Breeding Snowy Owls Are Obligate Lemming Predators in Utqiaġvik, Alaska: Results from 30 Years of Study

Abstract

For 30 years (1992–2021), we collected pellets and pellet fragments and recorded prey cached in Snowy Owl (Bubo scandiacus) nests during the breeding season in Utqiaġvik, Alaska. About 14,000 pellets from an estimated 700 Snowy Owls yielded 43,689 prey items, while caches in 284 nests yielded 3334 prey items. The owls ate thirty-seven species of vertebrates: one species of fish, five species of mammals, and thirty-one species of birds. Based on the pellet analysis, lemmings represented 99.0% of the total prey, with brown lemmings (Lemmus trimucronatus) representing 94.6%, collared lemmings (Dicrostonyx groenlandicus) representing 3.1%, and unidentified lemmings representing 1.3%. All other species were <1%. Based on the prey cached in nests, lemmings represented about 90.0% (89.9%) of the total prey (n = 3334), with brown lemmings representing 88.0% (87.9%), collared lemmings representing 1.9%, and unidentified lemmings representing <1%. Birds represented only 10.0% of the prey cached in nests, although many species were eaten. Food niche breadth (FNB) and dietary evenness (DIEV) scores from pellets were narrow for the prey identified within a group or species. FNB and DIEV scores from the prey cached in nests were also narrow for the prey identified within a group or species. There was almost complete dietary overlap when comparing the prey from pellets with the prey from caches. Biomass estimates from brown lemmings (178 kg) cached in nests were 59 times more than those from collared lemmings (3 kg). Biomass estimates for large birds were misleading, as the owls mainly ate the breast, humerus, and femur muscles. Our study supports a general consensus that Snowy Owls are obligate lemming specialists during the breeding season in Utqiaġvik. In fact, they depend almost entirely on one species of lemming—the brown lemming. Consequently, anthropogenic or natural factors that impact lemming populations and distributions will directly affect Snowy Owl populations.

1. Introduction

The Snowy Owl (Bubo scandiacus) is one of the largest and heaviest owls in the world. In its circumpolar Arctic breeding areas, Snowy Owls nest on the ground and are primarily dependent on the population of lemmings for successful reproduction (Holt et al. 1999 [1], Konig and Weick 2008 [2], Holt et al. 2025 [3]). The Snowy Owl is monotypic and has the most northern breeding and non-breeding distribution of any owl species in the world. Results from one genetic study using mtDNA (mitochondrial DNA) suggested a single global panmictic population (Marthinsen et al. 2008 [4]). A later study using SNPs (single-nucleotide polymorphisms) buttressed the mtDNA results (Gousy-LeBlanc et al. 2023 [5]).

Several studies have reported the breeding ecology of this species in Canada, Great Britain, Greenland, Russia, and the United States (see Holt et al. 2025 [3]). With the exception of Menyushina’s studies (1994 a, [6], b [7], c [8], 1997 [9], 2007 [10]) from Wrangel Island, Russia, and this study from Utqiaġvik (formerly Barrow), Alaska, most breeding season studies include one to four seasons (see the review in Holt et al. 2025 [3]).

Prior to the Arctic winter, many Snowy Owls migrate to southern latitudes, often being considered nomadic (Cramp 1985 [11], Parmelee 1992 [12]) or irruptive migrants (Holt and Zetterberg 2008 [13], Holt et al. 2025 [3]). Yet, some Snowy Owls winter within Arctic latitudes in both terrestrial and marine environments, and rarely on or near their breeding grounds (Holt et al. 2025 [3] for a historical review). It remains unknown if Snowy Owls wintering near their breeding grounds feed upon lemmings or other terrestrial animals. The satellite tracking of Snowy Owls since 1999 has clearly shown that some Snowy Owls spend the winter in Arctic marine environments (National Geographic 2002 [14], Fuller et al. 2003 [15], Therrien et al. 2011 [16], 2015 [17], Robillard et al. 2016 [18], Øien et al. 2018 [19]; also see https://www.projectsnowstorm.org/, accessed 10 March 2025). In fact, during the Arctic winter, Snowy Owls have been observed or satellite tracked on the edges of polynyas, leads, open areas of ice, and winter pack ice. With few observations made, our ideas of their main diet during the winter in these open-water Arctic habitats remain speculative, but the options are narrow, and it seems likely to be a combination of waterfowl and seabirds, and perhaps scavenging. For Snowy Owls that migrate to southern latitudes near human populations in Canada and the United States, their winter diet is well-known, and the owls’ have shown an ability to diversify their diet and eat whatever is most available (Holt and Zetterberg 2008 [13], Holt et al. 2025 [3], N. Smith, personal communication). This is in contrast to other open-country species of owls such as long-eared and short-eared owls, which are obligate small mammal predators throughout the year (Holt 1993 [20], Holt 1997 [21]).

Surprisingly, for a species believed to be dependent on lemmings for breeding, there are very few food habit studies in North America. In reviews of the North American Snowy Owl breeding season diet, previous studies provided reliable, but simple descriptive or simple quantitative assessments. Descriptive examples include the following. Fisher (1893 [22]) suggested “lemmings and arvicoline mice” comprise almost the entire summer diet. Bent (1938 [23]) suggested that Snowy Owls eat lemmings and small mammals during the Arctic summer. Simple quantitative examples include the following. On Baffin Island, Canada, of the 964 prey items from “a great number” of pellets, 946 (98.1%) were brown lemmings (Lemmus trimucronatus) and 18 (1.9%) were collared lemmings (Dicrostonyx groenlandicus) (Watson 1956 [24]). On South Hampton Island, Canada, of the 358 prey items from 149 pellets at 11 Snowy Owl nests, collared lemmings represented 96.0% (n = 344) (Parker 1974 [25]). On Bylot Island, Canada, over four breeding seasons, Therrien et al. (2015 [17]) collected 875 pellets from 26 nests. Of the 2263 prey, 1111 (49.1%) were brown lemmings and 1074 (47.5%) were collared lemmings. See Holt et al. (2025 [3]) for a complete review.

Although several studies have reported the breeding season diet from Europe and Russia (summarized in Mikkola 1983 [26], Cramp 1985 [11], Potapov and Sale 2012 [27]), the manner in which these data are reported complicates some species identification. Furthermore, see Holt et al.’s (2025 [3]) and Stenkewitz and Neilson’s (2019 [28]) studies for unusual or exceptional diets from islands without lemmings.

Herein, we provide a more in-depth descriptive assessment of Snowy Owl breeding season feeding ecology from our study in Utqiaġvik (formerly Barrow), Alaska, USA. We emphasize the importance of lemmings to the Snowy Owl breeding output at Utqiaġvik, but also the importance of lemmings to Snowy Owl survival throughout their Arctic breeding range. Utqiaġvik has a history of detailed lemming research conducted from 1955 to 1974 (see Pitelka 1973 [29], Pitelka and Batzli 1993 [30], Pitelka and Batzli 2007 [31], Pitelka and Batzli 2018 [32]). These authors suggest that lemmings at Utqiaġvik “cycle”, with population peaks occur every 2–6 years. Whether these cycles are predictable mathematical events or fit the original definition of the word found in the Merriam-Webster Online Dictionary (see www.merriam-webster.com, accessed 10 March 2025) is debatable. In 1990, DWH initiated a study on Snowy Owls at Utqiaġvik, Alaska, now in its 35th year. In 1992, and continuing today, DWH began a lemming snap-trap study in conjunction with the Snowy Owl breeding study to assess the relationship between lemmings and owls at Utqiaġvik (DWH, unpubl. data, this study 1992–2024).

2. Study Area

The North Slope of Alaska covers approximately 200,000 km² and drains water to the Arctic Ocean. The three main physiographic regions are the Brooks Mountain Range, Arctic Foothills, and Arctic Coastal Plain. This entire area, also known as the Beaufort Coastal Plain Eco-region, stretches from western Canada across Alaska and borders the Arctic Ocean, north of the Brooks Mountain Range (Nowacki et al. 2001 [33]).

Our study is located on the Arctic Coastal Plain, near the village of Utqiaġvik (71°18′ N; 156°40′ W). Utqiaġvik is the most northerly settlement in Alaska and bordered by the Chukchi Sea to the west and Beaufort Sea to the east (Figure 1). See Brown et al. (1980 [34]), Nowacki et al. (2001 [33]), Johnson et al. (2011 [35]), and Villarreal et al. (2012 [36]) for a review of the local habitat and landscape. Our study area encompasses approximately 214 km². Topographic relief is low, ranging from sea level to about 10 m. Overall, ice-wedge-created polygonal ground, shallow lakes, and underlying permafrost characterize the landscape.

3. Materials and Methods

From 1992 to 2021, and continuing, we have studied the breeding ecology and predator–prey relationship between Snowy Owls, Nearctic brown lemmings (Lemmus trimucronatus), and Nearctic collared lemmings (Dicrostonyx groenlandicus formerly rubricatus) at Utqiaġvik (Fuller et al. 2003 [15], Holt et al. 2008 [38], Seidensticker et al. 2011 [39], Holt 2022 [40]). Hereafter, we refer to these two species of lemmings as the brown lemming and collared lemming. We collected pellets from 1992 to 2016 and recorded prey cached at nests between 1992 and 2021. Our research season lasts approximately 3 months, from June to August, and continues today. This summation builds upon those 30 years of study by the same primary researcher (DWH).

In assessing Snowy Owl breeding season feeding ecology from Utqiaġvik, our objectives were to (1) describe overall food habits, (2) present a detailed examination of their feeding niche and relationship to lemmings, and (3) discuss the importance of lemmings to the Snowy Owl life history at Utqiaġvik.

3.1. Finding Nests

Snowy Owls nest on mounds or other promontories that become snow-free earlier than the rest of the tundra (Menyushina 1997 [9], Holt et al. 2008 [38]). To locate nests, we hiked and scanned the tundra for adult (white) male Snowy Owls. Adult males usually roost and hunt near nests, and their highly reflective bright white plumage can be seen for over 1 km without binoculars. Once males were observed, we scanned the area for cryptically colored and horizontally postured females on nests. Only females incubate and brood.

3.2. Pellet Collections

We collected pellets for 25 years (1992–2016) from 284 nests. For each nest, we collected pellets wherever we could find them within about 500 m of the nest. This represented samples from male and female owls at each nest. We also collected as many pellets as possible from individual non-breeding Snowy Owls throughout the study area. Most of these pellets were collected at roosting mounds or within the area of these mounds. So, at the end of each season, we knew approximately how many Snowy Owls resided in our study area. After breeding, however, some Snowy Owls linger or winter near Utqiaġvik. Thus, the following season, we separated out and discarded older dried-out winter pellets.

Some Snowy Owls return to our study area over their lifetime. We buttress this statement based upon a few observations of our satellite-tracked owls (Fuller et al. 2003 [15]) and owls banded on the left leg or recovered. We make an effort each season to identify banded individuals, especially breeders. However, given their nomadic tendencies, and the lack of banded owls we observe, it is unlikely most individual owls are the same from year to year. Furthermore, give our enormous samples, and the fact that lemmings dominate the diet, even if every individual was not an independent sample, it is very unlikely to have affected our results, as we are confident our pellet samples represent hundreds of individual owls.

We shipped most pellets back to our lab in Montana. In 2004 and 2014, pellets were lost in the mail. Thus, there are no data for those years. Pellets were soaked in a solution of Clorox bleach and water for about 30 min to dissolve and separate fur from bone. Fur was skimmed off and discarded, and bones were placed on a screen and flushed clean with water.

3.3. Prey Identification from Pellets

To identify mammalian prey from species for this region, we followed the key to skull and dental characteristics outlined in Bee and Hall (1956 [41]). For mammals, only skulls were used to tally prey. Of 100 breeding season pellets, jaws and skulls yielded the best numbers for counting the prey per pellet: 2.7 for skulls and 3 for jaws (see Holt et al. 2025 [3], who suggested using skulls to quantify prey, because skulls can be identified as species). For birds, we used skulls, legs, and feathers to identify and quantify groups or species. We used bird field guides, and museum specimens if needed.

3.4. Cached Prey

We recorded prey species cached at nests for 18 breeding seasons between 1992 and 2021. Whole bird carcasses were easy to identify as species, and for some, we identified the sex. If the prey was lemmings, we recorded the species, body mass, and sex if possible. We cut one leg off each lemming during nest visits to reduce double-counting.

3.5. Food Niche

To assess Snowy Owl feeding ecology at Utqiaġvik, we first identified prey as species from pellets and caches. Next, prey not identified as species were identified using a genus or group (e.g., ducks, shorebirds). We then reported the frequency and percentage of occurrence in the owls’ diet. In order to analyze proportions, we selected the Chi-square test, with the significance level set at 0.05. To meet the assumptions of the test, we eliminated species, genera, or groups that tallied less than five individuals. We also eliminated species, genera, or groups that represented <1% of the total prey eaten. We deemed those groups as non-significant in the owls’ diet.

To explore Snowy Owl feeding ecology [e.g., food niche breadth (FNB) and dietary evenness (DIEV)], we used information theory equations to examine the structure of the owls’ feeding niche. Also known as diversity indices, we chose the Shannon–Weaver and Simpson equations to examine species richness, and Hill’s equation to examine evenness (see Marti et al. 2007 [42], Holt et al. 2024 [37]). These equations have been modified over time to give more meaningful and interpretable metrics (see Marti et al. 2007 [42], and descriptions below). Ultimately, these equations allow large datasets to be expressed as single values. In turn, values can be compared from different studies, provided the same or similar methods are used. There are strengths and weaknesses to each equation, with some reviewers favoring one equation over another and other reviewers seeing no value in them (see Zar 1999 [43], Marti et al. 2007 [42], Collier and Schertner 2012 [44] for reviews). Therefore, we computed both the Shannon–Weaver and Simpson indexes to make comparisons. See the Information Theory below.

We defined the annual feeding niche as the relationship between the owls and their prey, and we followed Holt 1993 [20] and Holt et al. 2024 [37] in describing FNB and DIEV. Because FNB and DIEV are influenced by levels of resolution, we attempted to identify all prey as species. However, that was not always easy. For example, when dissecting pellets, or recording prey cached at nests, we often encountered skulls, legs, long bones, or feathers. Some could be identified as species; however, others could not, particularly bones from pellets. Thus, we assigned some prey to groups such as unidentified lemmings, ducks, shorebirds, and unidentified birds. For example, 10 brown lemmings, 8 collared lemmings, and 4 Red Phalaropes (Phalaropus fulicaria) were species, whereas 6 unidentified lemmings, 10 ducks, 8 shorebirds, 6 passerines, 4 unidentified birds, and 1 fish were groups. Logically however, at least one species occurred in each group.

Therefore, we calculated FNB and DIEV at two levels of resolution: (1) groups and species, as in the example above; and (2) we elevated some groups to species, such as ducks and shorebirds. For example, a duck or shorebird equaled at least one species, and the total number was retained. We also made other adjustments for the calculations. For example, only two species of lemmings occurred in our study, and when both were identified from a sample, the unidentified lemming group from that sample was removed for calculations. Similarly, when several groups of birds were identified (i.e., ducks, passerines) and an unidentified bird group, we removed the unidentified group from species-level assignments. We acknowledge that some biases may exist in this methodology, but not including the group categories at the species level would reduce the diversity of the prey eaten.

Next, to determine if our decisions to analyze prey at the group and species levels were biased and gave significantly different results, we compared the median value at the group level of resolution with the median value at the species level of resolution for pellets and prey cached at nests for H’, R, and D.

Finally, we compared the median FNB and DIEV breeding season scores of species identified from cached prey at nests (n = 16 years) with groups and species identified from pellets (n = 14 years) for H’, R, and D.

We used a Mann–Whitney non-parametric medians test because samples from cached prey at nests could also show up in pellets, thus violating the assumptions of independence of the parametric test for means. Furthermore, non-parametric tests are appropriate for comparing indices, such as diversity and evenness scores (Fowler and Cohen 1990 [45], Zar 1999 [43]). The alpha level was set at 0.05.

After these adjustments, we defined a broad FNB as a high number of prey species, relatively equally distributed in the owls’ diet (i.e., heterogeneous), and a narrow FNB as a low number of prey species, relatively unequally distributed in the owls’ diet (i.e., homogeneous). Because DIEV scores range from zero to one, we considered a score of one or approximately one a uniform representation of prey proportions in the diet, while scores <0.50 represented a non-uniform or shared distribution of prey in the diet.

3.6. Information Theory

Food niche breadth (FNB) was calculated using the antilog of the Shannon–Weaver index:

H’ = −Σp_i log p_i

where p_i represents the proportion of each species in the sample. The antilog of the Shannon–Weaver diversity index is linearly related to the number of prey categories in the sample. Some statisticians believe that H’ can underestimate diversity unless samples are large (Zar p. 41, 1999 [43]), whereas others believe that H’ can be biased if many species are represented (Collier and Schwertner p. 62, 2012 [44]).

Additionally, we calculated FNB using Simpson’s equation:

D = Σp_i²

where p_i is the proportion of each member of the assemblage being investigated. However, we used the reciprocal (1/D) of Simpson’s index, because it is believed to yield a more meaningful number (Marti et al. 2007 [42]). Some statisticians prefer D, as it is simple to calculate and understand, robust, and meaningful (Collier and Schwertner p. 62, 2012 [44]).

Dietary evenness (DIEV) scores were calculated:

F = (N₂ − 1)/(N₁ − 1)

where N₁ is the antilog of the Shannon–Weaver index (H’) and N₂ is the reciprocal of Simpson’s index, where (1/D). We used Spearman’s rank correlation to examine whether the FNB and DIEV scores were influenced by the sample size.

3.7. Dietary Overlap

Some authors suggest that pellet analysis is biased if other methods of prey identification are omitted (Marti et al. 2007 [42]). For example, pellets yield reliable numbers and species identification for small mammalian prey but may not yield accurate numbers and species identification for large prey, whose remains are often carcasses, feathers, or fur. Indeed, incorporating observations and photographs or recording cached prey at nests may yield different results than pellet analysis alone.

Thus, to determine if pellets and prey cached at nests in this study yielded similar results (i.e., non-biased), we used Pianka’s equation for dietary overlap (O) (see Marti et al. 2007 [42], Holt et al. 2024 [37]) and calculated the overlap (O):

O = Σp_ij p_ik √ √ (Σp_ij², Σ_ik²)

where p_ij and p_ik are the proportions of prey species or other taxa in the diet of Snowy Owl prey from pellets j and cached prey k, respectively. Values range from zero to one, where zero equals no overlap and one equals complete overlap. We then multiplied the value (O) by 100 to report a percent.

3.8. Prey Biomass

We did not determine the body mass of lemmings from pellets, but we calculated the biomass from intact carcasses cached at nests. This may not represent the average body mass of all lemmings roaming the tundra but is a reliable metric of prey biomass brought to the nests. Furthermore, several size and age classes of lemmings were found cached at the nests. For example, small lemmings just dispersing from natal nests, medium-sized sub-adults, and large adult individuals were represented. We then multiplied the mean weight of lemmings cached at nests by the total number.

We did not estimate the biomass for most birds because they were small, usually torn apart, could not be identified to the species level, meaning they were difficult to quantify, and birds only represented <1% and <10% of the total prey from pellets and caches, respectively. Biomass estimates for large birds such as ducks were easier to calculate; however, these estimates were biased. For example, traditionally, one would multiply the number of prey items by the average live body mass, as derived from some authority. However, this is a common miscalculation in feeding ecology studies of birds of prey (see Holt 1993 [20]). Our field observations revealed Snowy Owls only ate specific areas of large avian prey, such as major muscles—the arms, breast, and legs. They most often discarded the head, feet, guts, most flight feathers, and major skeletal elements. Therefore, we estimated the large bird biomass by dissecting and weighing the major muscles of freshly dead, intact carcasses we obtained from hunters.

4. Results

Between 1992 and 2021, we found 284 Snowy Owl nests within our study area. We collected pellets and pellet fragments and recorded prey cached at nests from about 700 Snowy Owls. We did not separate pellets to individual owls, but rather compiled them within a nesting area and the overall study area. By combining species identified from approximately 14,000 pellets and pellet fragments and 3334 prey cached at nests, we recorded 37 species of vertebrates: 1 species of fish, 5 species of mammals, and 31 species of birds (

Table 1. Thirty-seven species of prey eaten by about 700 Snowy Owls between 1992 and 2021. List derived from pellet contents and prey cached at nests. Although there is overlap, 23 species were identified from approximately 14,000 pellets, whereas 32 species were identified from 3334 caches. Lowercase x means prey where at least 1 prey species was represented.

Species	Pellets	Cached
MAMMALS
Nearctic Collared Lemming	x	x
Nearctic Brown Lemming	x	x
Least Weasel	x	x
Short-tailed Weasel	x	x
Arctic Fox	x
BIRDS
Greater White-fronted Goose	x	x
Brant	x	x
Northern Pintail	x	x
King Eider	x	x
Common Eider		x
Spectacled Eider		x
Steller’s Eider		x
Long-tailed Duck	x	x
Ptarmigan spp.	x	x
Red-throated Loon	x	x
American Golden Plover	x	x
Dunlin	x	x
Semipalmated Sandpiper		x
Western Sandpiper		x
Pectoral Sandpiper		x
Long-billed Dowitcher		x
Ruddy Turnstone		x
Red Phalarope	x	x
Glaucous Gull	x	x
Pomarine Jaeger		x
Parasitic Jaeger	x	x
Long-tailed Jaeger		x
Alcid (Aethia spp.)		x
Snowy Owl	x
Short-eared Owl	x	x
Common Raven	x
Common Redpoll		x
Lapland Longspur	x	x
Snow Bunting		x
Savannah Sparrow		x
Raptor spp.	x
FISH
Unidentified Fish	x

).

Pellet analysis from 1992 to 2016 yielded 43,689 prey (

Table 2. Prey (n = 43,689) identified from Snowy Owl pellets collected during the breeding seasons at Utqiaġvik, Alaska between 1992 and 2016. * Pellets lost. The short hyphen (-) means no data.

	BR	CO	UL	DU	SH	GU	LO	PT	RA	PA	UB	FI	Total
1992	3611 (97.3)	68 (1.8)	4 (<1)	4	9	4	0	0	5	6	0	0	3711
1993	8991 (98.2)	118 (1.3)	3 (<1)	21	14	2	2	1	0	4	0	1	9157
1994	2192 (99.3)	8 (<1)	0 -	2	2	2	0	0	0	0	0	0	2206
1995	4082 (96.1)	115 (2.7)	5 (<1)	3	16	3	0	0	0	0	24 (<1)	0	4248
1996	3295 (97.5)	82 (2.4)	2 (~1)	0	0	0	0	0	0	0	0	0	3379
1997	245 (78.5)	57 (18.3)	4 (1.3)	1	2	0	0	0	0	0	3 (<1)	0	312
1998	1041 (94.0)	41 (3.7)	14 (1.3)	0	0	0	0	0	0	0	11 (<1)	0	1107
1999	1605 (89.3)	130 (7.2)	35 (1.9)	3	1	0	0	0	0	0	23 (1.3)	0	1797
2000	2620 (80.1)	370 (11.3)	254 (7.8)	4	0	0	0	0	3	0	19 (<1)	0	3270
2001	244 (89.7)	1 (<1)	25 (9.2)	0	0	0	0	0	0	0	2 (<1)	0	272
2002	345 (95.3)	4 (1.1)	0 -	1	1	0	0	0	0	0	11 (3.0)	0	362
2003	328 (86.1)	38 (10.0)	0 -	1	1	0	0	0	1	1	11 (2.9)	0	381
2004 *	-	-	-	-	-	-	-	-	-	-	-	-	0
2005	451 (85.3)	31 (5.9)	45 (8.5)	1	0	0	0	0	0	0	1 (<1)	0	529
2006	1670 (95.9)	58 (3.3)	1 (<1)	0	0	0	0	0	0	0	12 (<1)	0	1741
2007	2993 (98.7)	23 (<1)	12 (<1)	0	0	0	0	0	0	0	3 (<1)	0	3031
2008	4390 (98.9)	28 (<1)	8 (<1)	0	0	0	0	0	0	0	12 (<1)	0	4438
2009	786 (86.9)	1 (<1)	116 (12.8)	0	0	0	0	0	0	0	2 (<1)	0	905
2010	433 (97.1)	11 (2.5)	0 -	0	1	0	0	0	0	1	0	0	446
2011	247 (74.2)	36 10.8)	40 (12.0)	1	1	0	0	0	0	3	5 (1.5)	0	333
2012	215 (71.4)	44 (14.6)	0 -	0	11	0	0	0	0	22	9 (3.0)	0	301
2013	177 (71.7)	45 (18.2)	0 -	0	6	0	0	0	0	0	19 (17.7)	0	247
2014 *	-	-	-	-	-	-	-	-	-	-	-	-	0
2015	298 (91.7)	23 (7.1)	0 -	2	1	0	0	0	0	0	1 (<1)	0	325
2016	1091 (91.6)	31 (2.6)	3 (<1)	0	1 (<1)	0	0	0	0	0	65 (5.5)	0	1191
Grand Totals	41,350 (94.6)	1363 (3.1)	571 (1.3)	44 (<1)	67 (<1)	11 (<1)	2 (<1)	1 (<1)	9 (<1)	37 (<1)	233 (<1)	1 (<1)	43,689

BR = Brown lemming; CO = Collared lemming; UL = Unidentified lemming; DU = Duck species; SH = Shorebird species; GU = Gull species; LO = Loon species; PT = Ptarmigan species; RA = Raptor species; PA = Passerine species; UB = Unknown bird species; FI = Fish species.

). Overall, lemmings represented 99.0% of the Snowy Owl diet over these 25 years. Of all lemmings, brown lemmings represented 94.6% of the total prey, collared lemmings 3.1%, and unidentified lemmings 1.3%. All other species or groups made up <1% of the total. The unidentified bird category was most likely made up of shorebirds and passerines (n = 233). Snowy Owls were present in all 30 years of our study; however, they only nested in 18 years between 1992 and 2021. Prey species cached at nests over these 18 breeding seasons yielded 3334 prey (

Table 3. Prey (n = 3334) cached at Snowy Owl nests during 18 breeding seasons at Utqiaġvik, Alaska, between 1992 and 2021.

	BR	CO	UL	DU	SH	GU	LO	PT	RA	AL	PA	UB	Total
1993	521 (98.9)	0	1 (<1)	0	0	1 (<1)	0	0	0	0	4 (<1)	0	=527
1995	585 (92.8)	20 (3.2)	1 (<1)	6 (<1)	5 (<)	6 (<1)	0	0	0	0	6 (<1)	1 (<1)	=630
1996	218 (80.4)	0	0	10 (3.7)	19 (7.0)	10 (3.7)	1 (<1)	0	0	0	12 (4.4)	3 (1.1)	=273
1999	78 (61.4)	3 (2.4)	0	16 (12.6)	12 (9.4)	7 (5.5)	0	0	0	0	10 (7.9)	1 (<1)	=127
2000	353 (84.8)	22 (5.3)	1 (<1)	24 (5.8)	6 (1.4)	4 (<1)	0	0	3 (<1)	0	3 (<1)	0	=416
2002	32 (74.4)	0	0	2 (4.6)	2 (4.6)	0	0	1 (2.3)	0	0	6 (13.9)	0	=43
2003	17 (63.0)	1 (3.7)	0	0	3 (11.1)	0	0	0	0	0	4 (14.8)	2 (7.4)	=27
2005	43 (84.3)	1 (2.0)	0	1 (2.0)	4 (7.8)	0	0	0	0	0	2 (3.9)	0	=51
2006	406 (94.2)	3 (<1)	0	4 (<1)	6 (1.4)	2 (<1)	0	0	0	0	3 (<1)	7 (1.6)	=431
2008	439 (98.6)	0	1 (<1)	3 (<1)	0	1 (<1)	0	0	0	0	1 (<1)	0	=445
2011	25 (67.5)	2 (5.4)	0	4 (10.8)	4 (10.8)	0	0	0	0	0	2 (5.4)	0	=37
2012	52 (61.2)	9 (10.6)	0	7 (8.2)	11 12.9)	1 (1.2)	0	0	0	0	5 (5.9)	0	=85
2014	26 (31.3)	1 (1.2)	0	24 (28.9)	15 (18.0)	13 (15.6)	0	0	0	1 (1.2)	3 (3.6)	0	=83
2015	25 (71.4)	0	0	4 (11.4)	4 (11.4)	0	0	0	0	0	2 (5.7)	0	=35
2016	0	0	0	1	0	0	0	0	0	0	0	0	=1
2018	89 (98.9)	0	0	0	1 (1.1)	0	0	0	0	0	0	0	=90
2019	21 (84.0)	2 (8.0)	0	0	0	1	0	6	0	0	1	0	=31
2021	2 (<1)	0	0	0	0	0	0	0	0	0	0	0	=2
Grand Totals	2932 (88.0)	64 (1.9)	4 (<1)	106 (3.2)	92 (2.8)	46 (1.4)	1 (<1)	7 (<1)	3 (<1)	1 (<1)	64 (1.9)	14 (<1)	3334

BR = Brown lemming; CO = Collared lemming; UL = Unidentified lemming; DU = Duck species; SH = Shorebird species; GU = Gull species; LO = Loon species; PT = Ptarmigan species; RA = Raptor species; AL = Alcid species; PA = Passerine species; UB = Unknown bird species.

). Overall, lemmings dominated, representing about 90.0% (89.9%) of the cached prey (Table 3). Brown lemmings dominated those, representing 88.0% (87.9%), collared lemmings 1.9%, and unidentified lemmings <1%. Recording caches allowed us to identify avian prey as species and groups. For example, ducks represented about 3.2% of the total prey, shorebirds 2.8%, passerines 1.9%, and gulls 1.4%. All other avian groups made up <1% of the total. Collectively, birds only represented about 10.0% of the prey cached at nests, although many species were eaten.

Brown lemmings dominated the owls’ diet from pellet analysis and prey cached at nests, in every year, ranging from 71.4 to 99.3% (Table 2) and 31.3 to 98.9% (Table 3), while birds ranged from <1 to 3.2% from pellets or caches (Table 2 and Table 3).

4.1. Food Niche Breadth from Pellet Analysis

Breeding season FNB and DIEV scores from pellets collected between 1992 and 2016 had a narrow range for prey identified as a group or species (

Table 4. Breeding season food niche breadth (FNB) and dietary evenness (DIEV) based upon prey recorded from pellets between 1992 and 2016. Shannon’s Diversity Index, based on H’ g = prey identified to groups and H’ s = prey identified to species; Simpson’s Diversity Index, based on R g = prey identified to groups and R s = prey identified to species; Hill’s dietary evenness, based on D g = prey identified to group, and D s = prey identified to species; number of prey items N g = identified to groups and N s = identified to species. * Pellets lost in 2004 and 2014. The short hyphen (-) means no data.

Year	H’ g	H’ s	R g	R s	D g	D s	N g	N s
1992	1.16	1.15	1.05	1.05	0.330	0.336	3711	3707
1993	1.11	1.11	1.03	1.03	0.320	0.324	9157	9153
1994	1.04	1.04	1.01	1.01	0.273	0.273	2206	2206
1995	1.22	1.21	1.08	1.07	0.361	0.368	4248	4243
1996	1.12	1.12	1.05	1.04	0.402	0.410	3379	3377
1997	1.91	1.80	1.53	1.49	0.585	0.619	312	308
1998	1.32	1.24	1.12	1.10	0.396	0.417	1107	1093
1999	1.55	1.41	1.24	1.19	0.444	0.471	1797	1762
2000	1.94	1.53	1.51	1.29	0.541	0.561	3270	3016
2001	1.45	1.07	1.22	1.02	0.507	0.323	272	247
2002	1.26	1.26	1.09	1.09	0.377	0.377	362	362
2003	1.68	1.68	1.32	1.32	0.479	0.479	381	381
2004 *	-		-		-		-	-
2005	1.70	1.30	1.35	1.14	0.502	0.477	529	484
2006	1.21	1.21	1.08	1.08	0.404	0.404	1741	1741
2007	1.08	1.05	1.02	1.01	0.313	0.321	3031	3019
2008	1.07	1.05	1.02	1.01	0.302	0.311	4438	4430
2009	1.50	1.02	1.29	1.00	0.592	0.275	905	789
2010	1.15	1.15	1.06	1.06	0.379	0.379	446	446
2011	2.35	1.74	1.73	1.37	0.541	0.506	333	293
2012	2.55	2.55	1.85	1.85	0.549	0.549	301	301
2013	2.30	2.30	1.80	1.80	0.617	0.617	247	247
2014 *	-		-		-		-	-
2015	1.39	1.39	1.18	1.18	0.495	0.495	325	325
2016	1.42	1.42	1.18	1.18	0.437	0.437	1191	1191
Total Prey							43,689	41,605
Mean	1.49	1.38	1.25	1.19	0.441	0.423
Ranges	1.04–2.55	1.02–2.55	1.01–1.85	1.01–1.85	0.273–0.617	0.273–0.617
SD	0.44	0.40	0.26	0.23	0.10	0.10

). For example, for the group level of resolution, Shannon’s Diversity Index H’ ranged from 1.04 to 2.55, mean = 1.49, SD = 0.44; Simpson’s Diversity Index R ranged from 1.01 to 1.85, mean = 1.25, SD = 0.26; and Hill’s Diversity Index D ranged from 0.273 to 0.617, mean = 0.441, SD = 0.10 (Table 4). Spearman rank correlation coefficients for H’, R, and D showed negative correlations for groups: r_s = −0.6432, r_s = −0.7603, and r_s = −0.7161, respectively, indicating sample sizes did not influence the results. Similarly, for the species level of resolution, Shannon’s Diversity Index H’ ranged from 1.02 to 2.55, mean = 1.38, SD = 0.40; Simpson’s Diversity Index R ranged from 1.01 to 1.85, mean = 1.19, SD = 0.23; and Hill’s Diversity Index D ranged from 0.273 to 0.617, mean = 0.423, SD = 0.10 (Table 4). Spearman rank correlation coefficients for H’, R, and D also showed negative correlations: r_s = −0.5729, r_s = −0.5637, and r_s = −0.5318, respectively, indicating sample sizes did not influence the results.

We then compared the median FNB and DIEV scores of prey groups (n = 23) versus species (n = 23) for H’, R, and D using Mann–Whitney U tests. There were no significant differences for either metric: H’ = U = 264.5, p > 0.05; R = U = 276.5 and D = U = 264.5, p > 0.05. Thus, prey identified as groups versus species yielded similar results from pellets (Table 4).

4.2. Food Niche Breadth from Cached Lemmings

Breeding season FNB and DIEV scores from prey cached at nests during 16 of the 18 years between 1992 and 2021 also had a narrow range from prey identified as a group or species (

Table 5. Breeding season food niche breadth (FNB) and dietary evenness (DIEV) based upon prey cached at nests between 1992 and 2021. Shannon’s Diversity Index, based on H’ g = prey identified to groups and H’ s = prey identified to species; Simpson’s Diversity Index, based on R g = prey identified to groups and R s = prey identified to species; Hill’s dietary evenness, based on D g = prey identified to groups and D s = prey identified to species; number of prey N g = identified to groups and N s = identified to species. * Sample sizes for 2016 and 2021 were too small for calculations. * Range at bottom of table is same for both categories in each column. The short hyphen (-) means no data.

Year	H’ g	H’ s	R g	R s	D g	D s	N g	N s
1993	1.07	1.07	1.02	1.02	0.308	0.308	527	527
1995	1.44	1.43	1.15	1.15	0.352	0.357	630	629
1996	2.25	2.14	1.54	1.51	0.432	0.447	273	270
1999	3.56	3.43	2.42	2.39	0.557	0.571	127	126
2000	1.91	1.91	1.37	1.37	0.410	0.410	416	416
2002	2.38	2.38	1.72	1.72	0.528	0.528	43	43
2003	3.10	2.55	2.28	1.98	0.610	0.632	27	25
2005	1.86	1.86	1.39	1.39	0.449	0.449	51	51
2006	1.37	1.27	1.12	1.09	0.334	0.330	431	424
2008	1.09	1.09	1.02	1.02	0.297	0.297	445	445
2011	2.89	2.89	2.05	2.05	0.560	0.560	37	37
2012	3.41	3.41	2.42	2.42	0.590	0.590	85	85
2014	4.70	4.70	4.15	4.15	0.852	0.852	83	83
2015	2.45	2.45	1.85	1.85	0.584	0.584	35	35
2016 *	-		-		-		1	1
2018	1.06	1.06	1.02	1.02	0.356	0.356	90	90
2019	2.66	2.66	1.98	1.98	0.594	0.594	31	31
2021 *	-		-		-		2	2
Total prey							3334	3320
Means	2.32	2.26	1.78	1.75	0.488	0.491
Range *	1.06–4.70	1.06–4.70	1.02–4.15	1.02–4.15	0.297–0.852	0.297–0.852
SD	1.03	1.01	0.80	0.79	0.148	0.149

). For example, Shannon’s Diversity Index H’ for prey identified as groups ranged from 1.06 to 4.70, mean = 2.32, SD = 1.03; Simpson’s Diversity Index R ranged from 1.02 to 4.15, mean = 1.78, SD = 0.80; and Hill’s Diversity Index D ranged from 0.297 to 0.852, mean = 0.488, SD = 0.148 (Table 5). Spearman rank correlation coefficients for H’, R, and D showed negative correlations for groups: r_s = −0.5764; r_s = −0.5816, and r_s = −0.8088, indicating sample sizes did not influence the results.

Similarly, for the species level of resolution, Shannon’s Diversity Index H’ ranged from 1.06 to 4.70, mean = 2.26, SD = 1.01; Simpson’s Diversity Index R ranged from 1.02 to 4.15, mean = 1.75, SD = 0.79; and Hill’s Diversity Index D ranged from 0.297 to 0.852, mean = 0.491, SD = 0.149 (Table 5). Spearman rank correlation coefficients for H’, R, and D also showed negative correlations: r_s = −0.5852, r_s = −0.6213, and r_s = −0.7941, respectively, indicating sample sizes did not influence the results.

We then compared the median FNB and DIEV scores of prey groups (n = 16 years) versus species (n = 16 years) for H’, R, and D using Mann–Whitney U tests. There were no significant differences for either metric: H’ = U = 123.5, p > 0.05; R = U = 124, p > 0.05; and D = U = 131.5, p > 0.05. Thus, prey identified as groups versus species yielded similar results from prey cached at nests (Table 5).

4.3. Comparing Pellets Versus Caches

We then compared the medians FNB and DIEV scores of groups identified from pellets (n = 14 years) and species identified from caches (n = 16 years). There were no significant differences in the medians H’ = U = 64.5, p > 0.05; R = U = 66, p > 0.05; D = U = 88, p > 0.05.

Finally, we compared the median FNB and DIEV scores of cached (n = 16 years) versus pellets (n = 14 years) at the species level for H’, R, and D (

Table 6. Comparison of food niche breadth (FNB) and dietary evenness (DIEV) scores of cached prey items (n = 3334) versus prey recorded from pellet analysis (n = 43,689) between 1992 and 2021. Shannon’s Diversity Index, based on H’ g = prey identified to groups and H’ s = prey identified to species; Simpson’s Diversity Index, based on R g = prey identified to groups and R s = prey identified to species; Hill’s dietary evenness, based on D g = prey identified to groups and D s = prey identified to species; number of prey N g = identified to groups and N s = identified to species. * Sample size too small for prey cached in 2016 and 2021. The short hyphen (-) means no data.

	H’			R			D
Year	Cache	Pellets		Cache	Pellets		Cache	Pellets
	s	g	s	s	g	s	s	g	s
1993	1.07	1.11	1.11	1.02	1.03	1.03	0.308	0.320	0.324
1995	1.44	1.22	1.21	1.15	1.08	1.07	0.352	0.361	0.368
1996	2.25	1.12	1.12	1.54	1.05	1.04	0.432	0.402	0.410
1999	3.56	1.55	1.41	2.42	1.24	1.19	0.557	0.444	0.471
2000	1.91	1.94	1.53	1.37	1.51	1.29	0.410	0.541	0.561
2002	2.38	1.26	1.26	1.72	1.09	1.09	0.528	0.377	0.377
2003	3.10	1.68	1.68	2.28	1.32	1.32	0.610	0.479	0.479
2005	1.86	1.70	1.30	1.39	1.35	1.14	0.449	0.502	0.477
2006	1.37	1.21	1.21	1.12	1.08	1.08	0.334	0.404	0.404
2008	1.09	1.07	1.05	1.02	1.02	1.01	0.297	0.302	0.311
2011	2.89	2.35	1.74	2.05	1.73	1.37	0.560	0.541	0.506
2012	3.41	2.55	2.55	2.42	1.85	1.85	0.590	0.549	0.549
2014	4.70	-	-	4.15	-	-	0.852	-	-
2015	2.45	1.39	1.39	1.85	1.18	1.18	0.584	0.495	0.495
2016 *	-	1.42	1.42	-	1.18	1.18	-	0.437	0.437
2018	1.06	-	-	1.02	-	-	0.356	-	-
2019	2.66	-	-	1.98	-	-	0.594	-	-
2021 *	-	-	-	-	-	-	-	-	-
Mean	2.32	1.54	1.42	1.78	1.26	1.20	0.488	0.439	0.440
Range	1.06–4.70	1.07–2.55	1.05–2.55	1.02–4.15	1.02–1.85	1.02–1.85	0.297–0.852	0.302–0.549	0.311–0.561
SD	1.03	0.46	0.38	0.80	0.26	0.21	0.148	8.18	0.078

). For H’ and R, there were significant differences: H’ = U = 55, p < 0.05; R = U = 59, p < 0.05; but not for D = U = 91.5, p > 0.05.

These results indicated that the method used to quantify prey (cache or pellets) may have influenced the FNB and DIEV scores, although they were close. The slight differences may be ecologically irrelevant because the percent of lemmings in the diet was about 90% for either method. Nonetheless, pellets reflect the most numerous prey items eaten, while caches increase our knowledge of prey species diversity.

4.4. Dietary Overlap Between Pellets and Prey Cached at Nests

To explore these methods of tallying prey, we compared prey recorded from pellets (n = 43,689) (Table 2) and prey cached at nests (n = 3334) (Table 3) at the group level of resolution. Pianka’s dietary overlap equation showed almost complete dietary overlap O = 0.998 (99.8%) (

Table 7. Dietary overlap; O = 0.998 (98.8%) calculated by comparing prey from pellets (n = 43,689; Table 2) and cached prey at nests (n = 3334; Table 3) between 1992 and 2021. Calculations based on group level of resolution.

Species/Groups	Pellets	Caches
Brown Lemming	41,350	2932
Collared Lemming	1363	64
Unidentified lemming	571	4
Duck spp.	44	106
Shorebird spp.	67	92
Gull spp.	11	46
Loon spp.	2	1
Ptarmigan spp.	1	7
Raptor spp.	9	3
Passerine spp.	37	64
Unidentified Birds	233	14
Unidentified Fish	1	0
Unidentified Alcid	0	1
Total	43 689	3334
Dietary Overlap	0.998 (99.8%)	0.998 (99.8%)

). This comparison suggests that recording prey from pellets gives almost the same results as recording prey cached at nests, given adequate sample sizes. However, recording prey cached at nests revealed higher species diversity in the owls’ diet (Table 1 and Table 3).

We do realize these may not all be independent samples, as some prey from caches can occur in pellets, and species such as shorebirds or unidentified birds were grouped, which may have influenced the results. Yet, the finding that collectively they made up <1% (pellets) and <10% (cached) of the diet supports our decisions. Nonetheless, we performed a comparison to determine which method yielded the most reliable results.

4.5. Prey Biomass Estimates

We used the rounded mean body mass of intact brown (71 g, n = 2508) and collared (61 g, n = 53) lemmings cached at nests to estimate the overall contribution to the owls’ diet (DWH, unpubl. data 1992–2024). Biomass estimates from brown and collared lemmings cached at nests yielded about 178 kg and 3 kg, respectively (DWH, unpubl. data 1992–2024). Biomass estimates for large birds, however, were biased. For example, the mean mass of six (two males; four females) freshly killed King Eiders (Somateria spectabilis) obtained from hunters was about 1662 g (1.6 kg) ± 214 g (range 1350–1950 g) (

Table 8. Comparisons of live body mass of King Eider and muscle mass actually eaten by the owls. Using the average adult live mass overestimated the three major muscle masses usually eaten by almost four (3.7). Essentially, about 73.5% of the eiders was not eaten. See Column 7 for the overestimate ratio.

Species	Actual Live Mass g (Sex)	Breast Muscle g (% of Total)	Humerus Muscle g (% of Total)	Femur Muscle g (% of Total)	All Muscle g (% of Total)	Overestimate Ratio
King Eider	1950 (M)	312 (16.0%)	67 (3.4%)	110 (5.6%)	489 (25.1%)	3.9
King Eider	1650 (M)	270 (16.4%)	55 (3.3%)	94 (5.7%)	419 (25.4%)	3.9
King Eider	1550 (F)	276 (17.9%)	42 (2.7%)	88 (5.7%)	406 (26.2%)	3.8
King Eider	1350 (F)	250 (18.5%)	52 (3.9%)	82 (6.0%)	384 (28.4%)	3.5
King Eider	1850 (F)	290 (15.7%)	70 (3.8%)	112 (6.0%)	472 (25.5%)	3.9
King Eider	1625 (F)	274 (16.9%)	87 (5.4%)	112 (6.9%)	473 (29.1%)	3.4
Total Mass	9975	1672 (16.8%)	373 (3.7%)	598 (6.0%)	2643 (26.5%)	3.7
Mean	1662	278	62	99	440
SD	214.3	20.8	15.9	13.4	43.0
Range	1350–1950	250–312	42–87	82–112	384–489

). The total biomass of these six eiders was 9975 g (9.9 kg). However, the owls mainly ate the breast, humerus, and femur muscles. The total weight of these three muscles was about 2643 g (2.6 kg) (mean = 440 g ± 43.0, range 384–489, n = 6), or about 26.5% of the total live weight of the six King Eiders (Table 8). Furthermore, although sample sizes were small, there were significant differences in the relative muscle mass from these ducks (chi-square = 15.890, df = 10, p < 0.0001). The owls primarily ate the breast muscles, which yielded about 4.5 and 2.8 times more meat than the humerus and femur muscles, respectively (Table 8). There were significant differences between the mean muscle mass of the breast versus the humerus and the breast versus the femur: t = 20.252, df = 10, p < 0.0001 and t = 17.773, df = 10, p < 0.0001, respectively. Thus, standard calculated biomass estimates based on live weights of these eiders would over estimate by a factor of almost 4 (3.7) what was actually eaten (Table 8). Indeed, about 73.5% of the eider ducks was not eaten.

5. Discussion

At Utqiaġvik, Snowy Owls appeared in 35 consecutive years from 1990 to 2024. The results from our food habit study cover 30 consecutive years (1992–2021) and support descriptive observations by early naturalists that Snowy Owls in this area are obligate lemming predators for breeding. Despite 23 species of prey being represented in pellets and 32 species in caches, brown and collared lemmings combined comprised 99% and 90% of all prey, respectively. Specifically, brown lemmings represented 95% and 88% of all prey eaten.

In order to breed successfully at Utqiaġvik, Snowy Owls need high densities of lemmings. In fact, the Snowy Owl at Utqiaġvik relies primarily on one species of lemming—the brown lemming—and is essentially a one-prey-species obligate lemming predator for successful reproduction. Also, at Utqiaġvik, populations fluctuations of Snowy Owls are correlated with population fluctuations of brown lemmings (DWH, unpubl. data, this study 1992–2024). The owls’ nest in large numbers, lay more eggs, and produce more offspring during brown lemming population highs (Holt et al. 2025 [3]). However, intervals between lemming population highs, their amplitude, and their density vary annually over time (DWH, unpubl. data, this study 1992–2024).

5.1. Why Such a Narrow Feeding Niche at Utqiaġvik?

Throughout the Snowy Owl Arctic breeding range, natural selection appears to favor a narrow feeding niche for breeding. The most reasonable explanation is the narrow diversity of prey species throughout the Arctic (Holt et al. 2024 [37]). At Utqiaġvik, for example, only three species of rodents occur. Of these, two species are lemmings, brown and collared, and the third species is the Arctic ground squirrel (Urocitellus parryii). The ground squirrel is uncommon, hibernates much of the year, and is in hibernation when Snowy Owls arrive on the breeding grounds. There are few species of shrews and weasels and no hares. Of the two lemming species, the collared is uncommon, its populations fluctuate somewhat at Utqiaġvik, though never dramatically, and it never reaches the extremely high density seen in brown lemmings (Pitelka and Batzli 1993 [30], Pitelka and Batzli 2007 [31], DWH, unpubl. data, this study 1992–2024). Instead, it is the dramatic population fluctuations and high densities of brown lemmings that dictate breeding opportunities for the Snowy Owls.

Even though historic data show dramatic population fluctuations of brown lemmings at Utqiaġvik, only periodically did lemmings reach extremely high densities. For example, depending on the data used and analysis, the intervals between lemming population highs (“cycles”) can be considered as 2–6 years, with an average of 3.8 years for peak densities (Pitelka 1973 [29]); both 2–4 and 5–6 years (Pitelka and Batzli 1993 [30]); or 3–5 years (Pitelka and Batzli 2007 [31]). Regardless, the Utqiaġvik region is believed to maintain the highest densities of brown lemmings on the North Slope of Alaska (Bee and Hall 1956 [41], Pitelka and Batzli 1993 [30]), and brown lemmings here are one of the most well-studied lemming populations in the world (Pitelka and Batzli 2007 [31], 2018 [32], DWH, unpubl. data, this study 1992–2024). Whether brown or collared lemmings at Utqiaġvik cycle in accordance with the strictest definition of the word, or in a predictable mathematical fashion, is debatable. Nonetheless, population fluctuations of both lemming species—but particularly brown lemmings—do exist, and patterns do emerge.

5.2. How Do Snowy Owls Find Lemming Hotspots for Breeding?

What Search Images do Snowy Owls at Utqiaġvik and elsewhere use to discover high densities of lemmings for breeding? The answer is unknown. However, a common hypothesis from both historic and present-day researchers is that during spring migration, Snowy Owls travel and search the tundra for lemmings. It is assumed the owls visually search for high densities of lemmings that would allow them to settle and breed, settle and not breed, or move on. Perhaps this is true; however, there is no validation through direct observations from field researchers, or boots-on-the-ground validation of inferences from satellite-tracked owls (Therrien et al. 2011 [16], and review in Holt et al. 2025 [3]). In reality, the idea of observing lemmings at the time of year when Snowy Owls are selecting nesting areas is difficult for researchers to validate. The tundra is mostly inaccessible and snow-covered, and importantly, lemmings are believed to remain subnivean until the snow-melt. However, this last statement remains to be proved. Do brown lemmings stay brown because they do not surface during winter? Inũpiat Eskimo people from Utqiaġvik report to DWH that brown lemmings are periodically seen scurrying about the town throughout the winter months. Does this same behavior occur on the surrounding tundra? Why would the collared lemmings turn white in winter if they did not surface and scurry about on a snowy landscape?

Alternative Search Image hypotheses to wandering the tundra searching for lemmings are feasible and could be age- and sex-linked. Perhaps older adult owls know through previous experience where lemming hotspots occur and visit these areas in sequence during spring migration or spring wandering. Perhaps they are also looking for other Snowy Owls. Perhaps adult male Snowy Owls arrive first, or they arrive near-simultaneously with females on potential breeding grounds. Males then assess lemming densities and establish territories. In turn, females search for adult males on these territories, rather than searching for lemmings directly. Females then assess the males with resources, and either they decide to settle with presumably high-quality males on high-quality territories or they move on. Why else would age-deferred adult male white plumage, adult sexual color dimorphism, and male courtship behaviors have evolved if not for males to attract females? Females at Utqiaġvik only breed with older adult white-plumaged males. Younger males, which resemble females in plumage until about two to three or more years of age, do not breed at Utqiaġvik. Perhaps these younger-aged owls are looking for other owls as an indication of lemming abundance, rather than looking for lemmings. See Holt (2022 [40]) and Holt et al. (2025 [3]).

In years when lemming populations were low, the owls did not breed or bred in very small numbers, and sometimes, there were zero females in our study area. Nonetheless, single adult white-plumaged males established themselves on traditional nesting sites for several weeks, even if not breeding. This suggests there are adequate numbers of lemmings to sustain individual owls, but not the densities needed for breeding. The lack of females buttresses this argument and further suggests that females may be more sensitive to the lemming population density than males, whereby they move through the area if the lemming population density or male nuptial gifts (i.e., lemmings) do not fit their reproductive needs. Overall, females may assess whether to stay or continue their migration using three or more strategies: (1) male color as an indication of age, territory held, and fitness, (2) male courtship offerings of lemmings as an indication of the lemming density for breeding, or (3) females hunting for themselves and assessing lemming densities for their own reproductive needs.

5.3. Conservation of Lemmings

Lemmings are small Arvicoline rodents, with the most northerly distribution of any rodent species in the world (Stenseth and Ims 1993 a [46], b [47]). Lemmings live much of their life under the snow and are known to have dramatic population fluctuations over time, often referred to as “cycles” (Stenseth and Ims 1993 a [46], b [47], Pitelka and Batzli 2007 [31], 2018 [32], Krebs 2013 [48], Erich et al. 2020 [49], Gauthier et al. 2024 [50], Krebs 2024 [51]). Throughout their worldwide range, lemmings drive the ecology of the tundra and have an enormous influence on survival and reproduction in direct and indirect ways among a host of avian and mammalian species (see Kasrud et al. 2008 [52], Gilg et al. 2006 [53], Gilg et al. 2009 [54], Karell et al. 2008 [55], Ims et al. 2011 [56], Schmidt et al. 2012 [57], Holt et al. 2025 [3]). For example, other researchers have provided convincing data that when local lemming populations decline, obligate lemming predators such as Snowy Owls and other predatory and non-predatory species also decline (Kasrud et al. 2008 [52], Gilg et al. 2006 [53], Gilg et al. 2009 [54], Karell et al. 2008 [55], Ims et al. 2011 [56], Schmidt et al. 2012 [57]). These suggestions hold true for Snowy Owls at Utqiaġvik (DWH, unpubl. data, this study 1992–2024) but remain to be quantified for other species (see Holt et al. 2025 [3] for an assessment).

On the other hand, Erich et al. (2020 [49]) and Gauthier et al. (2024 [50]) used trend analysis of several long-term lemming studies, and suggested no global declines in lemming populations throughout their Arctic breeding range over the past 40 or so years. If true, then how do we balance local lemming population declines reported by several other authors (Kasrud et al. 2008 [52], Gilg et al. 2006 [53], Gilg et al. 2009 [54], Karell et al. 2008 [55], Ims et al. 2011 [56], Schmidt et al. 2012 [57]) with Erich et al. 2020 [49] and Gauthier et al.’s 2024 [50] results? Which is the more important conservation issue, (1) local/regional lemming declines, or (2) apparent worldwide lemming populations’ stability?

Interestingly, too, Gousy-LeBlanc et al. (2023 [5]) used single-nucleotide polymorphism (SNP) methods to suggest that Snowy Owls have been declining for about 200 years. If we assume that Snowy Owls were near-obligate lemming predators 200 years ago, then have lemmings been declining in synchrony with Snowy Owls for 200 years? If present-day lemming populations are stable throughout the world, as suggested by Erich et al. (2020 [49]) and Gauthier et al. (2024 [50]), and lemmings are the major factor underpinning the successful breeding output for Snowy Owls (Holt et al. 2025 [3]), then is the purported decline of Snowy Owls worldwide (Gousy-LeBlanc et al. 2023 [5]) true? Could the SNP techniques used by Gousy-LeBlanc (2023 [5]) be used to assess if lemming populations have declined or not in synchrony with Snowy Owls over the past 200 years?

5.4. Conservation of Snowy Owls

Given our results, environmental perturbations that affect and result in effects on lemming populations at Utqiaġvik will have direct effects on Snowy Owl breeding populations there, and perhaps globally (Holt et al. 2024 [37], Holt et al. 2025 [3]). Recently, some authors have suggested that voles are moving north with a warming climate and invading or replacing lemmings in some areas (Golovnyuk 2017 [58], Dudenhoffer et al. 2021 [59], Sokolova et al. 2024 [60]). If true, how will vole range expansion affect and have an effect on lemmings and Snowy Owls? Will voles replace lemmings in some Arctic regions? Will voles meet the energetic requirements in terms of numbers, density per unit area, and body mass to sustain breeding Snowy Owls? Although voles were identified as important in a few Snowy Owl breeding studies, lemmings are clearly the driving force behind successful Snowy Owl breeding worldwide (Holt et al. 2025 [3]).

6. Conclusions

At Utqiaġvik, Snowy Owls have an ability to assess lemming population density rather quickly and decide whether to stay or move on. The owls never miss a spring lemming population high and stay and breed when densities reach the owls’ preferred numbers (DWH, unpubl. data, this study 1992–2024). It appears, however, that female Snowy Owls are more sensitive to lemming densities, likely for breeding purposes. We buttress this argument with the fact that few or zero females are found in the study area during years with low lemming populations, while adult males are always present, whether breeding or not. This scenario may also be applicable throughout the owls’ Arctic breeding range.

Food drives life, and Snowy Owls are locked in a near-obligate predator–prey relationship with lemmings. Studies of Snowy Owl food habits must continue to determine if environment changes affect the lemming species distribution and density, both intra-specifically and inter-specifically. Furthermore, food habit studies may also allow us to detect the range expansion of voles. How will these changes, if they occur, impact the owls? Although food habit studies for many species of organisms can be difficult to assess, and in some cases redundant, or even assumed, food resources must be monitored continuously over the long term. If possible, methods taken to effectively assess lemming density through some sort of monitoring program are advised. For example, grid sampling, snap-trap sampling, observational transects, or casual observation could be performed. The most common method worldwide is snap-trapping.

Lemmings can have an enormous influence on the local ecology of the tundra wherever they occur, and they may be the most significant indicator of a healthy Arctic. However, lemmings are difficult to observe and study. Therefore, by monitoring the Snowy Owl distribution and annual nesting production, conservationists may be able to assess annual and long-term hotspots for many species of Arctic fauna and determine the relative health of the Arctic environment.

In conjunction with any of the above, one must have some assessment of annual Snowy Owl nest production. Just counting nests without some measure of reproductive output has limited value in population monitoring. Changes in lemmings’ distribution and fluctuations in their populations and thus their availability will affect and produce effects on Snowy Owls. Ultimately, the future of Snowy Owls is tied to the future of lemmings.