Effects of climate change on animal behavior and cascading ecosystem responses are rarely evaluated. In coastal Alaska, social river otters (Lontra Canadensis), largely males, cooperatively forage on schooling fish and use latrine sites to communicate group associations and dominance. Conversely, solitary otters, mainly females, feed on intertidal-demersal fish and display mutual avoidance via scent marking. This behavioral variability creates “hotspots” of nutrient deposition and affects plant productivity and diversity on the terrestrial landscape. Because the abundance of schooling pelagic fish is predicted to decline with climate change, we developed a spatially-explicit individual-based model (IBM) of otter behavior and tested six scenarios based on potential shifts to distribution patterns of schooling fish. Emergent patterns from the IBM closely mimicked observed otter behavior and landscape use in the absence of explicit rules of intraspecific attraction or repulsion. Model results were most sensitive to rules regarding spatial memory and activity state following an encounter with a fish school. With declining availability of schooling fish, the number of social groups and the time simulated otters spent in the company of conspecifics declined. Concurrently, model results suggested an elevation of defecation rate, a 25% increase in nitrogen transport to the terrestrial landscape, and significant changes to the spatial distribution of “hotspots” with declines in schooling fish availability. However, reductions in availability of schooling fish could lead to declines in otter density over time.
Citation: Albeke SE, Nibbelink NP, Ben-David M (2015) Modeling Behavior by Coastal River Otter (Lontra Canadensis) in Response to Prey Availability in Prince William Sound, Alaska: A Spatially-Explicit Individual-Based Approach. PLoS ONE 10(6): e0126208. https://doi.org/10.1371/journal.pone.0126208
Academic Editor: Emili García-Berthou, University of Girona, SPAIN
Received: August 26, 2014; Accepted: March 30, 2015; Published: June 10, 2015
Copyright: © 2015 Albeke et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: The raw simulation data can be obtained from the University of Wyoming Data archive at the following DOI: http://dx.doi.org/10.15786/M2159X. The zipped file contains 20 tab-delimited text files containing the raw simulated data, 1 Excel spreadsheet describing the content within each data table, 1 PDF describing the relational data structure of the database, and 1 VB.NET application directory housing the entire code, GIS and database products required to run an individual simulation model. These data are 17.1 GB in size.
Funding: Funding for this project was provided by the National Science Foundation (http://www.nsf.gov/) DEB #0454474, the University of Georgia (http://uga.edu) Warnell School of Forestry and Natural Resources (http://www.warnell.uga.edu/), and the University of Wyoming (http://uwyo.edu). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Forecasting changes in species distributions, migration patterns, population dynamics, and resiliency in response to predicted alteration of global climate has been in the forefront of ecological studies for the past few decades [1,2,3,4,5,6,7]. These investigations range from correlative empirical studies to complex modeling, as well as combinations of the two [8,9,10]. For example, using empirical data on abundance, survival and habitat selection of polar bears (Ursus maritimus) [11,12], in conjunction with stochastic population models parameterized with sea ice loss based on global circulation models, Hunter et al.  projected a 0.8–0.94 probability of extinction of the Beaufort Sea population by the year 2100.
Evaluation of the effects of climate change on animal behavior is mostly limited to migration and breeding phenologies [4,6,14,15,16]. A notable exception is the study of gray wolves (Canis lupus) on Isle Royal, Michigan, where a relation between snow accumulation and social grouping was established. Hunting by larger wolf packs facilitated an increase in kill rates of moose (Alces alces). As a result, the moose population declined facilitating an increase in the growth of balsam fir (Abies balsamea) . The paucity of empirical and modeling studies on the potential effects of climate change on animal behavior and cascading ecosystem responses [17,18] is surprising. Because individuals are the building blocks of inherently complex ecological systems  and provide a natural scale at which to measure biotic and abiotic interactions , this seems an appropriate level for climate change investigations. Individuals are limited behaviorally and physiologically and their responses may be more predictable and easier to model than that of a population . Additionally, individuals respond to internal and external environments by seeking to maximize ‘fitness’ through adaptive behavior, leading to the emergence of system level properties [19,21,22].
Individual-based simulation models (IBM), which treat individuals as unique and discrete entities, have been used since the 1970’s . These discrete entities have several unique characteristics, such as age, that change during the cycle of the model . IBMs have several advantages over analytical and stochastic system models, including variability among individuals, local interactions, complete life cycles [19,23], and responses to previous and current states (i.e. Markovian dependencies). Also, an emergent property of an IBM is the overall system stochasticity, precluding the need to combine the effects of multiple variance components associated with dynamic system models [24,25]. Further, spatially explicit IBMs integrate individual responses with landscape heterogeneity by specifying the explicit location of entities and their spatial relationship to other landscape features . These advantages allow testing of theory under many different conditions, an attribute typically not available in natural systems . This attribute is especially desirable given the myriad of potential climate change scenarios.
Similar to seabirds [27,28,29,30], piscivory by coastal river otters (Lontra canadensis) provides a pathway for nutrient transport between sea and land . Marine-derived carbon (C), nitrogen (N), and phosphorus (P) transported by river otters to terrestrial latrine sites (specific locations along the shoreline) can be several orders of magnitude higher than other nutrient inputs in this system [32,33]. Uptake of marine-derived nutrients (MDN) associated with river otter activity increases photosynthetic capacity of the overstory layer of coastal conifer forests .
Coastal river otters exhibit atypical social behavior compared with other mammals [35,36]. In this system males occur in large groups (3–18 otters) that increase foraging efficiency on schooling pelagic fish within the nearshore environment [35,36,37,38]. Group size depends on the availability and spatial distribution of these pelagic fish . In contrast, female otters and some males remain solitary year round, foraging primarily on intertidal-demersal fish. These individuals occasionally join a male group to opportunistically forage on pelagic fish, most likely because schooling pelagic fish have a higher energy density than the intertidal-demersal ones [35,39].
In coastal Alaska, social otters frequently use specific latrine sites as communication centers, advertising group association and dominance [37,40]. In contrast, nonsocial otters visit numerous latrine sites at low frequency, likely facilitating mutual avoidance . These behavioral differences among individuals are determined by otter demography (abundance and sex ratio) and the distribution of pelagic fish in the nearshore environment. Extant spatial and temporal variation in the availability of pelagic fish leads to shifts in otter-mediated nutrient flux from sea to land, potentially making this system highly sensitive to future climate change.
The nearshore environment of coastal Alaska supports a diverse fish community composed of two distinct groups: resident intertidal-demersal species and migratory pelagic species. The intertidal species, primarily Cottidae, Scorpaenidae, Hexagrammidae, Cancridae, together with invertebrates such as mussels (Mytilus trossulus) and crabs (Metacarcinus gracilis, M. magister and others), are a ubiquitous, non-migratory prey [35,37,41,42]. In contrast, Salmonidae, Ammodytidae, Clupeidae, and Gadidae arrive in the nearshore environment to spawn [35,43]. These schooling pelagic fish species typically begin spawning in early May and return to the open ocean or expire (salmon) by November [42,44,45]. The responses of intertidal species to future climate change are unknown . Nonetheless, decadal surveys conducted in the Gulf of Alaska, as well as species specific studies, demonstrate that schooling pelagic fish, who are cold water specialists, disappear from shallow coastal areas accessible to river otters with increasing sea-surface temperatures [47,48,49]. Thus, warming temperatures may result in decreases in availability of these fish, leading to shifts in otter sociality and nutrient transport to the terrestrial landscape.
Our goal was to investigate the effects of variation in the availability of schooling pelagic fish on nutrient transport by coastal river otters via a spatially explicit IBM. We tested six simulation scenarios based on potential shifts to spawning patterns of schooling fish in relation to warming sea-surface temperatures. We compared model results with empirical data we collected in this system in 2006 and 2007 as well as those reported in previous studies [35,36,37].
The study area, of approximately 240 km2, encompassing 143 km of coastline, is located in the southwestern portion of Prince William Sound (PWS), Alaska and includes four islands: Knight Island (60.47 N, 147.75 W), Disk Island (60.49 N, 147.65 W), Ingot Island (60.53 N, 147.64 W) and Eleanor Island (60.55 N, 147.59 W; Fig 1). The region has a maritime climate with cool and wet summers followed by winters of deep snow accumulation . The coastal landscape is typically snow-free from early May to early November. The structure of the coastline is primarily steep and rocky with some flat, low gradient beaches and numerous bays and inlets . The coastal vegetation is predominantly old-growth forest of Sitka spruce (Picea sitchensis) and western hemlock (Tsuga heterophylla), with a well-developed under-story layer comprised of Oplopanax horridus, Vaccinium spp., Menziesia ferruginea, and Rubus spp. [31,34].
The study area coastline and the virtual lines are paths along which otters can move in the model. Outside area coast is not available to otters in the model.
The bathymetric gradient of the nearshore environment is highly variable, ranging from near vertical to slopes of only a few degrees. The substrate is also variable with sizes ranging from large boulders to fine sediment . Large tidal fluctuations in this habitat (annual maximum tide of 4.66 m and a minimum tide of -1.13 m ) greatly affect the vegetative community. Two kelp species (Agarum cribrosum and Laminaria saccharina) dominate within sheltered bays and less exposed coastline [41,53]. On exposed points, bull kelp, Cereocystis luetkeana, comprises the canopy and Laminaria bongardiana the understory [41,53]. Eelgrass, Zostera marina, grows on softer substrate usually found in inner bays [41,54,55,56]. The majority of the intertidal region is dominated by Fucus gardneri, interspersed with red and green algae [41,57,58].
As semi-aquatic mammals, coastal river otters rarely forage far from shore and even less commonly venture inland . Therefore, their home range sizes are often calculated based on length of shorelines rather than area [51,59]. To model movements of river otters in our study we created a network of paths along the coast with corridors connecting individual islands or large bays on the same island . This landscape network depicted the most probable paths used by otters (Fig 1).
To describe the terrestrial and marine habitats within the network, we obtained IKONOS 1m panchromatic stereo-pairs and 4m multispectral satellite imagery for the study area (GeoEye, Thornton, CO). Initially, using the Leica Photogrammetry Suite (LPS) within ERDAS IMAGINE (ERDAS, Inc., Norcross, GA), an existing Digital Orthro Quarter-Quad (DOQQ) aerial image was used as a reference for creating tie-points linking the two 1m panchromatic IKONOS stereo and multispectral images. From the georeferenced images, we derived three separate terrestrial datasets: 1) the coastline was digitized at a 1:1,500 view scale (245.3 km of shoreline), 2) supervised classification of five land cover classes (alder, conifer, muskeg, rock and water) was completed, and 3) a 10m digital elevation model (DEM) was derived . Additionally, the marine portion of the landscape used by otters was created from bathymetric sounding points obtained through the National Geophysical Data Center . We developed a 10m bathymetric model using the Inverse Distance-Weighting (IDW) algorithm.
We used the one marine and three terrestrial datasets to develop a suite of landscape variables describing the coastal and nearshore environment . To the best of our ability, we chose metrics following Bowyer et al. [50,51] and Larsen . Each metric was calculated for every 10m interval along the coastline (point-location). We employed Maximum Entropy [64,65] to estimate the probability (MEP) of each point-location used as an otter latrine . During each simulation, MEP was used as a surrogate for otter habitat quality.
To complete the construction of the landscape network, we appended an additional 80.3 km of ‘virtual lines’ to the coastline network (Fig 1) to act as travel corridors between islands and across large bays . The virtual lines were constructed through a multiple step process. First, we created Thiessen polygons using the 10m point-locations and converted into lines. For approximately every 2 km of coastline, the line connecting two islands or a bay was retained and the excess removed. The remaining virtual lines were slightly modified to create a straight line with only two vertices. Network Analyst tools within ArcGIS (ESRI, Redlands, CA) were used to identify network nodes and populate the adjacency table describing the connectivity of network edges. Portions of the landscape network were initially attributed as either within or outside the study area. The study area comprised 58% of the total available coastline within the landscape network.
We simulated six separate scenarios, based on abundance and spatial distribution of schooling pelagic fish, each replicated 100 times (Table 1). Each simulation was run using an hourly time step, beginning at 12:00am May 15 and running to 12:00am August 16 of the simulation year. Parameters included otter sex, activity-state (active or resting), number of hours in current activity-state, defecation-state (defecated or not), number of hours since defecating, satiation-state (fed on pelagic fish school or not), and spatial location (Table 2). Otter movements were simulated along the landscape network (Fig 1) which had three state variables assigned to each point-location (10m section of coastline): 1) the abiotic habitat quality (likelihood of being an otter latrine), 2) a radial-extent scaling factor (the ratio between the expected and actual network distances), and 3) potential pelagic fish spawning habitat (S1 Text).
In all simulations, we constrained female movements by delineating a 50% core area along the landscape network (i.e. ‘edge’), with a point representing the center of their home range acting as an attractant. That is, once a female ventured beyond the edge, her next movement oriented toward the central location. Female core areas were exclusive because empirical studies have shown that females have low spatial overlap and distinct core areas of use . Male movements along the network were unconstrained in that we allowed male movements to overlap female core areas, as well as areas occupied by other males .
The model simulates individual otter movement and behavior through foraging on both intertidal-demersal and schooling pelagic fish. Both male and female movements followed a Biased Correlated Random Walk (BCRW). Correlated random walks (CRWs) are those successive movements that have correlated directions . In contrast, successive movements characterized by a consistent directional bias are termed BRCW [67,68,69,70]. In our case, the bias was composed of directional movements toward the central location of the female’s home range and the nearest latrine site within 1km of a successful encounter with a fish school for both sexes. Therefore, foraging behaviors were influenced by the presence of otter feces and/or fish schools. For example, an otter would continuously move along the landscape network within a search distance drawn from a distribution for each hour of activity, or until it reached within 50m of a fish school (Table 2). If an otter did not encounter a fish school in a given activity bout, it was assumed to have consumed intertidal-demersal fish. The model also accounts for post-absorbance resting and olfactory communication processes (Fig 2). At each time step, the activity-state of each otter was assessed with simulated behaviors occurring only when otters were active (Fig 2). The direction of movement along the network was determined by the detection of feces within 1km of the current location, simulating olfactory communication among group members (Table 2).
Values within parentheses indicate model equations to determine individual otter choices (S1 Text).
Each foraging event was followed by defecation behavior driven by the defecation-state, satiation-state and spatial location of the individual. For example, an otter had an increased probability of defecating at the nearest highest-quality latrine after a successful encounter with a fish school, simulating successful feeding, or after 8–9 hours have elapsed in which the otter is assumed to have fed on intertidal-demersal resources (Table 2). The maintenance of an individual current activity-state (active or inactive) was driven by the hours within the current activity-state and the satiation-state. For example, duration of resting at a static location (i.e., den) was a function of the time elapsed since entering the inactive-state and whether the otter successfully encountered a fish school in the preceding foraging bout (Fig 2; Table 2). For an exhaustive description of model process, following the protocol by Grimm et al. , see S1 Text.
Parameter mean values were held constant for the entire simulation given the conditions of the tested scenario, while for each individual at each time step the actual values were randomly drawn from the appropriate distribution (Tables 1 and 2; S1 Text). Response variables included defecation rate (proportion of hours each individual experienced a defecation event during the simulation), hours of activity (number of hours in the active-state per otter), fish school foraging success (number of occurrences in which an otter successfully located a fish school), social group interaction (percent time within a social group; group size), home range size (calculated as the length of shoreline), and coastline use (total number of fecal deposits per latrine site; S1 Text). We compared the results for these response variables for each schooling fish scenario (see below) to published empirical data [35,36,37,51,71,72] and those we collected in 2006 and 2007 in the same area.
Schooling Fish Scenarios
To test the effects of potential climate change, we simulated six scenarios in which the availability and spatial distribution of schooling pelagic fish varied (Table 1). In the first scenario (Schools_100%), the hourly availability of fish schools to foraging otters was drawn from the maximal range of observed densities [35,44] based on the daily spawning patterns of these species as quantified in the late 1990s (S1 Text; eq. 7 and 8). The hourly landscape position of these schools was determined by point-locations characterized by spawning habitats with a depth ≤ 3m [73,74]. In the second scenario (Random_100%), availability was similar to the first scenario, but the spatial location was not limited by spawning habitat. In the remaining four scenarios, the availability of fish schools declined in 25% increments to 0% (Schools_75%, Schools_50%, Schools_25%, School_None) with locations constrained to spawning habitat. These scenarios were based on predictions that with increasing sea-surface temperatures, the abundance of spawning pelagic fish will decline in the nearshore environment [47,48,49].
The density of otters within the study area was previously estimated to range between 0.28 to 0.8 otters/km of coastline (40–115 otters) . Recent abundance estimates of resident otters suggested a range of 55–78 animals for the same area . Thus, at the beginning of each simulation, the number of animals within the study area was randomly drawn from a uniform distribution bounded by these values. Estimates of otter demographic parameters within the study area were then extrapolated to the entire landscape network. For example, the total number of otters in the simulation was calculated by adding individuals to the resident population by multiplying the randomly drawn abundance by 0.585, which represents the fraction of coastline outside the study area.
Prior to simulation, each otter was assigned as male or female based on sex ratio of 69% male, 31% female, derived from previous studies [36,76,77,78]. The activity-state was assigned using a Bernoulli trial with ‘Active’ probability calculated as the ratio of average hours Active:Inactive (1.43:11.69; Table 2; S1 Dataset). The number of hours at the current activity-state was randomly assigned using a uniform distribution bounded by 0 and the upper 95% CI (1.88:15.32; Table 2; S1 Dataset) for each activity-state, rounded to the nearest integer. Ormseth and Ben-David  found that captive otters defecate, on average, once every 4.9 hours. Thus, the number of hours since last defecating was randomly drawn from a uniform distribution bounded by 0 and 5.
The final initialization step for each otter was its placement onto the landscape. This step was performed separately for the study area and out-of-area portions of the network. Point-locations were filtered to include only habitat values > 0.464 MEP, reflecting habitat characteristics of latrine sites favored by otters . Male placement was unconstrained. Female home ranges were randomly drawn from a truncated normal distribution (value > 0) based on 50% core-area length of 4km (± 2SD) of coastline , with the central point-location adjusted using the radial extent scaling factor (S1 Text). Female 50% core areas on the landscape network did not overlap [71,80].
The timing and magnitude of spawning migration of pelagic fish to the nearshore environment varies annually by species [35,44,45,81,82]. Because of this variation, we did not differentiate between species of schooling fish. Using georeferenced, aerially identified fish school data provided in Ben-David et al. , the number of fish schools within 100m of the coastline, during a one-day period, were counted for years 1996–1999. The minimum (40) and maximum (98) number of schools were used to set the bounds for a uniform distribution depicting the maximum number of schools available during each simulation. The timing of fish school entry into the simulation is described by eq. 7 (S1 Text). The initial landscape position of fish schools differed for the six scenarios.
To estimate the relative importance of model parameters on simulated otter behavior, we conducted a sensitivity analysis using the School_100% scenario. In this analysis, we adjusted input values of the following parameters: foraging movement distance, visual detection distance, hours between defecations, hours actively foraging, hours inactive, scent detection distance, scent decay rate of feces, latrine location memory (attraction to the highest-quality latrine within 1km), change in activity state following fish school encounter, and change in defecation frequency following fish school encounter (Table 2; S1 Text). For all parameters for which we had empirical means and standard deviations (SD), the input values were adjusted by ± 10% by multiplying the mean by 1.1 and 0.9, respectively. To calculate the new SDs we multiplied the adjusted means by the coefficient of variation (CV). For those relationships in which we relied on expert knowledge (i.e., latrine location memory, change in activity state following fish school encounter, and change in defecation frequency following fish school encounter; Table 2) we adjusted the input variables by 50% to ensure a range of all possible conditions.
This study was conducted in the Chugach National Forest and authorized under a Special Use Permit #GLA832 (expires on 12/31/2016). This project did not involve the use of vertebrate animals and did not require authorization by Institutional Animal Care and Use Committee. In 2006, while surveying the study area shoreline we identified 320 active river otter latrine sites. Of these sites, we selected 100 for monitoring fecal deposition using a stratified random sampling. We first stratified sites by activity level (> 100 scats—high use, < 100—low use) and then by location within the study area (Herring Bay, Lower Passage, and Northwest Bay of Eleanor Island). We re-visited these sites nine times between May and August and counted all new deposits at each sampling visit [34,76]. To ensure that we only counted fresh feces, we marked all present ones with crafting glitter (Glitterex Corp., Cranford, NJ) and counted only unmarked deposits at every subsequent visit. In 2007, we again counted fecal deposits on the same 100 latrine sites during five visits between May and August. Using the total fecal counts per site per visit in each year we calculated the mean fecal counts per day (mean feces/day).
The data from each replicated simulation scenario were compiled into a single MS SQL Server database. Data, including spatial location, were stored for each otter and fish school for each simulated hour. To account for variability, data were initially summarized for each replication and then for the entire simulation scenario. We compared results from the six fish-school scenarios by evaluating overlap of 95% CI for defecation rate, hours of activity, fish school foraging success, percent time within a social group, group size and 50% home-range size by sex. Home range size was estimated using Brownian Bridges . We also compared values generated by these scenarios with published empirical data [35,36,37,51,71]. Similarly, we calculated mean and 95% CI for fecal counts per day for each scenario and compared those to counts we obtained in 2006 and 2007 in the same area.
To assess changes to patterns of nutrient deposition on the terrestrial landscape from fish school availability and distribution, we used Detrended Correspondence Analysis (DCA) [84,85]. In this analysis the mean number of feces deposited on each 50m of shoreline was calculated for every simulation by scenario. For each simulation (100 replicates), we repeated the DCA comparing the similarity in fecal deposition between scenarios. We calculated the mean and 95% CI and evaluated their overlap using values extracted from the first two axes of each DCA.
To assess the relative importance of the various model parameters, we created a tornado diagram using mean feces/day as the response variable. The diagram displays the range of mean response values and their associated 95% CI relative to the mean and 95% CI of the School_100% scenario. All statistical analyses were performed using Program R 2.15.1 .
Simulation scenarios were successfully completed for twenty-six separate parameterizations, each replicated 100 times. Each simulation replicate required approximately 15 hours to complete. The data from all simulation scenarios were captured within an SQL Server database composing a total of 2,695,027,086 records.
Comparison to Observed
Model results from all scenarios closely matched observed patterns from empirical studies (Table 3). For example, otter abundance for the entire landscape was within the range of estimated values from non-invasive genetic sampling in 2006 (Table 3) . Similarly, maximum observed group size ranged from 9–11 individuals ; our models yielded a range from 7–15 (Table 3). In addition, the percent of time social females spent in mixed-sex groups was 77.5% based on empirical data  and ranged from 82.1–83.0% in our simulations (Table 3). Finally, the percent of simulated feces containing pelagic fish for School_100% and Random_100% was equivalent to observed frequencies (Table 3) . The only discrepancy occurred in 50% core home-range size for males where model results were 2 to 4 times higher than observed (Table 3) .
Mean feces/day, as estimated for all scenarios, ranged from 298.5 (±6.7 95% CI) to 391.5 (±7.7). This was within the range of observed rate of fecal deposition of 345.4 feces/day during 2006 and 365.6 feces/day in 2007 (Fig 3). While the proportion of “hotspot” deposition sites (> 150 feces) in the model was similar to that observed, the proportion of intermediate deposition sites (50–150 feces) was underrepresented in the model results (Fig 3).
Mean feces/day was most sensitive to variation in activity-state following a fish school encounter and distance to nearest high-quality latrine (Activity State (fish) and Memory (m); Fig 4). The least influential parameter was Scent Decay Rate (Fig 4). Several parameters exhibited a skewed response to perturbation of the input values. Increasing the Hours Between Defecation events yielded a higher than expected reduction in the mean feces/day. Similarly, reducing the Hours Inactive disproportionately increased the mean feces/day (Fig 4).
The solid-black vertical line represents the mean (grey-dashed 95% confidence intervals) for the School_100% scenario (308 feces/day; baseline). For each adjusted parameter, the horizontal bar represents the mean values for two scenarios (lower and upper; see Tables 1 and 2). Each simulation scenario was replicated n = 100.
Schooling Fish Scenarios
Changes in the availability and distribution of pelagic schooling fish resulted in significant changes to sex-specific behaviors of coastal river otters and associated fecal deposition rates. As expected, with a reduction in fish school availability, encounters with schooling fish by simulated otters drastically declined (Fig 5). Female encounters were significantly lower than males only for the School_100% and School_75% scenarios, suggesting that male otters assumed foraging strategies similar to females when fish schools became scarce (Fig 5). There was no sex-related difference in encounter rates when fish schools were randomly placed along the coastline. In this scenario, female encounter rates were significantly higher than when fish schools occurred in spawning habitat only (Fig 5). Concurrently, foraging time increased with decreasing fish school availability and followed similar trajectories for both sexes (Fig 5).
Dark circles represent mean values for females and light circles for males, bars represent 95% confidence intervals (n = 100).
The mean number of otter-groups decreased as fish schools declined in availability, with the steepest rate of change occurring between School_50% and School_25% (Fig 6). Concurrently, the mean number of otters within social groups declined linearly (Fig 6). When fish schools were randomly placed along the coastline, the mean number of otters per group was significantly lower compared to School_100% and School_75% (Fig 6), although the difference amounted to only 2.4%. It is important to note that simulated otters had a tendency to aggregate into groups even without the presence of fish schools. The sexually dimorphic grouping behavior empirically observed in otters was replicated within the simulations, with males spending significantly more time in social groups than females for each scenario (Fig 6). As with group size and mean number of otters, time spent in groups declined with decreasing schooling fish availability, with no discernible difference between sexes.
Females represented by dark circle, male by light circle. Bars represent 95% confidence intervals (n = 100).
The mean defecation rate by otters significantly increased with decreasing availability of schooling pelagic fish, and was approximately 25% higher between School_None and School_100% (Fig 7). This translates into an increase of approximately 156 kg of MDN reaching terrestrial latrine sites annually (from estimated 580 kg School_100% to 736 kg School_None; assuming each deposit is equivalent to 5.15g of N as in Ben-David et al. ). In addition, the distribution of fecal deposition onto the landscape varied among scenarios, with an unexpected correspondence in the distribution of feces between Random_100% and School_25% (Fig 8). The proportion of “hotspot” sites with > 150 feces increased by 25% in response to declining availability of schooling fish (Fig 3). Concurrently, the proportion of sites with 50–150 feces increased by 125% although the distribution of “hotspots” on the landscape still deviated from observed patterns in both 2006 and 2007 (Fig 3).
Dark circles represent mean values for females and light circles for males, bars represent 95% confidence intervals (n = 100).
The emergent patterns from the IBM we developed appeared to closely mimic observed otter behavior and landscape use, suggesting that the decision rules and model parameters we chose were a close approximation of conditions experienced by wild otters. For most response variables, except for male 50% core home-range and the proportion of “hotspot” sites, model estimates were within observed empirical ranges. The advantages of developing this IBM are clearly evident from the emergence of otter sociality without an explicit imposition within the model. However, these advantages may be overshadowed by model complexity and enormity of the resulting dataset. Having developed this model, we were able to forecast the potential effects of climate alteration via changes in schooling fish availability and demonstrated a significant change in the transport of MDN to the terrestrial environment in both quantity and distribution. Nonetheless, because our model was restricted to a single season, some of the emergent properties, such as the surge in fecal deposition rate, are likely a short-term response. In this model we did not account for a potential decline in demersal-intertidal fish given increased predation by otters in the absence of schooling pelagic fish; nor did we account for the likely decline in otter density over time. To fully understand the dynamics of this system under differing climate scenarios, this model will need to be extended to include multiple years and account for otter population vital rates.
The IBM approach is grounded in the belief that adaptive behaviors of individuals emerge as patterns at the system level . To facilitate the expansion of IBMs in ecological theory, Uchmański and Grimm  proposed four criteria for a model to be considered an IBM: 1) the degree to which an individual’s life cycle is reflected; 2) the dynamics of individual resource use are explicitly represented; 3) real or integer numbers are used to represent population size; and 4) the extent to which variability between same age individuals is considered. We believe that our IBM meets these criteria: 1) although individual annual life cycle is not explicitly addressed, each individual’s daily activity is accounted for on an hourly basis. An individual uses stochastic rules, based on previous experience and states, to determine whether to forage or rest; 2) each individual may have the opportunity to locate and prey upon pelagic fish schools. Concurrently, pelagic fish schools are dynamically interacting with the physical environment instead of behaving in a static manner; 3) pelagic fish school densities are accounted for using real numbers and seasonal models. Otter density is estimated from a known distribution of the study area and; 4) each individual experiences both temporal and spatial variability throughout the modeling process. The variability may occur through spatial location (habitat quality), prey availability, and unique hourly experiences influencing decisions that affect current and future states. Thus, while we restricted this model to a single season and did not include otter vital rates in this simulation, it can be considered a special case of an IBM.
Our IBM produced separate social patterns for male and female otters in the absence of explicit rules of intraspecific attraction or repulsion. That is, foraging male otters were attracted to latrine sites containing fresh feces but not directly to each other, whereas females were attracted to the center of their core home ranges which did not overlap, indirectly creating repulsion. It appears that these simple rules are a self-organizing mechanism that generates sex-specific social groups in otters. Simulations of flocking and schooling behaviors in multiple species have been generated with relatively few rules of engagement, and groups were maintained without obvious fitness benefits [88,89,90]. Nonetheless, in such models the self-organizing rules include specific code for intraspecific synchronization and cohesion [88,89,90,91]. In contrast, in our modeled otter sociality and grouping behaviors emerged from foraging behavior loosely constrained by prey distribution on the landscape, suggesting that in some systems the formation of groups may be a by-product of resource distribution [92,93]. Nonetheless, while the number of simulated otter-groups and the time males and females spent in social groups declined with the reduction in schooling fish availability, sociality did not completely disappear with this resource. This agrees with observations of group formation in otters inhabiting freshwater systems. For example, Melquist and Hornocker  observed small groups of otters in Idaho and Serfass  recorded cooperative foraging in Canada. Thus it seems that in our model, as well as natural systems, other factors related to otter activity and landscape use foster aggregations.
To evaluate if other factors contribute to otter sociality, it will be imperative to assess whether simulated otters exhibited fidelity to certain groups. Although groups of wild river otters are formed at random with respect to kin [36,96], individuals exhibit high fidelity to their group . Blundell et al.  observed that during the mating season (which ends prior to the initialization of our model in May) several adult males dispersed to adjacent areas, but returned to their home range and re-joined the group they previously left approximately a month later. Similarly, male otters that exhibited high levels of dyadic interactions while in captivity were found in close spatial proximity post-release . Hansen et al.  hypothesized that familiarity was the process influencing group cohesion in wild otters and suggested that male otter pups may become familiar with neighboring male groups via olfaction when visiting latrine sites with their dames. If so, we would expect otters to produce individually distinct olfactory signals that are recognizable by others . Rostain et al.  demonstrated that river otters are able to distinguish male and female feces as well as recognize the social status of animals relative to their own. These observations suggest that river otters excrete individualistic scent and are able to recognize the scent of others. Kean et al.  have shown that the feces of Eurasian otters (Lutra lutra) contain compounds unique to adults and juveniles as well as sex-specific ones. It is reasonable to assume that river otter feces contain similar compounds and that both species also have individually-unique scent.
Despite the high fidelity, river otter groups in the wild exhibit numerous fission-fusion events while foraging  and the structure of the social network varies through time . In the future we will use results from our model to assess group fidelity, structure of the social network, and frequency of fission-fusion events of simulated otters and compare them to empirical data derived from radiotelemetry, non-invasive genetic analyses, and Encounternet proximity sampling [72,76,99]. Should we find that simulated otters show little group fidelity, we will develop an additional decision rule for the model creating attraction to familiar individuals based on individualistic scent. Such a rule may correct the discrepancy between observed and modeled male home range size.
Inherent to our modeling rules was the assumption that consumption of schooling pelagic fish confers fitness benefits, although no immediate advantage was observed among wild otters. Based on the observation that the number of offspring and relatives in the population did not differ between social and solitary animals, and otters adopting either strategy have similar size and condition, Blundell et al.  concluded that sociality did not produce fitness benefits. These authors hypothesized that the two social strategies (social and solitary), within the same population, persist because of large temporal fluctuations in the availability of schooling pelagic fish . Testing of this hypothesis with empirical data will be impossible because the simultaneous collection of fish and otter data over an extended period of time will be impractical and inordinately expensive. Our results suggest that indeed when the availability of schooling pelagic fish declines, male otters forage and use latrines in a fashion resembling the behavior of females. Thus, once extended to include multiple years and account for otter population dynamics, our model could be used to test this and associated hypotheses.
Nonetheless, our model will require several adjustments before it can be extended. First, if simulated otters do exhibit high group fidelity, we will likely need to add a constraint on male movement to correct for the larger than observed male home ranges. Indeed, it is possible that our olfactory decision rule for male otters may have not been correctly formulated. The assumption that all feces are the same may need to be revisited because feces containing no pelagic fish may indicate resource depletion instead of resource availability . Indeed, this may somewhat explain the drastic increases in male otter home-range sizes with decreasing availability of schooling pelagic fish.
Second, the model will require simplification because each single-season simulation took approximately 15 hours to complete. Our sensitivity analyses demonstrated that variation in Scent Decay Rate, Hours Active, and Visual (m) had relatively little influence on model outcomes, so these variables could potentially be made constant. It may also be possible to reduce the effect of the variable Memory (m), which forced the otters to travel to and potentially defecate at the nearest high-quality latrine within 1 km, rather than use any available latrine in their immediate vicinity. We created this decision rule because our field studies have shown that otters exhibit high fidelity to specific latrines, which are visited by multiple generations of otters . Spatial memory has been documented from sharks to primates [101,102,103,104,105] and inclusion of such a parameter in models of animal movements has improved their accuracy [106,107,108]. From this perspective it is not surprising that this parameter had such strong influence on our results, although we may still capture the effects of memory by allowing otters to travel to the nearest latrine.
Indeed, this latrine selection rule probably caused the emergence of lower number of “hotspot” sites with 50–150 feces on the landscape as compared with our observed data. These emergent spatial properties of latrines were likely affected by the otter movement rule rather than the spatial placement of fish schools because the mean feces/day and the proportion of “hotspot” sites was similar between the Random_100% and the Schools_100% models. Only the actual location of used latrines was different between these two scenarios likely because foraging success increased when fish school locations were not restricted to specific spawning areas. Specifically, females experienced significantly higher success in locating fish schools than when these prey were patchily distributed. Because females were driven by different decision rules with regards to use of landscape, the distribution of latrines differed in this scenario.
Two processes may have influenced the lower fecal deposition rate we recorded in the Schools_100% and Schools_75% scenarios relative to values we observed in 2006 and 2007. First, we likely generated population estimates that were slightly biased low. We imposed a strong repulsion rule on females, basically that no two female core areas overlap. A relaxation of the rule may yield a slightly larger otter population. Also, we used the number of resident animals identified from non-invasive genetic sampling in 2006 within the study area  and then divided that value by proportion of modeled landscape that composed of the study area. The number of residents included adult animals only, whereas the wild population was augmented in July with pups emerging from natal dens . Including recruitment in the model would have likely increased the overall population size and fecal deposition rate. In fact, the spatially-explicit simulated otter population could serve as an excellent source of information for designing fecal collection protocols that will yield unbiased estimates for the wild population. Because the location of each river otter during every hour of the simulation is known, and the location and timing of its fecal excretion is also recorded and saved, we could simulate any number of collection scenarios, and using mark-recapture modeling , test which protocol (in terms of number of collection days and number of latrines sampled) provides the least biased and most accurate abundance estimate.
Second, the higher than expected fecal deposition rate in 2006 and 2007 may have resulted from the fact that we simulated fish school availability with data collected in 1996–1999 [37,44], whereas we conducted the fecal deposition study nearly a decade later. Our observed fecal deposition rate suggests that in 2006 otters may have encountered a higher frequency of pelagic fish schools than in 2007. The mean sea-surface temperature in Prince William Sound was significantly different in June 2006 than in June 2007 (http://www.ndbc.noaa.gov/), likely increasing the availability of schooling fish for otter consumption during 2006 and decreasing their fecal deposition rate. Following this rational it is reasonable to assume that schooling fish abundance in our study area was higher in the 1990s than in the 2000s. A recent stock assessment for Pacific herring illustrated poor recovery of this fishery in coastal Alaska with oceanographic factors as one of three main contributors . Thus, it is possible that schooling fish availability was lower than modeled when we collected the empirical data. This hypothesis can be tested by conducting dietary analysis on the feces we collected to generate the abundance estimates in 2006 and calculating the percent containing pelagic fish.
With a decline in schooling pelagic fish, otters may exhibit a two-pronged response. They will likely switch to more heavily prey on intertidal-demersal fish and decline in abundance. In our model we did not account for the density and distribution of intertidal-demersal fish explicitly, and assumed that benthic resources are uniformly distributed. Instead, the marine abiotic conditions that usually affect benthic fish in this system were included as variables within the model predicting latrine quality (MEP) . Dean et al.  found differences in the distribution and abundance of benthic fish given characteristics of the marine environment. It may be an enlightening exercise to develop a model using only terrestrial variables for the otter latrine selection model and couple it with an intertidal-demersal fish habitat model. Such a model will allow us to assess the impact of varying benthic fish availability on otter behavior.
Our single-season model simulating the potential effects of climate change on nutrient transports from sea to land via the abundance of schooling pelagic fish and otter behavior resulted in several unexpected patterns. Foremost was the observation of elevated defecation rate and 25% increase in nitrogen transport to the terrestrial landscape. This occurred due to an increase in duration of the Active-State which reflected search of alternative prey and increased likelihood of visiting a latrine. This is clearly demonstrated within the sensitivity analysis in which Hours Inactive and Hours Between Defecation parameters had significant effects on mean feces/day. Also unexpected was the lack of decline in number of “hotspot” sites. While male group membership declined and their activity increased, their behavior in terms of attraction to latrines did not. A concurrent increase in nutrient deposition and the number of “hotspot” sites would have had significant implications to the vegetation at river otter latrines. Roe et al.  documented increased productivity of conifers growing at river otter latrines compared with those growing on non-latrine sites. Concurrently, enhanced river otter activity caused a reduction in shrub biomass and allocation of excess nitrogen to storage in shoots . The shading from densely foliaged conifers also resulted in declines in understory plant diversity on latrines . Thus enhanced latrine visitation and increased nitrogen fertilization could influence the terrestrial landscape in this system. Nonetheless, in response to reduced availability of schooling fish, male otters are likely to change their behavior. Reduction in schooling fish availability will likely result in increased predation on intertidal-demersal fish which occur at lower biomass  and have lower energy density than schooling pelagic fish . Reduced resource availability will likely lead to declines in otter density over time. Indeed, otter density is substantially lower in ecosystems where food availability is diminished [113,114]. Thus, changes to terrestrial vegetation will be short lived. In future models we will assess the effects of climate change on nutrient transports in this system, by extending this model to include multiple years and account for otter population dynamics.
S1 Dataset. Tab-delimited text file containing summarized observations of individual otter behavior over a 24-hour period.
Column descriptions are as follows: IDNum (radio collar ID); MinOfStartTime (date/time describing the beginning time of the behavior); MaxOfStartTime (date/time describing the ending time of the behavior); DateGroup (integer identifying the unique set of observations for the individual otter); ActivityGroup (text identifying whether the otter was ‘Active’ or ‘Inactive’); SumHr (float summarizing the total number of hours in the current state for the individual otter); TotTime (float the total number of hours of observation).
S2 Dataset. Tab-delimited text file containing summarized observations of individual otter movement rates in meters per hour.
Column descriptions are as follows: Freq_StopID (text ID describing the individual collar ID and the observation ID in a ‘from-to’ format); Activities (text describing the specific otter behaviors at the ‘from-to’ locations. The codes: ‘AL’ = ActiveLand, ‘AW’ = ActiveWater, ‘AS’ = ActiveShore, ‘D’ = Dive, ‘I’ = Intertidal, ‘IL’ = IntertidalLand, ‘L’ = Land, ‘NR’ = NoVisualRadioOnly, ‘S’ = Shore, ‘W’ = Water); TimeDiff_min (float number of minutes between ‘from-to’); Dist (float number of meters traveled between ‘from-to’); m/hr (float the rate of travel given the measurements).
S3 Dataset. Excel file containing all summarized data for each simulation scenario.
Each tab represents the data for a specific response variable (described by the tab label). These data were extracted from the SQL Server database that contains over 2.7 billion rows of data.
We thank C. Meyer, J. Gulledge, and T. Balser for help with designing this project. K.E. Ott, A. Roe, J. Herreman, M. Wood, B. Myers, T. Whitaker, and K. Pope assisted with field and lab work. M. Lan, J.T. Peterson and R.L. Hendrick provided comments on early drafts of this manuscript. Special thanks to H.N. Golden, A. Christ, J. Wells and T. Rinaldi from the Alaska Department of Fish and Game for their help in establishing camp and coordinating the delivery of supplies. Logistical support was provided by the Alaska Department of Fish and Game and Babkin Charters Inc.
Conceived and designed the experiments: SEA NPN MB. Performed the experiments: SEA. Analyzed the data: SEA NPN MB. Contributed reagents/materials/analysis tools: SEA NPN MB. Wrote the paper: SEA NPN MB. Designed the software used in analysis: SEA.
- 1. Araújo MB, Peterson AT (2012) Uses and misuses of bioclimatic envelope modeling. Ecology 93: 1527–1539. pmid:22919900
- 2. Cherry SG, Derocher AE, Thiemann GW, Lunn NJ (2013) Migration phenology and seasonal fidelity of an Arctic marine predator in relation to sea ice dynamics. Journal of Animal Ecology 82: 912–921. pmid:23510081
- 3. Colchero F, Medellin RA, Clark JS, Lee R, Katul GG (2009) Predicting population survival under future climate change: density dependence, drought and extraction in an insular bighorn sheep. Journal of Animal Ecology 78: 666–673. pmid:19245378
- 4. Jiguet F, Gadot A-S, Julliard R, Newson SE, Couvet D (2007) Climate envelope, life history traits and the resilience of birds facing global change. Global Change Biology 13: 1672–1684.
- 5. Hyvönen R, Ågren GI, Linder S, Persson T, Cotrufo MF, Ekblad A, et al. (2007) The likely impact of elevated [CO2], nitrogen deposition, increased temperature and management on carbon sequestration in temperate and boreal forest ecosystems: a literature review. New Phytologist 173: 463–480. pmid:17244042
- 6. Walther G- R, Post E, Convey P, Menzel A, Parmesan C, Beebee TJC, et al. (2002) Ecological responses to recent climate change. Nature 416: 389–395. pmid:11919621
- 7. Wiederholt R, Post E (2010) Tropical warming and the dynamics of endangered primates. Biology Letters 6: 257–260. pmid:19864277
- 8. Graham CH, VanDerWal J, Phillips SJ, Moritz C, Williams SE (2010) Dynamic refugia and species persistence: tracking spatial shifts in habitat through time. Ecography 33: 1062–1069.
- 9. Stewart JR, Lister AM, Barnes I, Dalén L (2010) Refugia revisited: individualistic responses of species in space and time. Proceedings of the Royal Society B: Biological Sciences 277: 661–671. pmid:19864280
- 10. Thomas CD, Hill JK, Anderson BJ, Bailey S, Beale CM, Bradbury RB, et al. (2011) A framework for assessing threats and benefits to species responding to climate change. Methods in Ecology and Evolution 2: 125–142.
- 11. Durner GM, Douglas DC, Nielson RM, Amstrup SC, McDonald TL, Stirling I, et al. (2009) Predicting 21st-century polar bear habitat distribution from global climate models. Ecological Monographs 79: 25–58.
- 12. Regehr EV, Hunter CM, Caswell H, Amstrup SC, Stirling I (2010) Survival and breeding of polar bears in the southern Beaufort Sea in relation to sea ice. Journal of Animal Ecology 79: 117–127. pmid:19754681
- 13. Hunter CM, Caswell H, Runge MC, Regehr EV, Amstrup SC, Stirling I. (2010) Climate change threatens polar bear populations: a stochastic demographic analysis. Ecology 91: 2883–2897. pmid:21058549
- 14. Jenni L, Kéry M (2003) Timing of autumn bird migration under climate change: advances in long–distance migrants, delays in short–distance migrants. Proceedings of the Royal Society of London Series B: Biological Sciences 270: 1467–1471. pmid:12965011
- 15. Marra P, Francis C, Mulvihill R, Moore F (2005) The influence of climate on the timing and rate of spring bird migration. Oecologia 142: 307–315. pmid:15480801
- 16. Saino N, Rubolini D, Lehikoinen E, Sokolov LV, Bonisoli-Alquati A, Ambrosini R, et al. (2009) Climate change effects on migration phenology may mismatch brood parasitic cuckoos and their hosts. Biology Letters 5: 539–541. pmid:19443508
- 17. Post E, Peterson RO, Stenseth NC, McLaren BE (1999) Ecosystem consequences of wolf behavioural response to climate. Nature 401: 905–907.
- 18. Wilmers CC, Estes JA, Edwards M, Laidre KL, Konar B (2012) Do trophic cascades affect the storage and flux of atmospheric carbon? An analysis of sea otters and kelp forests. Frontiers in Ecology and the Environment 10: 409–415.
- 19. Grimm V, Railsback SF (2005) Individual-based modeling and ecology. Princeton, New Jersey: Princeton University Press. 428 p.
- 20. Pascual M, Levin S (1999) From individuals to population densities: searching for the intermediate scale of nontrivial determinism. Ecology 80: 2225–2236.
- 21. Railsback SF (2001) Concepts from complex adaptive systems as a framework for individual-based modelling. Ecological Modelling 139: 47–62.
- 22. Grimm V (1999) Ten years of individual-based modelling in ecology: what have we learned and what could we learn in the future? Ecological Modelling 115: 129–148.
- 23. Grimm V, Berger U, Bastiansen F, Eliassen S, Ginot V, Giske J, et al. (2006) A standard protocol for describing individual-based and agent-based models. Ecological Modelling 198: 115–126.
- 24. Czembor CA, Morris WK, Wintle BA, Vesk PA (2011) Quantifying variance components in ecological models based on expert opinion. Journal of Applied Ecology 48: 736–745.
- 25. Vincenot CE, Giannino F, Rietkerk M, Moriya K, Mazzoleni S (2011) Theoretical considerations on the combined use of System Dynamics and individual-based modeling in ecology. Ecological Modelling 222: 210–218.
- 26. Dunning JB Jr., Stewart DJ, Danielson BJ, Noon BR, Root TL, Lamberson RH, et al. (1995) Spatially explicit population models: current forms and future uses. Ecological Applications 5: 3.
- 27. Anderson WB, Polis GA (1998) Marine subsidies of island communities in the Gulf of California: evidence from stable carbon and nitrogen isotopes. Oikos 81: 75–80.
- 28. Hobson KA, Drever MC, Kaiser GW (1999) Norway Rats as Predators of Burrow-Nesting Seabirds: Insights from Stable Isotope Analyses. The Journal of Wildlife Management 63: 14–25.
- 29. Mulder CPH, Keall SN (2001) Burrowing seabirds and reptiles: impacts on seeds, seedlings and soils in an island forest in New Zealand. Oecologia 127: 350–360.
- 30. Croll DA, Maron JL, Estes JA, Danner EM, Byrd GV (2005) Introduced predators transform subarctic islands from grassland to tundra. Science 307: 1959–1961. pmid:15790855
- 31. Ben-David M, Bowyer RT, Duffy LK, Roby DD, Schell DM (1998) Social behavior and ecosystem processes: river otter latrines and nutrient dynamics of terrestrial vegetation. Ecology 79: 2567–2571.
- 32. Giblin AE, Nadelhoffer KJ, Shaver GR, Laundre JA, McKerrow AJ (1991) Biogeochemical diversity along a riverside toposequence in Arctic Alaska. Ecological Monographs 61: 415–435.
- 33. Lilleskov EA, Fahey TJ, Lovett GM (2001) Ectomycorrhizal fungal aboveground community change over an atmospheric nitrogen deposition gradient. Ecological Applications 11: 397–410.
- 34. Roe AM, Meyer CB, Nibbelink NP, Ben-David M (2010) Differential production of trees and shrubs in response to fertilization and disturbance by coastal river otters in Alaska. Ecology 91: 3177–3188. pmid:21141179
- 35. Blundell GM, Ben-David M, Bowyer RT (2002) Sociality in river otters: cooperative foraging or reproductive strategies? Behavioral Ecology 13: 134–141.
- 36. Blundell GM, Ben-David M, Groves P, Bowyer RT, Geffen E (2004) Kinship and sociality in coastal river otters: are they related? Behavioral Ecology 15: 705–714.
- 37. Ben-David M, Blundell GM, Kern JW, Maier JAK, Brown ED, Jewett SC (2005) Communication in coastal river otters: creation of variable resource sheds for terrestrial communities. Ecology 86: 1331–1345.
- 38. Rock KR, Rock ES, Bowyer RT, Faro JB (1994) Degree of association and use of a helper by coastal river otters, Lutra canadensis, in Prince William Sound, Alaska. Canadian Field-Naturalist 108: 367–369.
- 39. Anthony JA, Roby DD, Turco KR (2000) Lipid content and energy density of forage fishes from the northern Gulf of Alaska. Journal of Experimental Marine Biology and Ecology 248: 53–78. pmid:10764884
- 40. Rostain RR, Ben-David M, Groves P, Randall JA (2004) Why do river otters scent-mark? An experimental test of several hypotheses. Animal Behaviour 68: 703–711.
- 41. Dean TA, Haldorson L, Laur DR, Jewett SC, Blanchard A (2000) The distribution of nearshore fishes in delp and eelgrass communities in Prince William Sound, Alaska: associations with vegetation and physical habitat characteristics. Environmental Biology of Fishes 57: 271–287.
- 42. Mecklenburg CW, Mecklenburg TA, Thorsteinson LK (2002) Fishes of Alaska. Bethesda, Maryland: American Fisheries Society. 1037 p.
- 43. Larsen DN (1984) Feeding habits of river otters in coastal southeastern Alaska. The Journal of Wildlife Management 48: 1446–1452.
- 44. Brown E, Wang J, Vaughan S, Norcross B (1999) Identifying seasonal spatial scale for the ecological analysis of herring and other forage fish in Prince William Sound, Alaska. In: Ecosystem approaches for fisheries management, editor. Alaska Sea Grant College Program, AK-SG-99–01. Fairbanks, Alaska: University of Alaska-Fairbanks. pp. 499–510.
- 45. Brown ED (2002) Life history, distribution, and size structure of Pacific capelin in Prince William Sound and the northern Gulf of Alaska. ICES Journal of Marine Science 59: 983–996.
- 46. Jewett SC, Hamazaki T, Danielson S, Weingartner T (2008) Retrospective analyses of Norton Sound benthic fauna in response to climate change. North Pacific Research Board Final Report 605. 44 p p.
- 47. Norcross BL, Brown ED, Foy RJ, Frandsen M, Gay SM, Kline TC, et al. (2001) A synthesis of the life history and ecology of juvenile Pacific herring in Prince William Sound, Alaska. Fisheries Oceanography 10: 42–57.
- 48. Cooney RT, Allen JR, Bishop MA, Eslinger DL, Kline T, Norcross BL, et al. (2001) Ecosystem controls of juvenile pink salmon (Onchorynchus gorbuscha) and Pacific herring (Clupea pallasi) populations in Prince William Sound, Alaska. Fisheries Oceanography 10: 1–13.
- 49. Anderson PJ, Blackburn JE, Johnson BA (1997) Declines in forage species in the Gulf of Alaska 1992–1995, as an indication of regime shift. Fairbanks, Alaska: Alaska Sea Grant Report 97–01. 531–544 p.
- 50. Bowyer RT, Testa JW, Faro JB (1995) Habitat selection and home ranges of river otters in a marine environment: effects of the Exxon Valdez oil spill. Journal of Mammalogy 76: 1–11.
- 51. Bowyer RT, Blundell GM, Ben-David M, Jewett SC, Dean TA (2003) Effects of the Exxon Valdez oil spill on river otters: injury and recovery of a sentinel species. Wildlife Monographs 153: 1–53.
- 52. NOAA (2007) Tidal station locations and ranges.
- 53. Dean TA, Stekoll MS, Smith RO. (1996) Kelps and oil: The effects of the Exxon Valdez oil spill on subtidal algae; 1996; Bethesda. American Fisheries Society Symposium. pp. 412–423.
- 54. McRoy CP (1968) The distribution and biogeography of Zostera marina L. (eelgrass) in Alaska. Pacific Science 22: 507–513.
- 55. McRoy CP (1970) Standing stocks and other features of eelgrass (Zostera marina L.) populations on the coast of Alaska. Journal of Fisheries Research Board of Canada 27: 1811–1821.
- 56. Dean TA, Stekoll MS, Jewett SC, Smith RO, Hose JE (1998) Eelgrass (Zostera marina L.) in Prince William Sound, Alaska: effects of the Exxon Valdez oil spill. Marine Pollution Bulletin 36: 201–210.
- 57. Gilfillan ES, Suchanek TH, Boehm PD, Harner EJ, Page DS, Sloan NA (1995) Shorline impacts in the Gulf of Alaska region following the Exxon Valdez oil spill. In: Wells PG, Butler JN, Hughes JS, editors. Fate and Effects in Alaskan Waters, ASTM STP 1219. Philadelphia: American Society for Testing and Materials. pp. 444–484.
- 58. Stekoll M, Deysher L, Highsmith R, Saupe S, Guo Z, Erickson W, et al. Coastal habitat injury assessment: intertidal communities and the Exxon Valdez oil spill; 1996. American Fisheries Society Symposium. pp. 177–192.
- 59. Blundell GM, Maier JAK, Debevec EM (2001) Linear home ranges: effects of smoothing, sample size, and autocorrelation on kernel estimates. Ecological Monographs 71: 469–489.
- 60. Vuilleumier S, Metzger R (2006) Animal dispersal modelling: handling landscape features and related animal choices. Ecological Modelling 190: 159–170.
- 61. Albeke SE (2010) Influence of individual animal behavior on spatial and temporal variability in nutrient deposition. Athens, Georgia: University of Georgia. 204 p. Available: http://athenaeum.libs.uga.edu/xmlui/handle/10724/26845. Accessed 15 January 2010
- 62. NGDC (2007) Bathymetry, topography and relief. NOAA Satellite and Information Center.
- 63. Larsen DN (1983) Habitats, movements, and foods of river otters in coastal southeastern Alaska. Fairbanks, Alaska: University of Alaska. 169 p. Available: Accessed
- 64. Phillips S, Dudík M (2008) Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31: 161–175.
- 65. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecological Modelling 190: 231–259.
- 66. Patlak C (1953) Random walk with persistence and external bias. The bulletin of mathematical biophysics 15: 311–338.
- 67. Codling EA, Plank MJ, Benhamou S (2008) Random walk models in biology. Journal of The Royal Society Interface 5: 813–834. pmid:18426776
- 68. Blackwell PG (1997) Random diffusion models for animal movement. Ecological Modelling 100: 87–102.
- 69. Alt W (1980) Biased random walk models for chemotaxis and related diffusion approximations. Journal of Mathematical Biology 9: 147–177. pmid:7365332
- 70. Othmer HG, Dunbar SR, Alt W (1988) Models of dispersal in biological systems. Journal of Mathematical Biology 26: 263–298. pmid:3411255
- 71. Blundell GM, Bowyer RT, Ben-David M, Dean TA, Jewett SC (2000) Effects of food resources on spacing behavior of river otters: does forage abundance control home-range size? Biotelemetry 15: 325–333.
- 72. Blundell GM, Ben-David M, Groves P, Bowyer RT, Geffen E (2002) Characteristics of sex-biased dispersal and gene flow in coastal river otters: implications for natural recolonization of extirpated populations. Molecular Ecology 11: 289–303. pmid:11928704
- 73. Bargmann G (1998) Forage fish management plan: a plan for managing the forage fish resources and fisheries of Washington. In: Washington Department of Fish and Wildlife, editor. Olympia, Washington: Washington Fish and Wildlife Commission. pp. 77.
- 74. Penttila D (2007) Marine forage fishes in Puget Sound. Seattle, Washington: Seattle District, U.W. Army Corps of Engineers. 2007–03 2007–03.
- 75. Testa JW, Holleman DF, Bowyer RT, Faro JB (1994) Estimating populations of marine river otters in Prince William Sound, Alaska, using radiotracer implants. Journal of Mammalogy 75: 1021–1032.
- 76. Ott KE (2009) Recolonization or local reproduction? An assessment of river otter recovery in previously-oiled areas of coastal Alaska via non-invasive genetic sampling [M.S.]. Laramie, Wyoming: University of Wyoming. Available: http://gradworks.umi.com/14/76/1476893.html. Accessed 1 Mar 2010
- 77. Chapman J, Feldhamer G (1982) Wild mammals of North America: biology, management, and economics: Johns Hopkins University Press Baltimore, MD, USA.
- 78. Lariviére S, Walton LR (1998) Lontra canadensis. Mammalian Species 587: 1–8.
- 79. Ormseth OA, Ben-David M (2000) Ingestion of crude oil: effects on digesta retention times and nutrient uptake in captive river otters. Journal of Comparative Physiology B 170: 419–428. pmid:11083525
- 80. Gorman TA, Erb JD, McMillan BR, Martin DJ (2006) Space use and sociality of river otters (Lontra canadensis) in Minnesota. Journal of Mammalogy 87: 740–747.
- 81. Haegele C, Schweigert J (1985) Distribution and characteristics of herring spawning grounds and description of spawning behavior. Canadian Journal of Fisheries and Aquatic Sciences 42: 39–55.
- 82. Robards MD, Piatt JF, Rose GA (1999) Maturation, fecundity, and intertidal spawning of Pacific sand lance in the northern Gulf of Alaska. Journal of Fish Biology 54: 1050–1068.
- 83. Horne JS, Garton EO, Krone SM, Lewis JS (2007) Analyzing animal movements using brownian bridges. Ecology 88: 2354–2363. pmid:17918412
- 84. Hill MO, Gauch HG Jr. (1980) Detrended correspondence analysis: an improved ordination technique. Vegetatio 42: 47–58.
- 85. Oksanen J, Minchin PR (1997) Instability of ordination results under changes in input data order: explanations and remedies. Journal of Vegetation Science 8: 447–454.
- 86. R Core Team (2012) R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
- 87. Uchmaski J, Grimm V (1996) Individual-based modelling in ecology: what makes the difference? Trends in Ecology & Evolution 11: 437–441.
- 88. Lopez U, Gautrais J, Couzin ID, Theraulaz G (2012) From behavioural analyses to models of collective motion in fish schools. Interface Focus.
- 89. Parrish JK, Edelstein-Keshet L (1999) Complexity, pattern, and evolutionary trade-offs in animal aggregation. Science 284: 99–101. pmid:10102827
- 90. Sumpter DJT (2006) The principles of collective animal behaviour. Philosophical Transactions of the Royal Society B: Biological Sciences 361: 5–22.
- 91. Lukeman R, Li Y-X, Edelstein-Keshet L (2010) Inferring individual rules from collective behavior. Proceedings of the National Academy of Sciences 107: 12576–12580. pmid:20616032
- 92. Carr GM, Macdonald DW (1986) The sociality of solitary foragers: a model based on resource dispersion. Animal Behaviour 34: 1540–1549.
- 93. Mangel M (1990) Resource divisibility, predation and group formation. Animal Behaviour 39: 1163–1172.
- 94. Melquist WE, Hornocker MG (1983) Ecology of river otters in west central Idaho. Wildlife Monographs 83: 3–60.
- 95. Serfass TL (1995) Cooperative foraging by North American river otters, Lutra canadensis. Canadian field-naturalist Ottawa ON 109: 458–459.
- 96. Hansen H, McDonald DB, Groves P, Maier JAK, Ben-David M (2009) Social networks and the formation and maintenance of river otter groups. Ethology 115: 384–396.
- 97. Tibbetts EA, Dale J (2007) Individual recognition: it is good to be different. Trends in Ecology & Evolution 22: 529–537.
- 98. Kean EF, Müller CT, Chadwick EA (2011) Otter scent signals age, sex, and reproductive status. Chemical Senses 36: 555–564. pmid:21444931
- 99. Barocas A, Golden HN, Ben-David M (2012) Measuring associations between river otters: Evaluation of an advanced ‘Encounternet’ tracking system. Alaska, USA: Kenai Fjords National Park.
- 100. Kruuk H, Conroy JWH, Moorhouse A (1991) Recruitment to a population of otters (Lutra lutra) in Shetland, in relation to fish abundance. The Journal of Applied Ecology 28: 95–101.
- 101. Clayton N (1995) Development of memory and the hippocampus: comparison of food-storing and nonstoring birds on a one-trial associative memory task. The Journal of Neuroscience 15: 2796–2807. pmid:7722629
- 102. Holding ML, Frazier JA, Taylor EN, Strand CR (2012) Experimentally altered navigational demands induce changes in the cortical forebrain of free-ranging Northern Pacific Rattlesnakes (Crotalus o. oreganus). Brain, Behavior and Evolution 79: 144–154. pmid:22237415
- 103. Lavenex P, Steele MA, Jacobs LF (2000) Sex differences, but no seasonal variations in the hippocampus of food-caching squirrels: a stereological study. The Journal of Comparative Neurology 425: 152–166. pmid:10940949
- 104. Gibeault S, MacDonald S (2000) Spatial memory and foraging competition in captive western lowland gorillas (Gorilla gorilla gorilla). Primates 41: 147–160.
- 105. Schluessel V, Bleckmann H (2012) Spatial learning and memory retention in the grey bamboo shark (Chiloscyllium griseum). Zoology 115: 346–353. pmid:23040178
- 106. Avgar T, Deardon R, Fryxell JM (2013) An empirically parameterized individual based model of animal movement, perception, and memory. Ecological Modelling 251: 158–172.
- 107. Shumway CA (2008) Habitat complexity, brain, and behavior. Brain, Behavior and Evolution 72: 123–134. pmid:18836258
- 108. Spencer WD (2012) Home ranges and the value of spatial information. Journal of Mammalogy 93: 929–947.
- 109. Crait JR, Blundell GM, Ott KE, Herreman JK, Ben-David M (2006) Late seasonal breeding of river otters in Yellowstone National Park. The American Midland Naturalist 156: 189–192.
- 110. Amstrup SC, McDonald TL, Manly BFJ, editors (2005) Handbook of capture-recapture analysis. New Jersey, USA: Princeton University Press. 296 p.
- 111. Pearson W, Deriso R, Elston R, Hook S, Parker K, Anderson J (2012) Hypotheses concerning the decline and poor recovery of Pacific herring in Prince William Sound, Alaska. Reviews in Fish Biology and Fisheries 22: 95–135.
- 112. Roe AM (2008) The effects of coastal river otters (Lontra canadensis) on the plant community of Prince William Sound, AK. Laramie, Wyoming: University of Wyoming. 123 p. Available: http://search.proquest.com/docview/304452478/fulltextPDF/DE71CEF1BDB94027PQ/1?accountid=14793. Accessed 1 Mar 2010
- 113. Mowry RA, Gompper ME, Beringer J, Eggert LS (2011) River otter population size estimation using noninvasive latrine surveys. The Journal of Wildlife Management 75: 1625–1636.
- 114. Guertin DA, Ben-David M, Harestad AS, Elliott JE (2012) Fecal genotyping reveals demographic variation in river otters inhabiting a contaminated environment. The Journal of Wildlife Management 76: 1540–1550.