Online Proofing System Instructions The Wiley Online Proofing System allows authors and proof reviewers to review PDF proofs, mark corrections, respond to queries, upload replacement figures, and submit these changes directly from the PDF proof from the locally saved file or while viewing it in your web browser. 1. For the best experience reviewing your proof in the Wiley Online Proofing System please ensure you are connected to the internet. This will allow the PDF proof to connect to the central Wiley Online Proofing System server. If you are connected to the Wiley Online Proofing System server you should see the icon with a green check mark above in the yellow banner. Connected Disconnected 2. Please review the article proof on the following pages and mark any corrections, changes, and query responses using the Annotation Tools outlined on the next 2 pages. 3. To save your proof corrections, click the “Publish Comments” button appearing above in the yellow banner. Publishing your comments saves your corrections to the Wiley Online Proofing System server. Corrections don’t have to be marked in one sitting, you can publish corrections and log back in at a later time to add more before you click the “Complete Proof Review” button below. 4. If you need to supply additional or replacement files bigger than 5 Megabytes (MB) do not attach them directly to the PDF Proof, please click the “Upload Files” button to upload files: Click Here 5. When your proof review is complete and you are ready to submit corrections to the publisher, please click the “Complete Proof Review” button below: Click Here IMPORTANT: Do not click the “Complete Proof Review” button without replying to all author queries found on the last page of your proof. Incomplete proof reviews will cause a delay in publication. IMPORTANT: Once you click “Complete Proof Review” you will not be able to publish further corrections. USING e-ANNOTATION TOOLS FOR ELECTRONIC PROOF CORRECTION TION Once you have Acrobat Reader open on your computer, click on the Comment tab at the right of the toolbar: This will open up a panel down the right side of the document. The majority of tools you will use for annotating your proof will be in the Annotations section, pictured opposite. We’ve picked out some of these tools below: 1. Replace (Ins) Tool – for replacing text. 2. Strikethrough (Del) Tool – for deleting text. Strikes a line through text and opens up a text box where replacement text can be entered. How to use it Strikes a red line through text that is to be deleted. How to use it • Highlight a word or sentence. • Highlight a word or sentence. • Click on the Replace (Ins) icon in the Annotations section. • Click on the Strikethrough (Del) icon in the Annotations section. • Type the replacement text into the blue box that appears. 3. Add note to text Tool – for highlighting a section to be changed to bold or italic. 4. Add sticky note Tool – for making notes at specific points in the text. Highlights text in yellow and opens up a text box where comments can be entered. How to use it Marks a point in the proof where a comment needs to be highlighted. How to use it • Highlight the relevant section of text. • • Click on the Add note to text icon in the Annotations section. Click on the Add sticky note icon in the Annotations section. • • Type instruction on what should be changed regarding the text into the yellow box that appears. Click at the point in the proof where the comment should be inserted. • Type the comment into the yellow box that appears. USING e-ANNOTATION TOOLS FOR ELECTRONIC PROOF CORRECTION TION 5. Attach File Tool – for inserting large amounts of text or replacement figures. Inserts an icon linking to the attached file in the appropriate place in the text. 6. Drawing Markups Tools – for drawing shapes, lines and freeform annotations on proofs and commenting on these marks. Allows shapes, lines and freeform annotations to be drawn on proofs and for comment to be made on these marks. How to use it • Click on the Attach File icon in the Annotations section. • Click on the proof to where you’d like the attached file to be linked. • Select the file to be attached from your computer or network. • Select the colour and type of icon that will appear in the proof. Click OK. How to use it • Click on one of the shapes in the Drawing Markups section. • Click on the proof at the relevant point and draw the selected shape with the cursor. • To add a comment to the drawn shape, move the cursor over the shape until an arrowhead appears. • Double click on the shape and type any text in the red box that appears. Journal of Ecology Forty years of seagrass population stability and resilience in an urbanizing estuary 1 Andrew O. Shelton1,2*, Tessa B. Francis3, Blake E. Feist1, Gregory D. Williams4, Adam Lindquist5 and Philip S. Levin1,6,7 CE: Gayathri K Conservation Biology Division, Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, 2725 Montlake Blvd. E, Seattle, WA 98112, USA; 2Earth Resources Technology, Inc., 14401 Sweitzer Lane Suite 300, Laurel, MD 20707, USA; 3Tacoma, Puget Sound Institute, University of Washington, 326 East D Street, Tacoma, WA 98421, USA; 4Pacific States Marine Fisheries Commission (Under contract to the Northwest Fisheries Science Center), NOAA Fisheries, 2725 Montlake Blvd E, Seattle, WA 98112, USA; 5Washington Department of Fish and Wildlife, 1111 Washington St SE, Olympia, WA 98504-3150, USA; 6 The Nature Conservancy, Washington Field Office, 74 Wall St., Seattle, WA 98121, USA; and 7School of Environmental and Forest Sciences, University of Washington, Seattle, WA 98195, USA PE: Sakthivel R. 1 2 Key-words: coastal development, eelgrass, Puget Sound, Resilience, shoreline armouring, spatiotemporal models, urbanization, Zostera Introduction Nearly 40% of the world’s population lives within 100 km of a coast (Sale et al. 2014), most of the world’s major cities are located in coastal zones (Timmerman & White 1997; *Correspondence author. E-mail: ole.shelton@noaa.gov Small & Nicholls 2003), and future population expansion will increase disproportionately in the world’s coastal zones (Seto, G€ uneralp & Hutyra 2012; Sale et al. 2014; Neumann et al. 2015). Human development and activities in coastal regions place enormous stress on natural systems and have negatively affected the structure and function of many nearshore and coastal ecosystems (Imhoff et al. 2004; Lotze et al. 2006; Published 2016. This article is a U.S. Government work and is in the public domain in the USA. No. of pages: 13 12682 Manuscript No. J E C 1. Coasts and estuaries contain among the most productive and ecologically important habitats in the world and face intense pressure from current and projected human activities, including coastal development. Seagrasses are a key habitat feature in many estuaries perceived to be in widespread decline owing to human actions. 2. We use spatio-temporal models and a 41-year time series from 100s of km of shoreline which includes over 160 000 observations from Puget Sound, Washington, USA, to examine multiscale trends and drivers of eelgrass (Zostera spp.) change in an urbanizing estuary. 3. At whole estuary scale (100s of km), we find a stable and resilient eelgrass population despite a more than doubling of human population density and multiple major climactic stressors (e.g. ENSO events) over the period. However, the aggregate trend is not reflected at the site scale (10s of km), where some sites persistently increase while others decline. 4. Site trends were spatially asynchronous; adjacent sites sometimes exhibited opposite trends over the same period. Substantial change in eelgrass occurred at the subsite (0 1 km) scale, including both complete local loss and dramatic increase of eelgrass. 5. Metrics of local human development including shoreline armouring, upland development (imperviousness) and human density provide no explanatory power for eelgrass population change at any spatial scale. 6. Our results suggest that the appropriate scale for understanding eelgrass change is smaller than typically assumed (approximately 1- to 3-km scale) and contrasts strongly with previous work. 7. Synthesis. Despite ongoing conservation concern over seagrasses world-wide, eelgrass in Puget Sound has been highly resilient to both anthropogenic and environmental change over four decades. Our work provides general methods that can be applied to understand spatial and temporal scales of change and can be used to assess hypothesized drivers of change. Dispatch: 21.10.16 Summary Journal Code 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 doi: 10.1111/1365-2745.12682 2 A. O. Shelton et al. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Greene et al. 2015). Anticipating and adapting to such changes requires an understanding of where and when such human activities occur, the mechanisms driving change in coastal ecosystems, and the spatial and temporal scale at which change is occurring (Rahbek 2005). Of particular importance in coastal marine ecosystems are foundational, habitat-forming species (Dayton 1972) like mangroves, saltmarshes and seagrasses that support productive marine communities. Seagrasses, in particular, play a vital role in temperate and tropical coastal marine systems, providing habitat for species of conservation concern (Heck et al. 2008), supporting fisheries (McArthur & Boland 2006; Orth et al. 2006), providing coastal protection and erosion control (Koch et al. 2009; Arkema et al. 2013), mediating ocean carbon chemistry (Fourqurean et al. 2012), improving water clarity (Gacia, Granata & Duarte 1999; Lee & Dunton 1999) and providing subsidies to other marine ecosystems (McGlathery et al. 2012). As the ecosystem functions provided by seagrasses have become increasingly appreciated over the past decade (Barbier et al. 2011), concern over both the threats to and population trajectories of seagrasses has grown (Orth et al. 2006; Waycott et al. 2009; Short et al. 2011). Threats to seagrasses are well documented (Orth et al. 2006; Grech et al. 2012). On large spatial scales threats include increased ocean temperature and changing ocean chemistry (Koch et al. 2013; Thomson et al. 2015) while localized threats include light limitation from overwater structures, sediment loading, and eutrophication associated with development and agriculture (Gallegos & Bergstrom 2005; Li et al. 2007; Rehr et al. 2014), direct disturbance from dredging, aquaculture, and fishing (Simenstad & Fresh 1995; Short & Wyllie-Echeverria 1996; Cardoso et al. 2010), and threats from shoreline modifications (Patrick et al. 2014; Patrick, Weller & Ryder 2016). In addition to these bottom-up drivers, top-down drivers have been identified recently as potentially playing an underappreciated role in seagrass population dynamics (Heck et al. 2000; Williams & Heck 2001). Topdown drivers include both direct herbivory on seagrasses (Hughes & Stachowicz 2004), and direct and indirect effects of meso-grazers on seagrass communities (Heck et al. 2000; Duffy et al. 2015). Nearly all threats to seagrasses are positively correlated with human population density and activities. Waycott et al. (2009) synthesized global seagrass time series to show a persistent global loss of seagrasses since at least 1940 (median annual loss rate between 0 29% and 0 84% annually; their Table S2, Supporting Information) and asserted that negative impacts on seagrasses due to humans were accelerating. While such global averages provide large-scale descriptors of seagrass trends, there is substantial variation in seagrass trends at local and regional scales (Waycott et al. 2009). Furthermore, both the drivers of seagrass change and management actions to increase, maintain or recover seagrass populations often occur at regional or local scales. Thus, detailed studies on local seagrass populations are required to link trends to environmental and anthropogenic drivers. To date, a major hindrance to understanding seagrass systems has been a lack of long-term monitoring data (Duarte 2002). As seagrass beds can persist over long periods of time and may respond to an array of factors that act on a range of spatial and temporal scales (from the scale of metres to 100s of km, and days to decades), understanding seagrass drivers and trends requires data collected across appropriate spatial and temporal scales. Here, we focus on understanding trends and drivers of eelgrass (Zostera spp.) populations in an urbanizing estuary in the north-east Pacific. We develop and apply novel statistical methods to understand large-scale spatial and temporal changes in eelgrass populations. We incorporate over 160 000 observations that span more than 500 km of shoreline and 41 years (1972–2012) to provide one of the only large-scale and longterm descriptions of seagrass habitats in the north Pacific. We document eelgrass change at regional and local scales and consider the role played by both regional and local drivers in generating trends. We extend available information about Puget Sound eelgrass by nearly 30 years (see WDNR 2015) and summarize eelgrass trends at three scales: regional (100-km scale), site (1–10 km) and local (0 1 km). We utilize a diverse suite of spatially explicit environmental and human covariates to ask how changes in shoreline and water attributes at regional, site and local levels contribute to changes in eelgrass. Additionally, our time series spans a wide range of well-documented environmental changes including several major ENSO events (1982– 1983, 1988–1989, 1997–1998 and 2010–2011; Clarke 2014). The spatial and temporal signatures of eelgrass change allow us to begin to disentangle the drivers of eelgrass populations in Puget Sound and provide actionable information for practitioners working to maintain and recover seagrasses both in Puget Sound and temperate systems world-wide. Materials and methods STUDY AREA AND SPECIES All sampling was conducted in Puget Sound, a large fjord-like estuary in the southern portion of the inland Salish Sea (Fig. 1). Puget Sound covers an area of approximately 2330 km2 at low tide (Burns 1985) with an average depth of 150 m and a maximum depth of nearly 300 m. Circulation is driven by tidal currents, seawater inflow from marine waters, freshwater outflow from rivers and winds (Ebbesmeyer & Barnes 1980; Thomson 1994). Human populations are concentrated on the eastern shore of the central basin around the commercial ports of Seattle, Tacoma and Everett and 60% of the state’s population lives within 20 km of the Puget Sound coast (Appendix S2). We focus on the native eelgrass Zostera marina. Eelgrasses are the most abundant temperate species of seagrass, occurring along the coasts of the north Pacific Ocean from Japan to Mexico and widely in the north Atlantic Ocean. Eelgrass beds generate considerable ecological and economic value (Irlandi, Ambrose & Orlando 1995; Costanza et al. 1997; Plummer et al. 2013), providing spawning substrate and nursery habitat for numerous fish and invertebrates (Heck & Thoman 1984; Heck, Hays & Orth 2003), and fuel primary production and detrital pathways in nearshore food webs (McConnaughey & McRoy 1979; Valentine & Heck 1999). Furthermore, eelgrass is frequently used as an indicator of the function of nearshore marine Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Colour online, B&W in print Resilient seagrass in an urban estuary 3 Fig. 1. Puget Sound study area and focal areas included for examination of eelgrass trends. Boxes with two letter identifiers are used throughout the paper. Grey shading in bottom left panel delineates watershed areas that drain into Puget Sound. Yellow, orange and brown colours identify impervious surfaces with darker colours indicating higher proportion imperviousness. [Colour figure can be viewed at wileyonlinelibrary.com] ecosystems (Partnership 2011). As with seagrasses world-wide, there are few long-term descriptions of eelgrass populations, or seagrasses in general, in the north Pacific (Duarte 2002). In 2015, the native eelgrass Z. marina covered approximately 20 000 ha in Puget Sound (WDNR 2015), where it grows in muddy to sandy habitats in intertidal and shallow subtidal zones. Shoot density may reach 800 shoots per m 2 and shoot length may exceed 2 m (WDNR 2015). SURVEYS AND DATA We use information gathered during herring spawning surveys conducted by the Washington Department of Fish and Wildlife (WDFW) to document nearshore habitat change within Puget Sound between 1972 and 2012. During herring spawning season in Puget Sound [January–March for all sites but Cherry Point (CP) which occurs in April–June; Fig. 1], WDFW personnel survey nearshore spawning habitats by dropping a specially designed metal rake (a modified grappling hook) to the shallow nearshore benthos from a small boat, ensnaring benthic vegetation and retrieving the rake. For each rake sample, the geographic coordinates, presence of 29 benthic vegetation categories (most identified to genera), abundance category of herring eggs on vegetation and water depth are recorded. During a single survey, samples are collected approximately every 200 m along the coast, though the frequency varies substantially by site and across years. Sampling occurs at all tidal stages and at depths from mean sea level to c. 10 m below mean sea level. Approximately weekly surveys are conducted at each survey location, resulting in hundreds to thousands of georeferenced observations at each spawning location in each year (Trumble et al. 1977; Stick 1994; Stick, Lindquist & Lowry 2014). Across all sites and years this data set includes more than 250 000 unique, georeferenced observations spanning 41 years. For this article, we focused exclusively on the most commonly observed vegetation category, Zostera spp. (eelgrass; almost exclusively Z. marina, though very rarely Zostera japonica may be observed). As we were primarily interested in detecting temporal trends and spatial patterns in eelgrass occurrence, we restricted our analyses to sites that had extensive temporal and spatial sampling coverage; we included sites first sampled in 1980 or earlier and had at least 4000 observations across the entire time series. This resulted in 160 932 observations from 11 survey locations for our analyses (Table S1.1). We used the data from surveys to generate trends for 14 sites of comparable size (representing between 12 5 and 27 0 km of shoreline; Fig. 1, Table 1); we divide the shoreline of the Cherry Pt. survey into four sites to make them comparable in size to other sites (SB, BB, CP and LM are derived in part from the Cherry Pt. data). Similarly data from surveys at Port Orchard-Port Madison were divided into Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 4 A. O. Shelton et al. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Table 1. Summary of trend areas Site Herring survey location SB BB Semiahmoo Bay Semiahmoo Bay and Cherry Point Cherry Point Cherry Point Fidalgo Bay Discovery Bay Holmes Harbor Port Susan Port Gamble Port Orchard-Port Madison Port Orchard-Port Madison South Hood Canal Quilcene Bay Quartermaster Harbor Sum Among-site mean CP LM FB DB HH PS PG PO BA HC QB QM Length of shoreline (km) Area eelgrass (ha; time series mean) Area eelgrass (ha; 2011–2012) Total potential area (ha) Mean proportion eelgrass cover (2011–2012) 15 1 12 5 841 550 876 651 977 890 0 896 0 731 14 7 27 3 14 5 27 0 20 1 21 6 20 6 22 2 22 1 11 2 18 2 26 9 274 0 108 1023 731 199 279 172 235 167 84 365 242 179 5,176 78 1154 672 169 290 229 202 151 58 369 238 183 5,321 474 1463 896 322 294 386 359 412 313 456 336 476 8,055 0 164 0 789 0 750 0 525 0 986 0 593 0 563 0 366 0 185 0 809 0 708 0 384 comparably sized sites: PO and BA. Site SB is surveyed by two different surveys (one in the late winter, one in the late spring as part of the Cherry Pt. survey), but is only counted as one site, resulting in a total of 14 trend sites. As most survey locations had missing data – all but one had at least 1 year in which no rake samples were collected – and to ease computation, we binned observations data into 2-year groups (e.g. we combined data from 1974 and 1975). Combining data into 2-year groups had the effect of halving the number of spatial fields estimated by the model, greatly sped model estimation and produced improved metrics of model fit. We explored alternate binning strategies using 4-, 5- or 10-year bins, but preliminary comparisons of models showed that 2-year bins provided the best combination of statistical tractability and model fit. To examine spatial and temporal changes in eelgrass, we divided the shore into 100-m-long segments and assigned covariate values for each 100-m-long segment. Each observation was associated with the nearest shoreline segment based on Euclidean distance. This approach acknowledges positional uncertainty about the location of individual observations – sampled locations were not identified using GPS until the 2000s and earlier locations were digitized from hand drawn charts. Additionally, this approach reflects the uncertainty about how terrestrial anthropogenic covariates are connected to specific subtidal locations. To account for spatial and temporal sampling variation, we only estimated trends in eelgrass for shore segments that had sufficient numbers of observations (≥3 observations within 0 5 km) in at least 10 of the 20 possible 2-year time periods. Thus, only shore segments with observations spanning at least 20 years were used to describe trends. Because the first year of data varies among sites (Table S1.1), we estimated eelgrass occurrence for all sites beginning in 1981– 1982, and thus, we present the total area of eelgrass for 1981–1982 to 2011–2012. 4 0 661 temporal scale at each site that can be aggregated to produce a picture of eelgrass change in Puget Sound at multiple spatial and temporal scales. We model the presence or absence of eelgrass for observation i in time t at shoreline segment s, Yit(s), using a binomial generalized linear mixed model (GLMM) with a logit link. Let pit(s) represent the probability of occurrence at segment s in year t, Xit(s) represent a matrix of fixed covariates (described below), b represent regression coefficients estimated for each covariate, and et(s) represent spatiotemporal random effects, then, Y it ðsÞ Bernouliðpit ðsÞÞ eqn 1 logitðpit ðsÞÞ ¼ X it ðsÞb þ et ðsÞ: Thus, the occurrence of eelgrass is described by the contribution of measured environmental covariates as well as a random effect to include the contribution of unmeasured – but potentially spatially and temporally varying – covariates. We provide alternative model structure for incorporating shared trends across shoreline locations in Appendix S1. Furthermore, we note that the model structure employed is extremely general and flexible and can be used to account for a wide variety of model structures (see also Cressie & Wikle 2011). This statistical approach assumes that the eelgrass collected by the rake provides a statistically unbiased sample of the presence of eelgrass. While we have no data about the sampling efficacy of the rake itself, it is possible that it either over- or under-represents the true occurrence of eelgrass. Therefore, while we present estimates of occurrence and total eelgrass area, these estimates may by biased by an unknown constant of proportionality. Since the same equipment and methodology have been used for all sites and the entire times series, this effect will be constant across our entire data set, making comparisons within and among sites appropriate. Thus, while absolute estimates of abundance may be statistically biased, estimates of trend are unaffected. STATISTICAL ANALYSES To understand changes in eelgrass, we constructed spatio-temporal statistical models of eelgrass occurrence for each survey location. The model provides estimates of eelgrass occurrence at a fine spatial and COVARIATES We included five continuous covariates as fixed effects in our model. The first, water depth (m), was recorded during the survey and is Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology Resilient seagrass in an urban estuary 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 available for each observation (Appendix S2). We expect an optimum depth to be associated with eelgrass distributions and thus include both linear and quadratic terms (depth and depth2) as covariates. Since we do not have the time at which measurements were taken, we cannot adjust recorded depths relative to a standard tidal datum (e.g. mean low water), and so we used the reported depth. The resulting observation error in the water depth measurements likely introduces uncertainty in our estimates of the effect of depth on eelgrass occurrence. We evaluated the effect of four additional covariates on eelgrass, focusing on physical attributes of, and anthropogenic modifications to, nearshore habitat hypothesized to influence eelgrass presence and distribution. We included two physical attributes important to seagrass: wave exposure (Stevens & Lacy 2012) and bathymetric slope (Koch 2001), measured as distance to the 5 48 m (3 fathom) isobaths (Appendix S2). We included two anthropogenic covariates: the proportion of shoreline that is armoured (bulkheads, riprap, concrete walls, etc.) and the mean imperviousness (a general proxy for human development) of a 150-m buffer zone nearest each shoreline segment (Appendix S2). As with the rake survey data, we assigned covariate values to each 100 m shoreline segment. Both armouring and impervious cover variables were fixed values for the entire time series; information about changes to armouring and impervious surfaces in Puget Sound was not available across the time series in adequate spatial or temporal detail for this analysis, and used values represent covariates c. 2005 (Shipman 2008; Simenstad et al. 2011). Additional anthropogenic covariates were initially considered, but available time series for proxies of likely stressors like eutrophication or sedimentation were unavailable. Publicly available information on marine water characteristics such as nitrate, chlorophyll a, phosphate and other nutrients in Puget Sound begin in 1999 (PSEMP 2015) and is unavailable for earlier years. As Puget Sound is a marine-dominated estuary, humans contribute proportionally less to marine nutrients than in many other estuarine systems (Steinberg et al. 2010; Mohamedali et al. 2011; Cope & Roberts 2013). To evaluate possible site-wide change in eelgrass across time, we also estimated model that included fixed offsets for each time period. We estimated models both with an intercept for each time as well as a model with a single, shared intercept. To reduce colinearity of the covariates, we centred the distance to the 5 48 m isobath (3 fathom, hereafter ‘bathymetric slope’) and weighted wave exposure (hereafter ‘fetch’) covariates before estimation and square-root-transformed fetch to reduce the statistical leverage of extremely large values. SPATIO-TEMPORAL RANDOM EFFECTS In eqn 1, the random effect et(s) represents a vector of spatiotemporal effects at time t and shoreline segment s, and accounts for processes not included in the fixed covariates. To match the structure of available covariates and improve computation, we used the shoreline to provide a one-dimensional background for the random effects. Thus, points that are proximate (either in space or time) will be more strongly correlated than points that are further apart. We write et(s) as a first-order autoregressive process, et ðsÞ Normalðqet 1 ðsÞ; RÞ eqn 2 where q defines the strength of temporal autocorrelation and Σ is the covariance matrix that represents spatial covariation in random effects. We constrained q to ensure stationarity: 1 < q < 1. We used a Mat ern covariance function with m = 3/2 to define Σ and estimated fixed effects via maximum marginal likelihood using the Laplace approximation to approximate the integral across random effects (see Appendix S1). We implemented our model using the R package ‘INLA’ (www.r-inla.org). To estimate the spatio-temporal random effects, we introduced a knot approximately every 0 5 km of shoreline and used second degree smoothing to describe the spatial field along the shoreline. Further model description and example code can be found in Appendix S1 and in the data archive accompanying this article. To define the spatial extent of each area, we included shoreline points that were observed regularly during the time series and we included a 0 5-km buffer along the extreme edges of the observed sample points. MODEL OUTPUTS AND PREDICTIONS We use the above statistical model for two main purposes. First, from model estimates we made predictions of eelgrass occurrence across all modelled years and combine these predictions to understand temporal changes to eelgrass populations at small (c. 0 1 km) to large (100s km) scales. Secondly, estimated parameters informed the effect of physical and anthropogenic covariates on eelgrass occurrence, and the spatial and temporal scales at which eelgrass respond to unmeasured covariates and stochastic forces. Since we estimated the model for each survey location independently (11 total), we had eleven estimates for each fixed covariate. We used these eleven estimates to ask how the direction and magnitude of each covariate compared across locations. We also examined estimates of spatial variance and temporal autocorrelation parameters at each site to understand the contribution of unmeasured spatial and temporal effects to eelgrass occurrence. To predict occurrence, we generated MCMC predictions for the probability of occurrence at the centroid of each 30 9 30 m grid of cells occurring with shallow coastal waters of Puget Sound. As the area in each grid cell is 900 m2, the estimated area of eelgrass area for MCMC sample g at time t and location s is 900pgt ðsÞ. The eelgrass area at each site is simply the sum across all grid cells. We produced predictions for a depth range from 0 to 6 m below mean low water at each site in each time period. We made 1000 predictions for each grid cell in each time period and used these predictions to calculate eelgrass trends and bound uncertainty about the trends. In total, we generated predictions for 554 3 km of shoreline representing 13 543 ha of benthic habitat (see Data Archive for a GIS data base of spatial predictions). We focus on understanding patterns of eelgrass change at three scales. Starting at the largest scale, we combine information from the 14 trend areas in Puget Sound to produce aggregate measures of eelgrass trend at the scale of hundreds of km of shoreline. Because some areas were not first surveyed until 1981, we provide aggregate estimates of eelgrass only for the period 1981–1982 to 2011–2012. We express our uncertainty for the total Puget Sound area as the sum of individual site variances (estimates for individual trend areas are approximately normally distributed). At the scale of tens of km scale, we assessed eelgrass and asked whether trends varied within Puget Sound. We estimated linear trends for each site using ordinary least-squares regression on the predicted time series and express the change at each site as a percentage change in area relative to the time series mean of area present at each site. We also used available GIS information to ask whether trends in eelgrass related to site-wide measures of anthropogenic change. We regressed the eelgrass population growth rate against two measures of human impact: the cumulative proportion of shoreline armoured at each site and the area weighted mean imperviousness for drainages leading into each site (Appendix S2). In addition, we use available Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 6 A. O. Shelton et al. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 human demographic data to calculate the human population density in the freshwater drainages flowing into each area between 1970 and 2000 and regressed rates of human density change against eelgrass growth (Appendix S2). Finally, at the 0 1-km scale, we asked whether changes in occurrence were homogeneous site-wide or whether intra-site variability underlied site-level changes. We visualized the distribution of eelgrass occurrence both by making maps to identify areas of change and by constructing frequency distributions of the occurrence of eelgrass across all grid cells within each site. To contrast time periods, we compared early (mean occurrence for each grid cell for available years before 1984) and recent (mean occurrence between 2009 and 2012) occurrence. For frequency distributions, we generated probability distributions of eelgrass occurrence using a Gaussian kernel smoother with a bandwidth of 0 02. Results We estimated eelgrass occurrence at a fine spatial and temporal scale in 11 Puget Sound survey locations and generated predictions about changes in eelgrass and associated environmental and human covariates for 14 sites of comparable size. Hereafter, we present result for sites in order from north to south to enhance the ability to detect geographic patterns in the results (Fig. 1). The sites can be classified into two groups: sites where eelgrass has never been abundant (CP, PO, BA, QM; time-average of eelgrass probability of occurrence <0 4) and the remaining 10 sites where more than half of the area is covered by eelgrass on average (Table 1). While the shoreline lengths are relatively similar among sites, the bathymetry of Puget Sound results in substantial variation in subtidal area among sites; each site represents between 294 and 1463 ha of area used for determining trend (Table 1). Taken together, this means that sites do not contribute equally to the total area examined in our study. Estimated eelgrass area ranged from a maximum estimate of 1154 ha in 2012 (LM) to only 58 ha (BA). AGGREGATE AND LOCAL TRENDS OF EELGRASS Across 274 km of shoreline representing 8055 ha used for detecting eelgrass trends, we found remarkable stability in the aggregate eelgrass area across survey areas between 1981 and 2012 (Fig. 2). There was no clear signature of decline in eelgrass; the linear regression estimated eelgrass area was flat over the time series (point estimate of 0 02% annual increase). During this same time period, the human population in counties bordering Puget Sound increased from about 2 63 to 4 45 million (Minnesota Population Center 2011). However, this aggregate flat trend masks substantial variation among sites (Fig. 3). We identified five sites with negative trends (≤ 0 5% annual change; CP, FB, DB, PG, BA; Fig. 3), two sites with strong positive trends (≥0 5% annual change: LM, PS), and the seven sites with annual change between 0 5% and 0 5%. There is no obvious geographic structure to the trend from north to south; indeed, an increasing area is occasionally directly adjacent to a decreasing area (compare LM and CP; Figs 1 and 3). Fig. 2. (a) Cumulative eelgrass area across the 14 trend areas and aggregate trend (annualized per cent area change). Vertical dotted lines indicate major El Ni~no events (1982–1983 and 1997–1998) and dash-dot lines indicate major La Ni~na events (1988–1989 and 2010– 2011). (b) Frequency distribution of annualized per cent area change. Vertical line indicates estimated cumulative change from the top panel. COVARIATE EFFECTS There were varying effects of physical and anthropogenic covariates on eelgrass occurrence. As expected, eelgrass occurrence was strongly negatively associated with water depth (Table S1.3) corresponding to increased light limitation at greater depths. Depth coefficients were strongly negative with all sites having either a negative quadratic (10 sites) or negative linear (one site) effect of depth on eelgrass occurrence (recall that covariate effects are derived for each survey – 11 total – rather than for the 14 trend areas) Physical variables had no consistent effect on eelgrass occurrence at the 0 1 km scale. Across all areas, most estimated coefficients for fetch overlapped 0, indicating no strong effect on eelgrass (Table S1.3). At two sites (HH, PG) eelgrass had increased occurrence with higher fetch and two sites (CP, QB) had a negative association between fetch and eelgrass. Bathymetric slope had minimal measurable effect on eelgrass occurrence; coefficients for 9 of 11 areas overlapped with zero, one was positive (CP), and one was negative (HH; Table S1.3). Anthropogenic variables at the 0 1-km scale were also poor predictors of eelgrass occurrence (Fig. S1.2). Shoreline Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology Resilient seagrass in an urban estuary 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Fig. 3. Time series for each of 14 trend areas. Points and lines indicate mean, shading indicates 95% CI. Dashed line shows linear trend for each site. Note that the y-axis varies among panels. armouring had a negative relationship with eelgrass at 8 of 11 areas but only three had confidence intervals that did not include 0. The magnitude of any armouring effect was small (all slopes are between 1 and 1), suggesting that even if there was a true effect of shoreline armouring on eelgrass, its magnitude was small. Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 8 A. O. Shelton et al. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 There was minimal evidence of any effect of impervious cover on eelgrass: 9 of 11 sites have coefficients that overlap 0 (Fig. S1.2). The notable outlier was QM with a negative, but highly uncertain, coefficient. Larger coefficients for impervious cover relative to shoreline armouring arise primarily because the range of observed impervious cover was much lower than the range of shoreline armouring (see Fig. S1.1). Estimates of spatio-temporal random effects et(s) were often large, indicating a substantial influence of spatial location on eelgrass occurrence. The estimated autocorrelation coefficient (q) was >0 90 for all sites but one (Table 2), indicating that the value at a segment at time t was a strong predictor of time t + 1. Additionally, the estimated spatial range (the longshore distance at which the correlation between two points falls to 0 05) was between 1 and 3 km at all sites (Table 2), indicating the spatial scale for factors that may be affecting eelgrass occurrence. For both autocorrelation and spatial range, HH was notably different than the other sites. HH was unique among our sites because there was very little spatial or temporal variation in eelgrass at HH; nearly the entire surveyed area was occupied by eelgrass and this did not change notably over the study period making the spatial field and associated parameters difficult to estimate. Site-level comparisons of anthropogenic drivers provide no evidence of strong effects of shoreline armouring, imperviousness or change in human density on eelgrass growth (Fig. 4). All examined relationships were not significant though there was a trend towards a negative relationship between human population growth rate and eelgrass population growth rate. This relationship was non-significant and driven by a single site (CP) with a large proportional increase in human density. of the predicted points on the 30 9 30 m grid for two periods: pre-1984 and 2009–2012 (Fig. 5). For example, sites DB and FB show substantial declines in median occurrence driven by generalized decline in eelgrass occurrence at many locations within the site. BB and PS provide examples of generalized increase in eelgrass. In contrast, while the median occurrence changes only slightly at PG, a substantial portion of the distribution in 2009–2012 is near 1 while another component is very near 0. The near 0 component was not present in the earlier period, indicating that eelgrass has been virtually lost from part of the site. A similar pattern of localized change is evident for BA, QB and QM. Other sites appear very stable (e.g. SB, HH). IDENTIFYING AREAS OF CHANGE Above we show minimal change in eelgrass at the level of Puget Sound (100-km scale), but substantial among-site variation (10-km scale). However, site variation was not generally driven by site-wide changes in eelgrass occurrence but rather shifts in occurrence that occur intra-site (0 1- to 1-km scale). We illustrate small-scale changes in occurrence in two ways. First, we provide the distribution of occurrence values for all Table 2. Estimates of spatio-temporal random effect parameters. Spatial range indicates the long-shore distance at which correlation falls to 0 05 Site Autocorrelation (q) Spatial range (km) Spatial variance (r2) BB CP FB DB HH PS PG QB PO HC QM 0 97 0 96 0 93 0 95 0 77 0 98 0 97 0 97 0 97 0 97 0 98 2 5 1 5 1 5 1 7 9 4 2 3 1 1 1 7 2 0 1 5 3 4 3 01 2 92 2 16 2 09 2 49 2 74 2 05 4 47 4 73 1 15 7 53 (0 94, (0 95, (0 89, (0 93, (0 55, (0 96, (0 95, (0 95, (0 96, (0 93, (0 97, 0 99) 0 97) 0 96) 0 97) 0 92) 0 99) 0 98) 0 98) 0 98) 0 99) 0 99) (1 5, (1 3, (1 0, (1 3, (5 1, (1 6, (0 7, (1 3, (1 6, (0 8, (2 6, 3 9) 1 8) 2 4) 2 3) 17 8) 3 4) 1 6) 2 3) 2 5) 2 4) 4 5) (1 64, (2 27, (1 34, (1 45, (1 28, (1 62, (1 42, (3 18, (3 43, (0 69, (4 28, 5 50) 3 66) 3 52) 3 01) 4 99) 4 78) 2 98) 6 34) 6 62) 1 93) 13 89) Fig. 4. Relationships between trend area eelgrass growth rate and shoreline armour (a), imperviousness (b) and proportional change in human density (c). Each point represents a trend area. Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology Resilient seagrass in an urban estuary 9 1·0 Colour online, B&W in print 0·8 Pre–1984 2009–2012 0·6 0·4 Eelgrass occurrence 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Fig. 5. Distribution of eelgrass occurrence at each site before 1984 (dark green) and in the most recent four years (2009–2012). Shaded areas represent a kernel smoothed probability density of eelgrass occurrence across all grid cells at a site. Grid cells included in the density are identical for each time period and all distributions integrate to 1. Horizontal lines indicate the median occurrence for each probability density at each time period. [Colour figure can be viewed at wileyonlinelibrary.com] 0·2 0·0 SB BB CP LM FB DB HH 1·0 0·8 Pre–1984 2009–2012 0·6 0·4 0·2 0·0 PS Secondly, we provide maps for three sites to highlight particular areas that have experienced large changes – both positive and negative – over the period (PG, PO, and BA; Fig. 6). We compare predictions from 1981–1982 and 2011– 2012. These maps illustrate the patchwork nature of eelgrass bed change in Puget Sound with areas of virtually no change interspersed with areas of both positive and negative change including areas where eelgrass has been entirely lost from stretches of shoreline. We provide figures comparing 1981– 1982 and 2011–2012 for the remaining sites in the online supplement (Appendix S3). Discussion Seagrasses provide some of the most productive and ecologically significant marine ecosystems in the world, which motivates understanding their status, trend and potential threats (Waycott et al. 2009). Seagrass beds thrive in estuarine and nearshore habitats, a nexus of human threats from use and development that are projected to increase in coming decades (Timmerman & White 1997; Seto, G€uneralp & Hutyra 2012; Neumann et al. 2015). To our knowledge, our analysis presents the longest data set on eelgrass in the north Pacific and among the longest eelgrass time series in the world. It provides an unparalleled opportunity to examine multiscale (both spatial and temporal) trends in eelgrass. Though the data we use were not collected for the purpose of monitoring eelgrass, we feel confident about our conclusions about eelgrass trends in Puget Sound as our basic trends post-2000 match a formal, PG QB PO BA HC QM stratified eelgrass monitoring programme initiated in the early 2000s (WDNR 2015). Additionally, some specific areas of eelgrass loss identified in our analyses have been previously identified by local experts (Thom et al. 2014) and increases confidence in our results. These results show remarkable large-scale stability in eelgrass across broadly dispersed sites in Puget Sound. This stability occurs despite a rapidly increasing human population (population more than doubled between 1970 and 2010; Minnesota Population Center 2011), multiple major oceanographic anomalies including several major El Ni~ no events (Trenberth et al. 2002; Clarke 2014). This stability is also coincident with the implementation of wide ranging of environmental legislation aimed at improving Puget Sound water quality [e.g. Clean Water Act (FWPCA 1972); Washington State Shoreline Management Act (RCW 1971)] and specific protection of eelgrass as a ‘habitat of special concern’ by WDFW. Our results have two main general implications. First, and most generally, our work presents general statistical framework for understanding spatial and temporal change in nearshore habitats. While we apply our methods to presence– absence data, the framework is directly applicable to continuous response variables (e.g. plant density), to more complicated two-dimensional spatial problems (Ward et al. 2015; Ono et al. 2016), and to alternative time series structures. Recent studies have developed various approaches to analysing the spatial distribution of seagrasses (Coles, McKenzie & De’ath 2009; Grech & Coles 2010; March et al. 2013; Schubert et al. 2015), using remote sensing data to Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 Colour online, B&W in print 10 A. O. Shelton et al. Fig. 6. Estimated mean eelgrass occurrence at sites in Puget Sound in 1981–1982 (left column) and 2011–2012 (right column). The top two panels show Port Gamble Bay (PG), and the bottom shows Port Orchard Bay incorporating parts of site PO and site BA. We make predictions for eelgrass occurrence between mean low water and 6 m. Locations labelled with numbers highlight three stretches with substantial loss (1 and 3) and gain (2) in eelgrass occurrence over the 30 years. Locations above mean low water are coloured in tan, and depths below 6 m are coloured in blue. [Colour figure can be viewed at wileyonlinelibrary.com] assess spatio-temporal change (Lyons, Roelfsema & Phinn 2013) and projecting seagrass distributions under future climate and development scenarios (Saunders et al. 2013), but none to our knowledge has explicitly used spatio-temporal models as we have here. While we were not able to ascribe changes in seagrass to specific measured covariates here, the model identifies locations that have experienced changes during the past 40 years (Fig. 6; Appendix S3). Identified areas of change provide excellent candidates for future studies to elucidate mechanisms underlying the loss or gain of eelgrass and as potential restoration targets. Our results do not coincide with expected responses of eelgrass to increases in human population size (Orth et al. 2006; Patrick et al. 2014) whereby light limitation of eelgrass via Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology Resilient seagrass in an urban estuary 11 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 algal overgrowth and shading (Burkholder, Tomasko & Touchette 2007) is associated with human presence and increased nutrients (Seitzinger et al. 2002). This is likely in part a result of the fjord-like structure of Puget Sound where nutrients are predominantly of oceanic origin and nutrients of terrestrial origin are frequently flushed from shallow nearshore areas (Mackas & Harrison 1997; Steinberg et al. 2010). These results also suggest that using aggregate metrics of human impacts, such as impervious surface or human population density in associated watersheds, may not be adequate for assessing the impacts of very localized stressors, such as local nutrient sources or shoreline hardening, on nearshore habitats. These results do reveal critical information about the spatial scale at which eelgrass responds to the environment. Despite the lack of a detectable signature of shared stressors on eelgrass across sites, research suggests that eelgrass populations are influenced by regional oceanic and climate forces (Brown et al. 2013; Valle et al. 2014). However, over the past 40 years, eelgrass in Puget Sound has proven resilient to large-scale climatic and anthropogenic change; we do not see coincident changes in eelgrass populations that would indicate a major shared climatic driver across sites (Fig. 3). Instead, our results show that substantial changes to eelgrass populations occur at the site and subsite level. Our sites show no obvious geographic coherence in trends of eelgrass – adjacent sites can have opposite trends – and geographic coherence would be expected if shared oceanographic or climate drivers controlled eelgrass trends. We conclude that the role of local drivers in determining eelgrass persistence is predominant. Our results suggest the relevant spatial scale for evaluating eelgrass change and identifying potential drivers is approximately 1–3 km of shoreline (Table 2). While we identified locations of significant eelgrass change and identified the appropriate scale for investigating changes, we did not find a clear signature of shoreline armouring or impervious surfaces on eelgrass at scales ranging from 0 1 to 10 km. A major limitation of our covariates were that they were static, reflecting the armouring present c. 2005; no information is available about how these variables changed during the course of our study, potentially leading to underestimates of anthropogenic effects. As most major changes to Puget Sound shorelines long predate 1970 – the ports of Seattle, Tacoma and Everett were developed largely in first decades of the 20th century and likely resulted in large losses of eelgrass – it is unlikely that additional data would radically change our broad scale results. However, we do note that none of our sites include nearshore habitats directly adjacent to dense urban areas like downtown Seattle (Fig. 1) and so we lack information about eelgrass change in response to high levels of urbanization. Eelgrass populations are stable and resilient in the face of substantial changes to the Puget Sound environment over the past 40 years. However, we strongly suspect the development of Puget Sound pre-1970 led to the widespread loss of eelgrass as has been shown in other systems (Serrano et al. 2016). Our work highlights how data on focal species without data on potential pressures of similar spatial and temporal scales can limit the ability to make inference about the drivers of trends. With access to a single snapshot of armouring in Puget Sound c. 2005 (Shipman 2008; Simenstad et al. 2011) and time series of land use and land cover that only include the past 20 years (Vogelmann et al. 2001; Xian et al. 2011; NOAA 2013), no data are available to systematically inform changes of the shoreline before 2000 at the c. 1-km scale. Time series of other potentially important covariates such as nutrient availability and sediments that may affect eelgrass through light availability are similarly unavailable for much of our time frame. This indicates clearly the need for joint monitoring of populations and threats to understand trends in seagrasses, and likely applies to understanding coastal habitats world-wide. Our results show that despite concerns over the potential for catastrophic, rapid changes in seagrass populations (Orth et al. 2006; Waycott et al. 2009; Thomson et al. 2015), eelgrass population change in Puget Sound is relatively slow and gradual. While very local population declines in Puget Sound have been reported (WDNR 2015) and are hypothesized to be driven by factors ranging from low seedling survival (Dooley et al. 2013) to wasting disease (Groner et al. 2014), shared, Puget Sound-wide drivers of eelgrass loss are not evident. We do note that because our data are limited to occurrence and do not include density information, our results do not reflect any changes in eelgrass shoot density that may have occurred over the period. We cannot preclude the possibility of future collapses but our analyses suggest that in general management and restoration at the local level can lead to incremental improvements and gradual increases in seagrass populations. Fortunately, as conservation and restoration project implementation largely occur on local spatial scales, our results suggest that localized problems may have local solutions. Acknowledgements We thank the many WDFW staff and researchers who have contributed to the WDFW herring surveys over the years and especially K. Stick for his advice and help in understanding herring survey data. We thank B. Huebner, H. Moran, D. Annis and C. Robertson for help digitizing survey data. R. Thom, K. Stick and D. Lowry provided helpful comments on preliminary results. J. Gaeckle, B. Christiaen, J. Samhouri, K. Stick and D. Lowry improved earlier versions of this manuscript. This document has been funded wholly or in part by the United States Environmental Protection Agency under assistance agreement 00J30301-0 to the University of Washington. The contents of this document do not necessarily reflect the views and policies of the Environmental Protection Agency nor does mention of trade names or commercial products constitute endorsement or recommendation for use. Data accessibility Data and GIS files associated with this article are accessible via Dryad Digital Repository, http://dx.doi.org/xxxxx (Shelton et al. 2016) References Arkema, K.K., Guannel, G., Verutes, G., Wood, S.A., Guerry, A., Ruckelshaus, M., Kareiva, P., Lacayo, M. & Silver, J.M. (2013) Coastal habitats shield people and property from sea-level rise and storms. Nature Climate Change, 3, 913–918. Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 6 12 A. O. Shelton et al. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 7 18 Barbier, E.B., Hacker, S.D., Kennedy, C., Koch, E.W., Stier, A.C. & Silliman, B.R. (2011) The value of estuarine and coastal ecosystem services. Ecological Monographs, 81, 169–193. Brown, C.J., Saunders, M.I., Possingham, H.P. & Richardson, A.J. (2013) Managing for interactions between local and global stressors of ecosystems. PLoS ONE, 8, e65765. Burkholder, J.M., Tomasko, D.A. & Touchette, B.W. (2007) Seagrasses and eutrophication. Journal Of Experimental Marine Biology And Ecology, 350, 46–72. Burns, R. (1985) The Shape and Form of Puget Sound. ?????, ?????. Cardoso, P.G., Leston, S., Grilo, T.F., Bordalo, M.D., Crespo, D., Raffaelli, D. & Pardal, M.A. (2010) Implications of nutrient decline in the seagrass ecosystem success. Marine Pollution Bulletin, 60, 601–608. Clarke, A.J. (2014) El Ni~no physics and El Ni~no predictability. Annual Review of Marine Science, 6, 79–99. Coles, R., McKenzie, L. & De’ath, G. (2009) Spatial distribution of deepwater seagrass in the inter-reef lagoon of the Great Barrier Reef World Heritage Area. Marine Ecology Progress Series, 392, 57–68. Cope, B. & Roberts, M. (2013) Review and Synthesis of Available Information to Estimate Human Impacts to Dissolved Oxygen in Hood Canal. Washington State Department of Ecology and Environmental Protection Agency, Olympia, WA, USA. Costanza, R., Arge, D., Groot, D. et al. (1997) The value of ecosystem service and nature capital in the world. Nature, 387, 253–260. Cressie, N. & Wikle, C.K. (2011) Statistics for Spatio-Temporal Data. John Wiley & Sons, Hoboken, NJ, USA. Dayton, P.K. (1972) Toward an understanding of community resilience and the potential effects of enrichments to the benthos at McMurdo Sound, Antarctica. Proceedings of the Colloquium on Conservation Problems in Antarctica, pp. 81–96. Dooley, F.D., Wyllie-Echeverria, S., Roth, M.B. & Ward, P.D. (2013) Tolerance and response of Zostera marina seedlings to hydrogen sulfide. Aquatic Botany, 105, 7–10. Duarte, C.M. (2002) The future of seagrass meadows. Environmental Conservation, 29, 192–206. Duffy, J.E., Reynolds, P.L., Bostr€om, C. et al. (2015) Biodiversity mediates top-down control in eelgrass ecosystems: a global comparative-experimental approach. Ecology Letters, 18, 696–705. Ebbesmeyer, C.C. & Barnes, C.A. (1980) Control of a fjord basin’s dynamics by tidal mixing in embracing sill zones. Estuarine and Coastal Marine Science, 11, 311–330. Fourqurean, J.W., Duarte, C.M., Kennedy, H. et al. (2012) Seagrass ecosystems as a globally significant carbon stock. Nature Geoscience, 5, 505–509. FWPCA (Federal Water Pollution Control Act). (1972) 33 U.S.C. §1251 Et Seq. Gacia, E., Granata, T.C. & Duarte, C.M. (1999) An approach to measurement of particle flux and sediment retention within seagrass (Posidonia oceanica) meadows. Aquatic Botany, 65, 255–268. Gallegos, C.L. & Bergstrom, P.W. (2005) Effects of a Prorocentrum minimum bloom on light availability for and potential impacts on submersed aquatic vegetation in upper Chesapeake Bay. Harmful Algae, 4, 553–574. Grech, A. & Coles, R.G. (2010) An ecosystem-scale predictive model of coastal seagrass distribution. Aquatic Conservation: Marine and Freshwater Ecosystems, 20, 437–444. Grech, A., Chartrand-Miller, K., Erftemeijer, P., Fonseca, M., McKenzie, L., Rasheed, M., Taylor, H. & Coles, R. (2012) A comparison of threats, vulnerabilities and management approaches in global seagrass bioregions. Environmental Research Letters, 7, 024006. Greene, C.M., Blackhart, K., Nohner, J., Candelmo, A. & Nelson, D.M. (2015) A national assessment of stressors to estuarine fish habitats in the contiguous USA. Estuaries and Coasts, 38, 782–799. Groner, M.L., Burge, C.A., Couch, C.S. et al. (2014) Host demography influences the prevalence and severity of eelgrass wasting disease. Diseases of aquatic organisms, 108, 165–175. Heck, K.L., Hays, G. & Orth, R.J. (2003) Critical evaluation of the nursery role hypothesis for seagrass meadows. Marine Ecology Progress Series, 253, 123–136. Heck, K.L. & Thoman, T.A. (1984) The nursery role of seagrass meadows in the upper and lower reaches of the Chesapeake Bay. Estuaries, 7, 70– 92. Heck, K.L., Pennock, J.R., Valentine, J.F., Coen, L.D. & Sklenar, S.A. (2000) Effects of nutrient enrichment and small predator density on seagrass ecosystems: An experimental assessment. Limnology & Oceanography, 45, 1041– 1057. Heck, K.L. Jr, Carruthers, T.J.B., Duarte, C.M., Hughes, A.R., Kendrick, G., Orth, R.J. & Williams, S.W. (2008) Trophic transfers from seagrass meadows subsidize diverse marine and terrestrial consumers. Ecosystems, 11, 1198– 1210. Hughes, A. & Stachowicz, J. (2004) Genetic diversity enhances the resistance of a seagrass ecosystem to disturbance. Proceedings of the National Academy of Sciences of the United States of America, 101, 8998–9002. Imhoff, M.L., Bounoua, L., Ricketts, T., Loucks, C., Harriss, R. & Lawrence, W.T. (2004) Global patterns in human consumption of net primary production. Nature, 429, 870–873. Irlandi, E.A., Ambrose, W.G. & Orlando, B.A. (1995) Landscape ecology and the marine environment: how spatial configuration of seagrass habitat influences growth and survival of the bay scallop. Oikos, 72, 307. Koch, E.W. (2001) Beyond light: Physical, geological, and geochemical parameters as possible submersed aquatic vegetation habitat requirements. Estuaries, 24, 1–17. Koch, E.W., Barbier, E.B., Silliman, B.R. et al. (2009) Non-linearity in ecosystem services: temporal and spatial variability in coastal protection. Frontiers in Ecology and Environment, 7, 29–37. Koch, M., Bowes, G., Ross, C. & Zhang, X.H. (2013) Climate change and ocean acidification effects on seagrasses and marine macroalgae. Global Change Biology, 19, 103–132. Lee, K.-S. & Dunton, K.H. (1999) Inorganic nitrogen acquisition in the seagrass Thalassia testudinum: development of a whole-plant nitrogen budget. Limnology and Oceanography, 44, 1204–1215. Li, X., Weller, D.E., Gallegos, C.L., Jordan, T.E. & Kim, H.-C. (2007) Effects of watershed and estuarine characteristics on the abundance of submerged aquatic vegetation in Chesapeake Bay subestuaries. Estuaries and Coasts, 30, 840–854. Lotze, H., Lenihan, H., Bourque, B. et al. (2006) Depletion, degradation, and recovery potential of estuaries and coastal seas. Science, 312, 1806–1809. Lyons, M.B., Roelfsema, C.M. & Phinn, S.R. (2013) Towards understanding temporal and spatial dynamics of seagrass landscapes using time-series remote sensing. Estuarine, Coastal and Shelf Science, 120, 42–53. Mackas, D.L. & Harrison, P.J. (1997) Nitrogenous nutrient sources and sinks in the Juan de Fuca Strait/Strait of Georgia/Puget Sound Estuarine System: assessing the potential for eutrophication. Estuarine, Coastal and Shelf Science, 44, 1–21. March, D., Al os, J., Cabanellas-Reboredo, M., Infantes, E., Jordi, A. & Palmer, M. (2013) A Bayesian spatial approach for predicting seagrass occurrence. Estuarine, Coastal and Shelf Science, 131, 206–212. McArthur, L.C. & Boland, J.W. (2006) The economic contribution of seagrass to secondary production in South Australia. Ecological Modelling, 196, 163– 172. McConnaughey, T. & McRoy, C.P. (1979) 13C label identifies eelgrass (Zostera marina) Carbon in an Alaskan estuarine food web. Marine Biology, 53, 263–269. McGlathery, K.J., Reynolds, L.K., Cole, L.W., Orth, R.J., Marion, S.R. & Schwarzschild, A. (2012) Recovery trajectories during state change from bare sediment to eelgrass dominance. Marine Ecology Progress Series, 448, 209–221. Minnesota Population Center (2011) National Historical Geographic Information System: Version 2.0. Minnesota Population Center, ?????. Mohamedali, T., Roberts, M., Sackmann, B. & Kolosseus, A. (2011) Puget Sound Dissolved Oxygen Model Nutrient Load Summary for 1998–2008. Washington State Department of Ecology, Olympia, WA, USA. Neumann, B., Vafeidis, A.T., Zimmermann, J. & Nicholls, R.J. (2015) Future coastal population growth and exposure to sea-level rise and coastal flooding–a global assessment. PLoS ONE, 10, e0118571. NOAA (2013) NOAA’s Coastal Change Analysis Program (C-CAP) 1992 to 2010 Regional Land Cover Change Data – Coastal United States. NOAA, ?????. Ono, K., Shelton, A.O., Ward, E.J. & Thorson, J.T. (2016) Space-time investigation of the effects of fishing on fish populations. Ecological Applications, 26, 392–406. Orth, R.J., Carruthers, T.J.B., Dennison, W.C. et al. (2006) A global crisis for seagrass ecosystems. BioScience, 56, 987–996. Partnership, P.P.S. (2011) Leadership Council Resolution 2011-01: Adopting an Ecosystem Recovery Target for Eelgrass. PDF http://www.Mypuget sound.Net/Index.Php?Option=Com_Docman&Task=Cat_View&Gid=323& Itemid=238 (accessed 29 March 2016). Patrick, C.J., Weller, D.E. & Ryder, M. (2016) The relationship between shoreline armoring and adjacent submerged aquatic vegetation in Chesapeake bay and nearby Atlantic coastal bays. Estuaries and Coasts, 39, 158–170. Patrick, C.J., Weller, D.E., Li, X. & Ryder, M. (2014) Effects of shoreline alteration and other stressors on submerged aquatic vegetation in subestuaries of Chesapeake bay and the mid-Atlantic coastal bays. Estuaries and Coasts, 37, 1516–1531. Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 8 9 10 Resilient seagrass in an urban estuary 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 11 12 Plummer, M.L., Harvey, C.J., Anderson, L.E., Guerry, A.D. & Ruckelshaus, M.H. (2013) The role of eelgrass in marine community interactions and ecosystem services: results from ecosystem-scale food web models. Ecosystems, 16, 237–251. PSEMP Marine Waters Workgroup (2015) Puget Sound Marine Waters: 2014 Overview (eds S.K. Moore, R. Wold, K. Stark, J. Bos, P. Williams, K. Dzinbal, C. Krembs & J. Newton). http://www.psp.wa.gov/PSEMP/PSmarinewa tersoverview.php. Contact email: marinewaters@psemp.org. Rahbek, C. (2005) The role of spatial scale and the perception of large-scale species-richness patterns. Ecology Letters, 8, 224–239. RCW Revised Code of Washington 90.58.06 23 (1971) Shoreline Management Act. ?????, ?????. Rehr, A.P., Williams, G.D., Tolimieri, N. & Levin, P.S. (2014) Impacts of terrestrial and shoreline stressors on eelgrass in Puget Sound: an expert elicitation. Coastal Management, 42, 246–262. Sale, P.F., Agardy, T., Ainsworth, C.H. et al. (2014) Transforming management of tropical coastal seas to cope with challenges of the 21st century. Marine Pollution Bulletin, 85, 8–23. Saunders, M.I., Leon, J., Phinn, S.R. et al. (2013) Coastal retreat and improved water quality mitigate losses of seagrass from sea level rise. Global Change Biology, 19, 2569–2583. Schubert, P.R., Hukriede, W., Karez, R. & Reusch, T.B.H. (2015) Mapping and modeling eelgrass Zostera marina distribution in the western Baltic sea. Marine Ecology Progress Series, 522, 79–95. Seitzinger, S.P., Kroeze, C., Bouwman, A.F., Caraco, N., Dentener, F. & Styles, R.V. (2002) Global patterns of dissolved inorganic and particulate nitrogen inputs to coastal systems: Recent conditions and future projections. Estuaries, 25, 640–655. Serrano, O., Lavery, P., Masqu e, P., Inostroza, K., Bongiovanni, J. & Duarte, C. (2016) Seagrass sediments reveal the long-term deterioration of an estuarine ecosystem. Global Change Biology, 22, 1523–1531. Seto, K.C., G€ uneralp, B. & Hutyra, L.R. (2012) Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools. Proceedings of the National Academy of Sciences of the United States of America, 109, 16083–16088. Shelton, A.O., Francis, T.B., Feist, B.E., Williams, G.D., Lindquist, A. & Levin, P.S. (2016) Data from: Forty years of seagrass population stability and resilience in an urbanizing estuary. Dryad Digital Repository, http://dx. doi.org/XXXXXX. Shipman, H. (2008) Geomorphic Classification of Puget Sound Nearshore Landforms. U.S. Army Corps of Engineers, Seattle, WA, USA. Short, F.T. & Wyllie-Echeverria, S. (1996) Natural and human-induced disturbance of seagrasses. Environmental Conservation, 23, 17–27. Short, F.T., Polidoro, B., Livingstone, S.R. et al. (2011) Extinction risk assessment of the world’s seagrass species. ????, 144, 1961–1971. Simenstad, C.A. & Fresh, K.L. (1995) Influence of intertidal aquaculture on benthic communities in Pacific Northwest estuaries: scales of disturbance. Estuaries, 18, 43–70. Simenstad, C.A., Ramirez, M., Burke, J. et al. (2011) Historical Change of Puget Sound Shorelines: Puget Sound Nearshore Ecosystem Project Change Analysis. Washington Department of Fish and Wildlife, Olympia, Washington, and U.S. Army Corps of Engineers, Seattle, WA, USA. Small, C. & Nicholls, R.J. (2003) A global analysis of human settlement in coastal zones. Journal of Coastal Research, 19, 584–599. Steinberg, P.D., Brett, M.T., Bechtold, J.S., Richey, J.E., Porensky, L.M. & Smith, S.N. (2010) The influence of watershed characteristics on nitrogen export to and marine fate in Hood Canal, Washington, USA. Biogeochemistry, 106, 415–433. Stevens, A.W. & Lacy, J.R. (2012) The influence of wave energy and sediment transport on seagrass distribution. Estuaries and Coasts, 35, 92–108. Stick, K.C. (1994) Summary of 1993 Pacific herring spawning ground surveys in Washington state waters. Washington Department of Fish and Wildlife Progress Report, 311. Stick, K.C., Lindquist, A. & Lowry, D. (2014) 2012 Washington State Herring Stock Status Report. Washington Department of Fish and Wildlife. Report No. FPA 14-09. Thom, R.M., Gaeckle, J.L., Buenau, K.E., Borde, A.B., Vavrinec, J., Aston, L. & Woodruff, D.L. (2014) Eelgrass (Zostera marina L.) Restoration in Puget Sound: Development and Testing of Tools for Optimizing Site Selection. Pacific Northwest National Laboratory, United States Department of Energy, Oakridge, TN, USA. Thomson, R.E. (1994) Physical oceanography of the Strait of Georgia e Puget Sound e Juan de Fuca Strait system. Proceedings of the BC/Washington Symposium on the Marine Environment. Review of the Marine Environment and Biota of Strait of Georgia, Puget Sound and Juan de Fuca Strait (eds J. Wilson, R. Beamish, R. Aitkens & J. Bell). ?????, ????? ¶ Thomson, J.A., Burkholder, D.A., Heithaus, M.R., Fourqurean, J.W., Fraser, M.W., Statton, J. & Kendrick, G.A. (2015) Extreme temperatures, foundation species, and abrupt ecosystem change: an example from an iconic seagrass ecosystem. Global Change Biology, 21, 1463–1474. Timmerman, P. & White, R. (1997) Megahydropolis: coastal cities in the context of global environmental change. Global Environmental Change, 7, 205– 234. Trenberth, K.E., Caron, J.M., Stepaniak, D.P. & Worley, S. (2002) Evolution of El Ni~no-Southern Oscillation and global atmospheric surface temperatures. Journal of Geophysical Research. Atmospheres, 107, 4065. Trumble, R., Penttila, D., Day, D.E. & McAllister, P. (1977) Results of herring spawning ground surveys in Puget Sound, 1975 and 1976. Washington Department of Fish Progress Report No. 21. Valentine, J.F. & Heck, K.L. Jr (1999) Seagrass herbivory: evidence for the continued grazing of marine grasses. Marine Ecology Progress Series, 176, 291–302. Valle, M., Chust, G., del Campo, A., Wisz, M.S., Olsen, S.M., Garmendia, (2014) Projecting future distribution of the seagrass Zostera J.M. & Borja, A. noltii under global warming and sea level rise. Biological Conservation, 170, 74–85. Vogelmann, J.E., Howard, S.M., Yang, L., Larson, C.R., Wylie, B.K. & Van Driel, N. (2001) Completion of the 1990s national land cover data set for the conterminous United States from landsat thematic mapper data and ancillary data sources. Photogrammetric Engineering and Remote Sensing, 67, ?????–?????. Ward, E.J., Jannot, J.E., Lee, Y.-W., Ono, K., Shelton, A.O. & Thorson, J.T. (2015) Using spatiotemporal species distribution models to identify temporally evolving hotspots of species co-occurrence. Ecological Applications, 25, 2198–2209. Waycott, M., Duarte, C.M., Carruthers, T.J.B. et al. (2009) Accelerating loss of seagrasses across the globe threatens coastal ecosystems. Proceedings of the National Academy of Sciences of the United States of America, 106, 12377– 12381. WDNR, N.H.P. (2015) Puget Sound Submerged Vegetation Monitoring Program: 2010–2013 Report. WDNR, Olympia, WA, USA. Williams, S.L. & Heck, K.L. (2001) Seagrass community ecology. Marine Community Ecology (eds M.D. Bertness, S.D. Gaines & M.E. Hay), pp. 317–337. Sinauer Associates, Sunderland, MA, USA. Xian, G., Homer, C., Dewitz, J., Fry, J., Hossain, N. & Wickham, J. (2011) Change of impervious surface area between 2001 and 2006 in the conterminous United States. Photogrammetric Engineering and Remote Sensing, 77, 758–762. 13 14 Received 5 July 2016; accepted 23 September 2016 Handling Editor: A. Randall Hughes Supporting Information Additional Supporting Information may be found in the online version of this article: Appendix S1. Detailed description of statistical models and data processing methods. Appendix S2. Description of GIS methods for deriving physical and anthropogenic variables. Appendix S3. Comparative maps of eelgrass in 1981–1982 and 2011–2012 for all 14 eelgrass trend areas. Published 2016. This article is a U.S. Government work and is in the public domain in the USA. Journal of Ecology 15 Graphical Abstract The contents of this page will be used as part of the graphical abstract of html only. It will not be published as part of main. 16 17 Despite ongoing conservation concern over seagrasses world-wide, eelgrass in Puget Sound has been highly resilient to both anthropogenic and environmental change over four decades. Our work provides general methods that can be applied to understand spatial and temporal scales of change and can be used to assess hypothesized drivers of change. Author Query Form Journal: Article: JEC 12682 Dear Author, During the copy-editing of your paper, the following queries arose. Please respond to these by marking up your proofs with the necessary changes/additions. Please write your answers on the query sheet if there is insufficient space on the page proofs. Please write clearly and follow the conventions shown on the attached corrections sheet. If returning the proof by fax do not write too close to the paper's edge. Please remember that illegible mark-ups may delay publication. Many thanks for your assistance. Query reference Query 1 AUTHOR: Please confirm that given names (red) and surnames/family names (green) have been identified correctly. 2 AUTHOR: Please check the edit made in the keywords. 3 AUTHOR: ‘Supplement S1, S2 and S3’ have been changed as ‘Appendix S1, S2 and S3’ respectively, similar changes made in supporting information files also. Please check. 4 AUTHOR: Please provide an appropriate table footnote to explain the bold values in Table 1. 5 AUTHOR: Clark 2014 has been changed to Clarke 2014 so that this citation matches the Reference List. Please confirm that this is correct. 6 AUTHOR: Please provide URL for ‘Data accessibility’ section and also update in reference list. 7 AUTHOR: Please provide the publisher name, publisher location for reference Burns (1985). 8 AUTHOR: Please provide the publisher location for reference Minnesota Population Center (2011). 9 AUTHOR: Please provide the publisher location for reference NOAA (2013). 10 AUTHOR: Please check all website addresses and the functionality of the underlying links and confirm that they are correct. (Please note that it is the responsibility of the author(s) to ensure that all URLs given in this article are correct and useable.) 11 AUTHOR: Please provide the publisher name, publisher location for reference RCW Revised Code of Washington 90.58.06 23 (1971). 12 AUTHOR: Please provide the journal title for reference Short et al. (2011). 13 AUTHOR: Please provide the publisher name, publisher location for reference Thomson (1994). 14 AUTHOR: Please provide the page range for reference Vogelmann et al. (2001). Remarks 15 AUTHOR: Please take this opportunity to review any supporting information files that you submitted via ScholarOne. If you find any errors, please upload the corrected file and add a note to the PDF to confirm which file should be replaced. Also, we ask that you check and confirm the file names and descriptions for all supporting information. This will be your last opportunity to make any changes to the Supporting Information files, these files cannot be amended after publication. 16 AUTHOR: Please confirm that the Graphical abstract has no errors and is suitable for online publication. 17 AUTHOR: The journal style is to include a graphical abstract for online publication. The graphical abstract requires a figure along with blurb content for the figure. Please designate the figure to be included as part of graphical abstract. 18 AUTHOR: Please provide missing details if any for reference FWPCA (1972).