Printer Friendly

Distribution of hauled-out ladoga ringed seals (pusa hispida ladogensis) in spring 2012.

(Received 5 October 2012; accepted in revised form 4 March 2013)

ABSTRACT. The spatial distribution and habitat selection of the Ladoga ringed seal (Pusa hispida ladogensis), an endangered freshwater seal, are poorly understood, particularly for the ice-covered period. A fixed-wing, strip-transect aerial survey conducted in early April 2012, before the Lake Ladoga breakup, provided data on seal density and distribution throughout the lake in relation to several environmental covariates: depth, distance to shore, recreational ice-fishing activity, and ice type. A predictive model was applied to combinations of covariates to estimate the total number of seals hauled out on ice of Lake Ladoga. The model estimate was 5068 (95% CI: 4026-7086) seals over an area of 16 827 km2. The mean observed seal density was 0.29 seals/km2 (SD = 0.351, range from 0 to 8.61), and density was highest (> 1 seal/km2) in regions that were relatively shallow (<50 m). Densities appeared to increase with distance from shore but dropped off again at the longest distances. The average density was lower in fast ice habitats (0.13 seals/km2) than in drifting pack ice habitats (0.44 seals/km2). Relatively high seal densities observed in "ice edge" zones (0.26 seals/km') could be explained by the ice formation pattern of large ridged and hummocked areas in the transition zone between shorefast ice and secondary ice. The presence of fishermen had a highly significant negative effect on seal presence (f3 =--7.8,p = 0.0014), resulting in a nearly twofold decrease in seal density in shorefast ice habitats (0.09 seals/km2 in fishing areas and 0.15 seals/km2 in areas without fishing activity). An extensive winter recreational fishery, in combination with potential negative trends in ice conditions on the lake, might reduce the amount of suitable habitat for the Ladoga ringed seal in the near future.

Key words: Ladoga ringed seal, ice associated seals, distribution, fisheries, ecological factors, modeling

RESUME. La repartition spatiale et la selection de l'habitat du phoque rnarbre de Ladoga (Pusa hispida ladogensis), phoque d'eau douce en vole de disparition, ne sont pas bien comprises, surtout en cc qui a trait a la periode d'englacement. Un leve aerien a base de transects en bandes effectue au debut du mois d'avril 2012, avant la debacle du lac Ladoga, a permis d'obtenir des donnees sur la densite et la repartition du phoque a l'echelle du lac par rapport a plusieurs covariables environnementales : la profondeur, la distance jusqu'au rivage, la Oche recreative sous la glace et le type de glace. Un modele predictif a ete applique a des combinaisons de covariables afin d'estimer le nombre total de phoques qui se hissent sur la glace du lac Ladoga. Le modele a permis d'obtenir une estimation de 5 068 (IC de 95 % : 4026-7086) phoques dans une aire de 16 827 [km.sup.2]. La densite moyenne des phoques observes t:s.tait de 0,29 phoque/[km.sup.2] (ecart-type = 0,351, ecart allant de 0 a 8,61), et la densite etait plus elevee (> 1 phoque/[km.sup.2]) dans les regions relativement peu profondes (< 50 m). Les densites semblaient augmenter en fonction de la distance du rivage, mais elles baissaient de nouveau lorsque les distances etaient plus longues. La densite moyenne etait moms elevee dans les habitats a glace rapide (0,13 phoque/[km.sup.2]) que dans les habitats a banquises en derive (0,44 phoque/[km.sup.2]). Les densites de phoques relativement elevees qui ont ete observees dans les zones de [much less than] lisieres de glaces [much greater than] (0,26 phoque/[km.sup.2]) pouvaient s'expliquer par le modele de la formation de glace des grandes zones de glaces tourmentees et moutonnies faisant partie de la zone de transition entre la banquise cotiere et la glace secondaire. La presence de pecheurs avait un effet considerablement negatif sur la presence des phoques ([beta] = -7,8, [rho] = 0,0014), cc qui se traduisait par une diminution de presque la moitie de la densite de phoques dans les habitats a banquise cotiere (0,09 phoque/[km.sup.2] dans les zones de Oche et 0,15 phoque/[km.sup.2] dans les zones oil il n'y avait pas de peche). L'intensite de la Oche recreative en hiver, allide aux tendances potentiellement negatives en matiere d'etat des glaces du lac, pourrait avoir pour effet de rapetisser la quantite d'habitat habitable par le phoque marbre de Ladoga dans un avenir rapproche.

Mots cles : phoque marbre de Ladoga, phoques en fonction des glaces, repartition, pecheries, facteurs ecologiques, model isation

Traduit par la revue Arctic par Nicole Giguere.

INTRODUCTION

The Ladoga ringed seal (Pusa hispida ladogensis, Nord-quist, 1899) is an endemic ringed seal subspecies inhabiting Lake Ladoga in northwestern Russia (Fig. 1). Its population has been isolated from the Baltic Sea population for more than 8000 years (Muller-Wille, 1969; Hyvarinen and Niem-Men, 1990), and during this time it has successfully adapted to a freshwater environment atypical for most marine mammals. Intensive harvesting during the 20th century caused a severe decline of the Ladoga ringed seal population (Tormosov and Philatov, 1973; Sipila et al., 2002), which decreased dramatically from an estimated 20 000 in the 1930s (Chapskiy, 1932) to estimates of 3000-5000 in the mid-1970s (Zheglov and Chapskiy, 1971; Antonuk, 1975) and 3000 in 2001 (Verevkin, 2003). Harvest of the Ladoga ringed seal was banned in Russia in 1980 as a conservation measure (Sipila and Hyvarinen, 1998) and the subspecies was included in the Red Data Book (Danilov-Danil'yan et al., 2000). The international scientific community's concern about the future of the Ladoga ringed seal led the International Union for Conservation of Nature to declare it an endangered subspecies (Kovacs et al., 2012).

There is concern about the interaction of ringed seal with fisheries in the area; high by-catch of seals is considered to be one of the major threats to the survival of this subspecies. Conflicts between seals and commercial fishermen are increasing because the seals cause significant damage to fishing gear and hauls, which makes fishing unprofitable in most areas of the lake (Trukhanova et al., 2012). Furthermore, a large portion of the, coastal fast ice is used by recreational ice fishermen during the seal pupping period, which may indirectly affect pup survival.

Aside from fishing-related effects, key features of Ladoga seals' behavior and seasonal movements are influenced by their freshwater environment. The Ladoga ringed seal and the Saimaa ringed seal (P. h. saimensis) are the only two subspecies of ringed seal that inhabit freshwater lakes year-round, which has led to marked differences from marine subspecies in habitat preferences and movements (Berta and Churchill, 2011). The marine breeding habitats of ringed seals are associated with landfast ice and pack ice areas (e.g., McLaren, 1958; Finley et al., 1983; Lydersen and Gjertz, 1986; Kunnasranta, 2001), while Saimaa ringed seals give birth in snowdrifts on the shorelines of islands or islets close to the shore (Sipila, 1990). The Ladoga ringed seal is thought to breed in hummocked areas located along the fast ice band or, more seldom, in snowdrifts in the northern part of the lake (Antonuk, 1975; Sipila et al., 1996; Kunnasranta, 2001). Because the entire lake does not freeze completely every year, variations in ice extent affect seal distribution from season to season (Antonuk, 1975; Philatov, 1990). Both Lake Saimaa and Lake Ladoga are relatively shallow water basins (Kuusisto, 1999; Kuderskiy, 2009) but regional differences in depth and temperature regimes influence seasonal fish migrations (Kuderskiy, 2009), which then influence seal distribution (Harkonen et al., 2006).

The Ladoga ringed seal, like all subspecies of the ringed seal, is pagophilic, using ice for pupping, pup nursing, resting, and the annual molt. The breeding behavior of the Ladoga seal has not been documented, but most likely the pupping season lasts from March to the beginning of April (Agafonova et at., 2007). It is thought that lactation lasts 1.5-2 months and that mating takes place during the mid to late stages of lactation (Chapskiy, 1932). In early April, seals begin to molt and tend to bask on the ice surface or rockeries until mid-May, ensuring better insulation for their skin (Feltz and Pay, 1966; Kelly, 1988). These facts suggest that the ringed seals, especially breeding animals, are likely to be sedentary during the ice season and tend to stay near their lairs, while continuously maintaining breathing holes in the ice and nursing pups. Subadult seals tend to travel much greater distances during winter (Smith, 1987; Gjertz et al., 2000; Freitas et at., 2008) since they are not constrained by establishing and maintaining birth lairs. During periods of severe ice conditions, however, they also depend on permanent breathing holes.

The seasonal activities of Ladoga seals are affected by anthropogenic and environmental factors that influence birth and survival rates, distribution, and abundance of the subspecies (Agafonova et at., 2007). The factors that are most influential in determining the spatial use, distribution, and ultimately, the abundance of the Ladoga ringed seals remain largely unknown. In this paper, we use results of an aerial survey of ice seals performed in Lake Ladoga in April 2012 to model densities and habitat selection of the Ladoga ringed seal on spring ice throughout the lake with respect to bathymetry, proximity to the shoreline, ice type, and use of the area by recreational fisheries in winter. Our specific goals are to identify and interpret the physical features of the lake environment that are most directly related to space use by seals, to generate density maps over the entire lake, and to quantify the effect, if any, of recreational fisheries. We also discuss implications for the management of this population.

MATERIAL AND METHODS

Aerial Survey

The Ladoga ringed seal aerial survey, conducted on 4, 5, 6, and 14 April 2012, encompassed ice-covered areas of Lake Ladoga at the beginning of the annual molt, using standard line transect methods (see, e.g., Skalski et al., 2005). Flights took place primarily between 1000 and 1500 (local time) under sunny or mainly clear weather conditions. Three observers collected the data from a fixed-wing, low-noise aircraft (Cessna-182T) flying along strictly east--west oriented linear transects. Observations were made from both sides of the aircraft to 413 m distance on either side, i.e., along a strip 826 m wide. Observers maintained a consistent strip width by aligning markers placed on the windows and the wing struts of the aircraft. The flight altitude and ground speed were controlled by the onboard navigation system and maintained as consistently as possible at 90 m and 190 km/h. In total, 24 survey transects (Fig. 1) covered 1748 [km.sup.2] (10.4%) of ice surface and 109 [km.sup.2] of water surface (the water surface was excluded from the analysis). Each seal detected on ice was photographed, and its coordinates were determined using hand-held Garmin cs62 GPS units attached to the cameras. For the purpose of the analysis, we used a set of waypoints, each of which marked individual animals seen within transects.

Physical Covariates

Lake Ladoga (Fig. 1) is a large (17 872 [km.sup.2]), relatively shallow subarctic lake in northwestern Russia. Its maximum extent is 125 km (E-W) and 219 km (N-S); mean depth is 48 m and maximum depth is 228 m. The coastline is represented by flat, sandy beaches and wetlands except on the northern shore, referred to as the "skerries region," which is marked by fjord-like inlets and rocky cliffs. Typically, ice forms on the lake in early December and reaches its maximum extent in February, usually covering the entire lake.

We subdivided Lake Ladoga into a grid of equal blocks (5 [km.sup.2] each), 491 of which were included in transect coverage. Three physical factors (depth, distance to the shore, and ice type) and the presence or absence of recreational fishermen were covariates chosen for the analysis and modeling of seal distribution. Depth for each individual block in Lake Ladoga was obtained from Garmin BlueChart map (ver. [3.01.sup.[C]] 2011 Garmin Ltd). Distance to the shore (proximity) was measured as the shortest straight line from the block center point to the nearest lakeshore, using Garmin BaseCamp (ver. [3.3.2.sup.[C]] 2008-12 Garmin Ltd).

We determined several ice types for the analysis on the basis of satellite imagery AERONET_Helsinki (for 5 April 2012) with 250 m resolution provided by NASA and downloaded in the 7-2-1 frequency band for easier interpretation (http://rapidfire.sci.gsk.nasa.gov). We assumed that the total ice area was constant over the survey period and that the ice did not move significantly. In winter 2011-12, ice began to form on the lake in late January, when a narrow (5-10 km) fast ice band appeared along the shoreline. Ice coverage was 100% by the beginning of March, and breakup began about a week later. When the survey started, the central part of the lake was covered with extensive ice floes or groups of ice floes frozen together (hereafter referred to as "drifting pack ice"). During the winter, this area had stable ice that was associated with fast ice zones. However, it was the first area to break up in early spring. The fast ice band still remained untouched along the western, eastern, and southern shores and completely covered the northern skerries region of the lake (to the north of Valaam Archipelago). Fast and drifting pack ice areas were divided by open water areas. Another specific habitat type was the ice edge. Any block that covered both ice and water areas was classified as "ice edge," and the area covered by ice in this category was averaged to be 2.5 [km.sup.2].

Twelve main winter fishing areas were defined on the basis of interviews with recreational and commercial fishermen and through data from an aerial survey that photographed and mapped concentrations of fishermen on the ice. This covariate was entered in the model as "true" if there was any ice fishing activity in a given block, and "false" in the opposite case.

ANALYSIS

Analysis and modeling included specifying and fitting a negative binomial generalized linear model (NBGLM) of seal count data as a function of several covariates in various combinations. The model is written as:

[N.sub.i]= NB ([[mu].sub.i], k)

log([[mu].sub.i]) = f(.)

where NB refers to the negative binomial distribution with fitted dispersion parameter k and mean [[mu].sub.i] is the expected number of seals in a given block, modeled as the exponential of a linear combination f(.) of the covariates. The NBGLM was chosen over standard Poisson GLM in order to account for over-dispersion in the count data (Zuur et al., 2009), and preferred over a quasi-Poisson model because a well-defined likelihood for the negative binomial facilitated comparison of models, and because a diagnostic plot of means against variances suggested a squared relationship more appropriate to a negative binomial model (VerHoef and Boveng, 2007).

The response variable in the fitted models was the number of seals in the blocks. Explanatory variables were the factors listed above: depth and proximity to the shore were computed at all locations of the seal sightings and included as continuous covariates. Because of observed non-linear responses of seal counts to depth and distance from shore (Fig. 2A, 2C), we also added square root terms, depthu [sup.0.5](d2) and proximity [sup.0.5](p2), to the set of fitted models. Ice type was a categorical covariate. The models we fitted ranged from the null model (MO) of a constant linear predictor to the most complex model (M36), which included linear and square root terms of depth and proximity, as well as their interaction with each other and with the ice-type covariate [in formula notation: (depth + [depth.sup.0.5])*(proximity + [proximity.sup.0.5])*ice-type]. Note that we used the square root term (rather than, e.g., a square term) to capture nonlinear structure because the square root transformation has the property of keeping predictions from increasing unrealistically at higher values of depth or distance from shore. We chose to use non-linear terms in the GLM over a generalized additive model (GAM) because the GLM provided a good fit without the additional levels of complexity and decisions required for fitting a GAM.

We compared all 37 models using Akaike's Information Criterion (Burnham and Anderson, 2002). As a measure of goodness of fit, we used proportion deviance explained (PDE), which is analogous to an adjusted [R.sup.2] measure (Wood, 2006):

PDE = 100 x (1--residual deviance / null deviance).

Because ice type was a highly significant interaction factor (see RESULTS), we modeled the seal densities in each different habitat separately and selected models as described above, comparing negative binomial GLMs that included various combinations of the depth, proximity, and non-linear terms. Because fishery activities were identified almost exclusively in the shorefast ice type, we additionally added fisheries presence as an explanatory covariate in that ice habitat.

We obtained an uncorrected estimate of total seal abundance by subdividing the lake into a grid of blocks of the same size as the transect blocks used in the modeling and obtaining a model prediction using a hybrid combination of the most parsimonious models in each habitat as selected by AIC. Confidence intervals around the point estimate were obtained by using a bootstrap method (Manly, 1997), whereby we resampled our observed data with replacement and re-estimated the total abundance 1000 times. We report the 2.5% and 97.5% quartiles of the bootstrapped distribution. The predicted abundance was converted into densities and mapped in order to compare the structure of relative densities across the lake region.

For all analyses, we used the R statistical programming language (R Development Core Team, 2008), including the "MASS" package (Venables and Ripley, 2002) for negative binomial generalized linear modeling and the "maps" and "mapdata" packages (Becker and Wilks, 1993) for mapping.

RESULTS

Lake Ladoga is characterized roughly by two regions: a shallow, sandy-bottomed zone in the southern part of the lake (mean depth ca. 30 m), and a deepwater zone in the northwest, characterized by reefs and skerries (mean depth ca. 130 m). For the purposes of modeling and fitting, we subdivided the lake into 3460 blocks of 5 k[m.sup.2] and determined depths, proximity to shore, ice type, and presence of fisheries for each block. Mean depth measurements ranged from 2 m along the shoreline to 200 m in the deepest area (mean 43.4 m, SD = 37.9 m, Fig. 2C). Distance to shore for each block was determined at 5 km resolution and ranged from 0 to 50 km in the central part of the lake. At the beginning of April 2012, out of a total area of 17 872 k[m.sup.2], the lake had only 1045 k[m.sup.2] (5.9%) of open water. The ice-covered portion had somewhat more drifting pack ice (8860 km[m.sup.2], 49.6% of total area) than fast ice (7367 k[m.sup.2], 41.2%). The remaining 600 k[m.sup.2] (3.6%) was classified as ice edge. We estimated that around 2800 k[m.sup.2] of fast ice was actively used for ice fishing in April 2012 (defining "use" as the presence of fishery activity within a 5 k[m.sup.2] block).

The seal survey data used for modeling consisted of 491 blocks, each with an area of 5 k[m.sup.2]. Seals were sighted in 179 (36.4%) of these blocks. Sightings ranged from 1 seal (81 blocks) to 32 seals (1 block), with a total of 558 individual seals sighted. The mean observed seal density was 0.29 seals/k[m.sup.2] (SD = 0.351, ranging from 0 to 8.61). We identified patterns in the relationship of the sightings to environmental covariates (Figs. 2, 3). In particular, densities appeared to increase with distance to shore but dropped off again at the longest distances (Fig. 2A), and similarly, densities were highest at intermediate depths (Fig. 2C). Finally, densities tended to be higher on the drifting ice (mean 0.44 seals/k[m.sup.2], Fig. 2B) than on the fast ice (mean 0.13 seals/k[m.sup.2]).

We fitted 37 NBGLMs and report the A1C, AA1C, and proportion deviance explained in Table 1. Most of the best models were among the more complex models. The model with the lowest AIC was:

TABLE 1. Model selection table for the Ladoga ringed seal
count data (April 2012) with all possible combinations of
distance to shore (proximity) and depth and their squares
(p2 and d2, respectively), the three ice types, and all
interactions considered as explanatory variables. Models
are sorted by increasing [DELTA]AIC, and the lowest AIC
and highest partial deviance explained (PDE) models are
shown in bold type.

Model  Formula            K     AIC  [DETA]AIC   PDE

36     ~ ((depth + d2)   15  1254.0        0.0  0.23
       + (proximity +
       p2)) * ice-type

31     ~ ((depth + d2)   12  1259.7        5.7  0.21
       + proximity) *
       ice-type

32     ~ ((depth + d2)   18  1263.5        9.5  0.23
       * proximity) *
       ice -type

34     ~ (depth +        12  1264.2       10.2   0.2
       (proximity +
       p2)) ? ice-type

37     ~ ((depth + d2)   27  1264.5       10.4  0.26
       * (proximity +
       p2)) * ice-type

30     ~ (depth + d2) *   9  1265.1       11.1  0.18
       ice-type

20     ~ ((depth + d2)    8  1265.4       11.4  0.18
       * proximity) +
       ice-type

25     ~ ((depth + d2)   11  1266.9       12.8  0.19
       * (proximity +
       p2)) + ice-type

8      ~ ((depth - d2)    6  1267.1       13.1  0.16
       * proximity)

35     ~ (depth *        18  1267.1       13.1  0.22
       (proximity +
       p2)) * ice-type

28     ~ (depth +         9  1269.0       15.0  0.17
       proximity) *
       ice-type

26     ~ depth *          6  1269.7       15.7  0.16
       ice-type

17     ~ (depth *         6  1269.8       15.8  0.16
       proximity) +
       ice-type

13     ~((depth + d2) *   9  1271.0       17.0  0.17
       (proximity +
       p2))

29     ~ (depth *        12  1271.7       17.7  0.18
       proximity) *
       ice-type

23     ~ (depth *         8  1272.3       18.3  0.16
       (proximity +
       p2)) + ice-type

5      ~ (depth *         4  1273.1       19.1  0.14
       proximity)

11     ~ (depth *         6  1275.5       21.5  0.14
       (proximity +
       p2))

19     ~ ((depth + d2)    6  1284.9       30.9  0.12
       + proximity) +
       ice-type

24     ~ ((depth + d2)    7  1286.3       32.3  0.12
       + (proximity *
       p2)) + ice-type

12     ~ ((depth + d2)    5  1287.2       33.2  0.11
       + (proximity +
       p2))

16     ~ (depth -         5  1287.9       33.9  0.11
       proximity) +
       ice-type

18     ~ (depth + d2) *   5  1287.9       33.9  0.11
       ice-type

7      ~ ((depth + d2)    4  1288.4       34.4   0.1
       + proximity)

10     ~ (depth +         4  1288.9       34.9   0.1
       (proximity +
       p2))

14     ~ depth +          4  1289.0       35.0   0.1
       ice-type

22     ~ (depth +         6  1289.1       35.1  0.11
       (proximity +
       p2)) + ice-type

4      ~ (depth +         3  1290.1       36.0  0.09
       proximity)

33     ~ (proximity +     9  1291.8       37.8  0.12
       p2) * ice-type

15     ~ proximity +      4  1293.7       39.7  0.09
       ice-type

27     ~ proximity *      6  1294.1       40.1   0.1
       ice-type

21     ~ (proximity +     5  1295.5       41.5  0.09
       p2) * ice-type

9      ~ (proximity +     3  1299.8       45.8  0.07
       p2)

3      ~ proximity        2  1299.9       45.9  0.06

2      ~ depth            2  1313.1       59.1  0.03

6      ~ (depth + d2)     3  1314.9       60.8  0.03

1      ~ 1                1  1321.7       67.7     0


M36: seal number~((depth + d2)+ (proximity + p2)) * ice-type

which explained 24% of the total variance. The most complex model, M37, ranked 5th, accounting for the absolute maximum of 26% of the PDE. Note that all of the top 10 models included a square root term for either proximity or depth. Further, all of the top six models included an interaction with ice type, suggesting that a unique set of parameters modeled the number of seals in each ice type.

Because the ice-type interaction was clearly an important one, we fitted the complete additive model (depth + [depth.sup.0.5] + proximity + [proximity.sup.0.5]) to the seal sightings in each of the three ice types. The values for the coefficients and their respective p-values (using Type III ANOVA, i.e., sequential chi-squared likelihood ratio tests after removing each of the terms) are displayed in Table 2. All the slope coefficients for the drifting pack ice are significantly different from zero, suggesting that there are strong, structural relationships between seal abundance and these physical covariates. There is a preference for shallower waters in the drift ice (negative first-order slope) that is exaggerated by a positive square root term. In contrast, there is a weaker positive relationship with distance to shore, which is attenuated by a negative square root term. Interestingly, the signs on the slope coefficients for the shorefast ice are exactly inverted compared to the drift ice. In particular, there is a negative first-order relationship with distance from shore, but a gradual increase with depth, though these effects are less significant (p-values between 0.05 and 0.10). The edge habitat displayed the highest magnitude of response to distance from shore, but no effects were significant, in part because of small sample sizes in this transitional category.

TABLE 2. Estimated regression coefficients, including the
mean and square-root effects of distance to shore (proximity)
and depth on seal abundance, for each of the three ice types.
P-values (in parentheses) are obtained from sequential
chi-squared comparisons of likelihoods (Type III ANOVA).
Marginally significant values (p < 0.10) are italicized,
significant values (p < 0.05) are in bold type. and highly
significant values (p <0.01) are italic and bold.

                     Shorefast            Edge  Drifting pack
                           ice  (transitional)            ice

N                          186              41            264

depth (m)                -0.04   0.051 (0.359)   0.064(0.009)
                       (0.075)

[depth.sup.0.5]           0.67   -0.96 (0.221)  -1.3 (<0.001)
                       (0.042)

proximity (km)           0.213    0.254(0.632)   -0.19(0.024)
                       (0.096)

[proximity.sup.0.5]       -1.3    -2.4 (0.574)     2.5(0.006)
                       (0.108)


Mean seal density for fishermen-free areas was 0.15 seals/k[m.sup.2] (SD = 0.32), compared to 0.09 seals/k[m.sup.2] (SD = 0.33) for the areas occupied by the fishermen. We modeled the effect of fisheries separately in the fast ice habitat where the fisheries occur by adding the fisheries presence categorical variable to the depth and proximity first- and second-order terms and guiding model selection with AIC as before. Again, there were 37 models to compare, of which the top 10 are presented in Table 3. Four of the five best models included the presence of fisheries, and the best model included both depth covariates, but not distance to shore (M30: seal number ~ (depth + d2) * Fisheries). The coefficients and Type III ANOVA results for this model are presented in Table 4.

TABLE 3. Model selection table for Ladoga ringed seal
abundance (top 10 models) in the shorefast ice with
additional fisheries presence variable (Fisheries).
Models are sorted by increasing AIC, and the lowest AIC
and highest partial deviance explained (PDE) models are in bold.

Model  Formula             K    AIC  [DETA] AIC    PDE

30     ~ (depth + d2) *    6  307.9         0.0  0.148
       Fisheries

31     ~ ((depth + d2) +   8  309.3         1.4  0.170
       proximity) *
       Fisheries

36     ~ ((depth + d2) +  10  311.5         3.6  0.186
       (proximity + p2))
       * Fisheries

6      ~ (depth + d2)      3  312.5         4.6  0.047

18     ~ (depth + d2) +    4  312.8         4.8  0.064
       Fisheries

1      ~ 1                 1  313.1         5.2  0.000

12     ~ ((depth + d2) +   5  313.8         5.9  0.073
       (proximity + p2))

3      ~ proximity         2  314.3         6.4  0.009

7      ~ ((depth + d2) +   4  314.4         6.5  0.048
       proximity)

2      ~ depth             2  314.6         6.7  0.006
TABLE 4. ANOVA table for model of seal abundance in the
shorefast ice. Significant values (p < 0.05) are in bold,
and highly significant (p < 0.01) are italic and bold.

Factor        [beta]  Likelihood ratio  Df         p (>
                            ehi-square      chi-square)

depth         -0.008               0.1   1       0.7553

d2             0.131              0.11   1       0.7351

Fisheries     -7.780             10.27   1       0.0014

depth:        -0.159              7.36   1       0.0067
Fisheries

d2:Fishcries   2.482              8.46   1       0.0036


The fisheries main effect was highly negative ([beta] = -7.8) and highly significant (Type III ANOVA, p = 0.0014), indicating that seals were avoiding the presence of fishermen on the fast ice. While the main effects of depth and square root of depth were weak and insignificant, the fisheries interaction with both of these covariates was highly significant, with coefficients suggesting an overall preference for deeper waters, especially in the presence of fishery activities.

An analogous analysis for the edge ice type suggested a best model with depth and square root of depth effects (PDE 17%, Type III ANOVA p = 0.06 and 0.04 for depth and [depth.sup.0.5] respectively) and no significant distance to shore effects.

We used these results to develop three different prediction models, one for each of the three ice types. The models can be written as:

[N.sub.s]([d.sub.i] [f.sub.i] | fast ice) = NB[[[alpha].sub.0] + (1 - [f.sub.i])([[alpha].sub.3] [d.sub.i] + [[alpha].sub.0] sqrt([d.sub.i])) ([[alpha].sub.3] [d.sub.i] + [[alpha].sub.4] sqrt([d.sub.i])))]

[N.sub.s]([d.sub.i] | edge ice) = NBA [[[beta].sub.0] + [[beta].sub.2] [d.sub.i] + [[beta].sub.3] sqrt([d.sub.i])]

[N.sub.s]([d.sub.i] [p.sub.i] | drift ice) = NB[[[gamma].sub.0] + [[gamma].sub.2] [d.sub.i] + [[gamma].sub.3] sqrt([d.sub.i]) + [[gamma].sub.4] [p.sub.i] + [[gamma].sub.5] scirt([p.sub.i])

where [N.sub.s] represents the number of seals in a block at depth [d.sub.i], distance from shore [p.sub.i], and fishery presence (0 or 1) denoted [f.sub.i]. The coefficients of the linear predictors are [alpha], [beta], and [gamma]. We applied the prediction model to the combinations of covariates in each of 3640 blocks covering the area of the lake and plotted the distribution of predicted densities (Fig. 4).

Maximum seal densities were predicted to be in the central part of the lake with a distinct spike (> 1.0 seal/k[m.sup.2]) in the zone with depths of less than 50 m (see Fig. 4). That prediction corresponds well with actual sighting data: most sightings were in the 20-30 m depth range (see Fig. 2). Density decreased to 0.4-0.6 seals/k[m.sup.2] in deeper areas (50-100 m). The northern deepwater skerries area covered with fast ice and the southernmost bays had the lowest seal density.

Estimation of total seal abundance was not possible because we do not have the necessary data on detection and seal hauling-out probabilities in a given time. However, the final model did allow estimation of the expected number of hauled-out seals based on a combination of covariates. The estimate of hauled-out seals for early April 2012 was 5068 seals (95% CI: 4026-7086) and to our understanding was likely to be biased low.

DISCUSSION

The survey was performed before the spring breakup of ice, so that the ice cover was stable during the survey period. We therefore consider our predicted density map to be a reasonable representation of ringed seal distribution throughout the lake during the ice season in 2012. There is, however, a possibility that seal counts in the skerries region were biased downward as lairs located close to the shoreline were still partly covered with snow at the time of the survey and therefore harder to detect.

Despite the fact that our density map corresponds well with the general location of seal sightings made during the survey, the model explains less than 25% of the count deviance. This fact suggests that other factors, which we did not observe or include in the model, were affecting seal distribution. One possible candidate is behavioral aggregation of individuals leading to unaccounted spatial auto-correlation (Dormann et al., 2007). Possible reasons for seals to aggregate in springtime include a preference for leads in the ice vs. breathing holes, food availability in specific areas, or avoidance of sources of disturbance. Also, male seals may aggregate close to mature females nursing their pups and wait until mating occurs; such a reproductive behavior strategy has been demonstrated, for example, for the Caspian seal (Phoca caspica; Wilson et al., 2012).

Landlocked seal populations have stricter constraints on their spatial distribution than most seals in marine environments. These constraints make them more susceptible to random variations in the environment, such as lack of suit-. able habitat arising from ice formation patterns in a given year, or trends in fish stock availability. The GLM framework we used for our habitat-based distribution modeling assumes to some extent that the entire lake area is accessible to the seals, without taking into account any constraints on the availability of those areas or the residual effect of movement from the ice-free period. In early winter when ice forms, the seals choose appropriate habitat on the basis of obvious ice characteristics and possibly other factors, such as prey availability. Fast ice zones have traditionally been considered the habitat most actively used by ringed seals for making lairs and pupping (Sipila et al., 1996; Kunnasranta et al., 2001). However, 38% of the fast ice area in Lake Ladoga is actively used for recreational ice fishing, which makes these areas less suitable for breeding seals. From early winter until late spring, thousands of recreational fishermen and some commercial fishermen use the fast ice for ice fishing, a traditional winter recreation in the region. Fishermen use skis, snowmobiles, and cars to reach remote areas and sometimes spend several days living in tents on thick ice. Winter fishing areas may vary from season to season depending on ice conditions, although in general, traditional fishing sites remain consistent. Even under ideal ice conditions, a seal is unlikely to breed close to high concentrations of fishermen or intense transport traffic on ice. This assertion is supported by our data and modeling results: very few seals were detected in these areas during the aerial survey, and the NBGLM suggested a strong avoidance of ice-fisherman areas, even controlling for distance to shore effects.

Local fishermen have told us that some seals use ice-fishing holes as breathing holes and even occasionally feed from the nets set under the ice. These behaviors are more likely typical of non-breeding seals, since breeding females tend to remain in their territories throughout the winter to maintain their lairs and breathing holes, while breeding males tend to concentrate near females during the lactation period until mating occurs. These hypotheses are based on observations and more detailed documentation of Caspian seal breeding behavior (Wilson et al., 2012).

Thus, we conclude that disturbance caused by humans has become a major factor influencing seal distribution and has likely shifted the core area of seal pupping from the shorefast ice to the central part of the lake. In combination with an apparently increased population size (compared to the previous aerial survey estimates of Verevkin, 2003), this shift could lead to higher seal densities in the central part of the lake, outside the fast ice zone and more distant from the shore. A number of pups detected during the survey in the central part of the lake demonstrate that the seals do breed in remote areas away from shore without a high level of dependence on fast ice, though it should be kept in mind that the ice cover, wherever it is formed, must be stable enough during the pupping season to provide a platform for pup rearing.

Relatively high seal densities observed in ice edge zones (0.26 seals/[km.sup.2]) in our opinion could be well explained by ice formation patterns. In early winter, after formation of shorefast ice, new ice floes are brought by the wind to the ice edge, where the ice breaks, freezes together, and forms hummocked zones suitable for lair construction. When the breakup process starts in spring, it begins in the boundary zone between primary and secondary ice, where seal density is originally high. As snow melts and a polynya appears in this particular area, we can expect high numbers of animals to stay along the ice edge.

Unfortunately, very little historical information is available on seal winter distribution patterns. Several aerial surveys conducted on the lake in the 1950s, 1970s, and 1980s either did not report registered seal locations at all or had rather poor coverage because of early ice melt or limited survey extent (Sorokin, 1957; Zheglov and Chapskiy, 1971; Antonuk, 1975; Philatov, 1990). In March--April 1973, a series of aerial surveys (Antonuk, 1975) showed that the majority of the seals were hauling out in the Svirskaya Bay area (0.33-0.76 seals/km2), while another dense grouping (0.14-0.20 seals/km2) was found in the far northeastern part of the lake, either on ice fields or on fast ice. Such high densities might have been a result of poor ice coverage, since less than 30% of the Lake Ladoga surface was frozen in that winter. Later, Bychkov and Antonyuk (1975) published a ringed seal spring-winter haul-out map identifying the eastern part of the lake (to the east of 32.0[degrees] E) as an area commonly used by the seals in the ice-bound season. Similar results were achieved by Zheglov and Chapskiy (1971), although they also reported occasional seal sightings in the skerries region (0.028 seals/[km.sup.2]) and on broken fast ice along the western shore (0.19 seals/[km.sup.2]). Although the ice coverage during their survey period was close to 80%, their survey transects did not include the central part of the lake, which left a gap in their seal density data. The only brief note on females and pups occupying the central part of the lake in severe winters is found in Philatov (1990), but without any further clarification. Hence we can see a significant data deficiency that does not allow a comparison of the modeled density distribution for winter-spring 2012 with density distributions of previous years. We assume that in mild winters, incomplete ice coverage makes seals concentrate in all available habitat types regardless of distance to shore or bathymetry, since the low ice coverage facilitates higher mobility for foraging trips. In colder winters, animals "choose" preferable areas in relation to environmental characteristics and might even have an opportunity to maintain comfort distances from other individuals on the ice.

The primary factor for the seals in any season is prey availability (Smith, 1987), though this is among the most difficult variables to assess directly. Because of constraints on long-distance movements during the ice season (discussed above), the ability to catch fish close to wintering sites becomes extremely important. The data available on fish biomass distribution (Rumyantsev, 2002) show a possible correlation between seal habitat use in winter and concentration of fish species such as European smelt (Osmerus eperlanus), vendace (Coregonus albula), and European whitefish (Coregonus lavaretus spp.). To test this hypothesis and quantify these relationships, however, we require additional data and a more dedicated study.

Among marine mammals, ice-associated seal species are among the most vulnerable to climate warming, at both global and regional scales, suffering from compounded habitat loss and changes in prey abundance (Laidre et al., 2008; Kovacs et al., 2011). Among the ice-associated seals, the landlocked species, such as the Caspian seal, the Baikal seal (Phoca sibirica), and the Ladoga and Saimaa ringed seal subspecies, are additionally limited by the constraints to dispersal or range-shifting that are intrinsic to landlocked bodies of water. The ringed seal in Lake Ladoga lives at the very southern edge of the global range of the ringed seal species, and therefore it is particularly vulnerable to shifts in the temperature regime, especially during the most critical stages of its life cycle. The lake is not completely covered with ice every winter (Usachev et al., 1985). Since 1947, maximum ice coverage has been trending negatively at an average rate of 0.18% less ice per year (Fig. 5, linear regression p= 0.066): it has attained 100% of the lake in only 36 of the 60 years, with a decreasing probability of total coverage (binomial GLM of probability 100% coverage against year: [beta] =-0.028, p= 0.038). Note that the maximum ice extent reached a historical low of 15% relatively recently, in 2008 (Fig. 5).

Light ice conditions, limited area of breeding habitat, and early breakup may lead to a decline in pup survival due to a higher number of stranded pups (Smith and Harwood, 2001; Alekseev et al., 2012) and greater exposure of pups to predation (Jussi et al., 2008). The combined effects of diminished ice conditions and extensive fishing activity on the lake in winter and spring might have a cumulative impact on the ringed seal birth rate. Despite current positive trends in population size, a changing ice regime and increasing fishing activity may put this subspecies at risk. On the basis of these considerations and the results of our analyses, we believe that areas with limited access for fishermen and other anthropogenic disturbances such as the Nizhnesvirskiy Strict Nature Reserve area and the restricted military zone that covers a large portion of the western shore between 60.4[degrees] and 60.8[degrees] N (Fig. 1) will be critical for conserving the population of Lake Ladoga seals.

There is an urgent need to assess the condition of Ladoga fish stocks (a variable we could not include in the model), to clarify fish species distribution, and to identify sustainable levels of fishing effort, both commercial and recreational. Population monitoring of seals, including number estimation, satellite tracking, and diet studies, should be performed on a regular basis. Ice conditions in any given year may ultimately be used to predict seal distribution and thus allow protection of breeding winter habitats. Further, understanding of the seals' seasonal movements, distribution patterns, and responses to increased anthropogenic pressures are also important to ensure conservation of the subspecies.

ACKNOWLEDGEMENTS

The study was conducted within the framework of the project "Assessment of ringed seal by-catch in commercial fisheries and its impact on seal populations in Lake Ladoga, Northwest Russia" supported by the Marine Mammal Commission (USA). The authors are grateful to the observers who contributed to data collection, M. Verevkin and V. Vysotsky. We thank aero club "Vzletim.ru" staff for making the aerial survey possible. We would like to acknowledge Dr. Kristin Laidre for her useful comments, which helped to improve the manuscript.

REFERENCES

Agafonova, E.V., Verevkin, M.V., Sagitov, R.A., Sipila, T., Sokolovskaya, M.V., and Shahnazarova, V.U. 2007. The ringed seal in Lake Ladoga and the Valaam Archipelago [in Russian]. Vammalan Kirjapaino Oy.

Alekseev, V.A., Andrievskaya, EM., and Trukhanova, I.S. 2012. Ringed seal and grey seal rehabilitation methodology [in Russian]. Actual Issues of Veterinarian Biology 3(15):72-85.

Antonuk, A.A. 1975. Estimation of total population number of the seal Pusa hispida ladogensis in Lake Ladoga [in Russian]. Zoological Journal 54(9):1371 -1377.

Becker, R.A., and Wilks. A.R. 1993. Maps in S. Statistics Research Report 93(2). Murray Hill, New Jersey: AT&T Bell Laboratories.

Berta, A., and Churchill, M. 2011. Pinniped taxonomy: Review of currently recognized species and subspecies, and evidence used for their description. Mammal Review 42(3):207 -234, doi:10.1111/j.1365-2907.2011.00193.x.

Burnham, K.P., and Anderson, D.R. 2002. Model selection and multimodel inference: A practical information-theoretic approach, 2nd ed. New York: Springer-Verlag.

Bychkov, V.A., and Antonuk, A.A. 1975. Ladoga ringed seal and its conservation [in Russian]. Zoological Journal 54(9):255-266.

Chapskiy, K.K. 1932. Ladoga seal and its harvest possibilities [in Russian]. Proceedings of Leningrad Scientific and Research Institute 13(2):147-157.

Danilov-Daniryan, V.I., Am irkhanov, A.M., Pavlov, D.S., Sokolov, V.E., Alimov, A.F., Darevskiy, I.S., Dezhkin, V.V., et al., eds. 2000. Red Data Book of the Russian Federation: "Animals" [in Russian]. AST, Astrel.

Dormann, C.F, McPherson, J.M., Araujo, M.B., Bivand, R., Bolliger, J., Carl, G., Davies, R.G., et al. 2007. Methods to account for spatial autocorrelation in the analysis of species distributional data: A review. Ecography 30(5):609-628, doi:10.1111/j.2007.0906-7590.05171.x.

Feltz, E.T., and Fay, F.H. 1966. Thermal requirements in vitro of epidermal cells from seals. Cryobiology 3(3):261-264.

Finley, K.J., Miller, G.W., Davis, R.A., and Koski, W.R. 1983. A distinctive large breeding population of ringed seals (Phoca hispida) inhabiting the Baffin Bay pack ice. Arctic 36(2):162-173.

Freitas, C., Kovacs, K.M., Ims, R.A., Fedak, M.A., and Lydersen, C. 2008. Ringed seal post-moulting movement tactics and habitat selection. Oecologia 155:193-204.

Gjertz, I., Kovacs, K.M., Lydersen, C., and Wiig, 0. 2000. Movements and diving of adult ringed seals (Phoca hispida) in Svalbard. Polar Biology 23:651-656.

Harkonen, T., Masi, M., Jussi., I., Verevkin., M., Dmitrieva, L., Helle, E., Sagitov, R., and Harding, K.C. 2006. Seasonal activity budget of adult Baltic ringed seals. PLoS ONE 3(4) e2006, doi:10.1371/journal.pone.0002006.

Hyvarinen, FL, and Nieminen, M. 1990. Differentiation of the ringed seal in the Baltic Sea, Lake Ladoga and Lake Saimaa. Finnish Game Research 47:21-27.

Jussi, M., Harkonen, T., Helle, E., and Jussi, I. 2008. Decreasing ice coverage will reduce the breeding success of Baltic grey seal (Halichoerus grypus) females. Ambio 37(2):80-85.

Kelly, B.P. 1988. Ringed seal. In: Lentfer, J.W., ed. Selected marine mammals of Alaska: Species accounts with research and management recommendations. Washington, D.C.: Marine Mammal Commission. 57-75.

Kovacs, K.M., Lydersen, C., Overland, J.E., and Moore, S.E. 2011. Impacts of changing sea-ice conditions on Arctic marine mammals. Marine Biodiversity 41(1):181-194, doi:10.1007/s12526-010-0061-0.

Kovacs, K.M., Aguilar, A., Aurioles, D., Burkanov, V., Campagna, C., Gales, N., Gelatt, T., et al. 2012. Global threats to pinnipeds. Marine Mammal Science 28(2):414-436, doi:10.1111/j.1748-7692.2011.00479.x.

Kuderskiy, L.A. 2009. Composition and commercial importance of fish stocks in Lake Ladoga. In: Lavrentieva, G.M., ed. Fisheries research of large lakes of northwestern part of European Russia. Collection of Scientific Papers of the State Research Institute of Lake and River Fisheries (GosNIORKh) No. 334. St. Petersburg: Dean. 138-212.

Kunnasranta, M. 2001. Behavioral biology of two ringed seal (Phoca hispida) subspecies in the large European lakes Saimaa and Ladoga. PhD Dissertations in Biology. University of Joensuu 7.

Kuusisto, E. 1999. Basin and balances. In: Kuusisto, E., ed. Saimaa: A living lake. Helsinki: Tammi. 21-39.

Laidre, K.L., Stirling, I., Lowry, L.F., Wiig, 0., Heide-Jorgensen, M.P., and Ferguson, S.H. 2008. Quantifying the sensitivity of Arctic marine mammals to climate-induced habitat change. Ecological Applications 18(2) Supplement: S97-S125.

Lydersen, C., and Gjertz, I. 1986. Studies of the ringed seal (Phoca hispida, Schreber 1775) in its breeding habitat in Kongsfjorden, Svalbard. Polar Research 4(1):57-63.

Manly, B.F.J. 1997. Randomization, bootstrap and Monte Carlo methods in biology, 2nd ed. Boca Raton, Florida: Chapman & Hall/CRC Press.

McLaren, 1.A. 1958. The biology of the ringed seal (Phoca hispida Schreber) in the Eastern Canadian Arctic. Bulletin 118. Ottawa: Fisheries Research Board of Canada.

Muller-Wille, L.L. 1969. Biometrical comparison of four populations of Phoca hispida Schreb. in the Baltic and White Seas and Lakes Ladoga and Saimaa. Commentationes Biologicae, Societas Scientiarum Fennica 31:1-12.

Philatov, I.E. 1990. Ladoga ringed seal [in Russian]. In: Sokolov, V.E., ed. Rare and endangered mammal species of USSR. Moscow: Nauka. 57-64.

R Development Core Team. 2008. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. http://www.R-project.org.

Rumyantsev, V.A. 2002. Lake Ladoga atlas [in Russian]. St. Petersburg: Institute of Limnology, Russian Academy of Sciences.

Sipila, T. 1990. Lair structure and breeding habitat of the Saimaa ringed seal (Phoca hispida saimensis Nordq.) in Finland. Finnish Game Research 47:11-20.

Sipila, T., and Hyvarinen. H. 1998. Status and biology of Saimaa (Phoca hispida saimensis) and Ladoga (Phoca hispida ladogensis) ringed seals. In: Heide-Jorgensen, M.P., and Lydersen, C., ed. Ringed seals in the North Atlantic. The North Atlantic Marine Mammal Commission Scientific Publication 1. 83-99.

Sipila, T., Medvedev, N.V., and Hyvarinen, H. 1996. The Ladoga seal (Phoca hispida ladogensis Nordq.). Hydrobiologia 322:193-198.

Sipila, T., Medvedev, N., Kunnasranta, M., Bogdanov, V., and Hyvarinen, H. 2002. Present status and recommended conservation actions for the Ladoga seal (Phoca hispida ladogensis) population. WWF Suomen Report 15. Metsahallitus: Natural Heritage Services.

Skalski, J.R., Ryding, K.E, and Millspaugh, J.J. 2005. Wildlife demography: Analysis of sex, age and count data. Oxford: Elsevier Academic Press.

Smith, T.G. 1987. The ringed seal, Phoca hispida, of the Canadian western Arctic. Canadian Bulletin of Fisheries and Aquatic Sciences 216.

Smith, T.G., and Harwood, L.A. 2001. Observations of neonate ringed seals, Phoca hispida, after early break-up of the sea ice in Prince Albert Sound, Northwest Territories, Canada, spring 1998. Polar Biology 24:215-219.

Sorokin, S.M. 1957. On winter concentrations of Ladoga ringed seal [in Russian]. Scientific and Technical Newsletter of the All-Union Scientific and Research Institute of Lake and River Fisheries 5:80-81.

Tormosov, D.D., and Philatov, I.E. 1973. On Ladoga ringed seal numbers and the rational use of this species. In: Sokolov, V.E., ed. Rare mammal species of SSRU and their conservation [in Russian]. Moscow: Nauka. 103-104.

Trukhanova, I.S., Sagitov, R.A., Verevkin, M.V., Alekseev, V.A., and Andrievskaya, E.M. 2012. Ladoga ringed seal and fisheries: Background for social ecological and economical strain development. Terra Humana 2(23):232-238.

Usachev, V.F., Prokacheva, V.G., and Borodulin, V.V. 1985. Assessment of dynamics of lake ice, snow cover and river overflows by remote sensing methods [in Russian]. Leningrad: Gidrometeoizdat.

Venables, W.N., and Ripley, B.D. 2002. Modern applied statistics with S. 4th ed. New York: Springer. http://www.stats.ox.ac.uk/pub/MASS4.

Verevkin, M.V. 2003. The results of the ringed seal aerial survey in Lake Ladoga [in Russian]. In: Proceedings of the Third International Symposium "Dynamics of Game Animal Populations in Northern Europe, 16-20 June 2002, Sortavala, Karelia, Russia. Petrozavodsk: Institute of Biology, Karelian Research Centre of RAS. 202-204.

Ver Hoef, J.M., and Boveng, P.L. 2007. Quasi-Poisson vs. negative binomial regression: How should we model overdispersed count data? Ecology 88(11):2766-2772.

Wilson, S., Dolgova, E., Trukhanova, I.S., and Goodman, S. 2012. The Caspian seal breeding behavior and pup development. Proceedings of Seventh International Conference "Marine Mammals of the Holarctic," 24-28 September 2012, Suzdal, Russia. 157-160.

Wood, S.N. 2006. Generalized additive models: An introduction with R. Boca Raton, Florida: CRC Press.

Zheglov, V.A., and Chapskiy, K.K. 1971. Experience of aerial survey of ringed and grey seals and their breathing holes in the Gulfs of the Baltic Sea and on the Lake Ladoga [in Russian]. Proceedings of AtlantNIRO (Atlantic Research Institute of Marine Fisheries and Oceanography) 39:323-342.

Zuur, A.E., Ieno, E.N., Walker, N.J., Saveliev, A.A., and Smith, G.M. 2009. Mixed effects models and extensions in ecology with R. New York: Springer.

(1.) Saint Petersburg State University, 199034, 7/9 Universitetskaya nab., St. Petersburg, Russia

(2.) SPbCP0 "Biologists for Nature Conservation," 199034, 8/10, Birzhevaya liniya, St. Petersburg, Russia

(3.) Corresponding author: irina_trukhanova@yahoo.com

(4.) Department of Statistics, University of Washington, Box 354322, Seattle, Washington 98195-4322, USA

(5.) National Marine Mammal Laboratory, Alaska Fisheries Science Center/NOAA, 7600 Sand Point Way NE, Seattle, Washington 98115-6349, USA

[c] The Arctic Institute of North America

IRINA S. TRUKHANOVA, (1), (2), (3) ELIEZER GURARIE (4), (5) and RUSTAM A. SAGITOV (1), (2)
COPYRIGHT 2013 Arctic Institute of North America of the University of Calgary
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2013 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:ARCTIC VOL. 66, NO. 4 (DECEMBER 2013) P. 417-428
Author:Trukhanova, Irina S.,; Gurarie, Eliezer; Sagitov, Rustam A.
Publication:Arctic
Geographic Code:4EXRU
Date:Dec 1, 2013
Words:8674
Previous Article:Residency Times and Patterns of Movement of Postbreeding Dunlin on a Subarctic Staging Area in Alaska.
Next Article:First records of the arctic moth gynaephora groenlandica wocke south of the arctic circle: a new alpine subspecies.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |