Snowy Owls in central North America have regular migration and high philopatry to wintering sites though not always to home ranges

,


INTRODUCTION
Regular migration, a seasonal, repeatable, and predictable movement, occurs in many bird species and is associated with strong connectivity between established breeding and wintering sites (Newton 2010).Long-distance migrants often show lifelong philopatry to a wintering site or small territory where they are resident after fall migration (Robertson and Cooke 1999, Remsen 2001, Holmes and Sherry 1992) but several species move unpredictably, or "nomadically" during the winter season without settling (Teitelbaum and Mueller 2019).Nomadism in the nonbreeding season appears to be driven by low predictability of food sources in time or space (Andersson 1980, Newton 2006, Lindberg et al. 2007).For example, species like Emberizidae finches, that depend on conifer seeds or raptors that rely on mobile prey, are often characterized by extensive movements in the non-breeding season that result in low philopatry to wintering areas (Förschler et al. 2006, Benkman and Young 2020, Dawson 2020).
Even within species characterized by long-distance movements in winter, there may be individual variation in the pattern of movement and settlement that is little understood.A few recent studies have focused on how factors such as age, sex, body condition, or environmental conditions may affect winter residency within or between populations.Oppel et al. (2009) found that intraspecific variation in the duration of winter residency for King Eiders (Somateria spectabilis) was influenced more by traits of the individual bird than by the coverage of sea ice in its environment.In wintering Eurasian Siskins (Carduelis spinus), resident individuals which settled at rich food patches were socially dominant and in better body condition compared to transients with nomadic behavior (Senar et al. 1992).
Here, our goal was to study the degree of philopatry to wintering range in the Snowy Owl (Bubo scandiacus), a top predator which breeds in the circumpolar Arctic and which is designated as vulnerable by the IUCN (IUCN 2020).Snowy Owls are well known for long-distance movements during their annual cycle (Potapov and Sale 2012, Brown et al. 2021).In summer, they specialize on lemming prey and travel widely in the high Arctic to search for hotspots of these rodents which vary cyclically on the tundra landscape (Holt et al. 2015).As a result, Snowy Owls show very low breeding philopatry (Fuller et al. 2003, Therrien et al. 2014).In the eastern Arctic of North America, the average breeding dispersal distance for female Snowy Owls was 710 km which was farther than between their wintering locations (Robillard et al. 2018).Because effective conservation strategies for migratory species should incorporate information about the regularity of habitat use during both the summer and winter periods (Norris et al. 2004, Webster andMarra 2005), we were interested in the degree to which individual Snowy Owls returned to wintering sites on the Canadian prairies.

Many aspects of what directs large scale movements of Snowy
Owls in winter remain a mystery, but several studies using satellite and GPS transmitters are starting to fill in the picture.Most (74%) of breeding female Snowy Owls tagged on Bylot Island, Nunavut, in the eastern Canadian Arctic remained in the Arctic during winter, moving along the marine coastlines and around polynyas in the sea ice while preying mainly on seabirds.The average distance between wintering areas of these 21 female owls in the east was 389 km (Robillard et al. 2018).In contrast, eight female Snowy Owls that were tagged in the far western Arctic of the Yukon (Doyle et al. 2017) and Alaska (Fuller et al. 2003), moved inland during winter to settle in Arctic terrestrial montane boreal biomes.In Fennoscandia, Snowy Owls tagged while breeding in northern Norway remained in the Arctic during winter, often migrating eastwards to Russia, while foraging both out on the sea ice but also moving inland to feed on Willow Ptarmigan (Lagopus lagopus; Øien et al. 2018).Hence, different populations of Snowy Owls across the circumpolar Arctic migrate in different directions, traverse variable distances, and overwinter in different habitat types with different types of prey.
The highest density of Snowy Owls overwintering regularly south of the Arctic treeline is on the prairie grasslands of central North America (Kerlinger et al. 1985), but aside from a few band recoveries, information about winter philopatry to this region is lacking.Here, we assessed philopatry at three spatial scales.At the largest (coarsest) continental scale, we predicted that Snowy Owls wintering on the prairies of central North America would be more philopatric than owls wintering along coastlines in eastern North America.This is because the main prey on the prairies, rodents such as deer mice (Peromyscus maniculatus) or voles in the genus Microtus, characteristic of cropland and grazed pastures (Boxall and Lein 1982), are relatively stable compared to lemmings in the Arctic.At temperate latitudes, most rodents do not show cyclic "boom-and-bust" cycles as in the Arctic (Hansson and Henttonen 1985) and deer mice rarely cycle, although local densities may vary annually (Galindo and Krebs 1987, Sullivan et al. 2004, Chang and Wiebe 2018a).In contrast, owls wintering along eastern coastlines often subsist on waterfowl or large seabirds; these are patchy prey which are mobile and sporadic in occurrence according to the location of polynyas in the sea ice (Gilchrist andRobertson 2000, Robertson andGilchrist 2003).
At a medium (regional) spatial scale, within the central prairie region, individuals vary in the degree to which they settle on a stable home range versus move nomadically over hundreds of kilometers during the winter period (Chang andWiebe 2018b, McCabe et al. 2021).Therefore, we were interested in whether the traveling path of a Snowy Owl during the overwintering winter period overlapped from one year to the next.Finally, at the smallest spatial scale (local), we were interested in whether areas that the owls settled in for a period of time, i.e., home ranges, overlapped between consecutive winters.At all scales, we were interested in behavioral differences between the sexes because previous studies have either followed only females or have not analyzed sexual differences in philopatry.Snowy owls have reversed sexual dimorphism, with males weighing 25-30% less than females on the Canadian prairies (Kerlinger andLein 1988, Chang andWiebe 2016).Because the larger females are socially dominant to males (Boxall and Lein 1982), we predicted that females would be more faithful to familiar wintering ranges than males.We have addressed factors affecting home range size elsewhere (Chang and Wiebe 2018b), so we focus here on the extent of spatial overlap of home ranges and of movement tracks of owls between successive winters.

Study site and field methods
On the prairies of south-central Saskatchewan, Snowy Owls begin to arrive from their breeding grounds in the Arctic around mid-November and many start to settle on winter home ranges in December.Most migrate north by early April (Brown et al. 2021), depending somewhat on the rate of snowmelt (Curk et al. 2020).Although densities of owls vary somewhat from year to year, the prairie population is not considered "irruptive" because relatively high numbers are present every winter (Kerlinger et al. 1985).The landscape is flat, or gently rolling, with less than 20% of the original short-and mixed-grass native prairie remaining (Samson and Knopf 1994).Today, agricultural lands predominate and include pasture (rangeland), as well as cropland growing mainly canola, pulses, and grains.Agricultural land is gridded in square miles forming mosaic-like rectangular blocks of different species of crop cover usually between 0.64-2.56km².Winters are cold in south-central Saskatchewan, with frequent sustained temperatures below -20 °C and some days as low as -40 °C.Snow depth is variable but may persist from late October to early April.The main prey of Snowy Owls wintering in the prairies is small rodents (76-91% of individual items, Boxall and Lein 1982); 90-99% of items (Detienne et al. 2008) but alternate prey such as Grey Partridge (Perdix perdix) and Rock Dove (Columba livia) are also hunted.
To monitor the movements of owls, we trapped adults (after second year [ASY]) starting in December, using bal-chatris or a remotely-triggered bow nets, in central Saskatchewan between 2015-2017 and in 2021.Owls were trapped roughly as far north as Saskatoon (52°7' N, 106° 40' W) and as far south as Kyle (50°4 9' N, 108° 02' W).We deployed GPS/GSM transmitters (Ecotone Telemetry, model Saker H, 25 g) which recorded locations that were accurate to 20 m every 6 hours.The transmitters were attached using a harness of Teflon ribbon (0.25 inch tubular tape, Bally Ribbon Mills Inc., USA) so the units weighed < 3% of the owl's body mass.Survival analyses indicate no increased mortality by the use of such tags (Heggøy et al. 2017) or reduction in reproductive success (Therrien et al. 2012).Because the transmitters relied on cell phone towers, we could not determine the status (dead vs. alive) of any owl that remained in the high Arctic all winter.Of the 27 owls originally tagged, 16 returned to the prairies in at least one subsequent year, allowing us to calculate philopatry.Some owls returned for more than one successive year: two males and three females with three winters; one female with four winters; two males with five; and three males and one female with six winters.Thus, we had 56 owl-winters based on nine female and seven male owls.

Measures of philopatry
Because the timing of migration varied among individuals and years, we defined the winter period for each owl-winter by examining the individual's tracking data for the full annual cycle.An owl was said to have arrived on the wintering grounds the first time its flight path from the Arctic changed from a unidirectional southward trajectory and turned at a ≥ 90° angle within the prairie biome.Similarly, its winter period ended when the flight path began to move in a unidirectional north trajectory out of the prairie biome.
During this winter period, we considered philopatry at three spatial scales: continental, regional, and local.At the continental scale, we considered philopatry during winter to the central prairie biome.At the regional scale, we characterized residency by estimating each owl's occurrence distribution (OD; Fleming et al. 2016) during the entire winter; ODs quantify where an animal may have traveled during the observation period given the data.Thus, an owl's OD included the area of any home range(s) plus the area of nomadic flight paths between them over the entire winter period.At the local scale, previous work showed that Snowy Owls in Saskatchewan often alternated between a period of residency on a local home range and a period of linear, nomadic movement before settling on a different local home range (Chang and Wiebe 2018b).Here, we defined periods of residency on local home ranges as when the owl remained in a restricted area, moving daily back and forth across a central point on the landscape for a time span of at least one week.This gave a minimum cluster of 28 GPS fixes (7 days of 4 fixes per day) and usually more (Appendix Table 1) to model home ranges.
Of the 16 owls that returned in at least one subsequent year and for which we could determine philopatry at the continental scale, 12 owls' transmitters provided data that were suitable for reliable estimation of home ranges or ODs at the local scale.

Statistical analysis
Traditional methods for calculating home ranges, maximum convex polygons, and kernel density estimates assume independent fixes but most GPS-tracking datasets are autocorrelated so such modeling techniques result in area estimates that are negatively biased (Noonan et al. 2019).Hence, we estimated home ranges by computing autocorrelated kernel density estimates (AKDE) which directly model the autocorrelation of the dataset, using the continuous-time movement modeling (ctmm) version 1.1.1package (Fleming and Calabrese 2017) for Program R version 4.2.2 (R Core Team 2022).In each year it was monitored, an owl used ≥ 1 discrete home range (s) for a period of at least one week; for each of these we computed a suite of movement models and used the Akaike Information Criterion (AICc) to select the best-supported model, optimizing the smoothing bandwidth of the kernel density estimate based on the selected model (Fleming and Calabrese 2017).Our suite of movement models included the IID process, which is the model assumed by conventional range estimators in which both positions and velocities are uncorrelated; the OU process which models correlated positions but uncorrelated velocities; the OUF process with both correlated positions and velocities (Calabrese et al. 2021); and the OUf process which is a special case of OUF where the two autocorrelation timescales cannot be distinguished.We fit isotropic and anisotropic versions of these models, which correspond to circular and elliptical covariances respectively.Using the best-supported model for each home range, we created AKDEs and calculated their 95% contours (Appendix Table 1).For each owl, we estimated the percent overlap between its joint home ranges (consisting of ≥= 1 discrete home range) in year x with those in year x + 1.We also estimated the proximity of home range(s) in year x to any range(s) in year x + 1 by calculating the distance between centroids of the ranges.If the owl used more than one home range in a year such that there were multiple distance estimates to range(s) in the subsequent year, we averaged the distances.
For the ODs, we used a 99% CI to obtain an area (a probability distribution) that encompassed all the movement tracks of an owl during its overwintering period.These distributions used a Dynamic Brownian Bridge Movement model (Kranstauber et al. 2012) which provide a confidence region of area of usage based on the actual movement pathways.Such models provide more accurate estimates of space use than conventional kernel or minimum convex polygons which can severely overestimate the area of use of an animal with a complex movement track (Silva et al. 2020).We calculated percent overlap between these ODs between year x and year x + 1 for each owl-year using the coefficient of areal correspondence (Minnick 1964), by dividing the area of overlap of the two ODs (numerator) by their total summed areas (denominator; code in Appendix 2).
To quantify the extent that owls moved during the winter period, we calculated two variables.As a measure of the geographic space an owl covered during its winter travels we calculated the maximum linear distance ("span") between two GPS fixes.Because the tracks of owls sometimes looped back on themselves during the winter period, we also calculated the total distance flown by the owl along its flight path which included any short movements while on home ranges and any flights between them.To test for an effect of sex on these two variables and on distances between the centroids of successive home ranges between years and the amount of overlap of ODs between years, we entered sex as a fixed factor in a Linear Mixed Model (LMM) with Owl ID as a random factor to account for individuals that provided more than one year of data.The frequency of each sex with overlapping home ranges vs. non-overlapping ranges was assessed with Chisquare test.LMMs and Chi-square tests were run in SPSS v. 28 with alpha set at 0.05.Means are reported with standard errors.

RESULTS
At a largest (continental) scale, all 16 owls (nine females and seven males) that were tagged in central Saskatchewan and which returned to winter south of the Arctic tree line (where the transmitters were detectable by cell phone towers) were philopatric to the central prairie region.Both males and females remained north of the United States border and did not drift east or west out of the provinces of Alberta or Saskatchewan (Fig. 1).A visual inspection of migration tracks for individual owls indicated greater philopatry to a wintering region than to a breeding region in the high Arctic (Fig. 2).The duration of wintering periods of owls ranged from 79 to 163 days (128 + 3.5, n = 40 owl-winters) and spanned the period from 1 November to 18 April.The maximum linear geographic span between fixes during these winter periods averaged 156.1 + 20.3 km for the five females and 201.2 + 17.1 km for the seven males; the longest span over the landscape was that of a male which roamed across 467 km during the winter of 2019-2020.There was extensive overlap between the sexes in the 95% confidence intervals around the maximum spans (LMM: F 1, 6.45 = 1.43,P = 0.27).The total flight paths of owls during winter averaged 1053 + 81.7 km for males and 805.0 + 79.1 km for females with the longest flight path recorded as 1931 km, for a male.The 95% confidence intervals for mean lengths of flight paths overlapped between the sexes (LMM: F 1, 10.1 = 2.40, P = 0.15).Most (72.2%) of ODs overlapped somewhat between consecutive owl-winters (n = 36) but the proportion of overlap was rather small (Fig. 3).For males it averaged 6.34% + 1.2 and for females 4.25% + 2.0 and the amount of overlap was considerable for the 95% confidence intervals between the sexes (LMM: F 1, 8.85 = 0.63, P = 0.44).
Owls had between 1-4 home ranges during a winter period, settling for durations between 7-121 days within these areas (Fig. 4; Appendix 1).Home ranges averaged 162 + 47 km² for males (n = 76 home ranges for seven males) and 94.3 + 36 km² for females (n = 39 home ranges for five females).More males (6 of 7, 86%) than females (1 of 5, 20%) had some overlap of home ranges between successive winters (Chi-square test: X²= 5.18, P = 0.023).When overlap occurred, the proportion of the ranges that overlapped was 21% for the female and was also 21% + 16 (range 5-49%) for the six males.However, with only one female showing overlap, it was impossible to test for a difference between the sexes.At this finest spatial scale, the average distance between centroids of home ranges in one year and those in the following year was shorter for males (88 + 14 km) than for females (149 + 20 km; LMM: F 1, 8.4 = 5.77, P = 0.041).Thus, successive home ranges of males were closer to each other and were more likely to overlap than were those of females.

DISCUSSION
At the large continental scale, wintering Snowy Owls in central North America were highly philopatric to the central prairie region whereas at the smallest spatial scale, home ranges of an individual owl did not always overlap between successive years.Because all owls in our study were initially tagged as adults, we cannot rule out the possibility that yearlings are the age class which shows large-scale, cross-continental dispersal movements between their first and second winters.However, once an adult Snowy Owl winters on the prairies, it appears to return there in a regular, annual migration to the same biome.There were no annual gaps in our detection of individuals across time, suggesting that an owl which is migratory maintains that overwintering strategy through its lifetime; however, a larger sample is needed to determine whether individuals in central North America ever alternate between remaining in the high Arctic all year versus migrating southward.Philopatry to an overwintering latitude would be consistent with Therrien et al. (2011) and Robillard et al. (2018) who documented that only two of 21 female Snowy Owls tagged in the eastern Arctic switched to wintering on the prairies in the subsequent year.In eastern North America, female Snowy Owls were also consistent in the general habitat type they used between winters, most returning to either inland (terrestrial) biomes versus to marine coastlines (Robillard et al. 2018).At this large geographical scale, perhaps it is most profitable for an individual owl to return to the same overwintering biome in order to take advantage of learned hunting skills needed to specialize on a certain prey type (e.g., small mammals versus seabirds).Although Snowy Owls overwintering on the prairies in the current study were quite philopatric to a winter region, they showed much greater breeding dispersal in the Arctic, consistent with other populations breeding to the east, in Nunavut (Therrien et al. 2014).The owls in Fennoscandia which move east and west within the Arctic during winter (Øien et al. 2018) probably also experience a similar type of biome each year.
After migrating to the central prairies, most male and female Snowy Owls alternated periods of nomadic wandering with periods of residency on one or more home ranges.We could not quantify prey abundance over this vast spatial scale, but it is reasonable to assume that such winter movements are linked to searching for prey hotspots on the landscape and that owls leave a home range when prey are depleted below a certain threshold.A previous study found that the distance moved between successive home ranges within a year did not differ between the sexes (Chang and Wiebe 2018b).Similarly, the results of the current study indicate that males and females covered a similarsized spatial area and flew similar total distances while searching for prey during their winter periods.Although the movement tracks of an individual (the OD) usually crossed over a place where it had traveled the previous winter, the total movement path did not overlap closely between years.Thus, Snowy Owls did not follow a memorized or systematic traveling route during nomadic searches for prey on the prairie landscape.We are unaware of another bird species that alternates periods of nomadic wandering with residency on a defended home range during the non-breeding period.
Home ranges where owls settled often did not overlap between one year and the next.However, contrary to our prediction, males were more philopatric to previous areas of settlement than females.Namely, a greater proportion of males than females settled on a range that overlapped (at least partially) with one that they had used the previous winter and the average distance between the centroids of annual home ranges was shorter for males compared to females.Although small rodents on the prairies in winter do not show predicable cycles, population densities at a certain location can vary by a factor of 4 or 5 between successive years (Mihok et al. 1985, Chang andWiebe 2018a).Thus, prey hotspots for owls on the prairies may shift on the landscape over distances of 100 km or more.Another reason why a home range at a certain location may not retain its quality for owls is that farmers on the prairie rotate most crop fields annually.For example, a particular area may lose its attractiveness for owls if much of the preferred hunting habitat, legume fields, are converted to canola fields which tend to be avoided by Snowy Owls (Chang and Wiebe 2018a).
Settlement patterns at the fine spatial scale of home ranges are consistent with the hypothesis that females, which are socially dominant to males (Kerlinger and Lein 1986), are more likely to monopolize prey hotspots that shift annually and which may be some distance from former home ranges.If the richest areas of prey become monopolized by females, males may be likely to "make the best of a bad situation" by returning to a former home range where their familiarity with the landscape structure would aid in hunting success (Lundberg 1979, Greenwood and Harvey 1982, Merkle et al. 2014) and perhaps compensate for somewhat lower prey densities.The fat reserves and body condition of overwintering males were less than that of females (Chang and Wiebe 2016) but both sexes usually maintained a winter body mass above the starvation threshold so males seem to be able to meet energy requirements even if they are excluded from areas of peak prey.
In other regions of North America, the sample size of Snowy Owls tracked over multiple years is small but none of the four females tagged in Alaska (Fuller et al. 2003) or the four females in the Yukon (Doyle et al. 2017) migrated in a regular north-south direction to settle south of the subarctic boreal region.These owls in western North America did not use the same home range in two successive winters, although some briefly visited locations from the previous winter.Based on two females, the average distance separating wintering home ranges in subsequent years was 873 km (Doyle et al. 2017), considerably longer than the 149 km for females in our study.In eastern North America, the average distance between centroids of wintering home ranges of 21 females in successive years was 389 km and the average overlap between successive home ranges was 29% (Robillard et al. 2018).However, the different methodology used to calculate home range size for the eastern owls resulted in estimates an order of magnitude larger than in our study, so quantitative comparisons of home range overlap between populations are difficult based on existing data.However, Snowy Owls on the central prairies of North America were more philopatric to a home range on a fine scale than were owls from either the west or the eastern parts of the continent.
The strong connectivity of a breeding population of Snowy Owls in the central Arctic to a particular wintering area on the central prairies may make it vulnerable to changes in land use, or to changes in climactic regimes that reduce access to small mammals over a large part of the central prairie region.For example, shifts in winter climate that affect the depth or density of snow could alter the density of small mammals wintering beneath protective snow cover (Bilodeau et al. 2013) or the ability of owls to hunt them (Chamberlain 1980).Changes to grassland vegetation as a result of fire, haying, or cropping practices also affect the density of mice and voles (Lafond et al. 2020).On a fine spatial scale, Snowy Owls, especially females, may be able to mitigate annual fluctuations in prey by tracking, and settling, at prey hotspots within the prairie biome although the ability to assess prey numbers may be negatively affected by deep snow cover (Therrien et al. 2015).Long-term information on the abundance and predictability of small rodents of the central prairie region, and the extent to which they are affected by changing winter climate, will be needed to model Snowy Owl numbers in central North America in relation to climate change.

Fig. 1 .
Fig. 1.Annual movements of adult Snowy Owls (Bubo scandiacus) to and from their wintering areas on the Canadian prairies showing linkage to the prairie biome within the population at the continental scale.To avoid clutter, only the first year of tracking an individual is depicted here, with different colors showing one year of movement for each of 12 owls.

Fig. 2 .
Fig. 2. Example of migratory tracks of a female owl (left panel) and of a male (right panel) with six successive years of data showing strong winter philopatry at the continental scale to the central prairie region.Breeding philopatry in the high Arctic is comparably much weaker.

Fig. 3 .
Fig. 3. Example of Occurrence Distributions of a male owl tracked during six consecutive winters in Saskatchewan.

Fig. 4 .
Fig. 4. Example of home range polygons (95% CI) of a male owl tracked during six consecutive winters in Saskatchewan.