Monitoring temporal and spatial trends of illegal and legal fishing in marine conservation areas across Canada's three oceans

Expansion of marine conservation areas (CA) necessitates resource‐efficient and achievable strategies for monitoring and evaluation of ongoing fishing activity at national levels. To demonstrate and explore such a strategy, we conducted the first extensive analysis of fishing activity within Canada's static, geographically defined marine CAs with fishing regulations (n = 264 areas). We used 8 years of Automatic Identification System data to estimate fishing effort across three oceans and conducted temporal and spatial comparisons specific to each CA's regulations and enactment date. We addressed questions on CA effectiveness, fishing displacement, fishing the line behavior, and relationships between fishing activity and spatial CA attributes. We estimated 22,000 h of fishing activity within CAs after enactments, 22% of which was identified as illegal. CA effectiveness appeared to be lowest for Atlantic CAs based on illegal fishing effort density within CAs. Fishing displacement and fishing the line was generally not apparent as buffer areas around CAs tended to already have higher fishing effort prior to enactments. CA effectiveness and responses to CAs varied considerably, as was visualized using timeseries plots and maps developed for each CA. Our evaluation of a nation's full suite of CAs provides managers with a foundation and approach for continued monitoring and reporting.


| INTRODUCTION
Monitoring marine conservation areas (CA) to determine whether the intended goals of protection are being achieved requires monitoring both human pressures and ecological components. Ongoing human pressures are often overlooked in CA monitoring frameworks, despite the fact that CAs can only be successful if regulated activities are mitigated and if non-regulated activities are not causing unforeseen impacts Gill et al., 2017). Marine CA monitoring instead has typically focused on measuring ecological performance to track the maintenance or improvement of biodiversity and fish stocks, primary goals of many CAs (Côté et al., 2001;Edgar et al., 2014). The main regulatory focus of marine CAs is to limit extractive impacts of fishing (Grorud-Colvert et al., 2021); we refer to all static, geographically defined areas with fishing restrictions as "conservation areas" (including Marine Protected Areas (MPA), which for Canada are those established under the Oceans Act, and other designated area types; see Iacarella et al., 2023). As such, tracking ongoing fishing activity is especially important for understanding management effectiveness of CAs and trends in ecological performance. Low ecological performance can only be understood if it is known whether fishing activity remains high as this will likely lead to reduced expectations of performance Gill et al., 2017;. Illegal fishing has often been the cause of unsuccessful ecological performance of marine CAs (Bergseth et al., 2015;Gill et al., 2017;Iacarella et al., 2021;Pollnac et al., 2010;Rife et al., 2013;Rojo et al., 2019;White et al., 2015). Unregulated, legal fishing within CAs can be similarly detrimental to protecting biodiversity and fish stocks (Aburto-Oropeza et al., 2011;Dureuil et al., 2018;Lester & Halpern, 2008;. Furthermore, meta-analyses of studies on CA ecological performance have found that no-take CAs have much greater increases in fish biomass than partial CAs where some extractive activity is still allowed (Giakoumi et al., 2017;Sala & Giakoumi, 2018).
Monitoring temporal trends in fishing activity before and after CA enactments is a key component of understanding management effectiveness and ecological performance of CAs. If illegal activity is successfully reduced, but there is no resulting improvement in ecological performance, then levels of legal activity (or other human pressures) may in part indicate why populations are not maintained or improved (Dureuil et al., 2018;Rife et al., 2013;. Shifts from regulated gear types to legal ones or an increase of pre-existing legal fishing within CAs is a form of fishing displacement that may have unintended consequences (Vaughan, 2017). For instance, temporal trends evaluated using Vessel Monitoring System (VMS) data found increases in fishing intensity in a partial CA, as well as increased fishing in the first year after enactment of a no-take CA (Magris, 2021). Analysis of the large, remote Phoenix Islands Protected Area using Automatic Identification System (AIS) data found a 130% increase in fishing effort (hereafter defined as time vessels spend fishing) relative to a control area between the time of CA announcement and enactment, followed by a reduction after enactment (McDermott et al., 2019). Another analysis using AIS data found little change in fishing effort after enactments of four other large, remote CAs owing to minimal pre-existing fishing (White et al., 2020). These temporal patterns shed light on the effectiveness of CAs in mitigating fishing activity and what may be reasonable expectations of ecological performance in response to CA enactments.
Spatial comparisons of fishing activity are another important element for identifying aggregation of activity within CA extents and shifts in fishing activity from CA enactments. For instance, hot spots of fishing activity can be visualized to determine high fishing density locations and how this may change over time (Boerder et al., 2017;Magris, 2021). Areas outside of CAs that historically had similar fishing effort, or broader surrounding marine extents, can also serve as controls for understanding fishing trends that are an effect of the CA (Dureuil et al., 2018;McDermott et al., 2019;White et al., 2020). Fishing activity in areas directly surrounding CAs (i.e., CA buffers) can further identify the impact of the CA (Rowlands et al., 2019), fishing displacement from the CA to surrounding extents (Bucaram et al., 2018;Vaughan, 2017), and "fishing the line" behavior (Boerder et al., 2017;Tassetti et al., 2019). Displaced fishing effort is important to detect and monitor as this may lead to failure of conservation goals if the target species being protected experience the same or higher rate of fishing pressure transferred to outside of the CA (Gilman et al., 2019). Likewise, fishing the line behavior is a result of fishers seeking higher catches from the spillover effect of CAs, and can lead to accidental crossing over into CAs (Kellner et al., 2007;Tassetti et al., 2019). Spatial attributes of CAs can also influence the likelihood of illegal fishing from accidental crossing into CAs (e.g., low area to perimeter ratios, complex boundaries) or purposeful noncompliance from less enforcement coverage (e.g., large, remote) (Crawford et al., 2004;Read et al., 2011;Rojo et al., 2019). Monitoring spatial trends of fishing activity within and surrounding CAs provides essential information on management effectiveness of CAs and subsequently altered fishing activity outside of CAs.
We present a nationwide evaluation of temporal and spatial patterns in illegal and legal fishing activity across Canada's three oceans and over 8 years to address a major gap in current conservation area monitoring. We assess fishing activity using AIS-based estimates of fishing effort applied to Canada's marine CAs, specific to their regulations and enactment dates, across the Arctic, Atlantic, and Pacific Oceans. We then evaluate the effectiveness of these CAs through temporal and spatial comparisons of fishing activity before and after enactment dates, within and surrounding CAs. We further explore how trends in fishing effort can be used to determine two forms of fishing displacement, gear type (i.e., increased legal fishing in CAs) and spatial (i.e., increased fishing in areas surrounding CAs) (Vaughan, 2017), and to detect fishing the line behavior. Finally, we evaluate spatial attributes of CAs to explore whether certain features such as boundary complexity and distance to shore influenced illegal fishing effort. The fishing effort estimates presented here can be directly applied to inform management effectiveness of the CAs, as well as to interpret the results of ecological performance monitoring.

| Marine conservation areas
We used the same criteria for CA inclusion as in Iacarella et al. (2023) extended to Canada's three oceans ( Figure 1). Specifically, any CA that was static, geographically defined, and enacted prior to 2020 (1981-2019) was included for analysis. This led to 5 CAs in the Arctic region, 47 in the Atlantic, and 185 in the Pacific (seasonal fishing closures were not considered) ( Figure 1). The areas were conserved as Oceans Act MPAs (n = 13), Marine Refuges (n = 59), Rockfish Conservation Areas (Pacific only, n = 162), Fisheries Closures (n = 2), and one Pacific National Marine Conservation Area. Fishing regulations specific to each CA were compiled and verified (see Appendix S1 for further information; Iacarella et al., 2023). Ten Oceans Act MPAs had either separately managed areas (i.e., Eastport MPA), spatially distinct designated zones, or multiple zones with different regulations; these were assessed as separate CAs (n = 29). Two Pacific CAs protected seamounts that were also included individually (n = 8), for a total of 264 assessed CAs.
Fishing activity within and outside of CAs was compared by creating four spatial boundaries for each CA including (1) the CA area; (2) the perimeter as 100 m inside and outside the CA; (3) a 1 km buffer extending outside the CA; and (4) a 5 km buffer area extending 1-5 km from the CA (Appendix S1). We used surrounding buffer areas instead of other unprotected locations within the Exclusive Economic Zone to compare to CAs so that we could evaluate fishing displacement and fishing the F I G U R E 1 Static, geographically defined conservation areas (blue polygons) with fishing regulations across Canada's (a) Arctic, (b) Pacific, and (c) Atlantic Oceans. line behavior. CA sizes varied greatly (median = 15 km 2 , mean = 2441 km 2 ± 1555, ±1 SE), but many were small, nearshore, and spatially aggregated ( Figure 1). We selected buffers that extended 5 km beyond the CA (1 and 5 km buffers combined: median = 170 km 2 , mean = 619 km 2 ± 148) as an area that could reasonably be applied across all CAs for standardized comparisons (Bucaram et al., 2018). This extent is in line with other studies examining fishing displacement from CA enactments (0-10 km bin in Boerder et al., 2017;3.5 km in Kleiven et al., 2019;1 km in Tassetti et al., 2019). When spatial overlap occurred between a CA (X) and the buffer of another CA (Y), the overlapping buffer area of Y was excluded from the time series based on the date when CA X was enacted; any overlap between buffers was retained and considered as separate areas as these would both be hypothesized to have increased fishing from displacement (see further details in Appendix S1). The 1 and 5 km buffers were selected to test for fishing displacement, whereas the perimeter area was designed to encompass the CA boundary and any potential fishing the line behavior. Other studies on fishing displacement and fishing the line have found higher fishing effort within 1 km of the CA (Boerder et al., 2017;Kleiven et al., 2019;Tassetti et al., 2019). We expected fishing displacement and fishing the line would result in fishing effort density to be highest in perimeters, followed by 1 km buffers, and 5 km buffers, particularly for regulated (i.e., "illegal") fishing.
Spatial attributes of CAs such as the complexity of the boundary and CA size have previously been shown to contribute to more illegal fishing (Crawford et al., 2004;Read et al., 2011;Rojo et al., 2019). We expected that more illegal fishing would occur in CAs that are more difficult to comply with (i.e., lower area to perimeter ratio, complex shape) or enforce (i.e., larger, further from shore). We calculated spatial attributes based on the CA boundaries using Polygon Complexity QGIS 2.0 plugin (https://github.com/pondrejk/PolygonComplexity; Brinkhoff et al., 1995). Attributes were calculated separately and averaged for CAs with spatially distinct areas and included size, perimeter length (e.g., circumference), area-perimeter ratios, distance to shore from the closest CA boundary, deviation from compactness (i.e., increases with deviation from a circle), and complexity (i.e., increases with shape irregularity; Brinkhoff et al., 1995).

| Satellite and terrestrial AIS data
Satellite and terrestrial data were processed and analyzed following Iacarella et al. (2023) and briefly summarized here. Historical fishing effort within and surrounding CAs was modeled from satellite AIS data from the Canadian Space Agency (2012-2016) and Global Fishing Watch (GFW; 2017-2019), and from terrestrial AIS data collected by the Canadian Coast Guard (2012-2019). Only vessels that had a point location within 20 km of a CA from 2012 to 2019 were retained for processing (satellite AIS Maritime Mobile Service Identities, n = 42,130; terrestrial AIS, n = 32,043). This reduction of data included vessels beyond the spatial extent of analysis (5 km buffers around CAs), but was initially set to provide flexibility in determining an appropriate buffer size. Once these vessels were selected, the full tracking history was used to estimate fishing activity from movement patterns.
Fishing effort (hours vessels spend actively fishing) was estimated from convolutional neural network algorithms and vessel registries as developed by GFW (Kroodsma et al., 2018;McCauley et al., 2016). Models determined whether a vessel was actively fishing and what gear class (of 16) it was likely using (Appendix S2). These gear classes were compared to each CA's regulations to classify estimated fishing effort as illegal, legal, or "remainder" (categories referred to as "fishing activity"; Appendix S3) . Remainder fishing occurred when the level of specificity of the GFW gear classifications and the CA regulations could not be matched (see further description in Iacarella et al., 2023). Across all gear classes and CAs, there were 1618 counts of gear classes matched to regulations as illegal, 1335 as legal, and 1265 as remainder.
All assessed CAs allowed Indigenous food, social, and ceremonial fishing and most allowed other forms of fishing. Eleven CAs (including zones separately) were otherwise no-take across the entire conservation area (the Arctic Tuvaijuittuq MPA) or within zones (nine zones within six Oceans Act MPAs and a National Marine Conservation Area in the Atlantic and Pacific). Any detected fishing activity within these no-take CAs was classified as illegal.

| Mapping and analysis
Mapped fishing locations that were within the CAs and buffers were selected for analysis and binned into 0.01 cells for the CAs, perimeters, and buffer extents (Appendix S1). Total fishing effort hours per month were calculated for each GFW fishing gear class from 2012 to 2019, and then for each fishing activity category (illegal, legal, and remainder) specific to the CA. These categories were applied consistently before and after enactment, and within and surrounding each CA to enable temporal and spatial comparisons. Temporal trends in fishing activity for each CA were visualized using timeseries plots of fishing effort before and after enactments (total hours/10 km 2 /# years) for the CA, perimeter, and buffer extents. In addition, gridded maps were produced showing the sum of fishing hours by 0.02 cells across CAs and buffers for each fishing activity category, as well as for before and after enactment, and the difference between the two (after-before, when available). Timeseries plots and gridded maps were made for all CAs with any detected fishing within the spatial boundaries (n = 245).
Quantitative comparisons of fishing effort were made between CA, perimeter, and buffer spatial extents encompassing the same timeframes. This enabled comparisons of spatial shifts in fishing activity without the confound of increasing AIS use and satellite coverage over time. For these comparisons, we applied a log response ratio of total fishing effort per km 2 in the CA compared to the perimeter, 1 km buffer, and 5 km buffer, with a detection limit (0.00001) added to 0 values. Before and after enactment periods were calculated separately to determine whether higher activity in an area was an effect of CA enactments or a pre-existing pattern. Thus, log response ratios were calculated for CAs enacted within the 2012-2019 timeframe and with some fishing detected before or after enactments (n = 63 of 68 enacted within the timeframe). Paired t-tests were conducted on the before/after log response ratios to test for significant differences with enactments for illegal, legal, and remainder fishing by ocean region. Furthermore, we plotted fishing effort density (h/km 2 /year) after enactments within CA and perimeter extents in relation to spatial attributes of all CAs. After visual assessment of trends, we determined statistical analysis was not merited. Spatial analyses were done using QGIS (V2.0), ArcGIS (V.10.8), Python (V3), and Google BigQuery. Data compilation, plotting, and statistical analyses were done using R (R Development Core Team, 2021) and plotly (Sievert, 2020).

| RESULTS
The number and density of vessels identified as fishing within Canada's CAs generally increased from 2012 to 2017, with the maximum monthly count reaching 14 in the Arctic, 191 in the Atlantic, and 232 in the Pacific (Figure 2a). The density of fishing vessels in CAs was notably higher in the Pacific than in the Arctic and Atlantic when accounting for total CA extents within ocean regions by 2020 (Figure 2b).
Of the 264 CAs, 73% had fishing effort detected within the CA extent (not considering enactment dates) from 2012 to 2019. Fishing effort after enactments was detected within 71% of CAs, with illegal fishing estimated after enactments in 33%. Half of all estimated fishing effort within CAs after enactments was legal (54%, 12,084 total hours of fishing), 22% was illegal (4822 h), and 24% was classified as remainder (5385 h). The highest . Illegal, legal, and remainder fishing were categorized based on CA regulations enforced after enactment. Note, "illegal" fishing was not illegal before enactment or in buffers, but was categorized as such to enable comparison to the CA after enactment. Lines represent fishing effort before and after enactment (total effort before or after/# years). Timeseries plots including monthly values for all CAs are in Appendix S5. estimated illegal fishing effort density within CAs after enactments was found for Atlantic CAs (Figure 3). Arctic CAs had low fishing effort density compared to the Atlantic and Pacific, with none detected in Arctic Oceans Act MPAs (Figure 3; Appendix S4). The highest estimated illegal fishing effort density after enactments was in Atlantic Marine Refuges and the Pacific National Marine Conservation Area. Legal fishing was also high in Atlantic Marine Refuges and Oceans Act MPAs, and Pacific Rockfish Conservation Areas (Appendix S4; further Pacific results are in Iacarella et al., 2023).
Timeseries plots and gridded maps of fishing effort for each CA provided visual comparisons of fishing activity before and after enactments and within CA and buffer extents. We highlighted six different CA examples to showcase interesting trends in activity and the usefulness of this presentation for evaluating and reporting on the management effectiveness of CAs (see Appendix S5 for   (c)) and surrounding 1 km and 5 km buffers. St. Anns Bank MPA buffers were only delineated for the portion outside of the CA as the rest of the CA is nested within another CA zone. Illegal, legal, and remainder fishing were categorized based on CA regulations enforced after enactment. Note, "illegal" fishing was not illegal before enactment or in buffers, but was categorized as such to enable comparison to the CA after enactment.
all CA timeseries plots, including monthly fishing effort values, and maps). The Pacific Hecate Strait and Queen Charlotte Sound Glass Sponge Reefs MPA Adaptive Management Zone is an example of a CA that had mostly remainder fishing effort identified, which declined in the CA and increased in the buffers following enactment (Figure 4a). The Atlantic Conservation Areas of Sambro Bank and Western Emerald Banks provided contrasting examples of shifts in estimated illegal and legal activity with enactment (Figure 4b,c). In the former, all fishing effort, including estimated illegal activity, increased in the CA and buffer extents after enactment, whereas in the latter, all effort declined. The maps of differences in fishing effort from before to after enactments highlighted locations of increased and decreased effort within CAs and buffer extents ( Figure 5). For Atlantic Conservation Areas, Emerald Basin showed areas of increased estimated illegal activity within all extents, whereas Jordan Basin had patches of both increased and reduced remainder activity particularly around the 1 km buffer (Figure 5a,b). Conversely, the Atlantic St. Anns Bank MPA Zone 3 had mostly legal activity within the CA with areas of increased and decreased fishing effort throughout (Figure 5c).
Fishing effort density within CAs compared to surrounding areas before and after enactments varied depending on the ocean region and fishing activity. Overall, levels of illegal and legal fishing activity did not change significantly from before to after enactments across the three oceans (paired t-tests, p > .10) (Figure 6). The 5 km buffers generally had the highest fishing effort density relative to the CAs before and after enactments across oceans, whereas perimeters tended to have lower levels. Most notably, higher illegal fishing was estimated in Arctic CAs than in the 1 km buffers after enactments ( Figure 6). In the Atlantic, remainder fishing effort tended to be higher in CAs than in the 1 km buffers before enactments, and the converse was found after enactments (t 25 = À1.85, p = .076). The higher proportion of remainder fishing in 5 km buffers than in CAs before enactments in the Atlantic tended to increase after enactments (t 25 = À1.94, p = .064) ( Figure 6). Remainder fishing was higher in Pacific 1 km buffers than in the CAs both before and after enactments, though the difference between the two extents tended to decrease after enactments (non-significant, p > .10). Mean fishing effort generally increased for all spatial extents and fishing activities after enactments (Appendix S6), though this must be understood in the context of likely increased AIS use by fishing vessels and satellite coverage over the timeframe (Figure 2). Contrary to our predictions, fishing effort after enactments within CA and perimeter extents did not exhibit any apparent trends based on spatial attributes of size, area-perimeter ratios, deviation from compactness, or complexity ( Figures S7-S10). Fishing effort in the perimeters also did not reveal any patterns that might be expected with fishing the line behavior ( Figure 6; Appendix S7-S10). However, the CAs with higher levels of legal and remainder fishing tended to be closer to shore, whereas the highest estimated illegal fishing was further from shore but in Fishing activity LRR of fishing effort (hr/sq km) Enactment before after LRR of fishing effort (hr/sq km) F I G U R E 6 Spatial comparison of fishing effort within conservation areas (CA) versus their respective perimeter and buffer extents before and after enactments as measured by the log response ratio of fishing effort (i.e., ln [CA/buffer]). Values above zero indicate fishing effort was higher in the CA than in the perimeter or buffers. A detection limit (0.00001) was added to extents or fishing activities (illegal, legal, remainder) with no measured fishing effort to enable comparison. Bars are mean ± 1 SE for CAs with some measured fishing activity before enactments (n = 60). a CA that also had high levels of legal fishing (Atlantic Lophelia Coral Conservation Area) (Appendix S10).

| DISCUSSION
For the 2012-2019 timeframe, almost 5000 h (more than half of a year) of illegal fishing effort was estimated after enactments of CAs across Canada's three oceans. There were over 17,000 combined hours (approximately 2 years) of additional fishing that included legal activity, as well as potentially illegal activity that could not be fully matched to regulations owing to further detail needed in gear classifications compared to CA regulations. Using these estimates, we explored questions of CA effectiveness, CA impacts on fishing displacement, fishing the line behavior, and relationships between fishing activity and spatial attributes of CAs. Overall, we found (1) CA effectiveness varied greatly but appeared to be lower in Atlantic CAs owing to higher illegal fishing effort density, (2) fishing displacement following enactments was not observed as shifts towards more legal fishing in CAs or regulated fishing in buffers, particularly as buffers often already contained higher fishing effort than CAs prior to enactments, (3) fishing the line was not indicated by higher fishing density in perimeters, and (4) spatial attributes of CAs did not explain fishing activity trends. These results inform the effectiveness of CA management, the impact of CAs on fishing behavior in surrounding areas, and reasonable expectations for ecological performance of CAs . Our analytical summaries for a nation's full suite of marine CAs represents a major step-forward for monitoring and evaluation of fishing activity that is critically needed to ensure that these areas are meeting their intended conservation objectives.

| Conservation area effectiveness and impacts on fishing activity
Comparisons of fishing effort before and after enactments, and fishing trends following enactments, provide insight on the management effectiveness of CAs, as well as the potential for ecological benefits. Fishing effort timeseries, as shown here, can be directly compared to measures of ecological performance to understand trends in indicators such as fish biomass after CA enactments (Bergseth et al., 2015;Giakoumi et al., 2017;Gill et al., 2017). In addition, estimates of ongoing legal fishing may be as important to monitor as illegal fishing to understand ecological performance as partial CAs are known to be less beneficial to biomass (Aburto-Oropeza et al., 2011;Lester & Halpern, 2008;Sala & Giakoumi, 2018).
Differences in fishing effort from before to after enactments and within to outside CAs identify how much change in pressure is conferred by the CAs. For instance, no change in fishing effort signifies little expected benefit from fishing reduction even if fishing effort is very low. We found that remainder fishing in Atlantic CAs declined compared to the 1 km buffer after enactments, suggesting shifts in illegal and/or legal fishing effort to outside the CAs and a benefit of CA enactments. Also, in particular examples such as the Hecate Strait MPA (Figure 4a), we found a decline in remainder fishing in the core and an increase in the buffers after enactment. This shift in remainder fishing indicates it was likely primarily activity that was made illegal following enactment. Others who have assessed fishing displacement from CAs have found shifts from the CA extent to directly outside (e.g., within 1 km, Kleiven et al., 2019;Tassetti et al., 2019), with fishing declining beyond 10-40 km (Boerder et al., 2017;Bucaram et al., 2018). However, aside from such exceptions, we did not find strong differences in fishing activity spatially after enactments as fishing tended to be higher 1-5 km outside the CAs both before and after enactments. Using the same AISbased method as shown here, White et al. (2020) similarly found four of five large offshore Pacific CAs had no change in fishing effort after enactments because fishing was already minimal in those areas. This may indicate that CAs were placed in areas that were less intensively fished, which can be a strategy used to gain support for CAs from fishers (Read et al., 2011). The socio-economic impact of CAs on fishers is an important consideration in marine CA design that influences compliance levels (Iacarella et al., 2021;Read et al., 2011), though it also presents a trade-off by potentially placing CAs in locations where they are less likely to improve ecological performance (Ziegler et al., 2022). However, the full benefit of CAs will increase over time if fishing and other human pressures are effectively mitigated in CAs while activities expand in surrounding waters (Grorud-Colvert et al., 2021;Magris, 2021;Miller et al., 2018;White et al., 2020). Ongoing monitoring and assessment of fishing effort, and other human pressures, surrounding CAs is important for understanding and tracking the benefit of CAs (Ziegler et al., 2022).
We did not find overarching evidence of fishing displacement, fishing the line behavior, or a strong influence of spatial CA attributes on fishing activity. These results are likely owing to a variety of factors including the detected presence of ongoing illegal activity within CAs, as well as some apparent placement of CAs in areas that tended to have less fishing originally. In particular, the lack of evidence for fishing the line-from lower fishing effort density in perimeters and higher effort in 5 km buffers relative to CAs-corresponded with no relationship with spatial attributes that are most relevant to fishing the line (i.e., CA area-perimeter ratios, deviation from compactness, and complexity). In the event of fishing the line, CAs that are high in perimeter to area ratio or complex in shape may have more illegal fishing owing to accidental crossing into the CA and from greater difficulty in surveilling complex boundaries (Read et al., 2011). The one attribute that matched predictions was the tendency for higher effort in nearshore CAs for legal and remainder fishing but not for illegal fishing, which is likely a result of ease of access and, in the case of illegal fishing, more exposure to surveillance and public witness (Crawford et al., 2004). Finally, individual assessments of CAs through the timeseries plots and fishing effort maps provide more nuanced evaluation of shifts in behavior than summaries across oceans, in light of the great variety of regulations and allowances across the 264 assessed CA areas (Appendix S5). Despite the general lack of evidence for fishing the line behavior or an effect of spatial attributes, CA design considerations such as boundary and zone complexity, and CA size, are still important as this can influence ease of compliance and surveillance (Read et al., 2011;Rojo et al., 2019).

| Monitoring and reporting on fishing activity in conservation areas
Patterns in fishing effort are not wholly captured by AIS, and ideally multiple vessel tracking methods would be compiled to comprehensively monitor ongoing activity in CAs Park et al., 2020;Rowlands et al., 2019). AIS is required by the International Maritime Organization for large vessels (over 300-500 tons) for the purposes of vessel communication and navigational safety, and some countries have additional carriage requirements for fishing vessels (Robards et al., 2016). The observed increase in fishing vessel detections within Canada's CAs, particularly from 2012 to 2017, may be from a combination of more AIS carriage and satellite coverage McCauley et al., 2016). Canada does not mandate AIS use for fishing vessels, and instead has relied on logbooks, Electronic Monitoring System (groundfish fisheries, Pacific region only), VMS (select fisheries, primarily Atlantic region), and aerial surveillance for fishing vessel tracking . These data sources can be used on a regional basis to better capture the full picture of vessel activity as there are a significant number of vessels in Canada's waters not using AIS (Burke et al., 2022;Iacarella et al., 2023;Serra-Sogas et al., 2021).
Aside from AIS, VMS is generally the only other tracking source that currently can be analyzed at the quantitative level shown here and by others (Dureuil et al., 2018;Kroodsma et al., 2018;McCauley et al., 2016;McDermott et al., 2019;White et al., 2020). VMS is solely required for commercial fishing vessels so fishing by any other vessel types is not captured, data transmission rates are lower (e.g., by hour instead of seconds), and data use can be highly restricted McCauley et al., 2016). However, VMS has been valuable for monitoring and enforcing CAs in countries with extensive carriage mandates (Read et al., 2019), and can provide important information on activity that may be missed in AIS data (Chang & Yuan, 2014;Lee et al., 2010;Magris, 2021). For instance, Magris (2021) found ongoing fishing activity and mixed CA effectiveness in Brazil CAs using VMS, whereas Kroodsma et al. (2018) detected minimal fishing activity in the area using AIS. As such, our estimates of fishing effort within and surrounding CAs are likely very conservative, and highlight the need for increased AIS carriage mandates.
The summaries and comparisons of fishing activity in CAs presented here provide a national level perspective of management effectiveness and an indication of behavioral responses to CAs. Temporal and spatial patterns can be viewed at the level of individual CAs for monitoring and management, and for CA-specific evaluation of fishing trends based on CA design and other performance questions of interest. Such monitoring and evaluation of ongoing fishing activity, and other human pressures, in CAs is imperative to ensure these areas are effective conservation tools , rather than "well intended" but ineffective measures (Rife et al., 2013).

ACKNOWLEDGMENTS
We thank many people across several branches of Fisheries and Oceans Canada (DFO; Marine Security and Operations Centre, Ecosystems and Oceans Sciences, Fisheries Management, Ecosystems Management, Canadian Coast Guard) and the Canadian Space Agency for providing data, advice, and information review. We also thank the reviewers who provided feedback that greatly improved 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
Processed 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.