Informing sea otter reintroduction through habitat and human interaction assessment

Sea otters Enhydra lutris have been absent from Oregon, USA, following their extirpation over a century ago. Stakeholder groups and native tribes are advocating for reintroduction to restore historic populations. We investigated the potential for successful reintroduction by: (1) estimating expected equilibrium sea otter densities as a function of habitat variables to assess sea otter habitat in Oregon; and (2) spatially relating areas of high expected densities to human activities (e.g. fisheries, recreation, vessel activity, protected areas) to anticipate potential disturbance or fishery resource competition. We estimated that 4538 (1742−8976; 95% CI) sea otters could exist in Oregon, with higher expected abundance (N = 1551) and densities (x– = 2.45 km−2) within the southern region. Most core habitat areas (97%), representing clusters of high expected densities, overlapped with some form of human activity. While commercial shipping and tow lanes overlapped little (1%) with core habitat areas, recreational activities (58%) and fisheries (76%) had a higher degree of overlap, posing higher disturbance risk. We anticipate higher resource competition potential with the commercial red sea urchin fishery (67% of harvest areas) than the commercial Dungeness crab fishery (9% of high-catch crabbing grounds). Our study presents the first published carrying capacity estimate for sea otters in Oregon and can provide population recovery targets, focus attention on ecological and socioeconomic considerations, and help to inform a recovery plan for a resident sea otter population. Our findings suggest current available habitat may be sufficient to support a sea otter population, but resource managers may need to further investigate and consider whether current human activities might conflict with reestablishment in Oregon, if plans for a reintroduction continue.


INTRODUCTION
Throughout history, humans have exploited wildlife populations, and these activities may partially explain Earth's sixth mass extinction (Barnosky et al. 2011, Dirzo et al. 2014. Well-known examples of exploitive practices include near collapses in global whale populations due to international whaling (Clapham et al. 1999) and over-fished cod stocks along the US northeast coast (Hutchings & Myers 1994, Myers et al. 1997. As biodiversity loss has accelerated, the importance of species diversity to ecosystem function, resilience, and services has become apparent (Cardinale et al. 2002, Elmqvist et al. 2003, Downing & Leibold 2010. Biodiversity loss thus represents a prominent threat to environmental ABSTRACT: Sea otters Enhydra lutris have been absent from Oregon, USA, following their extirpation over a century ago. Stakeholder groups and native tribes are advocating for reintroduction to restore historic populations. We investigated the potential for successful reintroduction by: (1) estimating expected equilibrium sea otter densities as a function of habitat variables to assess sea otter habitat in Oregon; and (2) spatially relating areas of high expected densities to human activities (e.g. fisheries, recreation, vessel activity, protected areas) to anticipate potential disturbance or fishery resource competition. We estimated that 4538 (1742−8976; 95% CI) sea otters could exist in Oregon, with higher expected abundance (N = 1551) and densities (x -= 2.45 km −2 ) within the southern region. Most core habitat areas (97%), representing clusters of high expected densities, overlapped with some form of human activity. While commercial shipping and tow lanes overlapped little (1%) with core habitat areas, recreational activities (58%) and fisheries (76%) had a higher degree of overlap, posing higher disturbance risk. We anticipate higher resource competition potential with the commercial red sea urchin fishery (67% of harvest areas) than the commercial Dungeness crab fishery (9% of high-catch crabbing grounds). Our study presents the first published carrying capacity estimate for sea otters in Oregon and can provide population recovery targets, focus attention on ecological and socioeconomic considerations, and help to inform a recovery plan for a resident sea otter population. Our findings suggest current available habitat may be sufficient to support a sea otter population, but resource managers may need to further investigate and consider whether current human activities might conflict with reestablishment in Oregon, if plans for a reintroduction continue.
KEY WORDS: Sea otter · Enhydra lutris · Oregon · Reintroduction · Habitat · Carrying capacity · Fisheries · Conservation OPEN PEN ACCESS CCESS sustainability. Recovery of at-risk species, particularly species vital to ecosystem function, can help maintain ecosystem integrity (Soulé et al. 2003). Environmental managers enlist a range of strategies to facilitate at-risk species recovery, such as establishing protected areas, moving threatened populations into captivity, or conducting reintroductions and reinforcements (Briggs 2009).
Sea otters Enhydra lutris were once distributed along most North Pacific Ocean coastlines from Japan to Baja California, Mexico, but were extirpated from most of their historic range during the peak of the maritime fur trade from the mid-1700s to mid-1800s (Kenyon 1969). Recovery occurred slowly over the first half of the 20th century; however, a significant boost to recovery occurred in the late 1960s when resource managers translocated sea otters from Amchitka Island and Prince William Sound, Alaska, to Southeast Alaska, British Columbia, Washington, and Oregon. Most of these translocation efforts were successful, and populations in these areas are now abundant and thriving. A notable exception was the translocation effort to Oregon, where the founding population gradually declined from reintroduction in 1970−1971 (N = 93 otters) to 1981, when only 1 otter was observed during routine surveys, after which the population was expected to disappear . No consensus exists for the cause of failure of the Oregon population, but several hypothesized factors include lack of appropriate habitat or prey, human disturbance, or sea otter emigration due to homing behavior . Presently, stakeholder groups and native tribes are advocating for a second attempt at sea otter reintroduction to Oregon, arguing that this action could achieve several objectives, including: (1) aiding recovery efforts for a species of conservation concern; (2) restoring coastal food web structure and function; (3) provisioning ecosystem services, including economic or intrinsic/recreational benefits; and (4) restoring lost cultural and tribal traditions and ecological connections.
Species reintroductions represent an important tool for managers charged with recovering at-risk species (Clark & Westrum 1989, Seddon et al. 2007). There have been several notable cases where translocations have contributed to species recovery, including the previously mentioned sea otter translocations across the North Pacific Ocean, red deer Cervus elaphus to central Portugal (Valente et al. 2017), and gray wolves Canis lupus to Yellowstone National Park, USA (Smith & Guernsey 2002, Ripple & Beschta 2003. Species reintroductions are also risky because uncertainty surrounds whether the species will reestablish in their release area (Sarrazin & Barbault 1996). Habitat suitability assessments can reduce this uncertainty by identifying areas of unoccupied habitats that are likely to sustain the introduced species and foster population growth over time (Cheyne 2006). Predator populations and population growth are often limited by prey availability. Therefore, habitat models can be used to identify and predict areas of unoccupied habitats that are likely to contain adequate prey to sustain the predator.
Sea otters have been absent from Oregon waters for more than 100 yr, during which time nearshore habitats have experienced substantial change. A variety of human activities now occur along the Oregon coast, including fisheries, recreation, and shipping (Norman et al. 2007, LaFranchi & Daugherty 2011, which could disturb sea otters or make habitats less hospitable. At present there has been no systematic assessment of the potential for sea otters to reestablish in Oregon. Using predictive models to evaluate the potential for sea otter recovery in different sites can help fill this knowledge gap; such models re quire an understanding of habitat features that facilitate effective sea otter foraging and knowledge of current nearshore habitats in Oregon. Luckily, sea otter habitat-use patterns and foraging activities are well documented in other regions (e.g. Ostfeld 1982, Laidre et al. 2009, Hughes et al. 2013, Lafferty & Tinker 2014, and this information can be leveraged for consideration of the potential for sea otter recovery in Oregon. Sea otters are typically found within shallow and intertidal rocky habitats, where they forage for benthic macroinvertebrates such as sea urchins, sea snails, bivalves, and crabs (Estes et al. 1982, Ostfeld 1982, Laidre & Jameson 2006, Newsome et al. 2009). Canopy-forming and understory macroalgae (i.e. kelp, seaweed) also provide important habitat for prey species as well as protected resting habitat for sea otters (Estes & Palmisano 1974, Estes et al. 1982, Nicholson et al. 2018). In addition to rocky and kelpdominated habitats, sea otters also use soft-sediment habitats on the outer coast and within estuaries (Riedman & Estes 1990, Hughes et al. 2013, Hale et al. 2019. Sea otters have been infrequently observed hauling out on shore to rest, groom, and forage. This behavior appears to be more common on marshes within Elkhorn Slough, California, and on sand and mud bars in Alaska, and is much less observed along outer coastal shorelines (Kenyon 1969, Garshelis & Garshelis 1984, Faurot 1985, Green & Brueggeman 1991, Eby et al. 2017. The seaward distribution of sea otters is limited by their maximum diving capacity of 100 m depth (Bodkin et al. 2004, Thometz et al. 2016, although most dives occur within 40 m depth. Within their nearshore distribution, sea otter densities have a non-linear relationship with depth, where densities peak around a model depth of 15 m, and gradually decline as depth increases or decreases (Tinker et al. 2017). The slope and width of the continental shelf can dictate how dense or spread out populations are across space (Tinker et al. 2021). Relative to other marine predators, sea otters have extremely high metabolisms and almost no capacity for energy storage in fat tissue, and thus require anywhere from 25 to 30% of their own body weight in food every day (Costa & Kooyman 1982, Riedman & Estes 1990). Their extreme dependency on high energy prey means identifying high quality foraging habitat with in their depth limits is imperative to facilitating successful reintroduction.
Population growth and survival are 2 metrics used to assess the performance and potential success of reintroduction efforts and species reestablishment (IUCN/SSC 2013). Both lethal (i.e. mortality) and non-lethal (e.g. human disturbance, resource competition) stressors may reduce or hinder population growth and survival. Some causes of sea otter mortality (e.g. white shark attacks, cardiac arrest, infectious diseases, fishing gear entanglements, etc.) are well studied and directly limit population growth , Kreuder et al. 2003). Yet, the population-level consequences of non-lethal stressors, such as human disturbance, are more difficult to assess. Conceptually, this understanding requires evidence that (1) exposure to a stressor causes a be havioral or physiological response, (2) those responses alter internal health (e.g. homeostasis), (3) the internal health alterations influence individual vital rates (e.g. survival, fecundity, growth), and (4) a sig nificant number of individuals experience these impacts to vital rates resulting in population-level effects (National Academies of Sciences, Engineering, and Medicine 2017, Pirotta et al. 2018). As with many other marine mammals that are sensitive to human disturbance (Williams et al. 2006, Tyack 2008, sea otters exhibit physiological and behavioral responses to disturbance: for example, recreational boating (e.g. kayaks, dive boats, jet skis) can cause sea otters to increase their activity and spend less time resting, with implications for their metabolic costs (Curland 1997, Barrett 2019. To meet their metabolic de mands, sea otters spend most of their daily time budget foraging and resting (Yeates et al. 2007, Thometz et al. 2014. Therefore, any deviation from these behavioral states impacts their internal health and could be energetically costly. It is unclear whether, and to what degree, disturbance-induced behavior and physiological responses in sea otters are great enough to produce population-level consequences. Regardless, human disturbance has been, and continues to be, a concern for sea otter survival and conservation (US Fish and Wildlife Service 2003) and should be accounted for when deciding if and where sea otters should be reintroduced in Oregon.
Fisheries add a further complication due to concerns regarding competition and ecosystem impacts of sea otter foraging on certain shellfish species important to fisheries (Johnson 1982). Sea otters exhibit strong top-down pressures by reducing prey densities and size via predation (Estes et al. 1978, Estes & Duggins 1995, and sea otter-driven reductions in fishery-dependent prey species have been documented (Garshelis & Garshelis 1984, Garshelis et al. 1986, Larson et al. 2013, Carswell et al. 2015. Importantly, a network of 5 no-take marine reserves was established along the Oregon coast in 2013; this reserve network restricts human activity and could alleviate or prevent potential disturbance to sea otters and resource competition with fisheries, in the event of a sea otter reintroduction. Here we summarize multiple data sets and conduct analyses aimed at informing management decisions related to sea otter reintroduction to Oregon. Our study objectives are (1) to assess habitat presence and quality along the Oregon coast; and (2) to determine the potential for recovering sea otter populations to spatially overlap with select human activities that might cause resource competition or disturbance to sea otters. We expect that the results of our study will help managers assess the feasibility for a successful sea otter reintroduction to Oregon and identify potential next steps in the process.

Study area
The study area includes all nearshore coastal waters in Oregon, USA, from the Columbia River in the north to the Oregon−California state border in the south. The Oregon coastline is comprised of alternating sandy beaches and complex rocky habitats, with several bays and estuaries. The shallow, gradual-sloping continental shelf extends 17 to 74 km from the shoreline and is comprised of hard and soft benthic substrates (Kulm & Fowler 1974). The outer coast supports several macroinvertebrate prey items for sea otters including urchins Strongylocentrotus spp., abalone Haliotis spp., Dungeness crab Metacarcinus magister, and razor clams Siliqua patual. A number of small and large coastal estuaries also support invertebrate prey for sea otters, including bay clams (Tresus, Saxidomus, Leukoma, Mya spp.) and various crab species (ODFW 2006). Kelp canopies along the outer coast are primarily composed of bull kelp Ne reocystis luetkeana and occur in rocky habitats (Mackey 2006, Springer et al. 2007) along the southern coastline. Eelgrass Zostera spp. is the dominant vegetation in estuaries (Sherman & DeBruyckere 2018). Both kelp forests and eelgrass beds provide habitat for important sea otter prey species as well as resting habitat for sea otters.

Habitat-based population model
To investigate the presence and quality of sea otter habitat in Oregon, we adapted and applied a recently developed model of habitat-specific population potential for sea otters in California (hereafter referred to as the CA model; Tinker et al. 2021). The spatial proximity and overall similarity of coastal habitats in Oregon and California suggested that results of the CA model can be reasonably extrapolated to Oregon. In brief, Bayesian methods were used to fit a state-space model of density-dependent population growth, in which local carrying capacity (K ) was predicted as a function of a suite of local habitat features and environmental variables (henceforth, habitat variables) from 0 to 60 m depth. Habitat -density relationships (henceforth, parameters; Table 1) were estimated by fitting the CA model to a time series of annual survey counts of sea otters at known geographic locations, collected using shore-based and aerial surveys (from 1983 to 2017, except 2011), and augmented by cause-of-death data from stranded animals (Tinker et al. 2021). Using the joint posterior distributions from the CA model (Table 1), expected density at K can then be projected at the scale of a 100 m spatial grid, based on local habitat characteristics that have been summarized over the same spatial grid. In addition to mean point estimates, model projections include the combined uncertainty associated with unexplained environmental and demographic variation, as well as parameter uncertainty. Here, we applied the parameters estimated from the CA model to spatial data layers of the same suite of habitat variables in Oregon to project localized sea otter densities and abundance at carrying capacity within the study area. The CA model parameters were applied to Oregon habitat variables and used to project sea otter densities in the same manner as the CA model, with identical variables, coefficients, and functions. Further details on the CA model design, development, and Bayesian methods are presented in Tinker et al. (2021).

Habitat variable data layers
We obtained spatial layers for each Oregon habitat variable from publicly available sources (Table 2) and converted all layers to a 100 m grid using standard lin- 2.9681 of kelp cover on log-K α DSR effect of deviations from mean 0.1816 0.0917 0.0006 0.3592 slope on log-K, linear response α DSR2 effect of deviations from mean 0.2051 0.0637 0.0787 0.3283 slope on log-K, quadratic response α OFSH effect of increasing distance −0.6058 0.1713 −0.9334 −0.2618 from shore beyond 1 km (i.e. 'far offshore effect') on log-K α NPP effect of increasing net primary 0.5537 0.1305 0.3002 0.8117 production on log-K σ K magnitude of random variation 0.9343 0.2769 0.4800 1.5610 in log-K among regions Table 1. Parameters estimated from the Bayesian state-space habitat model for sea otters in California and applied in the Oregon model, including a description of each parameter, and the mean (x -), standard deviation (SD), and 95% confidence interval (CI) of the fitted posterior distribution. Parameters were estimated in 2019 using sea otter survey data from 1983 to 2017, except 2011.  1990, 1969−1999, and 2010, and we used these layers to calculate proportional kelp cover (PK) for each grid cell. Benthic substrate was classified (i.e. hard, mixed, soft) following seafloor descriptions of Greene et al. (1999). To quantify the proportional cover of hard substrate (PR) at each grid cell, we reclassified mixed substrate to hard because we observed a high degree of overlap between kelp canopies and mixed substrate, suggesting mixed substrate may functionally act as hard. We estimated net primary productivity (NPP) using an index for mean monthly NPP, following methods identical to those of Tinker et al. (2021). NPP data derived from a chlorophyll-based Vertically Generalized Production Model and represented temperature-dependent, chlorophyll-specific photosynthesis (Behrenfeld & Fal kowski 1997). We filled in missing nearshore cells using k-d tree, a nearest neighbor interpolation method (Bhatia & Vandana 2010), to calculate an average NPP value based on the 5 nearest cell values. Sea otter densities within estuaries can differ from soft sediment habitats on the outer coast (Silliman et al. 2018), so we identified estuary habitats using a categorical switch variable 'EST' (where EST = 1 for areas within estuaries, 0 for outer coast). We only in cluded estuary reaches classified as water as potential sea otter habitat, and we did not include the Columbia River in the predictions as it is unclear if this large estuary will support high otter densities. Several land polygons disagreed on shoreline position, so we merged a rocky and sandy shoreline layer to create a more precise shoreline and land layer. We conducted all spatial analyses and interpolations in ESRI's Ar-cGIS v10.6.1. We incorporated depth and distance-to-shore effects following methods of Tinker et al. (2021). Because there is a strong, non-linear relationship between log distance-to-shore (Euclidean) and depth (D g ) at any grid cell (g), we de trended distance-to-shore (DS g ) values using the following equation: The values in the least-squares equation were estimated using maximum likelihood methods and fit to data for the California coast (Tinker et al. 2021); very similar values were obtained from a similar analysis in Oregon, but we use the California values so as to retain the same habitat -density relationship parameters. The resulting distance to shore residuals (DSR g ) are independent of depth and effectively provide an index of benthic slope: positive values correspond to areas where distance to shore is greater than average relative to depth (shallow slope), and negative values represent areas where distance to shore is lower than average relative to depth (steeper slope). In Oregon, 2 reefs (Orford and Blanco Reef) have offshore island clusters that cause the seafloor to decrease in depth as distance-to-shore increases, complicating the relationship between depth and distance to shore. To account for this, we calculated distanceto-shore from these islands to appropriately assign slope effects with in these reefs. We excluded any islands outside these reefs.
Parts of the Oregon continental slope extend far offshore, where shallow depths would theoretically be accessible to sea otters, but sea otters have not been observed to regularly use these areas in California (Tinker et al. 2021). Accordingly, both the CA (Tinker et al. 2021) and Oregon models include this additional variable to allow for an offshore effect (OFSH) that can mediate predicted densities further offshore.
This offshore variable has no effect within 1 km of shoreline, but can have increasingly large effects for areas > 5 km offshore.  To account for the non-linear relationship between otter densities and depth, we included the following depth function and variables from the CA model (Tinker et al. 2021): where D* represents the modal depth and β 1 and β 2 control the rates at which density changes as depth varies inshore and offshore (respectively) of this modal depth. Lastly, to account for the fact that equilibrium sea otter densities reflect the quality of habitat available to individual sea otters within their home ranges, not just at a single point in space, we applied a 4 km moving average smoothing window to all habitat variables, following Tinker et al. (2021). For each sequential 1 m isobath, habitat variables were averaged across all cells within a 4 km smoothing window (i.e. the smoothed cell values for each habitat variable were specific to depth). The width of the smoothing window (4 km) was based on observed sea otter core home range size (Ralls et al. 1995, Tarjan & Tinker 2016).

Projecting carrying capacity
To predict otter densities (independents [i.e. adults] km −2 , excluding dependent pups) at carrying capacity on the outer coast of Oregon, we solved the following equation: (4) Each grid cell (g) was assigned an expected otter density at carrying capacity (K g ) as a function of the mean log otter density in outer coast soft sediment habitat (intercept κ s ; Table 1), the above-described net effect of habitat variables (H j,g corresponds to EST, PK, PR, DSR, DSR 2 , NPP and OFSH; see Section 2.3 and Table 1), the non-linear depth function (Eq. 3), and a random effect (ζ g |p ) representing unexplained deviations from mean expected otter densities at grid cells within a region (P) (regions described below). The random effect term was normally distributed with mean of 0 and standard deviation parameter (σ K ), and we note that for the Oregon model, this term is centered on 0 for all regions (since the specific random effects for the CA model were conditioned upon the data used to fit that model), and thus does not affect the mean projected densities; however, its inclusion does add appropriate levels of predictive uncertainty to the model projections. The effects of habitat variables are controlled by parameters, α j , which can be interpreted as log ratios, or the log proportional increase or decrease in otter densities associated with a unit change in each habitat variable. In the CA model, inclusion of these habitat variables was found to reduce the unexplained variance in equilibrium density by 42% as compared to an intercept-only model, and by 17% as compared to an intercept plus depth model (Tinker et al. 2021). Therefore, inclusion of these habitat effects is expected to similarly improve our predictive power to estimate equilibrium densities in Oregon.
To estimate carrying capacity for Oregon, we evaluated Eq. (4) using the Oregon habitat variables, and with parameter values set by iteratively drawing 10 000 samples from the joint posterior distribution estimated for CA using MCMC methods (Table 1, Tinker et al. 2021). We thereby calculated a posterior distribution of K g values for the Oregon coast, which we summed across all grid cells to obtain a posterior distribution for total expected abundance at K for both the outer coast and estuaries. We then combined abundance estimates for estuaries and outer coast to determine total predicted sea otter abundance at carrying capacity for the entire Oregon coast. We divided the study area into 3 regions (north, central, and south; Fig. 1), of approximately similar sizes (see Table 5) using the same regional boundaries as Jameson (1974), and we compared predictions between regions. By using the full joint posterior distribution of the parameters from the CA model (Table 1), and including the additional variance associated with random effects (ζ g |p ), we were able to realistically quantify uncertainty for each region in terms of credible intervals (CI, α = 0.05). For the outer coast, we reported mean densities out to the 40 m isobath, for consistency with previous studies on sea otter density along the US West Coast (Laidre et al. 2001, Tinker et al. 2021).

Core habitat areas of high sea otter densities
To anticipate locations where sea otters and human activities may interact in Oregon, we identified core habitat areas where clusters of high sea otter densities are most likely to occur. To identify high density habitat areas, we log-transformed the predicted equilibrium density values for all grid cells to obtain a more normal data distribution and extracted those cells having log-densities > 2 standard deviations above the mean (4.36 otters per km 2 ). We grouped high-density cells within 1 km of each other to identify contiguous core habitat areas, which we delineated using the 'Raster to Polygon' tool in ESRIs ArcGIS. For each resulting core habitat area, we summed predicted densities to calculate total abundance. We then excluded core habitat area polygons whose combined abundance was lower than a threshold of 8 to identify core habitat areas likely to support relatively high sea otter abundances. We set this abundance threshold at 8 by (1) identifying all core habitat areas at or near historical sea otter foraging locations (Simpson, Orford, and Blanco Reefs) from the first translocation (Jameson 1974), and then (2) identifying the single core habitat area with the lowest abundance at those historical foraging locations, to represent or suggest a minimum viable population size.

Human activities
We assessed the potential for interaction (i.e. resource competition with fisheries and human disturbance to sea otters) between sea otters in Oregon and 3 types of human activities: fisheries, non-recreational vessel traffic, and protected areas. We collected logbook landings data for the 10 most recent fishing seasons for a few commercial and recreational fisheries (Table 3). To protect fishermen confidentiality, data do not include harvest from fishing grounds where relatively few vessels were present. However, harvest data do represent the vast majority of fishery landings over this time period. We selected fisheries for target species that (1) are commonly consumed by sea otters and likely to be consumed in Oregon (Ostfeld 1982), and (2) are valued by local economies and/or conservation and, therefore, present an opportunity for resource competition (ODFW 2017a, 2019). We identified 'high-catch crabbing grounds' as areas having harvests that were 2 standard deviations above the mean of the log-transformed commercial Dungeness crab logbook data. We included recreational data on human-powered (i.e. kayaking, surfing, swimming, scuba, snorkeling, and skimboarding) and wildlife-viewing activities reported through an opt-in internet survey where respondents identified the type and location of coastal activities they participate in (LaFranchi & Daugherty 2011). Responses were spatially joined and displayed in polygon planning units used in Oregon's Territorial Sea Plan. We assessed potential non-recreational vessel activity by combining commercial shipping lanes, tugboat tow lanes, and ports that provide facilities for large ships and commercial fishing boats (Hesselgrave et al. 2011). We included 1 additional port (Newport), which was missing from this dataset, with known commercial fishing processing facilities. We also assessed the 5 no-take marine reserves in Oregon (Redfish Rocks, Cape Perpetua, Cape Falcon, Cascade Head, and Otter Rock) as protected areas.

Interaction potential
We assessed interaction potential between core habitat areas and human activities by quantifying 2 interaction metrics: direct overlap and proximity (Table 4). We measured the percent overlap between human activities and core habitat areas as the proportion of the total abundance of sea otters within core habitat areas that spatially overlapped  Table 3. Interaction potential between sea otters and human activities, including potential sources of disturbance, in Oregon, USA. Data layer descriptions and sources provided. Direct overlap and proximity metrics represent how interaction potential between sea otters and human activities were measured for each activity, while interaction level reports the calculated interaction potential for those associated metrics. Any ratios reported under interaction level are the proportions of activity or habitat spatial units (  with human activity polygons. We also quantified proximity between core habitat area polygons and human activities, reasoning that activities were likely to interact with sea otters if they occurred within 2 km of core habitat areas, based on re ported daily dispersal patterns (i.e. 1 to 2 km) of sea otters at all age and sex classes (Ralls et al. 1995). Proximity measures were more appropriate than proportional overlap for certain human activities such as those with point locations (e.g. ports) or diffuse ac tivities. For fisheries, we highlighted potential interactions with relatively high-landing fishing grounds.
All associated datasets and spatial layers are available in an online public data repository (https:// figshare. com/ projects/Oregon_Sea_Otter_Carrying_ Capacity_Kone_et_al_2020_/78075).

Carrying capacity and core habitat areas
We predicted a total abundance of 4538 (1742− 8976; 95% CI) sea otters at carrying capacity within outer coast and estuarine habitats of Oregon. We predicted higher total abundance and average otter density in the south region on the outer coast. However, within estuaries, we predicted slightly higher abundances in the central region (Table 5). We predicted higher abundances along the entire outer coast (3781 otters) than in estuaries (757 otters). We identified 10 core habitat areas (Fig. 2), mostly in the south region (80% of habitats; 742/926 otters). Core habitat areas had an average abundance of 93 otters per polygon, ranging from 8 to 494 otters.

Human activities
All fisheries examined in this study either overlapped with, or were proximate to, core habitat areas, but the interaction potential varied between fisheries. A small proportion of primary Dungeness crabbing grounds, where 22% of crab are caught along the coast, overlapped with and/or were proximate to core habitat areas (Table 3). Most high-catch crabbing grounds occurred within the central region (Fig. 2), but a smaller proportion of these grounds (2%; 6.18/ 252.42 km 2 ) were within dispersal distance for sea otters, than in the north (11%; 20.40/ 178.30 km 2 ) or south (19%; 21/ 109.71 km 2 ) regions.
Abalone were harvested from 8 harvest zones in Oregon, primarily in the south region. All harvest zones overlapped with, and were proximate to, core habitat areas (Table 3). We found most abalone landings (91%; 1336/1467 individuals) came from just 2 harvest zones, but only 1.4% (13/926 otters) of core habitat areas occurred within these zones.
When we considered fisheries as a potential disturbance to sea otters, we found all core habitat areas overlapped with either high-catch crabbing grounds or red sea urchin harvest areas. In total, approximately 76% (699/926 otters) of core habitat  Table 5. Predicted total abundance and mean (x -) sea otter densities (km −2 ), with 95% confidence interval (CI), for each region in Oregon, USA. Estuary densities are identical due to uniform estuarine density parameter applied to all estuaries areas overlapped with fisheries. We did not include abalone harvest zones in this estimate due to lack of spatial resolution. By area, most recreation (i.e. human-powered and wildlife viewing) took place in the central region (45%; 606.25/1355.52 km 2 ), relative to the north (26%; 357.78/1355.52 km 2 ) and south (29%; 391.49/ 1355.52 km 2 ) regions. While core habitat areas did not overlap entirely (58%; 536/926 otters) with recreational activity, all core habitat areas did directly overlap with recreational activity to some degree. Most of this overlap occurred in the south (68%; 365/536 otters) and central (32%; 170/536 otters) regions. Commercial shipping lanes were located primarily offshore but extend to the shoreline at 5 ports, and tow lanes were scattered across all regions throughout the study area (Fig. A1 in the Appendix). We found no overlap with tow lanes, but a small degree of overlap with commercial shipping lanes (1% core habitat areas; 9/926 otters). Fishing ports were located across the entire study area. When we considered potential disturbance from all non-recreational sources of potential vessel activity (i.e. fishing ports, commercial shipping lanes, and tow lanes) to core habitat areas, we found most core habitat areas (N = 7) were proximate to some form of vessel activity. Importantly, all (2/2) core habitat areas in the central region, and most (5/6) in the south, could be disturbed by some form of vessel activity. By combining all potential disturbances (i.e. fisheries, recreation, shipping and tow lanes, ports), we found the vast majority (97%; 896/926 otters) of core habitat areas overlap with some form of disturbance. Among regions, approximately 1% (13/896 otters), 19% (170/896 otters), and 80% (714/896 otters) of this potential direct disturbance occurred in the north, central, and south regions, respectively. Within regions, direct disturbance from the evaluated factors could affect approximately 85% (13/15 otters), 100% (170/170 otters), and 96% (714/742 otters) of core habitat areas in the north, central, and south regions, respectively.
The marine reserves in Oregon were somewhat evenly distributed within the study area, including Cape Falcon (32 km 2 ; north region), Cascade Head (25 km 2 ; central region) Otter Rock (3 km 2 ; central region), Cape Perpetua (36 km 2 ; central region), and Redfish Rocks (7 km 2 ; south region). Two of these marine reserves overlapped with, or were proximate to, core habitat areas: Otter Rock and Redfish Rocks marine reserves. Two percent (19/926 otters) of core habitat areas overlapped with Otter Rock (<1 otter) and Redfish Rocks (19 otters). Interestingly, despite the relatively small sizes of Redfish Rocks and Otter Rock, we found they contained the most core habitat areas out of all 5 marine reserves.

DISCUSSION
We present here the first carrying capacity estimate for sea otters in Oregon. The estimates of potential abundance, based on habitat−density relationships described for California and the distribution of available habitat in Oregon, suggest that sea otters could form relatively abundant populations along the Oregon coast. We have also identified a range of potential human−sea otter interactions that could result in resource competition or disturbance to sea otters. While we show sea otters may inter act with people across the entire Oregon coast, we also show some areas of more limited interaction potential where these negative interactions may be limited.
Our carrying capacity and core habitat area findings provide a glimpse into potential future sea otter distribution and abundance in Oregon, but they also corroborate observations of sea otter space use and distribution following the previous translocation effort. Specifically, these results suggest sea otters are more likely to thrive along the southern coastline, supported by greater preponderance of high quality habitat. In the 1970s, sea otters were observed to routinely forage at Orford, Blanco, and Simpson Reefs along the southern coastline (Jameson 1974). Our current results also suggest that these reefs represent potentially important future habitats for sea otters, providing a total of 24 km 2 of core habitat areas within 52 km of each other. We identified another large core habitat area just south of Newport, OR (21.12 km 2 ; 133 otters). Together these findings suggest the southern coastline may be more suitable for sea otters, based on habitat alone, with some potentially important habitats along the central coastline, and little along the north coast. While our findings cannot conclusively address whether the 1970s failed sea otter translocation was due to lack of suitable habitat, they suggest this was probably not the case.
Human interactions and disturbance have been suggested as a potential cause of the 1970s failed translocation effort (Jameson 1974). In the present study, we show that sea otters could interact with humans and potentially face disturbance from fisheries, recreation, and various sources of vessel activity (i.e. commercial shipping lanes, tow lanes, and large ports), potentially impacting population re-establishment. These human activities are spatially variable, and we have shown that some activities, like commercial shipping lanes and tow lanes, have a much lower potential of disturbing sea otters, based on the lack of spatial overlap with core habitat areas. Other activities, such as recreation, may be greater contributors to potential sea otter disturbance. We found that most of the core habitat areas predicted to experience disturbance were located within the south region, probably a reflection of the fact that most core habitat areas occurred within this region. Within regions, disturbance could disproportionally impact sea otters in the central region, as all (100% by area) core habitat areas overlap with disturbance from potential vessel activity, fisheries, and recreation. Yet, sea otter habitat also has a high proportion of overlap with human activities in the north (85%) and south (96%) regions. Importantly, these disturbance assessments are only based on spatial overlap with habitats, not intensity. Specifically, commercial shipping lane data does not indicate how frequently ships pass through lanes, recreation data does not tell us how prevalent activities are within planning units, and fisheries landings data does not directly indicate fishing effort. To better understand the potential intensity of fishery disturbance on sea otters, future studies should focus on relating sea otter core habitat areas with fishing effort, not landings. Our fishery disturbance results assume landings are correlated with effort, but this may not be the case. If landings and efforts are not correlated, there may be other areas in our study area where sea otters may be disturbed by fishing activity.
Proximity to disturbance is an assessment of potential disturbance to sea otters while foraging within 2 km of core habitat areas. Our disturbance proximity results should not be interpreted as the distance between a sea otter and disturbance stimuli that elicits a behavioral or physiological response, but rather they should be considered as the areas of the marine environment beyond core habitat areas where sea otters could come into direct contact with humans while foraging. We recognize sea otters are more likely to elicit a behavior response within 54 m of human activities (Barrett 2019), but this distance estimates the Euclidean distance between the observed location of a sea otter and disturbance stimuli. Our analysis is precautionary as it considers all potential areas where sea otters may interact with humans given their dispersal potential (i.e. within 2 km). While foraging, sea otters can disperse further than 2 km (e.g. 4 km; Ralls et al. 1995, Tarjan & Tinker 2016), but we applied a 2 km threshold as a conservative estimate given our use of a 4 km smoothing window that already considers dispersal potential. Therefore, we intended to avoid overestimating dispersal, as this might unrealistically increase our disturbance potential results.
Furthermore, our assessments assume all human activities disturb sea otters to the same degree and, therefore, are equally likely to reduce or limit population reestablishment. This assumption makes our disturbance results highly speculative as (1) we lack knowledge on the relative importance of various forms of human disturbance on sea otter behavior and energetics and (2) research on the populationlevel consequences of human disturbance on sea otters is nascent. Most research has focused on recreation due to the proximity of sea otters to ecotourists (Curland 1997, Benham 2006, and distance from a sea otter to a disturbance stimulus is a good predictor of behavior response probability (Barrett 2019). Given that proximity is a key factor, we expanded our disturbance assessments to other forms of human activities that are proximate to core habitat areas and may elicit similar behavioral and physiological responses in sea otters as recreation. Given our assumptions, direct overlap may be a stronger indicator of potential disturbance and we recommend that further research should address the relative influence of different types of human disturbance on sea otter behavior and energetics.
Resource competition between sea otters and fisheries is a common concern across the sea otter range (Carswell et al. 2015). Our study directly addressed those concerns by assessing potential interactions between sea otters, based on core habitat area distribution, and the Dungeness crab and red sea urchin commercial fisheries in Oregon. We found very little spatial overlap between core habitat areas and crabbing grounds that produce the highest annual landings in the commercial Dungeness crab fishery, which is the most lucrative fishery in Oregon (ODFW 2017a). Based on these results, we suspect sea otters and commercial crabbers may experience relatively limited resource competition and interaction. Dungeness crab is a soft-sediment species (Holsman et al. 2006), which likely explains the lack of spatial overlap between important crabbing grounds and core habitat areas. Many of these crabbing grounds occur in areas where sea otters are predicted to be less dense, and in offshore areas that are beyond the diving capacity of sea otters (Bodkin et al. 2004). Sea otters may interact with the commercial crab fishery in isolated areas, but we suspect they are unlikely to impact or compete with the entire fishery. One limitation of these crabbing−otter interpretations is that they only represent potential interactions with potential adult Dungeness crab population distribution, inferred from fishery landings data. Juvenile Dungeness crabs concentrate in relatively shallow habitatsincluding intertidal zones and estuaries (Fernandez et al. 1993, Armstrong et al. 2003. If core habitat areas spatially overlap with or are proximate to shallow habitats inhabited by juvenile crab populations, sea otter predation on juvenile crabs could potentially reduce adult crab recruitment and eventually impact the commercial fishery. The results of our study do not address this hypothetical scenario, so more research on this potential impact is warranted.
In contrast with the minimal overlap with crabbing grounds, we found a high degree of overlap between the red sea urchin and abalone fisheries and core habitat areas. This finding is perhaps not surprising given the similarities in habitat preferences of all 3 species for rocky reefs (Tegner & Levin 1982, Kato & Schroeter 1985. Given the proximity of these highlanding harvest areas to core habitat areas, these results suggest a high potential for interaction with, and impacts from, sea otters for these fisheries. Importantly, sea otters are size-selective predators that target larger individuals within prey populations (Ostfeld 1982). Urchin fisheries also target large individuals, which is likely why there is no evidence of viable commercial red sea urchin fisheries occurring within areas occupied by sea otters in other regions. If sea otters are reintroduced to Oregon, it is highly likely Oregon could experience similar declines in large sea urchins, eventually making it difficult or impossible for a commercial urchin fishery to persist in areas where sea otters have recovered. Managers may therefore wish to consider alternative economic opportunities for commercial urchin fishermen and divers, specifically, before they decide whether to proceed with a reintroduction effort. Similarly, abalone population reductions via sea otter predation could also threaten the viability of Oregon's recreational fishery or may even be a conservation concern, given current abalone population declines (ODFW 2017b). However, it is worth noting that abalone in other regions have been found to persist within cryptic habitats, sometimes at elevated densities, in areas having high density sea otter populations (Lee et al. 2016, Raimondi et al. 2015. Marine protected areas represent one possible approach to minimizing human−sea otter interactions that may lead to disturbance or resource competition. Unfortunately, our analyses suggest that sea otters may be afforded little protection by current marine reserves in Oregon due to limited spatial overlap between core habitat areas and reserves. Protecting the types of habitats important to sea otters was unlikely to have been a priority for managers while establishing the Oregon marine re serves, which could explain these findings. Several other protected areas exist along the Oregon coast (i.e. marine gardens, limited-access protected areas, and national wildlife refuges), but we did not assess these protected areas as they are not fully protected, and monitoring and enforcement are limited. Even if these protected areas only prohibit some human activities some of the time, that exclusion could help protect and preserve potentially important sea otter habitat and could be investigated in future research.
The results of our study come with a few caveats and limitations. First, this study is an extension and an extrapolation of how habitats support sea otter populations in California, but it is uncertain whether Oregon habitats, especially kelp canopies, will support similar equilibrium sea otter densities as in California. The CA model indicates that the presence of kelp canopy is associated with higher sea otter densities; however, those results are likely driven primarily by giant kelp Macrocystis pyrifera, a species which is more persistent (Foster & Schiel 1985) than Oregon's bull kelp, which experiences intra-and inter-annual variability (Springer et al. 2007). Kelp provides an important food resource for otter prey (i.e. sea urchins, abalone). Reduced kelp could limit prey population size and quality (i.e. mass), which could limit otter densities. Kelp variability could also redistribute core habitat areas from where we have predicted. This limitation highlights the important bottom-up processes that support sea otters and how environmental variability may impact sea otter abundance and distribution. Yet, through their strong topdown pressures, sea otters can help maintain ecosystem function and important habitats, such as kelp, by controlling herbivores, like sea urchins. Recently, northern California (where sea otters have historically occurred, but do not currently) experienced a 90% reduction in canopy-forming bull kelp due to climatic and biological stressors, with purple urchin grazing being a major contributor (Rogers-Bennett et al. 2019). To date, Oregon has only experienced some kelp cover reductions in isolated locations, nowhere near the extent observed in northern California. Yet, if these events continue to unfold in Oregon, reintroducing sea otters might help limit largescale losses of kelp forests like that which has occurred in northern California. Despite these limitations, we feel our extrapolation of sea otter densities associated with key habitat variables from California to Oregon is appropriate given relative geographic proximity, data availability, and application of this novel approach. To address this limitation, however, future analyses could determine sea otter density and habitat functional relationships in other locations within the current range of sea otters (e.g. Washington, Alaska, British Columbia) to assess how representative the California data may be of Oregon.
A second caveat to our results relates to the projected abundances in estuaries. The densities of estuarine sea otter populations predicted by the CA model are informed by the few currently occupied estuaries in California, specifically Morro Bay and Elkhorn Slough. The former estuary supports a fairly low abundance of otters, while the latter (Elkhorn Slough) supports a very high abundance, apparently sustained by an abundant and productive prey base (Kvitek & Oliver 1988). The contrast in abundance between the 2 California estuaries leads to a high degree of uncertainty in our model estimate for estuarine habitats. Extrapolation of the model to Oregon estuaries effectively projects an average of the Morro Bay and Elkhorn Slough equilibrium densities, with very large associated standard error. While this approach may be reasonable as a first pass approach, further research is needed to elucidate the potential of Oregon estuaries to support thriving sea otter populations.
A third limitation of our analyses is that we only identified core habitat area and potential human interactions on the outer coast, not within estuaries and along shorelines. Our findings therefore do not reflect the potential role estuaries and shorelines may play in supporting future sea otter populations, including providing additional foraging habitat and resting areas to haul out (despite this behavior being rare), nor do they capture the potential for human− sea otter interactions in estuaries. Sea otters occur in Elkhorn Slough, an estuary in California, with high population density supported by locally abundant clam populations (Kvitek & Oliver 1988, Maldini et al. 2012, Hughes et al. 2013, Eby et al. 2017. There is also evidence that other California estuaries (e.g. San Francisco Bay, Drakes Estero, and Morro Bay) were historically occupied by sea otters, based on archaeological remains discovered in Native American shell middens and anecdotal accounts of sea otter estuarine habitat use (Schenck 1926, Odgen 1941, Broughton 1999, Jones et al. 2011, Hughes et al. 2019). The potential for California estuaries as future sea otter habitat has recently been considered (Hughes et al. 2019). Sea otters may also thrive in estuaries along the Oregon coast, with relative prey availability likely acting as the primary determinant of population potential. Many Oregon estuaries do contain populations of potential sea otter prey species, particularly bay clams and Dungeness crabs. In fact, in some of the estuaries identified by our model (Alsea Bay, Coos Bay, Netarts Bay, Siletz Bay, Tillamook Bay, Yaquina Bay), bay clam populations have been identified (Ainsworth et al. 2014). To better understand the potential for Oregon estuaries to play a role in supporting a resident sea otter population, future research should investigate prey availability, and human use and pre sence within and/or near potentially important estuaries.
A final caveat is that our examination of potential human−sea otter interactions in Oregon lacks spatial and temporal resolution. Carrying capacity predictions were calculated using a finer spatial resolution than was available for most of the human activities, specifically commercial Dungeness crab cells, abalone harvest areas, and recreation planning units. The available data indicate approximately where those activities are located but lack spatial precision and are thus representative of very general patterns. Additionally, these interpretations are temporally static. We only considered where core habitat areas are located relative to current human activities, and did not consider seasonal patterns, or the time course from reintroduction to equilibrium abundance. As previously discussed, habitat and biotic features can shift over time due to any number of forces (e.g. climate change, ocean acidification, etc.). Humans may redistribute their activities, such as fisheries and reserves, in response to these ecological shifts, and so the patterns of potential interaction presented here may not hold under such changes.

CONCLUSIONS
Reintroductions are a well recognized strategy to augment the recovery of at-risk species (Clark & Westrum 1989, Seddon et al. 2007. A sea otter reintroduction to Oregon could reestablish this once native species and help sea otters recover from previous human exploitation. However, given the risk of another failed reintroduction effort and lack of information to explain that failure, managers have not yet decided whether they will proceed with a reintroduction. To facilitate the decision, we have attempted to address some of the common uncertainties associated with species reintroductions, including habitat suitability. Our analyses indirectly address some of the hypotheses for the cause of failure of the previous translocation effort; for example, our results are not consistent with the 'lack of suitable habitat' hypothesis, suggesting instead that the available habitat could support a population of > 4500 sea otters. Our study also identified areas of particularly high density, and potential interactions with human activities. Managers could use this information to set reasonable population recovery targets that factor in both ecological and socioeconomic considerations. Lastly, we have investigated the potential to reintroduce sea otters to Oregon from a bottom-up perspective; moving forward, an important next step will be to consider how reestablishing sea otters could change the environment via top-down processes. Reintroducing sea otters could result in negative effects to certain commercial fisheries but could also lead to positive outcomes such as restored kelp habitats and ecological resilience that supports fisheries and tourism operations. As managers consider whether to proceed with a reintroduction, monitoring will be key to understanding how these bottom-up and top-down processes will play out, providing insight into the trade-offs associated with each process, and assessing the ultimate success of the reintroduction effort.