If you build it, they will come: Coastal amenities facilitate human engagement in marine protected areas

1. Calls for using marine protected areas (MPAs) to achieve goals for nature and people are increasing globally. While the conservation and fisheries impacts of MPAs have been comparatively well-studied, impacts on other dimensions of human use have received less attention. Understanding how humans engage with MPAs and identifying traits of MPAs that promote engagement is critical to designing MPA networks that achieve multiple goals effectively, equitably and with minimal environmental impact. 2. In this paper, we characterize human engagement in California's MPA network, the world's largest MPA network scientifically designed to function as a coherent network (124 MPAs spanning 16% of state waters and 1300 km of coastline) and identify traits associated with higher human engagement. We assemble and compare diverse indicators of human engagement that capture recreational, educational and scientific activities across California's MPAs. 3. We find that human engagement is correlated with nearby population density and that site “charisma” can expand human engagement beyond what would be predicted based on population density alone. Charismatic MPAs tend to be located near tourist destinations, have long sandy beaches and be adjacent to state parks and associated amenities. In contrast, underutilized MPAs were often more remote and lacked both sandy beaches and parking lot access.


| INTRODUC TI ON
Marine protected areas (MPAs)-places where human activity, especially extractive practices such as fishing, is prohibited or restricted-are a common ocean management tool used to achieve a mixture of conservation, fisheries and cultural objectives (Erskine et al., 2021;Grorud-Colvert et al., 2021;Marcos et al., 2021).By restricting extractive and destructive human activities, adequately designed, funded and regulated MPAs can increase the diversity and abundance of marine fish and invertebrates (Edgar et al., 2014;Gill et al., 2017;Goetze et al., 2021;Zupan et al., 2018) and the function and resilience of marine ecosystems (Cheng et al., 2019;Mellin et al., 2016).In the long term, and with concerted community participation and buy-in, well-designed MPAs can also yield fisheries benefits through increased productivity and spillover resulting from improved biomass and age structure of populations in the MPA (Di Lorenzo et al., 2020;Marshall et al., 2019).Furthermore, MPAs can facilitate and enhance other non-extractive human engagement in ocean ecosystems, such as cultural activities, recreation and tourism, education and outreach and scientific research (Angulo-Valdés & Hatcher, 2010;Ban et al., 2019;Erskine et al., 2021;Roncin et al., 2008).
While the ability and prerequisites for MPAs to achieve conservation and fisheries objectives have been comparatively wellstudied (e.g.Claudet et al., 2008;Edgar et al., 2014;Giakoumi et al., 2017;Goñi et al., 2010;Lester & Halpern, 2008;Wilson et al., 2020), the enabling conditions for achieving other human use objectives has received less attention (Ban et al., 2019;Erskine et al., 2021;Gerber et al., 2003;Naidoo et al., 2019;Turnbull et al., 2021).This is surprising given the frequency with which human engagement objectives-such as recreation, education and scientific research-are identified in international, national and regional MPA planning documents.For example, the Independent World Commission on the Oceans identifies the "provision of areas for scientific research, education and recreation" as a key benefit of MPAs (IWCO, 1998).Similarly, the U.S. Framework for the National System of Marine Protected Areas identifies the benefits of U.S.
Evaluating human engagement in MPAs is needed to track progress towards achieving these objectives and for identifying the design principles that determine human engagement in MPAs.Here, we use California's MPA network, the world's largest MPA network scientifically designed to function as a coherent network (Botsford et al., 2014), as a case study for identifying conditions that promote or limit human engagement in MPAs.
In 1999, the California state legislature passed the Marine Life Protection Act (MLPA), which directed the state to use the best available science to redesign and greatly expand its system of MPAs to function as a coherent network and to address six goals in service of conservation, fisheries and other cultural objectives (Gleason et al., 2013;Marine Life Protection Act, 1999).
In addition to goals to preserve biodiversity and ecosystem function and to sustain, conserve, protect and rebuild marine populations, including those of economic value, the MLPA also included a goal to "improve recreational, educational and study opportunities provided by marine ecosystems that are subject to minimal human disturbance and to manage these uses in a manner consistent with protecting biodiversity."From 2004 to 2012, a community-driven and science-guided design process led to a coordinated network of 124 MPAs, containing 16% of state waters, along California's 1300 km (840 mile) coastline.Following implementation, an extensive monitoring effort began to ensure that the network could undergo adaptive management (Botsford et al., 2014).While some monitoring programmes were developed around human engagement in MPAs (e.g. the MPA Watch citizen science programme; MPA Watch, 2022b), the majority 4. Synthesis and applications: These results suggest that achieving MPA goals associated with human engagement can be promoted by developing land-based amenities that increase access to coastal MPAs or by locating new MPAs near existing amenities during the design phase.Alternatively, human engagement can be limited by locating MPAs in areas far from population centres, coastal amenities or sandy beaches.Furthermore, managers may want to prioritize monitoring, enforcement, education and outreach programmes in MPAs with traits that predict high human engagement.Understanding the extent to which human engagement impacts the conservation performance of MPAs is a critical next step to designing MPAs that minimize tradeoffs among potentially competing objectives.

K E Y W O R D S
California, citizen science, community engagement, human dimensions, human use, marine protected areas, recreation, tourism of the monitoring effort was focused on the ecological goals of the MLPA and on elucidating ecological responses to MPA implementation.
Here, we characterize human engagement in California's MPA network and identify traits associated with high engagement.
We assemble and evaluate diverse indicators of engagement that capture a range of recreational, educational and scientific activities.We then relate levels of human engagement to population density, accessibility, amenities and other traits likely to influence engagement.This provides a rare quantification of the ways in which people engage with MPAs and the potential pathways for enhancing or limiting engagement based on management goals.These insights are helpful as California (Newson, 2020), the United States (Executive Order on Tackling the Climate Crisis at Home and Abroad, Executive Order 14008, 2021) and the world (CBD, 2021) aim to protect 30% of the ocean by 2030 (30 × 30) to meet an array of conservation, fisheries and other cultural objectives (Sullivan-Stack et al., 2022).

| Marine protected areas
California's coastal waters are protected by a mosaic of spatial management areas that vary in regulatory authority and protection status (Figure 1a; Table S1).State-managed areas include: ( 1 marine conservation areas (SMCAs), which restrict some types of fishing, except for within special no-take SMCAs, which prohibit all fishing; (3) state marine recreational managed areas (SMRMAs), which restrict fishing and allow hunting of waterfowl; (4) state marine parks (SMPs), which prohibit commercial fishing; and (5) special closures, which restrict activity around seabird colonies and marine mammal haulouts and are the only designation not defined as an MPA by the MLPA (Table S1).Federal marine reserves and conservation areas (FMRs and FMCAs respectively) extend certain SMRs and SMCAs around the Channel Islands into federal waters (Figure 1a).
We focus on the 124 MPAs that the MLPA identifies as being part of California's state-managed coastal MPA network (Figure 1a; Table S1).This excludes federally managed MPAs around the Channel Islands; SMRAs and SMPs in San Francisco Bay, which were established before the MLPA planning process and are not coastal; and special closures, which are not identified as MPAs by the MLPA.We refer to the resulting network of 49 SMRs, 60 SMCAs, 10 no-take SMCAs and 5 SMRMAs as California's state MPA network.While the Channel Islands MPAs were established before the MLPA planning process, they have been legally incorporated into the network.The four MLPA regions (South, Central, North Central and North Coasts; Figure 1) encompass a wide range of ecological dynamics, coastal features, oceanographic environments, cultures and economies.

| Surrounding human communities
We hypothesized that the number of people living near an MPA and the socioeconomic vulnerability of this population would contribute to engagement levels.In short, we expected that MPAs with larger and less vulnerable nearby human populations (i.e.populations with more disposable income and time for recreation) would experience greater human engagement.We characterized the human population living near MPAs using population demographics data from the 2010 U.S. Decennial Census (USCB, 2010a).The 2010 data is the most recent available data given extended delays in the release of the 2020 U.S. Census data (Schneider, 2023).We downloaded total population estimates by census block, the smallest geographic unit used in the census, using the tidycensus R package (Walker et al., 2022) and calculated the density of people living within each block.We rasterized (500 × 500 m resolution) these data and calculated the number of people living within a 50 km radius (~31 miles) of each MPA (Figure 1b).The number of people living within 50 km is generally correlated (r 2 > 0.8) with population densities using buffer distances ranging from 10 to 100 km (~6-60 miles) (Figure S1).
We estimated the social vulnerability of these populations using 12 indicators identified by Jepson and Colburn (2013) and collected by the U.S. Census American Community Survey (USCB, 2010b).
These indicators describe various metrics of poverty status, housing characteristics, labour force structure and population composition (Table S2; Figures S2-S4).We downloaded these indicators by census tract, the smallest geographic unit for which all of the indicators were available (one level larger than census block), also using the tidycensus R package (Walker et al., 2022).We combined these indicators into a single vulnerability index by averaging the z-scores of each indicator (i.e.indicators were centred on the statewide average and scaled to unit variance).Thus, a value of zero indicates average vulnerability across all of the various indicators, negative values indicate higher than average vulnerability and positive values indicate lower than average vulnerability.We rasterized the tract-level index to match the population raster and calculated the average vulnerability of the population within 50 km of each MPA as the populationweighted average of the social vulnerability index.

| Human engagement in protected areas
We developed indicators of human engagement in recreational, educational and scientific activities in California's state MPA network using a mixture of citizen science, naturalist and state agency datasets (Table S3).We focused on recreational, educational and scientific engagement given that they are specific objectives of the network (Marine Life Protection Act, 1999) and given the lack of data on other cultural, spiritual or emotional types of human engagement.
We used data from two citizen science programmes (MPA Watch and Reef Environmental Education Foundation) and two naturalist social networks (iNaturalist and eBird), which provide spatially referenced records of activities (e.g.surfing, swimming, boating, tidepooling, diving, etc.) or observations of wildlife submitted by individual users, as indicators of recreational and educational engagement in MPAs.
While popular social media platforms such as Instagram, Facebook, Flickr and Twitter may provide a better indicator of visitation rates than specialist platforms such as iNaturalist and eBird (Tenkanen et al., 2017), the volume of data generated by these platforms requires careful subsampling to be manageable (e.g. Hausmann et al., 2017).Although analysis of these social media indicators of engagement was outside the scope of this study, we encourage their use in future research.We used data from the California Department of Fish and Wildlife (CDFW) on the annual numbers of permits issued for scientific research in California's MPAs as an indicator of scientific engagement.Finally, we used CDFW data on regulatory citations as an indicator of regulatory compliance within the network.
We used MPA Watch survey data to measure consumptive and non-consumptive human activities in California's MPA network.
MPA Watch is a citizen science programme that trains volunteers to observe and collect data on human engagement in protected areas (MPA Watch, 2022b).Volunteers use a standardized survey protocol (MPA Watch, 2022a) to record consumptive (e.g.fishing) and non-consumptive (e.g.surfing, boating, tidepooling, running, etc.) activities occurring both on-and offshore of coastal sampling sites (Table S4).Consumptive activities are classified as either active (e.g.fishing line in water) or inactive (e.g.fishing pole on boat but not being used); we focus on active consumptive activities.
We caution that SMRMAs and some SMCAs allow some forms of harvest and that MPA Watch volunteers, while well trained, are not legal authorities on MPA boundaries and regulations.Thus, our ability to infer the legality of consumptive activities documented by MPA Watch volunteers is limited.MPA Watch has been in operation since 2011 and, as of writing, has conducted over 33,000 surveys in 49 MPAs (47 of which meet our inclusion criteria) and 60 control (non-MPA) locations (Figure S5).While some MPAs have been surveyed consistently since 2011, others did not receive consistent visits until 2015 or later (Figure S5A).
To allow comparison between sites with variable temporal coverage, we limited analysis to surveys that took place from 1 January 2015 to 31 December 2021.To eliminate spurious results from surveys that were conducted either early in the morning or late at night or were either shorter or longer than the official protocol (MPA Watch, 2022a), we also limited analysis to surveys that occurred between 6 AM and 8 PM and lasted between 10 and 60 min (Figure S5B,C).We quantified human engagement by MPA in terms of (1) the percent of surveys in which an activity was observed and (2) the median number of activities observed per hour for surveys in which activities were observed (zeros excluded because of high zero-inflation; Figures S6 and S7).
We used iNaturalist submission records to measure engagement in wildlife observation within and adjacent to MPAs.iNaturalist is a web-and app-based platform that allows observers to submit wildlife photos for identification by amateur and professional naturalists (iNaturalist, 2022).iNaturalist was launched in 2008 and as of writing, has more than 100 million observations, 2 million observers and 380,000 observed species globally.We used the rinat R package (Barve et al., 2021) to download all iNaturalist observations submitted by users in a bounding box spanning the California coastline from 1 January 2000 to 31 December 2021 (iNaturalist allows back submissions, hence the availability of pre-2008 observations).We defined MPA-associated observations as observations occurring within 100 m of an MPA and quantified human engagement from 2012 through 2021 by MPA in terms of the number of (1) unique observers (number of iNaturalist users who submitted wildlife observations) and ( 2) observations (number of entries submitted).More than 5800 observers have submitted >72,000 observations associated with 121 of California's state MPAs (Figures S8 and S9).
We used eBird submission records to measure engagement in birding within and adjacent to MPAs.eBird is a global programme that collates observations of birds submitted by birdwatchers (eBird, 2022).It was launched in 2002 by the Cornell University Lab of Ornithology and the National Audubon Society but allows back submissions from birding diaries.As a result, eBird contains observations dating back centuries in many locations.As of writing, the global eBird dataset includes over 69.7 million submissions from nearly 800,000 birders.We downloaded eBird observations from California and, as with the iNaturalist data, identified observations occurring within 100 metres of an MPA from 2012 through 2021.
We quantified human engagement by MPA in terms of the number of (1) unique observers and (2) observations.More than 19,000 birders have conducted >193,000 surveys and made >3.8 million submissions to eBird associated with 114 of California's state MPAs (Figures S10 and S11).
We used Reef Environmental Education Foundation (REEF) diver surveys as an indicator of engagement in diving and snorkeling in California's MPAs.REEF is an international marine conservation organization that trains volunteer SCUBA divers and snorkelers to collect and report information on marine fish and select invertebrate and algae species during recreational SCUBA dives and snorkels (REEF, 2022).The diver survey programme was launched in 1993 and, as of writing, has >250,000 surveys by 16,000 volunteers at 15,000 sites worldwide.We received records of >14,700 surveys conducted in California and identified 4085 surveys occurring within 41 of California's state MPAs from 2012 through 2021 (Figures S12   and S13).We quantified human engagement by MPA in terms of the (1) number of surveys conducted and (2) number of years in which a survey was conducted.
We used records of scientific permits issued by CDFW for research conducted within California's MPA network as an indicator of the contributions of MPAs to scientific knowledge.While permits are required for any extractive or manipulative research in California's coastal waters, purely observational research (i.e. research without capturing, handling, etc.) does not require permits; thus, the permit data may underestimate the amount of research occurring in the network.From 2012 to 2021, 5329 scientific permits were issued for research in all 124 of California's state MPAs (Figures S14 and   S15).We quantified human engagement by MPA in terms of the (1) number of permits issued and (2) number of years in which permits were issued.
We used records of citations issued by the CDFW Law Enforcement Division for regulatory violations occurring within California's MPA network as an indicator of compliance.From 2016 to 2021, 2812 citations were issued for violations occurring within 85 of California's state MPAs (Figures S16 and S17).We quantified non-compliance by MPA in terms of the (1) number of citations issued and (2) number of years in which citations were issued.We used generalized linear models assuming a Poisson distribution to evaluate the correlation between the total number of citations issued within an MPA and human population density, human engagement (defined using the iNaturalist observer data), and observations of active fishing (defined using the MPA Watch survey data).We caution that the lack of patrol effort information limits our ability to infer non-compliance rates (i.e.whether more citations corresponds to more effort or more illegal activity) and advise that, going forward, CDFW record information on effort (e.g.number of patrol hours) to improve ability to document patterns of non-compliance and target patrol strategies.
To compare human engagement across indicators (Figure 2), we selected key metrics for each indicator (Table S3) to display in an engagement scorecard (Figure 3).We centred each metric on its mean and scaled it to unit variance to facilitate comparisons across indicators.We also measured and compared the degree to which engagement is concentrated within specific MPAs, a metric of the selectivity of users, by developing the engagement accumulation curves shown in Figure 4. We developed these curves by first calculating the percent contribution of each MPA to network-wide engagement for each of the metrics selected for the scorecard.
We then plotted the accumulation of these contributions beginning with the MPA with the highest engagement and ending with the MPA with the lowest engagement.The steeper the resulting curve, the more network-wide engagement is dominated by a few MPAs.

| Drivers of human engagement
We hypothesized that human engagement in MPAs would be correlated with nearby population density (Cinner et al., 2018;Ravenstein, 1885) except for (1) "charismatic" MPAs that draw participation from afar and thus generate more engagement than would be predicted based on nearby population density and (2) "underutilized" MPAs that are difficult to access (e.g.located offshore, limited F I G U R E 2 Maps illustrating six indicators of human engagement in California's state marine protected area (MPA) network.Multiple metrics are used to measure engagement for each indicator; see Table S3 for definitions of these metrics.Across indicators, larger symbols and deeper colours indicate higher engagement.In (c-f), black x's mark MPAs without any reported engagement.Dark horizontal lines delineate the four Marine Life Protection Act regions.See Figure S17 for a map of the regulatory citations indicator.road access, etc.) and thus generate less engagement than would be predicted based on nearby population density.To distinguish charismatic and underutilized MPAs, we regressed human engagement (as measured by the number of iNaturalist observers) against population density and extracted the MPAs that fell above (charismatic) or below (underutilized) 75% of the fitted values (Figure 5).For this model, we used the number of iNaturalist observers as our measure of human engagement because it was the most spatially comprehensive indicator (i.e.describes engagement in the greatest number of MPAs) and it correlates with all of the indicators of non-extractive engagement (i.e. it is not correlated with citations or consumptive activities; Figure S18).
We used logistic regression to identify traits associated with charismatic and underutilized MPAs (Figure 6).We considered 13 traits describing a range of MPA design features (age, size, protection level), habitats (sandy beach, rocky intertidal, kelp, estuary), accessibility and amenities (distance to port, number of parks, parking lots, campgrounds and picnic areas within 1 km) and the social vulnerability index.See Table S5 for the source of each explanatory variable.We then used a series of logistic regressions to evaluate the association between engagement (charismatic vs. typical and underutilized vs. typical) and these traits.We defined the logistic target level for each model based on "typical" MPAs (response of 0) versus charismatic or underutilized (response of 1).Logistic models were constructed stepwise after a priori identifying relevant drivers of engagement.The best fitting models were selected using Akaike information criterion (Akaike, 1974) to identify the most parsimonious model of the relationship between engagement and the evaluated traits.

| Comparison to non-MPA areas
The methods described above were used to determine which MPAs within California's MPA network generate the most human engagement and to identify the factors that drive differences in the levels of engagement; however, they are unable to reveal whether MPAs generate more, less or equivalent human engagement as similar non-MPA areas.To understand the degree to which MPA designations impact human engagement in coastal areas, we compared engagement in MPA areas to similar counterfactual non-MPA areas.We identified similar counterfactual areas through statistical matching (Ferraro, 2009), which is increasingly used to elucidate the ecological impacts of MPAs (Ahmadia et al., 2015;Gill et al., 2017).In short, we rasterized California's state waters into 200 m raster cells and paired F I G U R E 4 Cumulative contributions of individual marine protected areas (MPAs) to network-wide engagement based on several indicators of human engagement.The diagonal dashed line indicates a theoretical accumulation curve in which individual protected areas contribute equally to engagement within the overall network.Curved lines above this reference line indicate accumulation curves in which some protected areas make larger contributions (higher performers) to network-wide engagement than others (lower performers); the steeper the curve, the more network-wide engagement is dominated by a few protected areas.The accumulation curve for population size (dotted black line) provides an additional frame of reference: if human engagement were proportional to population size, engagement would accumulate according to this curve.Thus, curves steeper than this line indicate that benefits are more concentrated than would be predicted by population density (i.e.engagement is more selective) whereas curves shallower than this line indicate a more even distribution of benefits than would be predicted by population density (i.e.engagement is less selective).The MPA Watch indicators are excluded because they are not available for all MPAs within the network.
each MPA cell with a non-MPA counterfactual cell with otherwise similar properties.We identified non-MPA counterfactual cells that were similar to their MPA reference cells in their depth (m), distance from shore (km), nearby population density, proximity to parks and proximity to public beaches.These matching variables were selected based on their association with engagement based on theory (Cinner et al., 2018;Ravenstein, 1885) and as revealed through the regression analysis (Figure 6).Ideally, we would also match based on pre-MPA visitation rates (Devillers et al., 2015), but the lack of sufficient pre-MPA visitation data (see limited pre-2007 data in Figure S8) precluded this gold standard.However, by controlling for these known and quantifiable drivers of MPA site selection and human engagement, we can isolate, to the greatest extent practicable, the impact of MPA designation on human engagement.We derived these values for both MPA and counterfactual cells using the sources listed in Table S6.We identified suitable counterfactuals through statistical matching using the MatchIt package (Ho et al., 2011), using one-toone Mahalanobis distance matching with replacement and propensity score callipers of 0.20 standard deviations (Ho et al., 2007).
After an appropriate counterfactual was identified for each MPA cell (Figure S19), we calculated the log-response ratio of the sum of activities within each MPA's cells and its paired counterfactuals cells for the three engagement indicators with activities reported inside and outside MPAs using GPS coordinates (i.e., the iNaturalist, eBird and REEF indicators).We tested whether the mean log-ratio of these sums differed from zero using t-tests (i.e.whether MPAs and non-MPAs generate different levels of human engagement).Logresponse ratios were calculated after adding 1 to the engagement values occurring in both the numerator and denominator to avoid non-finite ratio values.

F I G U R E 5
Correlation between human engagement in an marine protected areas (MPA) and the number of people living within 50 km of the area.Human engagement is measured as the number of iNaturalist observers submitting observations within 100 m of an MPA from 2012 through 2021.The grey line and 95% confidence interval illustrate a linear regression (r 2 = 0.14; p < 0.001) fit to all points.Blue points with residuals greater than 75% of the fitted values were classified as "charismatic" MPAs, whose engagement is higher than would be expected based on population density.Red points with residuals less than 75% of the fitted values were classified as "underutilized" MPAs, whose engagement is lower than would be expected based on population density.The charismatic and selected underutilized MPAs are labelled with their abbreviated names.The green line and 95% confidence interval illustrate a linear regression (r 2 = 0.62; p < 0.001) fit to the "typical" protected areas (green points), whose engagement is largely determined by population density.
All data analysis and visualization were done in R (R Core Team, 2021), and all data and code are available on GitHub here: https://github.com/NCEAS/ca-mpa

| Human engagement in protected areas
MPA Watch volunteers observed non-consumptive activities in the vast majority of surveys conducted coastwide and within all of the 47 surveyed MPAs (Figure 2a).MPA visitors were most commonly observed walking and recreating on the beach, often with their pets.
Offshore recreation included boating, surfing, bodyboarding and swimming.MPA visitors were also often observed viewing wildlife and exploring tidepools (Figure S6B,C).MPAs in the South Coast region were most popular, especially those near the metropolitan areas of San Diego and Los Angeles (Figure 2a).MPA Watch volunteers observed active consumptive activities (i.e.fishing and hand collection of organisms) in all but four of the 47 surveyed MPAs (Figure 2b) but at rates substantially lower than those observed for non-consumptive activities (Figure S7B,C).
Hook-and-line fishing was the most commonly observed consumptive activity and was observed in ~6% of surveys within SMCAs (MPAs in which certain types of fishing are often allowed).However, active hook-and-line fishing was also reported by volunteers in surveys of no-take SMCAs (~1.8% of surveys) and SMRs (~2% of surveys; Figure S7B).Hand collection of organisms, trap fishing and spear fishing were the next most frequently reported consumptive activities.Net fishing, dive fishing, commercial passenger fishing vessel (CPFV) fishing and kelp harvest were more rarely reported (Figure S7B,C).Observations of consumptive activities were more frequent in South Coast MPAs and within SMCAs, which allow some types of harvest.
The number of people submitting wildlife observations to iNaturalist from within California's MPA network increased through time (Figure S8B,C).The majority of observers submit observations from only one MPA per year, but some observers make submissions from up to 21 MPAs per year (Figure S8C).Observers are especially interested in plants (often land-based), shells (molluscs) and seabirds (Figure S8B).iNaturalist participation is especially high in the touristic Monterey Bay area and secondarily high in the densely populated F I G U R E 6 Marginal effects of significant predictors of "charismatic" (top row) and "underutilized" (bottom row) MPAs as identified through stepwise logistic regression.Marginal effects represent the predicted probability when varying the variable of interest while fixing the other variables at their means.Shading depicts 95% confidence intervals.See Table S5 for the list of predictors included in each model and Table S7 for the results of each model fit.

FREE et al.
San Diego, Los Angeles and San Francisco areas (Figure 2c).MPA engagement was less selective than predicted by human population density for this form of human engagement (Figure 4).On average, California's MPAs have not generated more iNaturalist engagement than counterfactual sites (p = 0.12), indicating that non-MPA areas with similar features generate just as much engagement as MPAs for this type of activity (Figure 7).The number of recreational divers and snorkelers contributing to the REEF citizen science survey programme from within California's MPA network increased from the programme's inception in 1994 to a peak in 2011, then decreased until a resurgence during the COVID-19 pandemic (2020-2021; Figure S13B,C).Participants visited a range of habitats and depths but generally favoured kelp forests and rocky reefs (Figure S13B,C).The majority of participation has come from MPAs with high profile dive sites including, in decreasing order of prevalence, Matlahuayl SMR, Edward F. Ricketts SMCA, Point Lobos SMR, Pacific Grove Marine Gardens SMCA and Carmel Bay SMCA (Figures 2e and 3).REEF divers have been more selective in their MPA visitation than any of the other evaluated user groups (Figure 4).California's MPAs have, on average, generated much more REEF survey engagement than counterfactual sites (p < 0.0001), indicating that MPA status attracts engagement for this type of activity (Figure 7).
The number of scientific permits issued for research within California's MPA network has been variable through time and decreased during the COVID-19 pandemic (2020-2021; Figure S15B).to the representation of the different MPA designations within the network (i.e.no bias towards no-take areas; Figure S15C).
The number of citations issued for regulatory violations was highest in MPAs in the South Coast region, especially in the MPAs around Catalina Island, a major tourist destination off the coast of Los Angeles (Figure S17A).In general, the number of citations is positively correlated with nearby human population size (p < 0.001; Figure S17B) and human engagement (p < 0.001; Figure S17C) in MPAs, where engagement is defined as the total number of people contributing iNaturalist observations from within an MPA from 2012 to 2021.Interestingly, the number of citations was negatively correlated with the observation of active consumptive activity by MPA Watch observers (Figure S17D), which could indicate that the active consumptive activity reported by MPA Watch observers is sanctioned or that active consumptive activity is more prominent in areas with less active enforcement.Citations were more highly concentrated in certain MPAs than would be predicted by human population density alone (Figure 4).

| Drivers of human engagement
Across all indicators, human engagement in MPAs was highest in the populous South Coast region and the touristic Monterey Bay area in the Central Coast region and lowest in the remote North Coast region (Figures 2 and 3).We found that human engagement in MPAs was correlated to nearby population density (r 2 = 0.14; p < 0.001) but that MPA traits can enhance or reduce engagement beyond what would be predicted based on population density alone (Figure 5).
Elevated engagement in 20 "charismatic" MPAs (MPAs whose engagement is greater than would be expected based on population density) was associated with older MPAs with long sandy beaches and many adjacent land-based parks (Figure 6; Table S7).Reduced engagement in 42 "underutilized" MPAs (MPAs whose engagement is lower than would be expected based on population density) was associated with remoteness (i.e.far from the nearest port), lack of sandy beaches and lack of parking lot access (Figure 6; Table S7).
Counter to our hypothesis, social vulnerability was not a significant driver of human engagement in MPAs (Table S7).

| DISCUSS ION
Understanding the ability and prerequisites for MPAs to achieve human use objectives is central to designing MPA networks that provide multiple benefits to people and nature.California's MPA network supports a diverse array of recreational, educational and scientific activities.MPAs are commonly used for recreational activities such as walking, playing or relaxing on the beach or boating, surfing, swimming or SCUBA diving in the ocean.Engagement in these activities makes important contributions to local economies (Pendleton & Kildow, 2006) and to cultural, emotional and physical health (Hipp & Ogunseitan, 2011;Jacobson, 2020).Wildlife viewing is also common within California's MPAs and provides a platform for education and research.Many visitors engage in MPAs through citizen science programmes that provide opportunities both to learn about the natural world and to contribute to meaningful scientific datasets (Freiwald et al., 2018;Rapacciuolo et al., 2021).Finally, scientific researchers have utilized the MPA network as a "largescale ecological experiment" (sensu Jensen et al., 2012) to derive globally-relevant insights into MPA performance, marine ecology and fisheries and conservation science (e.g.Starr et al., 2015;White et al., 2021;Ziegler et al., 2022).
However, not all MPAs generate equal levels of human engagement.In general, engagement is positively correlated with surrounding human population density: the more people living near an MPA, the more engagement an MPA generates.Charismatic MPAs, MPAs that receive more engagement than would be expected based on nearby population density, likely draw additional users because they have adjacent land-based attractions (i.e.parks) and associated amenities (e.g.parking lots, restrooms, campgrounds).These MPAs also have higher amounts of sandy beaches, which based on the MPA Watch surveys, tend to generate higher engagement than rocky beaches.Furthermore, many of the charismatic MPAs are located in areas spanning the Monterey Bay and Big Sur coastlines and the city of San Diego, which attract high numbers of tourists.These results are consistent with studies of land-based protected areas that find that visitation rates are driven primarily by the availability of amenities such as parking lots, walking paths and campgrounds and the accessibility of parks to human populations (see Heagney et al., 2018 and references within).Finally, engagement is moderated by the selectivity of different user groups.For example, whereas divers are highly selective in their choice of MPAs to visit, scientists have conducted research much more evenly across the statewide MPA network.Birders disproportionately visit estuarine MPAs, which tend to harbour large bird populations due to their high productivity (Paracuellos & Tellería, 2004).
It is also critical to understand patterns of unsanctioned use within California's MPA network.Overall, consumptive use was observed in a higher proportion of surveys conducted in MPAs that allow some types of harvest (i.e.SMCAs and SMRMAs) than in fully no-take MPAs that prohibit all fishing (i.e.SMRs and no-take SMCAs).
However, MPA Watch surveys, which we caution are conducted by citizen scientists and not by law enforcement officers, document fishing inside many of California's no-take MPAs.While observed much less frequently than non-consumptive activities, fishing was still reported in 10% of all MPA Watch surveys conducted in no-take MPAs.The vast majority of reported fishing in no-take areas was by recreational anglers using hook-and-line fishing gear.In most cases, we suspect this was due to a lack of education on the location of MPA boundaries by recreational anglers, as opposed to deliberate poaching activities.The rare observation of commercial fishing in MPAs suggests high compliance by the commercial fleet, which is highly informed about the location and regulations of MPAs.This is consistent with official summaries showing that, in 2011 (the most recent year with publicly available data), 271 citations were issued  to commercial fishers while 10,052 citations were issued to recreational fishers (~4 times larger than the number issued to recreational hunters) (CDFW, 2011).This suggests that outreach within the recreational fishing community could be especially effective at increasing compliance with MPA regulations.
Our findings have several key management implications.If promoting human engagement in MPAs is a management objective, our results suggest that MPA planners could improve access and promote engagement either by (1) locating new MPAs in areas with adjacent land-based parks and amenities or (2) investing in the development of new land-based parks and/or amenities adjacent to existing MPAs.Furthermore, aligning protections on land and sea could improve MPA performance by preventing pollution, sedimentation or eutrophication resulting from run-off from land-based activities (Cicin-Sain & Belfiore, 2005).Alternatively, if reducing human engagement is desired-for example, to enhance the protection of biodiversity or other ecosystem or cultural services sensitive to human visitation or to limit cumulative stressors to promote climate resilience-then planners could locate MPAs far from people or landbased parks and amenities (Campbell et al., 2020).Our results could also help guide decisions about where to invest in the monitoring, enforcement and outreach programmes required to ensure compliance (Murray & Hee, 2019).We found that the citation frequency for MPA rule violations increased with engagement and adjacent population size.These programmes may want to prioritize MPAs in areas of high population density and with adjacent land-based amenities and sandy beaches.However, remote MPAs can also be areas of elevated non-compliance due to lower levels of perceived risk of detection (Crawford et al., 2004;Rojo et al., 2019), and enforcement should not entirely abandon these areas.In addition to monitoring and enforcement, expanded education and outreach is needed to prevent non-compliance before it happens, especially among recreational anglers (Bergseth & Roscher, 2018).
Equitable human engagement in California's MPA network is also an important socioeconomic objective.Unfortunately, the indicators of engagement evaluated here do not include demographic information on the identity of human users, limiting our ability to evaluate the equity of engagement among different user groups.The collection of information in the identity of MPA users is thus a vital first step towards considering equity in future MPA planning and outreach.Knowledge of the representativeness of current users is necessary to design and implement programmes that promote access and engagement among underrepresented groups.This knowledge could be gained by interviewing MPA visitors in intercept surveys and assessing the composition of these users relative to that of surrounding communities (e.g.Scully-Engelmeyer et al., 2021).It could also be gained through focus groups with the various community organizations that engage with MPAs, such as fishing, diving and/ or birding clubs or direct interaction with communities (e.g.Diedrich et al., 2017).The equity of access and engagement should be considered at the outset of any additional MPA planning, including the identification of methods for tracking and benchmarking progress towards these objectives.As California prepares to expand its MPA network to meet 30 × 30 goals, it will be important to build on the successes and lessons of the original participatory planning process (Gleason et al., 2013) to further enhance the ability for ocean users, especially indigenous people, to ensure that their values are reflected in the objectives, regulations and design of the expanded network (Barclay et al., 2017;Voyer et al., 2015;Voyer & Gladstone, 2018).
MPAs with low human engagement can still provide valuable contributions to the human engagement, conservation and fisheries goals of the MPA network.While total engagement at some MPAs is low, these MPAs could be more important to small but underserved human populations in the neighbouring area.This is a key benefit of the MLPA's spacing requirements, which mandated that California's MPAs be placed within 50-100 km of each other (Saarman & Carr, 2013).This spacing ensures that coastal populations have relatively similar access to MPAs along the entire California coast.Thus, while MPAs in low population areas have lower engagement, the people living in these areas have opportunities for access similar to people living in higher population areas.Furthermore, MPAs also aim to achieve conservation and fisheries benefits and MPAs with low human engagement can be critical contributors to these goals.This is especially true given that human engagement with MPAs has the potential to negatively impact ecosystem function and MPA performance (Milazzo et al., 2002).Limiting human engagement can also reduce the cumulative impacts of multiple stressors on MPAs, including climate change, eutrophication and pollution (Mach et al., 2017).MPAs with low human engagement are thus key in the design of effective MPA networks, as they can buffer or offset the impacts of human activities in MPAs with greater engagement and limit cumulative impacts in a multi-stressor environment.A network of MPAs, like that in California, provides the opportunity to design individual MPAs that meet differing criteria and perspectives regarding human-nature relationships (Pereira et al., 2020) while contributing to overall network performance across a range of axes.
The methodological framework developed here presents a useful starting point for assessing human engagement in any MPA network.To start, the iNaturalist and eBird citizen science programmes already have wide global coverage and REEF has high participation in many regions.Other social media platforms, such as Instagram, Twitter and Flickr, may also be used to assess how, when and where people engage in MPAs (Retka et al., 2019;Tenkanen et al., 2017).
However, these indicators do not capture all types of human engagement or all of the information needed to understand the ecological impacts of human engagement or the equity of engagement among different human populations.Notably, our indicators do not capture information on: (1) user demographics, which are key for understanding equality in access (Nicholls & Shafer, 2001); (2) activities that have negative ecological impacts, such as anchoring (Creed & Amado Filho, 1999); or (3) money spent on licences, entry fees, food, gas and lodging, among other expenses associated with human engagement in MPAs, which are helpful in quantifying the broader impact of MPAs to local economies (Sala et al., 2013).Furthermore, the types of engagement evaluated here, especially engagement in science and tourism, likely undercount underserved and disadvantaged

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information can be found online in the Supporting Information section at the end of this article.Table S1.California marine protected area (MPA) designations.
Table S2.Social vulnerability indicators and metrics used to calculate the social vulnerability index.
Table S3.Indicators of human engagement evaluated in this paper.Table S4.Human use activities recorded by MPA Watch volunteers.
Table S5.Sources of explanatory variables included in logistic regressions evaluating traits associated with charismatic and underutilized MPAs.
Table S6.Matching variables used in the design of counterfactual areas and their sources.
) state marine reserves (SMRs), which prohibit all fishing; (2) state F I G U R E 1 Maps illustrating (a) California's marine protected area (MPA) network and (b) nearby human population density.In (a), greens indicate state MPAs established by the Marine Life Protection Act (MLPA), oranges indicate state MPA designations excluded from the analysis and purples indicate federal MPAs excluded from the analysis.See Section 2.1 and Table S1 for the definition of each MPA designation.Point size indicates MPA area (km 2 ).Dark horizontal lines delineate the four primary MLPA regions (labelled with month of implementation).MPAs in the San Francisco Bay region were established before 2007 and were not part of the MLPA planning effort.MPAs in the Northern Channel Islands were also established before the MLPA (2003 and 2007 in state and federal waters respectively) but have been officially incorporated into the network.The thin grey line indicates state waters (3 nautical miles offshore).In (b), point size indicates the number of people living within 50 km of each MPA.Colours indicate population density by census block in the 2010 U.S. Census.A few key coastal cities are labelled for reference.

FREE
et al.

F
I G U R E 3 A synthesis of human engagement indicators within California's state marine protected areas (MPAs).MPAs are sorted by population density within 50 km (first column of each plot) within each region.Engagement indicators are centred on the average of each indicator and scaled to unit variance to ease comparison across indicators; thus, colour indicates the number of standard deviations (SDs) from the mean where blue shades indicate MPAs with above average engagement and red shades indicate MPAs with below average engagement.Grey indicates MPAs without data and x's indicate MPAs with true zeros.MPAs with greater ("charismatic") and less ("underutilized") engagement than expected based on surrounding population density are marked in the population size column.See TableS3for definitions and metrics of the displayed indicators.25758314, 2023, 5, Downloaded from https://besjournals.onlinelibrary.wiley.com/doi/10.1002/pan3.10524 by University Of California, Santa Barbara, Wiley Online Library on [04/10/2023].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License | 1599 People and Nature FREE et al.
Birders have been visiting California's MPAs since before they were designated as protected areas (FigureS11B,C).The participation of birders in the eBird citizen science programme increased linearly from the 1960-2005 and exponentially since 2005 (FigureS11B).Participation has been greatest, in terms of number of birders submitting eBird observations, at popular birding hotspots such as Bolsa Chica Basin SMCA, Elkhorn Slough SMR, Matlahuayl SMR, Morro Bay SMRMA and Point Reyes SMR (Figures2d and 3).MPAs within estuaries-including Bolsa Chica Basin, Elkhorn Slough and Morro Bay-generate a disproportionate amount of eBird activity: despite representing only 2% of California's state MPA network by area (17% by count), around 40% of recent annual visits to the network logged by eBirders have been within estuarine MPAs (FigureS11C).Despite the tendency for eBirders to visit estuarine MPAs, the selectivity of birders was generally proportional to that predicted by population density (Figure4), suggesting that estuarine MPAs are located in areas with high population density.On average, California's MPAs have generated slightly more eBird engagement than counterfactual sites (p = 0.02), indicating that MPA status attracts engagement for this type of activity (Figure7).
The distribution of scientific research throughout the MPA network has been more even than any other type of human engagement (Figure 4).In general, fewer permits have been issued for research in the North and North Central Coast regions, and more permits have been issued for research in the Central (especially Monterey Bay) and South (especially Los Angeles and San Diego) Coast regions (Figures 2f and 3), where academic institutions and marine science non-profits are more highly concentrated.Scientific research in MPAs of different designations has generally occurred in proportion F I G U R E 7 The level of human engagement in marine protected areas (MPAs) compared to non-MPA counterfactuals for indicators with the required data.Log-response ratios were calculated after adding 1 to the engagement values occurring in both the numerator and denominator to avoid non-finite values.Log-response ratios greater than zero indicate MPAs where the MPA designation is associated with higher engagement relative to the counterfactual whereas ratios less than zero indicate MPAs where the MPA designation is associated with lower engagement relative to the counterfactual.Asterisks indicate indicators whose mean response ratio is significantly different from zero (*p < 0.05, **p < 0.01).p-values are shown parenthetically.In the boxplots, the solid line indicates the median, the box indicates the interquartile range (IQR; 25th to 75th percentiles), the whiskers indicate 1.5 times the IQR and the points beyond the whiskers indicate outliers.Points represent log-response ratios for each MPA and counterfactual pair.

FREE
et al.

Figure S1 .
Figure S1.The correlation between population density calculated using the selected 50 km buffer and population densities calculated using alternative buffer distances.

Figure S2 .
Figure S2.Maps of the social vulnerability indicator used to calculate the social vulnerability index by California US Census tract.

Figure S3 .
Figure S3.Distribution of the social vulnerability indicator values by California US Census tract used to calculate the social vulnerability index.

Figure S4 .
Figure S4.Social vulnerability index by US Census tract (polygons on land) and average social vulnerability index within 50 km of each MPA (points at sea).

Figure S5 .
Figure S5.The (A) coverage of usable MPA Watch surveys over time by marine protected area (MPA).A usable survey is a survey in which the duration was accurately recorded (i.e.end time occurs after start time).Note log-scale for fill color.San Francisco Bay MPAs are plotted in the North Central Coast region for simplicity.Only surveys occurring between 1 January 2015 and 31 December 2022 were considered in the analysis.We also excluded (B) surveys shorter than 10 min or longer than 60 min and (C) surveys ending before 7 AM or starting after 7 PM.

Figure S6 .
Figure S6.Non-consumptive activities in California's state marine protected areas (MPAs) based on surveys conducted by MPA Watch.

Figure S7 .
Figure S7.Active consumptive activities in California's state marine protected areas (MPAs) based on surveys conducted by MPA Watch.

Figure S8 .
Figure S8.Coverage of iNaturalist observation data over time by marine protected area (MPA).Note log-scale for fill color.MPAs are listed in order of overall sample size within each region.

Figure S9 .
Figure S9.Human interest in wildlife within California's state marine protected areas (MPAs) based on usage of the iNaturalist web-and app-based application.

Figure S10 .
Figure S10.Coverage of eBird observation data over time by marine protected area (MPA).Note log-scale for fill color.MPAs are listed in order of overall sample size within each region.

Figure S11 .
Figure S11.Human engagement in birding within California's state marine protected areas (MPAs) based on submissions to the eBird citizen science program.

Figure S12 .
Figure S12.Coverage of REEF survey data over time by marine protected area (MPA).

Figure S13 .
Figure S13.Engagement of recreational divers and snorkelers in the REEF citizen science survey program within California's state marine protected areas (MPAs).

Figure S14 .
Figure S14.Number of scientific permits issued annually from 2012 to 2021 by marine protected area (MPA).MPAs are listed in order of overall sample size within each region.

Figure S15 .
Figure S15.Number of scientific permits issued for research within California's state marine protected areas (MPAs) from 2012 through 2021.

Figure S16 .
Figure S16.Number of citations issued by CDFW Law Enforcement for regulatory violations occurring within California's MPAs from 2016 to 2021.MPAs are listed in order of overall sample size within each region.

Figure S17 .
Figure S17.Number of citations issued by CDFW Law Enforcement for regulatory violations occurring within California's state marine protected areas (MPAs) from 2016 through 2021.

Figure S18 .
Figure S18.Correlation between human engagement indicators.The lower section shows pairwise comparisons of engagement indicators.

Figure S19 .
Figure S19.The balance of matching variables (A) pre-and (B) postmatching and the (C) correlation between the values of MPA and matched non-MPA raster cells.In (C), the black line is the one-to-one line.