Migratory behaviours are risk-sensitive to physiological state in an elevational migrant

Interactions between physiological and behavioural adaptations to seasonal environments may be key to unravelling the mystery of partial migration. Physiological state, as indicated by body fat, influenced probability of migration in bighorn sheep as well as migration distance and probability of switching migratory tactics across years.


Accretion of body fat by animals is an important physiological adaptation that may underpin seasonal behaviours, especially
where it modulates risk associated with a particular behaviour.Using movement data from male Sierra Nevada bighorn sheep (Ovis canadensis sierrae), we tested the hypothesis that migratory behaviours were risk-sensitive to physiological state (indexed by body fat).Sierra bighorn face severe winter conditions at high elevations and higher predation risk at lower elevations.Given that large body fat stores ameliorate starvation risk, we predicted that having small body fat stores would force animals to migrate to lower elevations with more abundant food supplies.We also predicted that body fat stores would influence how far animals migrate, with the skinniest animals migrating the furthest down in elevation (to access the most abundant food supplies at that time of year).Lastly, we predicted that population-level rates of switching between migratory tactics would be inversely related to body fat levels because as body fat levels decrease, animals exhibiting migratory plasticity should modulate their risk of starvation by switching migratory tactics.Consistent with our predictions, probability of migration and elevational distance migrated increased with decreasing body fat, but effects differed amongst metapopulations.Population-level switching rates also were inversely related to population-level measures of body fat prior to migration.Collectively, our findings suggest migration was risk-sensitive to physiological state, and failure to accrete adequate fat may force animals to make trade-offs between starvation and predation risk.In complex seasonal environments, risksensitive migration yields a layer of flexibility that should aid long-term persistence of animals that can best modulate their risk by attuning behaviour to physiological state.

Introduction
In highly seasonal environments, animals have evolved a suite of adaptations to cope with dramatic changes in the amount, quality and distribution of available resources between seasons.Physiological adaptations, such as circannual and circadian rhythms (Erriksson et al., 1981;Loudon, 1989;Hiebert et al., 2000;Arnold et al., 2018), seasonality of metabolic rates (Chappel and Hudson, 1980;Smit and McKechnie, 2010), seasonal and physiological changes in gut morphology and volume (Zimmerman et al., 2006) and seasonal hypoor hyperphagia (Peltier et al., 2003;Barboza et al., 2006;Fuchs et al., 2019) contribute to changes in food acquisition and nutrient assimilation that underpin an annual cycle of body mass and fat (Mautz, 1978;Cook et al., 2004;Barboza and Parker, 2008;Taillon et al., 2013;Smiley et al., 2022).Body fat stores are particularly significant to large-bodied endotherms that rely on these to finance survival and reproduction in seasonal environments (Cook et al., 2004;Barboza and Parker, 2008;Monteith et al., 2013;Taillon et al., 2013;Smiley et al., 2022).The importance of body fat in seasonal environments may extend beyond direct effects on survival and reproduction, particularly where the relative value of body fat as a nutritional buffer varies across behavioural states (Denryter et al., 2022a), but this notion remains largely untested.
According to the risk-sensitivity hypothesis, animals should aim to optimize resource gains (or fitness) relative to constraints via two possible strategies: risk-averse and riskprone.Classically, risk-averse strategies were characterized by behaviours that result in a consistent, but modest reward, so there is little risk of not being rewarded (Caraco et al., 1980).In contrast, risk-prone strategies were characterized by behaviours that result in a variable reward value that can change and be more valuable or less valuable than the constant reward (Caraco et al., 1980).In risk-prone strategies, animals risk the constant (guaranteed) reward for the possibility of gaining a greater pay-off; the pay-off, however, may not be realized.Although classical work explored risk sensitivity relative to exogenous resources (Caraco et al., 1980;Stephens, 1980), more recent work has demonstrated risk sensitivity relative to endogenous resources also occurs.For example, female ungulates alter investment in survival and reproduction relative to body fat stores in a risk-sensitive (averse) manner (Bårdsen et al., 2008;Monteith et al., 2013;Smiley et al., 2022).For animals that rely on accretion of body fat to finance life in seasonal environments, risk sensitivity relative to body fat may be critical to determining the adaptive value of seasonal behaviours including migration (Smiley et al., 2022).
Migration occurs almost universally amongst vagile species, but is as diverse and complex as it is common, and comprises decisions at multiple spatial and temporal scales (Aikens et al., 2022).At coarse scales, animals choose a migratory tactic, whereas at finer scales, animals choose when to migrate, where to migrate (or how far to move) and how long to remain on their migratory range.Traditionally, migratory tactics were considered dichotomous-either animals migrated (i.e. made a single round-trip movement between two distinct ranges) or they were residents that remained on a single range year-round.We now understand migration and residency are ends of a continuum that include intermediate migratory tactics such as making multiple round trips between ranges (e.g.'vacillators' or 'commuters') or using more than two ranges (e.g.'mixed migrants' or 'multirange migrants') (Bunnefeld et al., 2011;Cagnacci et al., 2011;Denryter et al., 2021;van de Kerk et al., 2021;Enns et al., 2023).Some individuals and populations employ the same tactic year after year, exhibiting little migratory plasticity (Sawyer et al., 2018), whereas others have more mutable (plastic) migratory behaviours and choose different tactics across years (Spitz et al., 2018).Why some individuals and populations exhibit greater diversity in migratory tactics than others remains enigmatic, but understanding the relationship between migratory portfolios and demography may be important to conservation and management (Lowrey et al., 2020).Risk-sensitive decision-making relative to body fat, however, could be a plausible explanation because the relative value of body fat to survival differs amongst migratory tactics (Denryter et al., 2022a) and because the choice between residency and migration may itself reflect a trade-off between nutrition and predation.
In a risk-sensitive paradigm of migration for partially migratory populations, animals should choose a migratory tactic that affords them the greatest potential fitness pay-off.Where large quantities of body fat are critical for survival of individuals pursuing a given migratory tactic, animals should only choose that tactic if they have achieved adequate body fat stores (Cornelius et al., 2021;Denryter et al., 2022a).If an individual does not achieve adequate body fat stores for that migratory tactic, they should choose an alternative.For example, in altitudinal or elevational migrants, residents remain at high elevations during winter, where food supplies are depauperate compared with lower elevations, but predators are generally less common (Johnson et al., 2010;Greene et al., 2012;Spitz et al., 2020;Gammons et al., 2021).For those residents to survive winter, they rely strongly on body fat stores (Denryter et al., 2022a) and therefore, high levels of body fat should be a prerequisite for residency in a risk-sensitive framework of migration.In contrast, migrants that move to lower elevation ranges with less snow and more abundant food supplies should be less reliant on body fat for overwinter survival, but also face greater risk of predation (Johnson et al., 2010;Greene et al., 2012;Spitz et al., 2020;Gammons et al., 2021).If choice of migratory tactic is risk-sensitive to body fat, we predict that migratory propensity would increase with decreasing body fat stores.Essentially, body fat would allow residents to exhibit a risk-prone strategy to starvation, whilst being risk-averse to predation.In contrast, migrants would have lower body fat levels and be risk-averse to starvation (i.e.selecting a strategy with a more consistent food reward), but simultaneously risk-prone to predation.Further, animals may attune their migratory decisions at finer temporal and spatial scales relative to body fat to further reduce their risks.For example, body fat levels may dictate how far down in elevation an individual moves because predation risk is highest on low-elevation ranges (Johnson et al., 2013) and hence animals should stay as high in elevation as their body fat stores would allow (i.e.risk-averse to starvation, but riskprone to predation).
To test the hypothesis that choice of migratory behaviour was risk-sensitive to physiological state (as indicated by body fat), we used movement and body fat data from a migratory ungulate with high levels of migratory plasticity-Sierra Nevada bighorn sheep (Ovis canadensis sierrae; hereafter Sierra bighorn).In a risk-sensitive model of migration, differences in body fat also could underpin facultative migration and migratory plasticity.Hence, we also hypothesized that differences in body fat underpin facultative migration and predicted that at the population level, switching rates would increase with decreasing body fat because as body fat levels decrease, animals exhibiting migratory plasticity should modulate their risk of starvation by switching migratory tactics.We also assumed that decreasing body fat levels would require animals to be more plastic because at higher levels of body fat, they have a buffer against environmental conditions, but at lower levels of body fat, they lack that buffer and need to be more responsive to the immediate environment.Sierra bighorn face severe winter conditions at high elevations and higher predation risk at lower elevations.

Study area
Sierra bighorn live in 14 subpopulations in four recovery units (RUs; Central, Kern, Northern, Southern; i.e. metapopulations) throughout the Sierra Nevada Mountain Range (hereafter Sierra Nevada) in California.Bighorn sheep are highly gregarious, occurring in small bands that are sexually segregated for much of the year in groups of females and lambs (but sometimes include young males up to ∼2 years old) and bachelor groups of older males (Bleich et al., 1997;Schroeder et al., 2010).The eastern Sierra Nevada is in a rain shadow that supports xeric vegetation communities, including Great Basin sagebrush-bitterbrush scrub communities at low elevations (1500-2500 m); pinyon-juniper woodlands, subalpine meadows and subalpine forests at mid-elevations ( 2500 (Johnson et al., 2010).Summers in the Sierra Nevada are warm and dry; snow is the primary source of precipitation, with winter (Oct-Apr) snow depths ranging from 0 to 3.5 m and 0.4-14.9m in southern and northern portions of the study area during 2006-19 (Aspendell-Bishop and Lee Vining, CA, respectively; National Weather Service, accessed 25 December 2022).Strong winds (up to ≥240 kph) scour some alpine ridges, keeping them snow-free during winter (Spitz et al., 2020).During summer, Sierra Nevada bighorn sheep live in alpine habitats at moderate to high elevations (>3300 m).Some bighorn remain in alpine habitats at high elevations (sometimes >4000 m) during winter, but most live at more moderate elevations (1500-2700 m) (Spitz et al., 2018).Mountain lions (Puma concolor) are the most common predator of adult bighorn in this system, which also prey on mule deer and can result in apparent competition (United States Fish and Wildlife Service, 2007;Johnson et al., 2013;Gammons et al., 2021).

Animal capture, handling and monitoring
Animal research was conducted in accordance with US Fish and Wildlife Service Permit TE050122-6 and California Department of Fish and Wildlife Animal Welfare Policy (2018-02), and capture protocols (Sierra Nevada Bighorn Sheep Capture Plan 2006-10-2018-10) were approved annually.During 2006-18, we captured Sierra bighorn using a net gun fired from a helicopter.Animals were hobbled, blindfolded, bagged, slung from a helicopter and ferried to a staging area for processing and data collection.During handling of Sierra bighorn, we collected data on body mass and fat using manual palpation scores of body condition and ultrasound measurements of maximum depth of rump fat (Stephenson et al., 2020).We converted body condition scores and ultrasonic measurements of rump fat to estimate ingesta-free body fat (IFBFat) using equations validated for bighorn sheep (Stephenson et al., 2020) and then scaled these measurements following Denryter et al. (2022b).We also fit various models of GPS collars (various models manufactured by ATS, Isanti, MN; Lotek, Newmarket, Canada; North Star, Oakton, USA; Tellus, Lindesberg, Sweden; Telonics, Mesa, USA; Vectronic Aerospace, Berlin, Germany) programmed to collect 1-24 GPS locations per day on each Sierra bighorn.

Classification of migratory behaviours
We used migrateR (Spitz et al., 2017), to fit generalized linear models to elevation data in time series to classify migratory behaviours of Sierra bighorn.MigrateR extends net-squared displacement models often used for long-distance migration (Bunnefeld et al., 2011) to elevational migration (Spitz et al., 2017) and subsampled all data to the first location recorded each day for determination of migratory behaviours.We did not include data from animals in the first year following translocation to a new range because these animals lacked knowledge possessed by migrants (Jesmer et al., 2018) and therefore these movements would have been more exploratory in nature, rather than an intentional, directional migration.Following the methods of Denryter et al. (2021), we expanded the migratory classifications of Spitz et al. (2017) to classify migratory tactics as traditional migration, residency, abbreviated migration and vacillating migration.Traditional migrants were those individuals that migrated between seasonal ranges that were spatially separated and whose elevational-time series data showed a distinct drop in elevation, a continuous period of occupancy on the secondary range and a return to higher elevations after the period of occupancy.Residents were animals whose elevational-time series data were best described by a linear model because these individuals remained at approximately the same elevation year-round.Abbreviated migrants wintered at high elevations, but then made a 2-to 3-week movement to lower elevations in late spring when snow has subsided at lower elevations and green-up commenced.Vacillating migrants made multiple round-trip movements between high-and low-elevation ranges during winter.Final determinations of migratory tactics were based on Akaike Information Criterion corrected for small sample size (AICc) ranking (Anderson and Burnham, 2002;Burnham and Anderson, 2002) of migratory models and visual inspections of plots (Mysterud et al., 2011;Spitz et al., 2017;Denryter et al., 2021).We also estimated linear distance moved between high-and low-elevation ranges for migratory animals using migrateR and used this information in our analyses.

Statistical analyses
All statistical analyses were conducted in Stata 14 (StataCorp, College Station, TX), except for Chi-squared equality of proportions tests that we conducted in R (R Core Team, 2023).To evaluate how autumn body fat influenced migration, we used analysis of variance, multinomial logistic regression and linear regression.First, we used a Chi-squared equality of proportions test to evaluate whether composition of migratory tactics differed amongst populations (i.e. 14 individual herds of Sierra bighorn).We used RU, which essentially represents metapopulations whose habitats share similar characteristics, rather than herd in analyses with body fat owing to small sample sizes within populations.RU was used as a surrogate of availability of high-elevation and low-elevation winter ranges to account for the influence of the availability of different types of range on migratory propensity (Spitz et al., 2020).High-and low-elevation habitats differ amongst RUs (along a north-south gradient) and have different risks in terms of nutrition, snow and predation that could influence migratory decisions.Additionally, reintroduction timelines differ across RUs, which could affect migratory propensity given the length of time required for migration to become reestablished following reintroduction (Jesmer et al., 2018).
We used analysis of variance to evaluate differences in autumn body fat across migratory tactics (i.dents in terms of winter risk], traditional migrant, vacillating migrant).We used multinomial logistic regression to evaluate main and interaction effects of autumn body fat and RU on probability of migrating (any migratory tactic) versus residency; we grouped residents and abbreviated migrants into one category owing to small sample sizes.The Kern RU was excluded from analyses incorporating RU as a covariate owing to small sample sizes for body fat (n = 1).We used linear regression to evaluate the influence of autumn body fat on distance migrated, including RU as a covariate and evaluated potential interaction terms.We also used linear regression to examine whether switching rates were a function of mean population-level body fat across all years of data because inyear sample sizes were too small.To determine the most supported model amongst candidate models in regression analyses we used an information-theoretic approach and ranked models using Akaike's Information Criterion corrected for small sample size and examined models for uninformative parameters (Anderson and Burnham, 2002;Burnham and Anderson, 2002;Arnold, 2010).Where we did not have a priori competing models to compare, we used frequentist statistics rather than an information-theoretic approach.For all tests of significance, we used α = 0.10, rather than α = 0.05, owing to small samples sizes and our greater tolerance for committing a Type II error than a Type I error.Our small sample sizes reduced statistical power for our analyses and therefore increased the likelihood that we would fail to reject a false null hypothesis and miss important biological effects based on an arbitrary α.Traditional use of α = 0.05 has been noted to be negatively impacted by small sample size and use of α > 0.05 has received increasing acceptance for complex relationships or small sample size as has α < 0.05 for large sample sizes, partly in efforts to avoid Lindley's paradox (Thiese et al., 2016;Maier and Lakens, 2022).

Results
We classified migratory behaviours for 240 animal-years for 129 individual male Sierra bighorn.We classified migratory behaviour for 60 animals in 1 year, 40 animals in 2 years, 19 animals in 3 years, seven animals in 4 years and for three animals in 5 years.Migration was the predominant behavioural state as 171 animal-years were classified as migrants, which included traditional migrants (n = 101) and vacillating migrants (n = 70).Residency was documented for 69 animal-years, which included n = 48 instances of residency alone and n = 21 instances of residency with abbreviated migration.Traditional migration, residency and abbreviated migration occurred in all RUs, but vacillating migration was not documented in the Kern RU.Proportions of traditional migrants, vacillating migrants, residents and abbreviated migrants differed amongst the four RUs (χ 2 [9] = 73.12,P < 0.001; Fig. 1).
On average, resident males had the greatest body fat levels in autumn, followed by vacillating migrants, tradi-tional migrants and abbreviated migrants, but differences were not significant in analysis of variance (P = 0.860, df = 3, 61, F = 0.25) (Fig. 2A).Nevertheless, by RU, residents tended to have greater autumn IFBFat levels than migrants, except in the Southern RU (Fig. 2B).The top model for probability of migration included only an interaction term for body fat and RU (Table 2, Fig. 3).Probability of migration dropped off precipitously for animals in the Central and Northern RUs when autumn IFBFat ≥10% (Fig. 3).For every 1-percentage point increase in body fat, there was a 16% decrease in the likelihood that a Sierra bighorn in the Central RU would be a vacillating migrant, compared with a resident and a 21% decrease for Sierra bighorn in the Northern RU (Fig. 3, Table 2).Similarly, for every 1-percentage point increase in body fat there was a 17% decrease in the likelihood that an animal would be a traditional migrant in the Central RU and a 25% decrease in the Northern RU (Fig. 3, Table 2).There was no relationship between body fat and probability of migration for animals in the Southern RU (Table 4).
Sierra bighorn in the Central and Southern RUs moved further down in elevation than in the Northern RU (Table 3, Fig. 4).Distances separating high-and low-elevation ranges averaged 1176 ± 176 m for Sierra bighorn in the Central RU, 331 ± 182 m in the Northern RU and 1087 ± 91 in the Southern RU (marginal mean ± SE).Body fat influenced fine-scale migratory behaviours.Amongst vacillating and traditional migrants, body fat influenced the distance between migratory and resident ranges (Fig. 4, Table 3).For every 1-percentage point increase in autumn body fat, the difference between high-and low-elevation ranges decreased by ∼51 m, on average, for Sierra bighorn and differed amongst populations (Fig. 4, Table 3).
Male Sierra bighorn switched migratory tactics between years 60 times out of 111 opportunities to switch, for a switching rate of 54%.Males switched from migratory to residency behaviours n = 19 times (32% of switches) and from residency to migratory behaviours n = 7 times (12%).The remaining switches were from one migratory tactic to another or one residency tactic to another: vacillating migrant to traditional migrant (n = 13; 22%); traditional migrant to vacillating migrant (n = 9; 16%); abbreviated migrant to resident (n = 9; 15%); resident to abbreviated migrant (n = 3; 5%).Amongst 30 individuals tracked over ≥3 years, n = 14 switched tactics more than once, including n = 13 individuals that switched tactics twice, two individuals that switched tactics three times and one individual that switched tactics four times.Additionally, body fat influenced switching rates at the population level (P = 0.012, R 2 = 0.521, adj.R 2 = 0.468, n = 11; Table 4), with switching rate decreasing 7 percentage points for every 1 percentage point increase in IFBFat (Fig. 5).

Discussion
Body fat is an important adaptation to seasonal environments through its direct influence on reproduction and survival   ( Cook et al., 2013;Monteith et al., 2013;Stephenson et al., 2020;Denryter et al., 2022a).We hypothesized that the importance of body fat to animals living in seasonal environments would extend beyond direct effects by influencing migratory behaviours in a risk-sensitive manner.In our study area, residents remain at high elevations year-round, which means they experience harsh winter conditions and rely more on body fat for survival than migrants that winter at low elevations (Denryter et al., 2022a); migrants however incur increased predation risk (Johnson et al., 2010;Gammons et al., 2021).Because of the contrasting risks associated with choice of migratory tactic and migration distance, we predicted that probability of migration would be inversely related to body fat as would migration distance.decreasing body fat.Rates of switching migratory tactics across years increased as body fat decreased, reiterating that body fat acts as a buffer and ameliorates the needs of animals to be wholly responsive to their immediate environment.Collectively, our findings support the hypothesis that migration was risk-sensitive to body fat.
Our study adds to a growing body of work that relates body fat with migratory behaviours, including timing or initiation of migration (Monteith et al., 2011;Shry et al., 2019), as a fuel for migration (Pettersson and Hasselquist, 1985) or as a cue for departing stopover sites (Goymann et al., 2010).In our study, migratory behaviours were a risk-sensitive function of body fat, particularly where the relative value of body fat as a nutritional buffer varies across behavioural states (Denryter et al., 2022a).For example, animals with smaller body fat stores were more likely to migrate to low-elevation winter ranges likely because starvation is imminent at high elevations during winter without adequate fat reserves (Denryter et al., 2022a).High levels of body fat are a requisite nutritional buffer for high-elevation residents to persist in environments characterized by extreme winters (United States Fish and Wildlife Service, 2007;Spitz, 2015;Conner et al., 2018) that limit food availability and increase energy demands for locomotion and thermoregulation (Chappel and Hudson, 1978;Fancy and White, 1987)   residents require almost twice as much body fat as migrants to achieve >90% overwinter survival (Denryter et al., 2022a).Smaller body fat stores of migrants likely precluded them from pursuing residency because they lacked an adequate nutritional buffer against harsh winter conditions at high elevations.Instead, small fat stores may have forced animals to seek milder conditions at lower elevations where they would have more predictable access to food, but at the cost of predation risk (Johnson et al., 2010;Spitz, 2015;Gammons et al., 2021).Thus, migrants chose a migratory tactic that was risk-averse to starvation, but risk-prone to predation, in contrast with residents that were risk-averse to predation and risk-prone to starvation.
The elevational distance an animal migrated also was risk-sensitive to body fat.As expected, animals with less body fat migrated further down in elevation than animals with more body fat, presumably because as their nutritional buffer declined animals would have to move further down in  elevation to escape snow and access food supplies.Choosing migration distance relative to body fat stores likely increased the relative value of the nutritional buffer provided by their fat stores; for example a given amount of body fat should last longer in an environment with little snow, where food is more abundant and energy expenditures are lower, compared to one where snow limits food availability and increases locomotory costs (Denryter et al., 2022a).Choosing a migration distance relative to body fat stores likely enabled Sierra bighorn to attune their risk relative to the buffer provided by different levels of body fat without exposing themselves to unnecessary risk of predation.That is, Sierra bighorn could have chosen to migrate further to increase access to food, but they did not need to of the nutritional buffer afforded by their body fat stores.Thus, attuning migration distance to body fat stores may have allowed Sierra bighorn to balance trade-offs between starvation and predation.
Our study focused exclusively on migratory behaviours of males, which was atypical for studies of ungulate migration (Supplementary Information 1; Supplementary Table S1).Focusing on males allowed us to explore relations between physiological state and migration without the potentially confounding influence of having a lamb at heel (Festa-Bianchet, 1988) and allowed us to examine potential differences in migratory behaviour between sexes that have different annual physiological cycles.Elevational distance migrated by males did not differ appreciably from those reported for females (Spitz et al., 2018), but males switched tactics more frequently (Spitz et al., 2018).Higher switching rates of males versus females may reflect greater behavioural flexibility given their lack of parental investment in offspring and have been reported in other partially migratory ungulates (Peters et al., 2019).Whereas males may be able to choose a migratory tactic based solely on their needs, females with a lamb at heel must also consider the needs of the lamb.Although most female Sierra bighorn do not lactate over winter, they retain strong bonds with their offspring until they prepare to give birth to subsequent offspring in May, and given that the first year of life is one of greatest vulnerability in ungulates (Gaillard et al., 2000), mothers should enhance their fitness by supporting offspring through the first year.Keeping lambs away from predators by remaining at high elevations can reduce predation risk to lambs, but lambs invest in growth rather than accretion of body fat in their first summer and may be more susceptible to mortality from starvation if wintering at higher elevations.Thus, females may be more inclined to migrate regardless of body fat if winter conditions on high-elevation ranges increase risk of mortality for lambs.However, females below a body fat threshold of ∼14% may be forced to migrate (Denryter et al., 2022a) to survive, which also may contribute to their lower migratory plasticity compared to males.Greater migratory plasticity of males versus females may be an important consideration for re-establishing desired migrations.
In addition to individual-level effects of body fat on migratory behaviours, body fat also was related to migratory plasticity (i.e.switching rates).At the population level, switching became more common with decreasing body fat levels, in accordance with our hypothesis that body fat underpins facultative migration-an observation consistent in birds (Cornelius et al., 2021).In the Sierra Nevada, decreasing body fat levels are associated with lower levels of survival, especially for residents (Denryter et al., 2022a).Thus, as body fat levels decrease, animals exhibiting migratory plasticity should modulate their risk of starvation by switching migratory tactics.At the population level, fluctuating levels of body fat also could lead to changes in the composition of migratory tactics within the population.Not all taxa exhibit high levels of migratory plasticity and in less plastic species, such as mule deer (Odocoileus hemionus hemionus; Sawyer et al., 2018), the inability to be flexible may mean that they cannot modulate the value of body fat through migration and therefore must achieve greater levels of body fat prior to winter.Additionally, greater variation in body fat levels in less plastic species may result in more variable survival given the relationship between body fat and survival (Denryter et al., 2022a).In the absence of additional data, we caution interpretation of this finding until additional data can substantiate (or refute) it.Nevertheless, population-level effects of body fat and migratory tactic may be imminent, especially given that the nutritional buffer provided by body fat can buffer vital rates of populations from severe weather (Jesmer et al., 2021).Differences in body fat across migratory tactics andpopulations also may help explain why vital rates of migrants and residents often differ (Nicholson et al., 1997;Middleton et al., 2013;White et al., 2014;Hebblewhite et al., 2018;Denryter et al., 2022a).
Although our results consistently supported our hypotheses, several important caveats are present.Our sample size for males was relatively large for published studies of ungulate migration (Supplementary Information 1;  S1) and body fat, but small within RU and for switching rates for periods >2 years.Because of small sample sizes, we relaxed confidence levels and accepted a higher risk of a Type II error than a Type I error.Despite these limitations, consistency in our results across migratory behaviours at individual and population levels strongly support our hypothesis that migratory behaviours were risk-sensitive to body fat.Further, our findings can be explained mechanistically, which also bolsters our confidence that migration was risk-sensitive to body fat, with residents being risk-prone to starvation and migrants being risk-averse.

Figure 3 :
Figure 3: Relationship between autumn infesta-free body fat percentage (IFBFat (%)) in autumn of male Sierra Nevada bighorn sheep and predicted probability of migratory tactics for animals in the Central (A; n = 13), Northern (B; n = 17) and Southern (C; n = 35) RUs in California, USA, from 2006 to 2019.

Figure 5 :
Figure 5: Relationship between population-level IFBFat (%) in autumn of male Sierra Nevada bighorn sheep and switching rate (shown as predicted values (solid line) with 95% confidence intervals as dashed lines) where each circle represents one population in California, USA, from 2006 to 2019.

Table 1 :
(Spitz et al., 2017) models characterizing probability of male Sierra Nevada bighorn sheep choosing a migratory tactic (residency, traditional migration or vacillating migration; determined from classification of movement data in migrateR(Spitz et al., 2017)); and linear regression models characterizing the relationship between distance separating high-and low-elevation seasonal ranges of migratory male Sierra Nevada bighorn sheep relative to ingesta-free body fat in autumn (AutumnIFBFat) and Recovery Unit (RUCode; categorical variable with 3 levels (Central, Northern, Southern); Kern was excluded due to lack of Autumn IFBFat samples) in California, USA, from 2006 to 2019.K is the number of parameters, n is sample size, LL is log likelihod, AICc is Akaike Information Criterion corrected for small sample size, ΔAICc is the change in AICc from the top model, and ωi is the Akaike model weight

Table 2 :
Results from top multinomial logistic models, as indicated by Akaike Information Criteria corrected for small sample size (Table1), characterizing probability of male Sierra Nevada bighorn sheep choosing a migratory tactic (traditional migration or vacillating migration) relative to residency as a function of AutumnIFBFat and RU (RUCode; categorical variable with 3 levels (Central, Northern, Southern); Kern excluded due to lack of Autumn IFBFat samples) in California, USA, from 2006 to 2018

Table 3 :
Results from top linear regression models, as indicated by Akaike Information Criteria corrected for small sample size (Table1), characterizing the relationship between distance separating high-and low-elevation seasonal ranges of migratory Sierra Nevada bighorn sheep relative to ingesta-free body fat percentage in autumn (autumnIFBFat) and recovery unit (RU; categorical variable with 3 levels (Central, Northern, Southern); Kern excluded due to lack of Autumn IFBFat samples) in California, USA, from 2006 to 2019

Table 4 :
Results from a linear regression model characterizing the relationship between population-level ingesta-free body fat (IFBFat) in autumn and population-level switching rates for n = 11 populations of Sierra Nevada bighorn sheep in California, USA, from 2006 to 2019