Species recovery and recolonization of past habitats: lessons for science and conservation from sea otters in estuaries

Historical research effort on estuaries as sea otter habitat

We summarized research efforts in different ecosystems using a Web of Science search with the following terms: TOPIC: (sea otter*) AND TOPIC: (marsh* or salt marsh* or kelp* or seagrass* or rocky* or soft-bottom* or nearshore* or coast* or subtidal* or intertidal* or beach* or embayment* or * bay* or shore*). Results were accessed on March 3, 2017. We divided studies by two categories: (1) northern (E. lutris kenyoni) and southern (E. lutris nereis) sea otter populations, and (2) coastal ecosystem type (open coast, estuary/embayment, and both). We further subdivided population and coastal ecosystem type into habitats within the two ecosystems (beach, kelp, rocky intertidal, seagrass, soft-bottom, and multiple).

State-space model of sea otter growth and equilibrium abundance in a California Estuary

At the present time Elkhorn Slough is the only estuary in California in which a distinct and self-sustaining population of sea otters has become established (Kvitek et al., 1988; Maldini et al., 2012; Hughes et al., 2013; Eby et al., 2017), although with continued range expansion it is likely that other estuaries (including San Francisco Bay) will eventually be colonized given the current population is <40 km away from the mouth of San Francisco Bay (Tinker & Hatfield, 2017a). Sea otters initially moved into the Slough in 1984, however the early colonizers were mostly or entirely males (consistent with typical patterns of sea otter range expansion; Lafferty & Tinker, 2014). Reproductive females did not become established in Elkhorn Slough until the early 2000s, largely as a result of the addition of rehabilitated juvenile female otters by the Monterey Bay Aquarium (Mayer et al., 2019), after which intrinsic growth of the Elkhorn Slough population was rapid (Estes & Tinker, 2017; Mayer et al., 2019). It is therefore reasonable to characterize the first 15 years of sea otter occupation in Elkhorn Slough as a population-establishment phase, and we use data on population trends from 2000–2018 to assess the potential for growth and equilibrium of established populations in estuarine habitats. Annual survey data for Elkhorn Slough were extracted from U.S. Geological Survey population survey data available in an online repository (Tinker & Hatfield, 2017b) and are also provided in Dataset S1.

To estimate key parameters of sea otter population dynamics in estuaries, we fit a Bayesian state space model (SSM) to the time series of survey data for Elkhorn Slough using established methods (Stroud, Muller & Sanso, 2001). The rationale of an SSM is as follows: a process model of population dynamics (in this case, a stochastic logistic growth model of density-dependent population growth) is fit to a time series of population counts using Markov Chain Monte Carlo (MCMC) methods. This method enables disentanglement of process error (i.e., environmental stochasticity, the unexplained but real variation in annual growth rate, or λ) from observer error (i.e., measurement and sampling error, the unexplained year-to-year differences in sightability that do not reflect true abundance). We assume that true abundance, N_t, is a latent variable whose dynamics are described as: (1) $$N_t=N_{t-1}\cdot \exp \left(r\right[1-\frac{N_{t-1}}{K}]+{\mathrm{\epsilon }}_t),{\mathrm{\epsilon }}_t\sim N(0,{\mathrm{\sigma }}_p)$$

In Eq. (1), r is the intrinsic rate of population growth, K is the equilibrium abundance, or carrying capacity, and N(0, σ_p) represents a random normal variable with mean 0 and standard deviation σ_p, a fitted parameter representing process error or environmental stochasticity. We note that Eq. (1) reflects simple logistic growth, a special case of theta-logistic growth where the implied value of θ (a parameter determining the shape of the relationship between log(λ) and density) is set to 1. The assumption of simple logistic growth is recommended for data sets where there are insufficient data to reliably estimate θ (Clark et al., 2010) and is consistent with other simple models of sea otter population growth (Tinker, 2015; Tinker et al., 2019a).

In addition to obtaining estimates of r and K appropriate for estuarine habitats, we also derive estimates of equilibrium densities specific to habitat types, $k_h^d$. We define three general habitat types for sea otters in Elkhorn Slough: (1) nearshore open water, (2) eelgrass beds, and (3) salt marsh habitats (including both tidal creeks and adjacent pickleweed beds). To calculate $k_h^d$, we define A_h as the area of habitat type h available in the Slough (in units of km²) and p_h as the proportion of all surveyed sea otters occurring in each habitat type h. We then calculate the equilibrium densities for each habitat type as: (2) $$K_h^d={\displaystyle \frac{p_h\cdot K}{A_h}}$$

To fit the process model defined by Eq. (1), we relate the latent variable, N_t, to the observed survey counts at Elkhorn Slough, C_t, by assuming that the counts are distributed as a negative binomial distribution with mean N_t and variance σ²_o: (3) $$C_t\sim NB(a,b),a=\frac{N_t^2}{({\mathrm{\sigma }}_o^2-N_t)},b=1-\frac{({\mathrm{\sigma }}_o^2-N_t)}{{\mathrm{\sigma }}_o^2}$$ where σ_o represents observer error, another fitted parameter. Equation (3) implies that survey counts will, on average, be centered around true abundance, and thus does not account for imperfect detection. We recognize this is unrealistic, as some (small) proportion of animals are invariably missed on visual surveys (Estes & Jameson, 1988); however, the un-replicated complete count method used to census otters in California does not provide a means of estimating a sightability correction factor. Consequently, for consistency with previous analyses of these survey data (Lubina & Levin, 1988; Tinker et al., 2006; Lafferty & Tinker, 2014), we make the simplifying assumption that counts are centered around true abundance, recognizing that this means that our estimates of N_t are biased low to a small (but unmeasured) degree.

We estimated parameters in Eqs. (1–3) using Bayesian MCMC methods, implemented using JAGS (“Just Another Gibbs Sampler,” Code S1) and the R programming language (Plummer, 2003; R Development Core Team, 2015). We used vague half-Cauchy priors for K, N₁, and all variance parameters, and an informative normal prior for r with mean 0.2 and standard deviation 0.1, based on the results of numerous studies showing that r_max in recently established populations is consistently within the range 0.20–0.25 (Estes, 1990; Gerber et al., 2004; Lafferty & Tinker, 2014; Tinker, 2015; Tinker et al., 2019a). After model burn-in of 10,000 simulations we saved 500 samples for each of 20 chains, for a total of 10,000 samples used to calculate posterior distributions for all parameters and for the latent variable N_t (Fig. S1). Model convergence was evaluated by graphical examination of the CODA traces, and by ensuring that the R-hat statistic was <1.01 for all parameters. We also conducted posterior predictive checking (PPC; Gelman et al., 2000), creating a scatter plot of a discrepancy measure (summed deviance of observed counts) for replicate (simulated) vs actual (observed) data and reporting the associated Bayesian P-value (requiring 0.25 < P < 0.75; Gelman et al., 2014). We then summarized the mean, median, and 95% credible intervals (CI) for the posterior distributions of each parameter.

Applying state space model results to San Francisco Bay

We conservatively took the lower bound of the 95% CI of $k_h^d$ for Elkhorn Slough habitats as our point estimates for equilibrium densities of the equivalent habitats in the San Francisco Bay estuary. A fourth habitat type, offshore open water (defined as areas >2,000 m from shore but still within the 40 m depth zone), occurred in San Francisco Bay but not in Elkhorn Slough. Based on extensive data on the depths of sea otter foraging dives (Bodkin, Esslinger & Monson, 2004; Tinker et al., 2007; Thometz et al., 2016), individual habitat use data from Washington (Laidre et al., 2009) and California (Tinker et al., 2017), and survey data from large embayments in Alaska such as Prince William Sound, Yakutat Bay and Glacier Bay (Esslinger & Bodkin, 2009; Esslinger et al., 2015; Tinker et al., 2019a), it is clear that sea otters utilize such offshore habitats, but that the densities of animals in offshore areas are substantially lower than densities in the nearshore zone. For example, using published census data for southern sea otters from 1985 to 2018 (Tinker & Hatfield, 2017b), and restricting analysis to areas of soft sediment habitat along the open coast, we calculated that the number of otters observed >2,000 m from shore was just 15% of the number counted ≤2,000 m from shore. To obtain an estimate of equilibrium density for offshore areas <40 m depth in San Francisco Bay ($k_4^d$), we multiplied the equilibrium density for nearshore open water ($k_1^d$) by a correction factor, Φ = 0.1, thus conservatively assuming that offshore densities would be ~10% of nearshore densities. We used multiple GIS sources to characterize the four different sea otter habitats in San Francisco Bay, including EcoAtlas (ecoatlas.org), the California Aquatic Resources Inventory for wetland habitats, and the Eelgrass Aquatic Resources for eelgrass habitats.

We applied our conservative estimates of habitat-specific equilibrium densities derived from Elkhorn Slough to GIS layers of habitat cover in San Francisco Bay, then summed across habitats to obtain a total estimated abundance at K for San Francisco Bay (Table 1). We used this value of K, together with parameter values for r and of σ_p estimated from Elkhorn Slough, to parameterize 500,000 forward simulations of population growth in San Francisco Bay (achieved by iteratively solving Eq. (1)). We drew all parameter values from their respective posterior distributions to account for parameter uncertainty. For all simulations we initialized the population (N₁) with 20 sea otters: assuming a female:male ratio of 2:1, this corresponds (approximately) to the 14 rehabilitated juvenile female sea otters successfully introduced to Elkhorn Slough between 2002 and 2014, which contributed the majority of population growth during this period (Mayer et al., 2019). We also assumed there would be a population establishment phase similar to that seen for Elkhorn Slough, and thus forced the mean growth rate to 0 for the first 15 years. We summarized results of these simulations by plotting the mean expected trajectory and 90% CI for 50 years of population dynamics. Finally, we examined the sensitivity of the simulation results to our various assumptions and input parameters. Specifically, we sequentially reduced each model parameter ($k_h^d$, Φ, r and σ_p) by 10% of its default value, while holding other parameters constant, and measured the proportional response in N′₅₀, the mean projected abundance of the population after 50 years.

Prey availability in estuaries outside the sea otter range

To evaluate prey availability at sites within (Elkhorn Slough and Morro Bay) and outside (San Francisco Bay and Drakes Estero; Fig. 1A) the current southern sea otter range, we took two approaches: first, we assembled existing prey species lists (excluding Drakes Estero); and second, we sampled each estuary for crab, which have been documented to be important prey for successful sea otter recolonization in estuaries (Garshelis & Garshelis, 1984; Hughes et al., 2013). To generate sea otter prey species lists we assembled data from various monitoring programs (see Dataset S3 for references) that have sampled benthic organisms in each estuary.

To sample crab at each estuary we used standardized trapping techniques during summer 2016 (Elkhorn Slough), fall 2016 (Drakes Estero, San Francisco Bay) and winter 2017 (Morro Bay, California Department of Fish and Wildlife, Permit SC-6563). Due to logistical constraints we were unable to sample all four sites within the same season. We used shrimp pots (http://www.westmarine.com—Willapa Shrimp Pots) with modified openings that allow access to large crabs yet still prevent access or entrapment of any larger predators, such as sea otters, birds, and harbor seals. The dimensions of the shrimp pots were 61 × 61 × 23 cm, with four openings that were rigidly supported and ~20 cm in diameter. Each site had seven to eight locations sampled over a single tidal cycle in Zostera marina seagrass beds, with the exception of Elkhorn Slough, which had four locations that were sampled multiple times (n = 8–12 times during summer 2016). Each site received one trap per sampling effort, and traps were soaked for ~24 h. Each crab was identified and measured for carapace width, which was converted to edible biomass using power functions from Oftedal et al. (2007). San Francisco Bay crab catch per unit effort (CPUE) included the invasive European green crab (Carcinus maenas), caught in an April 2017 crab survey, which we have observed to be a prey item of sea otters in Elkhorn Slough (B. Hughes, 2015, personal observation), but was not included in the biomass conversion estimates in Oftedal et al. (2007). We applied a conservative biomass estimate for green crab using the biomass conversion for Metacarcinus magister. We consider this estimate to be conservative since green crab have a relatively narrow carapace width relative to biomass compared to other native Cancrid crab species.

To determine differences in crab prey availability in estuaries within and immediately outside the southern sea otter range we compared differences in edible crab biomass and carapace width of two common estuarine crab prey species: Romaleon antennarium and Cancer productus. For one site, Elkhorn Slough, we used a historic dataset that surveyed crab in the 1970s (1971–1976, see Hughes et al., 2013 for a description of the data), to compare to the contemporary crab trapping efforts at Elkhorn Slough and the other three estuaries. The 1970s Elkhorn Slough crab dataset characterized Elkhorn Slough crab populations prior to sea otter recolonization in the early 1980s, and thus helped support our interpretation of differences in crab populations for our 2017 surveys.

Differences in sea otter presence (fixed factor) and site (random factor) of CPUE and size were determined using linear mixed models with the nlme package in R v. 3.4.3. A Tukey multiple comparison test was used to compare differences among sites with and without sea otters using the multcomp package.

Historical research effort on estuaries as sea otter habitat

While many studies continue to be published on outer coast sea otters, our Web of Science search resulted in few investigations examining estuarine sea otters (Fig. 1B; Dataset S1; Code S2), and most of the estuarine studies were based on northern sea otters (E. lutris kenyoni). We found only nine studies that focused specifically on estuarine ecosystems for southern sea otters. Of all the studies we examined (n = 158), only two focused on seagrass (Z. marina) habitats (Fig. 1C), and only one to our knowledge has focused on salt marshes (Eby et al., 2017). Whereas, kelp was the most dominant single habitat studied for both northern and southern sea otter populations.

State-space model of sea otter growth and equilibrium

The state-space model fit to survey data from Elkhorn Slough (Dataset S2; Code S3) converged well and provided excellent fit to the data (Fig. S1). Trace plots of posterior samples showed well-mixed chains (Figs. S2 and S3), with all R-hat statistics <1.01 (Table 1), and the PPC scatterplot showed excellent agreement between replicate and observed data sets (Bayesian-P = 0.55; Fig. S4). The estimated instantaneous rate of growth at low density (r) was 0.22 (CI₉₅ = 0.10–0.33), while the estimated magnitude of environmental stochasticity (σ_p) was 0.30 (CI₉₅ = 0.16–0.47). The estimated value of K for Elkhorn Slough was 160 (CI₉₅ = 49–367), and using the lower bound of the 95% CI for K we estimate habitat-specific equilibrium densities ($k_h^d$) for three habitat types: nearshore open water (8.9 otters × km⁻²), eelgrass beds (19.2 otters × km⁻²) and saltmarsh habitats (4.5 otters × km⁻²; Table 1 and Fig. 2B).

Applying state space model results to San Francisco Bay

Our habitat-specific population growth model provides a conservative estimate for the equilibrium abundance of sea otters in San Francisco Bay of over 6,600 animals (Table 1) and shows that just 20 colonizers could grow to a population of 5,000 or more within 50 years (Fig. 2C). There is a great deal of uncertainty in growth projections, evident in the wide CI bounds around the mean projected trend in Fig. 2C: this variability reflects both the effects of process error (environmental stochasticity) and parameter uncertainty. To explore the latter source of uncertainty we conducted a sensitivity analysis, the results of which indicated that the parameter having the greatest impact on projected abundance was r, the rate of growth at low densities (Fig. S5). The estimate of $k_1^d$ (equilibrium density in nearshore open water) had the next greatest impact on N′₅₀, reflecting the predominance of this habitat type in San Francisco Bay. The magnitude of environmental stochasticity (σ_p) also had substantial influence, with a decrease in σ_p leading to an increase in projected abundance.

We note that our model projections are conservative in two respects: they assume lower equilibrium densities than currently seen in Elkhorn Slough, and estimates of the distribution of a key habitat type, eelgrass beds, are based on conservative estimates of the current distribution and do not include positive feedback loops whereby sea otter recovery could promote further seagrass expansion. Furthermore, we excluded certain estuarine habitats (beaches and subtidal flats), which are also known to be used by sea otters (Kvitek et al., 1988), but for which reliable density estimates could not be obtained. Despite these caveats, our model results suggest that San Francisco Bay could eventually support a sizeable southern sea otter population, 160% or more of the range-wide population as of 2018.

Prey availability in estuaries outside the sea otter range

Our comparison of prey availability in estuaries with sea otters (Elkhorn Slough and Morro Bay) with one of the focal recovery sites (San Francisco Bay) suggests that prey resources in San Francisco Bay are sufficient to sustain a healthy sea otter population. Out of the 52 prey species we identified in the two California estuaries with sea otters, 34 (65%) occurred in San Francisco Bay (Table 2; Dataset S3). Prey species include some that are of value to commercial fisheries, such as Dungeness crab (Metacarcinus magister), littleneck clam (Leucoma staminea), and mussels (Mytilus spp.), and also include invasive species such as European green crab (Carcinus maenas).

Crab survey efforts also indicate that our two focal sea otter recovery sites (San Francisco Bay and Drakes Estero) have an adequate supply of crab prey to support a sea otter population. Both estuaries without sea otters had significantly greater crab biomass caught in standardized traps compared to the estuaries (Elkhorn Slough and Morro Bay) with sea otters (Fig. 3A), and Drakes Estero had a significantly greater crab biomass than San Francisco Bay (Dataset S4; Code S4). When we compared sizes, crabs (R. antennarium and Cancer productus) from Drakes Estero were significantly larger than crabs from San Francisco Bay (no sea otters) and the two sites with sea otters (Fig. 3B; Dataset S5; Code S4). Furthermore, crabs of both species in Drakes Estero were similar in size to those in Elkhorn Slough prior to sea otter recolonization in the 1970s. Whereas, crab sizes in San Francisco Bay were intermediate between Drakes Estero (no sea otters), Morro Bay (with sea otters), and Elkhorn Slough (before and after sea otter recolonization).

Paradise lost: sea otters in California estuaries

Despite a substantial increase in population size over the past century, southern sea otters have recolonized only approximately 13% of their historic range (U.S. Fish and Wildlife Service, 2015), and long-term population growth has been slow (3–5% per year; Tinker & Hatfield, 2017a) compared to recovering populations of northern sea otters (E. lutris kenyoni) (Estes, 1990), where total population growth rates have been reported to be as high as 17–20% per year depending on location. Although this discrepancy in population growth rates has been interpreted as evidence of persistently high southern sea otter mortality, the observed population trajectory can largely be explained as the combined effect of localized density-dependence and the strong spatial structuring typical of sea otter populations. Adult females, in particular, rarely move more than 15 km per year from their home range center (Tinker, Doak & Estes, 2008; Tarjan & Tinker, 2016) along the predominantly long, narrow configuration of central California coastal habitat (Tinker, 2015). In such elongated and essentially one-dimensional habitats, a large proportion of the population occurs in resource-limited areas, with population growth and range expansion occurring only at the extreme ends of the range (Bodkin, 2015; Tinker, 2015). In more complex, multi-dimensional habitats (such as those occurring in southeast Alaska and British Columbia) there are typically multiple expansion fronts; therefore, a greater proportion of the population occurs at sub-equilibrium densities, resulting in faster growth rates at the regional scale. Because they afford a diversity of habitats and prey for range expansion, estuaries are among the ecosystems that are expected to be conducive to rapid sea otter population growth.

Two estuaries are located within the current range of the southern sea otter: Elkhorn Slough and Morro Bay. These estuaries are located to the north and south, respectively, of the central portion of the range, where recovering southern sea otters have resided the longest and are at or near equilibrium densities (Tinker, Doak & Estes, 2008; Thometz et al., 2016; Tinker et al., 2019b). Sea otters recolonized the outer portion of Morro Bay estuary in 1973 (Lubina & Levin, 1988) and have recently increased dramatically in number in Morro Bay Harbor (Tinker & Hatfield, 2017a), but they have not fully recolonized the inland portion of the estuary. In contrast, sea otters in Elkhorn Slough utilize inland estuarine habitats extensively. Over the past two decades, the Elkhorn Slough population has increased rapidly to more than 100 animals and has only recently shown signs of reaching equilibrium (Silliman et al., 2018). The initiation of the period of rapid growth in Elkhorn Slough can partly be attributed to the addition of rehabilitated sea otters from the Monterey Bay Aquarium, specifically to 14 females that survived for >1 year post release and were among the first animals to reproduce in the Slough (Mayer et al., 2019). Our initialization of San Francisco simulations with 20 animals was explicitly based on this scenario, although natural colonization would of course be expected to produce similar results.

Although some sea otters still leave Elkhorn Slough occasionally to forage along the outer coast, most appear to be entirely resident within the estuary, using different habitats for feeding, resting and interacting. This recovery demonstrates that estuarine habitats and prey resources can support a high-density, fully resident sea otter population. Estuaries such as Elkhorn Slough are home to diverse and abundant invertebrate communities (Wasson et al., 2002), and sea otter prey such as clams, crabs and worms can recruit and grow rapidly (Dataset S3). The high productivity of prey resources has supported decades of high foraging success of the dense sea otter population (Hughes et al., 2013; Silliman et al., 2018).

The two major estuaries currently inhabited by sea otters (Elkhorn Slough and Morro Bay) comprise 2,400 ha of total habitat and <50 ha of eelgrass habitat as of 2014 (Hughes et al., 2013, 2016; MBNEP, 2014). However, if estuaries immediately outside (100 km to the north and south) the current sea otter range are included, the total available estuarine habitat is much greater (Fig. 1A). Considering only those estuaries with eelgrass habitat (Bernstein et al., 2011), which could be a catalyst for sea otter colonization and is an indicator of good estuarine water quality and rich invertebrate prey resources, we found 11 more estuaries with a total area of 135,000 ha, which is a 56-fold increase in available estuarine habitat relative to that within the currently occupied range. In comparison, the total available outer coast habitat within the same region is 147,000 ha (Laidre, Jameson & DeMaster, 2001). When comparing the coverage of the two dominant habitat-forming species within the sea otter range—eelgrass in estuaries and kelp (e.g., Macrocystis pyrifera) on the outer coast—we conservatively estimated there to be 1,750 ha of eelgrass habitat within the projected range, and most of the habitat is on the northern side of the currently occupied range (see Bernstein et al., 2011), compared to about 5,700 ha of available kelp habitat along the outer coast (https://earthworks.stanford.edu/catalog/stanford-nf191bc1535). Even using our most conservative estimates, eelgrass alone could provide 23% of the total habitat available for sea otter range expansion over the next few decades within the 100 km projected stretch of mainland California. This estimate does not take into account the other important estuarine habitats used by sea otters, such as salt marshes, tidal creeks, and mudflats (Kvitek et al., 1988; Eby et al., 2017), which could greatly add to the total estimate of new sea otter habitat. Taken together, estuaries could provide important recovery habitat for sea otters as they expand to the north and south, and serve as a refuge from sharks and unfavorable outer coast conditions, like El Niño-driven storm and warming events, which reduce the abundance of kelp (Dayton & Tegner, 1984; Edwards, 2004).

If sea otters recolonize the large estuaries on the California coast, such as San Francisco Bay and Drakes Estero, or if rehabilitated otters were proactively reintroduced, the potential for population growth could be profound (Fig. 2C). As of 2014, eelgrass beds in San Francisco Bay cover about 1,100 ha (Merkel & Associates, 2014), and as of 2005 Drakes Estero eelgrass coverage is 300 ha (Brown & Becker, 2007). Drakes Estero has a relatively pristine watershed where eelgrass beds have been observed to be stable since a 2005 survey (B Becker, 2019, personal observation). San Francisco Bay’s watershed is highly developed with extensive anthropogenic impacts, including wetland conversion and loss (Gedan & Silliman, 2009). However, the Bay still supports many of the same habitats and prey resources utilized by southern sea otters in their currently occupied range (Table 2; Fig. 2A).

Sea otters are a prime example of a keystone predator in kelp forests: by regulating herbivorous urchins, sea otters increase the abundance and distribution of kelp (Estes & Palmisano, 1974; Estes et al., 1998). This well-established paradigm has recently been complemented by the discovery that sea otters play an analogous role in estuaries by enhancing the resilience of seagrass to anthropogenic nutrient inputs (Hughes et al., 2013, 2016). Through their high consumption rates of crabs in Elkhorn Slough, sea otters indirectly benefit seagrasses that are threatened by extreme nutrient loading and eutrophication (Hughes et al., 2011, 2013). By consuming crabs, sea otters can generate a trophic cascade, whereby small grazers (<5 cm) are released from predation by crabs, which enables them to consume shade-causing algal epiphytes that grow on the leaves of eelgrass (Z. marina) (Hughes et al., 2013, 2016). Prior to sea otter recolonization of the Elkhorn Slough estuary in 1984, eelgrass was nearly locally extinct. Since the recolonization of sea otters in the last three decades, eelgrass has recovered and expanded in area by >600%, and sea otter population growth positively correlates with eelgrass recovery (Hughes et al., 2013). At least in this ecological context, sea otters are capable of mediating the harmful effects of poor water quality on seagrasses without management intervention.

Results showing positive effects of sea otters in estuaries have regional conservation implications, as many estuaries toward the northern and southern limits of the current range of sea otters are highly threatened by anthropogenic stressors (Greene et al., 2015; Hessing-Lewis et al., 2017). Recently, severe losses of eelgrass to the south of Elkhorn Slough in Morro Bay (>95% since 2007; MBNEP, 2014) have been observed, with the exception of areas near the mouth where sea otters permanently reside. A wide range of ongoing human impacts to eelgrass in the north in San Francisco Bay (Boyer & Wyllie-Echeverria, 2010) have created the impetus for large-scale restoration. Eelgrass-facilitating trophic interactions may be one mechanism by which successful conservation and restoration outcomes are maximized.

An ecosystem-based management approach to decisions about estuarine restoration efforts

The restoration of estuaries has become a priority for land managers because of the shoreline protection, flood management, water filtration, wildlife habitat, recreational and educational opportunities, and other services that estuaries provide. San Francisco Bay, designated an “estuary of national significance” under the Clean Water Act, received $5.2 million in funding through the Environmental Protection Agency in 2017 for water protection and restoration projects in the San Francisco Bay area. Within the San Francisco Bay, the Don Edwards San Francisco Bay National Wildlife Refuge is carrying out the largest wetlands restoration project on the West Coast, the South Bay Salt Pond Restoration project, with the aim of improving the physical, biological, and chemical health of the estuary, including restoring habitat for birds, fish, benthic invertebrates, marine mammals, and plants, some of which are threatened or endangered. Re-occupation by sea otters could assist restoration goals (San Francisco Bay Subtidal Habitat Goals Report, 2010) by enhancing the resilience and functioning of estuaries through indirect effects. Potential ecological effects of sea otter recovery extend beyond the top-down impacts of sea otter trophic interactions on seagrass recovery (Hughes et al., 2013, 2016); they also encompass the well-established ecosystem services supported by seagrass systems (Waycott et al., 2009), including a growing recognition of seagrass as a sink for atmospheric carbon (Fourqurean et al., 2012; Greiner et al., 2013). Additionally, sea otter bioengineering effects may have positive feedbacks in soft sediment habitats. Sea otter consumption of clams and subsequent shell deposition can provide hard substrate in an otherwise soft sediment system (Kvitek & Oliver, 1992), which could promote the settlement and sub-surface growth of invertebrates and algae, altering community structure and increasing fish habitat. Finally, the physical disturbance created when sea otters forage for invertebrates within seagrass and other soft-bottom ecosystems (Kvitek et al., 1988) may have ecosystem feedbacks on benthic communities (Kvitek & Oliver, 1992) and ecosystem resiliency.

Based on previously reported rates of range expansion in southern sea otters (Lubina & Levin, 1988; Tinker, Doak & Estes, 2008; Lafferty & Tinker, 2014), sea otter recolonization of other California estuaries would be expected to happen gradually over coming decades. However, as mentioned above, white shark predation now increasingly curtails range expansion at both ends of the range (Tinker et al., 2016). In the absence of natural range expansion, an option to be considered is facilitated range expansion, focusing on under-utilized habitats such as estuaries, where shark predation is reduced. This could be accomplished by through the release of sea otters into San Francisco Bay, Drakes Estero or other California estuaries, as was done for Elkhorn Slough (Mayer et al., 2019).

The reintroduction of a keystone predator to the estuarine ecosystems it once occupied involves potential trade-offs. The strength of sea otters’ ecosystem effects is a result of their high rates of consumption of invertebrate prey. For example, estuaries with sea otters (Elkhorn Slough and Morro Bay) have a significantly lower crab size and crab biomass for the two most abundant species of cancrid crab prey in California estuaries (R. antennarium and Cancer productus) than estuaries without sea otters (San Francisco Bay and Drakes Estero; Fig. 3). To maintain homeostasis, sea otters must consume, 25% of their body weight per day in invertebrate prey (Costa & Kooyman, 1982). As such, sea otters can have very strong effects on their prey (Hughes et al., 2013, 2016; Needles et al., 2015), and have been documented to forage in a range of estuarine habitats, from seagrass beds to salt marshes (Fig. 4), exploiting a wide range of prey items, including crabs, clams, mussels, and worms (Table 2; Dataset S3).

While sea otters have diverse diets that include many non-commercial species, sea otter recolonization could affect commercial and recreational fisheries through competition or indirect effects. Sea otter predation can impact shellfish fisheries by reducing target species’ abundance and size (Garshelis & Garshelis, 1984). In estuaries where oyster aquaculture is practiced, such as Tomales Bay or Morro Bay, it is conceivable that sea otters could affect oyster production, although no data are currently available to assess the likelihood or magnitude of such interactions. Whereas far fewer fisheries occur in California estuaries than on the open coast, estuaries provide important nursery habitat for some commercially exploited species. Sea otter effects on spawning adult Dungeness crab would likely be largely negative, while effects on Pacific herring (Clupea pallasii) would likely be positive through beneficial effects on eelgrass habitat. The magnitude of positive or negative effects on commercial or recreational fisheries is highly contingent on the particular ecological setting, the fisheries involved, the number of sea otters present, and other factors.

Despite the positive or negative effects of sea otters on fisheries, sea otter viewing has high recreational value and some people value their intrinsic presence and would view reintroduction favorably. Sea otters are immensely popular with the public, in aquaria worldwide and in their natural habitats. Visitors to Elkhorn Slough cite sea otter-watching as the number one reason to visit this coastal habitat (Kildow & Pendleton, 2010), and in situ cameras (www.elkhornslough.org/ottercam) showing real time footage of sea otters in the estuary receive tens of thousands of unique views each year. If sea otters become common along the shores of San Francisco Bay, millions of additional people could have the opportunity to observe and enjoy sea otters in habitat they historically occupied, and ecotourism opportunities and revenues could increase accordingly. Sea otters could also build support for ecosystem restoration of degraded seagrass and salt marsh habitats within estuaries (Van Dyke & Wasson, 2005; Hughes et al., 2013; Wasson et al., 2017). The linkage of sea otters to seagrass and salt marsh health could help motivate public support for conservation and restoration of the entire ecosystem.

While estuaries provide suitable habitat for sea otters, they also pose different threats to sea otters than the open coast because of their proximity to areas of intensive human use. Pollutant inputs from adjacent land use activities are often high, and pollutants can be concentrated in shallow estuarine waters with long residence times. For instance, extremely high dichlorodiphenyltrichloroethane (DDT) levels were measured in male sea otters from Moss Landing at the mouth of the Elkhorn Slough Estuary (Jessup et al., 2010), although these high contaminant loads have not been linked to lower rates of survival. Sea otters at Elkhorn Slough are also affected by roads that disrupt the connectivity of historic estuarine wetlands (Fig. 4C). At two sites near the mouth of the Elkhorn Slough Estuary, sea otters routinely cross roads that bisect estuarine habitats; one sea otter was killed after being hit by a car in 2016, leading to efforts to slow traffic and enhance driver awareness in these areas. Depending on the size of the estuary, speeding boats and shipping traffic may also pose a risk to sea otters, either through direct strikes or through the potential for an oil or other chemical spill. Finally, repeated disturbance throughout the day caused by kayakers or eco-tourism boats may deprive sea otters of needed rest, deplete their energy reserves, and increase their vulnerability to other stressors (see https://www.seaottersavvy.org).

Elkhorn Slough provides a unique perspective when considering suitable estuarine habitats for sea otters outside their range. However, there are distinct challenges of human overlap in estuaries depending on the region and the individual estuary. San Francisco Bay, for example, has undergone dramatic physical and biological alterations as a result of human uses both past and present (San Francisco Bay Subtidal Habitat Goals Report, 2010). The Bay hosts four major shipping ports, two crude oil refineries, and a network of commuter ferries, in addition to many recreational boaters and wind surfers. Sea otters in San Francisco Bay would be exposed to the threat of injury from vessel traffic and could incur chronic fitness costs associated with repeated disturbance or displacement from preferred habitat. Heavy metal and organochlorine contaminants such as methylmercury, polychlorinated biphenyls, and DDT persist in high concentrations in San Francisco Bay (SFEI, 2017), and sea otters could be exposed to these chemical pollutants through consumption of contaminated invertebrate prey (Jessup et al., 2010). Large oil spills, although infrequent, are a potentially devastating threat to sea otters. Two oil refineries (Chevron and Phillips 66) operate on the Richmond waterfront of San Francisco Bay, receiving shipments of crude oil via tankers. San Francisco Bay experienced a major oil spill event in 2007 when the M/V Cosco Busan struck the Bay Bridge and spilled 53,569 gallons of fuel oil into the bay. Other threats associated with this urbanized estuary include an influx in bacterial disease associated with wastewater treatment plant discharges, entanglement in recreational or commercial fishing gear, exposure to harmful algal toxins, and destruction of foraging habitat and invertebrate prey by underwater dredging and sand mining operations.

If the reintroduction of sea otters to one or more estuaries were formally proposed, the engagement of the public and stakeholders could be used to fully explore the implications, including potential direct and indirect effects. Decision support tools and analyses that consider the full spectrum of trade-offs would be valuable to accompany and inform management decisions. Marine spatial planning, a process that enables decision-making around conservation trade-offs, is one such tool that can aid in the gathering and disseminating of information regarding the issues surrounding sea otter reintroduction into estuaries. This process could be especially useful when the needs and values of stakeholders become incompatible with conservation objectives. In Australia, Atwood et al. (2015) demonstrated the trade-offs between the recovery of shark predators, the abundance of herbivore bioturbators, and the seagrass system’s ability to store carbon. Similarly, in the northeast Pacific, key trade-offs include the recovery of listed species (sea otters), the abundance of their prey (including shellfish resources), and the ecosystem-level effects of sea otter reintroduction. Some of these ecological effects, while complex and difficult to monetize, may tip the balance between human values and top predator recovery that have historically pitted top predators against the humans competing with them for prey, and human harvest thereof. For example, the reintroduction of grey wolves into Yellowstone National Park has resulted in ecosystem benefits, such as increased woody vegetation and stream functioning, that could outweigh the costs (i.e., predation on livestock; Ripple et al., 2014). Comparison of estuaries occupied (Elkhorn Slough and Morro Bay) and unoccupied (San Francisco Bay and Drakes Estero) by sea otters offer the potential for valuable insight on socio-ecological effects of coastwide recovery. In order to realize the prospects of healthy coastlines, including the ecological restoration of estuarine systems such as seagrasses and salt marshes, differential human values and the ecological, cultural and economic trade-offs of sea otter recovery are crucial to consider.