Twenty years of coastal waterbird trends suggest regional patterns of environmental pressure in British Columbia, Canada

Waterbirds are often used as indicators of ecosystem function across broad spatial and temporal scales. Resolving which species are declining and the ecological characteristics they have in common can offer insights into ecosystem changes and their underlying mechanisms. Using 20 years of citizen science data collected by the British Columbia Coastal Waterbird Survey, we examine species-specific trends in abundance of 50 species in the Salish Sea and 37 species along the outer Pacific Ocean coast that we considered to form the core wintering coastal bird community of British Columbia, Canada. Further, we explore whether ecological commonalities increase the likelihood of a species undergoing declines by testing the hypotheses that waterbird abundance trends are influenced by dietary specialization and migration distance to breeding grounds. Results suggest that most populations are stable (i.e., temporal trends are not significant) in both the Salish Sea (36 of 50 spp.) and Pacific coast (32 of 37 spp.) regions. Twelve species displayed significant decline trends in the Salish Sea, whereas two had significant increases. Along the Pacific coast, only three species displayed significant decline trends, and two significant increases. This result is corroborated by guild-specific trends indicating that waterbirds occupying the Salish Sea are faring significantly worse than those residing along the outer coastal regions, almost irrespective of dietary specialization or migration distance. Our results provide evidence that differential environmental pressures between the inner and outer coastal regions may be causing overall loss of wintering waterbirds within, or abundance shifts away from, the Salish Sea. Potential mechanisms responsible for these observed patterns are discussed, including environmental (e.g., climate) and human-induced (e.g., nutrient and chemical pollution) pressures. Collaborative, inter-disciplinary research priorities to help understand these mechanisms are suggested. Vingt ans de suivi d'oiseaux aquatiques côtiers montrent des tendances régionales de pressions environnementales en Colombie-Britannique, Canada RÉSUMÉ. Les oiseaux aquatiques sont souvent utilisés comme indicateurs du fonctionnement des écosystèmes à de larges échelles spatiales et temporelles. La détermination des espèces en diminution et des caractéristiques écologiques qu'elles ont en commun permet de se faire une idée des changements dans les écosystèmes et des mécanismes sous-jacents. Au moyen de 20 ans de données de science citoyenne collectées lors du Coastal Waterbird Survey de la Colombie-Britannique, nous avons examiné les tendances dans l'abondance de 50 espèces dans la mer des Salish et de 37 espèces plus au large de la côte de l'océan Pacifique que nous avons considérées comme formant le noyau de la communauté d'oiseaux côtiers hivernant de la Colombie-Britannique, au Canada. En outre, nous avons étudié si les traits communs écologiques augmentaient la probabilité qu'une espèce subisse une baisse en testant les hypothèses selon lesquelles les tendances de l'abondance des oiseaux aquatiques étaient influencées par la spécialisation alimentaire et la distance de migration vers les sites de reproduction. Selon nos résultats, la plupart des populations étaient stables (c.-à-d. que les tendances temporelles n'étaient pas significatives) dans la mer des Salish (36 sur 50 sp.) et le long de la côte du Pacifique (32 sur 37 sp.). Douze espèces ont montré une tendance significative de diminution dans la mer des Salish, tandis que deux ont connu des augmentations significatives. Le long de la côte du Pacifique, seules trois espèces ont affiché une tendance à la baisse significative, et deux des augmentations significatives. Ce résultat est corroboré par les tendances propres à chaque guilde indiquant que les oiseaux aquatiques qui occupent la mer des Salish se portent nettement moins bien que ceux qui se tiennent plus au large des côtes, presque indépendamment de leur spécialisation alimentaire ou de leur distance de migration. Nos résultats montrent que les pressions environnementales variables entre les parties intérieures et extérieures des régions côtières peuvent être à l'origine de la perte globale des oiseaux aquatiques hivernant dans la mer des Salish ou d'un déplacement de ces oiseaux vers d'autres régions. Nous avons examiné les possibles mécanismes responsables des tendances observées, y compris les pressions environnementales (p. ex. le climat) et humaines (p. ex. la pollution par les nutriments et les produits chimiques). Enfin, nous proposons des priorités de recherche collaborative et interdisciplinaire pour faciliter la compréhension de ces mécanismes.


INTRODUCTION
Human impacts on marine ecosystems are complex and widespread. Activities such as commercial fishing (e.g., Thrush et al. 1998), pollution (e.g., Kennish 2002), and climate change (e.g., Harley et al. 2006) affect a diverse suite of flora and fauna, leading to complex and varying consequences for marine biodiversity. Unravelling these dynamic relationships and identifying the mechanisms driving declines in marine biodiversity continue to challenge ecologists (Möllmann andDiekmann 2012, Gamfeldt et al. 2015). Comprehensive long-term monitoring programs that track ecosystem-wide trends in biodiversity and abundance are invaluable to this process because they provide information about species in decline, their commonalities, and correlations with the hypothesized responsible drivers. Waterbirds are often considered to be ideal indicators of changes in marine productivity and ecosystem function across broad spatial and temporal scales because they are typically long-lived, migratory, and higher trophic-level feeders (Ainley and Hyrenbach 2010, Vilchis et al. 2015. Trends in waterbird abundance can therefore offer insights into ecosystem changes happening locally and beyond and the conservation measures necessary to mitigate them. The coastline of British Columbia (BC), Canada, is home to diverse and plentiful marine life, including many globally important staging, molting, and wintering areas for waterbirds that migrate along the Pacific Flyway (e.g., Important Bird and Biodiversity Areas in Canada: https://www.ibacanada.com/). The Salish Sea supports 172 bird species that are dependent on the region's habitats, of which 72 are highly dependent on intertidal and marine habitats (Gaydos et al. 2008, Gaydos andPearson 2011). This coastline also supports large urban populations, specifically in the south (e.g., Greater Vancouver Region), and maintains a strong economy based on ocean-based industries, which contribute $11 billion (8%) to the provincial gross domestic product (Stocks and Vandeborne 2017). Over the past two decades, the coastal human population of BC has increased by approximately 15,000 people/yr (1.21%; BC Stats 2020), which is coupled with the expansion of ocean-based industries, including (in order of importance): recreation, transportation, and seafood (Stocks and Vandeborne 2017). Along with the effects of global climate change on ocean productivity (e.g., Moore et al. 2018), BC's coastal and marine ecosystems are experiencing increasing pressures. Waterbird species that overwinter along BC's coastline have similarly experienced significant changes in abundance in recent decades (Anderson et al. 2009, Bower 2009, Crewe et al. 2012, the causes of which remain largely unresolved. In 1999, Birds Canada (formerly Bird Studies Canada) began the British Columbia Coastal Waterbird Survey (BCCWS) to collect baseline information on the status of waterbirds along the BC coast so that the impacts of natural and human-induced environmental change could be assessed through time.
Here, we provide an update to the species-specific trend analysis done by Crewe et al. (2012) that marks 20 years of data collection by the BCCWS. In addition to analyzing abundance trends of species occupying the Canadian extent of the Salish Sea (inner coastal), we also assess trends from the outer coastal region of the Pacific Ocean (see Methods: Study area) to gain insight into the differential pressures facing birds using the two regions. Further, our analysis assesses the ecological characteristics associated with species-and guild-specific trends in the Salish Sea and along the outer Pacific coast. Specifically, we hypothesized that abundance trends in waterbirds are influenced by dietary specialization and migration distance. First, we predicted that abundance declines will be more prevalent in higher tropic feeders because this guild is generally more responsive to shifts in the availability and quality of prey (e.g., Ballance et al. 1997, Hyrenbach and Veit 2003, Ainley and Hyrenbach 2010, Vilchis et al. 2015. Second, we predicted that abundance declines will be more pronounced in migrants because these populations are able to shift their range in response to resource availability more readily than local breeders can (e.g., Burger and Gochfeld 1991, Marks and Redmond 1994, Willie et al. 2020). Finally, we discuss potential mechanisms to stimulate future research and to inform conservation planning.

Study area
The survey area extends from the north of Graham Island (54°0 2'36.3" N, 131°54'38.3" W) to the southern tip of Vancouver Island, Canada (48°19'04.3" N, 123°33'28.6" W; Fig. 1). There are 326 survey routes in total, which cover a wide variety of habitat types, including: rocky shores, sandy beaches, salt marshes, mudflats, fast-flowing channels, and other inshore waters. The BC coast experiences the mildest winters in all of Canada, with temperatures rarely dropping below freezing (https://www. welcomebc.ca/Choose-B-C/Explore-British-Columbia/Climate-of-B-C). As a result, offshore and most inshore waters never freeze completely, making the coastline and freshwater inflows attractive to many wintering waterbirds that breed inland and further north.   1A) and are comparable to those used by Crewe et al. (2012) to permit direct trend comparison. This region faces disproportionately high levels of environmental pressure caused by the region's large and rapidly growing human population and ocean-based industries (Georgia Strait Alliance, About the Strait: https://georgiastrait.org/issues/about-thestrait-2/). The Pacific coast survey routes are those along the west coast of Vancouver Island and routes within the Johnstone and Hecate Straits (Fig. 1B). These coastal regions are sparsely populated by rural coastal communities and experience substantially less commercial and industrial pressure (Stocks amd Vandeborne 2017).

Data collection
From 1999-2019, surveys were conducted by volunteers using a standardized protocol and data collection sheets (Birds Canada 2018). Shore-based counts were completed monthly on or near the second Sunday of each month from September to April. Surveys were completed within approximately 2 h of high tide to maximize the opportunity for close observation. All waterbirds observed to a distance of 1 km from the high tide line were counted, except those that flew through without stopping. In the case of larger flocks, numbers were estimated by counting individuals and species in groups and scaling up (see Birds Canada 2013). Surveyors varied within and among years, and because skill level was not recorded, observer skill was assumed to vary randomly and not systematically over time. Data were entered through a customized online data entry system available on the Birds Canada website, NatureCounts (https://www.birdscanada. org/birdmon/default/main.jsp). Observations were processed using the eBird data filters to flag rare species and high counts during observer data entry (Cornell Laboratory help center: https://support.ebird.org/en/support/solutions/articles/48000795278the-ebird-review-process), and records were manually reviewed for form accuracy.

Analysis metric
The basic statistical unit for all analyses was the mean count of each species on a survey route across months within the speciesspecific survey window. Species-specific survey windows were assigned based on local knowledge of the months in the nonbreeding season that each species is most prevalent (Appendix 1). For example, the mean route-level count of Canada Goose (Branta canadensis) in a given year was calculated based on the sum of monthly counts in December, January, and February divided by the total number of months (N = 3). Only routes surveyed a minimum of 5 years and with 100% monthly coverage within the species-specific survey window were included in the analyses (e.g., December-February for most wintering birds). Because our aim was to focus on species that are persistent, abundant, and biologically associated with the study area, we excluded rarely detected species from our analyses (i.e., species with > 85% zero counts across all routes and years, species that had a mean abundance per year < 10, and species that were detected in fewer than one-half of the survey years). Counts of Greater Scaup (Aythya marila) and Lesser Scaup (A. affinis) were combined for the analysis (Scaup spp.) because of model convergence issues when they were run independently, which were resolved when counts were combined. In general, A. marila is much more numerous than A. affinis along BC coastlines in winter (e.g., Badzinski et al. 2006). The resulting data set included an ecologically and phylogenetically diverse suite of 50 species in the Salish Sea and 37 species along the outer Pacific coast that we considered the core wintering coastal waterbird community of the BCCWS (Table 1). Based on Davidson et al. (2015) and expert review, we categorized migration distance for each species as: (1) local, i.e., regularly breeding in at least small numbers within the study area catchment, (2) short distance, i.e., regularly breeding in at least small numbers in the interior of BC, adjacent northwestern United States, or the southeast Alaskan coast, or (3) long-distance migrant, i.e., breeding in the northern Boreal and Arctic regions of North America. Species were also classified based on their dietary specialization following Bower (2009)

Trends analysis
Twenty-year trends in counts were estimated independently for each species and region using a Bayesian framework with integrated nested Laplace approximation (INLA; Rue et al. 2009) in R (version 3.5.3; R Core Team 2014). Model comparison using leave-one-out cross-validation in INLA was used to select between a Poisson and negative binomial distribution (Held et al. 2010). Count data were fit using a log-linear regression model for each species that included a continuous effect for year (t) to estimate the change in population size on a given route (j) over time (i.e., linear trend), a hierarchical first-order autoregressive term to model the temporal autocorrelation structure of residuals among years (γ), and an independent and identically distributed hierarchical term to account for random variation in counts among routes (η).
log(μ tj ) = α + β 1 * year tj + γ t + η j (1) The number of routes detecting a species each year was included as an offset to account for variable count effort and distributional differences across the study area. The continuous year coefficients and credible intervals were transformed into constant rates of population change using 100 × (exp(estimate) − 1). Annual indices of population size for each species were derived from the model for the full 20-yr monitoring period. The analytical approach applied here follows that of Crewe et al. (2012) to enable more direct comparison of trends, with some adjustment to improve analytical rigor. Specifically, both methods used a log-linear regression model with continuous year and categorical route-level effects. Our model differed in that we deployed our analysis in a Bayesian framework and included additional hierarchical terms to account for residual variation and an offset to account for variation in sampling effort. The inclusion of a hierarchical term here better accounts for temporal autocorrelation structure of residuals among years, thus allowing the model to better meet the assumption of sampling independence. Ignoring this dependency can make the estimates of standard error too small. Further, the addition of an offset better ensures that changes in effort are accounted for during the analysis of trends. Table 1. Annual population trends for 50 coastal waterbird species overwintering in the Salish Sea (inner coastal) and 37 species in the outer coast regions of the Pacific Ocean in British Columbia, Canada, from 1999and 1999-2011(Crewe et al. 2012). Bold font indicates species for which a statistically significant annual trend (%/yr) is implied, i.e., lower credibility intervals (LCI) and upper credibility intervals (UCI) do not contain zero. The guild analysis was done using the same framework as previously described for species-specific trends, fitting a separate model for migration distances and dietary categorizations for each region. An additional independent and identically distributed hierarchical term was included to account for random variation in counts among species. The number of routes surveyed within a given year was included as an offset to account for differences in sampling effort.

DISCUSSION
Our results highlight several themes, including species-specific trends of importance to management agencies, 12-yr vs. 20-yr trajectory comparisons within the Canadian Salish Sea, and regional differences in guild-specific abundance patterns, which can better inform future targeted research and conservation initiatives. . These species are all local breeders or short-distance migrants. In summary, 7 of the 12 negatively trending taxa are benthivorous on their Salish Sea wintering grounds; Common Loon (piscivorous) and Scaup spp. (omnivorous) also feed in the benthic environment in winter (Evers et al. 2020, Kessel et al. 2020). The guild-specific analysis further corroborates the observed species-specific trends found in benthivores in the Salish Sea (Fig. 2). Notwithstanding the fact that many of the benthivorous taxa show declining trends in other parts of their continental and global ranges (Billerman et al. 2020, IUCN 2020, these results suggest that benthic habitat quality in the Salish Sea may deserve further targeted research attention (see Discussion: Guild-specific trends). The only two species to show increases in the Salish Sea were a lower abundance generalist (omnivorous) and a population augmented by previous introductions (Canada Goose;e.g., Isaac-Renton et al. 2010). The piscivore guild has previously been shown to be at highest risk of decline within the Salish Sea, linked to declines in forage fish (Vilchis et al. 2015), so it is interesting to note the increases in these two piscivores along the Pacific coast.

Species-specific trends
Comparison of our 20-year Salish Sea results with those from Crewe et al. (2012) also show some notable positive differences in species-specific abundance trends, which can be expected given that waterbird populations fluctuate quite widely over decadal time frames (e.g., Wilkins and Otto 2003). Ten of twenty species  (Crewe et al. 2012). However, along the Pacific coast, this subspecies had a 20-yr declining trend, which suggests that differential pressures affect this species within the Canadian Salish Sea and along the Pacific coast. The BCCWS is considered to have the most robust data for tracking abundance trends in this taxon because of the survey's standardized data collection procedures and robust statistical design (COSEWIC 2008).
Twenty-two species sampled by this survey occur in globally significant numbers along the BC coast (BirdLife International, global IBA criteria: http://datazone.birdlife.org/site/ibacritglob; Table 1); seven of these species experienced 20-yr declines in the Salish Sea, and just one, Red-necked Grebe (Podiceps grisegena), a piscivore, showed an increasing trend on the Pacific coast only. Several of these abundance trends are noteworthy. Declining sea ducks of global significance (e.g., Surf Scoter, White-winged Scoter) are part of the benthivore community considered above. A significant decline in Trumpeter Swan in the Salish Sea (−8.24; −12.01, −4.40) differs from regional trends reported elsewhere, which have noted abundance gains in recent decades (e.g., the Pacific Coast population increased 5.5%/yr from 1968-2005 and 1.5%/yr from 2005-2010; Northwest Swan Conservation Association: https://nwswans.org/trumpeter-swans/). The BCCWS is not well designed to sample this swan (Crewe et al. 2012), which uses southern coastal agricultural lands for feeding by day and nearshore waters for roosting. However, these results may indicate a change in trend trajectory in the Canadian Salish Sea after a period of conservation-driven increase and stabilization noted elsewhere (Rees et al. 2019). Continuing loss of winter habitat is identified as a major threat to this species (Trumpeter Swam Society, Pacific coast population: https://www.trumpeterswansociety. org/what-we-do/your-society-at-work/pacific-coast-population.html) as agricultural land use shifts toward more greenhouse and berry crops instead of field crops that provide wintering waterfowl habitat (British Columbia Ministry of Agriculture 2017). The significant decline of Dunlin (−9.70; −13.06, −6.01) wintering in the Salish Sea flags a potential conservation issue for local management agencies given that up to 20% of the Pacific Dunlin (Calidris alpina pacifica) population winters here (Morrison et al. 2006; Important Bird and Biodiversity Areas in Canada: https:// www.ibacanada.com/). Dunlin (all subspecies) is also identified as a high priority candidate species for assessment by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC), which comprises taxa not yet assessed but where new information suggests they may be at risk of extinction or extirpation (COSEWIC 2020). Black Turnstone, a rocky shore specialist (benthivorous), and Pacific Loon (piscivorous) both showed 20-yr declines in the Salish Sea but stable trends on the outer Pacific coast, contributing to the contrasting pattern between benthic feeders and piscivores on inner and outer coasts. One further taxon to draw attention to is Thayer's Gull (Larus glaucoides thayeri); a large proportion of the global population of this Canadian Arctic breeder winters in the Salish Sea and showed a narrowly nonsignificant declining trend, after both 12 years (Crewe et al. 2012) and 20 years (this study).
Although species-specific trends are informative and, in some instances, likely represent true trends in total population abundance, it is inherently difficult to interpret local changes in the abundance of longer distance migratory waterbirds. This difficulty is in part because local counts do not constitute an exhaustive inventory of the population, nor do they cover the species' entire range. Specifically, abundance trends in one part of a species' range cannot discern if local fluctuations are driven by shifting distributions or changing population numbers (Hyrenbach and Veit 2003). For example, the decline in Western Grebe in the Salish Sea is ongoing but is explained at least partly by a southward nonbreeding range shift of nearly 900 km during the past three decades (Wilson et al. 2013). To discern better if trends are a result of range-wide population change or redistribution, the Migratory Shorebird Project was established in 2013 to sample nonbreeding populations of shorebirds along the Pacific Coast of the Americas (Reiter et al. 2020). This hemispheric, hypothesis-driven monitoring framework, to which the BCCWS is a contributing partner, will provide the context to better assess local causes of declines in shorebird species of conservation concern.

Guild-specific trends
Within the Salish Sea, our results support the prediction that abundance declines are more prevalent in higher trophic-level feeders, with significant declines experienced by all guilds except herbivores (Fig. 2). Our findings align with those of Vilchis et al. (2015) for piscivores and guilds without local breeding populations and, for the first time, highlight benthivores as the guild showing steepest 20-yr declines. Putative mechanisms behind these changes in abundance could include: redistribution in response to prey (demonstrated for Western Grebe; Wilson et al. 2013), redistribution in response to predators (e.g., Middleton et al. [2018] found that dive-feeding birds in the Salish Sea, including scoters, moved away from shore in response to the threat from Bald Eagle Haliaeetus leucogaster), factors operating in other parts of the ranges of longer distance migrants (e.g., climate-driven changes in aquatic invertebrate prey on boreal or arctic breeding grounds; e.g., Corcoran et al. 2009, Arzel et al. 2020, and changes to benthic and epibenthic food sources in the Salish Sea. Duck species that have benefited from ongoing conservation investment in > 80,000 km² of wetlands and adjacent lands in Canada, under the North American Waterfowl Management Plan (NABCI-Canada 2019), showed stable or increasing species-specific trends in the Salish Sea, including American Wigeon (Mareca americana), Northern Pintail (Anas acuta), and Green-winged Teal (Anas crecca). At the guild level, these herbivores were the only group that showed a (narrowly) stable trend (Fig. 2).
When combined, results from the Salish Sea and Pacific coasts did not lend support to our guild-level hypotheses. Specifically, there was no evidence that abundance declines were more prevalent in higher trophic-level feeders along the Pacific coast, where piscivores (8.11; 4.08, 12.30) experienced significant abundance gains (Fig. 2). The prediction that longer distance migrants would be more susceptible to declines than local breeders was also unsupported; all three guilds (local breeders, short-distance and long-distance migrants) experienced abundance gains along the Pacific coast (Fig. 2). These guildspecific results provide evidence of consistently contrasting patterns within and outside the Salish Sea, suggesting that differential pressures may be causing overall loss of wintering waterbirds within, or abundance shifts away from, the Salish Sea, a pattern already demonstrated for Western Grebe (Wilson et al. 2013 Ward et al. (2015) showed evidence of local increases in 14 of 18 species in Puget Sound, especially those with local breeding populations, and some commonalities in declining species with our 20-yr results (e.g., White-winged Scoter, Western Grebe), raising the possibility that localized active restoration is a mechanism to consider in understanding changes in coastal waterbird abundances in other regions of the Salish Sea.

CONCLUSIONS
Results from this study further demonstrate that citizen science data provide robust insight into regional abundance trends of coastal waterbirds and the commonalities that influence their susceptibility to environmental pressures (Bower 2009, Vilchis et al. 2015. These abundance trends are important for evaluating the status of vulnerable and ecologically significant populations (e.g., COSEWIC, Pacific America's Shorebird Conservation Strategy, Important Bird and Biodiversity Areas), and establishing baseline statistics against which policy changes can be evaluated (e.g., oil spill monitoring; O'Hara et al. 2009). Longterm citizen science monitoring of coastal ecosystems provides a foundation from which to tease apart whether local population fluctuations are a result of true changing abundance or shifts in species distributions over time (Vilchis et al. 2015, Reiter et al. 2020) and to resolve the mechanisms responsible (e.g., Bianchini et al. 2020). Organisms living in the Salish Sea are exposed to differential environmental and human-induced pressures than those along the outer Pacific coast. For example, nutrient pollution from rivers, nonpoint source runoff, and wastewater sources have been recognized as primary threats to the ecological health of the inner waters of the Salish Sea (Khangaonkar et al. 2019). The effects of nutrient pollution on coastal ecosystems are exacerbated when coupled with projected future climate stressors, the negative effects of chemical pollution from industry and municipalities, oil spills, invasive species, and increased pressures from shipping, fishing, recreation, and shoreline development, all of which are likely to affect the Salish Sea to a greater extent than the outer coastal waters of the Pacific Ocean (e.g., Stocks and Vandeborne 2017

LITERATURE CITED
Ainley, D. G., and K. D. Hyrenbach. 2010. Top-down and bottom-up factors affecting seabird population trends in the California current system