Effects of Anthropogenic and Climate-Induced Habitat Changes on Adult Female Mule Deer Survival

ABSTRACT Increased wildfire size and frequency, invasion of non-native annual grasses, extensive fire suppression, climate change, and widespread juniper expansion have led to dramatic changes in sagebrush ecosystems, which provide habitat for a suite of wildlife species, including mule deer (Odocoileus hemionus). These landscape-level disturbances can cause habitat loss and degradation, which influence the quantity and quality of available forage and can negatively influence adult female mule deer survival. We used Program MARK to estimate survival rates using a known-fate model. Our dataset included 111 radio-collared female mule deer to estimate monthly survival rates and to investigate factors that may affect these rates, including movement behavior (migratory or resident) and habitat characteristics (wildfire and juniper canopy cover) in mule deer habitats supporting sagebrush potential vegetation types (PVTs). The dataset included 116 421 Global Positioning System (GPS) location points from 2015 to 2017 in the John Day Basin, Oregon, United States. Survival averaged 0.78 (95% confidence interval = 0.61–0.82) across the study period but was 0.69 (95% confidence interval = 0.59–0.77) in 2015. Our results indicated survival was positively influenced by the presence of small amounts of recent fire (< 15% of total winter range burned; 1–20 yr postfire) within an individual's winter range. Annual survival (0.78) was low compared with survival rates for adult female mule deer in other parts of their geographic range. These findings demonstrate how postfire vegetation response can have a positive effect on mule deer survival, regardless of their migration strategy. Our findings also indicate how climate change poses a growing threat to mule deer populations as prolonged periods of drought increase the spread of fatal diseases.


Introduction
Mule deer populations began declining in Oregon in 1981 from 30 0 0 0 0 individuals to 164 0 0 0 in 2022 ( Oregon Department Fish and Wildlife 2021 ).They have also declined throughout their geographic range (4.6 million in 20 0 0 to 3.7 million in 2021) ( White and Bartmann 1997 ;Robinson et al. 2002 ;Bender et al. 2007 ;Bishop et al. 2009 ).There may be multiple or interacting causes of this decline, including habitat degradation, expansion of predator populations, and changes to land management practices ( Ballard et al. 2001 ;Peek et al. 2002 ;Sawyer et al. 2017 ).Habitat quality has declined across much of the species' range due to increases in invasive grasses ( Bishop et al. 2001 ), changes in fire regimes ( Miller and Heyerdahl 2008 ), and land use change ( Stewart et al. 2002 ;Sawyer et al. 2006 ).Inadequate nutrition is often a limiting factor for mule deer population growth ( Bishop et al. 2009 ).Changes to habitat can modify the nutritional carrying capacity, destabilizing mule deer population dynamics.For example, Forrester and Wittmer (2013) found that the ultimate cause of mortality for most mule deer populations was poor nutrition.In Texas, malnutrition was the leading cause of mortality for adult mule deer ( Lawrence et al. 2004 ) and starvation was the major cause of fawn mortality in Colorado ( White et al. 1987 ).Forage quality will decrease in areas where invasive grasses (i.e., cheatgrass [Bromus tectorum], medusahead [Taeniatherum caput-medusae] ) have increased, which allows for lower nutrient availability for mule deer populations ( Schuyler et al. 2021 ).
Ungulate populations respond to resource limitations though a predictable series of population-level responses.Low juvenile fecundity can be an early indicator of an unstable population.This is often followed by low juvenile survival, and the last demographic to be affected is a reduction in adult female survival ( Gaillard et al. 1998 ).Mule deer experience lower and more variable fawn survival rates compared with other ungulates ( Forrester and Wittmer 2013 ), and mule deer populations rely more on high, stable adult survival rates than other ungulates to prevent long-term population declines ( Forrester and Wittmer 2013 ).Consequently, small changes in adult female survival can have major impacts on the overall population growth rate ( Morris and Doak 2002 ).Fluxes in adult survival rates are fundamentally related to individual nutritional condition ( Bender et al. 2008 ), which can be influenced seasonally ( Nicholson et al. 1997 ;Gaillard et al. 1998 ;Unsworth et al. 1999 ;DelGiudice et al. 2002 ;Brown et al. 2006 ) and by changes in pressures from predator populations, harvest, or poaching ( Ballard et al. 2001 ;Mulligan 2015 ).Determining the impact that these effects have on seasonal and annual adult survival can illuminate causes of population-level declines.
Many mule deer populations are migratory or exhibit partial migration in which some individuals migrate while others remain in the same area year-round ( Eberhardt et al. 1984 ;Garrott et al. 1987 ;Hebblewhite and Merrill 2009 ).In the Intermountain West, migratory deer spend their winters in low-elevation sagebrush and other arid and semiarid plant communities and then move to higher elevations to follow seasonal forage and water availability ( Nicholson et al. 1997 ;D'Eon and Serrouya 2005 ).However, migration is a trade-off between moving to higher-quality forage ( Pierce et al. 2004 ;Hebblewhite and Merrill 2009 ;Schuyler et al. 2021 ) and increased risk of predation or anthropogenic mortality while moving through a less familiar landscape ( White et al. 1987 ;Nicholson et al. 1997 ).The difference in survival rates between migratory deer and resident deer ( Nicholson et al. 1997 ;Hebblewhite and Merrill 2009 ;Gogan et al. 2019 ;Schuyler et al. 2019 ) provides evidence of this trade-off.However, for a population of mixed movement strategies to exist, there must be interannual variation in survival in which one strategy, migration, improves individual fitness some years, whereas the other strategy, being a resident, improves individual fitness in other years ( Hebblewhite and Merrill 2009 ).For example, annual survival among migratory deer in Oregon was higher than that of resident deer due to better quality forage on migratory summer ranges ( Schuyler et al. 2019 ).Conversely, a similar study in Yellowstone found no difference in survival rates between migratory and resident deer ( Gogan et al. 2019 ), but agricultural land uses at lower elevations during the summer likely benefited resident deer diets.
Potential vegetation type (PVT; Burcsu et al. 2014 ) classifications have been used extensively in connecting the effects of wildfire to the underlying habitat conditions across the western United States ( Ager et al. 2017 ;Charnley et al. 2017 ;Spies et al. 2017 ;Defrees et al. 2020 ;Povak et al. 2020 ).PVTs do not represent existing vegetation but instead characterize the underlying ecological site potential of an area and provide context for how native and nonnative vegetation communities respond to disturbances, like wildfire ( Chambers et al. 2014( Chambers et al. , 2017( Chambers et al. , 2019 ) ). Mule deer winter ranges in the Blue Mountain Ecoregion largely consist of Wyoming big sagebrush ( Artemisia tridentata ssp.wyomingensis ) PVTs.These communities have experienced dramatic landscape-scale changes in a relatively short ( < 150 yr) temporal period due to fire suppression ( Keeley et al. 1999 ;Baker 2006 ), conifer expansion ( Miller 2005 ;Coultrap et al. 2008 ), and the invasion of exotic grasses ( Pyke 1999 ;Davies and Johnson 2008 ).Wyoming big sagebrush PVTs are highly vulnerable to conversion to cheatgrass, when burned and are not likely to recover to a functional native composition post fire ( Pyke et al. 2022 ).Key shrubs used as winter forage, such as bitterbrush (Purshia tridentata) and big sagebrush (Artemisia tridentata), often become functionally extirpated, and the loss of these shrubs can reduce winter survival and ultimately reduce carrying capacity of winter ranges (Hobbs and Swift 1985;Bates et al. 2020 ;Ellsworth et al. 2020 ).Alternatively, when Wyoming big sagebrush plant communities experience patchy, low-severity fires and cheatgrass is not present, plant growth of perennial bunchgrasses and forbs can be invigorated due to the cycling of nutrients that can be beneficial for mule deer forage ( Green et al. 2012 ;Ellsworth et al. 2016 ).
Fire suppression and changes in land management practices have also led to the expansion of Western juniper (Juniperus occidentalis) into shrublands ( Miller 2005 ).Juniper encroachment has added to the loss of valuable shrub forage as increased tree cover outcompetes shrubs and herbaceous plants for light, water, and nutrients ( Miller 2005 ).In Colorado, an increase in juniper cover negatively influenced fawn mule deer survival ( Bergman et al. 2014 ).Across different areas of their range, mule deer are more likely to use juniper treatment areas than reference sites ( Bender et al. 2013 ;Morano et al. 2019 ;Platte and Torland 2023 ).Therefore, assessing juniper canopy cover in low-elevation shrublands is one metric of habitat quality that ultimately influences mule deer survival.
Migratory mule deer summer ranges within the Blue Mountain Ecoregion are found at higher elevations within Ponderosa pine (Pinus ponderosa) PVTs and mixed conifer forest PVTs (hereafter, "dry forests") and respond much differently to wildfire.Dry forests contain a number of fire-tolerant plant species and fires cycle nutrients from foliage to soil, which promotes plant growth ( Hessburg et al. 2005 ).This, in turn, has a positive influence on mule deer summer nutrition by increasing palatable shrub density and cover ( Roerick et al. 2019 ).Similarly, in the early postfire years, burned areas may improve summer forage due to the increased light availability and nutrients that facilitate growth of forage biomass ( Mackie et al. 2003 ;Greene et al. 2012 ).The increased forage quality and quantity ( Schuyler et al. 2021 ) can have a positive effect on individual fitness and survival ( Hobbs and Spowart 1984 ;Walter et al. 2009 ).
Our objective was to determine how spatial occurrences of wildfire, migration behavior, and vegetation characteristics influenced mule deer survival.We focused on sagebrush PVTs on winter range and dry forest PVTs on summer range.We predicted that 1) seasonal survival would be lowest during the fall (1 September −30 November), as individuals expend energy in migration and recover energy expended during lactation but are unable to find adequate forage late in the summer ( Pettorelli et al. 2011 ;Hurley et al. 2014 ;Schuyler et al. 2019 ; Table 1 , P1 ); 2) annual survival would be higher for migratory deer than resident deer due to access to higher forage quality and quantity available in dry forest PVTs compared with Wyoming big sagebrush PVTs ( Bender et al. 2007 ;Schuyler et al. 2019 ;Schuyler et al. 2021 ;P2 ); 3) survival would be positively influenced by burned areas within an individual's summer range (migratory deer, dry forest PVTs; resident deer, Wyoming big sagebrush PVTs) due to an increase in forage quality and quantity ( Proffitt et al. 2019, Roerick et al. 2019, Schuyler 2020 ; P3 ); 4) survival would be negatively influenced by recently burned areas within the individual's winter range in Wyoming big sagebrush PVTs due to a decrease in shrub cover ( Green et al. 2012 ;Schuyler 2020 ; P4 ); and 5) survival would be negatively influenced by an increase in western juniper canopy cover on an individual's home range due a decrease in shrubs in the sagebrush PVTs with juniper expansion ( Bergman et al. 2014 ;P5 ).
The climate across the study area had dry warm summers (average maximum August temperatures of 31 −33 0 C) and cold winters (average minimum temperatures of −2 to 0.5 0 C) (PRISM Climate Group 2019).Elevation ranged from 2 756 m at the peak of the Strawberry Mountain Range to 616 m along the John Day River.Annual precipitation averaged 30 cm at low elevations and 78 cm at high elevations over the past 30 yr (PRISM Climate Group 2019).Most of the study area occurred on private land (72%) with federal land administered by the Bureau of Land Management and US Forest Service (21%) and state land (6%) dispersed throughout.In 2014, a lightning strike ignited the South Fork Complex Fire, which burned nearly 10 117 ha of the Phillip W. Schneider Wildlife Area (PWSWA).PWSWA is managed for wintering mule deer populations, as well as other wildlife species.In addition to mule deer, the study area supported populations of elk (Cervus canadensis nelsoni), pronghorn (Antilocapra americana), feral horses (Equus ferus caballus), and predators including cougar (Puma concolor), black bear (Ursus americanus), and coyote (Canis latrans).

Capture and handling
The Oregon Department of Fish and Wildlife captured and fitted adult ( ≥ 1.75-yr-old) female mule deer with GPS collars (4400S, LOTEK Engineering Ltd., Newmarket, ON, Canada) between March 2015 and August 2017.The collars had a remote download function and were programmed to record GPS locations every 13 h.The collars had a 97% success fix rate (i.e., number of locations success-fully acquired divided by the number of locations attempted to be acquired) and < 10 m accuracy ( Forin-Wiart et al. 2015 ).Deer were captured using helicopter net guns ( Krausman et al. 1985 ), and personnel followed the Guidelines for the Capture, Handling, and Care of Mammals as approved by the American Society of Mammalogists ( Gannon and Sikes 2011 ).Collars remained on the animals, and data were remotely collected until the animal died, the battery was exhausted, the collar was lost, or the collar malfunctioned.Agency personnel were notified via email if a collar was motionless > 12 h (i.e. , mortality signal).
Research personnel captured and collared deer east of the Cascades in Oregon to delineate herd ranges and measure annual survival for the mule deer population at the state level.From the statewide collaring effort, we used a subsample ( n = 111) of deer that were captured within the John Day Upland ecoregion (Level IV; Commission for Environmental Cooperation 1997 ) to estimate survival for a population of deer that shared the same winter range.This subsample of deer was captured during March 2015.

Explanatory variables for survival
We estimated monthly survival rates for adult female mule deer from March 2015 through August 2017.We estimated yearly survival by biological year (1 June-31 May) ( Unsworth et al. 1999 ).We defined four seasons that reflected energetic demands of adult females: summer (1 Jun-30 Aug; fawning), fall (1 Sep-30 Nov; migration), winter (1 Dec-28 Feb; maintaining energy expenditure), and spring (1 Mar-31 May; migration) ( Wallmo 1981 ).We considered an individual migratory if its seasonal ranges (summer and winter) did not overlap, and the edge of the seasonal ranges was > 10 km ( Brown 1992 ) ( Fig. 1 ).Deer that had both winter and summer ranges in the same location were considered resident.
There was a range of 141 −1 378 GPS location points per individual.For every GPS location, we extracted a suite of environmental variables including wildfire occurrence (summer and winter range locations) and western juniper canopy cover (winter range locations) ( DeCesare et al. 2014 ).Wildfire spatial data included all fires > 405 ha from 1995 to 2016 using data from Monitoring Trends in Burn Severity (MTBS; Eidenshink et al. 2007 ).Individual winter and summer use were estimated by comparing the number of locations in burned areas to the total locations and were treated as continuous variables.
We mapped and estimated the percent area of each PVT present in winter and summer ranges ( Burcsu et al. 2014 ).Knowledge of the underlying PVTs provides critical context for the welldocumented effects of burning on native and non-native vegetation.PVTs can also provide insight on community responses in terms of community resilience, resistance, associated probability of recovery, and time to recovery in sagebrush, juniper, and dry-forest PVTs that composed our study area (Hasseburg et al. 2005 ;Chambers et al. 2014Chambers et al. , 2017Chambers et al. , 2019 ; ;Downing et al. 2019 ).
Western juniper canopy cover was derived from remotely sensed LIDAR and aerial imagery data sourced from the Institute of Natural Resources.Juniper canopy cover was treated as a continuous covariate in survival models.

Survival analysis
We generated monthly survival estimates for adult female mule deer using known-fate models, with estimates and model selection statistics generated in Program MARK ( White and Burnham 1999 ).Annual survival rates were estimated by multiplying the monthly survival rate by 12 mo ( White and Burnham 1999 ).This approach allowed the incorporation of temporal variation and individualspecific covariates within a standardized model selection and multimodel inference framework ( Murray 2006 ).All deer entered the analysis at the same time based on their capture month and year (March 2015).
We used a tiered approach to evaluate covariates and develop competing models and used corrected Akaike's Information Criterion (AIC c ) to select the most supported models in all tiers.This modeling approach results in the same model selection outcome as an all-possible-combinations modeling approach but includes fewer models with uninformative parameters ( Arnold 2010 ;Doherty et al. 2012 ;Morin et al. 2020 ).In the first tier, we fit models that tested temporal effects on survival (S) including biological year and season (e.g., S [Spring], S [Fall], S [Yr]).In the second tier, we fit univariate models for movement behavior on survival including two groups: migratory and resident.In the third tier, we fit models that evaluated fire effects for the summer and winter range and individual winter range western juniper canopy cover ( Table 1 ).In the final tier, we combined the highest-ranking covariates (lowest AIC c ) from the previous tiers (1 −3) together to com- We considered models within 2 AIC c to be competitive with the top model ( Burnham and Anderson 2002 ).We assessed correlations between all model covariates using a Spearman rank-order analysis ( Zar 1999 ).Highly correlated variables (|r| ≥ 0.60) were not included in the same model; however, we did not have correlations meeting this threshold.We also included the most general model that analyzed separate estimates of survival for each month and year ( S [Yr * Mon]) and included a model with no effects for comparison at all tiers ( S [.]).

Results
Over the 30-mo study (Mar 2015 to Aug 2017), 111 adult female mule deer had adequate location data for inclusion in annual survival models (minimum 1 mo and maximum 30 mo of location data, average duration 20 mo).During the study, 17 deer died from predation (30% summer, 30% fall, 30% winter, 10% spring), 3 died of disease or injury, 2 died from vehicle collisions, 2 died from illegal harvest, and 20 had unknown causes of death (45% winter, 30% fall, 15% summer, 10% spring).Sixty-seven deer remained in the study until there was loss of signal or collar failure.
The total area of the winter range was 203 974 ha and was predominantly composed of Wyoming big sagebrush (35%) and western juniper (25%) ( Table 2 ).The total area of the migratory summer range was 559 256 ha and was mostly compromised of mixed conifer forests PVT (67%) followed by Ponderosa pine (21%) (see Table 2 ).The total area of the resident summer range was 169 825 ha, and most of the percent area of PVTs for the resident summer range were Wyoming big sagebrush (35%) and western juniper (24%) (see Table 2 ).
During our first stage of modeling, we found the following parameters to be uninformative (i.e., betas overlapped zero): temporal effect month, movement behavior, summer range fire effect, juniper cover.Of the 20 candidate models ( Table 3 ), only one model emerged as a top model because no other models were within 2 AIC.Adult female survival in this study was best predicted by year and the percent of an individual's winter range that burned.Survival was lower during 2015 compared with the rest of the study ( ˆ β year2015 = −0.88,95% confidence interval [CI] = −1.58 to −0.24).We calculated the annual survival by the product of monthly survival rates for the biological year (June-May), which was 0.69 (95% CI = 0.59-0.77) in 2015 and 0.86 (95% CI = 0.77-0.91) in 2016.Annual survival rate, as an average over 2.5 yr, was 0.78 (95% CI = 0.61-0.82).Survival was also positively influenced by burned areas within the individual winter range ( ˆ β Firewinter = 0.11, 95% CI = 0.01 to 0.21; Fig. 2 ).

Table 3
Model selection results for the top 10 of 20 a priori models investigating survival probability (S) of radio-collared mule deer in the Blue Mountains, Oregon, United States, 2015-2017, relative to time effects, wildfire covariates, and vegetation covariates.Models are ranked according to Akaike's Information Criterion adjusted for small sample sizes (AIC c ). Difference in AIC c ( AIC c ), Akaike weight ( w i ), number of parameters ( K ), and deviance are also listed for each model.Model set includes the intercept-only (null) model of constant survival over time, S (.), and the most general model with survival variation by mo and yr, S (Yr • Mon).Letter abbreviations for percent of individual's winter range burned (Fire winter ), percent juniper canopy cover on winter range (JUOC).

Discussion
The average probability of adult female annual survival for our study was 0.78, which is lower than survival rates reported for adult female mule deer in other individual populations (0.86, Bleich et al. 2006 ;0.81, Bender et al. 2007 ;0.91, Bishop et al. 2009 ;0.89, Monteith et al. 2010 ;0.89 Hurley et al. 2011 ;0.81, Gogan et al. 2019 ), including the weighted mean annual survival rate from a meta-analysis that estimated adult female survival over the past 30 yr across the geographic range (0.84, CV = 0.06; Forrester and Wittmer 2013 ).The age class of the deer in this study were considered prime-aged (2 −7 yr), which have the highest annual survival when compared with yearling (1 −2 yr) or senescent individuals ( > 7 yr) ( Marescot et al. 2015 ).The population in this study has been steadily decreasing for decades ( Oregon Department Fish and Wildlife 2021 ) and the low survival rate for prime-aged females provides an estimate of the magnitude and potential source of the population's current rate of decline ( Bishop et al. 2005 ).Low adult female survival rates are attributed to poor body condition, which is likely a result of limited food and long-term changes in plant communities (Bender et al. 2010).During the time of this study (2014 −2015), the study area had experienced an extreme drought ( US Drought Monitor ;Svobada et al. 2002 ).In a waterlimited ecosystem, drought can severely reduce food resources for mule deer which will lower adult survival and overall herd productivity causing population declines ( Bender et al. 2011 ;Woods et al. 2018 ).
Even during nondrought years, forage availability plays a big role in adult female survival ( Forrester and Whittmer, 2013 ).Forage availability at lower elevations is relatively scarce in sagebrush steppe ecosystems as woody plants are sparse in these systems (Nicholson et al. 2006;O'Connor et al. 2020 ).Mule deer rely heavily on shrub species at lower elevations, particularly antelope bitterbrush (Burrell 1982;Hobbs 1989) due to the elevated nutritional content of the current yr's growth (Dietz et al. 1976).Changes in diet, particularly loss of valuable nutritious forage, can adversely affect annual mule deer survival (Hobbs 1989).Shrubs provide not only vital forage throughout the year but also cover.Shrub cover is a critical component to mule deer survival because it influences microclimate, predation risk, and snow depth (Mysterud and Ostbye 1999).The loss of shrubs on the winter range can make deer more susceptible to predation and thermal protection from wind and precipitation, ultimately lowering their survival (Reeve and Lindzey 1991).The negative effect of the loss of shrubs species from juniper encroachment may be compounded when the landscape is already degraded, as is the case for this study area, which has a history of overgrazing and invasion of annual grasses.Eventually, this lowers the capacity of the range to support mule deer populations and ultimately decreases adult female survival.
Nutritionally stressed deer are also more likely to suffer from parasites, disease, and predation ( Bender et al. 2007 ;Bender et al. 2011 ).Outbreaks of viral hemorrhagic diseases are some of the most common and significant to mule deer health ( Heffelfinger and Krausman 2023 ).We found that mule deer had particularly lower survival in 2015, which coincided with an Adenovirus Hemorrhagic Disease (AHD) outbreak across large areas of eastern Oregon (Jeremy Thompson, personal communication September, 2022).AHD, as the name implies, causes hemorrhaging and bleeding in various tissues and the outcome of the disease ranges from sudden death to inapparent infection ( Heffelfinger Krausman 2023 ).Infected deer with even mild forms of AHD can be fatal through secondary infections or starvation ( Woods et al. 2018 ).The disease is spread by direct contact between deer, resulting in high transmission in areas where deer congregate ( Woods et al. 2018 ).In drought years, AHD can spread through a population of mule deer rapidly due to high concentration of individuals near water sources, resulting in high mortality rates ( Woods et al. 2018 ).Almost 50% of the deer that died in our study were categorized as unknown, leading managers to suspect that many of these deer had died directly or indirectly from AHD. Due to the remote locations of many of the deer in this study, it was not feasible for biologists to get samples from the carcasses to test for AHD before being scavenged by predators.In 2016, the drought was reduced to a moderate level, which helped stop the spread of the AHD outbreak in the population ( US Drought Monitor ;Svobada et al. 2002 ).
Counter to our prediction, mule deer survival was positively influenced by small patches of burned areas within an individual's winter range.In many western juniper woodlands, plant succession has advanced to the point where there is limited understory vegetation, offering little available forage for mule deer, and a dense tree canopy that supports canopy fire spread ( Miller and Rose 1999 ).Wildfire, in this case, can reduce tree density and canopy cover, which will promote understory growth ( Chambers et al. 2005 ).These sites are also cooler and moister compared with Wyoming big sagebrush communities because they occur at slightly higher elevations and are more resistant to invasion by annual grasses ( Miller at al. 2013 ).It is likely that these burned patches within the western juniper PVTs offered better quality forage during the time-frame of this study as the postfire understory reestablished.Mule deer in this study used 1-to 10-yr burned areas on the winter range more often when compared with unburned areas ( Schuyler 2020 ), which supports our finding that these areas have a positive influence on survival.Mule deer use in juniper reduction treatments has been positively related to body size and condition for adult females across their geographic range ( Bender et al. 2013 ;Bergman et al. 2014 ).These findings support the current conservation concepts and management strategies within the sagebrush biome that encourage prescribed burning in forest PVTs to enhance habitat for wildlife, including mule deer ( Chambers et al. 2017 ).
Counter to our prediction, movement behavior (resident or migratory) and juniper canopy cover did not have an effect on survival.Other studies have found differences in survival due to movement behavior ( Nicholson et al. 1997 ;Schuyler et al. 2019 ), but we found no support for migration behavior influencing survival.This is likely because the time frame of this study (2.5 yr) was much shorter than the lifespan of a deer (10 yr) and therefore may not be long enough to detect differences between the two groups.Alternatively, given the low annual survival for this population, it appears neither movement strategy was more advantageous and could suggest that differences in the condition of the summer ranges are negligible (i.e., shared winter range and both summer ranges are all in poor condition).More research is needed to understand factors that might be affecting the entire population across all ranges, such as habitat loss and degradation or predation.
While we did not find a negative effect of juniper canopy, increases in western juniper tree density and canopy has been well documented as having a negative effect on understory plant communities by outcompeting shrub and herbaceous species ( Miller 2005 ;Coultrap et al. 2008 ;Dittel et al. 2018 ).Juniper encroachment has coincided with mule deer population declines ( Aldon 1993 ) and has also been found to negatively influence fawn survival rates due to lack of winter forage ( Bergman et al. 2014 ).Equally, studies have found that mule deer will select for areas with some juniper cover in the winter, which demonstrates a threshold where juniper cover can be beneficial to the species before it has a negative effect ( Coe et al. 2018 ;Schuyler 2020 ).This study area had a number of burned areas on the winter range that could provide better foraging opportunities in patches adjacent to the dense juniper stands.Future research using juniper canopy cover in tandem with understory vegetation measurements (i.e., shrub cover, plant community composition) and adult body condition would aid in elucidating the effect of juniper expansion on adult survival.
The PVTs in our study area have adapted to a range of fire return intervals depending on soil type, elevation, and current plant communities.For example, Wyoming Big Sagebrush PVT are less resilient to fire and have fire return intervals every 100 −150 yr compared with Ponderosa pine PVTs, which have a high proportion of fire-tolerant species and fire return intervals between 20 and 50 yr ( Chambers et al. 2014( Chambers et al. , 2017( Chambers et al. , 2019 ; ;Ellsworth et al. 2017 ).Therefore, we ran two additional models that tested the 1) interaction between migratory behavior and the percent area burned on an individual's summer range and 2) interaction between migratory behavior and the percent burned on an individual's winter range to see if survival was influenced by postfire vegetational response in relation to season ranges (i.e., PVTs).However, neither model was considered competitive.It is also important to acknowledge that we did not account for burn severity, which can have a strong effect on the vegetation responses to fire.For example, burn severity is typically much higher for the more arid, warmer, low-elevation Wyoming big sagebrush plant communities that dominated the winter ranges.These high-intensity wildfires often lead to the conversion from sagebrush and perennial bunchgrasses to invasive annual grasses, such as cheatgrass, medusahead, or ventenata (Ventenata dubia).By contrast, burn severity tends to be variable and patchy for the less arid, colder, higher-elevation dry-forest plant communities found on the summer range.We encourage future studies to conduct field research that incorporates burn severity when exploring postfire plant effects in postfire (10 yr or greater) plant communities on summer and winter ranges to evaluate habitat effects on mule deer survival.
This study provides managers with more information about causes for mule deer population declines.Repeated wildfires, or multiple interacting disturbances (i.e., the invasion of annual grasses, juniper expansion, and logging), can lower resilience, causing these communities cross thresholds into new plant community compositions, offering lower quality habitat for wildlife ( Stringham et al. 2003 ;Bates et al. 2014 ;Stevens-Rumann et al. 2018 ).Severe, extended periods of drought amplify poor forage conditions ( Bender et al. 2011 ), which can make deer more susceptible to diseases ( Heffelfinger and Krausman 2023 ).Therefore, managers should consider broad spatial and long temporal scales when developing and implementing habitat recovery plans.

Conclusions and Implications
Annual mule deer survival can vary significantly in response to environmental conditions.The combined effects of anthropogenic and climate-induced habitat changes within the geographic range of the species have reduced carrying capacity, as reflected by overall herd abundance.Habitat restoration projects focused on establishing perennial grasses and forbs and planting drought-tolerant species is necessary to increase forage quality and quantity for mule deer following a fire, as well as on areas that are prone to experiencing prolonged or increasingly frequent drought.Improving the nutritional availability across the landscape is key to improving mule deer herd productivity.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Figure 1 .
Figure 1.The potential vegetation types within study area (top) and GPS location data of 111 adult female mule deer (Odocoileus hemionus) and wildfire boundaries (bottom) used in known fate survival analysis in Blue Mountains, Oregon, United States, 2015 −2017.

Figure 2 .
Figure 2. Influence of winter range burned on mule deer survival including 95% confidence interval for female radio-marked mule deer (Odocoileus hemionus) in the Blue Mountains Oregon, United States, 2015 −2017.

Table 1
Sequential modeling approach used to develop model set for known-fate survival analysis of female mule deer (Odocoileus hemionus) in the Blue Mountains, Oregon, 2015 −2017.S is the survival rate estimate.Prediction column refers to a priori predictions located in text.Mon) Survival (S) is fully time dependent varying by month (Mon) and biological year (June-May).S (Yr + Mon) S varies by month with an additive effect of the biological yr (June -May).

Table 2
Potential vegetation types (PVTs) as percent of seasonal ranges of radio-collared mule deer in the Blue Mountains, Oregon, United States, 2015-2017.Percent PVTs for seasonal ranges included winter, summer: migratory, and summer: resident.