Application of AIS- and flyover-based methods to monitor illegal and legal fishing in Canada's Pacific marine conservation areas

New approaches are required to undertake the substantial task of monitoring ongoing fishing activity in marine conservation areas to ensure conservation goals are achieved. To address this need, we applied previously developed, yet currently underused, vessel tracking methods based on Automatic Identification System (AIS) and aerial surveillance ( “ flyovers ” ) to Canada's Pacific marine conservation areas from 2012 to 2019. We used satellite and terrestrial-based AIS receivers and flyover-based visual observations to estimate illegal and legal fishing activity after 185 conservation area (CA) enactments (i.e


| INTRODUCTION
Monitoring ongoing fishing activity in marine conservation areas (CA) is a critical but challenging task for ensuring the effectiveness of conservation measures. Marine CAs are often used as a tool for conservation and fisheries management by regulating fishing pressure and other extractive activities (Grorud-Colvert et al., 2021;-these static and geographically defined areas, referred to here as "conservation areas", include Marine Protected Areas (MPA; specifically established under the Oceans Act in Canada) and other designated areas that restrict fishing. Fishing activity in marine CAs often remains prevalent either through non-compliance of regulations or continued permitted fishing (Arias et al., 2015;Iacarella et al., 2021;Lester & Halpern, 2008;Pollnac et al., 2010;. At the beginning of 2022, 92% of globally implemented marine CAs allowed moderate to extensive extraction (MPAtlas.org). Monitoring continued illegal and legal fishing activity is an essential, though often overlooked, component of CA monitoring . However, the logistical challenge of monitoring fishing activity is growing with the rapid expansion of marine CAs to meet international agreements to protect 25% of the ocean by 2025 and 30% by 2030 (Convention on Biological Diversity, 2021). Advancement and application of fishing activity monitoring methods will be imperative to ensure expansion of CAs can achieve the goal of protecting marine biodiversity.
Methods using remotely detected Automatic Identification System (AIS) messages to track fishing vessel activity are highly promising for monitoring marine CAs as they provide spatially and temporally continuous coverage of large, remote ocean spaces, as well as coastal areas de Souza et al., 2016;Dureuil et al., 2018;Kroodsma et al., 2018;McCauley et al., 2016). Large vessels (over 300-500 gross tonnage) are required internationally to carry AIS transponders for vessel communication and navigational safety, while some countries impose additional carriage requirements for smaller fishing vessels (Robards et al., 2016). Other vessels such as recreational boaters may voluntarily use AIS; for instance, 796 recreational vessel pathways were assessed within Canada's Pacific waters using 2016 AIS data (Iacarella, Burke, Davidson, et al., 2020a). AIS-based methods have significantly advanced in recent years beyond presence detections and vessel speed assessments (de Souza et al., 2016;Lee et al., 2010) with accelerating capabilities for big data processing and deep machine learning McCauley et al., 2016;White et al., 2020). Convolutional neural network algorithms, a form of deep machine learning, have been developed as a cutting-edge method to detect and quantify apparent fishing effort and the gear type used based on the movement patterns observed through AIS with an overall >90% accuracy . Example findings from the AIS-based method include a substantial reduction in commercial fishing activity after enactment of a large offshore protected area (McCauley et al., 2016), prevalent trawling in CAs in the European Union (encompassing different regulations and designations; Dureuil et al., 2018), and illegal commercial fishing in inshore artisanal fishing areas in Africa . These methods present a significant opportunity to monitor illegal and legal fishing activity and target surveillance resources towards high non-compliance areas (Bergseth et al., 2015;Iacarella et al., 2021).
Additional vessel tracking methods are important to capture fishing activity that is missed by AIS (see review by Iacarella et al., 2021). In particular, the utility of AIS for monitoring CAs is more limited for countries that do not mandate carriage by fishing or small vessels (Exeter et al., 2021;Kanjir et al., 2018), as well as in cases of illegal fishing where AIS message transmission may be blocked or falsified Rowlands et al., 2019). A complementary approach to the AIS-based method is the use of information collected on vessels by aerial surveillance (hereafter, "flyover") programs. Flyovers are often conducted for real-time enforcement of marine CA regulations, but have also been successfully applied to monitor spatial and temporal trends in vessel and shore-based fishing activity in marine CAs (Haggarty et al., 2016;Rojas-Bracho & Reeves, 2013;Smith et al., 2015). Surveillance planes also carry AIS receivers which enables an estimate of vessel activity missed due to gaps in AIS coverage. For instance, a study in the southern portion of Canada's Pacific coast found 70% of 756 flyover vessel detections from 2015 to 2017 were of vessels without AIS (Serra-Sogas et al., 2021). Most non-AIS-based detections were for recreational and commercial fishing vessels, which are not mandated to carry AIS in Canada's waters.
Flyover information can greatly help supplement fishing activity missed using AIS-based methods, though it has its own limitations with constraints on surveillance frequency and coverage as substantial resources are required to operate planes and staff surveillance officers (Ford et al., 2022;Kelleher, 2002). Flyover information is also generally restricted to fishing detection counts (e.g., Smith et al., 2015), whereas AIS-based algorithms estimate fishing effort as the number of hours a vessel is actively fishing (Dureuil et al., 2018;Kroodsma et al., 2018;McCauley et al., 2016). Conversely, flyovers can provide visual confirmation of gear use compared with model estimates by AIS-based algorithms McCauley et al., 2016). A comprehensive picture of illegal and legal fishing activity in CAs requires multiple vessel tracking methods; a number of methods are available (e.g., Vessel Monitoring System (VMS), Synthetic Aperture Radar (SAR), onshore surveillance cameras; Lee et al., 2010;Chang & Yuan, 2014;Harasti et al., 2019;Rowlands et al., 2019;Magris, 2021;Burke et al., 2022), but currently the AIS-and flyover-based methods are the most extensive, accessible, and complementary in Canada (Iacarella, Clyde, & Dunham, 2020b).
We apply and compare AIS-and flyover-based vessel tracking methods for estimating illegal and legal fishing for Canada's Pacific marine CAs across 8 years. Our approach is more in depth for CAs than most to-date (e.g., Belhabib et al., 2020;Dureuil et al., 2018;White et al., 2020) in that we included several types of marine CAs with fishing regulations legislated in Canada and accounted for the many different regulations, zoning rules, and enactment dates to identify what constitutes illegal and legal activity specific to each CA ( Figure 1a and Table 1) (for AIS-based analysis applied to all of Canada see Iacarella et al., 2023). We determined the amount of fishing activity across CA types using both AIS-and flyover-based methods. We further applied the flyover-based method to evaluate gaps in fishing activity estimation from vessels without AIS based on user group (e.g., commercial, recreational). We then compare and discuss the strengths and weaknesses of both methods and how they can be used in a comprehensive monitoring strategy for marine CAs. These methods provide an advanced, yet underused, approach to monitoring fishing activity that is urgently needed to ensure marine CAs can achieve their intended conservation outcomes.

| Marine conservation areas
We included all geographically defined, coastal and oceanic CAs with static fishing regulations within Canada's Pacific Exclusive Economic Zone that were enacted prior to 2020 (1990-2019) for analysis. Seasonal fishing closures were excluded from the scope. This resulted in a total of 185 designated CAs that consisted of three Oceans Act MPAs (Endeavor Hydrothermal Vents, Hecate Strait and Queen Charlotte Sound Glass Sponge Reefs, SG̱ aan Ḵinghlas-Bowie Seamount), 18 Marine Refuges (Offshore Pacific Seamounts and Vents, and 17 Strait of Georgia and Howe Sound Glass Sponge Reef closures), 162 Rockfish Conservation Areas, one static Fisheries Closure (Race Rocks), and one National Marine Conservation Area (Gwaii Haanas) ( Figure 1a). Different CA types (e.g., Marine Refuge, Rockfish Conservation Area) have different purposes for protection (e.g., recreational use, species or habitat specific conservation), protection levels (i.e., ranging from fully to minimally; Grorud-Colvert et al., 2021), and IUCN categories (Table 1). Protection levels and IUCN categories have only been evaluated for the Oceans Act MPAs and the National Marine Conservation Area (Barron & Groulx, 2021). Regulations vary for each CA type and within some CA types (e.g., Oceans Act MPAs and Marine Refuges). The details on regulations for restricted ("illegal") and permitted ('legal') fishing were compiled and then verified by experts within Fisheries and Oceans Canada (DFO; see Appendix S1 for a list of resources and further information). Hecate Strait and Queen Charlotte Sound Glass Sponge Reefs MPA had two, spatially distinct designated zones with different regulations that

Oceans Act
To protect and conserve marine species, habitats, and ecosystems that are ecologically significant and distinct.

Fisheries Act
To decrease fishing mortality of inshore rockfish, provide opportunities for these species to rebuild, and protect their habitat. NA 2004NA , 2005NA , 2006 FC (1)

Fisheries Act
To conserve and protect fish and fish habitat. (1) Fully Protected-minimal low impact activities occur, no extractive or destructive activities, (2) Highly Protected-very limited extraction, (3) Lightly Protected-some protection but moderate extraction allowed, (4) Minimally Protected-moderate to heavy extraction with significant impacts for biodiversity, (5) Incompatible-activities with impacts so great areas were considered incompatible with biodiversity conservation (Barron & Groulx, 2021;Grorud-Colvert et al., 2021). b MPA enacted in 2008, but fisheries restrictions assessed in this paper were enacted in 2018.

Race Rocks Fisheries
were treated as separate CAs (i.e., Adaptive Management Zone, Vertical Adaptive Management Zone; Appendix S1 and Table S1). For seamounts, we included those within the Offshore Pacific Seamounts and Vents Closure as separate CAs (n = 5). Similarly, for the SG̱ aan Ḵinghlas-Bowie Seamount MPA in the Pacific, we focused spatially on the three seamount extents that have fishing restrictions. Seamount extents were delineated by a 1500 m depth contour provided by Fisheries and Oceans Canada. This resulted in the total number of assessed CAs increasing to 192.

| Satellite and terrestrial AIS data
Historical fishing effort estimates (i.e., time spent by vessels actively fishing) within CAs were produced based on satellite and terrestrial AIS data. AIS data were compiled from global satellite data provided to the Canadian Space Agency (2012-2016) and Global Fishing Watch (GFW; 2017-2019), and from coastal terrestrial data collected by the Canadian Coast Guard (2012-2019). Satellite AIS data provide spatially extensive coverage, but contain some temporal gaps as reception of AIS messages is dependent on orbital passes. Conversely, terrestrial AIS data are spatially restricted depending on the extent of land-based stations and to within 50 nautical miles of the coastline, but have no temporal gaps as long as vessels are within line-of-sight (Canada's AIS data are detailed in Iacarella, Clyde, & Dunham, 2020b Target species Flyover-based gear classes (143) AIS-based gear classes (16) F I G U R E 2 Process of determining whether fishing activity detected within conservation areas (CA) from AIS-and flyover-based methods was likely illegal or legal based on gear class details and conservation area regulations. Inability to directly match gear classes to some regulations led to identification of "remainder" fishing that could either be illegal or legal.
that entered a CA from 2012 to 2019 was used to develop movement patterns for identification of fishing activity. Convolutional neural network algorithms were applied to predict whether a vessel was fishing and, together with vessel registries, what gear type it was using among 16 gear classes (Appendix S2). Gear classes were then matched to each CA's regulations to determine whether estimated fishing effort was illegal or legal by identifying the common water column location, user group, and target species of the gear classes ( Figure 2; see Appendix S3 for gear classes matched to CA regulations). In some cases, gear classes could not be directly linked to particular regulations and therefore could not be parsed into illegal or legal fishing. This problem arose when there was a mismatch between the GFW gear classifications and varying levels of specificity for each CA's regulations, and led to identification of "remainder" fishing (illegal/legal/ remainder categories are hereafter referred to as "fishing activity"). Gear classes were more often identified as remainder when regulations were more specific and less strict (e.g., some gear types or fishing certain species allowed) compared with when regulations were broad and strict (e.g., all commercial and recreational fishing prohibited). For example, trawling detected in a CA that only restricted bottom contact fishing was identified as remainder fishing since the operating depth of trawlers was unknown.

| Flyover data
Flyover data were used to compare vessel monitoring effectiveness and address the gap in detecting fishing by vessels without AIS. These data provided vessel counts instead of fishing effort estimates, but highlighted the amount of fishing that was missed by satellite and terrestrial AIS through comparison of the number of fishing vessels detected with AIS (plane-received) and without AIS (visual and radar) during flyovers. Flyover data were compiled from DFO's Conservation & Protection aerial surveillance program database for the same 2012-2019 timeframe and included a total of 1386 flyovers ( Figure 1b). Mean flight times were 6.3 h ±0.06 (±1 SE) for a total of 8482 h from 2012 to 2019. Vessels with AIS transponders were detected by an AIS receiver attached to the plane, whereas vessels without AIS were detected with inverse synthetic-aperture radar and from visual observation by aircraft observers ("non-AIS" detections). Vessel detection using radar was set to 150 km during low altitude flyovers to target smaller fishing vessels. The vessel's gear type, user group (i.e., Aboriginal, commercial, recreational, government/research), and target species were determined visually or by looking up observed vessel information (e.g., vessel name) in a DFO database of fishing vessels. Details on flights and tracked vessels were contained in multiple reports and stored in a database. Five vessel reports were compiled for each flyover and provided information on the vessels detected and their mode of detection (i.e., AIS, radar, and visual) . We focused analysis of flyover vessel counts with and without AIS on those actively fishing to match the AIS-based method. Vessels with AIS were counted as actively fishing based on information provided in transmitted messages (11% of 56,998 fishing vessels), which was verified by aircraft observers in 87% of cases. Vessels without AIS were identified as actively fishing from visual observations made during flyovers. Further fishing activity categorization and user group summary statistics only included non-AIS fishing vessels that had visually confirmed gear type, user group, or target species. This information was also recorded for many AIS fishing vessels, but we focused on non-AIS vessels as this was the primary added value of flyovers to the AIS-based method. Specific details provided for gear type, user group, and target species for each fishing vessel were compiled and used to create 143 gear classes (Appendix S4). Like the AIS-based gear classes, flyover-based gear classes were binned into illegal, legal, and remainder fishing specific to each CA's regulations (Figure 2 and Appendix S4).

| Mapping and analysis
Georeferenced fishing effort hours from the AIS-based method and fishing activity counts from the flyover-based method within the spatially defined CAs and postenactment specific to each CA were selected for analysis (Appendix S1). AIS-based fishing effort values (i.e., hours vessels spent fishing) were binned into 0.01 cells for the CAs' extents and the total hours per month were calculated from 2012 to 2019 for each fishing gear class. Fishing effort per month post-enactment was then further summed into illegal, legal, and remainder fishing activity categories specific to each CA. Using the flyover-method, the total number of vessels actively fishing within CAs for each gear class and fishing activity were summed per month. Note, our identification of illegal fishing activity was not verified as flyover details on confirmed illegal activity, as well as reports on violations, are classified. We calculated summary statistics for non-AIS fishing vessel detections across the timeframe and by CA type and user group identified from visual observation. Spatial analyses were done using QGIS (V2.0), ArcGIS (V.10.8), Python (V3), and Google BigQuery. Data compilation and plotting were done using R (R Development Core Team, 2021).

| RESULTS
The number of vessels identified as fishing within Canada's Pacific CA extents (not considering enactment dates) using the AIS-based method tended to increase from 2012 to 2017, with the highest monthly counts generally occurring in March (2013March ( , 2015March ( -2019 and the lowest in December (2013,(2015)(2016)(2017)(2018) (Figure 3). Flyovers detected substantially more vessels actively fishing without AIS than with AIS within CA extents from 2012 to 2019 (93% of 889), with the highest monthly counts of non-AIS fishing vessels generally occurring from May-August and lowest counts from November to February (Figure 3). An increase in non-AIS vessel observations starting in 2018 did not coincide with an increase in the number of flyovers conducted (Figure 3). The user groups with the highest percentage of non-AIS detections within CA extents-noting that many of these vessels were not required to carry AIS-were Aboriginal (100% of 63), government or research (100% of 1), multiple (i.e., commercial or recreational; 100% of 33), followed by recreational (99.5% of 230), and commercial (89% of 562). Commercial fishing vessels represented the largest user group (61%) of vessels in CAs without AIS.
Using the AIS-based method, 76% of the 192 CAs had fishing effort detected within the CA extent (before or after enactment) from 2012 to 2019 (Figure 4a). Fishing effort was detected within 74% of CAs after enactment, with illegal fishing activity estimated in 31% of CAs. One third of estimated fishing effort within CAs after enactment was legal (32%, 1045 h), 22% was illegal (448 h), and 55% was identified as remainder (1810 h). Using the flyover-based method for non-AIS vessels, 56% of CAs contained active fishing vessels from 2012 to 2019, all of which had continued fishing after enactment (Figure 4b). Illegal fishing activity was estimated in 32% of CAs after enactment. More than half of all non-AIS fishing vessels within CAs after enactment were identified as legal (66%, 469 vessels), 20% were illegal (145 vessels), and 14% were remainder (102 vessels). Notably, a similar proportion of illegal activity was detected between the two methods, though the flyover-based method had much less remainder fishing with subsequently a higher proportion of legal fishing identified than the AIS-based method. Conversely, the AIS-based method detected fishing activity in more CAs.
The type and amount of fishing activity varied greatly depending on CA type. AIS-based fishing effort postenactment of CAs totaled 2952 h in Rockfish Conservation Areas, followed by 190 h in Oceans Act MPAs, 81 h in the recently enacted (2019) National Marine Conservation Area, 80 h in Marine Refuges, and less than an hour in the Fisheries Closure. The highest estimated fishing effort density (per CA area and year) was legal fishing within Rockfish Conservation Areas, followed by illegal fishing in the National Marine Conservation Area (Figure 5a). Flyover-based, non-AIS fishing vessel densities also indicated the highest level of activity in Rockfish Conservation Areas as legal, and reported one vessel potentially fishing illegally in the National Marine Conservation Area (Figures 5b and Figure 6). In Rockfish Conservation Areas, proportionally more illegal activity was detected by flyovers counts of non-AIS fishing vessels (16% of counts) than by AIS-based fishing effort estimates Flyover−based F I G U R E 3 Monthly number of actively fishing vessels within conservation area extents, not considering enactment dates, as detected using terrestrial and satellite Automatic Identification Systems ("AIS-based method") and during aerial surveillance ("flyover-based method") using planebased AIS receivers or by visual observation and synthetic-aperture radar ("non-AIS"). Gray shading represents the number of monthly flyovers conducted.
(5% of effort). The highest mean density of non-AIS fishing vessels detected by CA type was remainder fishing within Marine Refuges (Figure 5b).
The user groups identified most frequently by flyovers as actively fishing without AIS and within CAs after enactment were commercial vessels fishing legally F I G U R E 4 Total fishing effort estimated using the AIS-based method (a) and total counts of actively fishing vessels without AIS using the flyover-based method (b) from 2012 to 2019 (0.02 grid). Conservation areas are indicated as black polygon outlines, and data are clipped to Canada's Pacific Exclusive Economic Zone (dashed outline). (n = 361), recreational vessels fishing illegally (n = 120), recreational vessels classified as remainder (n = 57), recreational vessels fishing legally (n = 50), and Aboriginal vessels fishing legally (n = 49). All but one of the potential illegally fishing recreational vessels were reported within Rockfish Conservation Areas ( Figure 6). All legally fishing commercial vessels were also found within Rockfish Conservation Areas. In addition, several commercial vessels were potentially fishing illegally in Rockfish Conservation Areas (n = 15), Oceans Act MPAs (n = 7), and Marine Refuges (n = 2). All other non-AIS vessels fishing within Marine Refuges were recreational vessels classified as remainder fishing (n = 32; Figure 6).

| DISCUSSION
The AIS-and flyover-based vessel tracking methods revealed substantial fishing activity within Canada's Pacific marine CAs. The AIS-based method enabled measures of fishing effort, which resulted in 3312 h of estimated fishing activity from 2012 to 2019 in CAs after enactment. The flyover-based method provided vessel counts with visual verification of vessels without AIS. This method showed that most vessels identified as fishing in CAs were not carrying AIS (93%), indicating that there was substantially more fishing effort in CAs in addition to that found using the AIS-based method. In particular, commercial fishing vessels were the most prevalent vessels without AIS (61%) across user groups. The application of AIS-and flyover-based methods demonstrates the need to better align vessel tracking information and CA regulation specifications as a non-trivial amount of fishing activity could not be determined as illegal or legal, particularly for the AIS-based method (55% remainder), but also for the flyover-method (14% remainder).

| Estimated illegal and legal fishing in conservation areas
Many of Canada's Pacific CAs had ongoing fishing detected after enactment, and a third had illegal activity detected by both methods. The National Marine Conservation Area had the highest density of illegal activity estimated from AIS compared with other CA types, though only one identified illegally fishing recreational vessel was noted from flyovers. This high level of illegal activity may be from the recent enactment of this CA (2019), which could correspond with a lack of public awareness, as well as less surveillance and enforcement as indicated by the one illegal vessel detected from flyovers. Lack of public awareness and insufficient enforcement are key drivers influencing non-compliant fishing in marine CAs globally (Iacarella et al., 2021;Kirkman et al., 2021;Rojo et al., 2019). For instance, the initiation of a geofencing alert system for commercial fishing vessels entering Australian Marine Parks tackled both awareness and surveillance issues simultaneously (Read et al., 2019). The alerts reduced recorded illegal fishing to zero over 4 years compared with six prosecutions in the prior 3 years (Read et al., 2019). Conversely, flyovers found the most prevalent illegal activity to be recreational fishing in Rockfish Conservation Areas, a problem that has been highlighted before (Haggarty et al., 2016). Flyovers have been shown to be particularly useful for detecting illegal fishing by recreational vessels (Smallwood & Beckley, 2012;Smith et al., 2015;Zellmer et al., 2018), a user group often not carrying AIS as shown here and by others (Serra-Sogas et al., 2021).
There was also a high density of legal and remainder fishing activity in both Marine Refuges and Rockfish Conservation Areas detected by AIS, with primarily remainder recreational fishing in Marine Refuges and legal commercial fishing in Rockfish Conservation Areas detected by flyovers. The legal and remainder fishing in Rockfish Conservation Areas may not directly impact the intended conservation outcomes, but the permitted fishing activity in part accounts for the reason these CAs are currently not included towards Canada's marine conservation targets to meet international agreements (Convention on Biological Diversity, 2021;Thornborough et al., 2020). Conversely, the Oceans Act MPAs, Marine Refuges, and National Marine Conservation Area are included in Canada's marine conservation targets (https://www.dfo-mpo.gc.ca/oceans/conservation/areaszones/index-eng.html). These also had ongoing illegal and legal fishing detected by both methods, which is known to reduce the ecological performance of marine CAs (Bergseth et al., 2015;Lester & Halpern, 2008;Pollnac et al., 2010;Rife et al., 2013;White et al., 2015).
Well enforced, no-take CAs have repeatedly been shown to be more effective conservation tools than partially protected areas or areas that lack sufficient enforcement (Giakoumi et al., 2017;Sala & Giakoumi, 2018;Ziegler et al., 2022). Our results on the amount of ongoing fishing in Canada's Pacific CAs, which are highly conservative based on detection limitations, signify that some of these areas likely need further regulation and enforcement to achieve conservation goals (see also Iacarella et al., 2023). In particular, Kuempel et al. (2018) show that investment in enforcement should at least equal the investment in expansion of CAs to successfully protect exploited species.
When comparing levels of fishing activity by different user groups, it is important to note that a single commercial vessel can retain significantly more biomass than a recreational vessel. For instance, hundreds to thousands of poached sharks from the Galapagos Marine Reserve have been seized from single industrial vessels (Carr et al., 2013). Marine recreational fishing is estimated to comprise just under 1% of total global marine catches, a smaller but non-negligible contribution in part given the types of species that are targeted by recreational fisheries (Freire et al., 2020). Our comparison of user groups fishing in CAs provides an indication of which groups may be useful to focus on for management efforts and public outreach, but does not indicate which are causing the greater impact on conservation outcomes.

| AIS-versus flyover-based methods
We found that the fishing effort estimates provided by the AIS-based method enabled more meaningful quantification of fishing activity that is required for monitoring, evaluation, and reporting on CA performance. The flyover-based method provided fishing vessel counts which were less useful for quantitative evaluation, but helped fill the critical gap of detecting vessels without AIS carriage requirements. This is particularly important for countries with fewer carriage mandates. For instance, the number of recreational fishing vessels without AIS is substantial (Serra-Sogas et al., 2021). In Canada, commercial fishing vessels are also exempt from carrying AIS, and clearly some operators chose not to use it as evidenced by the percentage of commercial vessels without AIS detected by flyovers (89%). Fisheries management instead has traditionally relied on Electronic Monitoring System, VMS, and logbook report data to monitor commercial fishing. VMS data for commercial fisheries have been applied in the same way as AIS to determine fishing effort and illegal fishing activity, and can provide more data than AIS for commercial fishing in countries that have extensive carriage requirements (Chang & Yuan, 2014;Lee et al., 2010;Magris, 2021). However, these other tracking sources can have limitations from restricted data access and use, short term archival (e.g., Electronic Monitoring System data), and minimal carriage requirements depending on the region and fishery (McCauley et al., 2016;Kanjir et al., 2018;Iacarella, Clyde, & Dunham, 2020b;Exeter et al., 2021). Conversely, AIS data are publicly accessible, usable without restriction, and well archived (Iacarella, Clyde, & Dunham, 2020b;McCauley et al., 2016). Our results of increased AIS fishing vessels detected in CAs over the timeframe is likely from a combination of more AIS carriage and satellite coverage (Iacarella, Clyde, & Dunham, 2020b;McCauley et al., 2016). Additional AIS carriage requirements would greatly benefit CA monitoring and evaluation. In particular, we suggest that AIS should be mandated for all vessels operating within CAs.
Reliance on AIS can be problematic however when there are gaps in data from lost signal transmissions or purposeful disabling or masking of AIS transmitters (Welch et al., 2022). Detection of when a vessel "goes dark" can be used as an indication of illegal fishing events, particularly when the transmission gap spans the time a vessel is within a CA or near an Exclusive Economic Zone boundary Rowlands et al., 2019;Welch et al., 2022). Statistical approaches have been developed to determine whether loss in signal transmission is purposeful to highlight areas of potential illegal activity Welch et al., 2022). Spatial and temporal gaps in AIS data, or conversely high estimated fishing effort, can be used to identify CAs or other areas where additional surveillance methods, such as flyovers, may be most useful.
Direct observations of vessels through flyovers, cameras, and shore observation are commonly applied approaches for monitoring non-compliance in marine CAs that avoid the problem of "dark" vessels (i.e., without AIS or VMS, or those disabling or masking signal transmission) (Bergseth et al., 2015;Iacarella et al., 2021). The flyover-based method is more suitable than cameras and shore observations when many CAs need to be monitored as it provides vessel detections across large extents, as shown here and elsewhere (Haggarty et al., 2016;Smallwood & Beckley, 2012;Smith et al., 2015;Zellmer et al., 2018). Aerial surveillance is highly useful for visual verification of vessel activity and identification of all vessel types and sizes on the water, and has been successfully applied to evaluate vessel traffic and fishing within CAs (Haggarty et al., 2016;Smallwood & Beckley, 2012;Smith et al., 2015;Zellmer et al., 2018). However, this method has limited spatial and temporal coverage relative to remote sensing methods, is more useful for nearshore monitoring, and is highly resource intensive (Ford et al., 2022;Kelleher, 2002).
Remote surveillance methods that can detect dark vessels include the use of SAR Kurekin et al., 2019;Rowlands et al., 2019), optical satellite images (Kanjir et al., 2018) including Visible Infrared Imaging Radiometer Suite of nighttime fishing vessel lights (Exeter et al., 2021;Park et al., 2020), and creative approaches such as fitting albatrosses with geolocating loggers (Weimerskirch et al., 2020). SAR has been the best available satellite imaging sensor for ship detection to-date and has the same broad spatial coverage as satellite AIS (Kanjir et al., 2018). It has been used to effectively capture vessels that may be missed by AIS or flyovers Rowlands et al., 2019) and is often combined with AIS or VMS methods for Maritime Domain Awareness (Kanjir et al., 2018). Important limitations of SAR include low temporal resolution, misidentification or lack of detection depending on sea state, and a minimum size of vessels detected (Kanjir et al., 2018). Furthermore, illegal activity in CAs can only be determined from SAR where there are full restrictions on vessel entry or on vessels of a certain size Rowlands et al., 2019). Resource requirements of these various methods is an important consideration as monitoring and enforcement is costly (Ford et al., 2022;Kelleher, 2002), and many CAs are resource limited (Campbell et al., 2012;Gill et al., 2017). However, satellite radar and high resolution optical imagery costs are decreasing, and the data are becoming more available and accessible to the public (Park et al., 2020). Using multiple methods is advantageous to fill in the various gaps and limitations as shown here, by other comprehensive illegal fishing studies (Park et al., 2020;Rowlands et al., 2019), and by reviews of non-compliant fishing and detection methods (Bergseth et al., 2015;Iacarella et al., 2021;Kanjir et al., 2018).

| Challenges and recommendations for improving fishing activity monitoring in conservation areas
Applying the AIS-and flyover-based methods to identify illegal and legal fishing activity required matching their respective gear classifications to CA regulations, which we found was unachievable in some cases and led to the category of "remainder" fishing. This presents a significant problem for accurately monitoring and evaluating compliance and the effectiveness of CAs, but can be addressed by improving both gear classifications from tracking data and CA regulations. In the case of the AISbased method, gear classes were restricted to those that can be identified based on vessel movement patterns so matching to regulations that specify water column position (e.g., benthic trawling) for gear classes that can be operated at different depths (e.g., midwater or benthic trawling) was not possible with the current algorithms.
The additional specification of user group in many CA regulations also cannot be distinguished with the AIS-based method unless the gear type is only used by a particular group. The flyover-based method had some of the same challenges despite being conducted at least in part for the purpose of monitoring these CAs. However, we were able to apply user group identification from flyover information which aided in better matching to CA regulations, and enabled further exploration of which user groups were predominately fishing within CAs. Such analyses highlight information needs and can help inform and guide effective data collection during future flyovers.
The challenge of linking vessel tracking gear classes to CA regulations also emphasized the importance of creating regulations that are simple, clear, and defined with consistent language and level of specificity across CA types, for more feasible monitoring and evaluation. CAs with clearly understandable and enforceable regulations have been found to exhibit improved compliance and enforcement, and subsequently better ecological performance (Cinner & Huchery, 2014;Mascia, 2003). Regulations that cannot be assessed for compliance lead to an inability to evaluate management effectiveness and thus may also compromise ecological performance assessments. The simplest type of CA regulation-no-take areas-provide the easiest means of enforcement, compliance, and monitoring as all fishing is illegal. Illegal fishing has still been found to occur in these areas (Harasti et al., 2019;Magris, 2021;Rowlands et al., 2019), particularly without sufficient enforcement (Rife et al., 2013). However, when successful, no-take CAs are the most effective for biodiversity conservation (Lester & Halpern, 2008;. Our results of high levels of ongoing legal fishing and challenges in determining illegal versus legal activity lend further support for designation of no-take CAs.
Our application of AIS-and flyover-based methods to Canada's Pacific marine CAs builds on the utility of these methods through demonstration of how they can be used to evaluate fishing activity specific to CA regulations. The AIS-based method can be used to quantitatively assess management effectiveness of CAs, whereas the flyoverbased method can identify how much additional fishing is conducted by non-AIS vessels. Finally, mandating AIS carriage for most or all vessels operating within CAs would greatly benefit monitoring and evaluation.

ACKNOWLEDGMENTS
We thank the many people who provided input on initial project ideas and who facilitated data collection and data review including managers, scientists, and surveillance and enforcement officers from Fisheries and Oceans Canada (DFO; Conservation & Protection, Marine Security and Operations Centre, Ecosystems and Oceans Sciences, Fisheries Management, Ecosystems Management, Canadian Coast Guard) and the Canadian Space Agency. We also thank the reviewers for providing helpful feedback on expanding the original paper. The project was funded by the DFO National Conservation Plan and the DFO Strategic Program for Ecosystem-based Research and Advice.
DATA AVAILABILITY STATEMENT AIS-based monthly fishing effort estimates by fishing activity for all conservation areas, shapefiles of gridded yearly fishing effort by gear class, and shapefiles of closure and buffer extents are provided on Dryad: https:// doi.org/10.5061/dryad.70rxwdc3d. Flyover-based data are not publicly accessible.