A human impact metric for coastal ecosystems with application to seagrass beds in Atlantic Canada

Coastal biogenic habitats are particularly vulnerable to cumulative human impacts from both terrestrial and marine realms. Yet the broad spatial scale used in current global or regional approaches of quantifying multiple anthropogenic stressors are not relevant to the local or bay-wide scales affecting most coastal biogenic habitats. To fill this gap, we developed a standardized human impact metric to quantify the magnitude of anthropogenic impacts to coastal ecosystems more broadly, and biogenic habitats in particular. We applied this metric to 180 seagrass beds (Zostera marina), an important biogenic habitat prioritized for marine protection, across Atlantic Canada. Our impact metric includes five bay-scale and four local-scale terrestrial and marine-based impacts. Results show that seagrass beds and coastal bays in Atlantic Canada exist across a wide gradient of human impacts. Considerable differences in the range and intensity of impacts within and between regions provide insight into where coastal bays and seagrass ecosystems are expected to be most and least affected by individual or cumulative human threats. We discuss implications for management and conservation planning, and the general application of our impact metric to other coastal regions and habitats in Canada and beyond.


Introduction
Over past decades and centuries, the magnitude, spatial extent and variety of human impacts have substantially increased in coastal ecosystems around the world , Halpern et al. 2008. Nearshore biogenic habitats, such as seagrass meadows, kelp forests, rockweed beds, and oyster reefs, are especially vulnerable as they are subject to anthropogenic threats from both the terrestrial and marine realms (Orth et al. 2006, Worm and Lotze 2006, Beck et al. 2011, Krumhansl et al. 2016. Coastal management strategies have recognized nearshore biogenic habitats as areas of high conservation value, and their inclusion in marine protected areas (MPAs) is a conservation priority worldwide (DFO 2007, Cullen-Unsworth andUnsworth 2016). Despite this, it remains unclear how to prioritize areas for protection given the multitude of anthropogenic stressors impacting these ecosystems. Metrics of anthropogenic stressors used to inform management and conservation have previously been applied to ocean ecosystems across broad global and regional scales (Halpern et al. 2008, Ban and Alder 2008, Murray et al. 2015. However, these assessments are not relevant at smaller spatial scales, such as specific coastal bays, nearshore ecosystems or biogenic habitats. Impact metrics useful for management and conservation planning of coastal ecosystems, particularly for specific biogenic habitats, should quantify impacts at both bay-wide and local scales. Human impacts are widely recognized as having scale-dependent effects on ecosystem processes (Powell et al. 2013), and spatial scale may also be relevant when assessing the magnitude of stressors influencing marine ecosystems (Thrush et al. 1999). For example, human activities were found to influence seagrass ecosystems mainly at the local scale, typically within 1-3km of the seagrass bed (Shelton et al. 2017, Cullain et al. 2018a, Iacerella et al. 2018). Yet, given that coastal biogenic habitats are often confined to bays and estuaries they are also influenced by stressors that operate at the bay-scale, such as nutrient and sediment loading and pollution run-off from the surrounding watersheds (McIver et al. in review). Additionally, various biogenic habitat types may be differentially affected by human activities, suggesting the need for habitat-specific assessments. A comprehensive assessment of human impacts in coastal ecosystems should thus consider impacts to specific habitats at both the local and bay-wide scale.
In Canada, eelgrass (Zostera marina) has been designated an Ecologically Significant Species (ESS), and the inclusion of eelgrass beds within MPA networks is a central conservation priority (DFO 2009a). Given their global distribution and critical role in ecosystem functioning (Duarte 2002, Hughes et al. 2009), eelgrass beds are an ideal case study for the development of a human impact metric relevant for coastal ecosystems. Global declines in seagrass cover and associated ecosystem health has been attributed to various human activities, including nutrient pollution (Orth et al. 2006), spread of invasive species (Wong andVercaemer 2012, Williams 2007), coastal land alteration (Grech et al. 2012), construction of overwater structures (Fresh et al. 2006, Thom et al. 2011, and aquaculture (Skinner et al. 2013, Cullain et al. 2018a. While natural events are also sometimes responsible for large-scale seagrass losses (i.e., ice scour, storms; Duarte 2002), human activities are recognized as a significant driver of seagrass ecosystem degradation (Short and Wyllie-Echeverria 1996, Hemminga and Duarte 2000. A metric that quantifies multiple human impacts at spatial scales relevant for seagrass beds will aid conservation planning by identifying priority areas with low human impacts and highlight areas where management measures should be considered. The objectives of this study were to (1) develop a general standardized human impact metric that can quantify the magnitude and range of anthropogenic impacts on various coastal ecosystems and biogenic habitats, and (2) apply this metric to seagrass beds in Atlantic Canada.
First, we selected relevant anthropogenic impacts known to influence biogenic habitats, in particular seagrass beds, on local or bay-wide scales based on the published literature. We then compiled empirical data for these impacts for 180 seagrass beds inhabiting 52 bays across two biogeographic regions in Atlantic Canada and assessed the distribution of these impacts across the different seagrass beds, coastal bays, and biogeographic regions. Finally, we explored the potential general utility of our human impact metric for other biogenic habitats, coastal ecosystems, and geographic locations to inform coastal management and conservation.

General human impact metric development
We began by selecting relevant anthropogenic impacts for inclusion in the human impact metric for coastal ecosystems and biogenic habitats. Here, we focused on impacts known to influence seagrass beds on local or bay-wide scales and relevant indicators or measurable proxies based on the published literature (Table 1). We then assessed data availability to quantify the extent of each impact across our study region. To standardize the human impact metric across all sites, we focused on impacts and indicators for which comparable data were available from the three relevant provincial governments (Nova Scotia, New Brunswick, Prince Edward Island) or the two marine management regions of Fisheries and Oceans Canada (Maritimes region and Gulf region), which are also considered separate biogeographic regions (Scotian Shelf and Gulf of St.
Lawrence, respectively) with unique oceanographic and hydrodynamic processes (Spalding et al. 2007, DFO 2009b. We separated impacts into those acting at a bay-scale or local-scale (0-1km). The bayscale was defined by the geographic boundary of the bay (or estuary) and considered impacts that influence the entire waterbody of the bay, for example through tidal mixing or indirect effects, thereby also influencing the biogenic habitats within. In our case, this included watershed land use, human population density, nutrient pollution, invasion extent, and coastal commercial fishing activity (Fig. 1). In contrast, local-scale impacts were defined as those more directly affecting the biogenic habitat, whereby the considered local range depends on both the impact and the habitat in question. For our seagrass beds, we included aquaculture activity, water quality, overwater structures, and riparian land alteration ( Fig. 1) within 0-1km of the habitat.
This smaller scale has been suggested most relevant for assessing human impacts to seagrass beds (Shelton et al. 2017, Iacarella et al. 2018).
Our derived human impact metric therefore included five bay-scale and four local-scale impacts, with three of these impacts (land use, nutrient pollution, and aquaculture activity) each further subdivided, resulting in a total of 12 impact scores (Fig. 1). In addition, we included coastal land protection as a measure expected to benefit our biogenic habitat. Depending on the coastal ecosystem or biogenic habitat in question, these impacts and indicators and the scale at which they are assessed can be easily adapted. See Table 1 for detailed rationale for the inclusion of these impacts in our human impact metric for seagrass beds in Atlantic Canada. • Prevents erosion, reduces sediment, nutrient and pollution run-off, and maintains coastal buffer zones . May be more beneficial to coastal ecosystems than marine protection (Quiros et al. 2017). Watershed land use Watershed land alteration: 1) agriculture cover 2) urban cover • Predictor of water quality in receiving bays and estuaries (Uriarte et al. 2011, Huang et al. 2015. • Reliance on high water clarity makes biogenic habitats vulnerable to pollution and sediment runoff that follow conversion of natural to human dominated lands (Quiros 2016, Bryce et al. 2010, Leitao et al. 2018). Human population density Watershed human population density • General predictor of intensity of human impacts in coastal ecosystems (Shelton et al. 2017, Iacarella et al. 2018).

Nutrient pollution
Human-derived nitrogen input: 1) N loading rate 2) delta-N • Nutrients from point-and non-point sources and resulting eutrophication is a major driver of seagrass declines (Hauxwell et al. 2003, Orth et al. 2006). • N loading rate represents intensity of nutrient pollution (McIver et al. 2015), delta-N accounts for dilution from freshwater inflow and tidal exchange (Bugden et al. 2014). Invasion extent

Extent of invasive biofouling species
• Non-native species increasing in coastal waters  • Biofouling invaders affect biogenic habitats through shading and breakage (Brush andNixon 2002, Wong and. • Represents general invasion intensity as non-native species can facilitate others (Simberloff andVon Holle 1999, Braga et al. 2018). Coastal commercial fishing activity

Intensity of lobster fishing
• Proxy for general fishing intensity as lobster is largest coastal fishery in Atlantic Canada. • Affect biogenic habitats through direct destruction from gear, disturbance from boat traffic, and release of wastewater or chemical pollutants (Uhrin et al. 2005, Anderson et al. 2011. Aquaculture activity Presence of shellfish aquaculture facilities: 1) near-field (<25m) 2) mid-field (<1km) • Has rapidly expanded in Atlantic Canada and range often overlaps with biogenic habitats (Bastien-Daigle et al. 2007, Comeau 2013). • Shading and physical damage occur at near-field scale (Vance 2014). Dispersed negative effects occur within 1km radius (Skinner et al. 2013). • Far-field (bay scale) effects not considered as these include complex suite of negative and positive effects (Guyondet et al. 2013). • Finfish aquaculture accounted for in N loading estimates.

Water quality
Fecal coliform counts • Associated with reduced water clarity, decreased oxygen, and are indicative of disease causing pathogens for coastal fauna (Parker et al. 2016, Arasamuthu et al. 2017, Lamb et al. 2017). Overwater structures Human-made overwater structures • Structures supported above or floating on water (docks, marinas, causeways, etc.).

Riparian land alteration
Human altered riparian land • Defined as land within 0-50m from shoreline (Quiros et al. 2017).
• Causes sediment instability, fine sediment loading, and negative consequences for biogenic habitats and associated fauna (Quiros 2016, Van Katwijk et al. 2011, Bryce et al. 2010, Leitao et al. 2018). Figure 1. All bay-and local-scale human impacts included in the standard human impact metric for seagrass beds in Atlantic Canada. Note that coastal land protection represents a benefit as opposed to a stressor. Also note that three impacts (land use, nutrient pollution, and aquaculture) were further separated into sub-measures.

Data collection
Our goal was to apply the human impact metric to seagrass beds across Atlantic Canada. To do so, we first compiled the locations of 180 seagrass beds (Fig. 2) from field surveys conducted over the past decade (Schmidt et al. 2012, Skinner et al. 2013, Cullain et al. 2018b, Wong 2018, Weldon et al. 2009   Coastal land protection: We obtained GIS shapefiles of provincial, federal, and private conservation areas from Environment and Climate Change Canada. We clipped shapefiles to the 0-200m coastal land surrounding each bay to determine the percentage of protected land within each bay's coastal land zone. We defined coastal land as 0-200m from the coastline because this distance has previously been identified as important for pollution input to receiving waters (Valiela et al. 1997 (Nagel et al. 2018), and 6 bays in PEI (Palmer 2018). For the remaining 19 bays, we estimated total nitrogen load (kg N yr -1 ) using linear regression models specific to each province (Supplementary Methods, Fig.   S1 and S2). Briefly, the NLM estimates watershed and atmospheric derived total dissolved nitrogen loads from several point (i.e., direct atmospheric deposition, wastewater discharge, seafood processing plants, finfish aquaculture) and non-point sources (i.e., indirect atmospheric deposition, septic systems, and fertilizer addition) with appropriate loss parameters (Valiela et al. 1997, McIver et al. 2015, McIver et al. 2018. Final nitrogen loading rates were calculated by summing across all sources and standardized to the area of each bay (kg N ha bay -1 yr -1 ) to account for dilution effects. We assessed how many bays had nitrogen loading rates above the 50 kg N ha bay -1 yr -1 threshold identified by Latimer and Rego (2010) as being detrimental to seagrass coverage.
The DN combines total nitrogen load (kg N yr -1 , as calculated above) with a bay's tidal flushing time and freshwater recharge volume to estimate the increase in nitrogen concentration above ambient oceanic nitrogen concentration after dilution factors (Bugden et al. 2014). We calculated tidal flushing and freshwater recharge for each bay as described in Nagel et al. (2018).
We assessed how many bays had DN values above the 0.06 threshold identified by Bugden et al. (2014) as being likely to experience anoxic events as a result of excess nitrogen input.  (Garbary et al. 2014), we were unable to obtain quantitative estimates of green crab densities for each of our sites to include in our quantification of invasion extent but note that green crabs are known to occur in all 52 bays.
Our metric of invasion extent is a measure of invasive biofouling species richness over 10 years. We used a 10-year average (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015) of invader presence/absence for each of the nine invasive biofouling species and averaged across species to obtain a measure of invasion extent for each bay. When AIS monitoring stations were not present in a bay we used the closest station within 50km. When no monitoring station was located within 50km for a specific year we excluded that year from the 10-year average. Since the AIS monitoring stations are not located within seagrass beds, our measure of invasion extent is a proxy for how likely fouling is expected within a bay and is not indicative of invaders specifically fouling seagrass at that site. Overwater structures: We manually classified the total area of overwater structures (i.e., wharfs, bridges, causeways, etc.) within a 1km radius of each seagrass bed using Google Earth (Google Inc. 2018). We standardized the total area of overwater structures to the water surface area within a 1km radius.
Riparian land alteration: Using the land use data described above, we estimated human altered riparian land within 0-50m from the coastline due to urban, agriculture, and forestry land use in contrast to unaltered natural wetland, forest or conservation land areas. We define riparian land as 0-50m from the coastline because this distance has previously been used to define riparian land and human land alteration within this range has been shown to be detrimental to seagrass health (Quiros et al. 2017). We measured the area of riparian land alteration within a 1km radius of seagrass, which was standardized to the total land area within the 1km radius of each seagrass site. We used the same land alteration thresholds for watershed land use (>10%) to assess how many seagrass beds had adjacent riparian land considered a moderate to high risk for ecosystem degradation in receiving waters (WWF 2017). This threshold is based on total watershed land alteration, as opposed to strictly riparian land, as no threshold estimates exist for riparian land alteration. Given that coastal land alteration near seagrass beds is expected to be more detrimental than alteration in the entire watershed (Quiros et al. 2017), we expect this to be a conservative threshold.

Human impact standardization
We used the above data to first calculate the intensity of each human impact at each of the 180 seagrass beds (raw data). Each impact was then standardized to range from 0-1, and the multiple impacts at each site were compiled into a petal diagram to illustrate the overall human impact level. Standardization was performed at two different spatial scales: across all 52 bays or 180 sites and across each region.

Results
Comparing the intensity of human impacts across all 52 coastal bays (Fig. 3) and 180 seagrass beds (Fig. 4) revealed a wide gradient of human impacts in Atlantic Canada. We observed differences in the distribution of each impact between the Scotian Shelf and Gulf of St.
Lawrence, and within the Gulf region between the mainland coast (NB and NS) and PEI ( Fig. 3 and 4 insets). Below we describe and assess these differences among regions (Atlantic NS, Gulf NB+NS, and PEI) based on 95% confidence intervals ( Fig. 3 and 4 insets).

Bay-scale impacts
Land protection and use: The percentage of coastal land protection was highly variable, ranging from 0% to 100% protection (median = 3.21%; Fig. 3a). Coastal land protection was highest in PEI (10.04% ± 5.42% 95%CI), much greater than in Gulf NB+NS (0.68 ± 4.32%) and Atlantic NS (1.65 ± 5.53%), although not statistically significant. PEI watersheds also had the greatest land alteration for urban (Fig. 3b) and agricultural uses (Fig. 3c), although urban land use was similar between PEI and Atlantic NS (8.3 ± 1.49% and 7.1 ± 2.23%, respectively), but significantly greater than Gulf NB+NS (3.5 ± 1.35%). Agricultural land use in PEI (39.8 ± 7.11%) was significantly greater than in the Gulf NB+NS (8.4 ± 2.87%) and Atlantic NS (3.3 ± 0.32%). According to the threshold of >10% watershed land alteration (WWF 2017), 65% of the 52 bays were at risk for ecosystem degradation in receiving waters, most of them in Gulf NB+NS and PEI, and only 2 in Atlantic NS.
Human population density: Similar to the regional variation in land protection and use, the median human population density was higher in PEI (24.1 persons km -2 ) than in Atlantic NS (13.2 km -2 ) and Gulf NB+NS (16.9 km -2 ), although not statistically different among regions (Fig.   3d).

Local-scale impacts
Aquaculture activity: Near-field shellfish aquaculture (<25m from seagrass bed) was present at 20 seagrass beds and mid-field (<1km) at 87 beds across all 180 sites (Fig. 4a). Among regions, Atlantic NS had no shellfish aquaculture within 1km of any seagrass beds, and mid-field aquaculture was more prevalent in PEI (71% of beds) and Gulf NB+NS (41%) than near-field (8% and 15% respectively).
However, 35 (19%) of seagrass beds had water quality above this threshold, with 1 in Atlantic NS, 9 in PEI and 25 in Gulf NB+NS.
Riparian land alteration: Similar to human land alteration in the entire watershed, the median percentage of coastal land altered in close proximity to seagrass beds was significantly higher in PEI (5.4 ± 0.5%, Fig. 4d) than in Atlantic NS (2.3 ± 1.9%) and Gulf NB+NS (3.3 ± 0.6%).
However, the range of coastal land alteration was greater in Atlantic NS (0-25%) compared to PEI (1-18%) and Gulf NB+NS (0-8%). Of the 180 seagrass beds, 3% were affected by >10% riparian land alteration. Most of these were in Atlantic NS, accounting for nearly 25% of all Atlantic NS seagrass beds.  Boxplots show the median, the first and third quartiles, and outliers. Notches in boxplots show 95% confidence intervals for medians. Note that in (a) inset the percentage can be greater than 100% as seagrass beds with both near-field and mid-field aquaculture facilities are included. Note different scales on axes.

Human impacts at different spatial scales
Our human impact metric can be compared across all coastal bays and seagrass beds in Atlantic Canada ( Fig. 3 and 4) as well as within biogeographic regions (e.g. Scotian Shelf and Gulf of St. Lawrence, Fig. 5 and 6). Standardizing the impacts within each region allows for better comparison of the importance of each impact relative to other sites subjected to similar biogeographic conditions. For example, compared to all other seagrass beds, those on the Atlantic coast of NS had relatively low human impact ( Fig. 3 and 4 insets); yet when standardized within the Scotian Shelf region, relative impact patterns became more apparent (Fig. 5). This revealed that several seagrass beds were minimally impacted by human activities relative to other Atlantic NS sites, including Port Joli, Cable Island and Taylor's Head, which have higher coastal protection and lower overall bay and local-scale impacts (Fig. 5). In comparison, other seagrass beds were more impacted by human activities, with Second Peninsula highly impacted by urban and agriculture land use, poor water quality, and commercial fishing, St. Margaret's Bay by riparian land alteration, poor water quality, and commercial fishing, Sambro by high population density and urban land use, and Musquodoboit by nutrient pollution (Fig. 5).
Our human impact metric can also be used to assess the degree of impacts among individual seagrass beds within a bay. For example, Pugwash Bay (Fig. 7a) had relatively low human impacts at the bay-scale (i.e., almost all impacts within the 20-40 th percentile), but high heterogeneity in local-scale impacts among the individual seagrass beds. For example, Bed "C" (Fig. 7a) was only minimally impacted by poor water quality and overwater structures, while beds "A" and "B" were more highly impacted by overwater structures and reduced buffer zones.
In comparison, Bedeque Bay in PEI (Fig. 7b) was highly impacted by human activities at the bay-scale (i.e., most bay-scale impacts in the 80-98 th percentile) with high agricultural land use and nitrogen loading. Again, high heterogeneity in human impacts was evident among the individual seagrass beds (Fig. 7b). For example, Bed "A" had extensive overwater structures in its vicinity and poor water quality, while beds 'B' and 'C' were influenced by a relatively low degree of human impacts at the local-scale. Figure 5. Standardized human impact metric for seagrass beds on the Atlantic coast of Nova Scotia. Each coloured petal represents one human impact measure with the outer ring of each circle representing the maximum possible impact score (i.e. petals that fill up more space represent a higher score of that impact). Impact scores are shown for 7 selected sites relative to all 17 sites (green points) in Atlantic NS. The human impact metric was calculated for all Atlantic NS region sites but is only shown for 7 selected sites. Refer to Fig. 1 for legend. Figure 6. Standardized human impact metric for seagrass beds in (a) Prince Edward Island and (b) the Gulf coast of New Brunswick and Nova Scotia. Each coloured petal represents one human impact measure with the outer ring of each circle representing the maximum possible impact score (i.e. petals that fill up more space represent a higher score of that impact). Impact scores are shown for 9 sites in PEI and 7 in NB and NS but were calculated relative to all other 163 sites (green points) in the Gulf region. The human impact metric was calculated for all Gulf region sites but is only shown for 15 selected sites. Refer to Fig. 1 for legend. Bedeque Bay, PEI in the Gulf region. The large petal diagrams in the corner of each plot represent the bay-scale impacts and the small petal diagrams represent the local-scale impacts for each individual bed, with the respective other impacts removed (grey petals). Each coloured petal represents one human impact measure with the outer ring of each circle representing the maximum possible impact score (i.e. petals that fill up more space represent a higher score of that impact). Impact scores are relative to all other sites in the Gulf region of NB, NS, and PEI.
Green points indicate the locations of seagrass beds. Refer to Fig. 1 for legend.

Discussion
Quantifying the magnitude and range of human impacts in a standardized, comprehensive and comparable way is urgently needed for large-scale scientific assessments and for management and conservation planning in coastal ecosystems. Previous metrics of human impacts for marine ecosystems have typically been developed on global, national, or broad regional scales, with little effort towards metrics appropriate for coastal ecosystems or specific biogenic habitats. Our human impact metric incorporates a suite of anthropogenic stressors from both terrestrial and marine sources relevant for coastal bays and estuaries as well as biogenic habitats and can be adapted to and standardized across multiple spatial scales.
Despite including multiple human impacts into our metric, we did not to combine them into one cumulative impact score (sensu Halpern et al. 2008), as this requires scientificallyinformed vulnerability weightings that are not readily available. Although research on human impacts to seagrass ecosystems in Atlantic Canada has increased over the past decade (Coll et al. 2011, Schmidt et al. 2012, Hitchcock et al. 2017, Cullain et al. 2018a in review), we still know too little about the responses of seagrass ecosystems to individual and cumulative human impacts to allow effective ranking or evidence-based weighting of their importance.
Despite not explicitly accounting for the magnitude and relative importance of all impacts combined, our approach still provides a useful overall qualitative summary of the total human impact as visualized by the petal diagrams. Furthermore, providing individual impact scores instead of collapsing these into a single value allows influential impacts to be identified for specific bays, the relative impact to be compared, and also allows managers to identify and track impacts of interest. In the following, we discuss our application of the human impact metric to seagrass beds in Atlantic Canada, and then its potential for more general application to other habitats and regions and uses for management and conservation planning.

Application to seagrass beds in Atlantic Canada
Applying our human impact metric to 180 seagrass beds in 52 coastal bays in Atlantic Canada revealed several insights relevant for scientific assessment, management and conservation. We found considerable regional variation in impact strength that provide insight into where coastal bays and seagrass ecosystems are expected to be most and least affected by individual or cumulative human impacts. Also, our results highlight the importance of considering human impacts at multiple spatial scales when assessing threats to coastal ecosystems and biogenic habitats.
One of the most damaging human impacts to the extent and condition of seagrass beds is eutrophication from human-derived nitrogen loading (Orth et al. 2006). We found that 64% of all bays in our study are at risk of seagrass decline based on nitrogen loading rates (Latimer and Rego 2010). However, only 13% of bays, most of which are in PEI, are at risk for anoxic events based on DN values (Bugden et al. 2014). While both nitrogen loading and DN can indicate eutrophication risk, DN can indicate the potential for more severe anoxic events to occur by accounting for dilution from freshwater input and tidal flushing (Bugden et al. 2014).
Thus, while human-derived nitrogen loading can be high in many bays and lead to primary (e.g. enhanced phytoplankton, epiphytic or benthic algal growth) or secondary (e.g. reduced shoot density and biomass) eutrophication effects in seagrass beds (Schmidt et al. 2012, Cullain et al. 2018a, more severe loss of seagrass cover from anoxic events may only be of concern in PEI ( Bugden et al. 2014).
Despite relatively low land use change within Atlantic NS watersheds, the wide range of riparian land alteration in close proximity to seagrass beds suggests that local-scale impacts may be important for seagrass health in Atlantic NS. Additionally, 42% of Atlantic NS bays have no coastal land protection. This is similar to bays in Gulf NB+NS but is much higher than PEI where only 11% of bays have no coastal land protection. Seagrass beds bordered by land managed for conservation purposes usually have higher temporal stability than beds associated with unprotected land (Breininger et al. 2016). Not surprisingly, our results show that coastal land protection at the bay-scale correlates well with the extent of riparian land alteration adjacent to seagrass beds i.e. local-scale (p < 0.001, R 2 = 0.42, Fig. S3). This suggests that implementation of coastal conservation areas at the bay-scale may be a useful strategy to reduce the number of seagrass beds at risk of reduced buffer zones from riparian land alteration (Quiros et al. 2017).
Compared to the Gulf region, Atlantic NS bays have a higher extent of invasive biofouling species. This is not surprising given that the closer proximity of Atlantic NS to the open Atlantic Ocean and United States allows greater susceptibility to range expansions of invaders (Sephton et al. 2017). The difference in invasion extent between the two regions is driven by recent invasion of several tunicate species in NS that are absent from the Gulf (Moore et al. 2014. However, species distribution models predict northward range expansion of most invasive tunicates, with many seagrass inhabited bays in the Gulf region becoming hotspots for invasive biofouling species over the next 50 years (Lowen and DiBacco 2017). Fouling by invasive tunicates reduces seagrass growth and survival due to shading and premature breakage (Brush andNixon 2002, Wong and. Since our human impact metric identified many highly impacted seagrass beds in the Gulf region, investigating how seagrass beds respond to the cumulative effect of invasive biofouling species and other human impacts should be a research priority since this may become a reality in the near future. Overwater structures are recognized as having one of the largest negative impacts on seagrasses, yet no threshold level has been determined (Fresh et al. 2001, Rehr et al. 2014 (Fig. 6b). This illustrates the importance of quantifying numerous threats at multiple scales to fully understand impacts to biogenic habitats. Future research that identifies thresholds for the effects of overwater structures on coastal biogenic habitats would improve assessments of ecosystem health.

Assessment of human impact at multiple spatial scales
Our human impact metric is unique as it assesses impacts at multiple spatial scales. This allows evaluation of within-bay variation and can reveal discrepancies about the relative magnitude of impacts at different spatial scales. A metric that accounts for different spatial scales of impact is critical for biogenic habitats in coastal ecosystems given the uncertainty in what human impacts, and at what spatial scales these impacts, are most detrimental to these habitats. Seagrass responses to anthropogenic stressors typically exhibit a high level of within-bay variability (Hitchcock et al. 2017), and our results show that there is also often high within-bay variation in stressor intensity (Fig. 7). By considering impacts at both local and bay-wide scales our metric can highlight locations where there is spatial mismatch between the intensity of bay-wide and local-scale impacts at individual seagrass beds. For example, Bedeque Bay, PEI is a highly impacted bay based on bay-scale impacts, however, there is high variability among local-scale impacts within the bay (Fig. 7b). Observation of only local-scale impacts would suggest certain beds are relatively non-impacted when this might not be the case. This emphasizes the necessity of quantifying impacts at multiple scales to fully understand the degree of impact to coastal biogenic habitats.

General application to other habitats and ecosystems
Along with seagrass, several other species create complex biogenic habitats in coastal waters in Canada and elsewhere, including rockweed beds, kelp forests, oyster reefs, and others (Beazley et al. 2017 Canada. This database will be valuable for the application of our human impact metric to other biogenic coastal ecosystems in Atlantic Canada, as many of the seagrass inhabited bays are also home to kelp and rockweed habitats. Therefore, the same bay-scale impacts already calculated for the current database apply and only local-scale impacts need to be adjusted.
Our impact metric has been specifically tailored for impacts relevant to seagrass beds in Atlantic Canada. Applications to other ecosystems and regions should begin with an assessment of the impacts present and their potential influence (e.g. Table 1). For example, boat moorings are not an important threat to Atlantic Canada seagrass beds yet should be included in impact metrics for seagrass in parts of the coastal United States, United Kingdom, and Australia (Hallac et al. 2012, Demers et al. 2013, Unsworth et al. 2017. Also, some impacts may be more relevant for one habitat type than another within the same region, such as commercial plant harvesting in Atlantic Canada rockweed beds (DFO 2013). Furthermore, our human impact metric does not consider climate change related stressors (i.e. increased sea surface temperature and sea level rise), which impact seagrass through loss of suitable habitat and direct effects on survival (Waycott et al. 2007, Olsen et al. 2012, Valle et al. 2014. Large-scale climate change impacts were not relevant for the application of our metric to seagrass beds in Atlantic Canada but could be included for more spatially broad applications.

Application to management and conservation
Our metric is valuable for conservation planning by highlighting areas of low impact and high naturalness, which is a conservation priority for marine protected area placement in Canada and around the world (DFO 2004, Yamakita et al. 2015. For example, in the Scotian Shelf our impact metric highlights three seagrass beds in the Eastern Shore Archipelago with a high degree of coastal land protection and among the lowest impact scores relative to other beds in this region (Fig. 5, Table S1 and S2). This provides evidence for the Eastern Shore Archipelago as an ideal candidate for marine protection given the low human impact and high naturalness at both the bay scale and in close proximity to a valuable biogenic habitat (DFO in press). By assessing impacts in both the terrestrial and marine realms our impact metric will also be valuable for landsea conservation planning by identifying coastal ecosystems adjacent to protected land (Alvarez-Romero et al. 2015). For example, in the Scotian Shelf, Port Joli has high coastal land protection and low riparian land alteration relative to other bays in the region (Fig. 5, Table S1 and S2). In the Gulf, Kouchibouguac and St. Louis de Kent (Fig. 6, Table S3) have 100% coastal land protection due to the presence of a National Park.
The comprehensive, standardized, and multi-scale nature of our metric can also assist in developing management strategies for individual or multiple human impacts. For example, in the Scotian Shelf, Second Peninsula ranks among the most highly impacted sites (Fig. 5, Table S1 and S2) with nitrogen loading close to the threshold detrimental to seagrass cover (Latimer and Rego 2010), high levels of land use that exceed the threshold for risk of ecosystem degradation in receiving waters (WWF 2017), poor water quality, and no coastal land conservation. Thus, our impact metric identifies risks from multiple human impacts that could be addressed by specific management strategies.
The impacts can also be linked to key mechanisms of growth and survival. For example, light availability is a key determinant of seagrass condition (Hauxwell et al. 2003), thus, knowledge of which seagrass beds are subject to multiple light limiting impacts will be beneficial for management. In the Gulf, our metric identified several seagrass beds at risk from multiple forms of light limitation. For example, nitrogen loading for Bedeque PEI is among the highest in the Gulf region (Table S3), and one seagrass bed in Bedeque Bay also has among the highest impact scores for overwater structures and riparian land alteration (Table S4). The combination of bay-and local-scale impacts specifically influencing light availability suggests that this bed may be an ideal candidate for management actions targeted at maintaining water clarity, such as reductions in land-derived nutrient inputs (Leschen et al. 2010, Greening et al. 2014, Lefchek et al. 2018, removal of overwater structures (Thom et al. 2005, Rehr et al. 2014, and conservation of coastal land (Stoms et al. 2005, Alvarez-Romero et al. 2015.

Caveats and the way forward
One current limitation of our metric is that temporal variation could only be accounted for with some impacts (nutrient pollution, invasion extent, and water quality), which represent 10-year averages. The remaining impacts reflect intensity at the present time. This may be problematic if impacts been historically higher but are currently reduced through management strategies.
Inclusion of temporal impact data would be useful to evaluate the response time of biogenic habitats to impacts and should be included where possible.
Our application of the human impact metric is also limited by the disparity in spatial distribution of seagrass sites among regions (n = 163 and 17 for Gulf and Scotian Shelf regions, respectively). Seagrass beds included in our study were those sampled in summer field surveys over the past decade. Inclusion of sites from a large-scale monitoring program in the Gulf region (e.g. Community Aquatic Monitoring Program; Weldon et al. 2009) resulted in a much larger sample size relative to Atlantic NS, where data were compiled from small, individual research projects. The larger dataset in the Gulf region highlighted the importance of within-bay variability, which could not be assessed for the Scotian Shelf region. Further application of our human impact metric as more studies are conducted would be beneficial to fully map the distribution of anthropogenic stressors in this region.
Finally, further work is necessary to allow the individual and combined impacts to be ranked according to importance for seagrass health. An understanding of the relative importance of each impact will allow the calculation of cumulative scores based on vulnerability weightings, producing one overall metric that is easily interpretable and useful for management and conservation planning.

Conclusions
Given that coastal ecosystems are highly susceptible to human activities, the framework we have developed for quantifying human impacts is valuable for conservation and management planning. Our impact metric allows the finer-scale dynamics of anthropogenic stressors in coastal ecosystems to be explored by uniquely focusing on stressors relevant for coastal biogenic habitats across two spatial scales. Our metric can be applied to coastal ecosystems worldwide and can be used to prioritize areas for protection and management by identifying areas of low and high impact, highlighting prominent impacts within and between regions, and describing within-bay variation. The application of our human impact metric to seagrass beds across Atlantic Canada reveals a gradient of human impacts and identifies several bays and seagrass beds that may be good candidates for protection based on high naturalness, and others at risk for future degradation if management strategies are not implemented. Our results can also be used to apply the impact metric to other biogenic habitats in Atlantic Canada and as a baseline of human activities for future comparisons.