777525 0.001838661 64
3 3 505 0.03285953 0.158897684 0.017898385 21
4 4 477 0.85650140 0.145950313 0.493916485 556
5 5 0 0.00000000 0.000000000 0.000000000 0
We can then use this stratification to evaluate its how efficient the sampling stratification and
sampling fractions for each stratum would be with, for example, mammals. We can supply the
stratification and sampling fractions in each strata to EMsample().
> set.seed(0)
> ans2 <- EMsample(EM, objective.label = "LL_mammal", strata.label = "ManagementUnits",
sampling.fractions = ans1$parameters$fraction)
> ans2$N
[1] 649.572
> ans2$cv
[1] 3.147359
> ans2$parameters
strata N.population fraction mu sd N
1 1 1342 0.01191153 0.0005685337 2.321188e-05 8
2 2 14770 0.09872754 0.0004237071 1.953779e-05 64
3 3 505 0.03285953 0.0000000000 0.000000e+00 21
4 4 477 0.85650140 0.0018315841 2.432610e-04 556
5 5 0 0.00000000 0.0000000000 0.000000e+00 0
Here, the overall expected CV with the same sample size (n = 650) allocated to the
ManagementUnits strata in the same proportions for marine mammals was CV = 3.15.
Iterating over a range of sampling rates can be used to evaluate a given stratification for its
performance against an objective. To evaluate the expected CV for a given number of samples
(or sampling rates) with a given strata use the function EMiterate(). This takes arguments of
the EMobject along with an objective label.
69
For example, optimizing the sampling rate for seabirds in the longline fishery (and with 26
cores using parallel processing to reduce the time for the iterations to be undertaken), then
summarizing the results (i.e., obtaining the expected CV for each sampling rate)
> LL_bird <- EMiterate(EM, objective.label = "LL_bird", strata.label = "ManagementUnits",
parallel=TRUE, cores = 16)
Optimisation using parallel = TRUE. Using 26 cores
> ans3 <- EMsummary(EM, EMiterations = LL_bird)
$objective
[1] "LL_bird"
$species
[1] "LL_bird"
$fleet
[1] "LL"
$encounter
[1] "LL_bird"
$target.cv
[1] 0.1
$data
strata sampling.rate N cost.total cost.fleet cost.species cv
1 ManagementUnits 0.01000000 170.9400 100854.6 85470 15384.60 1.3229521
2 ManagementUnits 0.01908163 326.1814 114826.3 85470 29356.33 1.1922891
3 ManagementUnits 0.02816327 481.4229 128798.1 85470 43328.06 1.0800177
4 ManagementUnits 0.03724490 636.6643 142769.8 85470 57299.79 0.9323695
5 ManagementUnits 0.04632653 791.9057 156741.5 85470 71271.51 0.9003333
6 ManagementUnits 0.05540816 947.1471 170713.2 85470 85243.24 0.8390496
7 ManagementUnits 0.06448980 1102.3886 184685.0 85470 99214.97 0.8216530
8 ManagementUnits 0.07357143 1257.6300 198656.7 85470 113186.70 0.7832370
9 ManagementUnits 0.08265306 1412.8714 212628.4 85470 127158.43 0.7273614
10 ManagementUnits 0.09173469 1568.1129 226600.2 85470 141130.16 0.7186518
11 ManagementUnits 0.10081633 1723.3543 240571.9 85470 155101.89 0.6642151
12 ManagementUnits 0.10989796 1878.5957 254543.6 85470 169073.61 0.6642138
13 ManagementUnits 0.11897959 2033.8371 268515.3 85470 183045.34 0.6098161
14 ManagementUnits 0.12806122 2189.0786 282487.1 85470 197017.07 0.6056442
15 ManagementUnits 0.13714286 2344.3200 296458.8 85470 210988.80 0.5928917
16 ManagementUnits 0.14622449 2499.5614 310430.5 85470 224960.53 0.5701600
17 ManagementUnits 0.15530612 2654.8029 324402.3 85470 238932.26 0.5669137
18 ManagementUnits 0.16438776 2810.0443 338374.0 85470 252903.99 0.5671152
19 ManagementUnits 0.17346939 2965.2857 352345.7 85470 266875.71 0.5347539
20 ManagementUnits 0.18255102 3120.5271 366317.4 85470 280847.44 0.5229706
…etc…
The resulting data table can be plotted with plotEMsummary(), and is given in Figure 6.
The optimal sampling coverage for the target CV can be approximated using EMoptimise().
EMoptimise() applies a linear approximation to the output of EMiterate(), and then re-runs
the simulator with this value to evaluate the sampling CV for the approximated sample size. In
this case a sampling rate of about 90% will achieve the target CV of 0.10, with the number of
samples in each stratum given below.
> opt1 <- EMoptimise(EM, EMiterations = LL_bird)
$N
[1] 15229.36
> opt1$sampling.rate
[1] 0.8909184
> opt1$parameters
strata N.population fraction mu sd N
1 1 1342 0.01191153 0.017713459 0.002441512 181
2 2 14770 0.09872754 0.003777525 0.001838661 1504
3 3 505 0.03285953 0.158897684 0.017898385 500
4 4 477 0.85650140 0.145950313 0.493916485 13044
5 5 0 0.00000000 0.000000000 0.000000000 0
The consequences of using this sampling design on other objectives can also be evaluated. For
example, we can investigate how the ManagementUnits stratification and sample size allocation
70
for seabirds (opt1) would perform on another species, such as marine mammals (ans2) or
yellowfin tuna (ans3).
> set.seed(0)
> ans2 <- EMsample(EM, objective.label = "LL_mammal", strata.label = "ManagementUnits",
sampling.fractions = opt1$parameters$fraction, sampling.rate = opt1$sampling.rate)
> ans2$N
[1] 15229.36
> ans2$cv
[1] 0.8685586
> ans2$parameters
strata N.population fraction mu sd N
1 1 1342 0.01191153 0.0005685337 2.321188e-05 181
2 2 14770 0.09872754 0.0004237071 1.953779e-05 1504
3 3 505 0.03285953 0.0000000000 0.000000e+00 500
4 4 477 0.85650140 0.0018315841 2.432610e-04 13044
5 5 0 0.00000000 0.0000000000 0.000000e+00 0
> set.seed(0)
> ans3 <- EMsample(EM, objective.label = "LL_yellowfin", strata.label = "ManagementUnits",
sampling.fractions = opt1$parameters$fraction, sampling.rate = opt1$sampling.rate)
> ans3$N
[1] 15229.36
> ans3$cv
[1] 0.006737339
> ans3$parameters
strata N.population fraction mu sd N
1 1 1342 0.01191153 44.564837 19.732969 181
2 2 14770 0.09872754 153.507436 58.517783 1504
3 3 505 0.03285953 60.328047 34.340631 500
4 4 477 0.85650140 7.559763 7.129218 13044
5 5 0 0.00000000 0.000000 0.000000 0
In this example, the sampling stratification with a sampling rate optimized for seabirds (89%)
performed poorly for marine mammals (resulting in an expected CV = 0.86) and significantly
oversampled yellowfin tuna (expected CV = 0.007).
If we required the sampling rate, using the ManagementUnits stratification, that were required
for yellowfin, we could repeat the above optimization with the yellowfin objective, i.e.,
> set.seed(0)
> LL_yellowfin <- EMiterate(EM, objective.label = "LL_yellowfin", strata.label =
"ManagementUnits", parallel = TRUE, cores = 26)
Optimisation using parallel = TRUE. Using 16 cores
> opt2 <- EMoptimise(EM, EMiterations = LL_yellowfin)
> opt2$sampling.rate
[1] 0.07990016
> opt2$parameters
strata N.population fraction mu sd N
1 1 1342 0.029051801 44.564837 19.732969 40
2 2 14770 0.948192381 153.507436 58.517783 1295
3 3 505 0.019025135 60.328047 34.340631 26
4 4 477 0.003730683 7.559763 7.129218 5
5 5 0 0.000000000 0.000000 0.000000 0
This shows that, instead of requiring a sampling rate of allocating 86% of the samples to the
ManagementUnits stratum 4 as was required for the optimization for seabirds, 95% of samples
are allocated to stratum 2. The change of allocation resulted in a different optimum; there was a
considerable reduction in the sample size required for yellowfin tuna, with the number of
samples required dropping from 15,229 to 1,365 (i.e., the coverage rate decreased from 89% to
8%) and the expected CV increased from 0.007 to ~0.08.
To compare the level of improvement in sampling efficiency that was from the stratified
allocation, we can use the same approach, but with the definition of a single stratum to the
71
region. Here, we update the EM definition file to include a new strata definition (labelled None)
with the stratum label equal to 1 in every cell. Then re-run the optimization, i.e.,
> set.seed(0)
> LL_yellowfin2 <- EMiterate(EM, objective.label = "LL_yellowfin", strata.label = "None",
parallel = TRUE, cores = 26)
> ans5 <- EMsummary(EM, EMiterations = LL_yellowfin2)
> opt3 <- EMoptimise(EM, EMiterations = LL_yellowfin2)
> opt3$sampling.rate
[1] 0.2568284
This suggests that, without any stratification, the required sampling rate for yellowfin tuna for
the same expected CV was much higher, at 26%.
Similarly, the effect on other sampling objectives can be found by applying the stratification and
sampling fractions to each objective respectively.
Figure 6: Expected CV with sampling rate, assuming the ManagementUnits stratification for seabirds in the western and
central Pacific Ocean longline fishery. The target CV (shown by the horizontal dashed line) is 0.2. The baseline cost to
assess the EM sampling frame is the light blue line. The dark blue line indicates increased review cost, above the baseline,
as the sampling rate increases.
Evaluating a stratification for multiple objectives
Multiple objectives can be evaluated using genetic algorithms, from the ‘SamplingStrata’ R
package1,2. For example, the list of objectives defined in the WCPFC example can be listed with
getObject(EM,"objective"), and the longline objectives passed to EMoptimiseStrata(). At least
two objectives must be used, but note as the number of objectives increases, the stratification
will tend to be optimized with an increasing number of strata.
72
Each solution (i.e., a particular specification of strata across the map of spatial cells) is
considered as an individual in a population with the fitness of all individuals evaluated by
applying the Bethel-Chromy algorithm to calculate the sampling size to attempt to meet the
precision requirements of the target estimates. ‘SamplingStrata’ uses a modified version of the
functions in the ‘genalg’ package6 to implement the genetic algorithm.
With EMoptimiseStrata(), an additional strata label needs to be defined to hold the resulting
optimal estimated stratification. Optimizing for the longline shark, seabird, and mammal
objectives, and assigning the resulting stratification to the additional strata with label new,
> getObject(EM, "objective")
[1] "LL_yellowfin" "LL_porbeagle" "LL_bird" "LL_shark" "LL_mammal" "PS_yellowfin"
"PS_shark" "PS_mammal"
> set seed(0)
> EM <- EMoptimiseStrata(EM, objective.labels = c("LL_shark", "LL_bird", ”LL_mammal”),
new.strata = "new")
This example resulted in 5 strata, numbered 0:4. An “empty” stratum (stratum 0) is indicated
where no sampling would take place as there was no effort reported in those cells. A
representative image of the stratification is shown in Figure 7. The strata map can be printed to
see what cells were allocated to what stratum for the application of the actual electronic
monitoring review. Note that there is no requirement that stratum be contiguous or a neighbor
to be included within a stratum. Also note that the genetic algorithm may require more
iterations (default value iter = 300) or populations (default value pops = 50) to successfully
converge, and different random number seeds can also produce slightly varying results. For a
specific application, these values may need to be increased. Evaluation of suitable convergence
can be achieved by retesting with different random number seeds and/or by testing with larger
values of iter and pops.
Once the stratification and associated sampling fractions are extracted, these can then be
evaluated against any other or all objectives, using the approach in ‘Evaluating a pre-defined
stratification’ above, i.e.,
> new <- EM$strata[["new"]]
> plotEMmap(EM, type="strata", label="new", xlab="Longitude (5 degree cells)", ylab="Latitude
(5 degree cells)", as.image = TRUE)
> new$fraction$fraction
[1] 0.0000000 0.4835681 0.3145540 0.0657277 0.1361502
> LL_yellowfin <- EMiterate(EM, objective.label = "LL_yellowfin", strata.label = "new", quiet
= FALSE, sampling.fractions = new$fraction$fraction)
Optimisation using parallel = TRUE. Using 26 cores
> LL_bird <- EMiterate(EM, objective.label = "LL_bird", strata.label = "new", quiet = FALSE,
sampling.fractions = new$fraction$fraction)
Optimisation using parallel = TRUE. Using 26 cores
> LL_shark <- EMiterate(EM, objective.label = "LL_shark", strata.label = "new", quiet = FALSE,
sampling.fractions = new$fraction$fraction)
Optimisation using parallel = TRUE. Using 26 cores
> LL_mammal <- EMiterate(EM, objective.label = "LL_mammal", strata.label = "new", quiet =
FALSE, sampling.fractions = new$fraction$fraction)
Optimisation using parallel = TRUE. Using 26 cores
And then summarized using EMsummary() and plotted using plotEMsummary().
Figure 8 gives the expected CV for different levels of sampling for yellowfin tuna with the new
stratification. This shows, that with the stratification optimized over the shark, seabird, and
mammal objectives the target CV for yellowfin tuna (CV = 0.10) is met with a low level of
targeted sampling (i.e., an overall sampling rate of about 2%).
73
The optimal coverage for each species with the resulting stratification and sampling fractions
can be obtained, as earlier, using EMoptimise().
> opt1 <- EMoptimise(EM, EMiterations = LL_yellowfin)
> opt2 <- EMoptimise(EM, EMiterations = LL_bird)
> opt3 <- EMoptimise(EM, EMiterations = LL_shark)
> opt4 <- EMoptimise(EM, EMiterations = LL_mammal)
> opt1$sampling.rate
[1] 0.01908163
> opt2$sampling.rate
[1] 0.8909184
> opt3$sampling.rate
[1] 0.1968998
> opt4$sampling.rate
[1] 0.8727551
This gives the sampling rates required, using the stratification and sampling fractions
determined above with the new stratification, as about 2% to obtain an estimate for yellowfin
tuna with a target CV of 0.10; with 90% coverage required for an estimate of seabirds with a
target CV of 0.10. Less coverage (20%) was required for sharks with a target CV of 0.30; and
about 90% coverage for marine mammals with a target CV of 0.10 (Figure 9).
In this case, a user may wish to reduce the target CV for some species or increase it for others
and rerun to determine the optimal stratification. Or alternatively, sampling at the highest rate
could be implemented to ensure that all objectives are met.
Figure 7. Locations of stratum in the new optimized strata for yellowfin and seabird objectives using longline fleet in the
WCPFC. Axes show latitude (y axis) and longitude (x axis) in 25o increments.
74
Figure 8. Expected CV with sampling rate, assuming the new stratification optimized for the combined bird, shark, and
mammal objectives for yellowfin tuna in the western and central Pacific Ocean longline fishery. The dashed grey line
shows the target CV. The baseline cost to assess the EM sampling frame is the light blue line. The dark blue line indicates
increased review cost, above the baseline, as the sampling rate increases.
75
Figure 9. Expected CV with sampling rate, assuming the new stratification optimized for the combined seabird, shark,
and mammal objectives for sharks in the western and central Pacific Ocean longline fishery. The dashed grey line shows
the target CV. The baseline cost to assess the EM sampling frame is the light blue line. The dark blue line indicates
increased review cost, above the baseline, as the sampling rate increases.
76
References
EMoptim
1. Barcaroli, G. 2014. SamplingStrata: An R Package for the Optimization of Stratified Sampling.
Journal of Statistical Software 61: 1–24. https://doi.org/10.18637/jss.v061.i04
2. Barcaroli, G., Ballin, M., Odendaal, H., Pagliuca, D., Willighagen, E. and Zardetto, D. 2020.
SamplingStrata: Optimal Stratification of Sampling Frames for Multipurpose Sampling
Surveys. R package version 1.5-2, https://CRAN.R-project.org/package=SamplingStrata.
[Accessed 2 July 2022]
3. Pierre, J., Clough, P. and Debski, I. 2021. Making money and saving seabirds: An exploratory
economic analysis of seabird bycatch reduction. Tenth Meeting of the Seabird Bycatch
Working Group SBWG10 Inf 15. Agreement on the Conservation of Albatross and Petrels,
Virtual meeting, 17 - 19 August 2021.
4. R Core Team. 2021. R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. Available at: https://www.R-project.org/
[Accessed 15 April 2022]
5. Vidal, T., Wichman, M.-O-T.-A., Hamer, P., Pilling, G and the PNAO. 2021. Effort creep within
the WCPO purse seine fishery. WCPFC-SC17-2021/MI-IP-06. 17th Regular Session of the
Scientific Committee, Online meeting, 11 – 19 August, 2021. Western and Central Pacific
Fisheries Commission.
6. Willighagen, E. and Ballings, M. 2022. genalg: R based genetic algorithm for binary and
floating point chromosomes. Version 0.2.1. Available at: https://cran.r-
project.org/web/packages/genalg/genalg.pdf [Accessed 15 April 2022]
Literature sources for zero-catch set rate estimates:
• Oceanic whitetip sharks:
Tremblay-Boyer, L. and Neubauer, P. 2019. Historical catch reconstruction and CPUE
standardization for the stock assessment of oceanic whitetip shark in the western and
central Pacific Ocean. WCPFC-SC15-2019/SA-IP-17. 15th Regular Session of the Scientific
Committee, Pohnpei, Federated States of Micronesia. 12 – 20 August 2019. Western and
Central Pacific Fisheries Commission.
Walsh, W.A. and Clarke, S.C. 2011. Analyses of Catch Data for Oceanic Whitetip and Silky
Sharks Reported by Fishery Observers in the Hawaii-based Longline Fishery in 1995−2010.
Pacific Islands Fisheries Science Center, National Marine Fisheries Service, (NOAA),
Honolulu.
• Silky sharks:
Common Oceans (ABNJ) Tuna Project. 2018. Pacific-wide silky shark (Carcharhinus
falciformis) stock status assessment. WCPFC-SC14-2018/SA-WP-08.
• Porbeagle:
Common Oceans (ABNJ) Tuna Project. 2017. Southern Hemisphere porbeagle shark (Lamna
nasus) stock status assessment. WCPFC-SC13-2017/SA-WP-12 (rev. 1). 13th Regular
Session of the Scientific Committee, Rarotonga, Cook Islands. 9 – 17 August 2017. Western
and Central Pacific Fisheries Commission.
Francis, M.P., Clarke, S.C., Griggs, L.H. and Hoyle, S.D. 2015. Indicator based analysis of the
status of New Zealand blue, mako and porbeagle sharks. WCPFC-SC11-2015/ EB-IP-12. 11th
Regular Session of the Scientific Committee, Pohnpei, Federated States of Micronesia. 5 – 13
August 2015. Western and Central Pacific Fisheries Commission.
• Turtles:
Barlow, P.F. and Berkson, J. 2012. Evaluating methods for estimating rare events with zero-
heavy data: a simulation model estimating sea turtle bycatch in the pelagic longline fishery.
Fisheries Bulletin 110: 344–360.
77
Molony, B. 2007. Overview of purse-seine and longline bycatch issues in the western and
central Pacific Ocean. Paper prepared for the Inaugural Meeting of the Asia and Pacific
Islands Bycatch Consortium Honolulu, 15–16 February 2007.
• Seabirds:
Gilman, E., Chaloupka, M., Wiedoff, B. and Willson, J. 2014. Mitigating seabird bycatch during
hauling by pelagic longline vessels. PLoS ONE 9(1): e84499. doi:
10.1371/journal.pone.0084499.
Lawrence, E., Giannini, F., Bensley, N. and Crombie, J. 2009. Estimation of seabird bycatch
rates in the Eastern Tuna and Billfish Fishery. CCSBT-ERS/1203/Info14. Eighth Meeting of
the Ecologically Related Species Working Group, Busan, Republic of Korea. 1 – 3 September
2009. Commission for the Conservation of Southern Bluefin Tuna.
• Marine mammals, whale sharks:
WCPFC 2019 observer information, purse seine fishery (Version 27 July 2021; Available
here: https://www.wcpfc.int/scientificdatadissemination)
