Monitoring data for a new large offshore marine protected area reveals infeasible management objectives

Predicting and measuring changes resulting from marine protected areas (MPAs) has posed a challenge for practitioners, partly because ecosystems are complex and can change in unanticipated ways, but also due to MPA characteristics such as design factors, conservation objectives (COs), and monitoring programs, that can leave little chance of meeting stated goals. We consider these design factors for the Laurentian Channel MPA, a large offshore Canadian protected area established to protect against fishing impacts. Specifically, in this study we evaluated (1) whether it is realistic to expect improvements in the MPA for four previously established taxa‐specific COs, and (2) whether existing scientific surveys are capable of detecting changes in these CO taxa even if they occurred. Three CO species were sampled in scientific multispecies research vessel trawl surveys (Black Dogfish, Smooth Skate, and Northern Wolffish) and a fourth CO, sea pen taxa, were enumerated using seafloor imagery. Simulations indicate that trawl surveys have very little chance of detecting change in the abundance of the three fish species examined, while seafloor imagery data had higher statistical power for sea pen taxa. Moreover, we show that expecting change related to the removal of fishing is unrealistic due to the fact that the MPA was established in an area of minimal fishing pressure. While positive change is unlikely to be induced by the MPA, or be detected if they occurred, this MPA could provide conservation benefits if COs and monitoring approaches were realigned to match the unique features of this area that represents largely unimpacted sensitive benthic habitats.

caused by harmful anthropogenic impacts such as overfishing or overuse (Jamieson & Levings, 2001).The International Union for Conservation of Nature (IUCN) has led the global marine conservation agenda and fostered political support towards marine protection (Brooks et al., 2006;Harrop & Pritchard, 2011).Consequently many countries, including Canada, support the Convention on Biological Diversity (CBD), Global Biodiversity Framework (GBF), which aspires to protect 30% of marine habitats by 2030 (GBF, 2022).Towards these targets the global coverage of MPAs is currently 8.19% and increasing, and while 18.69% of national waters are protected only 1.44% of areas beyond national jurisdiction are protected (Protected Planet, 2021).Unfortunately, MPA establishment does not guarantee conservation success towards established goals.Indeed, some reports indicate that less than one third of existing MPAs are considered effective, with the majority failing to achieve their expressed management goals and objectives (De Santo, 2013;Kelleher et al., 1995;Mascia et al., 2014;Pomeroy et al., 2005).There are numerous reasons MPA expectations are not met, such as inadequate habitat for target species (García Charton et al., 2000), unrealistic conservation objectives (COs) (Claudet, 2018;Muntoni et al., 2019;Nickols et al., 2019), inadequate or unrealistic monitoring study design (Ahmadia et al., 2015), or inability to enforce MPA restrictions (Giakoumi et al., 2017) among others.Identifying such challenges and alternate pathways to success are important to maximizing the probability that MPAs will provide societal value.
The Laurentian Channel MPA (LCMPA) is one of Canada's largest MPAs, protecting 11,580 km 2 , and is among the largest 5% of no-take MPAs worldwide (UNEP WCMC and IUCN, 2022).It was designated in 2019 to alleviate the impact of commercial fishing activities on six taxa that were considered ecologically and/or economically valuable and in need of protection (Canada Gazette, 2017).These taxa include corals (in particularly significant concentrations of sea pens -Superfamily Pennatuloidea), Black Dogfish (Centroscyllium fabricii), Smooth Skate (Malacoraja senta), Porbeagle Shark (Lamna nasus), Northern Wolffish (Anarhichas denticulatus), and Leatherback Sea Turtle (Dermochelys coriacea) (Canada Gazette, 2017).With the LCMPA established, Fisheries and Oceans Canada (DFO) is now tasked with science-based monitoring to evaluate the success of the LCMPA's COs.However, the likelihood of achieving and detecting success related to established COs has yet to be assessed.Reductions in the prospective MPA's size during the establishment process undermined potential for ecological improvement (Muntoni et al., 2019).MPAs are often assessed against comparable unprotected habitats outside their boundaries (Rowley, 1994;S anchez Lizaso et al., 2000), but the exclusion of important fishing grounds from the MPA during the establishment process caused uncertainty as to whether such an inside-outside approach is now appropriate for the LCMPA.Though the LCMPA region has long been monitored with DFO's standardized multispecies trawl survey (McCallum & Walsh, 1997;Rideout & Ings, 2018), it remains unclear whether this foundational method has sufficient statistical power to detect change and inform adaptive management of COs or whether alternate methods (e.g., camera surveys) would better serve CO monitoring objectives.
In this study we (1) use research vessel multispecies trawl survey data, collected before the establishment of the MPA, to compare fish CO abundance and community metrics inside and outside of the MPA, with the goal of assessing the appropriateness of using such comparisons to evaluate MPA success; (2) use these multispecies trawl data, in addition to existing imagery data, to inform simulations that evaluate their statistical power to detect change in CO taxa inside the MPA; and (3) compare the statistical power of alternate community metrics (e.g., species richness) for detecting change.

| Data collection
DFO, in collaboration with several countries participating in the Northwest Atlantic Fisheries Organization (NAFO), has been conducting annual fish stock assessments through data collection via standard research vessel (RV) multispecies trawl surveys across its continental shelf and slope habitats since the 1970's (Doubleday, 1981).These standard foundational datasets are the most extensive long-term systematic data available for the LCMPA and surrounding areas that DFO have used to measure the abundance, biomass, and distribution of fishes and invertebrates following a randomstratified sampling methodology using a bottom trawl (McCallum & Walsh, 1997;Rideout & Ings, 2018).These surveys also collect bycatch taxa information for sea pens and other corals, and since these surveys are expected to continue inside the LCMPA (DFO 2022) they are considered a potential primary data source for MPA monitoring.
Among the six COs for the LCMPA, only three fish species are reliably sampled in the multispecies trawl surveys: Black Dogfish, Smooth Skate, and Northern Wolffish.The other CO taxa (e.g., Leatherback Turtles, Porbeagle Sharks) require alternative sampling methods to monitor them and are therefore not considered further here.Briefly, annual surveys have been conducted using a Campelen trawl onboard Canadian Coast Guard Ships (CCGS) Wilfred Templeman (decommissioned in 2008), Alfred Needler, and Teleost (Figure 1).Set positions are randomly assigned within each existing depth stratum, with effort proportionately weighted to the area of each stratum (with a minimum effort of 2 sets per stratum).Large strata are divided into approximately 12 km 2 units with a single trawling-set per unit (Table 1).Species' biological data are collected onboard research vessels to provide abundance and biomass estimates for all fish species and some invertebrate taxa collected from successful trawling sets.The number of each species (count data) caught per haul is standardized to 15 min tow durations at 3.0 knots speed ($0.8 nautical miles) defined as the Catch-Per-Unit-Effort (CPUE).We used CPUE data from these trawl surveys to describe and compare spatial and temporal differences in CO fish species abundance, and to parameterize analysis simulations.
The entire LCMPA is sampled on average by 23 trawl sets per year.The spatial footprint of the multispecies trawl surveys inside this MPA is therefore small, for example, 0.025 km 2 per trawl and 0.575 km 2 across all trawls per year (Rideout & Ings, 2018).The impact of this research-based bottom trawling on benthic habitats is considered to have a small footprint in the LCMPA (Benoît, Asselin, et al., 2020;Benoît, Dunham, et al., 2020;DFO, 2018).This is in contrast to commercial trawling, which banned in Canadian MPAs (Clark et al., 2016;DFO, 2019;Thrush & Dayton, 2002).
In situ camera surveys (e.g., seafloor imagery) can yield more accurate abundance data for sea pens than trawls (e.g., Chimienti et al., 2018;de Mendonça & Metaxas, 2021).Sea pen (and other coral) abundance data are not consistently collected as part of DFO trawl surveys (e.g., Gullage et al., 2022), and therefore not suitable for the trawl-based analysis conducted here.However, imagery data available from a separate camera survey that was conducted in June 2015 is included in our analysis as part of this study.The CCGS Hudson was used to collect high resolution still camera images using 4 K camera systems at 10 transects in the LCMPA (Figure 1) (Beazley et al., 2015).The camera transects ranged between 0.9 and 2 km in length, although most were $ 1 km, at surveyed depths of 328-467 m (Table 1).With this system, a photograph was taken with a downward-looking camera every time the system's trigger weight came in contact with the seafloor, with the number of photos per transect ranging between 30 and 50 (mean of 40 photos per transect) (Table 1) and spaced at $25 m intervals.However the distance between photos was variable.For each image, sea pens were counted  and identified to the lowest possible taxonomic level.Imagery annotation was conducted by the same individual for all photos.We incorporate these data within a similar power analysis approach as applied to the trawl data.

| Assessing the appropriateness of inside-outside comparisons
The establishment of the LCMPA was anticipated to promote the replenishment of depleted stocks within its boundaries and generate potential spillover to surrounding areas (Canada Gazette, 2019).Measuring success requires comparisons of CO taxa within the LCMPA to those outside.However, for such comparisons to be appropriate, a control area with common environmental conditions and species assemblages is required (Underwood, 1996).We assessed differences (catch per unit effort, CPUE) in select CO taxa inside and outside (North Atlantic Fisheries Organization Convention areas 3Pn and 3Ps) the LCMPA before the establishment over two time periods (3Pn 1996-2013, 3Ps 1996-2019) using generalized linear mixed-effect models (GLMMs).We also analyzed a subset of this dataset from 3Ps from a recent five-year time period (2015-2019) to evaluate when change may be detectable in the context of ongoing LCMPA reporting, expected to occur on 5-year cycles.The model included CPUE as the response variable, year as the explanatory variable, and depth strata as a random effect.The model was run using the glmmTMB package in R (Bolker, 2016).First, the data were analyzed using a Poisson error distribution.In cases of under-dispersion (dis-persion<1) we fitted a quasi-Poisson error structure, whereas overdispersion (dispersion >1.0) was addressed using a negative binomial model (Zuur et al., 2009).Zero inflation was examined by simulation using the DHRMa package (Hartig, 2017).Models were derived separately for each protected fish species considered (Black Dogfish, Smooth Skate, Northern Wolffish).

| Power analysis
Power analyses were used to assess the likelihood of detecting an effect within the LCMPA over a range of simulated change scenarios for the targeted CO fish species (Black Dogfish, Smooth Skate, and Northern Wolffish) and species richness based on multispecies trawl survey data, and sea pen abundance based on seafloor imagery data.Only data from inside the LCMPA were considered when establishing the variance parameters of baseline conditions that were applied using the GLMM modeling approach.In an ideal scenario, we would consider areas outside of the MPA as well, and conduct analysis as part of a Before-After-Control-Impact (BACI) design.However, following our Objective 1 assessments, we did not consider the areas outside the MPA to be a suitable control.Therefore, we restricted our analysis to data scenarios related to Before-After MPA establishment comparisons.These generated datasets were based on empirical data that were manipulated to create simulated effects over time.Statistical power varies in response to sample size, the accepted Type I error rate, α (e.g., the likelihood of falsely detecting a change), the detectable effect size, the statistical test employed, and the variability of the data.All things being equal, statistical power (the inverse of Type II error rates, 1 -β) improves when one or more of sample size, detectable effect size, and Type I error increase.In contrast, statistical power decreases as variability increases.We set standard power (0.8) and acceptable Type I error rates (0.05) (Di Stefano, 2003;Peterman, 1990), and explored the relationship between sample size (to inform monitoring feasibility) and detectable effect size across CO change scenarios.
Two simulation processes were conducted, one informed by data from the multispecies trawl surveys from 2015 to 2019, and a second that included sea pen abundance from seafloor imagery.These simulated data sets represent the simulated baseline ("before") and simulated change ("After") data used in the before/after statistical comparisons and related power analysis (further described in the next section).The multispecies trawl survey data analysis simulation scenarios included four sample sizes (15, 23, 50, and 100 trawls), where 15 trawls is considered a low survey effort, 23 trawls represents the current average number of trawl sets per year inside the LCMPA, and 50 and 100 trawls as high and very high levels of survey effort.The analysis modeled reductions in fish abundance relative to simulated baselines (e.g., CPUE reductions of 40%-90% from baseline) needed to detect significant change over monitoring time periods (3-10 years) relevant to management decision making.A decrease in the CO species abundance was chosen as our trend direction because measuring a negative impact would be required to trigger adaptive management.

| Generating simulated datasets
For each of the four sample size scenarios considered for each CO species, 1000 simulated datasets were generated by randomly sampling a negative binomial distribution parameterized by the empirically-based strata-specific mean CPUE and the fixed negative binomial dispersion parameter acquired from existing multispecies trawl survey data.The strata-specific means were derived for each stratum from 2015 to 2019 multispecies trawl survey data from inside the LCMPA (N = 9 strata) but artificially adjusted over time across a range of effect size scenarios (0.1 = very small effect, 0.5 = very large effect).The dispersion parameter was derived from the residual variation remaining from the empirical comparison of two time periods of multispecies trawl survey data: 2010-2014 and 2015-2019 using GLMM in lme4 package (Bates et al., 2015); with abundance (e.g., CPUE) as a response variable, time period as a fixed effect, and strata as a random effect.The two 5 year periods represent the duration of suggested MPA management plan cycles.

| Modeling simulated datasets
These datasets generated from simulated temporal changes to CPUE were then tested with a GLMM and associated likelihood ratio tests, incorporating a negative binomial error structure using the lme4 package (Bates et al., 2015), with CPUE as the response variable and time period (Before/After) as the explanatory variable.Significance values associated with each iteration of the simulated data in each scenario were stored and the scenario-specific proportion of significant likelihood ratio test results (alpha = 0.05) across all iterations were used to represent that scenario's statistical power.The CPUE of tested species and taxa in multispecies surveys varied across sets and strata, with some strata rarely serving as habitat for some CO taxa.Therefore, we conducted the power analysis for two scenarios: one in which data from all strata were used, and a second, in which we restricted analyses to core strata that excluded strata with a mean baseline abundance in the lowest quartile.
When change accumulates over time, the ability to detect it increases over that time period because the cumulative effect size and the number of data points in the analysis become larger.This is relevant, for example, when Canadian MPAs are examined based on the recommended 5-year management plan cycle (Canada Gazette, 2017).To explore changes within such a timeframe, and in addition to a Before-After comparison, we explored a time-series model for which we evaluated whether, and when, the slope of the temporal variable significantly differed from zero.Having an ability to spread sampling effort over time is a practical monitoring consideration, especially if stressors are likely to cause incremental change.Therefore, we simulated such time series as we described previously for Before-After comparisons, except our predetermined change accumulated over time.Through this approach we could assess how many years of monitoring would be required to detect a specific rate of change.
The power analysis of sea pen imagery data was performed in the same way as the multispecies trawl survey data analyses, with a few differences limited by the type and availability of sea pen data.The sea pen data were collected in 2015 only within the MPA.Like other metrics, the field data were used to establish the baseline and dispersion parameter estimate, and simulations were used to generate the power analysis datasets.The simulation included sample sizes ranging between 10 and 50 transects, each consisting of 40 photos.Simulated effect sizes were the same as that used for multispecies trawl survey simulations (sea pen density reductions of 20%-80% of baseline needed to detect change).The scenarios were analyzed following the same Baseline-Impact approach as that used for multispecies trawl survey data and the resulting simulated datasets were tested with the same GLMM approach, via the likelihood ratio test using the negative binomial error structure from the lme4 package (Bates et al., 2015).The dependent variable was the sea pen abundance and the independent variable was the temporal condition (baseline/impact).Transect was treated as a random effect of the intercept.Each scenario was simulated with 1000 iterations and was assessed for statistical significance (at P = .05level) within the GLMM framework.

| Black Dogfish
The multispecies trawl survey data indicate that Black Dogfish are distributed widely across 3Pn and the LCMPA.CPUE data for Black Dogfish indicate greater decreases in abundance inside the MPA than in areas adjacent the MPA over the entire time period considered (1996-2019) (Figure 2; Table 2).While there was a general increase in abundance in more recent years, from 2015 to 2019, the increased CPUE was much greater outside the prospective MPA than inside (Figure 2, Table 3).The mean CPUE of Black Dogfish in the prospective MPA decreased from 200.6 in 1996 to 40.2 individuals per haul in 2019, averaging a 9% decrease per year; however, it increased 11.3% between 2015 and 2019.

| Smooth Skate
The regional abundance of Smooth Skate was higher within the LCMPA than in areas along the south coast of Newfoundland (3Pn, and 3Ps).However, the amount of change observed for Smooth Skate abundance appears greater in areas outside the MPA based on multispecies trawl survey data considered (Figure 2; Table 2).For example, during the most recent period, from 2015 to 2019, the CPUE data suggest that Smooth Skate abundance tended to decrease, declining 2.2% inside the MPA and as much as 20.5% in 3Ps from year to year (Table 3).The CPUE of Smooth Skate caught in the MPA slightly increased from 2.3 to 2.9 (2.2%) from 1996 to 2019.

| Northern Wolffish
The CPUE of Wolffish was very low throughout the years and areas considered (Figure 2).Of 3613 hauls conducted, Northern Wolffish were only sampled in 0.9% of the sets conducted inside the MPA, 0.2% in 3Pn, and 0.7% in 3Ps (Table 1), offering limited data to consider changes in abundance.

| Sea pens
The seafloor imagery analysis detected sea pens in 30% of the 418 analyzed images (Table 1).Sea pen taxa include Pennatula spp.(82.6%),Anthoptilum sp.(13.5%),Kophobelemnon sp.(3.5%), and sea pens identified as sea pen sp.(0.4%) due to difficulties with species level identification from imagery.Number of sea pens per image ranged between 1 and 24, and number of sea pens per transect ranged between 0 and 458 (transects CON_048 and CON_044, respectively, Table 1).The high number of sea pens in Transect CON_044 included high proportions of juveniles (likely Pennatula aculeata).

| Power analysis of trawl and imagery data to detect change in CO species
Power to detect change in CPUE varied substantially among the species considered and with the number of trawls conducted per stratum.As expected, power increased with greater sample sizes, but in most cases the gains in power reached asymptotes at very high, and likely unrealistic, levels of sampling effort.For example, increases in power associated with increasing trawls from 23 to 50 sets per stratum was higher than the gain from 50 to 100 trawls per stratum, and in some cases the analysis using the highest number of trawls (e.g., 100 trawls) still failed to reliably detect simulated change.For Black Dogfish, change in abundance could only be detected with target power (0.8) when 50 or more The standardized trawling CPUE data for Black Dogfish (top panels), Smooth Skate (middle panels), and Northern Wolffish (bottom panels) in the LCMPA (left panels), and adjacent areas of 3Pn (center panels) and 3Ps (right panels).Red squares denote the annual mean CPUE.Black solid lines and shaded areas indicate the model prediction of CPUE and associated 95% confidence interval, respectively.simulated trawls per year were conducted, and when the effect size (the decline in abundance [CPUE]) was large, approximately 80% decline (Figure 3).The statistical power to detect change based on average sampling of 23 trawls per year was less than 0.5, even when the effect size exceeded an 80% decline in abundance.Power to detect change in Black Dogfish was improved slightly when only core strata were included, but still would require more than doubling the existing survey effort to reach the accepted level of power.Of the CO fish species examined, we observed the greatest power to detect change in Smooth Skate CPUE (Figure 3), for which under existing sampling effort (23 trawls per year), change could be reliably detected (≥0.80 power) for an effect size of >60% decline in CPUE.Restricting the simulation to core strata did not affect statistical power for Smooth Skate.The power to detect change in CPUE was the lowest for Northern Wolffish.Using the existing mean number of annual multispecies trawl survey sets (n = 23), a large effect size representing a 90% decline in abundance would still only have a statistical power of 0.25 (Figure 3).Similar to Smooth Skate, power was not improved with restrictions of sampling to core strata.For comparison to these single species metrics using the multispecies trawl survey data, we also examined whether a common indicator of biodiversity (species richness), using all the fish species in the catch data (Appendix 1 and 2), might have a higher likelihood of detecting change, and found that species richness provided higher power to detect change (Figure 3).For species richness, the lowest level of sampling effort (n = 15) produced acceptable statistical power (0.8) to detect change at smaller effect sizes (50% decline) than that of single species simulations.
For sea pens, we found that seafloor imagery transects data performed better than the multispecies trawl survey data used for CO fish species.Not surprisingly, power increased with more transects (Figure 4).Using 40 transects, a 50% decline in abundance would likely be detected with the target power (0.8) (Figure 4).Using 30 or even 10 transects would require 55% and 75% change in abundance respectively, to reliably detect effects.

| Temporal effectiveness of trawl surveys to detect change in CO species
The trend analysis over multiple years showed that the statistical power to detect change in fish abundance increased with time and sample size, as expected Abbreviations: CPUE, Catch-Per-Unit-Effort; GLMM, generalized linear mixed-effect models; MPA, marine protected area.
(Figure 5).Results from the trend analysis were similar to the time-period comparisons in that COs with the greatest power to detect changes in abundance were Smooth Skate, followed by Black Dogfish, then Northern Wolffish (Figure 5).For Smooth Skate, a decline of 20% per year would be reliably detected after 4 years using current sampling efforts of 23 hauls per year (Figure 5).For Black Dogfish, change is not likely to be detected in less than 5 years even if sampling rates are increased to more than 50 trawls per year.The power to detect change is <0.30 after 5 years with current sampling intensity of 23 hauls per year.Reliable power for this species was only achieved with existing sampling intensity over 8 years and 20% annual change in CPUE.For Northern Wolffish, the power to detect the simulated change in abundance never exceeded 50% at the mean sampling effort of 23 trawls per year (Figure 5).Reliable power to detect change was only achievable for Northern Wolffish when sample sizes exceeded 50 hauls per year, monitoring periods reached 9 years, and simulated rates of change were very high (80% change per year).For species richness, the power was high for all simulations.For example, an acceptable power of 0.8 could be obtained after 3 years at a decline level of 15% per year using current sampling effect of 23 hauls per year (Figure 5).Sea pens were not assessed temporally, as the available data were collected in the same year.

| DISCUSSION
In the Laurentian Channel MPA, efforts to minimize economic impact during establishment resulted in a reduced MPA size that excluded the vast majority of habitat that was considered valuable to fish harvesters (Muntoni et al., 2019).The implications of these choices were apparent in our inside-outside MPA comparisons of our CO taxa in which CO taxa were often more abundant outside the MPA even in the presence of fishing.This greatly limits the potential of the LCMPA to achieve its regulatory COs (Icheli et al., 2004).While the remaining MPA habitat is unique and valuable in its own right, as an area that has experienced minimal fishing pressure (Figure 6), it is not suited to the original MPA goal of reducing fishing-related mortality of CO species or creating spillover into adjacent areas.Rather than relying on six individual species, specified in the MPA regulations, to indirectly support biodiversity objectives, we advise that CO objectives and associated monitoring are realigned to the LCMPA's ecological potential (e.g., conservation of intact habitats and communities).Such alignment between objectives and ecological potential is crucial to having successful MPAs, as positive outcomes are more likely.

| Appropriateness of inside-outside comparisons to assess LCMPA conservation objectives
The inherent differences of CO taxa inside and outside the MPA will confound monitoring assessments that aim to isolate the effects of MPA restrictions through comparisons to adjacent unprotected habitats.Perhaps more troublesome is the fact that since outside areas are more productive for some CO species, there is significant risk that comparisons between inside and outside would foster the (misleading) belief that MPAs are detrimental to CO taxa.Consequently, monitoring of the LCMPA will need to focus on trends that are occurring within the regionally unique habitats of the MPA.While the lack of spatial controls reduces our ability to assess the reason for any observed changes (e.g., MPA-related vs. large scale environmental change), it does have the benefit of allowing limited resources to be focused on assessing change within the MPA.

| Statistical power of existing data to monitor change in CO taxa
Assessing trends within the MPA is also problematic with the existing CO taxa due to poor statistical power.Poor statistical power impairs the ability to accurately identify MPA status and trigger adaptive management of the LCMPA.Since MPA monitoring programs can incur significant costs (Balmford et al., 2004), multi-purpose data collection programs are an attractive cost-saving approach.However, DFO's multispecies trawl survey has poor statistical power to measure change for any of the LCMPA's fish CO'seven for those routinely captured in the trawls (e.g., Black Dogfish, Smooth Skate, and Northern Wolffish).Should the abundance of any fish CO taxa decline for example, the trawl data are ill suited to detect those changes in a timely manner (e.g., within a typical 5 year management plan cycle).Increases in abundances will be similarly difficult to detect with current survey methods.Some CO taxa are inherently challenging to use as indicators of change.For example, the CO taxa we examined include those with naturally low densities (e.g., Northern Wolffish) and those with inherently high variability (e.g., Black Dogfish).Other CO taxa not addressed in this study include sharks and turtles that occur at low densities and are transient through the MPA and not well sampled by multispecies trawl surveys.While sampling methods, study design, and sampling effort can improve statistical power, such changes are not always feasible.In such cases, alternative indicators that are ecologically relevant and more statistically suitable should be considered.For example, community-based trends, such as those described by Koen-Alonso and Cuff (2018), may be better aligned with recent shifts to ecosystem-based management.Certainly, our study showed that multispecies trawl survey data had improved statistical power to measure change in species richness, using all species data available, compared to changes in the abundance of individual CO taxa within the MPA.
Sea pens were a different case, in which the power analysis of imagery data indicated a higher potential to detect change in their abundance in the LCMPA.The results provide confidence that monitoring of these taxa will be useful for detecting change and informing management decisions relevant to MPA objectives.Moreover, imagery surveys are less invasive and can be conducted off a variety of platforms, making it more feasible to increase sampling effort and statistical power than trawl surveys.Current developments in seafloor imagery equipment (Clayton & Dennison, 2017;Dominguez-Carri o et al., 2021) and annotation (e.g., machine learning methods, Ayyagari et al., 2023;Piechaud & Howell, 2022), are expected to facilitate data collection and analysis.We acknowledge that our empirical data for sea pens are based on a single survey and do not capture temporal variation.However, sea pens are sedentary and relatively long-lived (e.g., likely decades, Neves et al., 2015Neves et al., , 2018;;Murillo et al., 2018) and we expect the abundance and distribution of these taxa to be less temporally variable than more mobile taxa such as fish.

| Using power analysis in MPA design and monitoring
Power analysis is an important analytical tool for evaluating monitoring program effectiveness, including those of MPAs (Bacheler et al., 2016;García Charton et al., 2000;Mascia et al., 2017), and should be key element of MPA planning and evaluation.Peterman and M'Gonigle (1992) demonstrated the usefulness of power analysis as a core procedure in fisheries research and management, and indeed wherever statistics are used to test hypotheses.Given the significant cost of establishing and maintaining MPAs, we believe power analyses are a critical step needed to refine and/or validate MPA program design decisions before resource-intensive monitoring programs are initiated.For example, power analyses in this study identified CO taxa for which assessing status and trends, and in turn informing adaptive management, will be inherently difficult.These analyses also manage expectations on the resources, time periods and effect sizes needed to detect change.Fortunately, power analysis tools have become more available and diverse in recent years.Even more complex study designs associated with ecological monitoring can be addressed using data simulation approaches (e.g., Morris et al., 2018) such as used in this study.

| Adaptive management options for the LCMPA
It is not uncommon for MPAs to fall short of conservation expectations, in part because of a disconnect between established MPA objectives, conservation potential, and/or ability to detect desired change (Claudet & Guidetti, 2010;Ferraro & Pattanayak, 2006).The establishment of large MPAs, created to safeguard against stressors that were never prevalent, has been met with criticism and skepticism (Clark et al., 2011) since they do little to address the goals for which they were created (Magris & Pressey, 2018).The LCMPA is a good example of this, and our results further suggest that even if CO's were appropriate the currently used trawl survey monitoring would fail to detect changes in CO species if they occurred.Power analysis simulations such as conducted herein are particularly useful early in the design and establishment process to inform the likely potential of a MPA monitoring program to achieve its objectives.
MPAs can have unanticipated benefits, such as increasing abundance of non-target species (Alemany et al., 2012;Allard et al., 2022;Côté et al., 2001).Due to a low fishing pressure prior to establishment (Figure 6), the LCMPA maintains its value as an important defacto refuge for sensitive benthic habitats.For instance, the prevalence of longlived and sedentary sea pens (Murillo et al., 2018;Neves et al., 2015Neves et al., , 2018) ) in significant concentrations (Kenchington et al., 2016) in the LCMPA is an indication of the suitability of the existing MPA area for their continued persistence.Preservation of these habitats and biological communities within the LCMPA is a more reasonable and measurable CO (Dunham et al., 2020) than the existing COs focused on fish and turtles for which the LCMPA does not serve as core habitat.But until MPA features are recognized and valued by stakeholders through adaptive management processes and tracked with monitoring programs, these values cannot contribute to perceptions of MPA success.
Adaptive management of an established MPA is a time-consuming process that often includes stakeholder consultations and regulatory hurdles, even when potential science-based improvements have been identified.However, it is necessary to refine MPA management over time due to the complex nature of restoring or enhancing ecological communities.When the evaluation of MPA performance against expected outcomes identifies negative outcomes, or if objectives are not easily measurable, adaptive management processes should be readily available and implemented to improve MPAs.Morris and Green (2014) recognized the difficulty of introducing adaptive management actions after an MPA has been created, and recommended that future MPAs consider adaptive management options during the MPA planning stage.The ability to implement regulatory MPA change, as new scientific information becomes available, should be a priority in MPA design and planning, rather than something to be addressed after a problem is identified (Morris & Green, 2014).This is particularly relevant as we enter an era of unprecedented environmental and technological change.Society must adapt to these emerging and often unpredictable changes, and MPA management is no exception.

F
I G U R E 1 Map of sampling site located on the east coast of Canada (red rectangle in inset), including the Laurentian Channel marine protected area (MPA) boundary (thin red thin) and adjacent fishing areas (3Pn and 3Ps).Each dot represents trawl locations conducted from 1996 to 2019 and green crosses indicate sea pen sampling transect locations conducted in 2015.Acronyms of eastern provinces of Canada include: NL: Newfoundland and Labrador; QC: Quebec; ON: Ontario; NB: New Brunswick; PEI: Prince Edward Island; NS: Nova Scotia; and SPM: Saint Pierre and Miquelon (France).

F
I G U R E 3 Statistical power of research vessel trawling to detect a statistically significant change in abundance (CPUE: the number of individual caught in 15 min tow durations at 3.0 knots speed) relative to baseline for Black Dogfish (a), Smooth Skate (b), Northern Wolffish (c) within core strata (i.e., habitats with typically high abundance), and species richness (i.e., using all the fish species in the RV survey data) using different sample size scenarios (15-100 trawls per stratum) and change (abundance relative to baseline).Only trawls from core strata (i.e., excluding strata that typically do not serve as habitat for CO taxa) were included in the analysis.Results are based on simulations informed by data collected from 2015 to 2019.Target statistical power of 0.8 is indicated by the dotted red line.F I G U R E 4 Statistical power to detect declines in the abundance of sea pens.Simulated data were informed by 2015 camera surveys.Simulated scenarios included 40 photos per transect and included a different number transects.The dashed horizontal line indicates statistical target of 0.8 power.

F
I G U R E 5 Statistical power to detect significant change in abundance (CPUE) in core strata for Black Dogfish (top-left panels), Smooth Skate (top-right panel), Northern Wolffish (bottom-left panels), and species richness (bottomright panels) using different simulated trawling effort per year.Facet number 3-10 represents the number of years simulated in the time series.The dashed horizontal lines indicate a 80% target power.There were 1000 simulations for each scenario.

F
I G U R E 6 Commercial fishing effort across all fisheries and gear types, where vessel monitoring and logbook reporting are available, prior to marine protected area (MPA) designation (2005-2018 combined).Fishing effort (hrs/ km 2 ) is standardized into percentiles (0-100).The MPA boundary and zones within are indicated in black.No data are provided for St. Pierre-Miquelon waters (red polygon).Data used to create this figure is available at: https://open.canada.ca/data/en/dataset/273df20a-47ae-42c0-bc58-01e451d4897a.
Summary of RV survey collections in different regions during spring from 1996 to 2019, and seafloor transect data collected at the Laurentian Channel MPA in 2015 using the T A B L E 1 Model (GLMM) parameters of CPUE temporal trends of Black Dogfish, Smooth Skate, and Northern Wolffish caught in the LCMPA and 3Ps divisionand 3Pn division (1996-2013) using strata as a random effect.
T A B L E 2 The GLMM estimated regression parameters of CPUE comparison for Black Dogfish, Smooth Skate, and Northern Wolfish caught in the MPA and 3Ps from 2015 to 2019.
T A B L E 3