Demographic parameters are key to understanding population dynamics. Here, we analyse the survival and reproduction of the German wolf population in the 20 years following recolonization. Specifically, we analysed the effects of environmental, ecological, and individual characteristics on i) the survival probability of the population; ii) annual survival rates of age classes; iii) reproduction probability; and iv) reproductive output, measured as the number of detected pups/juveniles. Using the Cox proportional hazards model, we estimated a median survival time of circa 3 years for wolves. Annual survival probabilities were found to be 0.75 for juveniles, 0.75 for subadults, and 0.88 for adults. Survival was lower for juveniles in winter and for subadult males in summer, probably associated with dispersal events. Low habitat suitability was clearly associated with lower survival in juveniles and subadults, but not in adults. Local territory density was related to increased survival. Reproduction probability within a territory was 0.89, but explanatory variables had no effect. Reproductive output was four pups/juveniles on average, positively related to habitat suitability and female experience, but negatively related to territory density. Survival values were very high for the species when compared to other regions. We hypothesize that carrying capacity has not been reached in the study area, thus the survival may decrease in the future if the landscape becomes saturated. Furthermore, our results highlight a spatial pattern in survival and reproduction, with areas of better habitat suitability favouring faster population growth. Thus, targeting conservation measures to low habitat suitability areas will have a strong population effect in the short term by boosting the survival and reproduction of the individuals, while long-term viability should be carefully planned with high suitability areas in mind, as those contain the territories with higher survival and reproduction potential. Methods Wolf individual and territory data for survival and reproduction analyses were provided by the Federal Documentation and Consultation Centre on Wolves (DBBW, www.dbb-wolf.de) and by the Senckenberg Centre for Wildlife Genetics. Information about individuals and territories was grouped into monitoring years (from the 1st of May to the 30th of April next year), starting in 2000 until 2020 (April 2021). Individuals were identified genetically and for the survival analysis, the original dataset was filtered to retain only reliable information on the lifespan of the animals. Thus, individuals with NA ("not available") in the variables 'sex' or 'date of birth' as well as individuals born or died outside the German border were removed, as the environmental data included in the demographic analyses were only available for Germany. Consequently, the status of the individuals (dead, alive) was assessed until April 2021. The age classes were defined as juveniles including pups (0-12 months), subadults (13-24 months), and adults (> 24 months) (Mech and Boitani, 2003). The final dataset contained a total of 1054 individuals. Reproduction data was analysed at the territory level. The number of juvenile counts might be less than the actual number of pups born, thus we defined this variable as ‘minimum reproductive output’. Territories with more than 10 observed pups/ juveniles were removed from analyses to account for the fact that such a high number of pups might stem from a double reproduction and thus belong to one or more females (n = 4). In addition, territories from the first year of pair formation were removed (n = 227), because pairs typically form shortly before or during the breeding season (in autumn or winter) and therefore, there is no opportunity for reproduction in the months prior to the pair formation, which would correspond to the reproduction in the first year in the dataset. The final dataset consisted of 723 entries comprising 205 different territories with data from 1-16 years per territory. Explanatory variables We analysed the survival and reproduction of wolves in relation to environmental and ecological conditions and individual characteristics. For the survival analysis, we used as environmental variables the wolf habitat suitability (Planillo et al., 2024) in an 8 km radius of the territory centroid, wolf local territory density for each year in a 50 km radius, season defined as summer (May-Oct) or winter (Nov-Apr), individual sex and age, with the latter being classified as age classes: juveniles < 12 months, subadults 12 to 24 months, and adults > 24 months old. For the reproduction analysis, environmental and ecological conditions were described by habitat suitability values and local territory density around each breeding territory. As individual characteristics, we included the experience of the reproductive female in the models, measured sequentially as the number of years the same breeding female had reproduced, i.e., the first year that the female reproduced was considered year 1, the second year 2, and so forth. Data Analysis

Survival analysis: Survival analysis was calculated for the whole population and for each of the age classes using Cox Proportional Hazards Regression (Therneau and Grambsch 2000).

Reproduction analysis: We analyzed reproduction patterns for i) the probability of reproduction in a territory and ii) the total number of juveniles per reproductive event. In both cases, data were analysed using generalised linear mixed-effects models (GLMMs) with binomial error distribution and logit link for reproduction probability and Poisson error distribution and log link for the reproductive output. Territory identity was included as a random effect. The total number of years that a territory was monitored in the dataset was included as a weighting variable to avoid an inflated effect of the territories observed only for one year. As explanatory variables, we included the mean habitat suitability of the territory, local territory density in a buffer of 50 km, and the quadratic effect of the experience of the female as fixed effects.

Population growth: We used the values of survival and reproduction to estimate population growth (λ) and contrast it with the observed data. We developed a population matrix model using three age classes, based on obtained values of reproduction and annual survival for the age classes, and used the eigenvalue of the matrix as our λ. We explored the observed population growth values with respect to the effects of the minimum and maximum values of habitat suitability. To compute the lambda for the latter cases, we predicted the survival of juveniles and subadults and the number of pups per reproduction in areas with the lowest and highest observed values of habitat suitability.

We gathered information on wolves that were struck by vehicles on roads in Poland from 2003 to 2023, primarily individuals who were killed, but also those (several individuals) seriously wounded that would have died without the intensive help of veterinarians provided in the rehabilitation centers. Data were collected according to the protocol developed by the authors, which included a form to be completed and a checklist for detailed photo documentation. Reports have been provided by the staff of State Forests, officers of regional directorates for environmental protection, the Police, the regional branches of the General Directorates of National Roads and Motorways, Regional Veterinarian Inspections, the staff of companies removing carcasses of animals from roads for utilization, as well as car drivers who found dead wolves on roads. The photo documentation was sent to us to confirm the species recognition and estimate the individual's age based on the teeth wear. The provided reports contained information on the date and cause of death, the individual's sex, the geographic coordinates of the death site, and the road number and section. Breeding females were recognized based on the evidence of lactation, pregnancy, and nipple length. Reports were verified by the authors and co-workers in the field whenever possible. In cases when doubts appeared, the cause of death was also verified during a necropsy performed by veterinarians or experienced wildlife biologists to exclude instances where wolves were illegally shot and dropped off on the road to simulate an accident. To ensure that as many cases of mortality as possible were registered, we also used an Internet search engine to obtain information on wolves killed and/or injured by cars. Altogether, we collected data on 447 wolves struck by vehicles from 2003 to 2023; however, the exact locations were known for 441 of these. In six cases, the wolves were found between two neighboring roads; therefore, it was impossible to determine the precise location of the accident. We were able to distinguish between adults and juveniles for 436 individuals and assess their exact age based on tooth wear for 362 individuals. Additionally, we obtained information on sex for 403 individuals. We assigned every killed wolf to the Carpathian, Baltic, or Central European wolf management units/subpopulation (see Linnell et al., 2008, for the justification of management units’ delimitation), taking into consideration the results of a study on the wolf genetic structure (Szewczyk et al., 2019, 2021). For all individuals with known exact locations (n = 442), we obtained the habitat type (forest, watercourse, open land, or urbanized area) adjacent to the accident site and the average vehicle traffic on the corresponding section of the road. In Poland, traffic data are measured every five years and are publicly available on the Polish General Directorate for National Roads and Motorways website (https://www.gov.pl/web/gddkia/generalny-pomiar-ruchu).The dataset contains the following columns:Species: common name of the grey wolf Canis lupus, i.e. wolfN: number of individuals (every entry is for single individual)Age_category: pup (1-12 months-year-old); adult (≥1 year-old)Sex: male, femaleBreeder: indicates females with visible singns of pregnancy (presence of fetuses) or lactation (extended nipples)Age: for pups age is given in months, for adults age is given in yearsDate: date in format DD.MM.YYYYLongitude: longitude - a geographical coordinate in format 00.00000Latitude: latitude - a geographical coordinate in format 00.00000Population: name of the wolf population, i.e. Baltic, Central European, CarpathianRoad_Category: general category of the road, i.e. forest_road, lokal_road, regional_road, national_road, express_road, motorway

The recovery of wolves (Canis lupus) across Europe is a notable conservation success in a region with extensive human alteration of landscapes and high human population densities. We provide a comprehensive update on wolf populations in Europe, estimated at over 21,500 individuals by 2022, representing a 58% increase over the past decade. Despite the challenges of high human densities and significant land use for agriculture, industry, and urbanization, wolves have demonstrated remarkable adaptability and increasing population trends in most European countries. Improved monitoring techniques, although varying in quality and scope, have played a crucial role in tracking this recovery. Annually, wolves kill approximately 56,000 domestic animals in the EU, a risk unevenly distributed and differently handled across regions. Damage compensation costs 17 million EUR every year to European countries. Positive economic impacts from wolf presence, such as those related to reducing traffic accide..., In fall 2022, the authors of this paper compiled estimates of wolf population size, trends and damages within their country, as well as details of the monitoring methodology used, the quality of the data, and other information on the legal status and main conservation measures, following a similar method as Chapron et al. (2014), restricting the compilation to existing information and without new analyses of raw monitoring data. Information was obtained from the most reliable sources available at the national level. This joint compilation effort was facilitated by the Large Carnivore Initiative for Europe, a Specialist Group of the IUCN’s Species Survival Commission. Collectively we covered all the European continent, except for the Russian Federation, Belarus and the Republic of Moldova, representing 34 countries (Albania, Austria, Belgium, Bosnia & Herzegovina, Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Kosovo, Latvia, Li..., , Manuscript: Continuing recovery of wolves in Europe

The two data files are:

1. Depredations and compensation costs for each of the 34 countries. The following fields document depredations: depredation unit (i.e. n animals killed for all countries, except for Poland), domestic animal (sheep and goat, cattle, horse and donkey, semi-domestic reindeer, dog) killed, total depredations, year of depredation data, amount of compensation for wolf damages (EUR), and year of compensation data. These figures are taken from Table 3 of the LCIE report by Boitani et al. (2022), except for few cases where the number was updated by the country-responsible author. The dataset also reports the wolf abundance (in number of individuals) and the year of wolf abundance data, as well as the calculated depredations per wolf (total depredations/wolf abundance) and the calculated compensation per wolf (compensation/wolf abundance). The wolf abundance figures are taken from Table 1 of the LCIE report by Boitani ...,

Killrate-dataThis file contains for each study period ("pack-winter") data associated to kill rate and prey availabilty. It is the basis for the functional response models in the article. Variables are described in the ReadMe file.weight_puppiesThis file contains data on the time of marking and weight of all wolf pups marked during winters 2000 - 2013 by the Scandinavian Wolf Research Project. Variables are described in the ReadMe file.

We analyzed data acquired from the USFWS Red Wolf Recovery Program (hereafter ‘Recovery Program’) on radio-marked (hereafter collared), monitored red wolves (i.e., within the NC recovery area). The Recovery Program data include the monitoring history for all collared and monitored adult red wolves released to the wild since the beginning of releases in 1987 to March 1, 2020; n=526. The Recovery Program survival dataset contains the following individual covariates employed in our analyses: endpoint (i.e., final wolf fate by cause of death or disappearance [lost-to-follow-up, LTF]), capture date (beginning of monitoring), and date of endpoint (end of monitoring). The dataset also provided data on sex and other individual covariates not used for our study and hence not described. For recovered wolf carcasses, cause of death was estimated by USFWS using standard methods following necropsy and radiography. Agency removal occurred when a red wolf was captured and removed permanently (lethally or not) from the NC population, generally because the wolf was considered a problem animal by USFWS but also some wolves were removed to supplement other wild populations (n=5). The LTF endpoint occurred when a wolf in the wild disappeared from monitoring because the affixed radio-collar stopped functioning due to either mechanical/battery failure or tampering/destruction by external causes including humans. We reclassified the marked animals’ fates obtained from the Recovery Program survival data into the following mutually-exclusive endpoints, following previous studies: agency removals (lethal or not, by agency personnel; n=40, 7.6%), collision (trauma by vehicle(s); n=68, 12.9%), reported poached (n=150, 28.5%), nonhuman (unrelated to humans; e.g., disease, intraspecific strife, n=66, 12.5%), and unknown (unable to discern cause of death in necropsy; n=82, 15.6%). We include LTF (disappeared individuals; n=117, 22.2%) as one of multiple mutually-exclusive endpoints. We estimated the time between collaring (capture date) and endpoint in days (t) for each red wolf (n=526) in our dataset. We calculated time collared (t) differently for surviving, dead, removed to captivity and LTF endpoints, following previous studies. For our mortality endpoints, we estimated t for wolves monitored by telemetry until death. For wolves relocated to captivity, we used the date of final removal to captivity by agency action. For LTF wolves, we used the last date of telemetry contact. We censored any wolves who were alive at the end of our study period (March 1, 2020, n=3 wolves). In our study, we exploit the complementarity of both models: our joint stratified Cox model allowed us to test the hypothesis that our management covariates affected the rate of occurrence (i.e., hazard) of specific endpoints, and endpoint-specific FG models allowed us to test if and how much these same covariates affected the probability and incidence of said endpoints. Methods Data was collected by the US Fish and Wildlife Service Red Wolf Recovery Program between 1987-2020, and processed following available statistical code (see study SM).

European wolves (Canis lupus) show population genetic structure in the absence of geographic barriers, and across relatively short distances for this highly mobile species. Additional information on the location of and divergence between population clusters is required, particularly because wolves are currently recolonizing parts of Europe. We evaluated genetic structure in 177 wolves from 11 countries using over 67K single nucleotide polymorphism (SNP) loci. The results supported previous findings of an isolated Italian population with lower genetic diversity than that observed across other areas of Europe. Wolves from the remaining countries were primarily structured in a north-south axis, with Croatia, Bulgaria, and Greece (Dinaric-Balkan) differentiated from northcentral wolves that included individuals from Finland, Latvia, Belarus, Poland and Russia. Carpathian Mountain wolves in central Europe had genotypes intermediate between those identified in northcentral Europe and the Dinaric-Balkan cluster. Overall, individual genotypes from northcentral Europe suggested high levels of admixture. We observed high diversity within Belarus, with wolves from western and northern Belarus representing the two most differentiated groups within northcentral Europe. Our results support the presence of at least three major clusters (Italy, Carpathians, Dinaric-Balkan) in southern and central Europe. Individuals from Croatia also appeared differentiated from wolves in Greece and Bulgaria. Expansion from glacial refugia, adaptation to local environments, and human-related factors such as landscape fragmentation and frequent killing of wolves in some areas may have contributed to the observed patterns. Our findings can help inform conservation management of these apex predators and the ecosystems of which they are part.

Description

This repository contains the supplementary materials (Supplementary_information.pdf, Supplementary_figures.pdf, Supplementary_tables.pdf) of the manuscript: "Spatio-temporal dynamics of attacks around deaths of wolves: A statistical assessment of lethal control efficiency in France". This repository also provides the R codes and datasets necessary to run the analyses described in the manuscript.

Datasets

We provide the spatially anonymized R datasets to respect confidentiality. Therefore, the preliminary preparation of the data is not provided in the public codes. These datasets, all geolocated and necessary to the analyses, are:

• Attack_sf.RData: 19,302 analyzed wolf attacks on sheep
  - ID: unique ID of the attack
  - DATE: date of the attack
  - PASTURE: the related pasture ID from "Pasture_sf" where the attack is located
  - STATUS: column resulting from the preparation and the attribution of attacks to pastures (part 2.2.4 of the manuscript); not shown here to respect confidentiality
• Pasture_sf.RData: 4987 analyzed pastures grazed by sheep
  - ID: unique ID of the pasture
  - CODE: Official code in the pastoral census
  - FLOCK_SIZE: maximum annual number of sheep grazing in the pasture
  - USED_MONTHS: months for which the pasture is grazed by sheep
• Removal_sf.RData: 232 analyzed single or multiple wolf removals
  - ID: unique ID of the removal
  - OVERLAP: are they single removal ("single" in the manuscript => "NO" here), or not ("multiple" in the manuscrit, here "SIMULTANEOUS" for removals occurring during the exact same operation or "NON-SIMULTANEOUS" if not).
  - DATE_MIN: date of the single removal or date of the first removal of a group
  - DATE_MAX: date of the single removal or date of the last removal of a group
  - CLASS: administrative type of the removal according to definitions from 2.1 part of the manuscript
  - SEX: sex of the removed wolves if known
  - AGE: class age of the removed wolves if known
  - BREEDER: breeding status of the removed female wolves, "Yes" for female breeder, "No" for female non-breeder. Males are "No" by default, when necropsied; dead individuals with NA were not found.
  - SEASON: season of the removal, as defined in part 2.3.4 of the manuscript
  - MASSIF: mountain range attributed to the removal, as defined in part 2.3.4 of the manuscript
• Mountain_range_buff_sf.RData: one row for each mountain range, corresponding to the buffered mountain ranges where removal control events could be sampled, as defined in part 2.3.3 of the manuscript
• Area_to_exclude_sf.RData: one row for each mountain range, corresponding to the area too close from other mountain ranges, terrestrial and maritime limits, where removal control events could not be sampled, as defined in part 2.3.3 of the manuscript
• Overlapping_removal_sf.RData corresponds to the spatial dataset necessary to run the supplementary figures about overlapping removals (S9)

You can also find the object called Subset_lt.RData, which gives the ID of removals for each dataset (single/multiple removals) or subsets of single removals.

The other RData resulted from the analysis. How to read their names:

• First part:
  - Buffer: they link the attacks to removals or of control events.
  - Kernel: they give the results of the kernel density estimation (z), according to the spatial (y) and temporal (x) coordinates, for each dataset.
  - Int: they give, for each distance and time, the total amount of attack intensities before (INT_BEF) and after (INT_AFT) the combined locations and days of removals or control events. 1 unit of distance = 50 meters, 1 unit of time = 1 day.
  - Trend: they give, for each distance and time, the trends in % of the attack intensities when comparing before and after, with their uncertainties and significance. Files with spatshift are designed for figures about spatial shift (unnested scales), contrary to files without spatshift (nested scales).
• Middle parts: (noted with an X after)
  - obs: results from removals
  - jack: results from jackknife samples of the removals
  - ctl: results related to control sets
  - cor: results corrected for livestock presence
  - raw: results uncorrected for livestock presence
  - sim: results for simulated attacks according to livestock presence

Lengths of lists or sublists correspond to:

- 20 elements: the two main datasets (single/multiple) and the 18 subsets of single removals.
- 100 elements: the 100 control sets.
- 1000 elements: the 1000 simulations of attacks (for the livestock presence correction).
- Other lengths and jack within the name: the jackknife samples.

Structure of the repository

Code 1: file Buffer.R

We keep only removals within geographic zones for the analysis (Removal_analyzed_sf), and sample their control events (100 simulations = control sets, Removal_ctl_sf_lt).

We start by delimiting the spatio-temporal buffer for each row of the removal and control datasets.
- We identify the attacks from Attack_sf.RData within each buffer, thanks to the function Buffer_fn, giving the data frames Buffer_X_df (one row per attack)
- We select the pastures from Pasture_sf.RData within each buffer, thanks to the function Buffer_pasture_fn, giving the data frames Buffer_X_sf (one row per removal or control event)

We calculate the spatial correction:
- We spatially slice each buffer into 200 rings with the function Ring_fn, giving the data frame Ring_sf (one row per ring)
- We add the total pastoral area of the ring of the attack ("SPATIAL_WEIGHT") with the function Spatial_correction_fn, for each attack of each buffer, within Buffer_X_df

We calculate the pastoral correction:
- We create the pastoral matrix for each removal or control event with the function Pastoral_matrix_fn, giving a matrix of 200 rows (one for each ring) and 180 columns (one for each day, 90 days before the removal date and 90 day after the removal date), with the total pastoral area in use by sheep for each corresponding cell of the matrix (one element per removal, Pastoral_X_mx_lt.RData)
- We simulate, for each removal or control event, the random distribution of the attacks from Buffer_X_df.RData according to Pastoral_X_mx_lt.RData with the function Buffer_sim_fn. The process is done 1000 times (one element per simulation, Buffer_X_sim_lt.RData).

Code 2: file Kernel.R where we estimate the attack intensities

We classified the removals into 2 main datasets and 18 subsets, according to part 2.3.4 of the manuscript (Subset_lt.RData) (one element per set).
We compute the jackknife samples for each dataset or subset (Removal_id_jack_lt).
We perform the kernel estimations with the function Kernel_fn (Kernel_X_lt).
We sum the intensities of attacks before and after the removals or control events, with the function Intensity_fn, giving Int_X_df_lt.RData.

Code 3: file Trend.R where we calculate the trends of attack intensities after removals

We focus on the nested trends first:
- We calculate them (Trend_X_df) with function Trend_fn
- We set the result significance test by observing the overlapping of confidence intervals of removals and control sets (Trend_X_comparison_df)

We focus on the spatial shifts (trends for each specific distance):
- We calculate them (Trend_X_spatshift_df) with function Trend_fn
- We set the result significance test by observing the overlapping of confidence intervals of removals and control sets (Trend_X_comparison_spatshift_df)

- Code 4: file Functions.R where can be found all custom functions called in the other analysis codes

- Code 5: file Figures.R produces part of the figures from the manuscript

Detailed comments are included in each code.

Support

If you have any question or request, do not hesitate to contact us at: oksana.grente@gmail.com

Authors and acknowledgment

Grente Oksana (CEFE, CNRS), Opitz Thomas (INRAE), Duchamp Christophe (OFB), Drouet-Hoguet Nolwenn (OFB), Chamaillé-Jammes Simon (CEFE, CNRS) and Gimenez Olivier (CEFE, CNRS).

License

GNU GENERAL PUBLIC LICENSE 3.0

The threat posed by large carnivores to livestock and humans makes peaceful coexistence between them difficult. Effective implementation of conservation laws and policies depends on the attitudes of local residents toward the target species. There are many known correlates of human attitudes toward carnivores, but they have only been assessed at the scale of the individual. Because human societies are organized hierarchically, attitudes are presumably influenced by different factors at different scales of social organization, but this scale dependence has not been examined. We used structured interview surveys to quantitatively assess the attitudes of a Buddhist pastoral community toward snow leopards (Panthera uncia) and wolves (Canis lupus). We interviewed 381 individuals from 24 villages within 6 study sites across the high-elevation Spiti Valley in the Indian Trans-Himalaya. We gathered information on key explanatory variables that together captured variation in individual and village-level socioeconomic factors. We used hierarchical linear models to examine how the effect of these factors on human attitudes changed with the scale of analysis from the individual to the community. Factors significant at the individual level were gender, education, and age of the respondent (for wolves and snow leopards), number of income sources in the family (wolves), agricultural production, and large-bodied livestock holdings (snow leopards). At the community level, the significant factors included the number of smaller-bodied herded livestock killed by wolves and mean agricultural production (wolves) and village size and large livestock holdings (snow leopards). Our results show that scaling up from the individual to higher levels of social organization can highlight important factors that influence attitudes of people toward wildlife and toward formal conservation efforts in general. Such scale-specific information can help managers apply conservation measures at appropriate scales. Our results reiterate the need for conflict management programs to be multipronged.

Abruzzo: Abruzzo, Lazio e Molise National Park; Mercantour: Mercantour National Park; Yellowstone: Yellowstone National Park; ind.: individuals.a Abruzzo: mean estimated value [66]; Mercantour: calculated as the mean number of wolves per pack divided by the mean estimated size of packs' territory in the park (estimated territory size: 260–350km2, [67,68]); Yellowstone: information for the northern range of the park [69].b Wolf fecal samples collected in Abruzzo and Mercantour were submitted to dietary analyses (Abruzzo: P. Ciucci and collaborators, [66]; Mercantour: C. Duchamp and collaborators, [70,71]). In Yellowstone, main prey species were assessed through close monitoring of packs [48–50,67].c Abruzzo: [60]; Mercantour: Millischer pers. comm.; Yellowstone: [72].d Abruzzo: [73]; Mercantour: Millischer pers. comm.; Yellowstone: [74].e Wolves recovered in the study areas and adjacent areas expected to belong to the territory of wolf packs resident of the park. Cause of death: poaching / collision / natural (intraspecific strife)/ unknown. Abruzzo: 2006–2008 (Gentile pers. comm.); Mercantour: 2005–2007 (Millischer pers. comm.); Yellowstone: 2007–2009 (Smith pers. comm.). Five of the 32 individuals who died from natural causes in Yellowstone were members of the studied packs. No such information is available for the two other study areas.Note: No wolf was legally destroyed in Mercantour or in Abruzzo in the years of sample collection. In Yellowstone, wolf hunting and trapping season first opened in September 2009, a few months after our sample collection was finished; four members of resident packs where legally shot before the end of the year.f Dogs (Canis lupus familiaris) travelling with tourists are prohibited in Abruzzo, allowed in the buffer zone but excluded from the core area of Mercantour, and restricted to a range of 100 yards off roads and parking lots in Yellowstone. Working dogs are shepherd dogs and livestock-guarding dogs.g Abruzzo: [64,65]; Mercantour (Millischer pers. comm.); Yellowstone: [75].Ecological characteristics of the investigated wolf populations at the time of sample collection.

Prey vulnerability to predation can vary by life history stage, and prey population stage structure determines the strength a predator species wields on their community. Prey population stage structure can vary over time, yet little is known about how temporal change in prey stage structure influences predator-prey interactions. We used data of wolves hunting adult female elk in Yellowstone National Park to demonstrate that stage-selective wolf predation of old individuals (>11 years old) was more additive than wolf predation of young individuals (2–11 years old). Additive predation of older elk coupled with an aging female elk population increased the strength of wolf predation over time. When vulnerable prey comprise an increasing proportion of a population, their early demise may decrease population growth. Accounting for temporal variation in predation risk across a prey population is therefore critical to understanding the community-level consequences of predator-prey interactions. Methods Please refer to the corresponding publication for methods of elk telemetry, determination of cause of mortality, elk age structure, and recovery of wolf-killed elk. Elk telemetry and survival data were processed into the format required by R package Flexsurv for competing risk analysis. Elk have records for each 'elk year' they were observed during the study period (2000–2009, 2011–2016), where 'elk year' runs from June to May. Start and stop times refer to their ages at the beginning and end of each year. The start age is mid-year if they were captured and marked that year, the stop age is at the end of the year unless they died, their collar failed, or they went missing. For elk marked as yearlings, we excluded their yearling year of data. Elk telemetry and survival data were processed into the format required by R package Wild1 for CIF analysis. These elk are the same as those in the other dataset, but the format was changed. Here time refers to days of the year rather than elk age. Please see the README file for additional information.