A Worldwide Review of Snowy Owl Feeding Ecology: The Importance of Lemmings and Voles in a Changing Climate

Simple Summary: It is common knowledge that Snowy Owls eat lemmings during the nesting season, and it is believed that lemming populations fluctuate dramatically over several years, often termed a cycle. When lemming populations reach high densities, Snowy Owls have a highly successful breeding season, producing numerous young per nest. This successful nesting results in large numbers of young Snowy Owls migrating south during their first winter. This

Nonetheless, it has been common knowledge for >100 years, that Snowy Owls are dependent on high densities of lemmings for successful breeding (Holt et al., 2020 [4], Holt et al., 2024 in press [5]).But surprisingly few studies have quantified this common knowledge with large sample sizes from the entire Snowy Owl breeding range (Holt et al., 2020 [4], Holt et al., 2024 in press [5]).And, given the importance of lemmings to Snowy Owl breeding, there has been no overall review of the owls' feeding ecology and relationship with lemmings in a changing climate.
Herein, we provide an assessment of the Snowy Owl breeding season feeding ecology throughout its circumpolar breeding range.We emphasize the importance of lemmings for the successful breeding of Snowy Owls.We also suggest how a changing climate might affect, and have an effect on, lemmings, with related consequences to Snowy Owls.To emphasize our assessment, we also compare breeding season feeding ecology with non-breeding season/winter feeding ecology.
Our objectives were as follows: (1) compare the Snowy Owl breeding season feeding niche throughout the Arctic; (2) compare the breeding season feeding niche with the winter feeding niche; (3) discuss the importance of lemmings to Snowy Owl breeding success and conservation; (4) discuss the potential affects and effect of lemmings being replaced by voles in Arctic regions as the climate warms, and (5) suggest future research directions.

Materials and Methods
We reviewed breeding season food habits of 15 studies throughout the Snowy Owls' worldwide breeding range.These areas include Canada, Finland, Greenland, Norway, Russia, Sweden, and the United States (Figure 1).We only included studies with >200 prey items to ensure we had reasonable sample sizes of prey, and a reasonable sample of studies.For some studies, we back-calculated prey reported as percentages to numbers.

Materials and Methods
We reviewed breeding season food habits of 15 studies throughout the Snowy Owls' worldwide breeding range.These areas include Canada, Finland, Greenland, Norway, Russia, Sweden, and the United States (Figure 1).We only included studies with >200 prey items to ensure we had reasonable sample sizes of prey, and a reasonable sample of studies.For some studies, we back-calculated prey reported as percentages to numbers.We summarized and discussed the percent of total prey that were small mammals (i.e., lemmings, voles) for the combined studies.To describe the owls' feeding ecology, and make comparisons among studies, we calculated Food Niche Breadth (FNB) and Dietary Evenness (DIEV), see below.We also compared breeding season feeding ecology from these 15 studies, with seven winter season feeding ecology studies, reviewed by Detienne et al. (2008) [45].We did this to show the dietary diversity and flexibility of Snowy Owls during winter, but the obligatory narrowness during breeding.
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, [46]).Ultimately, these equations allow large data, such as those presented here, to be weighted in order of "importance" and expressed as single values.For example, although several species of birds or small mammals could be in the diet, they are such a low percentage that they are not important, while others are.In turn, values can be compared from different studies, provided that the same or similar methods are used.There are strengths and weaknesses to each equation, with some reviewers favoring one equation over another, and some reviewers see little or no value (see Zar, 1999 [47], Marti et al., 2007 [46], Collier and Schertner, 2012 [48] for reviews).For example, the antilog of Shannon-Weaver's diversity index is linearly related to the number of prey categories in the sample.Some statisticians believe H ′ can underestimate diversity unless samples are large (Zar, 1999 [47] (p.41), whereas others believe H ′ can be biased if many species are represented (Collier and Schwertner, 2012 [48] (p.62).For Simpson's equation, we used the reciprocal (1/D) because it is believed to yield a more meaningful number (Marti et al., 2007 [46]).Some statisticians prefer Simpsons D over Shannon-Weaver H', as it is simple to calculate and understand, robust, and meaningful (Collier and Schwertner, 2012 [48] (p.62).Due to differing opinions, we provide both values for comparisons.
We defined the annual feeding niche as the relationship between the owls and their prey, and followed Holt, 1993 [49], Cromrich et al., 2002 [50], Detienne et al., 2008) [45] in describing FNB and DIEV.Because FNB and DIEV are influenced by levels of resolution, we attempted to identify all prey to species.However, that was not always easy, because of different methods used to identify prey by different researchers, and characteristics of prey were not always identifiable.
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 [51], Zar, 1999 [47]).Alpha levels were set at 0.05.
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 distribution in the owls' diet (i.e., homogeneous).Because DIEV scores range from zero to one, we considered a score of one or approximating one, a uniform representation of prey proportions in the diet.And, scores < 0.50 represent non-uniform or shared distribution of prey in the diet (see Detienne et al., 2008 [45]).

Information Theory
Food niche breadth (FNB) was calculated using the antilog of the Shannon-Weaver index, where H ′ = −Σp i log p i and p i represents the proportion of each species in the sample.Food niche breadth FNB using Simpson's equation was calculated, where D = Σp i where p i is the proportion of each member of the assemblage being investigated.Hill's dietary evenness (DIEV) scores were calculated, where and N 1 is the antilog of the Shannon-Weaver index (H ′ ) and N 2 is the reciprocal of Simpson's index, where (1/D).We used Spearman Rank Correlation to examine if FNB and DIEV scores were influenced by sample size.
We then discuss our findings on the predator-prey relationship between Snowy Owls and small mammals, particularly lemmings.Finally, we suggest future research directions.

Comparison with Other Breeding Season Studies
The combined 15 studies yielded 59,923 prey (Table 1).Although various methods were used to tally prey among studies, we tried to standardize these data with our methods.Thus, 59,585 prey were used for these calculations.Overall, the mean feeding niche calculated from the 15 breeding season studies was narrow: H ′ = 1.95;R = 1.60,D = 0.526.The lowest values were H ′ = 1.07;R = 1.01;D = 0.266, from Greenland, and the highest: H ′ = 4.01; R = 3.24; from Finland, while D = 0.873 from Canada (Table 1).In 10 of the 15 studies, lemmings (Lemmus and/or Dicrostonyx) were >71.8% of the owls' diet.Lemmings dominated the diet in most studies (Table 1).In four studies, Lemmus was >90% of the diet, and in three studies Dicrostonyx was >90% of the diet.In three other studies, Lemmus represented 71.8%, 73.8%, and 84.0% of the diet.In one study, Lemmus and Dicrostonyx were about equally represented: 49.1% and 47.5%, respectively.In the remaining four studies, Microtus and Clethrionomys voles were important.For example, in one study from Norway, Microtus made up 80.5% of the diet, while in two other studies from Finland, prey was more evenly distributed among three species, with Lemmus (41.8%),Clethrionomys (35.7%), and Microtus (20.5%) having higher proportions in one study, and Lemmus 58.1%, Clethrionomys 23.3%, and Microtus 13.5% in the other study.In the final study from Russia, Lemmus and Microtus were 53.0% and 37.8% of the diet (Table 1).

Comparison with Other Non-Breeding Season Studies
Feeding ecology of wintering Snowy Owls in North America was conducted and reviewed by Detienne et al. (2008) [45].They only calculated Shannon-Weaver H', and Hill's Dietary Evenness D, for their study.Of 5888 prey items reported from 7 studies, the mean winter FNB (H ′ = 4.61) was more than twice that of the 15 breeding seasons reported herein, FNB (H ′ = 1.95).Interestingly, however, DIEV between winter; D = 0.531 (n = 7 studies) and breeding DIEV, D = 0.526 (n = 15 studies) were similar (Table 2).

Discussion
Our comparisons of studies from Nearctic and Palearctic regions buttress a consensus that Snowy Owls are primarily obligate lemming predators for successful breeding.In fact, lemmings are the primary prey from most areas, although northern species of voles can be important in certain years, and regions.Overall, however, lemmings constitute the majority of prey in all but one study.And, only in unusual circumstances do breeding Snowy Owls deviate from this predator-prey relationship.Thus, environmental perturbations that affect, and have an effect on, populations of lemmings could have direct effects on Snowy Owl breeding populations-whether negative or positive.
The fact is, lemmings drive the ecology of the tundra wherever they occur, and influence the reproductive output of numerous species of birds and mammals (see Behavior; Social and Interspecific Behavior; Unique Snowy Owl interactions with other species/Association with Non-predators: in Holt et al., 2024, in press [5]).
Nonetheless, why has natural selection favored such a narrow feeding niche for breeding Snowy Owls?Foremost, the most reasonable explanation is the narrow diversity of prey species, particularly small rodents, throughout the Snowy Owls' Arctic breeding range.Next, the Snowy Owl is one of the largest and heaviest owls in the world (Holt et al., 1999 [1], Holt et al., 2024, in press [5]).Thus, its individual food requirements are likely high.And, feeding a family-female and brood probably require an enormous amount of food.Consequently, preying on the sparsely distributed stable population of small tundra nesting birds apparently does not fulfill these energetic requirements.However, preying upon abundant small rodents such as lemmings, during population highs, do.Indeed, lemmings and voles can exhibit dramatic population fluctuations, where intervals between population highs, amplitude, and density, vary annually.For example, during times of high populations, lemmings can increase by factors of 25, 50, 200, or more (see Krebs, 1964 [60], Krebs, 1993 [43], Pitelka and Batzli, 1993 [61]; Pitelka and Batzli, 2007 [41], Krebs, 2013 [44], Ehrich et al., 2020 [40], D. Holt, unpubl.data, this study 1992-2021).
Each Snowy Owl and lemming annual population fluctuations have their own signature, and numerous factors-abiotic and biotic-are believed to influence those fluctuations.So how do Snowy Owls find, and assess, regional populations of lemmings, and decide to stay and breed, move on, or forego breeding for that season?It has long been suggested Snowy Owls wander the Arctic in search of high lemming densities in order to breed.This hypothesis remains to be verified annually on a large geographic scale.Nonetheless, what cues/mechanisms do Snowy Owls use to search and discover high densities of lemmings or voles for breeding?Currently, this is unknown, but at our study site in Utqia ġvik, Alaska, Snowy Owls never miss a high lemming population, and immediately respond to annual spring numbers (D.Holt, unpubl.data, this study 1992-2021).The owls appear to have the ability to assess lemming density quickly-perhaps a day or a few days-and stay or move on, as suggested by Holt et al. (2015) [37], Holt et al., 2024 [5], in press).

Climate Change
Will climate change affect future lemming populations?In one of the first reports implicating the effects of snow conditions, Shelford (1943) [62] recorded a decline in collared lemmings associated with the absence or reduced amounts of snow cover at Churchill, Manitoba, Canada.And, approximately 40-45 years ago, researchers at Utqia ġvik, Alaska, cautioned that climate change could affect snow quality, which is important for winter breeding in lemmings.Thus, any climate change-related effects on winter snow quantity and quality could affect lemming populations (Fuller, 1967 [63], MacLean et al., 1974 [64], Batzli, 1981 [65]).
In an interesting inferential study using SNP methods, Gousy-LeBlance et al. ( 2023) [8] suggested Snowy Owls have been declining for about 200 years.If true, then have lemmings been declining since then too?Given the conflicting results from papers cited within this section, we suggest a need for another worldwide lemming population assessment.And, Birds 2024, 5 348 if possible, we also suggest a SNP study be conducted on lemmings to assess if their populations have declined in step with the purported Snowy Owl decline.

Nutrition
Does the nutritional value of lemming food fluctuate, and thus affect lemming population fluctuations, as suggested by Pitelka andBatzli (2007, 2018) [41,42]?If the nutrition of lemming forage is changing, will this affect lemming growth and body mass?In turn, could this affect successful Snowy Owl breeding?
Of the several multifactorial events driving lemming population fluctuations suggested by Krebs (2013, pp. 190-208) [44], Pitelka and Batzli (2018) [42] stressed that nutrition-the nutrient recovery hypothesis-likely plays the "most important role" in seasonal changes, with predation playing a "smaller role", and social strife having "little effect".Whether the nutrient-recovery hypothesis, predation, or social strife in lemming population dynamics has a hierarchal order is not clear.In a thorough review of predation on small mammal population structure in rodents, Krebs (2013, pp.160-161) [44] emphasized that predation can play a major and a minor role in influencing population structures.

Vole Incursions
What about northern latitude species of voles?Are they declining or stable in response to climate change?Voles in some latitudes appear to be expanding north.In fact, three studies have suggested vole species are moving northward with a warming climate, and infiltrating lemming habitats (Golovnyuk, 2017 [70], Dudenhoeffer et al., 2021 [71], Sokolova et al., 2023 [72]]).If this is a continuing trend, how will vole range expansion affect, and have an effect on, lemmings and Snowy Owls?Will voles eventually replace lemmings in some Arctic regions as the Arctic warms?Will vole population fluctuations ("cycles") be similar to lemming population fluctuations?Will voles be an adequate food source in terms of numbers/density, and body mass, to maintain breeding Snowy Owls, and other obligate lemming predators?Will voles be as easy to catch as lemmings are believed to be?In a few studies cited herein, voles were important to Snowy Owls' breeding; however, lemmings are clearly the driving force behind Snowy Owl breeding worldwide (Holt et al., 2024, in press [5]).

Conclusions
Snowy Owl and lemmings appear to have a long evolutionary predator-prey history.We recommend long-term studies that monitor lemming population fluctuations while simultaneously monitoring Snowy Owl population fluctuations, and nesting production.This is the only way to assess factors that might influence the predator-prey dynamics between these species.These studies should cover multiple lemming population fluctuations, perhaps 20 or more years at a minimum.Any environmental perturbations that affect, and have an effect on, lemming and vole populations throughout the Arctic, whether negative, neutral, or positive, will have direct effects on Snowy Owl breeding populations.
The Snowy Owl is the avian icon of the Arctic.However, it is lemmings that drive Snowy Owl breeding success.Unfortunately, lemmings are rodents, and have rodent stigma.Indeed, they do not fit the charismatic social appeal that other animals do.However, given Snowy Owls' enormous popularity and good looks, it renders them the poster animal to help galvanize Arctic conservation.By using Snowy Owl population fluctuations as an indicator of lemming population fluctuations, Snowy Owls could be the harbinger/indicator of a healthy Arctic environment, both locally and wide-ranging.Finally, one must conclude-how lemmings go-so do Snowy Owls.

Figure 1 .
Figure 1.Approximate locations of Snowy Owl food habits studies cited in this manuscript.Note there are 16 red pins for the 15 studies.The Wrangel Island site is included in the Wiklund et al. (1998) [25].

Table 1 .
Food niche breadth (FNB) and dietary evenness (DIEV) of breeding Snowy Owls from 15 studies worldwide.Total prey items were 59,923.However, because we used only prey identified to species for niche calculations, the total number was 59,585.H' = Shannon's Diversity Index; R = Simpson's Diversity Index; D = Hill's Dietary Evenness Index.Table Code: Location = country of study; # prey = total number reported from study; # prey used = total number of prey items identified to species at least once; # species = total number of species identified from each study; % L = Lemmus lemmings; % D = Dicrostonyx lemmings; % M = Microtus voles; C = Clethrionomys voles; % O = Other prey species.

Table 2 .
Food niche breadth (FNB) and dietary evenness (DIEV) of Snowy Owls from seven nonbreeding/winter studies from North America, representing 5888 prey items.Prey were identified as species for comparison.Table Code: Location = USA state and/or country of study; total prey = # total number reported from the study; H ′ = Shannon's Diversity Index; D = Hill's Dietary Evenness Index.