Study area

We explore the risk of habitat degradation from human activities along the Atlantic coast of the United States, focusing on two regions of ocean governance: The Northeast region that stretches from Maine to Connecticut, and the Mid-Atlantic region that continues from New York to Virginia. These regions are the first two areas along the coast of the United States to establish Regional Planning Bodies (RPBs) under the U.S. National Ocean Policy. These areas critically rely on adjacent marine habitats for livelihoods and well-being: in 2013, economic activity connected to the ocean in these states represented 935 thousand jobs and $63 billion in GDP [48]. Successful marine spatial planning efforts at the state level, such as those in Rhode Island [49] and Massachusetts [47] provided the preconditions for the U.S. Northeast to establish the first RPB, followed soon thereafter by the Mid-Atlantic RPB. The planning processes for these regions resulted in final ocean plans in December of 2016 that focus on ecosystem health, data collection (including extensive data portals), compatibility among uses, and effective decision making [34,35]. These plans are living documents, supported by ongoing engagement by the RPBs and supporting institutions, including the Northeast Regional Ocean Council (NROC) and the Mid-Atlantic Regional Council on the Ocean (MARCO), and are meant to provide actionable information to coastal decision-makers on an ongoing basis.

In this paper, we advance the science priorities of the Northeast and Mid-Atlantic regions by synthesizing the wealth of human use data available on regional portals and then assessing the impact these uses have on marine ecosystems. The Northeast Ocean Plan details an array of science priorities, including a priority to advance EBM and to improve our understanding of human uses and how these uses affect marine resources. The Mid-Atlantic plan echoes these priorities, and emphasizes cataloguing and synthesizing human uses through the Human Use Data Synthesis project [50]. Moreover, this work will provide spatial anthropogenic ecosystem risk data for NOAA’s Northeast U.S. Continental Shelf Integrated Ecosystem Assessment Program, in particular as part of its Ecosystem Status Report effort to characterize baseline ecosystem condition and ecosystem service provision across our study area.

Estimating risk

To assess the influence of multiple human activities on coastal and marine ecosystems, we apply the InVEST Habitat Risk Assessment (HRA) model. The HRA model is a quantitative approach to evaluating the cumulative influence of stressors associated with human activities on habitats [23,51] and is available as part of the open-source InVEST ecosystem service modeling software [52]. HRA uses a well-established approach [51,53–55] from the risk literature that originates from fisheries vulnerability assessment [37,39,56] and was subsequently adopted for ecological risk assessment [23,51]. A central feature and strength of the HRA model is that it explicitly breaks down exposure and consequence in a risk framework and, in so doing, offers insight into effective management strategies. To assess the exposure of ecosystems to activities and the consequence of that exposure, synthesizes information about interactions between multiple human activities and multiple habitats using both spatial and non-spatial data from the peer-reviewed literature, grey literature, and expert opinion. Below, we explain the concept of exposure and consequence, the specific criteria we used in this analysis, where we derived scores for each criterion, and then walk through how these pieces fit together to estimate risk.

The HRA model assesses the risk of habitat degradation from each stressor based on several exposure and several consequence criteria [51]. Exposure is the degree to which a habitat encounters a stressor and consequence is the habitat-specific result of that exposure. Both exposure and consequence are made up of multiple criteria that can be defined by the user. In this analysis, exposure criteria include spatial data of the footprint (i.e. extent) and intensity of the stressor as well as non-spatial information about the frequency of occurrence and scale of effect of the impact (see Supporting Information and Table A in S1 File for additional details). Consequence criteria include change in biomass, trophic impact, and expected recovery time (see Supporting Information and Table A in S1 File for additional details). For each habitat-stressor combination, the model includes unique information (a.k.a., ‘scores’) for each exposure and consequence criteria. As an example of one score, we specify in the model that for seagrass beds, the recovery time is longer after dredging than recreational boating (see the data repository for the full compilation of scores).

We use the scores developed through the Massachusetts cumulative impact assessment (detailed in Kappel et al. [57]), which was based on expert elicitation, and modify them for use in the HRA exposure and consequence framework. This expert elicitation asked scientists to assess, for each cumulative impacts criterion (which match those exposure and consequence criteria explained above), the effect of each stressors on their habitat of expertise. Although inputs to the Habitat Risk Assessment model are traditionally informed by peer-reviewed literature, the breadth of this analysis in the Northeast and Mid-Atlantic (i.e., 13 habitats and 31 stressors) made this particularly challenging. Further, the use of scores from Kappel et al. [47] allows for the two approaches to be compared and used together. Past comparisons of scores between Massachusetts and California [57,58] demonstrate the transferability of scores beyond the area for which they were originally derived and provide us with the confidence to extrapolate cumulative impacts scores from Massachusetts to the broader Atlantic coast. We modify Kappel et al. [57] by weighting the components of exposure and consequence equally (as described in Eq 3) and by binning their continuous scores into three categories. This approach formats the scores for HRA and improves interpretability while preserving the relative rank of scores and the relative importance of high scores (see S1 File for additional details).

We use these scores—for each criterion of exposure and consequence for each habitat-stressor combination—in the model to estimate cumulative risk for each habitat and across ecosystems. Here we use a 1km cell size as the spatial processing unit and resolution of habitat risk outputs based on a conservative consideration of the resolution of the human activity inputs (Table C in S1 File). We first average (1) the set of exposure criteria and (2) the set of consequence criteria for each habitat-stressor combination (Eqs 1 & 2): (Eqs 1 & 2) where E_jk is the exposure score specific to habitat j, from stressor k; C_jk is consequence score, e_ijk is the exposure rating criterion i, specific to habitat j and stressor k; c_ijk is the consequence rating. When the data were available (Table C in S1 File), we incorporated spatial variation in intensity into exposure. These equations result in a single exposure score and single consequence score for each habitat-stressor combination.

We then estimate risk associated with each habitat-stressor combination by calculating Euclidean distance to the origin, where the average exposure score is on one axis and the average consequence score is on the other (Eq 3): (Eq 3) where risk, R_jkl, to habitat j caused by stressor k in each location (cell) l is the Euclidean distance from the origin in an exposure-consequence plot. We map this habitat-stressor specific risk score where the habitat and stressor overlap in space (based on data presented in Tables B-C in S1 File and described in ‘habitats’ and ‘human activities and stressors’ sections below).

Cumulative risk to each habitat is the sum of the risk scores related the human activities co-occurring in each location containing that habitat: (Eq 4)

Ecosystem risk for each grid cell is the sum of habitat risk scores in that cell: (Eq 5)

Ecosystem risk (Eq 5) and cumulative risk to each habitat (a.k.a. ‘habitat risk’; Eq 4) provide the foundation of the results outlined in corresponding results sections, ‘Ecosystem Risk’ and ‘Habitat-Specific Risk’. We use a Tukey honest significant difference test with 95% family-wise confidence level to evaluate inter-regional differences in habitat risk (computed using HSD.test in the agricolae package for R). In addition, we use the breakdown of exposure and consequence within the risk equation (Eq 3) to inform the ‘Causes of Risk’ section of the results.

Habitats

To map risk to habitats, we first classify the Atlantic coastal area—from Maine to Virginia—into 13 habitat types, ranging from nearshore habitats to deeper, offshore zones. To build on and to allow for comparisons with previous work in the region [47,57], we use the habitat classification scheme from the cumulative impact assessment of Massachusetts waters. We include eight nearshore habitats—beaches, tidal flats, salt marsh, rocky intertidal, seagrass, algal, nearshore soft-bottom, nearshore hard-bottom—and five offshore habitats—soft-bottom shelf, hard-bottom shelf, bathyal shelf (waters deeper than 200m), shallow pelagic, and deep pelagic. Spatial data for these habitats come from a combination of trusted regional data portals (i.e. Northeast Ocean Data and Mid-Atlantic Ocean Data portals) and federal data sources (i.e. NOAA’s Environmental Sensitivity Index, National Wetlands Inventory, NOAA’s U.S. Coastal Relief Model, U.S.G.S Continental Margin mapping of sediment). As necessary and in order to use the most recent data and ensure full spatial coverage, we merge data from multiple sources. In the case of habitats defined by sediment and depth (e.g. soft- and hard-bottom nearshore and shelf), we intersect sediment and depth data to define the spatial extent. The use of recent, reliable, and publically available data allow for replication in new locations and adaptive management as data improve (see S1 File, Table B in S1 File, and Fig A in S1 File for additional details on these data).

Human activities and stressors

To assess the cumulative risk of human activities to coastal and marine habitats in the region, we incorporate the impact of 31 stressors. It is standard practice in the risk-assessment community to use the term “stressor,” which is a value-laden term. It is important to recognize that the human activities represented in this analysis as “stressors” also enhance people’s lives by providing food, recreational opportunities, livelihoods, and more. The approach we present here acknowledges that the activities used to derive benefits can also adversely affect the very habitats that are the source of these benefits to people.

Using Kappel et al.’s analysis [47] as a starting point, we include stressors available in regional and national data portals and those specified as important by regional planning bodies. We roughly group these stressors into six themes: (1) fishing and aquaculture, (2) human structures, (3) coastal and marine uses/impacts, (4) land-based impact, and (5) climate change (Table C in S1 File). Within the fishing and aquaculture theme, we coalesce fishing into five discrete categories of use: artisanal, non-destructive demersal fishing, destructive demersal fishing, pelagic, and recreational, based on similarity of impact. Aquaculture includes three categories consisting of finfish predators, shellfish, and marine plants (Fig B in S1 File). The second theme, human structures, includes benthic structures, coastal engineering, and three types of energy-related structures (wind, liquid-natural gas, and tidal; Fig C in S1 File). For the third theme, the most inclusive, we categorize additional marine uses and impacts into human trampling, dredging, military activity, shipwrecks, ocean dumping, ocean mining, oil spills, shipping, and five types of tourist activities (wildlife viewing, recreational boating, surfing, kayaking, and SCUBA; Figs D-F in S1 File). The fourth theme, land-based impacts, includes power plants, inorganic, sediment, nutrient, and light pollution (Fig G in S1 File). Finally, to explore climate change, the last theme, we capture the relative rate of increasing sea surface temperatures based on temperature anomalies relative to historical means (Fig H in S1 File). While other measures of climate change are important, we focus on sea-surface temperature because it affects all included habitat classifications, these effects are well documented and understood relative to other measures of climate change (e.g., increasing UV light), and the spatial data are readily available (see S1 File, Table C in S1 File, and Figs B-H in S1 File for additional details on these data).

We selected data for this analysis to maximize reliability and reproducibility (see Table B in S1 File). Most data come from regional portals or federal sources, are publically available (with the exception of commercial fishing), are fewer than five years old (with the exception of nutrient loading and coastal engineering for some states), and have full spatial coverage across our area of interest (with the exception of aquaculture). For many human activities we use the data as provided; in other instances, we merge multiple data sets to create complete spatial coverage, add spatial buffers following Kappel et al.’s [47] protocols, convert raster to vector for use in the model, and/or bin spatial variation to define categorize intensity (see S1 File and Table B in S1 File for additional details). In the absence of consistently reported observed data throughout the study region, we model inorganic pollutions (using impervious surfaces as a proxy for non-point source pollution) and nutrients loading (using the USGS SPARROW model [59]) from upland land-use to coastal pour-points and estuaries (see Supporting Information).

The data reveal spatial variation in the intensity of several stressors. As applicable and as data allow, we model spatial variation in intensity for each stressor. To do this we assign the lowest quartile of use or impact (e.g., fishing effort, nutrient loading) the lowest intensity and the upper quartile the highest intensity, with the middle quartiles composing moderate intensity. This spatially-explicit intensity score is included as an exposure criterion in the risk equation (Eq 1). See the S1 File for additional details on this approach and how it was applied to various stressors.

Ecosystem risk

The highest levels of risk are concentrated along the coastline and along the continental shelf, though all areas throughout the Northeast and Mid-Atlantic region experience some degree of risk (Fig 1A). Indicating a skewed distribution of risk in which fewer areas are at particularly high risk, ecosystem risk ranges from 0 to 106 across both regions, with a median risk score of 15.8 and the upper quartile of risk at 18.7. Areas of particularly high risk occur along the entirety of the coastline. However, this zone of high coastal risk visible across both regions varies at finer spatial scales. For example, high risk areas occur in wider swaths along the coastlines of Delaware, Massachusetts, New Hampshire, and Maine and occur in a narrower, more variable stretch along the coastlines of Virginia and Maryland (Fig 1)[60]. A second area of higher risk is a 40 to 100-km wide zone at the edge of the continental shelf, stretching from a distance of 125km offshore (at the Virginia-North Carolina border in the south) to 425km (off the coast of Maine in the north; Fig 1A). Areas of lower risk (less than the median of 15.8) occur primarily in the Mid-Atlantic from distances just beyond the high-risk coastline (from 5 to 20km off shore) (Fig 1A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Ecosystem risk and overlapping stressors for the Northeast and Mid-Atlantic at different scales.

In all maps, warmer colors represent greater risk or a greater number of overlapping human activities. (A) Ecosystem risk for the entire study area. Median risk is 15.8. (B) Number of overlapping stressors for the entire study area. The number of stressors is broken down by decile to allow for comparison with risk deciles. (C) Ecosystem risk map for Cape Cod, Massachusetts. The northern star is Boston Bay and the southern star is Plymouth Bay. (D) Ecosystem risk map for Delaware and Chesapeake Bay. (E) Ecosystem risk map for the Maine coastline. (F) Number of overlapping stressors for Cape Cod, Massachusetts. (G) Number of overlapping stressors for Delaware and Chesapeake Bay. (H) Number of overlapping stressors for the Maine coastline.

https://doi.org/10.1371/journal.pone.0188776.g001

The risk assessment approach provides spatially explicit information that highlights variation in in risk at multiple scales. While greater risk in the nearshore may be an intuitive result, our results show that nearshore risk is generally lower and less concentrated in the Mid-Atlantic than in the Northeast. Within the Mid-Atlantic, high risk areas are concentrated in northern New Jersey and at the mouths of the Delaware and Chesapeake Bays, while the rest of the Mid-Atlantic coastline is at lower risk. The Northeast coastline contains greater area at the highest risk levels, with risk concentrated along the coastlines of Massachusetts, New Hampshire, and Maine (Fig 1). The high risk area extends further offshore throughout the Gulf of Maine and Long Island Sound than it does along other sections of the coast. Illustrative of variation at the state scale, ecosystem risk is highest in Boston and Plymouth Bays (Fig 1C). And within a single bay, for example Cape Cod, risk is greater off the tip of the Cape Cod and along the coast than in the interior bay (Fig 1D). Beyond the nearshore area, areas of risk are even more spatially variable, with higher risk patches east of Long Island, along the northern extent of the U.S. continental shelf, and sporadically through the Gulf of Maine. Each of these spatial scales—comparable to regional, state, and bay-wide planning processes—reveals areas of particular concern and adds detail to our general understanding of coastlines at risk.

Ecosystem risk maps, which incorporate habitat locations as well as the exposure and consequence of each stressor on each habitat, go beyond simple un-weighted overlays of human activities and identify different areas (Fig 1). While overlay analysis illuminates the nearshore area as a place of potential conflict, it misses the especially high-risk waters (i.e. top decile) of Massachusetts and Maine, where the multitude of coastal habitats in close proximity suggest additional attention by managers may be needed. Risk analysis also highlights the high-risk offshore areas (Fig 1A) where there are few overlapping stressors (Fig 1B), demonstrating the relative importance of fewer, higher consequence stressors in the offshore area. Conversely, for areas 5 to 45km offshore (just beyond the immediate high risk coastline), and especially in the southern Mid-Atlantic, risk is lower (Fig 1A and 1D) than the number of activities would suggest (Fig 1B and 1G) because local activities are less impactful. Differences also emerge at finer spatial scales, for example in Delaware and Chesapeake Bays. While a simple overlay of stressors highlights much of Delaware Bay as a zone of many activities (Fig 1G), our cumulative risk analysis suggests the center of the bay faces the greatest threat (Fig 1D) resulting from high consequence activities like shipping and dredging. In the Chesapeake Bay, risk analysis highlights the importance of small bays and tributaries more than the simple overlay of stressors does (Fig 1D and 1G), reflecting the importance of land-based stressors in this bay. Patterns of population density along the coast [18] are evident in both sets of maps, but the overlay of human activities and risk analysis tell an increasingly rich story. Risk analysis provides more information than population density or overlap analysis because it includes the distribution of habitats as well as interactions between habitats and stressors. The following sections further explore the added value of these components.

Habitat-specific risk

Patterns of risk further vary between habitats and across regions. The RPBs did not formally delineate boundaries between the Northeast and Mid-Atlantic planning areas, choosing to have some overlap in the waters off of New York. For the purposes of quantitative comparison between regions, we bisect the two planning areas with a line running from New York Harbor to the continental shelf near Hudson Canyon. This delineation is consistent with other research efforts supported by the RPBs under the National Ocean Policy [61].

Across regions, risk is highest for rocky intertidal, hard-bottom shelf, tidal flats, and soft and hard-bottom nearshore habitat. Higher median scores for these habitats indicate a greater area at high risk. Between regions, habitats are generally at greater risk in the Northeast than in the Mid-Atlantic. Risk is higher in the Northeast for tidal flat, seagrass, marsh, algal, soft-bottom nearshore habitat, soft-bottom shelf habitat, shallow and deep pelagic, and bathyal shelf habitats. Habitat risk is higher in the Mid-Atlantic only for rocky intertidal, hard-bottom shelf, hard-bottom nearshore, and beach habitat (Fig 2). These inter-regional differences in habitat risk are significant for all habitats except soft- and hard-bottom nearshore, hard-bottom shelf areas, and tidal flats (based on a Tukey honest significant difference test with 95% family-wise confidence level). The variation in risk scores indicates that risk to particular habitats is not homogenous across the area of analysis. For example, the wide range of risk to tidal flats in both regions indicates spatial variation in risk, whereas the narrow range in risk to soft-bottom shelf area indicate relatively homogenous risk across the area of analysis (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Distribution of cumulative risk for each habitat in each region.

Boxes represent 25^th to 75^th percentiles, the bar represents the median, and dashed lines represent minimum and maximum values, having excluded outliers. Gray boxes represent habitats in the Mid-Atlantic and white boxes represent habitats in the Northeast. Letters represent groups of habitats that have statistically similar means from a Tukey honest significant difference test with a 95% family-wise confidence level.

https://doi.org/10.1371/journal.pone.0188776.g002

Tidal flats are one of the most at-risk habitats and this risk varies spatially. There are over 550sq. km of tidal flats in the Northeast and an additional 190sq. km in the Mid-Atlantic. Variation in risk is evident in the Northeast along the coastline of Connecticut, Rhode Island, and New Hampshire, where risk to tidal flats is predominantly in the upper quartile (quartile defined by both regions; Fig 3). In Maine, tidal flats south of Portland are at greater risk than those further to the north. Along the Massachusetts coast, tidal flats around Massachusetts Bay are at the greatest risk, while those along Cape Cod Bay and Nantucket Sound face lower risk. In the Mid-Atlantic, tidal flats at the greatest risk in Maryland along the northern section of Assateague State Park and in Virginia at the mouth of the Chesapeake Bay near Virginia Beach and Newport News (Fig 3). While the multitude of nearshore human activities affects risk to tidal flats everywhere, risk to these high risk areas is driven in particular by additional tourism (recreational boating, wildlife viewing, and SCUBA diving) as well as land-based stressors (e.g. power plants, nutrient loading, and light pollution), which do not occur in combination in lower risk areas. Given the importance of tidal flats for flood mitigation and fisheries, among other benefits [62], habitat risk maps offer an opportunity for regional, state, and bay-wide efforts to focus management towards specific locations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Habitat risk map for tidal flats and shelf habitats.

Tidal flats occur along the coastline and shelf habitats are distant from the coastline. Note that risk deciles are scaled for each habitat independently. Shelf habitats include both soft and hard-bottom shelf habitat categories.

https://doi.org/10.1371/journal.pone.0188776.g003

Soft- and hard-bottom shelf habitat, often some of the most productive habitats in the ocean, are the offshore habitats most at risk. Hard-bottom shelf is the habitat most at risk after tidal flats and covers over 4,500sq. km in the Northeast and just 150sq. km in the Mid-Atlantic. Soft-bottom shelf habitat covers nearly 72,500sq. km in the Northeast and 22,230sq. km in the Mid-Atlantic. Mean risk values are similar between the two regions for both soft- and hard-bottom shelf (Fig 2). In the Northeast, these habitats are at the greatest risk in the interior Gulf of Maine; risk decreases with increasing distance from the coast. A second area of relatively high risk exists offshore off Long Island and Connecticut, with lower, more variable risk areas spanning to the north and to the south into the Mid-Atlantic (Fig 3). Risk in these locations is driven, in part, by the higher intensities of non-destructive demersal fishing in combination with recreational fishing, wildlife viewing, shipping, and rising sea surface temperatures.

Causes of risk

Across habitats and locations, rising sea-surface temperatures and commercial fishing are the individual stressors responsible for the greatest amount of risk, yet the unique combination of stressors that affect categories of offshore and nearshore habitats as well as each individual habitat offer a more nuanced and informative explanation. For offshore habitats (depths > 30m; bathyal shelf, soft-bottom shelf, hard-bottom shelf, shallow pelagic, and deep pelagic), 75% of risk is explained by eight human activities: warming sea surface temperatures, four kinds of commercial fishing, shipping, recreational fishing, and wildlife viewing (Fig 4A). Specifically, for soft and hard-bottom shelf habitats, rising sea surface temperatures, four kinds of commercial fishing, and shipping collectively account for half to three-quarters of total risk across the area of analysis. In contrast to offshore habitats, for nearshore habitats (i.e., beaches, tidal flats, salt marsh, rocky intertidal, seagrass, algal, nearshore soft-bottom, nearshore hard-bottom), a larger suite of stressors contributes to risk; 15 stressors, each contributing a smaller percentage to overall risk, are collectively responsible for 75% of risk. Human trampling, light pollution, coastal engineering, and nutrient loading—in addition to the key stressors identified for offshore regions—each contribute more than 5% to risk, and many more contribute smaller percentages (Fig 4A). The greatest contributors to risk to tidal flats across both regions are warming sea surface temperatures, adjacent human populations (e.g. through trampling), coastal engineering, and land-based nutrient loading, sediment, and inorganic pollution; these seven stressors are collectively responsible for almost half the total risk to tidal flats. The effect of these stressors is evident and important in specific locations: the area of high risk off the Massachusetts, New Hampshire, and Maine coasts corresponds with the highest intensities of non-destructive demersal fishing; the area of high risk off of Long Island and Connecticut is associated with non-destructive demersal, artisanal, and pelagic fishing; areas of fine-scale low risk to the east of Cape Cod are associated with lower shipping density; and areas of low risk to the north, southeast, and south of Cape Cod are areas where fishing is restricted (Fig 3 and Supporting Information).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Causes of risk for nearshore and offshore habitats.

(A) Relative contribution of different human activities and stressors to total risk for nearshore and offshore habitats types. The relative contribution is averaged across all nearshore or offshore habitats. Nearshore habitats include beach, salt marsh, tidal flats, rocky intertidal, algal, seagrass, soft-bottom nearshore, and hard-bottom nearshore habitats. Offshore habitats include soft-bottom shelf, hard-bottom shelf, bathyal shelf, shallow pelagic, and deep pelagic zones. (B) Risk plot for tidal flats and (C) hard shelf habitat (combined across Northeast and Mid-Atlantic); each point represents a unique human activity or stressor. For each habitat, included human activities (points) are cumulatively responsible for 80% of risk. (D) Risk plot for nearshore (green) and offshore (blue) habitats. Points represent the unique exposure and consequence from a particular stressor. Points shown for each habitat category cumulatively contribute 80% of risk. Individual human activities or stressors contributing more than 5% to risk are labeled (SST = increasing sea surface temperatures, Fish DD = demersal destructive fishing, Fish NDD = non-destructive demersal fishing, Fish A = Artisanal fishing, Tramp = human trampling, C Eng = coastal engineering).

https://doi.org/10.1371/journal.pone.0188776.g004

Risk can be further deconstructed to determine how habitat types are variably affected by the exposure and consequence of stressors (Fig 4B–4D). Of the 15 stressors (ranging from rising sea-surface temperatures to coastal engineering to trampling) that cumulatively account for 75% of risk to tidal flats, average exposure (i.e., from spatial extent, intensity, frequency of occurrence, and scale of effect) varies between these stressors, ranging from 1.52 to 2.74 (where 3 is the maximum exposure score). In contrast, the consequence (e.g., change in biomass, trophic impact, and recovery time) of these multiple stressors to tidal flats is consistently lower (ranging from 1.21 to 1.7 with one outlier at 2.1) (Fig 4B). Because each individual stressor results in a relatively minor change to habitat quality (i.e., has a low consequence score), reducing risk to tidal flats will require reducing exposure to a number of individual stressors (Fig 4B). In contrast, the eight stressors—from warming sea surface temperatures to fishing and shipping—that account for 75% of the risk to hard-bottom shelf habitat have greater exposure (e.g., from greater temporal and spatial overlap or higher intensity) and consequence (i.e., greater impact on habitat quality) (Fig 4C). Reducing exposure from a single stressor, either completely or by limiting overlap, intensity, or frequency of occurrence, will reduce risk by a relatively large amount for these habitats. Nearshore and offshore habitats follow a similar pattern to tidal flats and hard-bottom shelf, respectively, where risk to nearshore habitats is caused by a multitude of stressors with relatively low consequence, and risk to offshore habitats is caused by fewer stressors with greater exposure (Fig 4D). Exposure and consequence information can be used to improve habitat quality; reducing risk to nearshore habitats will require simultaneously addressing the frequency of occurrence and intensity of multiple human activities and reducing risk to offshore habitats requires more targeted efforts to reduce exposure from specific threats.