Electronic monitoring in fisheries: Lessons from global experiences and future opportunities

Since the beginning of the 21st century, electronic monitoring (EM) has emerged as a cost‐efficient supplement to existing catch monitoring programmes in fisheries. An EM system consists of various activity sensors and cameras positioned on vessels to remotely record fishing activity and catches. The first objective of this review was to describe the state of play of EM in fisheries worldwide and to present the insights gained on this technology based on 100 EM trials and 12 fully implemented pro‐ grammes. Despite its advantages, and its global use for monitoring, progresses in implementation in some important fishing regions are slow. Within this context, the second objective was to discuss more specifically the European experiences gained through 16 trials. Findings show that the three major benefits of EM were as follows: representative coverage of the fleet any the enhanced registration of fishing activ‐ ity and location. Electronic monitoring can incentivize better compliance and discard but the fishing managers and industry are often reluctant to its uptake. EM ability, sustainability claims and market access, implementation on a larger scale. conclusion, EM as a monitoring tool embodies various solid strengths that are not diminished by its weaknesses. Electronic monitoring has the opportunity to be a powerful tool in the future monitoring of fisheries, when inte‐ grated within existing monitoring programmes.


| INTRODUC TI ON
Historically, fishing has largely been an unregulated industry, with fishers operating as independent explorers of the sea (Johnsen, Holm, Sinclair, & Bavington, 2009;Stevenson & Oxman, 1974). It was primarily governed by affective relations, often in local fishing communities (Johnsen et al., 2009). However, over the course of the 20th century, awareness of the impact of fishing on marine resources has grown, resulting in an increase in rules and regulations (Botsford, Castilla, & Peterson, 1997;Johnsen et al., 2009). Fisheries-dependent data collection has also increased, as more data are needed to assess fish stocks, and to monitor and regulate the environmental impact of fishing.
The value of fishery-dependent information in estimating the status of fish populations has regularly been called into question (Cotter & Pilling, 2007). Information may be biased because fisheries do not randomly sample fish populations and because fishing methods vary from place to place and time to time. Furthermore, landings do not provide information about all fish that are caught, since catch that is discarded at sea can represent a large proportion of the total catch (Borges, Zuur, Rogan, & Officer, 2004;Fernandes et al., 2011;Poos et al., 2013;Uhlmann et al., 2014;Ulleweit, Stransky, & Panten, 2010). Finally, misreporting may occur when fishers under-report problematic interactions with by-catch and quota-limited or "choke" species (Borges, 2015).
Despite the rapid increase in availability of new technology, such as GPS, network communication, digital cameras and image analysis software, the implementation of these innovations to monitor fisheries catches at sea has not evolved much. For instance, the vast majority of discard estimates are based on expensive fisheries observer programmes, and are associated with low coverage, often less than 1% of the fishing activities (Benoît & Allard, 2009;Depestele et al., 2011;Poos et al., 2013;Rochet, Péronnet, & Trenkel, 2002), often using subsamples of catches where fish are measured one by one on a measuring board and recorded with pencil and paper. Only within the last two decades, electronic monitoring (EM) has emerged as an additional approach for documenting catches in fisheries (Ames, Leaman, & Ames, 2007;Kindt-Larsen, Kirkegaard, & Dalskov, 2011;McElderry, Beck, & Anderson, 2011;Stanley, McElderry, Mawani, & Koolman, 2011). While the initial development of EM systems was largely an industry-led process to cope with management reforms and gear theft in the British Columbia crab fishery (Ames, 2005), it was quickly recognized that EM could also be used for monitoring and control in fisheries challenged by poor coverage by at-sea observations (McElderry, Schrader, & Illingworth, 2003). Electronic monitoring systems generally consist of various activity sensors, GPS, computer hardware and cameras (Figure 1) which allow for video monitoring and documentation of catches and detailed fishing effort estimation without requiring additional on-board personnel, unless additional biological data, for example otoliths, are needed (e.g. Needle et al., 2015;Ulrich et al., 2015). The data recorded can be reviewed at a later stage to obtain catch information, for example species composition, numbers, volume and lengths. In North America, the first EM trial was implemented in the Area "A" crab fishery in 1999 in British Columbia, Canada, to monitor vessel trap limits and to control catch and gear theft. As a result, the fisheries authorities implemented a full EM programme involving 50 vessels with a 36,000 fleet-wide trap limit. Subsequently, in 2002 EM was tested in the Alaskan longline fisheries to register catch and effort in the Pacific halibut (Hippoglossus stenolepis, Pleuronectidae) fishery and to test for compliance with regulations on seabird catch mitigation devices (Ames, Williams, & Fitzgerald, 2005;. In 2006, one of the largest EM programmes was introduced in the groundfish hook and line and trap fishery in British Colombia, Canada, to monitor compliance with self-reporting responsibilities on about 200 vessels.
In New Zealand, an EM programme was started to monitor marine mammals' and seabirds' interactions in gill net and trawl fisheries in (McElderry, McCullough, Schrader, & Illingworth, 2007.
In 2005, EM trials started in Australian waters, monitoring fish handling and by-catch mitigation measures in several fisheries. Since 2012, EM has been tested in tropical tuna fisheries in the Atlantic and Indian Ocean, and during the same period, EM technology was introduced in trials on similar fisheries in the Western and Central Pacific Ocean with the aim to enhance sampling coverage of observer programmes for these vast fishing grounds.
European EM trials started in 2008, with the rising awareness of the vicious circle in which North Sea demersal fisheries were trapped (Rijnsdorp, Daan, Dekker, Poos, & Densen, 2007). A recovery plan for Atlantic cod (Gadus morhua, Gadidae) in the region had evolved into a complex and micromanaged regulation with multiple gear categories and exemptions (Kraak et al., 2013;Ulrich et al., 2012).
Eventually, this resulted in the establishment of a new cod plan that included severe effort reductions. Several EU member states tried to incentivize cod discard reductions by making volunteer fishers accountable for their total catches rather than for their landings, in exchange for increased quota shares and, in some cases, exemptions from the effort reductions . Consequently, several EM trials were funded in order to verify declared catches, also known as "Fully Documented Fisheries" (FDF).
Electronic monitoring seems to be a good candidate for full catch documentation. However, in spite of the obvious advantages of EM, European managers have so far remained reluctant to use it because of its unpopularity among fishers. The fishers consider EM an intrusion in their private workspace (Baker, Harten, Batty, & McElderry, 2013;Plet-Hansen et al., 2017) and argue that camera surveillance reflects a governmental mistrust against them (Mangi, Dolder, Catchpole, Rodmell, & Rozarieux, 2013). This paper aimed to review the current status of EM worldwide and to discuss whether EM is a viable monitoring tool for fisheries. In addition, we summarize experiences with EM trials in northern Europe, where uptake of EM in monitoring programmes is slow, and compare them with experiences worldwide.

| RE SULTS
The comprehensive review collected information on 100 EM trials and 12 fully implemented EM programmes worldwide (Tables   1 and 2 (Table 1). Below, we summarize the findings of the review for different areas and fisheries.

| North America
The majority of fully implemented comprehensive EM programmes, 9 out of 12 (75%) worldwide, run in both Canada and the United States (  (James et al., 2019;Ruiz et al., 2015). The trials showed that EM was a promising tool to replace or to supplement current observer programmes (Briand et al., 2017;Ruiz et al., 2016

| Australia and New Zealand
In 2015

| South and Central America
In total, three EM studies were conducted in South and Central EM systems for this region (www.world wildl ife.org). Such low-cost EM systems will help address the more challenging but globally significant fishing regions, for example Asia and Southern Europe (Michelin, Elliott, Bucher, Zimring, & Sweeney, 2018). For example, a very basic low-cost EM application, just using a camera mounted on a small fishing vessel and video recording the complete fishing trip, also proved to be successful in other regions, for example monitoring protected species interactions in the Indonesian hand-line fishery (Kennelly & Borges, 2018). Along the development of low budget, the Chilean government is in the process of implementing EM in a fleet-wide programme to monitor compliance as part of the "by-catch law and mitigation plans" (Cocas, 2019).

| Europe
In total, 23 published studies describing 16 different trials from 6 different nations (Scotland, England, Denmark, the Netherlands, Germany and Sweden) were reviewed (Table 3). Trials were mainly conducted in demersal fisheries using active gears (trawls and seines), although some passive gears (gill net and longline) have also been monitored. Different types of vessels have been involved, from larger beam trawlers and seiners to small-scale fisheries with vessels less than 10 m in length. The trials often lasted several years and generated large amounts of data. The first trials started in Sweden, Denmark and Scotland in 2008, and a spin-off of the Scottish trial was still ongoing at the time of writing. The number of vessels participating in each trial varied between 1 and 27 vessels. Evaluating the usefulness of EM as a monitoring tool was the most common research objective among the studies and countries, with 17 out of 23 (74%) studies sharing this objective (Table 3). In 7 (30%) cases, this objective was combined with an evaluation and feasibility study of a catch quota management (CQM) regime or landing obligation. Other studies' objectives focused on EM as an alternative method for, for example, scientific data collection, testing increased flexibility in technical fisheries measures, monitoring by-catches, analyses of high grading or estimation of discards. One study investigated the possibilities to use computer vision technology to automate the process of data collection in EM (French, Fisher, Mackiewicz, & Needle, 2015). Even though several studies briefly

| Review of European EM operations
In the period 2008-2016, results of European EM trials were reported in a manner that allowed a detailed review of EM on an operational level. The trials were summarized and compared for efficiency for EM set-up and data flow, EM analyses, EM performance and EM costs. In addition, levels of acceptance and objective for the trials were described.

| EM set-ups and data flow
In all trials, the EM system set-up consisted of (a) a GPS recorder supplying information on vessel location, (b) cameras supplying visual information on fishing activities and catches, and (c) hydraulic and drum-rotation sensors to mark deployment and retraction of gears.
All data are conveyed into a computer, which saves the information ( Figure 1). Vessels in all trials were initially equipped with the technology developed by the Canadian company Archipelago Marine Research (www.archi pelago.ca). This system uses hard discs to store sensor data, geographical location and video recording. These hard discs were replaced manually before reaching data storage limits. The Danish and German trials switched to another provider that allowed the transmission of data using 4G cellular networks (www.ancho rlab.dk).
In all trials, the cameras were usually installed in a way that crew workflow was minimally affected. The number of cameras deployed depended on the size and the specific characteristics of the vessels.
The layout and selection of camera models and settings was the result of an optimization between quality and data storage requirement. The number of cameras, their field of view, the resolution (pixel density) and the frame rates were considered against the spe- Bryan (2015) Develop a methodology to use EM to confirm full retention of catch on-board a freezer trawl vessel (compliance with discard ban). ( Figure 5). The general systems among the reviewed trials had at least one camera pointed directly at the discard chute and sorting belt, one camera to cover the processing area or the deck on smaller vessels, one camera to observe net hauling and one camera to cover the catch in the hoppers. Meanwhile, recent EM systems have been able to store data from up to eight cameras. These additional cameras have been used for larger vessels in Scotland and Denmark to get a better coverage of the vessel and to limit blind spots Needle et al., 2015;Ulrich et al., 2015). On smaller vessels, the sorting areas may be small or absent and positioning the cameras was often challenging. Installing custom mounting infrastructure to improve camera positions was useful in trials on small vessels with open decks (Marine Management Organisation, 2013b; Needle et al., 2015).
Also, the availability of electrical power on small vessels may be limited by battery capacity when the engine is not running, thereby limiting the scope for implementation on some smaller inshore vessels. Meanwhile, autonomous systems have been developed that are powered by solar panels and batteries (Bartholomew et al., 2018).
Cameras can be set to record at different resolutions. For many applications, low resolution may be adequate. In current systems, low-resolution camera feeds are able to record at higher frame rates, which offers a smoother view and allows for the detection of abnormal behaviour in the handling process or when counting fish.
However, using low-resolution images hampers species recognition and measuring fish lengths. High-resolution camera feeds have lower frame rates and use considerably more hard disc space than low-resolution camera feeds. In several studies, for example #10 and #18 in Table 3, the cameras directed at the discard chute or processing area were set to record at maximum resolution. This resulted in high-quality images, but frame rates were limited to 5 frames per after 40 min because all catches in this fishery were processed rapidly and continuous recording was unnecessary (Course et al., 2011).
In all EM set-ups, GPS information was collected with high frequency (generally every 10 s) Ulrich et al., 2015). This is a much higher temporal resolution than the typical 0.5-to 2-hr interval used in the obligatory EU vessel monitoring system (VMS) (Deng et al., 2005;Hintzen et al., 2012;Lee, South, & Jennings, 2010). The high spatial and temporal resolution of GPS position data, combined with the hydraulic and drum-rotation sensors, allows for accurate effort calculation for vessels equipped with EM. This was demonstrated in the study by Needle et al. (2015), pointing out the differences in perceived fishing activity as indicated by either VMS or EM data for a Scottish seine vessel.
The VMS-derived fishing path underestimated the area impacted by the vessel, whereas the true path was accurately recorded by the EM data, showing the characteristic triangular pattern of seine fishing. Similarly, Götz, Oesterwind, and Zimmermann (2015) showed that haul durations indicated in fishing logbooks were imprecise when compared to those estimated using EM information.
In their trial for two vessels, the towing times listed in the logbooks for one vessel were generally longer than the times recorded by

| Data storage
Data collected from the various sensors and cameras are all linked to a central computer, which files the data onto a hard drive. All trials started with EM data being stored on exchangeable hard drives.
Once full, hard drives were replaced by empty drives to continue recording. Drives were usually replaced by authorized persons, for example fisheries inspectors (Götz et al., 2015;Needle et al., 2015) or by staff of the institutes responsible for the projects (Dalskov & Kindt-Larsen, 2009;Kindt-Larsen et al., 2011), although in some cases fishers were instructed to change hard drives themselves (Course et al., 2011;van Helmond et al., 2015). Particularly, in case of compliance monitoring data encryption is provided to ensure data protection in the chain of custody.
To avoid the manual replacement of hard drives,

| Supplementary information
Supplementary catch information, for example logbook, haul-byhaul catch and observer data, was collected in all trials, with the purpose to evaluate and compare the efficacy of EM in a variety of management and scientific objectives. In the case of catch quota management trials for cod, all catches, including undersize individuals, were recorded. During trials in Germany and Denmark, extra information on discards was provided in official electronic logbooks (Götz et al., 2015;Ulrich et al., 2015). In several trials, data from on-board observer programmes were used in comparison with EM data (Marine Management Organisation, 2013b; Needle et al., 2015). In the Netherlands and England, fishers were requested to record catches by species or size category on a haul-by-haul basis (Course et al., 2011; Helmond, Chen, & Poos, 2017).

| EM data analysis
Most of the EM studies have collected thousands of hours of video footage, thus requiring a structured approach for the review and interpretation of sensor and image data. Data analyses have been conducted by video observers, whose training have ranged from small introductory courses and cooperative training  to more formal training courses .
Video observers were often trained at-sea fisheries observers (van Helmond et al., 2015(van Helmond et al., , 2017 or have systematically been trained to recognize species and to operate the EM software. In some trials, they have also been trained in length measurement . This training improved the quality of the video review . recorded catch data in logbooks (Course et al., 2011;van Helmond et al., 2015;Kindt-Larsen et al., 2011;Needle et al., 2015;Ulrich et al., 2015). Attempts to identify all fish and invertebrates discarded from one trip of a Scottish trawler resulted in prohibitively long review times: the trip took 1 week and the analysis took 3 months . This would clearly not be sustainable for ongoing monitoring purposes and budgets.
Different procedures have been used in improving estimates of catches from EM video material in the different trials (Table 4). The first approach required crews to sort discards into baskets ( Figure 6) and show the baskets to the cameras before dis-

| EM performance
Most trials studied the performance of EM as a reliable source of catch information (Table 3). This performance depends on the technical reliability of the EM systems and the ability to correctly estimate catches. Technical EM failures and loss of data due to poor video quality were reported in 11 (out of 15) trials. However, not all technical errors were reported in similar detail. During the review, reported errors were classified in three different categories: system failure, storage failure and obstructed view. Where possible, errors were quantified as a percentage of data loss (Table 5). System failures were recorded in seven trials, with the main reason being broken cameras and non-functional drum-rotation sensors. Two studies (#12 and #22) mentioned system failure caused by power supply issues. Storage failure was recorded in three trials, caused by corrupted EM data, mainly video data, on the exchangeable hard drives.
During the German trial, a hard drive began to burn during the copy process in the Institute and data were lost (Götz et al., 2015).
Another form of storage failure occurred in the Dutch CQM trial; storage failure occurred because full hard drives were not replaced in time. This was not related to a technical failure of the EM system itself, but due to insufficient management of exchanging hard drives when vessels entered ports. A similar situation was described in the German trial where logistical and technical problems were encountered in relation to the exchange of hard drives, when vessels entered distant ports (Götz et al., 2015). Nevertheless, no data losses were reported in this trial because of these situations.
Obstructed view was reported in six trials. In these situations, 35% EM data loss in total, system failure was mentioned as one of the reasons (Dutch CQM trial) 21% data loss in total, system failure was mentioned as one of the main reasons (Dutch sole EM trial) 17% due to failure of cameras, 12% due to rotation sensors, 7% due to control boxes, also insufficient power supply was mentioned (English CQM trial for trawls and gill nets) 2.5%, rotation sensor and camera failure (English EM trial for vessels < 10 m) 0.7% of catch processing set for audit had camera breakdowns or video gaps either rendering the video useless or hampering the audit. trials have been sufficiently reliable to fulfil the goals of the studies, provided there was ongoing attention to maintenance.
All European trials had the objective to evaluate the ability of EM to estimate catches in commercial fisheries (Table 3).
Different methods were used to estimate catch from video footage (Table 4). To test the efficiency of EM, catch estimates based on video review were compared with recordings of fishers and/ or on-board observers. In the Danish and German CQM trials, catch weights were obtained from EM with the use of fishing crews that collected catches in baskets and showed those to the cameras ( Table 4). The Danish CQM trial observed discrepancies between fishers' and video observers' discard estimates that were often less than 5 kg per haul, without systematic bias and with clear improvements of the accuracy over time .
The Scottish, Dutch, German, English and in some years Danish CQM trials estimated catch directly from sorting belt or discard chute (  Ulrich et al., 2015).
The Scottish trial was able to estimate discards with no effective change to the catch processing systems used on each vessel . This was not the case in all trials, and protocols were developed to improve the registration of catches for vessels participating in EM in Denmark and in the Netherlands (van Helmond et al., 2017;Ulrich et al., 2015). Fishers were able to follow the protocols to improve video review, and when mismatches occurred, it has generally been sufficient to point to the issue in order to get the return to full compliance. These protocols substantially increased the accuracy of EM. However, for both trials it was reported that the protocol could be a burden for the crew. For example, the Danish basket system has been criticized by fishers, because it imposes additional work on crews. Moreover, baskets take much space on deck and they are heavy to move. In the Dutch case, the protocol required Also, the use of EM video data to provide length-frequency data is not always straightforward, as it is not always possible to view the full body of each fish due to occlusion by other fish or waste materials . However, a morphometric length inference model for fish of which the full body was not visible on footage was successfully tested in the Scottish trial . Also, developments in automated measurement of fish by computer vision may improve length measurements based on video data even further (French et al., 2015;Huang, Hwang, Romain, & Wallace, 2018;White, Svellingen, & Strachan, 2006). Nevertheless, even fully accurate length measurements would have to be converted into weight using lengthweight relationships rather than being weighed directly on-board, which could contribute to some discrepancies with observer estimates.
In summary, the EM performance depends critically on whether the operating specifications of the technology, the monitoring objectives, the vessel layout and the responsibilities of the vessel personnel in supporting the monitoring effort are considered.

| Cost-efficiency
The price of an EM system per vessel, including installation, in the trials has been around 9-10.000 €, and systems in the trials have  Needle et al. (2015) concluded that, although the initial costs of EM are high, EM is a more cost-effective monitoring method than an on-board observer programme in the mid-to-long term as running costs are much lower, consequently, that would allow for a wider sampling coverage for a given monitoring budget along with truly random sampling. Another important aspect regarding the cost-benefit of EM is the involvement of fishers in reporting their catches. Electronic monitoring is often used to validate self-reported catches or discards. Even though only a minority of these reports are audited with video, the fishers do not know which hauls will be audited and when, which creates an incentive to report all catches accurately. Consequently, even with a low audit rate, observation costs are expected to be largely internalized by fishers (James et al., 2019). It should be noted, however, that these cost analyses were based on EM trials and that we did not encounter cost analyses based on large-scale monitoring programmes. of interviewed EM-experienced fishers expressed positive views on EM. In contrast, fishers without any first-hand experience with EM remain largely negative about it; 90% of the interviewed fishers without EM experience were against it. Whether this division resulted from participating fishers being more in favour of EM prior to trial participation or whether participation in the trial had changed the opinion of the fishers was not studied. The fact that fishers were rewarded to fish with EM in most trials may also have been an influence. In addition, some studies indicated that protocols to improve video review can be a burden on the crew (van Helmond et al., 2017;Ulrich et al., 2015). The success of monitoring the landing obligation with EM likely depends, at least for a large part, on the workload that it imposes on skippers and crews for monitoring and registration of catches. Similar observations were made during the process of EM data review and analysis of Götz et al. (2015) and . However, the development of technologies to improve the implementation and reduce this burden of EM has been ongoing in the Scottish trial (French et al., 2015;Needle et al., 2015).

| EM acceptance
It is noteworthy that the first decisions to use EM in the EU did not come from the fishing industry, but from a strong political will. Based on the results of the first CQM trials in Denmark and fishers in case of differences between logbooks and EM will be counterproductive, a continuous dialogue about these differences may help improve data quality and acceptance of EM as a monitoring tool.
In the context of the adoption of EM in Europe, there is still no obligation for EU Member States to use EM as a verification or monitoring tool. If EM is required in some Members States but not in others, there will be no "level playing field" between European fishers. This concept of a "level playing field" potentially imposes an extra obstacle for the implementation of EM in European fisheries management .
The acceptance of EM will improve if benefits of EM for the fishing industry are greater than just improving compliance (Michelin et al., 2018). Such benefits could include improved data quality through EM, allowing for more efficient management measures and, eventually, improved financial performance for industry, and increased flexibility in regulations as a result of improved accountability from EM.
The Danish trial on free gear selection  is a good example of this, alternative uses for EM data, for example, improved business analytics, such as identifying and avoiding by-catch hotspots, support of (eco-) certifications by increasing traceability in seafood supply chains.

| EM objectives
Of the reviewed studies, 9 studies had the objective to evaluate the efficacy of EM as a monitoring tool (Table 3). Of these 9 studies, 8 concluded that EM is an effective monitoring tool compared with other existing monitoring methods such as at-sea observers, VMS and electronic logbooks (eLogs). One study of the 9 mentioned was not conclusive of the efficiency of EM as a monitoring tool compared with other methods, but indicated that EM delivered an appropriate coverage of fish catches and fishing time.
In addition, EM proved to be a successful tool to test alternative management regimes, for example catch quota management (CQM) trials and "unrestricted gear" trials .
In several studies, changes in fishers' behaviour were observed In two of the reviewed trials, the Dutch CQM and the Danish MINIDISC trials (studies #4 and #20,  (2018): EM has been widely tested and proven effective in monitoring protected species interactions in fishing gears.

| Summary of European trials, operational benefits of EM
The three major benefits of EM perceived in the European trials were as follows: (a) cost-efficiency, (b) the potential of EM to provide much wider (and more representative) coverage of the fleet than any observer programme will likely achieve and (c) EM registration of fishing activity and position of much greater detail.
With the potential to enhance data collection programmes, EM has the ability to improve the scientific stock assessment and risk assessment processes. In particular, the assessments of data-limited stocks (DLS) would benefit from a system like EM, the wider coverage of the fleet enabling data collection from less abundant species or specific fisheries, for example long-distance or smallscale fisheries, which are notably difficult to cover with a traditional observer programme. However, age and maturity data can only be collected through direct physical sampling. Observers can also collect sex data for some species by external observation (e.g. plaice, Elasmobranchs and Nephrops) which is not possible with existing EM systems. Therefore, EM cannot fully replace all the data needs currently provided by observers and it should be explored how observer and EM programmes could be integrated, as this would enable the benefits from both approaches to be utilized. An alternate possibility would be to continue development of length-based assessment methods, which would not require age data to the same extent as currently used in stock assessment methods .
In addition, EM species identification for similar-looking species was difficult for small species and when large concentrations of fish were processed (van Helmond et al., 2015). In contrast, observers can accurately identify all fish, crustacean and cephalopod species to the species level as required for stock assessments. However, there is potential for improving species identification in EM by making use of computer vision technology (Allken et al., 2019;French et al., 2015;Hold et al., 2015;Storbeck & Daan, 2001;Strachan, Nesvadba, & Allen, 1990;White et al., 2006).
The results of the EU review are summarized using a SWOT (Strengths-Weaknesses-Opportunities-Threats) analysis in the context of the current data collection framework ( initial installation costs can be overcome, EM offers the potential for fleet-wide monitoring coverage, with substantially more data than currently gathered in the various monitoring schemes, including the potential for length-distribution estimation of target species and a mapping of by-catch. In summary, EM as monitoring tool contains a range of solid strengths, that are not diminished by its weaknesses and EM has the opportunity to be a powerful tool in monitoring fisheries, integrated with existing data collection programmes, as long as a range of issues are addressed.

| Review of EM studies
There has been only limited coordination between the various trials between different regions in the world, and therefore, this review represents a step forward into synthetizing the outcomes of the various studies. Results of the studies have been documented in scientific peer-reviewed journals and technical reports. A challenge in this review was that not all trials have been well reported: some trials may never be documented, while others may not yet be documented because of a time delay in reporting results. Hence, it is not possible to include all trials in a global review. Another challenge in evaluating the performance of EM is that the technology has evolved over trials. Likewise, EM performance will evolve within trials and a perspective on the potential for EM may be more informed at the end of a trial rather than across a trial. Also, there is a difference in the level of detail in the methodology and results published in manuscripts or reports. Direct comparison between studies is, therefore, not always straightforward.

| Successes of EM worldwide
Based on continuity and expansion, EM has been successful in several different regions around the globe. Currently, EM programmes in Alaska, British Columbia, West and East Coasts of the United States and Australia are already well developed with comprehensive sampling schemes covering up to 100% of fleets, in some cases involving hundreds of vessels and thousands of fishing days. Clearly, the technical weaknesses of EM that were revealed in European trials have been encountered and solved in these examples where EM has been operationalized. In those cases, acceptance from the fishing industry was a crucial element for successful implementation of a full EM programme. Fully implemented programmes are often driven by the existence of a strong compliance or management issue that needs to be solved, for example gear theft or rampant discards, an example being the British Columbia, "Area A" crab fishery programme. In this case, EM is the best cost-effective solution and the efficiency of EM for these fisheries is demonstrated (McElderry, 2006). Full programmes can be adopted optimally if three components are present: (a) acceptance in the industry, (b) a strong incentive to monitor and (c) proven efficiency of EM.
Another component of successful EM implementation is government support. Electronic monitoring trials in the United States are subsidized by the government. A good example is the EM programme on the US Atlantic Highly Migratory Species longline fishery that was designed, approved and implemented in a little over a year (Michelin et al., 2018); such speed can be attributed to this being a fully government-funded EM programme. This initial investment by the government can help EM programmes develop, even if the long-term plan is to transition to industry cost allocation once a programme is fully implemented. On the other hand, system maintenance and longevity tend to be increased when fishers are investing in the systems themselves. A general factor in all fully implemented programmes (Table 3) is that EM cannot work in isolation and is often integrated with other monitoring elements, such as dockside monitoring, self-reported logs, observers and dealer reports. Various data types can provide useful information each with different strengths and weaknesses (Stanley et al., 2015).
In the field of research on interactions or by-catch of marine megafauna in commercial fisheries, EM is generally accepted as a reliable tool (Kindt-Larsen et al., 2012;Pierre, 2018). The high level of spatial and temporal coverage and the fact that megafauna is easily spotted on video records makes EM a very efficient tool for this purpose. This efficiency of EM in the field of by-catch registration of cetaceans is also reflected in the increasing number of activities organized by the Agreement on the Conservation of Small Cetaceans of the Baltic, North East Atlantic, Irish and North Seas (ASCOBANS).
The US regulatory programme to mitigate impacts on marine mammals in commercial fisheries potentially will also have an impact on the uptake of EM in the future (Michelin et al., 2018).
A fast-growing area of EM application is fisheries in remote areas, where monitoring fisheries is challenging, inefficient and costly.
Examples are the West and Central Pacific Islands, Indian Ocean and South Georgia. Electronic monitoring is a solution for enhancing existing observer programmes in these fisheries where extreme weather conditions, high safety risks and long distances make administering observer programmes difficult and EM is much less of a financial burden than an on-board observer (Ruiz et al., 2015;Stanley et al., 2015).
Also, issues of on-board accommodation, food, getting an observer in and out of remote locations do not exist with EM. In situations where the fishing industry has the responsibility, also financially, to monitor fishing activities, and where monitoring coverage is high, monitoring costs are a factor for an increased adoption of EM. In addition, EM put less constraints on the planning of fishing trips. Of course, when monitoring levels are minimal, the cost of buying and installing EM is higher than having an observer once every other year.

| Uptake of EM worldwide
Despite the apparent advantages of using EM systems in pilot studies, and successful EM programmes in some areas, fleet-wide implementation in globally important fishing regions is progressing slowly.
This slow uptake of EM can be attributed to several factors: 1. EM is often proposed as a compliance tool. This works well in situations when there is a common need to solve a compliance issue in the industry, for example the British Columbia, "Area A" crab fishery programme (McElderry, 2006) and the Groundfish Hook and Line Catch Monitoring programme in British Columbia (Stanley et al., 2015). However, in several cases EM was presented as a promising tool to monitor compliance in situations where full accountability seemed like an existential threat to the viability of the fishing industry (Michelin et al., 2018). This is especially true in fisheries with strong restrictions on discards and by-catches, like fisheries under the landing obligation in the EU, where fishers have become dependent on discarding the most limiting quota that would lead to early closures of the fishery, the "choke" species. Not surprisingly, EM has faced significant opposition from parts of the fishing industry in this region (Michelin et al., 2018;Plet-Hansen et al., 2017). 5. There is a strong perception of intrusion on the fishers' privacy.  point out that a large proportion of the fishing industry is not supportive in using EM for this reason. Besides privacy issues, the industry fears sensational use of footage, for example dolphin by-catch, liability and video manipulation (Michelin et al., 2018). Also, liability issues in the context of safety standards of work environment on-board can be an issue for vessel owners in cases where government institutions are requiring footage to monitor occupational health and safety regulations. Reluctance against EM regarding privacy issues and mistrust of data use is stronger for the proportion of the fishing industry without experience with EM . Once EM is implemented and fishers have actual exposure to EM, they generally have a more positive perception of the tool and it is easier to have an informed dialogue about applications (Michelin et al., 2018;Plet-Hansen et al., 2017). In other words, most fishers that are familiar with camera set-ups on their vessels did not experience an intrusion of privacy because of EM.
6. In some cases, EM raises concerns about employment impacts, especially when it is likely that at-sea observer sampling schemes will be scaled back with EM. These concerns are more concrete in regions with higher unemployment levels and where observer programmes enhanced job creation, but can be mitigated by employing experienced observers for video review, fisher liaison, data processing and following up on anomalies in imagery (Michelin et al., 2018). This may be preferable in the context of work-life balance, health and safety, since it allows staff to remain onshore.

| EM and the European Landing Obligation
A phased implementation of a landing obligation (LO) (EU, 2013) is implemented in the context of the European Common Fisheries Policy (Borges, 2015;Holden, 1994). Fully implemented and enforced the LO require fishers to report all catches of TAC species to be deducted from the quota. However, in practice non-compliance is potentially introduced (Batsleer, Poos, Marchal, Vermard, & Rijnsdorp, 2013;Borges, Cocas, & Nielsen, 2016;Condie, Grant, & Catchpole, 2013;Msomphora & Aanesen, 2015). Fishers are incentivized to continue to illegally discard low-valued fish to retain quota to fish for more valuable catches of the same species later and to prevent exhaustion of the most limiting quota that would lead to early closures of the fishery, the so-called "choke" effect (Batsleer, Hamon, Overzee, Rijnsdorp, & Poos, 2015;Baudron & Fernandes, 2015;Eliasen, Papadopoulou, Vassilopoulou, & Catchpole, 2014;Hatcher, 2014;Ulrich, Reeves, Vermard, Holmes, & Vanhee, 2011). Without additional or alternative tools for control and monitoring and/or a different set of incentives for fishers to fish more selectively, it has been anticipated that the LO will thus introduce more uncertainty into stock assessments and potentially jeopardize the chances of success of achieving the maximum sustainable yield (MSY) objective.
Electronic monitoring is often considered a potential candidate and, more importantly, the only financially affordable alternative, for full catch documentation under the LO (Aranda et al., 2019). An important constraining factor of implementing a full EM programme, within the context of the LO, is that EM is considered as a mechanism to monitor compliance. Such compliance-driven measures involving EM were only successful when there was support from the fishing industry. Incentives to gain support for EM would potentially improve the situation under the LO. For example, experiments with increased flexibility in gear choice , individual quota uplifts (van Helmond et al., 2016;Kindt-Larsen et al., 2011;Needle et al., 2015) and permission to enter closed areas (Needle & Catarino, 2011) have proved that incentives can make EM successful.
With regular feedback to the fishers, EM data can be used to inform on discard avoidance, and spatial distribution of unwanted catches, and could be disseminated on knowledge sharing platforms (Bergsson & Plet-Hansen, 2016;Bergsson et al., 2017;Needle et al., 2015). Electronic monitoring systems would have the potential to become a valuable information stream, for example, for the fishing industry to enable them to avoid unwanted catches or inform each other about real-time move-on rules.

| Enhancing the implementation of EM
Electronic monitoring as a monitoring tool contains a range of solid strengths that are not diminished by its weaknesses and EM has the opportunity to be a powerful tool in the future monitoring of a wide range of different types of fisheries. Electronic monitoring can be used to fully document a fishery or be integrated with existing data collection programmes, for management and compliance purposes or scientific data collection. Nevertheless, the viability of EM depends largely on how a range of threats are dealt with. Changes in the political landscape make the future of EM unpredictable; the end of the Fully Documented Fisheries programme in Denmark was the result of governmental change with a different view on fisheries management. Another important liability is its very low acceptance by the fishing industry. If EM is to be implemented as a monitoring tool, then turning this threat into an opportunity is the biggest challenge for EM, shifting the perception that EM is only fit for fisheries management and compliance objectives. In other words, changing the association of EM from being a "Big Brother" perspective to "giving the responsibility back to the fishing industry" in a results-based approach.
During the whole process of implementation, including the design and planning phases, involvement and participation of fishers are crucial (Stanley et al., 2015). In such a results-based approach, fishers are accountable for the impact they create on the marine environment (full documentation of catches), and EM should be used as a way for them to prove the reliability of their documentation, in the spirit of the "black boxes" used in trucks and flights.
Also, a marketing role is foreseen for EM: consumers would like to know the provenance or sustainability of the product they are buying. A growing number of seafood retailers are planning to link EM with traceability systems that allow for complete and transparent "net-to-plate" origin stories (Michelin et al., 2018). As part of this paradigm shift, additional issues such as hacking and data misuse will need to be addressed before a wide implementation can be completed, which requires discussions on data ownership, data storage facilities and access. Another underlying threat is the lack of evidence that EM is, in fact, less expensive than on-board observers in large-scale monitoring programmes.
In summary, EM as monitoring tool contains a range of solid strengths, that are not diminished by its weaknesses and EM has the opportunity to be a powerful tool in the future monitoring of the fisheries, integrated with existing data collection programmes.

ACK N OWLED G EM ENTS
The authors of this review would like to acknowledge and thank all the fishers, video inspectors, fisheries inspectors and support who have worked on the various EM projects. In addition, we thank two anonymous reviewers for their constructive comments on an earlier version of the manuscript. The participation of authors to this review has been funded through various sources, including the European Union's Horizon 2020 research and innovation programme under Thomas Catchpole https://orcid.org/0000-0001-7708-0967 Stephen Mangi https://orcid.org/0000-0001-8233-2612

DATA AVA I L A B I L I T Y S TAT E M E N T
The data, summarized information subtracted through literature review, that support the findings of this study are available from the corresponding author upon reasonable request.