Changes in ocean health in British Columbia from 2001 to 2016

Effective management of marine systems requires quantitative tools that can assess the state of the marine social-ecological system and are responsive to management actions and pressures. We applied the Ocean Health Index (OHI) framework to retrospectively assess ocean health in British Columbia annually from 2001 to 2016 for eight goals that represent the values of British Columbia’s coastal communities. We found overall ocean health improved over the study period, from 75 (out of 100) in 2001 to 83 in 2016, with scores for inhabited regions ranging from 68 (North Coast, 2002) to 87 (West Vancouver Island, 2011). Highest-scoring goals were Tourism & Recreation (average 94 over the period) and Habitat Services (100); lowest-scoring goals were Sense of Place (61) and Food Provision (64). Significant increases in scores over the time period occurred for Food Provision (+1.7 per year), Sense of Place (+1.4 per year), and Coastal Livelihoods (+0.6 per year), while Habitat Services (-0.01 per year) and Biodiversity (-0.09 per year) showed modest but statistically significant declines. From the results of our time-series analysis, we used the OHI framework to evaluate impacts of a range of management actions. Despite challenges in data availability, we found evidence for the ability of management to reduce pressures on several goals, suggesting the potential of OHI as a tool for assessing the effectiveness of marine resource management to improve ocean health. Our OHI assessment provides an important comprehensive evaluation of ocean health in British Columbia, and our open and transparent process highlights opportunities for improving accessibility of social and ecological data to inform future assessment and management of ocean health.


Supporting Methods: Goal models and data Habitat Services
The Habitat Services score is the average of Coastal Protection (CPP) and Carbon Storage (CSS) subgoals: The CPP and CSS subgoals are described below.

Coastal Protection
To determine coastal protection within a region, we sum the protective potential of each spatialized unit of coastal habitat based on the protective value of that habitat type and the exposure of the habitat site, and compare this total protective potential to that of an historic baseline.
Habitat extent for coastal forests !" and salt marsh !" are based on 30 m land use rasters [1], clipped to forest and marsh habitat within 1 km of the shoreline and no more than 5 m elevation. Coastal exposure for a given cell !"## is based on exposure classes from the British Columbia Marine Conservation Analysis project (BCMCA) [2]; raw values from 1 ("highly protected") to 6 ("highly exposed"), are rescaled from 0 to 1. Protective capacity weights for coastal forest !" and salt marsh !" are calculated as (1 -Vulnerability) / 4, i.e. rescaled 0 to 1.
Coastal forest and salt marsh exposure are exposure-weighted area of each habitat, based on 30 m land use rasters and exposure class of each cell: Gapfilling: Since land use rasters were available only for 1990, 2000, and 2010, values for intervening years are based on linear interpolation, e.g. !""# = 0.6 !""" + 0.4 !"#" . For values after 2010, the 2010 value is carried forward.

Carbon Storage
Carbon storage potential is scored based on the current extent of all carbon sequestering habitats, weighted by the amount of carbon effectively sequestered in a unit of each habitat. Scores for this goal compare carbon storage potential to an historic baseline.
As in the Coastal Protection subgoal, habitat extent for coastal forests !" and salt marsh !" are based on 30 m land use rasters [1]. For salt marsh, all wetland cells within 1 km of the shoreline are included. For coastal forests, we included all forest cells found within sub-watersheds incident with the coastline, and did not consider elevation.
Carbon sequestration potential is based on carbon burial rates for each habitat, measured in gC m -2 yr -1 [4].
• Salt marsh !" : 218 +/-24 gC m -2 yr -1 (mean +/-SE) • Coastal boreal forests !" : 4.6 +/-2.1 gC m -2 yr -1 Coastal forest and salt marsh area are based on number of cells for that habitat in 30 m land use rasters within the appropriate buffer zone: Gapfilling: Since land use rasters were available only for 1990, 2000, and 2010, values for other years were gap filled in the same manner as for the Coastal Protection subgoal.

Food Provision
The Food Provision goal is calculated as the sum of wild-capture fisheries, aquaculture, and wildcapture salmon subgoals divided by the number of non-NA subgoals available for that region and year.
The wild-capture fisheries, aquaculture, and wild-capture salmon subgoals are described below.

Wild-Capture Fisheries
Wild-capture fisheries are scored as a catch-weighted average of the health and management status of all stocks assessed against an MSY reference point (i.e. those with reported catch per DFO [6] and either / !"# or / !"# or both in the RAM database [35], hereafter "assessed," vs. "unassessed" stocks lacking an MSY reference point) within a region, modified by a penalty to account for unassessed stocks targeted within the region. Spatially explicit landings information for 20 species, representing 47 different stocks (S9 Table), for the years 2007 to 2015 were provided by DFO [6]. Scores for assessed fishery stocks are based on both the total biomass of the stock relative to biomass at maximum sustainable yield (MSY), !"# , and the fishing mortality relative to that at MSY, !"# , as reported by the RAM Legacy database [5]. These / !"# and / !"# values are rescaled from 0 to 1 based on a goal of maximizing sustainable yield, resulting in ′ ∈ [0,1] and ′ ∈ [0,1] (S5 Fig.). Unassessed stocks are given a score of half the catch-weighted average score of the region's assessed stocks. The overall score for a given region and year is the catch-weighted mean of all assessed and unassessed stocks within a region for that year.
Note / !"# data were unavailable for some assessed stocks, in which case the stock score was based on the ′ term.
See S9 Table for a list of all OHIBC stocks including assessment status.
Rescaling / Rescaled fishing mortality ′ for each stock is calculated based on / !"# , smoothed using a rolling four-year mean. A DFO harvest control rule indicates no targeted catch for / !"# below a critical threshold of 0.4, increasing to / !"# = 1 for / !"# ≥ 0.8. Our calculation allows a buffer around this to account for uncertainty in setting annual management targets, and incorporates an overfishing penalty ( ′ = 0 for / !"# ≥ 2) as well as an underfishing penalty to account for lost opportunity for additional sustainable catch.
When / !"# ≥ 0.8 (healthy stock): When / !"# < 0.8 (overexploited stock):  Gapfilling: Due to high variance of annual catch estimates used to determine weighting of stock status scores within each region, estimates per stock and region were gapfilled by carrying back the mean of the first three available years, and carrying forward the mean of the last three available years. Gaps in stock assessment values were simply carried forward from the last observation.

Aquaculture
The Aquaculture (AQC) model compares the aquaculture harvest within a region to its total harvest potential , for both finfish ( ) and bivalve ( ) aquaculture (weighted by harvest of each aquaculture type). !"#,!" Using aquaculture growth potential index data from Gentry et al. [9], we determined a reference harvest potential for finfish and bivalves for each region, in tonnes/km 2 . Designated aquaculture tenures [10] outline areas approved for aquaculture production, which we take to be a proxy for management targets. Multiplying the harvest potential by the area of designated aquaculture tenures for finfish and bivalves, we estimate the total sustainable harvest potential, in tonnes, for each region.
For both finfish and shellfish, a score of 100 reflects a harvest equal to the lower bound on potential calculated from (mean -1 sd) of the growth potential index: While harvest far above the estimated production potential may indicate unsustainable practices, particularly for high stocking densities of finfish, we did not apply an overproduction penalty due to the uncertainty inherent in production potential estimates and site-specific production methods.
Gapfilling: Because the time series of available data is short and shows high variance, we do not gapfill these layers, as assumptions are not likely to be valid. As such, this goal is scored only for 2011-2015.

Salmon
The Salmon sub-goal of Food Provision compares annual catch for = 13 indicator fisheries (S10 Table) to catch target !"# for that year for that fishery, scoring 100 when the catch is between 60% and 100% of the catch target (the standard deviation of / ! across all stocks and years is 0.4, so 60% of catch target allows a 1 standard deviation buffer), dropping to a score of 25 as catch falls from 60% of target to zero, and dropping to 0 when the catch exceeds twice the target. Score is calculated separately for each unit and then all scores for all indicator stocks are averaged; a single score is applied equally across all OHIBC regions.

First Nations Resource Access Opportunities
The First Nations Resource Access Opportunities (AO) goal examines access to four marine resources of broad FSC importance to First Nations communities across the British Columbia coast: wild-capture salmon, shellfish beds, herring spawn-on-kelp, and access to commercial fisheries.
The goal score for a given region and year is determined by an average of all available component scores for that region and year.

Commercial fisheries access
As a proxy for fisheries access, we compare the proportion of commercial fishing licenses allotted specifically for "aboriginal" license types or holders within each region by DFO [15] to the proportion of the region's population living in First Nations communities (based upon 2016 population of census subdistricts ( ) identified as First Nations communities (i.e., = ), [17]). A score of 100 in this component indicates the proportion of FN-allocated licenses meets or exceeds the proportion of FN population, or 15%, whichever is greater.
Fishing licenses allow access to specific Pacific Fisheries Management Areas (PFMAs). For each region and year, we count the number of licenses (First Nations, and all) that allow access to each OHIBC region.

Shellfish harvest
To determine access to safe shellfish harvests, we determined the number of contamination-related shellfish closure days in each fishery management subarea for each year (data were available for 2009 to 2015) [14], and calculated an area-weighted mean number of closure-free days throughout each region. A score of 100 in this component indicates no closures due to contamination in the region.

Herring spawn abundance
Herring spawn index data [16] estimate the mean density of herring spawn available in each herring section. We aggregate these to OHIBC region, applying a rolling three-year mean to smooth typical interannual variations to calculate herring spawn abundance for each region for the years 1940-2016. A score of 100 in this component indicates a smoothed herring spawn index value that meets or exceeds the reference value !"# as the mean value within each region across a 20-year reference period from 1940-1960. The reference period was selected to estimate historic abundance prior to a crash in herring stocks in the 1960s. Gapfilling: none.

Wild caught salmon
Wild salmon escapements near a defined escapement targets ensure access to healthy salmon stocks in the future. We calculate the ratio of annual escapement against escapement targets for twelve indicator salmon stocks across four species (chum, coho, pink, and sockeye) [11-13] (See Table S10 for a list of stocks included in this component). For each stock, score increases linearly from 0 when that stock's escapement to target ratio is at or below 0.4 (approximately one standard deviation below target) to 1 when the ratio is at or above 1.0. All stock scores are averaged into a single salmon score that is applied equally for all regions.

First Nations Livelihoods
The First Nations Livelihoods model is based on job and wage data for coastal First Nation communities. Employment data by industry within British Columbia was not sufficiently detailed to identify jobs and wages for marine-dependent sectors. Instead we use a population-weighted average of employment rates ( = 1 -unemployment rate ) [17] and inflation-adjusted median wage [17] within the First Nation-specific census subdistricts ∈ {FN} [17] that fall within OHIBC inland boundaries.
Because no objectively defined reference point for employment rate was available, we calculate a reference point as a relative value on a moving baseline: the value in the current year relative to the mean value in a moving 5-year reference period, starting 5 years prior to the current year. To enable comparison between First Nations and non-First Nations employment rate, we use the higher of the two rolling means as the reference point for both. This reflects an implicit goal of maintaining coastal livelihoods and economies on short time scales, allowing for decadal or generational shifts in what people want and expect for coastal livelihoods and economy.
We defined the wage reference point as the highest observed inflation-adjusted wage across all OHIBC regions and all years for both First Nation and non-First Nation communities.
!"# = max( !" , !"!!!" ) across all regions and years Employment and wage information are reported at the census subdistrict level, and a populationweighted mean value for each region is calculated based upon census subdistricts identified as First Nations.

Non-First Nations Livelihoods
The non-First Nations Livelihoods model is identical to the First Nations Livelihoods model, except that employment and wage data are based on non-First Nation communities, identified as non-First Nation-specific census subdistricts ∈ {non-FN} [17] that fall within OHIBC inland boundaries.
and !"# are identical to those used in the LVF subgoal.
Employment and wage information are reported at the census subdistrict level, and a populationweighted mean value for each region is calculated based upon census subdistricts not identified as First Nations.

Tourism and Recreation
For this goal, we use number of visitors to coastal parks [18] and visitor centers [19] within a defined coastal region as a measure of tourist activity. Our area of "coastal interest" is defined by a buffer extending 15 km inland from the coastline. Visitors in a given year are compared to a moving reference point of mean visitation over the prior five year period. Park visits and visitor center visits are scored separately then averaged. Regions with no park or visitor center data were given NA scores.
Gapfilling: Due to the "no net loss" reference point, we chose to not backfill values prior to 2007. Therefore this goal was not scored prior to 2007.

Sense of Place
Sense of Place score is the average of Lasting Special Places and Iconic Species subgoals.
The Iconic Species and Lasting Special Places subgoals are described below.

Iconic Species
The Iconic Species model measures the percentage of iconic species in each extinction risk category. Species list is based upon input from Karin Bodtker and Andrew Day of the Vancouver Aquarium/CORI project. A full list of included iconic species can be found in S11 Table. Threat weights were assigned based on the COSEWIC province-level threat status where available [23] (covering most species on the list) and IUCN threat status [20] (for species without an available COSEWIC assessment).
In addition to including only a subset of species in the Species goal, the Iconic Species subgoal is not based on area-weighted average of species within a region, only on whether a species is present within a region. A score of 100 indicates all iconic species are at "Least Concern" status.
where !"",!" score corresponds to IUCN extinction risk categories: "Least Concern" = 1.0, "Near Threatened" = 0.8, "Vulnerable" = 0.6, "Endangered" = 0.4, "Critically Endangered" = 0.2. S11 Gapfilling: Conservation status for each species is based on a last observation carried forward, i.e., the current status is based on the most recent prior assessment. For years prior to the first assessment, the first assessment status is carried backward.

Lasting Special Places
Lasting Special Places measures the percentage of protected coastal marine and coastline area in each region, against a 30% reference target [44]. We include protected areas within coastal waters (MPAs, within 3nmi of shore) for marine special places assuming that sense of place is limited to areas readily accessible to or visible from the shoreline. For land-based protected areas (PAs), we include coastal sub-watersheds as they are intrinsically connected to the marine system. To determine protection status using a variety of sources, including the World Database of Protected Areas [24], British Columbia parks and protected areas, and tribal parks. While MaPP Special Management Zones communicate areas of deep historical, traditional, and cultural importance, they are not yet formally protected, so these regions were excluded from the analysis.

Biodiversity
Biodiversity status averages the condition of species (Species subgoal) and biodiversity-supporting biogenic habitats (Habitats subgoal).
The Species and Habitats subgoals are described below.

Species
The Species model measures the average threat status, defined by COSEWIC province-level threat assessments [23] where available and IUCN Red List threat assessments [20] elsewhere, of all species found in each region, weighted by each species' area of distribution within the region. See S12 Table for a count of species by taxa included in this assessment. Note that the species included in the Iconic Species subgoal are also represented here, making up approximately 9% of the species included in this subgoal.
The upper reference point for the Species sub-goal is to have all species at a risk status of Least Concern. As in OHI global assessments, we scale the lower end of the goal to be 0 when 75% of species are extinct, a level comparable to the five documented mass extinctions that would constitute a catastrophic loss of biodiversity.
Gapfilling: Conservation status for each species is based on a last observation carried forward, i.e., the current status is based on the most recent prior assessment, rather than a linear interpolation. For years prior to the first assessment, the first assessment status is carried backward. S12

Habitats
Habitats score is the mean condition of each biodiversity-supporting habitat for each region and year. Habitats included in the assessment are soft-bottom benthic habitats [2], salt marsh [1], and ecologically/biologically significant areas (EBSAs). EBSAs are determined by the DFO as areas with oceanographic, physical, or ecological conditions with special significance [27]; for our purposes we include only EBSAs related to biodiversity-supporting structure, including sponge reefs, deep water corals, hydrothermal vents, and seamounts.
The Habitats sub-goal assess the health condition of each habitat present in a region, based upon external trawling pressures (for soft-bottom habitat and EBSAs) and coverage area relative to historical baseline (for saltmarsh habitats).
Subtidal soft bottom habitat health is the inverse of average trawl effort across the region (i.e. hours of trawl per km 2 ) on soft-bottom habitat areas [2], relative to a reference point of the maximum trawl effort ! observed in any 4 km x 4 km cell for any year in the dataset [6].
EBSA health is the inverse of the average trawl presence in a given year (i.e. trawled area relative to total EBSA area) on EBSA-associated areas [27]. Note this is not effort-based as for soft-bottom habitats, since these slow-growing structures are far slower to recover than soft bottom sediment. !"#,!" Saltmarsh condition is calculated as the extent of a region's saltmarshes within 1 km of the shoreline, as noted by a 30 m resolution land use raster [1]. The reference point is the saltmarsh extent according to the 1990 land use raster.
Gapfilling: EBSA and soft bottom habitat pressures values prior to 2005 were gapfilled using next observation carried forward, while 2016 values were carried forward from 2015. Saltmarsh condition was gapfilled in the same manner as described for Coastal Protection subgoal.

Clean Waters
The Clean Waters goal score is calculated as the geometric mean of its four components: eutrophication (nutrients), chemicals, pathogens and marine debris. Each component layer estimates the pressure !"#$ due to that component on the system, so each component score is calculated The chemical, nutrient, pathogen, and marine debris components are described below.

Chemical pollution
Chemical pollution was measured as the average of land-based organic and inorganic pollution from agricultural pesticide use and runoff from impervious surfaces, respectively, and ocean-based pollution from commercial shipping and ports [45]. Organics are based on rasters of modeled plumes at 934 m resolution and are available for 2002-2013 [28]; these rasters are masked to the OHIBC study region, log transformed (log( + 1)), and rescaled from 0 to 1 where 1 indicates the 99.99%ile of the log-transformed values. Inorganics and ocean-based pollution are similar, though the layers are for a single year based on Halpern et al. [45]. These layers are already log-transformed to a global reference point; here they are masked to the OHIBC study region and rescaled where 1 indicates the highest observed value in the OHIBC study region. The chemical pressure score for each region is the mean chemical pressure score of all cells within the region.

Nutrient pollution
Modeled land-based nitrogen input for 2002-2013 [28] was used as a proxy for nutrient input. As for organic chemical pollution, it was masked to the OHIBC study region, transformed as log( + 1), and rescaled 0 to 1 based on the 99.99%ile of values.

Pathogens
Due to a lack of information on direct measurements of human pathogens in coastal waters, we used a proxy measure for pathogens: the population density of inland regions with unimproved wastewater treatment (i.e. population density on septic, storage-and-haulage, or no treatment) relative to the highest population density of any OHIBC region.
At risk densities were based on Municipal Water Use Report surveys from 2004, 2006, and 2009 [30][31][32]. These reports estimate the percent of population served by sewers, private septic systems, and sewage hauling, based on municipality size. We defined "at risk" as population not on sewer systems. For First Nations communities, we relied upon the National Assessment of First Nations Water and Wastewater Systems [29]. A digitized map of inspected wastewater systems classified as high, medium, and low risk was used to estimate the average risk for First Nations communities (as determined by census subdivision) within each OHIBC region.

Marine debris
The status of marine debris was estimated using modeled mass density of marine plastics (in kg/km 2 ) from Van Sebille et al. [33] on a 1° global grid. We interpolated using a thin-plate spline method to extend this grid into the Strait of Georgia and coastal fjords, then reprojected to BC Albers projection at 1000 m resolution, masked to the OHIBC region of study. The data were then rescaled from 0 to 1 based on the highest value found within the OHIBC study region. There is no time series for this layer.
Gapfilling: As there is no time series for this component, all years were scored the same.

Supporting Methods: Trend
Trend represents the proportional change in status over a recent past period, and is used to infer likely changes in status in the near future. For most goal models (except SPP and ICO, noted below), trend is calculated as the slope estimate of a linear regression of status for the prior five-year period, divided by the status in the earliest year of the five-year period; this result is multiplied by five to indicate the likely change in status over the next five years.
In general, trend is constrained to a range of +1 to -1. If a goal status reaches 100, trend is limited to non-positive values; similarly, if a goal status reaches zero, trend is limited to non-negative values.
For the Species (SPP) and Iconic Species (ICO) subgoals, we converted IUCN species-specific trend information (e.g., "increasing", "decreasing", "stable") to numeric values, based on a regression of species status (only for species whose status has been assessed multiple times) against these categories. For ICO, trend is the average species-specific trend of all species found within a region; for SPP, trend is the area-weighted average of these species specific trends.

Supporting Methods: Pressures
The pressure score, , describes the cumulative impact of ecological and social stressors in a given year and region which tend to depress the goal score in future years. Pressure scores range from 0 to 1, and include both ecological ( ! ) and social pressures ( ! ), such that: where = 0.5 is the relative weight for ecological vs. social pressures categories. We default to equal weighting as little evidence was available to justify or quantify unequal weights between ecological and social pressures categories. It may be that future work can inform unequal weighting terms for individual goals.
For each goal, subgoal, or goal element (e.g. specific habitat), we calculated pressures as an impactweighted cumulative impact for each pressure category !"#$ !"#$ and !"# !!"# . Impact weights are based on a goal's sensitivity ! ! !"#$ to specific stressors ! ranked as low , as determined by peer-reviewed literature and expert judgment (S6 Fig. shows the matrix of stressors, goals, and weights). The denominator represents the maximum stressor impact weight for that category and goal. If cumulative pressure load for a goal/component combination exceeds the maximum possible stressor intensity, we cap it to 1.0, i.e. the equivalent to an individual stressor at maximum stress and intensity.

Ecological pressure
We included five subcategories of ecological stressors relevant to British Columbia: fishing pressure, habitat destruction, climate change, water pollution, and species introductions (invasive species and genetic escapes). Each pressure category may include several stressors in individual layers. The intensity of each stressor within each OHI region is scaled from 0 to 1, with 1 indicating the highest stress relative to a defined reference point, often the highest observed stress within the OHIBC study area.
The overall ecological pressure, ! , acting on each goal for each region and year was calculated as the weighted average of the pressure scores, , for each subcategory, , acting on that goal, with weights set as the maximum rank in each pressure category ( !,!"# ) for each goal, such that: Stressors that have no impact (i.e. ! ! !"#$ = ) drop out of the calculations and do not affect the pressures score.
A note on ecological pressures not included in this assessment: A number of likely significant pressures on BC's coastal ecosystems were not able to be included in this assessment. For example, we were unable to include impacts of terrestrial mining or log boom presence due to lack of data availability at a usable spatial and time series resolution. In future assessments, additional stressors can easily be incorporated into the pressures matrix as new data become available, though consideration should also be given to potential resilience measures that might ameliorate the impacts of those additional stressors.

Social pressures
Social pressures describe the lack of effectiveness of government and social institutions. Social stressors are described for each region and year on a scale of 0 to 1 (with one indicating the highest pressure).
The Community Well Being (CWB) Index [39] produced by Indigenous and North Affairs Canada combines indicators including education, labour force activity, income and housing to provide insights into the social well being of Indigenous and non-Indigenous communities in Canada. We calculate social pressures as the population-weighted average of community-level CWB scores, subtracted from 1 to indicate that low community well being indicates ineffective governance and social institutions.
This component is calculated separately for all communities in each OHIBC region, applied to goals describing benefits to all BC residents, and for First Nation communities specifically (as noted by census subdistrict designation), applied to goals only applicable to First Nation communities. Maximum pressure ( !"# = 1) occurs when all CWB indicators are at 0 out of 100, while minimum pressure ( !"# = 0) occurs when all indicators are at 100.
The CWB is also used as an indicator of social resilience, as described below.
A note on the assumption of linear and additive response to pressures: As in the global OHI pressures model, we assume for this OHIBC assessment that is that each goal responds to changes in intensity of ecological stressors in a linear and additive fashion. Such an assumption obviously fails to capture likely non-linear responses and synergistic or antagonistic interactions among stressors, but such responses remain poorly characterized so we could not justify including such responses in our model.

Supporting Methods: Resilience
Resilience for each goal and region, , is based on three components: ecological integrity, !"#$ ; regulatory efforts that target specific ecological pressures, !"# ; and social integrity, !"# . The !"#$ and !"# combine to address resilience to ecological pressures, while !"# addresses social pressures. Each resilience category contains one or more layers reflecting the magnitude of resilience within each region for each year; layers are "activated" to address specific pressures acting on specific goals based on a resilience matrix (S7 Fig.), and active layers are summed to determine a score for each resilience category. Each layer is constrained from 0 to 1.

Ecological integrity
An intact biodiverse ecosystem provides general resilience to ecological pressures by ensuring the system's ability to maintain functionality in the face of stressors imposed by human activity and climate change. For OHIBC, we consider the area-weighted average conservation status of all species found in the coastal zone (3 nmi offshore) (as resilience to coastal pressures) and found within the entire EEZ (as resilience to pressures not limited to the coast). The area-weighted average conservation status is calculated in the same manner as the Species subgoal.

Regulatory resilience
Regulatory resilience describes the institutional structures, rules, and regulations that directly address ecological pressures from human interactions with the marine system. For OHIBC we examined regulatory resilience to address three categories of pressure that correspond with : fisheries/biomass removal, habitat destruction, and aquaculture.
Where possible, we scored regulatory resilience based on a combination of a) existence of meaningful regulation, b) enforcement of regulation, and c) compliance with regulation.

Aquaculture regulatory resilience
Aquaculture regulatory data [34,40] are not spatialized to the OHIBC region level, so scores are calculated across the entire BC EEZ and applied equally to all regions.
• Presence: The existence of these data implies existence of regulation; score of 1 across all years.
• Enforcement: This is based on frequency of audits relative to some reference point.
-For enforcement we combine scores for (fish health audits)/(active facilities), (sea lice audits)/(active facilities) and (benthic surveys)/(active facilities) using a geometric mean: enforcement across all facets must be high to achieve a strong resilience score. Poor enforcement on any facet indicates weakness in regulatory enforcement. -For fish health and benthic surveys, the reference point is the max seen across all years for each metric. Score is Overall resilience score is calculated as: !" = reg presence + reg enforcement + reg compliance 3

Marine Protected Area regulatory resilience
Marine protected areas shield biodiversity from pressures due to fisheries exploitation and habitat destructive practices. Data on MPAs comes from UNEP WDPA [24], BC Province Parks and Ecologically Protected Areas, and tribal parks. Reference point for MPA coverage is 30% of marine area [44]. Specific enforcement and compliance data are not readily available, so we use existence of a management plan for each MPA [42] as a proxy for management effectiveness; ideally, all MPAs would be subject to a published management plan.
MPA resilience is calculated at two scales to account for pressures that act at different scales: system-wide pressures (entire EEZ) and coastal pressures (the 3 nmi coastal zone).
MPAs with management plan MPAs total

Fishing management regulatory resilience
Fishing regulations increase ecological resilience by limiting unintended biomass removal. Regionspecific data were not available, so we calculated scores for the overall BC EEZ and applied scores equally to all regions.
• Presence: For all years of the study, the Fisheries Act has been in place; therefore . = 1 for all years. • Enforcement: We use as a metric of enforcement the number of fisheries officers per fishing vessel !""#$%&' / !"##"$# for each year; as a reference point as the maximum observed officers per vessel for any year. License data was supplied by DFO [15] and fisheries officer count is based on groundfish enforcement [41].
• Compliance: As a metric of compliance, we use observer coverage in groundfish fishery [41]. For all years, observer coverage is reported as 100%, so . = 1 for all years.

Social resilience
Social resilience describes the social integrity of coastal communities that allow for adaptive responses to social and ecological pressures. We calculate social resilience scores by region separately for First Nation communities only (for First Nations-specific goals and subgoals) and for all communities (for all other goals and subgoals).

Community Well Being Index
The Community Well Being Index (CWB) [39] informs both our social pressures (as low scores indicate lack of effective social institutions) and our social resilience (conversely, high scores indicate functional social structures).
As for pressures, this component is calculated separately for First Nation communities only and for all communities in each OHIBC region.

MaPP
The Marine Plan Partnership involved eliciting input and advice from member First Nations and BC Province experts to develop marine plans based on the best available science and local and traditional knowledge. MaPP resilience accounts for the adaptive benefits of engaging in the planning process beginning in 2011, as well as a presumption of improved compliance and selfmonitoring once the plans were announced in 2015.

Supporting Methods: Data Selection Criteria
OHIBC incorporates 76 layers, constructed from dozens of datasets across social, economic, and environmental domains, to calculate status, pressures, and resilience for each goal. Ideally, every dataset would be an excellent "fit" for the needs of the calculation. In addition, each dataset would ideally provide the spatial and temporal resolution to allow OHIBC scores to distinguish differences in each goal among regions and from year to year, and the spatial and temporal extent to adequately assess the entire region across the entire study period. We ranked each OHIBC dataset across three dimensions to identify strengths and weaknesses, and to highlight data gaps where effort and resources could increase the utility of a dataset to this OHIBC assessment. Note that these rankings are based on criteria specific to OHIBC, and may not reflect the utility of the dataset to an assessment at a different scale.

Methods
We identified three dimensions of data that affect the ability to calculate meaningful goal scores and one dimension that pertains to OHI's open science philosophy. For each of these dimensions, a dataset was scored 0.0, 0.5, or 1.0 on multiple facets as applicable (S13 Table). The dimensions and methods are loosely based upon methods described in Fritz et. al. [46].
• Spatial dimension: OHIBC aims to identify differences and patterns in goal status, pressures, and resilience across the seven regions included in the assessment.
-Spatial extent: Ideally, spatial data encompass the entire area of interest, i.e., Canada's Pacific EEZ out to the shelf break. Scored as 1.0 if the dataset includes data across the entire study area of interest; 0.5 if the dataset includes most but not all OHIBC regions (e.g., MaPP regions only); and 0 if the dataset includes a minority of OHIBC regions. -Spatial resolution: Ideally, spatial data have sufficient resolution to distinguish between two neighboring OHIBC regions. Scored as 1.0 if the average spatial resolution is less than half the average area of OHIBC regions (e.g., census subdistricts; 4 km rasters of groundfish catch; Pacific Fishery Management Subareas); scored as 0.5 if the spatial resolution is on the order of the area of OHIBC regions (e.g, Pacific Fishery Management Areas; 0.5° rasters of species range and marine debris); and 0.0 if the data do not provide sufficient information to distinguish among regions (e.g., salmon stocks spatialized by river systems that do not indicate distribution in marine waters; provincelevel data on fisheries officers). Scored as 1.0 if the dataset resolution is less than or equal to one year; 0.5 if the resolution is less than or equal to 10 years; and 0 if the dataset is static. -Temporal baseline: For those goal models that compare current condition to a historic reference point (note, not the same as trend). Scored 1.0 for data that allow comparisons to a benchmark at least 50 years prior; 0.5 for data that allow a benchmark at least 10 years prior. • Fit dimension: OHIBC aims to capture a broad range of benefits afforded by a healthy marine social-ecological system, as well as the pressures and resilience that moderate those benefits. This dimension assesses how closely the available data "fit" the needs of the OHIBC target calculation. This is rather subjective, as in some cases the available data were chosen to fit a goal model, while in others, a goal model required modification to accommodate the available data.
-Fit extent: Does the dataset adequately capture the full range of conceptual understanding required by the target calculation? Scored as 1.0 for data that inform understanding across the entire system (e.g., species condition information was available for nearly all the iconic species identified for the Iconic Species goal); scored 0.5 for data that may represent only a portion of benefits (e.g., species condition for the Species goal is limited to a subset of taxa assessed by IUCN and COSEWIC; the Salmon goal is based on a limited selection of indicator stocks). No datasets were scored 0.0. -Fit resolution: Does the dataset allow for detailed exploration of goal status, pressures, or resilience within the broader context? Scored as 1.0 for datasets with a rich breakdown of categories or sectors (e.g., stock assessment and harvest data available for individual stocks; fishing license data can discriminate between First Nations and non-First Nations types); scored as 0.5 for datasets with some internal detail (e.g., census data include income and employment by very broad sectors in addition to overall); and 0.0 for data where finer-scale divisions are not available (e.g. aquaculture production potential is based on global averages but cannot be separated to identify potential for BC-specific species).

Inclusion/exclusion of datasets based on these dimensions
Scoring datasets in this manner provides a useful heuristic to guide selection of datasets, by enabling comparison of the tradeoffs betwen two sets that may convey similar information. As an example, we can examine two datasets that were considered but not used in the OHIBC assessment. To inform our Wild-Capture Fisheries goal calculation, we used species-level catch data from DFO [6], available at fine resolution across the BC EEZ but spanning only a portion of the study period. We also considered data from the Sea Around Us Project [47] which provides catch reconstruction data at 0.5° spatial resolution, annually going back decades. Scoring the two data sets side by side, we see identical scores, trading spatial resolution for temporal resolution, at which point second-order criteria can come into play, in this case a preference for direct catch estimates over modeled catch reconstructions. Similarly, we can compare aquaculture production datasets: we used DFO aquaculture production by Pacific Fishery Management Area [8], available for a short span of years, but also considered province-level estimates [48] spanning the entire study period. Again, the two datasets earn identical scores, trading spatial extent for temporal extent. Here we chose the spatially explicit data as more compatible with our production potential dataset [9]. Note that this methodology as applied here implicitly places equal weights on each dimension, but preferentially weighting temporal qualities over spatial qualities (for example) could suggest different data selection decisions.

Goal-level scores
Goal-level scores are the average across all layers and facets used to calculate the goal (Table S14).
As some layers provide more information than others (e.g. spatial-temporal vs. simply spatial), these layers contribute a greater weight to the goal score. For this reason, a goal's overall score may not be equal to the mean of its facet scores. In some cases, complementary datasets are included to improve extent (e.g. tribal parks to supplement parks and protected areas from WDPA). These complementary layers are combined prior to calculating the goal-level facet scores. The combined layer sums the full facet scores for the primary layer with half the facet scores for the secondary layer(s), with a maximum total value of 1. S14a