I have been working on the side-lines of EM in the Pacific and writing about it for a while now… and I say EM (Electronic Monitoring) as it is more than just a camera like the ones in the dash of a car… is a combination of cameras, computers, GPS and gear sensors on vessels, and a lot more of systems inland as to provide data to help managers, scientists and vessel owners effectively manage fisheries.
With the suppression of observer coverage in the WCPO and politics in NZ, a lot is being written about EM internationally and nationally.
this is the set up for 6 vessels that take 2-3 weeks trips each… it take in avg a 5 to 7 days to “observe” a full trip.
The cameras onboard are just the visible tip of the iceberg when it comes to EM, this blog is just to explore some of the other key infrastructure and services that are needed for cameras/sensors on board to work as a solution. As with other technologies, I fear that the hyping in the current media discourse pinning them as the solution to multifaceted seafood value chain problems (from IUU fishing, birds/sea mammals conservation, labour issues and maritime crime) risks hyperinflating expectations on what his EM can offer, with potential operators then walking away because it does not deliver on the hype built around it.
Don't get me wrong EM is a great tool, is the next step in a road started by ER (electronic reporting) less than a decade ago and it would be ubiquitous in the next decade. Yet, to rush EM into the vessels at the present with all the logistic challenges we have is pushing the advances made by EM so far into the risk of failure and losing a lot of the good ground gained so far.
EMonitoring should be seen as what it is, one of a range of tools that can be used to monitor commercial fishing activities, in addition to onboard observers, port state measures, CDS, E-Reporting and the lot. A tool box full of different tools helps governments collect a wider range of data to help make informed decisions and combat illegal, unregulated and unreported (IUU) fishing.
What are we talking about?
A camera e-monitoring system usually consists of wide-angle digital cameras, a GPS receiver, gear usage sensors, storage and processing devices, and a display screen. Sensors can collect data about boat locations and when fishing gear is being used. Cameras record imagery that can be analysed for determining the effort (number of sets, hooks or baskets), the total catch (species, approximate length, life status and fate) and other noticeable events (e.g. transhipments, crew behaviours). They are “closed systems” that do not allow for manual input or changes to data that is stored.
image by Archipelago Marine Research Ltd
Sensor data can be sent in real-time via satellite to national Review Centres, and to regional databases, yet normally footage from cameras is stored on a removable storage device that are provided to a Review Centre once the vessel returns to port. There is flexibility around whether analysis is conducted at the national level and then provided to regional agencies or vice versa.
image by FFA/WWF
A feedback report can be sent to the captain of the vessel, to ensure that they keep the systems maintained with cameras and sensors operating effectively. A trip report is sent to the fishery manager and compliance officers to alert them of any issues.
Where are we in the WCPO terms of EM?
Trials of EM systems for monitoring the activities of longline fishing vessels licenced to operate in the Pacific Island Countries and Territories began in 2014. There are now 73 longline vessels equipped with EM systems. The table below summarises the number of longline vessels per country equipped with an E- Monitoring System, the number of EM analysts trained and working, the number of review computers available and the number of analysed sets received in the EM database at SPC.
WCPFC-SC16-2020/ST-IP-08
The EM records are mainly analysed by national fisheries authorities and sometimes by EM providers. Some of the resulting EM data (vessel, catch, set, haul and location details) have been sent to SPC and are available for its member countries to query and report on using the online tool DORADO.
Currently for FIMS data is being received from a provider to the FIMS EM module. The EM data received is then displayed in FIMS and available for review. At this time the data does not relate to the video footage.
Standardised calibration and use of digital measuring tools are needed towards being able to use EM data (species and length) in stock assessments. These needs can be addressed through the establishment of an EM certification programme detailed further below
The EM application in the region focuses on identifying many activities. Cameras may identify interactions with bycatch species, and are especially useful when recording bycatches of protected species. The viewed data can also provide a secondary source of data, for example, to validate catch and bycatch logsheets.
The goal of EM is to provide a cost-effective monitoring solution capable of collecting data for scientific, management, and compliance purposes. Surveillance cameras installed on vessels have proven to be effective at recording crew and fishing activities, which can be checked for compliance with fisheries regulations. As such, camera monitoring that is integrated with Electronic fisheries information solutions (EFIS) can provide a useful means of validating vessel catch and gear reporting (hook numbers, use of wire tracers on longliners or FAD deployment and setting on FADs during the prohibition period).
The sensor data is especially useful in identifying steaming, setting and hauling. Sensors may also be inserted into hooks, or added to brails to weigh the fish on transhipment.
Cameras can also be an effective tool to monitor Health and Safety issues. A skipper can monitor crew activity from the deck. It may also be possible to detect any possible threat to observers, noting the loss of life of three PNG observers in the past four years.
EM has certain limitations and cannot replace the roles of other data gathering techniques entirely. For example, current camera technology does not offer an affordable and reliable means of identifying the sex, age and species composition of a catch sample. Some observations are difficult to make from camera footage/images, even with human input and specialist knowledge. Furthermore, camera setups (unless extremely elaborate) will generally have ‘blind-spots,' where the crew could discard bycatch, hide interactions with TEP species, or perform other IUU activities which could then go unreported.
As a result, while some see camera monitoring as a viable alternative to onboard human observers, others feel that on-board observers will remain necessary for the foreseeable future, at minimum to perform biological sampling and compliance monitoring where cameras are insufficient. The reality is that EM systems can complement the role of human observers and enhance overall observer coverage most particularly in the longline fishery.
Comparability of data EM vs Human
Lot of work has been one in the region by comparing the data obtained from the cameras with the data from traditional observers working in the same boat, and as my friend Malo Hosken (the SPC researcher on this topic) said: very few differences were noted between the overall species composition of the main tuna species between the two observer types indicating that the E-Monitoring footage is providing sufficient information for identification to the species level. There are several examples where it is evident that the species identification and catch coverage from the E-Monitoring video analysis are better than the on-board observer data.
One benefit of E-Monitoring is that it provides a means of reviewing footage of the video repetitively and by a number of people (e.g. including experts in species identification). With more time, further review of the data compiled from these trials could be undertaken to resolve differences between the on-board observer's record and the original E-Monitoring video analysis record to determine where the problem lies.
Furthermore, substantial advances are being made at the present regarding software tools that cannot only identify length but as well species, opening the door for fully automated analysis of video data. In these cases, the human observer analysing the footage comes only to resolve the events that the software hasn't. This technology has the potential of speeding up the analysis massively, allowing verified to enter the databases in near real-time at some stage.
EM service providers have been researching and developing software to use Artificial Intelligence (AI) solutions to aid in the analysis of EM records. Given that it is expected that more longline vessels will be equipped with EM systems, the need for efficient analysis of a growing volume of EM records means the use of AI software will be essential. However, for EM service providers to develop AI software which meet members’ needs, a large volume of EM records and corresponding EM data are needed. For example, if AI is going to be used in aiding the identification of species, at least 5000 images for each species are needed (at minimum).
These EM records and EM data are currently available and belong to the respective members who have produced them. There needs to be a discussion on how, collectively, members can benefit from ensuring EM service providers gain the records and data they need to pursue AI developments (which will result in proprietary software).
The Nature Conservancy (TNC) has launched an open library of training data to accelerate the implementation of AI in fisheries. For further details please go to: https://www.fishnet.ai/home
In early 2020, one EM service provider and one member country began trials of a Machine Learning Server. Equipped with an Artificial Intelligence algorithm, the Machine Learning Server points the EM analyst towards the start and end of setting and hauling operations as well as when a catch event occurs. This removes the need for the EM analyst to review video records where no activities are taking place, in turn augmenting the analysis rate.
A further challenge is environmental variability, which can make visual recognition difficult. Conditions at sea are particularly challenging – variable light, splash and spray, different vessel and background configurations. Similarly, fish may be in different orientations or stacked on top of each other – presenting additional challenges for accurate classification. This may make the use of AI more likely on belts and chutes as the technology is rolled out.
Finally, the market for EM systems is currently small, which makes it difficult for an EM vendor to make a large investment in research and development when AI may only have applicability to a small number of systems. And in cases when someone else is doing the video review (e.g., in a government center), the vendors have a little financial incentive to invest in AI development.
Despite these challenges, there is a lot of ongoing effort to advance AI for EM video analysis. While the hope is that it will eventually enable complete capture of species, length, and volume data, the next steps are going to advance in AI-assisted review (e.g., identifying key events).
This all exciting stuff but it comes at a huge cost, since a lot of this technology is proprietary and system providers sell you the systems and the software but then you have to pay for updates, or they sell you the subscription… and you tied up to a continuous budget for it that can change… is not so easy
Footage delivery logistics
Live video footage through satellite transmission is not cost-effective at present, so footage is stored on a hard drive and sent monthly, or after each trip, to the provider for analysis. The Marine Instruments e-eye system does provide for an integrated Iridium modem, which allows for real-time data transfer. This also allows for less HD space needed for photos and longer periods at sea.
While there is provision for this facility in most systems, the cost of transmitting still frames, as opposed to live footage is still prohibitively expensive, and quite impractical because of the high volume required for transmission.
Due to the size requirements of video footage, large on-board data storage facilities are required. Cameras film at 5 frames per second (24-30fps is movie standard), and use between 60 and 100 MB per hour of footage. A four-camera setup requires 240 to 400 MB per hour, which results in around 6 to 10 GB for each full day of recording. 1280x720 @ 24FPS on board (HD quality) and minimum recording capacity of the system is 13-14 weeks.
Due to these dataset sizes, video surveillance footage cannot feasibly be sent in real-time via a satellite feed. Instead, it is usually transferred directly from the hard drive after retrieval. The HDDs are changed very easily onboard and can be examined at the control centre. Videos are stored onboard and encrypted. Videos are extracted locally from the encrypted HDD for analysis ashore by the owner or the Observer Program.
The process of hard drive data retrieval and footage review and analysis is currently relatively slow, taking a few days for dispatch, forensic recording, and then data viewing (~ 3-6 hours) depending on the recording requirements.
Since hard drive exchange is the most trailed option, it is the primary model being considered for video retrieval for current EM development efforts any form of EM implementation can expect complex logistics associated with this option due to the many ports across multiple countries with varying degrees of infrastructure visited by thousands of participating vessels.
For the EM system to work effectively, vessel owners and operators who submit the hard drive should have confidence that the data seen by the reviewer accurately reflects what was recorded on the vessel, as confirmed by a reliable chain of custody and supported by data encryption.
In a place like NZ, this could be worked out, vessels could turn over hard drives to a dedicated collector stationed at each major port. These collectors could be trained staff affiliated with MPI or a 3rd party observer program.
The collector should be able to meet vessels, swap hard drives, and upload video in a local office to transmit via Wi-Fi or physically pass off the hard drives to the appropriate centralized video review office. A locker system could also be used so that pickups and drop-offs would not require a person to be available the moment a vessel comes into port.
Hard drive retrieval methods do not limit the risks associated with mailing physical hard drives from remote ports with limited infrastructure.
There is a challenge for FFA or WCPFC secretariat to determine whether hard drives can be sent through certified mail services or via courier throughout the entire region. But even for papers this present challenge in the WCPFC, in COVID times without flights, this option is not available for most ports
Furthermore, vessels fishing in the HS that may not come to port for periods longer than a year would depend on the various carriers at sea to interchange HD, which is not realistic at the present.
Footage review
The time it takes to analyse footage is dependent on various variables. A present estimate (with the present level; of AI/Machine Learning capabilities is 1:15 (1 hour takes 15 mins to view/analyse), assuming the analyst is someone that knows how to operate the proprietary software and has a good understanding of the licence conditions, fish and species of special interest ID and fishing operations. Reality is that It takes time even in the best of cases…
At the present in the pacific, we have reviewers from third-party vendor and government fishery agencies.
Government Fisheries Agency video review has been undertaken by the fisheries with the support of external funding. This option can be attractive to states that prefer to maintain control of the video footage from fishing in their waters, and it can generate jobs for the local economy. This review option may also be attractive for at-sea observers who would be strong candidates to become EM analysts ashore in a review centre, although the jobs are quite different.
A data review centre is a complex operational system. Establishing such a centre will require countries to increase budgets, purchase review stations, hire and train staff, to review a huge volume of video particularly if it is centralised
If not centralised, there are inefficiencies to having different ports or regions set up their own review centre instead of consolidating the review process. Experience from pilots in the pacific has demonstrated the challenges of building efficient in-country review capacity, with cases of long backlogs, and difficulty getting local reviewers to meet the same level of review efficiency as third-party vendors. EM pilots have also shown a wide variation in the performance of national review centres, and this type of variability across a state or worst across RFMO member states could be detrimental to an EM program.
The draft FFA Longline EM Policy commits the FFA Secretariat, SPC, and Parties to the Nauru Agreement (PNA) to collaborate to provide technical support and training to EM review centres, a signal that there is a recognition of the challenge of building and operating these centres and that many countries will need assistance. One possibility to mitigate the cited difficulties is to create sub-regional review centres.
Third-Party Review – Other countries contract the service through a third party. This could be a commercial EM vendor or a quasi-governmental agency. Under this model, the government or industry contracts with a third-party entity to review the video and deliver processed data that meets the set of specified standards (e.g., data fields, quality, timing, review rates) (we have that at the WCPFC). This can be a more efficient way to handle the review of data as the government can act solely as a contract manager instead of building its own capacity to review EM video from scratch. If local jobs are an important concern, it may be possible to structure contracts to require the hiring of in-country reviewers. For low wage countries, the third-party approach could be more expensive than analysing the data themselves, but the start-up costs and inefficiency of building a review centre from scratch could limit the cost savings
Costs
To date, most of the costs for EM programs in tuna fisheries have been paid by NGOs and international organizations, but this model will not continue forever. Currently, much of the enthusiasm by coastal states for EM is related to the idea that in the future, industry will be responsible for paying most, or all, of the costs. The draft Regional Longline Fisheries Electronic Monitoring Policy formulated by FFA member countries states as a guiding principle: “User pays - full cost recovery as a default.” Many segments of the fishing industry feel that costs could be high and are also uncertain about how an EM program will affect their business. As the group that will be most impacted, they may believe that it is unfair for them to be entirely responsible for funding an EM program. This difference of opinion on who should pay for EM is seen by many as the most significant impasse for EM implementation.
Many reports on EM state that a major advantage of EM is its cost relative to human observers. However, human observers can have a cost advantage where observer wages are low (e.g., in Pacific Island countries national observer programs).
The costs of EM are also quite clear and immediate, while the benefits of EM can be more uncertain and diffuse. For example, there are costs of inadequate knowledge of fishing activity (e.g., depleted fisheries) but those costs are not very clear, especially in fisheries with limited data.
EM costs fall into four categories:
Type 1: On vessel costs. These costs are associated with the installation and operation of EM hardware and supporting systems on board fishing vessels.
Type 2: Program administration and operational costs. These costs are associated with the administration and operation of the EM program, usually undertaken by national (or regional) fisheries administrations. These costs typically form the ‘core’ of the annual EM program budget, and would be a main focus for cost recovery.
Type 3: Policy and regulatory development costs. These costs are associated with the establishment of relevant regulatory and policy arrangements to support effective EM systems.
Type 4: Analytical costs. These costs are associated with the analysis of EM generated information to produce outputs in support of the administration and management of fisheries by national fisheries administrations (e.g., production of reports analyzing annual trends in EM information).
For example: the EM pilot project funded by FAO for the 50 tuna longliners based in Fiji operated for three years (2015- 2018). The total fixed and variable costs for the three years were US$986,575, which came out to US$6,577 per vessel per year. The pilot was able to review approximately only 44 % of the EM trips over the pilot’s duration.
Servicing EM Hardware Systems
The experience from EM pilots and fully implemented programs has contributed to improved reliability of EM hardware. Nevertheless, EM hardware may break or malfunction, so it is essential to have a good servicing plan in place that clearly articulates responsibilities and minimum levels of service.
In the EM pilots, major repairs were often undertaken by EM technicians who needed to fly to the repair site – a time consuming and expensive approach. Even when only a phone consultation was needed, time zone differences between the vessel location and the EM vendor could result in significant delays.
With these challenges in mind, EM programs will need to set clear service expectations with the EM vendor, and ensure that maintenance issues are promptly addressed. EM vendors can use a variety of approaches to meet these requirements, such as training local capacity (e.g., marine electronic technicians), engineers, or deck crew to undertake common repairs, placing EM technicians in-region, or providing 24/7 remote consultation.
Given the length and remoteness of many fishing trips in the region, empowering captains and crew to make EM repairs at sea will be important. This will include supplying spare parts, providing additional training on certain maintenance functions, and establishing communication protocols for technical support at sea.
Equally important to creating a servicing plan is setting the obligations for vessels that experience equipment malfunctions. Vessel operators should be required to perform basic functions to keep EM hardware in working order at sea, such as ensuring that the lenses remain clean and that camera views remain unobstructed. If a system malfunction, operators should be required to report it immediately to management authorities. The vessel would then need to follow an agreed procedure when the EM system malfunctions.
The strictest option would be to require vessels to return to port immediately. This would be a strong deterrent for intentional sabotage of EM systems, but could also place significant economic burdens on the fleet, and could potentially be a disincentive for fleets and flag states to participate in an EM program.
Other options could include a risk-based approach, where vessels that have demonstrated strong compliance may be allowed to continue fishing, while less compliant vessels would be required to return to port.
A limited number of exemptions for malfunctions may also be considered (e.g., a vessel can keep fishing for the first malfunction in a calendar year).
With some vessels taking trips several months in length and far away from port, the decision of how vessels need to respond to a malfunctioning EM system is likely to have a larger impact in the WCPO than in other EM programs in which trips are typically shorter and closer to shore.
EM Policy
The importance of policy cannot be underestimated, as you need to compare apples with apples, be able to define for what purpose the data is going to be used. Know the limitations you have with it, particularly if you going used for compliance! Do the fisheries law have the needed strength to deal with this type of e evidence? Do we need to think what ty[pe of prosecutable offences may not be covered by EM? I imagine based on some of the very hardcore lawyers I know from my fishing days, that they could drag prosecution for months based on technical issues if there was to be some crystallized sea salt in the lens particularly when it comes to undersize fish… and so on.
In October 2019, a regional workshop was held at FFA to draft the Regional Longline Fisheries Electronic Monitoring Policy. The policy was drafted with inputs from members, secretariats, industry and NGOs. The draft policy was presented at the 2020 MCS WG and FFC meetings and was adopted on 19 June 2020. The purpose of this EM Policy is to describe a regional framework that, inter alia:
supports collective action at a strategic level;
promotes a level playing field in relation to the implementation of EM and mitigates against market distortions; and
facilitates economies of scale for national and regional benefit.
EM minimum data field standards and training of analysts/observers
The adoption by WCPFC of ER and EM standards is expected to support and accommodate those CCMs that have commenced implementation of a range of EM and ER technologies in their fisheries and will ensure that the Commission's databases and systems are ready to exchange electronic data in an orderly and efficient manner.
The SPC/FFA/PNAO Data Collection Committee met in February 2020 to revise the draft DCC Longline EM minimum data fields standards. These standards are proposed for member countries to use when embarking on trials or implementation of E-Monitoring (EM) for longline vessels licensed to operate in members’ EEZs (and adjacent waters). These standards should be provided to the EM technical provider to ensure the minimum data fields specified within are generated from the EM system, according to the EM Protocol notes provided. These standards are in draft format and will be presented to the WCPFC Scientific Committee (SC16-ST-IP-07) and ER and EM Working Group meetings in 2020.
Standards for the training, assessment and certification of EM Analysts and observers using E-Reporting tools
Units of competencies for EM analysts and observers using ER tools were presented to the PIRFO Certification Management Committee in 2018 and subsequently adopted by FFC in 2018. To build on these, Members have identified that there is a need to develop curricula to guide the training, assessment and certification of EM Analysts (staff responsible for analysing EM records) and of observers using ER tools. Building from the success of the Pacific Islands Regional Fisheries Observer (PIRFO) competency development programme, SPC and FFA members support the concept that a training and development programme for EM Analysts and observers using ER tools be established within PIRFO programme.
Strengths, Challenges, and Opportunities for EM Tuna Fisheries
There have been numerous trials and fully implemented EM programs in the region. These trials have covered both longline and purse seine fisheries. From these trials, some general conclusions can be reached about the efficacy of EM as a monitoring and compliance tool
Strengths of EM
Provides accurate data on the location and time of fishing activity.
Accurately assesses the set type in purse seine fisheries.
Accurately estimates total catch per set in purse seine fisheries.
Provides good estimates of the catch of main target species in longline and purse seine fisheries.
Identifies most Endangered, Threatened, or Protected (ETP) species interactions.
Incentivizes more accurate reporting of data in logbooks.
Covers multiple views of the vessel at the same time, does not require breaks, and video can be reviewed multiple times.
Is less prone to intimidation, bribery, or interference in order to falsify reported data.
Review of much of the fishing activity can happen at high speed (e.g., >8x speed).
A space efficient solution for longline vessels with limited room for a human observer.
Can sometimes provide cost savings relative to human observers.
Helps document conformity with management measures and international obligations.
Scalable option to implement on various vessels with different gear types.
Challenges of EM
Accurate estimates of non-target species in purse seine and longline fisheries can be challenging with EM depending on catch-handling techniques and camera placement
Identification of ETP species may only be accurate at higher taxonomic levels (e.g., shark or turtle, but not at the species level). However, additional or higher resolution cameras may be a solution.
Accurate identification of juvenile tuna (e.g., small yellowfin and bigeye) is difficult, although this is similarly difficult for human observers.
EM systems are not linked to fish aggregating device (FAD) buoy identification systems.
EM is not currently suitable for biological data collection (e.g., sex identification, otolith measurement), which could be addressed by complementing EM with dockside sample collection.
EM cannot be used to accurately assess the condition or life status of fish.
Hard Drive Collection logistics are very complex and will hamper the application on HS vessels.
Fisheries Legislation may need to be changed as to incorporate new forms of electronic evidence and even types if fisheries offences where EM evidence can be used.
Emerging Opportunities of EM
EM has been explored to monitor labour practices onboard vessels, but no results of comprehensive studies have been published in the literature.
EM is being explored as a tool to monitor transhipments.
In general, it is easier to extract detailed information about catch in longline fisheries where the catch is brought on board one fish at a time, but EM has proven successful in purse seine fisheries as well. In some cases where EM has not been able to match reported data from human observers (e.g., species-level identification of non-target catch), changes in catch handling procedures or additional camera views may be able to overcome these challenges.
Many of the design choices are operational or technical in nature (e.g., what hardware should be selected, how much footage should be reviewed), but perhaps the most important lesson from the development of other programs is that “[getting an EM program adopted] is a people challenge not a technical challenge.
The reality is that, like most new solutions, there will be opposition along the way. Industry members will likely have concerns about additional monitoring, privacy, the costs of the program, and a general fear of the unknown. Fisheries managers may be concerned about the cost and complexity of the program and whether they will be able to effectively manage it. These concerns are legitimate and have emerged in the development of most, if not every, EM program. These issues cannot be sidestepped, and stakeholders need to be integrated into the design process so that their concerns are recognized and addressed. In particular, the industry needs to be involved as they will be the ones most impacted and their acceptance of the program will be critical to its success.
There are real challenges to developing an EM program, and the characteristics of particular fisheries can make this a bit more complex, but these are solvable challenges. Yet, to rush EM into the vessels at the present with all the logistic challenges we have is pushing the advances made by EM into the risk of failure.
References:
Roadmap for Electronic Monitoring in RFMOs by Mark Michelin, Nicole M. Sarto, Robert Gillett
Analysis of the costs and benefits of electronic tracking, monitoring and reporting systems applied in FFA countries and identification of the required legislative, regulatory and policy supporting requirements. Banks R, Muldoon G and Fernandes V.