Cyril Milleret

Spatial Capture-Recapture (SCR) models using least-cost path distance offer an unified framework to estimate landscape connectivity and population size from individual detection data, while accounting for individual and spatial heterogeneity in space use. In our case study on the Pyrenean brown bear population, two “outliers” individuals were detected more often over very large spatial extent compared to other individuals. The integration of such individuals made SCR modelling challenging, especially since it remains unclear how unmodelled heterogeneity in space use may bias connectivity and population size estimates. To address this gap, we used simulations reflecting the Pyrenean brown bear population, with two groups of individuals differing in their space use due to individual characteristics but also in their spatial responses to landscape structure. We compared six SCR model formulations that varied in whether individual and spatial heterogeneity in space use were (1) ignored, (2) explicitly modelled, or (3) handled by removing outliers. We then applied the same models to the empirical bear dataset. By combining simulations and sensitivity analyses, we highlight the challenges of choosing the appropriate modelling approach when multiple sources of heterogeneity in space use occur simultaneously. Our results show that the treatment of heterogeneity in space use should match the research objective: removing outliers supports accurate population size estimation, while explicitly modelling heterogeneity is essential for reliable connectivityassessment. Overall, we provide a practical framework for identifying and addressing heterogeneity in SCR applications, to guide practitioners in deciding when additional model complexity is warranted.

Death is an inherently spatial process. It happens to someone, somewhere, but often remains undetected in nature. Death is also the primary means by which humans regulate wildlife populations. Using a novel analytical method that accounts for the cryptic nature of the fate of individuals and one of the world’s most comprehensive non-invasive genetic monitoring datasets, we were able to map cause-specific mortality of the entire Scandinavian grey wolf ( Canis lupus ) population despite the fact that most mortality events (mean = 65%; [95% CrI: 50.3–76.8%]) remained undetected. Our analysis revealed strong spatial variation in mortality with, for example, areas with a high risk of mortality linked with the current wolf management policies. Furthermore, we showed that the risk of legal mortality increased, while the risk of mortality due to causes other than legal mortality decreased with local wolf population density. This illustrates the complex interactions between spatial determinants and cause-specific mortality and therefore the importance of considering spatial variation when estimating mortality. Maps of mortality can inform wildlife management and conservation by capturing an elusive process in population dynamics as it unfolds in time and space.

Wildlife populations are not static. Intrinsic and extrinsic factors affect individuals, which lead to spatiotemporal variation in population density and range. Yet, dynamics in density and their drivers are rarely documented, due in part to the inherent difficulty of studying long-term population-level phenomena at ecologically meaningful scales. We studied the spatiotemporal density dynamics in a recolonizing large carnivore population, the wolverine Gulo gulo , across the Scandinavian Peninsula over nine years. We fitted open-population spatial capture-recapture models to noninvasive genetic sampling data collected across Norway and Sweden to estimate annual density surfaces and their drivers. This approach allowed us to model sex-specific changes in wolverine density and the effect of landscape-level environmental determinants over time. Our results revealed that, as wolverines successfully recolonized many parts of their historical range in Scandinavia, the relationship with spatial determinants of density has changed over time. We also found support for sex-specific responses of the Scandinavian wolverine to the environmental determinants of density and differences in the temporal dynamics of their relationships, indicating disproportionate recolonization ability and anthropogenic pressures. We observed significant changes in the relationship of female wolverine density with several determinants during the study period, suggesting still ongoing expansion of female wolverines whereas males might have already reached the range limits. These findings show that the Scandinavian wolverine population is still recovering from centuries of persecution and severe range contraction. Our study sheds light on the dynamics and challenges of recolonizing large carnivores in human-dominated landscapes across time and space.

The ongoing expansion of wolf (Canis lupus) populations in Europe has led to a growing demand for up‐to‐date abundance estimates. Non‐invasive genetic sampling (NGS) is now widely used to monitor wolves, as it allows individual identification and abundance estimation without physically capturing individuals. However, NGS is resource‐intensive, partly due to the elusive behaviour and wide distribution of wolves, as well as the cost of DNA analyses. Optimisation of sampling strategies is therefore a requirement for the long‐term sustainability of wolf monitoring programs. Using data from the 2020–2021 Italian Alpine wolf monitoring, we investigate how (i) reducing the number of samples genotyped, (ii) reducing the number of transects, and (iii) reducing the number of repetitions of each search transect impacted spatial capture‐recapture population size estimates. Our study revealed that a 25% reduction in the number of transects or, alternatively, a 50% reduction in the maximum number of repetitions yielded abundance estimates comparable to those obtained using the entire dataset. These modifications would result in a 2046 km reduction in total transect length and 19,628 km reduction in total distance searched. Further reducing the number of transects resulted in up to 15% lower and up to 17% less precise abundance estimates. Reducing only the number of genotyped samples led to higher (5%) and less precise (20%) abundance estimates. Randomly subsampling genotyped samples reduced the number of detections per individual, whereas subsampling search transects resulted in a less pronounced decrease in both the total number of detections and individuals detected. Our work shows how it is possible to optimise wolf monitoring by reducing search effort while maintaining the quality of abundance estimates, by adopting a modelling framework that uses a first survey dataset. We further provide general guidelines on how to optimise sampling effort when using spatial capture‐recapture in large‐scale monitoring programmes.

Range expansion is a common feature from invasive to reintroduced or recovering populations. This process is mainly driven by population growth and dispersal and, consequently, different species’ intrinsic characteristics and dispersal mechanisms will result in contrasting population structures in space. How individuals of different sex and age classes are spatially distributed is key to understand and forecast range expansions, but remains largely unexplored. Here, we developed an age-structured open population spatial capture recapture (OPSCR) model to understand how spatial sex and age structures can shape population dynamics, as well as recovery processes. We use the expanding and endangered Pyrenean brown bear population as case study, taking advantage of a comprehensive non-invasive monitoring dataset collected between 2017 and 2021. The expansion front was dominated by adult (>4 years) and subadult males (2-4 years), while females and juveniles (<2 years) prevailed at the core. Overall, density declined with distance to the population core (area where last bears remained in the ‘90s and where translocated females established). Bear expansion was affected by the locations of the remnant range (i.e., the core areas) and constrained by female philopatry. Future projections suggest that increasing reproduction events at the periphery could speed up future population spread, and that this population will continue growing even under some level of demographic stochasticity. Our novel OPSCR model opens new pathways to understand complex spatiotemporal patterns of expanding populations and, in the case of large carnivores, could inform conservation action, for example, anticipating the adoption of damage prevention measures and raising awareness campaigns on how to coexist with bears in future expansion areas.

The ongoing expansion of wolf (Canis lupus) populations has led to a growing demand for up-to-date abundance estimates. Non-invasive genetic sampling (NGS) is now widely used to monitor wolves, as it allows individual identification and abundance estimation without physically capturing individuals. However, NGS is resource-intensive, partly because of the wolf elusive behaviour and wide distribution, but also because of the cost of DNA analyses. Optimization of sampling strategies is therefore a requirement for the long-term sustainability of wolf monitoring programs. Using data from the 2020-2021 Italian Alpine wolf monitoring, we investigate how (i) reducing the number of samples genotyped, (ii) reducing the number of transects, and (iii) reducing the number of repetitions of each search transect, impacted spatial capture-recapture population size estimates. Our study revealed that a 25% reduction in the number of transects or, alternatively, a 50% reduction in the maximum number of repetitions yielded abundance estimates comparable to those obtained using the entire dataset. These modifications would result in a 2,046 km reduction in total transect length and 19,628 km reduction in total distance searched. Further reducing the number of transects resulted in up to 15% lower and up to 17% less precise abundance estimates. Reducing only the number of genotyped samples led to higher (5%) and less precise (20%) abundance estimates. Randomly subsampling genotyped samples reduced the number of detections per individual, whereas subsampling search transects resulted in a less pronounced decrease in both the total number of detections and individuals detected. Our work shows how it is possible to optimise wolf monitoring by reducing search effort while maintaining the quality of abundance estimates, by adopting a modelling framework that uses a first survey dataset. We further provide general guidelines on how to optimise sampling effort when using spatial capture-recapture in large-scale monitoring programmes.

Survival of juvenile ungulates represents an important demographic parameter that influences population dynamics within ecosystems. In many ecological systems, the mortality of juvenile ungulates is influenced by various factors, including predation by large carnivores, human hunting activities and weather. While wolves Canis lupus are known to prey on moose Alces alces throughout all seasons, brown bears Ursus arctos primarily engage in predation during early summer, while human harvest primarily occurs in autumn and early winter. Hence, understanding the impacts of predation, harvest, and weather on the survival of juvenile moose is crucial for adaptive population management and the determination of sustainable harvest rates. To investigate the summer and autumn–winter survival of moose calves in relation to carnivore occurrence (wolf presence and bear density), summer habitat productivity, winter severity, human harvest, and migratory behaviour (migratory versus resident), we analysed data collected from 39 GPS‐collared female moose in south‐central Scandinavia. Our findings revealed significant interannual variation in summer survival rates, with areas with relatively higher bear densities exhibiting calf mortality rates twice as high as those in regions with low bear density. During the autumn–winter period, calf survival was lowest in the presence of wolves and deep snow, and it exhibited a negative correlation with the proportion of clearcuts and young forests within the mother's home range. Additionally, calf survival was negatively correlated with the risk of human hunting, and calves of stationary females displayed ten times higher survival rates compared to migratory individuals. Our study provides valuable insights into the survival of moose calves coexisting with two large carnivores and humans. Improving our understanding of the mechanisms causing calf survival to fluctuate has become increasingly important as many local moose populations in Scandinavia are declining and exposed to expanding predator populations, intense hunting pressure, and other threats associated with climate change.

After centuries of intense persecution, several large carnivore species in Europe and North America have experienced a rebound. Today's spatial configuration of large carnivore populations has likely arisen from the interplay between their ecological traits and current environmental conditions, but also from their history of persecution and protection. Yet, due to the challenge of studying population‐level phenomena, we are rarely able to disentangle and quantify the influence of past and present factors driving the distribution and density of these controversial species. Using spatial capture‐recapture models and a data set of 742 genetically identified wolverines Gulo gulo collected over ½ million km ² across their entire range in Norway and Sweden, we identify landscape‐level factors explaining the current population density of wolverines in the Scandinavian Peninsula. Distance from the relict range along the Swedish–Norwegian border, where the wolverine population survived a long history of persecution, remains a key determinant of wolverine density today. However, regional differences in management and environmental conditions also played an important role in shaping spatial patterns in present‐day wolverine density. Specifically, we found evidence of slower recolonization in areas that had set lower wolverine population goals in terms of the desired number of annual reproductions. Management of transboundary large carnivore populations at biologically relevant scales may be inhibited by administrative fragmentation. Yet, as our study shows, population‐level monitoring is an achievable prerequisite for a comprehensive understanding of the distribution and density of large carnivores across an increasingly anthropogenic landscape.

Survival among juvenile ungulates is an important demographic trait affecting population dynamics. In many systems, juvenile ungulates experience mortality from large carnivores, hunter harvest and climate-related factors. These mortality sources often shift in importance both in space and time. While wolves (Canis lupus) predate on moose (Alces alces) throughout all seasons, brown bear (Ursus arctos) predation and human harvest happen primarily during early summer and fall, respectively. Hence, understanding how the mortality of juvenile moose is affected by predation, harvest and climate is crucial to adaptively managing populations and deciding sustainable harvest rates. We used data from 39 female moose in south-central Scandinavia to investigate the mortality of 77 calves in summer/fall and winter/spring, in relation to carnivore presence (defined as wolf presence and bear density), summer productivity, secondary road density, winter severity and migratory strategy (migratory versus resident) using logistic regressions. Summer mortality varied significantly between years but was not correlated to any of our covariates. In winter, calf mortality was higher with deeper snow in areas with wolves compared to areas without and increased more strongly with an increasing proportion of clearcuts/young forests in the presence of wolves compared to when wolves were absent. Lastly, increasing hunting risk was associated with higher calf mortality, and migratory females had higher calf mortality compared to stationary ones. Our study provides useful insight into mortality rates of moose calves coexisting with two large carnivores and with an intensive harvest pressure. Increasing our understanding of the mechanisms driving calf mortality both in summer and winter will become increasingly important if the populations of wolves and bears continue to expand and the moose population declines, and both summers and winters become warmer.

Non‐invasive genetic sampling (NGS) methods are becoming a mainstay in wildlife monitoring and can be used with spatial capture‐recapture (SCR) methods to estimate population density. Yet SCR based on NGS remains relatively underused for ungulate population monitoring, despite the importance of robust density estimates for this ecologically and economically important group of species. This may be in part attributed to biological characteristics of ungulate species and data collection methods that lead to violations of SCR model assumptions. We conducted a simulation study to evaluate the robustness of SCR methods to spatially heterogeneous density (i.e., configuration of individuals into groups of variable sizes and composition), individual heterogeneity in space‐use patterns, and adaptive sampling (i.e., variation in detectability across space that correlates with density). We evaluated each violation separately and in combination. We parameterized our simulations based on published information and preliminary analyses of NGS data sets of 3 ungulate species: chamois (Rupicapra rupicapra), red deer (Cervus elaphus), and wild boar (Sus scrofa). While SCR estimates were robust to grouping and adaptive sampling, abundance estimates could be negatively biased (up to 10% in our simulations) in the presence of unaccounted individual heterogeneity in space use. The degree to which abundance estimates were underestimated depended mostly on the amount of variation in space use and detectability among age classes. This bias was also accompanied by a reduction in precision and coverage probability of the SCR estimators. We discuss the implications of these findings, possible approaches to identify problematic violations in available data sets (goodness‐of‐fit tests), and potential further developments of SCR models to ensure reliable abundance estimates for ungulate populations from NGS data.