You are viewing the site in preview mode

Skip to main content
Download PDF
Download ePub

Travel specific resource selection by female Kodiak brown bears during the sockeye salmon spawning season

Movement Ecology volume 12, Article number: 77 (2024) Cite this article

Abstract

Background

Access to salmon resources is vital to coastal brown bear (Ursus arctos) populations. Deciphering patterns of travel allowing coastal brown bears to exploit salmon resources dispersed across the landscape is critical to understanding their behavioral ecology, maintaining landscape connectivity for the species, and developing conservation strategies.

Methods

We modeled travel behavior of 51 radio-collared female Kodiak brown bears (U. a. middendorffi) from 2008 to 2015 during the sockeye salmon (Oncorhynchus nerka) stream spawning season to identify landscape patterns associated with travel pathways. To accomplish this, we first identified behavioral states of marked individuals, and then developed a resource selection function (RSF) to evaluate environmental covariates that were predictors of selection during travel behavior.

Results

Landcover edges, elderberry-salmonberry stands, lowland tundra, elevation, terrain position, and stream length influenced selection for travel corridors. The RSF validated well and was comparable to corridors identified by pathways used by bears while travelling.

Conclusions

Models identified spatial predictions of the relative probability of selection while bears were travelling during the salmon spawning season and identified areas that contained potential movement corridors important for bears inhabiting Kodiak Island. Our results characterized factors influencing travel, identified important movement corridors, and provided managers with information to make informed resource management decisions.

Background

Understanding how animals navigate landscapes to exploit heterogeneously distributed resources is critical for their conservation and management, especially as human encroachment and environmental change potentially disrupt animal movements and landscape connectivity. The ability of an animal to access resources and their decisions affecting movement patterns have implications for individual fitness, conservation, and management of populations [1,2,3,4]. Factors that influence resource selection and movement patterns often vary by season and animal behavior because animals move deliberately across the landscape depending on their current needs and the distribution of resources [5,6,7,8,9]. Patterns of resource selection can largely depend on an animal’s behavioral state or activity, which can be broadly categorized as resting, foraging or travel, as resource needs often vary greatly among behaviors and seasons [8, 10,11,12]. For example, selecting sites for rest or risk avoidance likely involves selection for environmental characteristics that differ from those for foraging or travel.

Only recently have studies differentiated among behaviors that may drive different patterns of resource selection. Studies that do not differentiate selection patterns among behavioral states may provide insights into general patterns of space use, but not behavior-specific patterns, which if ignored may lead to erroneous or incomplete inferences regarding space use [12, 13]. Several analytical techniques have been developed to better understand animal behavior from GPS location data, and researchers have successfully identified behavioral states to more accurately understand patterns of resource selection [10, 14,15,16]. For example, a study of African wild dogs demonstrated that aggregating all movement data (i.e., failing to account for behavior-specific resource use patterns) can be misleading [13]. Identifying how resource selection varies across different behavioral states can provide key insights into how landscape patterns or human activity foster or impede the ability of wildlife to access critical resources [17,18,19].

Coastal brown bears (Ursus arctos) rely heavily on salmon resources to meet their caloric needs [20,21,22,23,24,25,26,27]. Fecundity, body size, and population density of coastal brown bears are strongly correlated with salmon consumption [21, 28]. Recent research has also demonstrated that Kodiak brown bears prolong their access to spawning sockeye salmon (Oncorhynchus nerka) by exploiting spatial variation in spawning phenology [29,30,31]. Specifically, the ability of Kodiak Island brown bears (U.a. middendorffi) to track and exploit a suite of salmon spawning habitats spread across the landscape, both temporally and spatially, has a greater effect on the amount of salmon in the diet than interannual variation in salmon abundance [29]. The ability of salmon to maintain dense bear populations on the Kodiak Archipelago emphasizes the importance of ecosystem-based management of salmon escapement [28]. However, there is a lack of understanding of the habitats necessary to facilitate bear movement among salmon foraging sites.

While landscape patterns influencing travel behaviors for wildlife has been investigated [17,18,19, 32], an understanding of patterns of Kodiak brown bear movement to meet resource needs and the environmental features affecting these patterns is essential for a meaningful interpretation of resource selection and for effective conservation and management efforts [14, 33,34,35,36,37]. Human actions that could hinder the ability of bears to travel freely across the landscape to access salmon streams could adversely impact Kodiak brown bears. For example, outdoor recreation activities have the potential to displace brown bears and increase energetic costs [38, 39]. Therefore, knowledge of areas used by brown bears to travel among important resources scattered across the landscape is critical for their conservation and future resource planning [35,36,37].

Our work sought to better understand travel patterns of Kodiak brown bears and the environmental conditions influencing these patterns during the sockeye salmon stream spawning season (1 July through 21 August). Our specific objectives were to (1) identify the behavioral state of travel for GPS-collared female brown bears, and (2) develop resource selection models to evaluate environmental covariates that characterize travel behavior. To address our objectives, we first used a segmentation clustering approach [40] to detect change points between behavioral states in a time series of GPS locations for each bear. We considered 3 behavioral states which corresponded to user-defined behaviors: resting, foraging, and travel. We then used locations associated with travelling behavior to identify travel corridors and to evaluate resource selection in a resource selection function (RSF) framework [40, 41]. Our analyses identified environmental factors that bears selected while traveling during the sockeye salmon spawning season and provide practitioners with guidance to make informed decisions about the spatial juxtaposition of habitats that facilitate bear movement during the spawning season.

Methods

Study area

The study area encompasses the southwestern region of Kodiak National Wildlife Refuge and includes three primary salmon stream-lake systems (Karluk, Frazer, and Red), dozens of spawning tributaries and rivers, and a dense population of brown bears [42] (~ 250 independent bears/1000 km2; Fig. 1). The region has a maritime climate characterized by cool temperatures, overcast skies, and heavy precipitation. The topography of southwestern Kodiak is composed of a series of long fjords, a mixture of steep mountains, broad and short steep river valleys, gently rolling terrain, and wetlands. A mixture of deciduous tree species grow along rivers and streams. The lower elevation mountain slopes are covered in alder (Alnus crispa), salmonberry (Rubus spectabilis), elderberry (Sambucua racemosa), and forb-graminoid meadows. Alpine and subalpine vegetation and barren areas are common at elevations above 760 m. Large areas of lowland heath and shrub-graminoid wetlands are common in the subdued rolling lowlands [43]. Human activity within the study area is limited and dominated by recreation, primarily hunting, sport fishing, and bear-viewing.

Fig. 1
figure 1

Kodiak Island, Alaska study area used to evaluate female Kodiak brown bear resource selection while traveling during the salmon spawning season, 2008–2015

Full size image

Each lake-stream system included in our study is somewhat unique. Karluk Lake is long (20 km) and narrow (ranging from 1.4 km to 3.1 km wide) and has 11 tributaries, most of which are short and shallow and have spawning salmon runs during July and August. The exceptions are O’Malley and Thumb Creeks, which have relatively larger flows and drain larger valleys. Thumb Creek has sockeye salmon runs from July through August. Sockeye salmon transit through O’Malley Creek on their way to spawning grounds in Canyon and Falls Creeks during July and August. During September and October sockeye salmon spawn in O’Malley River. Karluk Lake is drained by the Karluk River, which is approximately 40 km long and flows into the ocean on the western side of the island near Karluk Village. Spawning sockeye and pink salmon (Oncorhynchus gorbuscha) are available to bears in the upper 4.3 km of Karluk River during September and early October [44]. Sockeye salmon spawn on suitable shoals along the shore of Karluk Lake largely from September through November, and to a lesser degree during July-August.

Frazer Lake is 14 km long and approximately1.3 km wide, and has one primary tributary, Pinnell Creek. Dog Salmon Creek drains Frazer Lake and runs south approximately 14 km emptying into Olga Bay. The Frazer drainage is unique among the two other lake-river systems because it supports an introduced sockeye salmon population. Brood stock sockeye salmon were introduced into Frazer Lake from the Red Lake and Karluk Lake drainages during 1951–1971 [45]. From 1956 to 1962 returning adult sockeye were transported by backpack around a waterfall approximately one km below the lake outlet that presented a barrier to salmon migration. In 1962 a fish ladder was constructed at the waterfalls (Frazer Fish Pass) allowing spawning salmon to pass the waterfall and establish a self-sustaining population. This was improved during 1971 and 1972 by building a diversion weir to help guide returning salmon towards the ladder,

The third drainage consists of the Ayakulik River and the smallest of the three lakes, Red Lake. Red Lake is 6.4 km by 1.3 km and has two significant tributaries with sockeye salmon runs, Connecticut Creek and Southeast Creek. Red Lake is drained by Red Lake River (6.3 km long) flowing into the Ayakulik River which runs for 23 km to the ocean on the west side of the island. Salmon spawn in Southeast and Connecticut Creeks in July and August. Salmon transit through Red Lake River to Red Lake and Connecticut and Southeast Creeks during July and August and spawn in Red Lake River itself during September.

The Alaska Department of Fish and Game (ADF&G) monitors salmon escapement into each of these systems with fish weirs on the Karluk River, Ayakulik River, and Dog Salmon Creek and the Frazer Fish Pass. From 2006 to 2018 sockeye salmon returns to Karluk Lake averaged 481,000 (181,000 during the early run and 300,000 during the late run) ranging from 53,000 in 2009 to 295,000 in 2007; Frazer Lake averaged 139,000 and ranging from 90,000 in 2006 to 219,000 in 2015; and Ayakulik/Red Lake averaged 163,000 ranging from 59,000 in 2006 to 218,000 in 2015 [46].

Sturgeon River, in the far west of the study area, consists of two branches – the East Fork and the Main Stem. The Main Stem Sturgeon is approximately 35 km from the headwaters to the Sturgeon Lagoon. The East Fork Sturgeon is approximately 17 km from its headwaters to the Sturgeon Lagoon. Both branches have chum salmon (Oncorhynchus keta) runs roughly between mid-June through mid-July [47]. Runs on the Sturgeon River have been monitored only intermittently by fixed-wing surveys since 2000, but observations during annual aerial bear surveys from 2006 to 2018 suggest a substantial decrease in the size of salmon runs on both the Main Stem Sturgeon and East Fork Sturgeon.

Capture and monitoring

We captured adult female brown bears in the Southwest region of Kodiak Island, Alaska during the last two weeks of May 2008–2015. All bears were anesthetized via darting from helicopter using a 1:1 mixture of tiletamine hydrochloride and zolazepam hydrochloride (Telazol®, Fort Dodge Laboratories, Fort Dodge, Iowa, USA) at a dose rate of 10–12 mg/kg [48]. Standard zoological measurements, weights, and age estimates based on cementum annuli of premolars [49] were taken for each bear. Capture and handling procedures were approved by the Institutional Animal Care and Use Committees (ADFG ACUC 07–08, USFWS IACUC Permit 2012008, USFWS IACUC Permit 2012008 Renewal, and USFWS IACUC Permit 2015-001). We fitted each bear with a GPS radio collar (Telonics Model #TGW-3790) programmed to record a location every hour from 15 May through 15 November. Collars contained an ultrahigh frequency (UHF) radio transmitter and locations were downloaded using an airplane fitted with a UHF receiver. Collars also collect activity data synchronous with each GPS location with an onboard mercury switch. Activity levels were expressed as the percentage of time being active during the data collection interval. We screened GPS locations for accuracy and removed relocations with a positional dilution of precision (PDOP) greater than 10 [50]. We restricted bear locations to the period of 10 July through the end of August each year to coincide with the annual sockeye salmon run.

Defining bear movement behavior

We identified travel behavior of radio-collared female Kodiak brown bears from 2008 to 2015 during the sockeye salmon stream spawning season to identify environmental factors associated with selection of travel pathways. To identify travel behavior compared to foraging or stationary behavior of female brown bears, we used a segmentation clustering method with the seglust2d package in R [51]. This approach followed a modified version of Lavielle’s method [52] to detect change points between behavioral states in a time series of GPS locations for each bear-year [52, 53]. We used speed and relative turning angle between successive 1-hour re-locations to differentiate behavioral states. We considered three behavioral states in our models that we defined as resting, foraging, and travel (Table 1). Resting behavior was characterized by little movement and low activity. Foraging behavior was indicated by high tortuosity and low speed, and travel behavior was characterized by low tortuosity and higher speed [53]. The minimum number of locations for each segment, Lmin, was set to 5 [51]. As a validation of this clustering technique to identify behavioral states, we used the activity sensor data collected in conjunction with GPS locations for each bear. We expected that activity would be higher while travelling compared to resting or foraging. We used an analysis of variance to evaluate whether mean activity sensor values from the transmitters differed significantly from each behavioral state identified with the segmentation clustering method. Activity sensor values were log transformed to ensure that data fit a normal distribution. Locations identified as travelling were used in subsequent resource selection analyses (Fig. 2). In addition, to identify potential bear travel corridors, we connected consecutive locations by individuals that were identified as travelling. Individual line segments were buffered by 200 m and buffered lines were then overlayed to sum the number of bears using respective movement pathways.

Table 1 Step length (m/h) and turning angle (standard error in parentheses) for the three behavioral states identified from female Kodiak brown bear location data
Full size table
Fig. 2
figure 2

Locations of female Kodiak brown bears associated with travel behavior identified during the salmon stream spawning season, 2008–2015 southwest Kodiak Island

Full size image

Resource selection analysis

We used a RSF at the seasonal home range scale of the population of radio-collared bears to evaluate resource selection while in traveling behavior. We defined the seasonal home range (hereafter, study area) as a 100% fixed kernel of all bear locations collected from 10 July through the end of August each year using the adehabitatHR package in R [54]. Our models enabled spatial predictions of the relative probability of selection while bears were traveling and identified factors influencing bear movements across the landscape. We used environmental predictor variables, including both landcover and topographic features, that we felt had biological relevance to Kodiak brown bears based on land-based field observations and aerial surveys. Landcover covariates were derived from Kodiak Land Cover Classification 30 m raster data [43], for which we aggregated landcover into 13 classes from 63 classes based on bear ecology (Fig. 2). Each landcover class was created by binary reclassification such that values of 1 equaled the landcover class of interested and values of 0 equaled other landcover classes. We considered elderberry/salmonberry and lowland tundra landcover in our analyses. We also calculated an edge diversity metric for low willow and tree landcover classes. Edge diversity metrices were calculated as the absolute difference between the mean value of each raster cell and the 8 cells surrounding the center cell. Edge diversity ranged from 0 (all 8 surrounding raster cells were the same landcover class as the center cell) to 0.89 (all 8 surrounding cells represented a different landcover class than the center cell) and represented edge diversity at a local scale.

For topography, we considered elevation, and a topographic position index (TPI) derived from a 30-meter digital elevation model [55]. We calculated TPI with the raster package in R [56]. TPI is a method to identify the relative position of a point along a topographic gradient and was calculated by comparing the elevation of each cell to the mean elevation of the 8 surrounding cells [57]. Positive TPI values correspond to ridges or hilltops, negative TPI values are indicative of topographic depressions within the surrounding landscape such as valleys, gullies, and mountain passes, and values approaching 0 are indicative of flatter topography and mid-slopes [58, 59].

We merged the water landcover class [43] with manually developed salmon stream layers to create a comprehensive water layer that represented the length of stream. A field crew walked each stream, and recorded stream depth, width, and substrate size at 5 m increments along the stream. The upstream limit of spawning activity was obvious on several streams because there was a salmon impermeable barrier such as a waterfall or dense downed vegetation. On unobstructed streams we determined the upstream limit of spawning by surveying for evidence of past spawning activity by looking for sockeye jawbones, which tend to persist through winter, bear trails, and other bear sign. If no evidence of previous spawning activity was observed over 100 consecutive meters of stream, we discontinued the stream morphology survey.

We estimated a travel-specific RSF for female brown bears within the study area using binomial generalized mixed models with package glmmTMB in R [60]. We specified a random intercept as individual bear-year and included individual-year specific random slopes for all covariates [61]. We randomly generated 10 available locations within the study area per bear-use location to represent available habitat but ensured that available locations were not located in open water. Available locations were assigned an id associated with a given bear-year and were given a weight of 1000 [62]. We extracted mean covariate values at each used and available location within a 275 m radius circular buffer. This buffer represented half of the average distance traveled by female brown bears between relocations while in travel behavior. Values averaged within 275 m are likely better representations of the environment at the scale of bear selection during travel [63]. For landcover covariates, we calculated the proportion of each landcover class within the circular buffer. For edge diversity and topography predictors, we calculated the mean values within the circular analysis buffer. We centered and scaled variables to facilitate model convergence and to make slopes of parameter estimates comparable [64, 65]. The RSF took the following form:

$$w(x) = {\mkern 1mu} \exp ({\beta _1}{x_1} + {\beta _2}{x_2} + \ldots + {\beta _n}{x_n})$$

where w(x) was proportional to the probability of female brown bear selection, and β1 represented the coefficient describing selection strength for covariate x1, and n represented the number of covariates in the model. We ensured that variables were note highly correlated (|r| > 0.70), and considered variables informative when coefficients had 95% confidence intervals that did not include zero [66, 67].

We used 5-fold cross validation to evaluate the RSF by randomly partitioning data by individual bear-years. We estimated predictions based on 4 of the 5 groups (training data) and compared them to the withheld group and repeated this until the 5 withheld groups were evaluated [41]. We binned predictions into 6 equal-area (quartile) intervals [68]. Validations were performed by running linear regression models on the number of observed locations from the test group compared to expected locations generated from each RSF bin [41, 69]. We considered models to be good predictors when linear regression models had high coefficients of determination (R2 > 0.9), and 95% confidence intervals of slope estimates excluded zero and included 1 [69]. We mapped the RSF across the study area by using coefficients from the top model and distributed predictions into 6 equal area bins corresponding to increasing relative probability of selection.

Results

We used data from over 175,000 GPS locations collected from 51 female brown bears with 76 unique bear-years spanning 2008–2015 (some bears carried collars for more than 1 year). We estimated a travel behavior RSF using 11,623 locations (mean, 153 locations per bear-year). During the study period, bears moved an average of 550 m/h while engaged in travel behavior (95% CI: 501–598 m/h; Table 1). Visual inspection suggested that this method fit the data well and identified travel corridors that radio-marked female brown bears used on Kodiak Island (Fig. 3). In addition, we found that activity sensor values recorded from radio-collars were concordant with the behavioral states identified in the segmentation clustering method (F = 795.9, P < 0.001). Activity sensor values were lowest for locations identified as resting behavior (x̅ = 57.86, 95% CI = 52.64–63.08), intermediate for foraging behavior (x̅ =111.79, 95% CI = 106.43–117.15), and highest for travelling behavior (x̅ =161.62, 95% CI = 154.12–169.12), providing additional justification for the use of the clustering approach.

Fig. 3
figure 3

Intensity of female Kodiak bear use of travel corridors during the salmon stream spawning season, 2008–2015 southwest Kodiak Island. Corridors were generated from bear locations while traveling and were overlaid to sum the number of bears using respective corridors

Full size image

Resource selection analysis

While engaged in travel behavior, landcover, habitat edges, and topography were predictive of resource selection. Bears tended to avoid elderberry-salmonberry and lowland tundra but selected for low willow and tree edge (Table 2). Bears also avoided higher elevation and TPI and selected for greater stream length. The predicted relative probability of selection decreased by approximately 5.3% when the proportion of elderberry-salmonberry increased by 10%. Relative probability of selection decreased by approximately 20.9% as lowland tundra increased by 10%. Relative probability of selection increased by approximately 6.4% and 9.6% as low willow and tree edges increased from zero to 0.33, respectively. A 100 m decrease in elevation was predicted to increase relative probability of selection by approximately 4% and a 10% decrease in TPI was predicted to increase relative probability of selection by 1.4%. Finally, the model predicted that a 0.1 km increase in length of stream increased the relative probability of selection by 1.5%. The interpretation of change in relative selection probabilities per unit change in variables were calculated using unstandardized selection coefficients in the RSF model, with other variables held at their mean.

Table 2 Parameter estimates, standard errors (SE) and 95% confidence intervals for predictor variables used in resource selection models describing female Kodiak brown bear resource selection while traveling
Full size table

The spatial prediction of the RSF performed well at predicting brown bear habitat selection while traveling. When we partitioned validation testing and training groups by individual bear, the average R2 = 0.94 ± 0.01 (SE), and confidence intervals of slope estimates included 1 and excluded zero in all folds (Table 3). Visual inspection of the highest 2 predicted bin values, which corresponded to the highest relative probability of selection (Fig. 4), were congruent with corridors identified by radio-marked bears (Fig. 3), and identified other areas across the study area that potentially facilitate bear movements.

Table 3 Five-fold cross validation results from female brown bear season home range resource selection models. We considered models good predictors of resource selection when they had a high coefficient of determination, and 95% confidence intervals (CI) surrounding slope estimates (B1) that excluded zero and included 1. RSF models were considered acceptable when slope estimates excluded both zero and 1
Full size table
Fig. 4
figure 4

Predicted relative probability of selection during travel by female Kodiak brown bears during the salmon stream spawning season, 2008–2015 southwest Kodiak Island. Predictions were binned into 6 quantiles from low (bin 1) to high (bin 4) relative probability of selection

Full size image

Discussion

Our work represents a biologically meaningful method to evaluate factors associated with travel behavior of Kodiak brown bears. We used travel behavior models based on GPS locations of female Kodiak bears to determine landscape features associated with travel and identify travel corridors selected by bears during the stream sockeye spawning season in southwestern Kodiak Island. Our RSF analysis demonstrated that bears select relatively open terrain, along habitat edges, and at lower elevations while travelling. In particular, bears exhibited strong selection for the margins of lakes and rivers where these environmental conditions coalesce. These findings are consistent with the authors’ observations of bears in the wild: bear trails tend to follow the edges of disparate habitats, the edges of streams and lakes, and through concave terrain like mountain passes.

We focused on GPS locations associated with travel behavior, rather than an aggregation of all locations which would include resting and feeding behavior. Consequently, habitat selection appeared to consistently reflect the relative ease or difficulty of moving through different habitats, rather than selection for feeding or resting sites. Our findings are supported by an innovative study that found brown bears likely reduced energetic costs of movement by selecting movement paths with lower resistance [70]. At the seasonal home range scale, bears avoided traveling through elderberry-salmonberry stands, lowland tundra, higher elevations, and TPI. Though elderberry-salmonberry stands provide important food when in season [71, 72], salmonberry stands are very dense and present relatively impenetrable obstructions for travel. Low values for TPI indicated that bears selected for valleys, low mountain passes, or similar topographic depressions for traveling within the study area. Avoidance of lowland tundra may be because it is largely covered by hummocks which makes travel difficult. This result may partly be an artifact of low levels of use of lowland tundra during the salmon season because most of the lowland tundra is in the west-southwest region of the study area which is more distant from salmon streams. This is a bit surprising because bear trails were often observed in the field where willow and/or alder stands meet lowland tundra communities. Selection for low willow edge and tree edge emphasizes the importance of edge habitat for travel. Edges along these different landcover types often offer narrow travel paths that are clear of obstructions and near escape cover. Selection for stream length reflects the relative ease of travel alongside streams and rivers and the fact that bears often move along streams to exploit stretches with high salmon concentrations.

A challenge in the effort to understand how habitat selection varies with behavior has been to accurately identify behaviors from animal location data. For example, Sorum [71] investigated seasonal diets of Kodiak brown bears (Ursus arctos middendorffi) with a combination of activity data and patterns of GPS locations to identify recent bedding sites. Several studies have employed hidden Markov modeling (HMMs) approaches to analyze radio-telemetry locations to differentiate behavioral states [10, 14, 17]. For example, Franke et al. [15] successfully identified bedding, feeding, and travel behavioral states of woodland caribou (Rangifer tarandus). However, using tracking data to identify biologically meaningful behavioral states when using HMM does come with difficulties [73]. A strict focus on statistical procedures to identify behavioral states can lead to models with superior goodness of fit that cannot necessarily be linked to biologically meaningful or interpretable behaviors [73, 74]. HMMs require user identified initial state-dependent probability distribution parameters, whereas the segmentation-clustering approach we employed does not [53]. A user-friendly approach, such as this, that is readily accessible to practitioners will be useful for identifying behavioral states with an array of tracking data.

Our modeling efforts highlight the importance of behavior-specific resource selection and connectivity across the landscape. Collectively, the results of this analysis accurately identified factors influencing bear travel during the salmon stream spawning season and provide managers with information to make informed decisions to conserve bear connectivity. Our results demonstrated that during a critical period of foraging, the stream spawning season, bears used relatively few paths to travel among foraging sites and these corridors were often centered near lake edges and areas with narrow topographic relief that facilitated movement. Maintaining the integrity of these travel corridors is critical to the conservation and management of the Kodiak brown bear. To achieve the management goals of maintaining a high population level and body size of Kodiak brown bears, ADF&G and the Kodiak NWR must take steps to ensure that travel corridors are maintained by identifying travel routes and ensuring the future integrity of those routes. Although we were unable to assess how flexible bears were in their use of distinct movement corridors in response to disruption, our findings suggest that bears depend on specific corridors repeatedly. Future studies investigating selection across all three behavioral states (resting, foraging, and travel), across all seasons, and among different classes of bears would add greatly to our understanding of the spatial ecology of coastal brown bears and allow agencies to better manage these populations and the lands they depend on.

Conclusions

Kodiak brown bears are currently able to exploit salmon resources widely spread across the landscape. But increasing recreation and development within some key areas in the study area could impede this ability to move freely among important salmon foraging areas. Specifically, recreation activities and infrastructure in the study area and elsewhere on Kodiak Island are often concentrated along lake shores, streams, rivers, and ocean beaches. In this study, we confirmed that bears indeed preferentially travel along lakeshores, streams, rivers, and ocean beaches, which creates a potential conflict between human activities and bear conservation. These travel corridors must be managed carefully to prevent increasing human activities on the Refuge from hindering bear movement among critical resources, which allow Kodiak bear populations to achieve some of the highest densities and largest sizes on earth.

Our research also emphasized the need to manage the salmon populations in a manner that maintains the ecological integrity of the salmon-bear ecosystem. Recent research has shown that variation in spawning phenology among salmon populations across the landscape has a profound effect on foraging opportunities for the Kodiak brown bear [29,30,31]. Not only is there population-level variation among spawning times across spawning and within spawning sites [29], there is also evidence for variation in salmon migration phenology [75, 76]. Maintaining and managing this phenological diversity poses some challenges. Although ecosystem-based fisheries management is a goal, phenological and population diversity is not explicitly considered within current salmon fisheries management protocols [24, 77, 78]. Management frameworks that use salmon escapement as the sole indicator of bear foraging opportunities may not be adequate. Research determining the relationship between these phenological subpopulations and when they enter river systems could contribute to developing fisheries management protocols that would help maintain this diversity and to help establish ecosystem-based fisheries management goals.

Data availability

Data are available from the corresponding author on reasonable request.

References

  1. Mysterud A, Langvatn R, Yoccoz NG, Stenseth NC. Plant phenology, migration, and geographical variation in body weight of a large herbivore: the effect of a variable topography. J Anim Ecol. 2001;70:915–23.

    Article  Google Scholar 

  2. Wang G, Hobbs NT, Boone RB, Illius AW, Gordon IJ. Spatial and temporal variability modify density dependence in populations of large herbivores. Ecology. 2006;87:95–102.

    Article  PubMed  Google Scholar 

  3. Hebblewhite M, Merrill E, McDermid G. A Multi-scale Test of the forage maturation hypothesis in a partially migratory Ungulate Population. Ecol Monogr. 2008;78:141–66.

    Article  Google Scholar 

  4. Ruff CP, Schindler DE, Armstrong JB, Bentley KT, Brooks GT, Holtgrieve GW, McGlauflin MT, Torgerson CE, Seeb JE. Temperature-associated population diversity in salmon confers benefits to mobile consumers. Ecology. 2011;92:2073–84.

    Article  PubMed  Google Scholar 

  5. Thurfjell H, Ciuti S, Boyce MS. Applications of step-selection functions in ecology and conservation. Mov Ecol. 2014;2:1–12.

    Article  Google Scholar 

  6. Zeller KA, McGarigal K, Beier P, Cushman SA, Vickers TW, Boyce WM. Sensitivity of landscape resistance estimates based on point selection functions to scale and behavioral state: pumas as a case study. Landscape Ecol. 2014;29:541–57.

    Article  Google Scholar 

  7. Zeller KA, McGarigal K, Cushman SA, Beier P. Using step and path selection functions for estimating resistance to movement: pumas as a case study. Landscape Ecol. 2016;31:1319–35.

    Article  Google Scholar 

  8. Wilson RR, Gilbert-Norton L, Gese EM. Beyond use versus availability: behaviour-explicit resource selection. Wildl Biology. 2012;18:424–30.

    Article  Google Scholar 

  9. Ellington EH, Muntz EM, Gehrt SD. Seasonal and daily shifts in behavior and resource selection: how a Carnivore navigates costly landscapes. Oecolgia. 2020;194:87–100.

    Article  Google Scholar 

  10. Clontz LM, Pepin KM, VerCauteren KC, Beasley JC. Behavioral state resource selection in invasive wild pigs in the Southeastern United States. Sci Rep. 2021;11:6924.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Beyer HL, Haydon DT, Morales JM, Frair JL, Hebblewhite M, Mitchell M, Matthiopoulos J. The interpretation of habitat preference metrics under use-availability designs. Philos Trans Royal Soc. 2010;365:2245–54.

    Article  Google Scholar 

  12. Roever CL, Beyer HL, Chase MJ, van Aarde RJ. The pitfalls of ignoring behaviour when quantifying habitat selection. Divers Distribution. 2014;20:322–33.

    Article  Google Scholar 

  13. Abrahms B, Jordan NR, Golabek KA, McNutt JW, Wilson AM, Brashares JS. Lessons from integrating behaviour and resource selection: activity-specific responses of African wild dogs to roads. Anim Conserv. 2015;19:247–55.

    Article  Google Scholar 

  14. Karelus DL, McCown JW, Scheick BK, van de Kerk M, Bolker BM, Oli MA. Incorporating movement patterns to discern habitat selection: black bears as a case study. Wildl Res. 2019;46(1):76–88.

    Article  Google Scholar 

  15. Franke A, Caelli T, Hudson RJ. Analysis of movements and behavior of caribou (Rangifer tarandus) using hidden Markov models. Ecol Model. 2004;173:259–70.

    Article  Google Scholar 

  16. Hance DJ, Moriarty KM, Hollen BA, Perry RW. Identifying resting locations of a small elusive forest Carnivore using a two-stage model accounting for GPS measurement error and hidden behavioral states. Mov Ecol. 2021;9. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-021-00256-8.

  17. Abrahms B, Sawyer SC, Jordan NR, McNutt JW, Wilson AM, Brashares JS. Does wildlife resource selection accurately inform corridor conservation? J Appl Ecol. 2016;54(2):412–22.

    Article  Google Scholar 

  18. Monteith KL, Hayes MM, Kauffman MJ, Copeland HE, Sawyer H. Functional attributes of ungulate migration: landscape features facilitate movement and access to forage. Ecol Appl. 2018;28(8):2153–64.

    Article  PubMed  Google Scholar 

  19. Osipova L, Okello MM, Njumbi SJ, Ngene S, Wester D, Hayward MW, Balkenhof N. Using step-selection functions to model landscape connectivity for African elephants: accounting for variability across individuals and seasons. Volume 22. Animal Conservation: Animal Conservation; 2018. pp. 35–48.

    Google Scholar 

  20. Barnes VG. The influence of salmon availability on movements and range of brown bears on Southwest Kodiak Island. Bears: Their Biology and Management. 1990; 8:305–313.

  21. Hilderbrand GV, Schwartz CC, Robbins CT, Jacoby ME, Hanley TA, Arthur SM, Servheen C. The importance of meat, particularly salmon, to body size, population productivity, and conservation of north American brown bears. Can J Zool. 1999;77:132–8.

    Article  Google Scholar 

  22. Ben-David M, Titus K, Beier LR. Consumption of salmon by alaskan brown bears: a trade- off between nutritional requirements and the risk of infanticide? Oecologia. 2004;138:465–74.

    Article  PubMed  Google Scholar 

  23. Gende S, Quinn T. The relative importance of prey density and social dominance in determining energy intake by bears feeding on Pacific salmon. Can J Zool. 2004;85:75–85.

    Article  Google Scholar 

  24. Schindler D, Armstrong J, Bentley K, Jankowski K, Lisi P, Payne L. Riding the crimson tide: mobile terrestrial consumers track phenological variation in spawning of an anadromous fish. Biol Lett. 2013;9:20130048.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Hilderbrand GV, Gustine DD, Joly K, Mangipane B, Leacock W, Cameron MD, Sorum MS, Mangipane LS, Erlenbach JA. Influence of maternal body size, condition, and age on recruitment of four brown bear populations. Ursus. 2019;29(2):111–8.

    Article  Google Scholar 

  26. Mangipane LS, Laferty DJR, Joly K, Sorum MS, Cameron MD, Belant JL, Hilderbrand GV, Gustine DD. Dietary plasticity and the importance of salmon to brown bear (Ursus arctos) body size and condition in a low Arctic ecosystem. Polar Biol. 2020;43:825–33.

    Article  Google Scholar 

  27. Rogers MC, Hilderbrand GV, Gustine DD, Joly K, Leacock WB, Mangipane BA, Welker JM. Splitting hairs: dietary niche breadth modelling using stable isotope analysis of a sequentially grown tissue. Isot Environ Health Stud. 2020;56(4):358–69.

    Article  CAS  Google Scholar 

  28. Van daele MB, Robbins CT, Semmens BX, Ward EJ, van Daele LJ, Leacock WB. Salmon consumption by Kodiak brown bears (Ursus arctos middendorffi) with ecosystem management implications. Can J Zool. 2013;91:164–74.

    Article  Google Scholar 

  29. Deacy WW, Erlenbach JA, Leacock WB, Stanford JA, Robbins CT, Armstrong JB. Phenological tracking associated with increased salmon consumption by brown bears. Sci Rep. 2018;8:11008.

    Article  PubMed  PubMed Central  Google Scholar 

  30. Deacy W, Leacock WB, Armstrong JB, Stanford JA. Kodiak brown bears surf the salmon red wave: direct evidence from GPS collared individuals. Ecology. 2016;97:1091–8.

    Article  PubMed  Google Scholar 

  31. Deacy WW, Leacock WB, Stanford JA, Armstrong JB. Variation in spawning phenology within salmon populations influences landscape-level patterns of brown bear activity. Ecosphere. 2019;10(1):e02575.

    Article  Google Scholar 

  32. Holdo RM, Fryxell JM, Sinclair AR, Dobson A, Holt RD. Predicted impact of barriers to migration on the Serengeti wildebeest population. PLoS ONE. 2011;6(1):e16370.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Michelot T, Blackwell PG, Matthiopoulos J. Linking resource selection and step selection models for habitat preferences in animals. Ecology. 2018;100(1):e02452.

    Article  PubMed  Google Scholar 

  34. Abrahms B, Aikens EO, Armstrong JB, Deacy WW, Kauffman MJ, Merkle JA. Emerging perspectives on resource tracking and animal movement ecology. Trends Ecol Evol. 2020;36:308–20.

    Article  PubMed  Google Scholar 

  35. Clark JD, Laufenberg JS, Davidson M, Murrow JL. Connectivity among subpopulations of Louisiana black bears as estimated by a step selection function. J Wildl Manage. 2015;79(8):1347–60.

    Article  Google Scholar 

  36. Panzacchi M, Van Moorter B, Strand O, Saerens M, Kivimäki I, St. Clasir CC, Herfindal I, Boitani L. Predicting the continuum between corridors and barriers to animal movements using step selection functions and randomized shortest paths. J Anim Ecol. 2016;85:32–42.

    Article  PubMed  Google Scholar 

  37. Almasieh K, Rouhi H, Kaboodvandpour S. Habitat suitability and connectivity for the brown bear (Ursus arctos) along the Iran-Iraq border. Eur J Wildl Res. 2019;65:57.

    Article  Google Scholar 

  38. Fortin JK, Rode KD, Hilderbrand GV, Wilder J, Farly S, Jorgensen C, Marcot BG. Impact of human recreation on brown bears (Ursus arctos): a review and new management tool. PLoS ONE. 2016;11(1):e0141983.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Penteriani V, Lopez-Bao JV, Bettega C, Dalerum F, del Mar Delgado M, Jerina K, Kojola I, Krofel M, Ordiz A. Consequences of brown bear viewing tourism: a review. Biol Conserv. 2017;206:169–80.

    Article  Google Scholar 

  40. Manly BFJ, McDonald LL, Thomas DL, McDonald TL, Ericson WP. Resource selection by animals: statistical design and analysis for field studies. 2002. Second edition, Kluwer Academic, Boston, Massachusetts, USA.

  41. Johnson CJ, Nielson SE, Merrill EH, McDonald TL, Boyce MS. Resource selection functions based on use-availability data: theoretical motivation and evaluation methods. J Wildl Manage. 2006;70:374–357.

    Article  Google Scholar 

  42. Barnes V, Smith R, van Daele L. Report to the Kodiak Brown Bear Trust: density estimates and estimated population of brown bears on Kodiak and adjacent islands; 1988.

  43. Fleming M, Spencer P. Kodiak Archipelago land cover classification users guide. USGS (Alaska Science Center) and National Park Service, 77 pp; 2007.

  44. Leacock WB, Sobota B, Helling H. Kodiak brown bear and human fall use of the upper Karluk River: 2006–2008. Refuge Report. Kodiak National Wildlife Refuge; 2014.

  45. Blackett RF. Establishment of sockeye (Oncorhynchus nerka) and Chinook (o, tshawytscha) salmon runs at Frazer Lake, Kodiak Island. J Fish Res Board Can. 1979;36(10):1265–77.

    Article  Google Scholar 

  46. McKinley TR, Schaberg KL, Witteveen MJ, Foster MB, Wattum ML, Vincent TL. Review of salmon escapement goals in the Kodiak management area. Fishery Manuscript No. 19 – 07. Alaska Department of Fish and Game; 2019.

  47. Price MA. Abundance and run timing of adult chum Salmon and Steelhead Kelt emigration in the Sturgeon River, Kodiak, Alaska. U.S. Fish and Wildlife Service Kenai Fishery Resource Office; 1999.

  48. Taylor WP, Reynolds HV III, Ballard WB. Immobilization of grizzly bears with tiletamine hydrochloride and zolazepam hydrochloride. J Wildl Manage. 1989;53:978–81.

    Article  Google Scholar 

  49. Matson GM, Van Daele LJ, Godwin L, Aumiller E, Reynolds L, Hristienko H. A laboratory manual for cementum age determination of Alaska brown bear first premolar teeth. Alaska Department of Fish and Game; 1993.

  50. Lewis JS, Rachlow JL, Garton EO, Vierling LA. Effects of habitat on GPS collar performance: using data screening to reduce location error. J Appl Ecol. 2007;44:663–71.

    Article  Google Scholar 

  51. Patin R, Etienne MP, Lebarbier E, Benhamou S. Segclust2d: Bivariate segmentation/clustering methods and tools. R package version 0.2.0; 2019, https://CRAN.R-project.org/package=segclust2d

  52. Lavielle M. Detection of multiple changes in a sequence of dependent variables. Stoch Process Appl. 1999;83:79–102.

    Article  Google Scholar 

  53. Patin R, Etienne MP, Lebarbier E, Chamaille-Jammes S, Benhamou S. Identifying stationary phases in multivariate time series for highlighting behavioral modes and home range settlements. J Anim Ecol. 2020;89:44–56.

    Article  PubMed  Google Scholar 

  54. Calenge C, Formann-Roe S, adehabitatHR. Home range estimation. R package version 0.4.21. 2023. https://CRAN.R-project.org/package=adehabitatHR

  55. US Geological Survey. The national map. TNM Download V2.0. Top Map data, 3 DEP products, Lidar, 1fSAR, NHD (Hydrography Dataset), NAIP Plus Imagery, National Structures Dataset. 2023. https://apps.nationalmap.gov/downloader/#/

  56. Hijmans RJ, Raster. Geographic Data Analysis and Modelling. R package version 3.3–13; 2020. https://CRAN.R-project.org/package=raster

  57. Guisan A, Weiss SB, Weiss AD. GLM versus CCA spatial modeling of plant species distribution. Plant Ecol. 1999;143:107–22.

    Article  Google Scholar 

  58. Weiss AD. Topographic position and landforms analysis. Poster, The Nature Conservancy; 2000. http://www.jennessent.com/downloads/tpi-poster-tnc_18x22.pdf

  59. Knitter D, Brozio JP, Hamer W, Duttman R, Müller J, Nakoinz O. Transformations and site locations from a landscape archaeological perspective: the case of neolithic Wagrien, Schleswig-Holstein, Germany. Land. 2019;8:68.

    Article  Google Scholar 

  60. Brooks ME, Kristensen K, van Benthem KJ, Manusson A, Berg CW, Nielsen A, Skaug HJ, Maechler M, Bolker BM. glmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. R J. 2017;9:378–400.

    Article  Google Scholar 

  61. Gillies CS, Hebblewhite M, Nielson SE, Krawchuk MA, Aldridge CL, Frair JL, Saher DL, Stevens CE, Jerde CL. Application of random effects to the study of resource selection by animals. J Anim Ecol. 2006;75:887–98.

    Article  PubMed  Google Scholar 

  62. Muff S, Signer J, Fieberg J. Accounting for individual-specific variation in habitat selection studies: efficient estimation of mixed-effects models using bayesian or frequentist computation. J Anim Ecol. 2020;89:80–92.

    Article  PubMed  Google Scholar 

  63. Smith KT, Taylor KL, Albeke SE, Beck JL. Pronghorn winter resource selection before and after wind energy development in south-central Wyoming. Rangel Ecol Manage. 2020;73:227–33.

    Article  Google Scholar 

  64. Becker RA, Chambers JM, Wilks AR. The new S language: a programming environment for data analysis and graphics. Wadsworth and Brooks Cole, Belmont, California, USA; 1988.

  65. Schielzeth H. Simple means to improve the interpretability of regression coefficients. Methods Ecol Evol. 2010;1:103–13.

    Article  Google Scholar 

  66. Arnold TW. Uninformative parameters and model selection using Akaike’s Information Criterion. J Wildl Manage. 2010;74:1175–8.

    Google Scholar 

  67. Burnham KP, Anderson DR. Model selection and multi-model inference: a practical information-theoretic approach. Second edition. Springer-Verlag, New York, New York, USA; 2002.

  68. Wiens TS, Dale BC, Boyce MS, Kershaw GP. Three way k-fold cross-validation of resource selection functions. Ecol Model. 2008;212:244–55.

    Article  Google Scholar 

  69. Howlin S, Erickson WP, Nielson RM. A validation technique for assessing precitive abilities of resource selection functions. Pages 40–51 in Proceedings of the First International Conference on Resource Selection. Western EcoSystems Technology, Laramie, Wyoming, USA; 2004.

  70. Finnegan SP, Svoboda NJ, Fowler NL, Schooler SL, Belant JL. Variable intraspecific space use supports optimality in an apex predator. Sci Rep. 2021;11:1–3.

    Article  Google Scholar 

  71. Sorum M. Behavior-specific resource selection by Kodiak brown bears. (Master’s Thesis), University of Idaho, Moscow Idaho, USA; 2013.

  72. Deacy W, Armstrong JB, Leacock WB, Robbins CT, Gustine DD, Ward EJ, Erlenbach JA, Stanford JA. Phenological synchronization disrupts trophic interactions between Kodiak brown bears and salmon. Proceedings of the National Academy of Sciences. 2017; 114:10432–10437.

  73. Pohle J, Langrock R, van Beest FM, Schmidt NM. Selecting the number of states in hidden Markov models: pragmatic solutions illustrated using animal movement. J Agricultural Biol Environ Stat. 2017;22:270–93.

    Article  Google Scholar 

  74. Edelhoff H, Signer J, Balkenhol N. Path segmentation for beginners: an overview of current methods for detecting changes in animal movement patterns. Mov Ecol. 2016;4:21.

    Article  PubMed  PubMed Central  Google Scholar 

  75. Doctor KK, Hilborn R, Rowse M, Quinn T. Spatial and temporal patterns of upriver migration by sockeye salmon populations in the Wood River system, Bristol Bay, Alaska. Trans Am Fish Soc. 2010;139:80–91.

    Article  Google Scholar 

  76. McGlauflin MT, Schindler DE, Seeb LW, Smith CT, Habicht C, Seeb JE. Spawing habitat and geography influence population structure and juvenile migration timing of sockeye salmon in the Wood River Lakes, Alaska. Trans Am Fish Soc. 2011;140:763–82.

    Article  Google Scholar 

  77. Hilborn R, Quinn TP, Schindler DE, Rogers DE. Biocomplexities and fisheries sustainability. Proceedings of the National Academy of Science. 2003; 100:6564–6568.

  78. Schindler DE, Hilborn R, Chasco B, Boatright CP, Quinn TP, Rogers LA, Webster MS. Population diversity and the portfolio effect in an exploited species. Nature. 2010;465:609–12.

    Article  CAS  PubMed  Google Scholar 

Download references

Acknowledgements

We are indebted to M. Sorum, C. Cheung, G. Perry, D. Perry, T. Leacock, B. Svoboda, (A) May, (B) Dunker, T. Melham, T. Tran, M. Jamison, L. Pless, E. Stephany, F. Cannizzo, I. Kelsey, M. Melham, J. Murawski, P. Weldon, S. Flemming, A. Orlando, and K. Hsu for assisting with field efforts and data collection and entry. We thank helicopter pilot Joe Fieldman and fixed-wing pilots Kurt Rees, Kevin Vanhatten, Kevin Fox, and Issac Bedingfield for their skilled flying during bear captures and aerial telemetry and surveys. We thank Kodiak National Wildlife Refuge administration for field support and logistics.

Funding

Funding was provided by Kodiak National Wildlife Refuge and the USFWS Inventory and Monitoring and Refuge Programs.

Author information

Author notes

  1. Kurt T. Smith

    Present address: Department of Zoology and Physiology, University of Wyoming, Laramie, WY, 82071, USA

Authors and Affiliations

  1. U.S. Fish and Wildlife Service, 101 12th Avenue, Room 236, Fairbanks, AK, 9970, USA

    William B. Leacock

  2. 27955 Blixit Creek Road, Potomac, MT, 59823, USA

    William B. Leacock

  3. Western Ecosystem Technology, 1610 Reynolds Street, Laramie, WY, 82072, USA

    Kurt T. Smith

  4. U.S. National Park Service, 1000 US Hwy 36, Estes Park, CO, 80517, USA

    William W. Deacy

Authors
  1. William B. Leacock
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Kurt T. Smith
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. William W. Deacy
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

WL conceptualized the article and wrote the original draft. KS developed the models with input from WL, analyzed the data, and created figures. KS and WD provided critical feedback in article revisions.

Corresponding author

Correspondence to William B. Leacock.

Ethics declarations

Ethics approval and consent to participate

Capture and handling procedures were approved by the Institutional Animal Care and Use Committees (ADFG ACUC 07–08, USFWS IACUC Permit 2012008, USFWS IACUC Permit 2012008 Renewal, and USFWS IACUC Permit 2015-001).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Leacock, W.B., Smith, K.T. & Deacy, W.W. Travel specific resource selection by female Kodiak brown bears during the sockeye salmon spawning season. Mov Ecol 12, 77 (2024). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-024-00513-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s40462-024-00513-6

Keywords

Movement Ecology

ISSN: 2051-3933