[Brucellosis in Alpine ibex: 10 years of research and expert assessments]. / La brucellose du bouquetin des Alpes - Un exemple de dix années de recherche et d'expertise.

Brucellosis due to Brucella melitensis affects domestic and wild ruminants, as well as other mammals, including humans. Despite France being officially free of bovine brucellosis since 2005, two human cases of Brucella melitensis infection in the French Alps in 2012 led to the discovery of one infected cattle herd and of one infected population of wild Alpine ibex (Capra ibex). In this review, we present the results of 10 years of research on the epidemiology of brucellosis in this population of Alpine ibex. We also discuss the insights brought by research and expert assessments on the efficacy of disease management strategies used to mitigate brucellosis in the French Alps.

Title: La brucellose du bouquetin des Alpes - Un exemple de dix années de recherche et d'expertise. Abstract: La brucellose à Brucella melitensis touche les ruminants domestiques et sauvages, ainsi que d'autres mammifères, dont les humains. Bien que la France soit officiellement indemne depuis 2005, deux cas humains reportés en Haute-Savoie en 2012 ont conduit à la découverte de l'infection dans un élevage bovin et chez les bouquetins des Alpes (Capra ibex) du massif du Bargy. Nous présentons dans cette synthèse les principales découvertes de ces dix dernières années sur le système brucellose-bouquetins. Nous discuterons également de l'apport de la recherche et de l'expertise sur l'évaluation de l'efficacité des mesures de gestion sanitaire mises en place dans le massif du Bargy pour lutter contre la brucellose.

Estimating the population size of a mountain galliform in the context of multi-stakeholder adaptive management.

We designed a participatory monitoring program for the capercaillie population in the French Pyrenees based on lek censuses conducted during the breeding season. This program was implemented by a consortium of stakeholders interested in the conservation of French galliforms. The program, carried out since 2010, relied on a dual frame sampling approach: The first sampled frame was the list of all known leks in the study area. We distinguished two types of known leks: leks known to be active before the onset of the program (with at least one cock detected since 2000) and leks with an indeterminate activity status at the time of the onset of the program. The monitoring program also accounted for the existence of leks that were unknown due mainly to incomplete expert knowledge. We therefore built a complementary area frame by discretizing the study area into a set of 4-km2 grid cells. These cells were then sampled and searched to find unknown leks. When unknown leks were found, cock censuses were organized. An additional field experiment allowed us to estimate the detection probability of unknown leks during these cell searches. We then fitted two hierarchical models: (i) An N-mixture model fitted to the lek census data set allowed us to estimate the mean number of cocks on the three types of leks (known active, known indeterminate, and unknown leks); and (ii) another model fitted to the cell search data set allowed us to estimate the number of unknown leks in the studied mountain range. By multiplying the estimated mean numbers of cocks associated with the three types of leks by the number of leks of each type (an estimated value in the case of unknown leks), we obtained estimates of the total numbers of cocks on all leks at different spatial scales in the study area every 2 years. Our model suggests that the capercaillie cock population was stable from 2010 to 2017 over the whole range but decreased slightly in the foothill area and western part, a decrease that worsened in 2018-2019.

Parasitized or non-parasitized, why? A study of factors influencing tick burden in roe deer neonates.

Ixodes ricinus, the most common species of tick in Europe, is known to transmit major pathogens to animals and humans such as Babesia spp. or Borrelia spp. Its abundance and distribution have been steadily increasing in Europe during recent decades, due to global environmental changes. Indeed, as ticks spend most of their life in the environment, their activity and life cycle are highly dependent on environmental conditions and therefore, on climate or habitat changes. Simultaneously, wild ungulates have expanded their range and increased dramatically in abundance worldwide, in particular roe deer (Capreolus capreolus), allowing tick populations to grow and spread. Currently, tick infestation on newborn wild ungulates is poorly documented. However, newborn ungulates are considered more sensitive to tick bites and pathogen transmission because of their immature immune systems. Thus, improving knowledge about the factors influencing tick infestation on newborns is essential to better understand their health risks. This study was conducted at Trois-Fontaines Forest, Champagne-Ardenne, France (1992-2018). Based on a long-term monitoring of roe deer fawns, we used a novel Bayesian model of the infestation of fawns to identify which biotic or abiotic factors were likely to modify the level of infestation by ticks of 965 fawns over time. We show that tick burden increased faster during the first days of life of the fawns and became constant when fawns were five days old and more, which could be explained by the depletion of questing ticks or the turnover of ticks feeding on fawns. Moreover, despite the known positive influence of humidity on tick activity, the tick burdens were weakly related to this parameter. Our results demonstrate that tick infestation was highly variable among years, particularly between 2000-2009. We hypothesize that this results from a modification of habitat caused by Hurricane Lothar.

Assessing spatiotemporal variation in abundance: A flexible framework accounting for sampling bias with an application to common pochard (Aythya ferina).

Assessing trends in the relative abundance of populations is a key yet complex issue for management and conservation. This is a major aim of many large-scale censusing schemes such as the International Waterbird Count (IWC). However, owing to the lack of sampling strategy and standardization, such schemes likely suffer from biases due to spatial heterogeneity in sampling effort. Despite huge improvements of the statistical tools that allow tackling these statistical issues (e.g., GLMM, Bayesian inference), many conservationists still prefer to rely on stand-alone turn-key statistical tools, often violating the prerequisites put forward by the developers of these tools. Here, we propose a straightforward and flexible approach to tackle the typical statistical issues one can encounter when analyzing count data of monitoring schemes such as the IWC. We rely on IWC counts of the declining common pochard populations of the Northwest European flyway as a case study (period 2002-2012). To standardize the size of sampling units and mitigate spatial autocorrelation, we grouped sampling sites using a 75 × 75 km grid cells overlaid over the flyway of interest. Then, we used a hierarchical modeling approach, assessing population trends with random effects at two spatial scales (grid cells, and sites within grid cells) in order to derive spatialized values and to compute the average population trend at the whole flyway scale. Our approach allowed to tackle many statistical issues inherent to this type of analysis but often neglected, including spatial autocorrelation. Concerning the case study, our main findings are that: (1) the northwestern population of common pochards experienced a steep decline (4.9% per year over the 2002-2012 period); (2) the decline was more pronounced at high than low latitude (11.6% and 0.5% per year at 60° and 46° of latitude, respectively); and, (3) the decline was independent of the initial number of individuals in a given site (random across sites). Beyond the case study of the common pochard, our study provides a conceptual statistical framework for estimating and assessing potential drivers of population trends at various spatial scales.

Estimating disease prevalence and temporal dynamics using biased capture serological data in a wildlife reservoir: The example of brucellosis in Alpine ibex (Capra ibex).

The monitoring of the disease prevalence in a population is an essential component of its adaptive management. However, field data often lead to biased estimates. This is the case for brucellosis infection of ibex in the Bargy massif (France). A test-and-cull program is being carried out in this area to manage the infection: captured animals are euthanized when seropositive, and marked and released when seronegative. Because this mountainous species is difficult to capture, field workers tend to focus the capture effort on unmarked animals. Indeed, marked animals are less likely to be infected, as they were controlled and negative during previous years. As the proportion of marked animals in the population becomes large, captured animals can no longer be considered as an unbiased sample of the population. We designed an integrated Bayesian model to correct this bias, by estimating the seroprevalence in the population as the combination of the separate estimates of the seroprevalence among unmarked animals (estimated from the data) and marked animals (estimated with a catalytic infection model, to circumvent the scarcity of the data). As seroprevalence may not be the most responsive parameter to management actions, we also estimated the proportion of animals in the population with an active bacterial infection. The actual infection status of captured animals was thus inferred as a function of their age and their level of antibodies, using a model based on bacterial cultures carried out for a sample of animals. Focusing on the population of adult females in the core area of the massif, i.e. with the highest seroprevalence, this observational study shows that seroprevalence has been divided by two between 2013 (51%) and 2018 (21%). Moreover, the likely estimated proportion of actively infected females in the same population, though very imprecise, has decreased from a likely estimate of 34% to less than 15%, suggesting that the management actions have been effective in reducing infection prevalence.

Should I stay or should I go? Determinants of immediate and delayed movement responses of female red deer (Cervus elaphus) to drive hunts.

Hunting can be used as a tool for wildlife management, through limitation of population densities and dissuading game from using sensitive areas. The success of these approaches requires in depth knowledge of prey movement. Indeed, movement decisions of game during hunting may affect the killing success of hunters as well as the subsequent location of surviving animals. We thus investigated red deer movement responses to drive hunts and their causal factors. We studied 34 hunting events in the National Estate of Chambord (France) and thereby provided a fine-scale characterization of the immediate and delayed movement responses of red deer to drive hunts. Red deer responded to drive hunts either by immediately fleeing the hunted area, or by initially remaining before ultimately fleeing after the hunters had departed. A few hours after the hunt, all individuals were located in distant areas (> 2 kilometres) from the hunted area. Immediate flight responses were less common when drive hunts occurred in areas with dense understorey. However, neither beater/dog densities nor site familiarity influenced the immediate flight decision. Following a drive hunt, red deer remained outside the hunted areas for periods twice as long compared to periods when no hunting occurred (34 hours vs. 17 hours). Such knowledge of game movement rates in response to drive hunts may help the development of informed management policy for hunted red deer populations.

Altitude shapes the environmental drivers of large-scale variation in abundance of a widespread mammal species.

AIM: Habitat quality and heterogeneity directly influence the distribution and abundance of organisms at different spatial scales. Determining the main environmental factors driving the variation in species abundance is crucial to understand the underlying ecological processes, and this is especially important for widely distributed species living in contrasting environments. However, the responses to environmental variation are usually described at relatively small spatial scales. Here, we studied the variation in abundance of a widely distributed mustelid, the European badger (Meles meles), across France. LOCATION: The whole metropolitan France. METHODS: We used (a) direct detections of 9,439 dead and living badgers, from 2006 to 2009, to estimate badger relative abundance in 703 small agricultural regions of metropolitan France and (b) a Bayesian modeling approach to identify the main environmental determinants influencing badger abundance. RESULTS: Despite a continuous distribution of badger in France, we found large variation in badger abundance between regions, explained by environmental factors. Among a set of 13 environmental variables, we demonstrated that badger abundance in lowlands (<400 m a.s.l.) was mostly driven by biotic factors such as potential food resources (earthworm abundance and fruit crops) and forest fragmentation. Conversely, in mountainous areas, abiotic factors (i.e., soil texture and climate) drove the variation in badger relative abundance. MAIN CONCLUSIONS: These results underline the importance of mapping the abundance of wildlife species based on environmental suitability and highlight the complexity of drivers influencing species abundance at such large spatial scales. Altitude shaped the environmental drivers (biotic vs. abiotic) that most influenced relative abundance of a widespread species. In the case of badger, such abundance maps are crucial to identify critical areas for species management as this mustelid is a main wild vector of bovine tuberculosis in several countries.

Many, large and early: Hunting pressure on wild boar relates to simple metrics of hunting effort.

Wild boar populations have increased dramatically over the last decades throughout Europe and in France in particular. While hunting is considered the most efficient way to control game populations, many local conflicts persist after the hunting period due to remaining high densities of wild boar despite the large number of animals culled every year. Therefore, increasing the efficiency of hunting is a timely issue. Herein, we assessed how hunting effort can be measured, and we determined whether the hunting effort carried out by hunters explains the observed hunting pressure. We measured the characteristics and results of all hunts that occurred in the experimental forest of Châteauvillain-Arc-en-Barrois (Northeastern France), and we modelled the number of animals culled as a function of the hunting effort, measured by the number of beaters, hunters, and dogs, as well as the size of the hunting area. We also accounted for variables suspected to affect the hunting efficiency achieved with a given effort, such as time of day (AM/PM), the month during which hunting occurred. We found that more posted hunters, larger hunted areas, and hunts carried out early in the season, i.e. before February, increased the number of culled animals. Our model can be used by wildlife managers to adjust hunting effort in order to reach the hunting pressure expected to meet management objectives.

Capitalizing on opportunistic data for monitoring relative abundances of species.

With the internet, a massive amount of information on species abundance can be collected by citizen science programs. However, these data are often difficult to use directly in statistical inference, as their collection is generally opportunistic, and the distribution of the sampling effort is often not known. In this article, we develop a general statistical framework to combine such "opportunistic data" with data collected using schemes characterized by a known sampling effort. Under some structural assumptions regarding the sampling effort and detectability, our approach makes it possible to estimate the relative abundance of several species in different sites. It can be implemented through a simple generalized linear model. We illustrate the framework with typical bird datasets from the Aquitaine region in south-western France. We show that, under some assumptions, our approach provides estimates that are more precise than the ones obtained from the dataset with a known sampling effort alone. When the opportunistic data are abundant, the gain in precision may be considerable, especially for rare species. We also show that estimates can be obtained even for species recorded only in the opportunistic scheme. Opportunistic data combined with a relatively small amount of data collected with a known effort may thus provide access to accurate and precise estimates of quantitative changes in relative abundance over space and/or time.

The spatial distribution of Mustelidae in France.

We estimated the spatial distribution of 6 Mustelidae species in France using the data collected by the French national hunting and wildlife agency under the "small carnivorous species logbooks" program. The 1500 national wildlife protection officers working for this agency spend 80% of their working time traveling in the spatial area in which they have authority. During their travels, they occasionally detect dead or living small and medium size carnivorous animals. Between 2002 and 2005, each car operated by this agency was equipped with a logbook in which officers recorded information about the detected animals (species, location, dead or alive, date). Thus, more than 30000 dead or living animals were detected during the study period. Because a large number of detected animals in a region could have been the result of a high sampling pressure there, we modeled the number of detected animals as a function of the sampling effort to allow for unbiased estimation of the species density. For dead animals -- mostly roadkill -- we supposed that the effort in a given region was proportional to the distance traveled by the officers. For living animals, we had no way to measure the sampling effort. We demonstrated that it was possible to use the whole dataset (dead and living animals) to estimate the following: (i) the relative density -- i.e., the density multiplied by an unknown constant -- of each species of interest across the different French agricultural regions, (ii) the sampling effort for living animals for each region, and (iii) the relative detection probability for various species of interest.

Bayesian modelling of hunting data may improve the understanding of host-parasite systems: wild boar diseases and vaccination as an example.

Wildlife diseases are often studied using hunting data. In such studies, inferences about diseases are often made by comparing raw disease prevalence levels, ignoring complications like stochasticity in recruitment. We carried out a field trial to study the effectiveness of oral vaccination of wild boar (Sus scrofa) against classical swine fever (CSF) in the Vosges mountains (Northeastern France) for 3 years (2008-2010). Since August 2004, hunters had carried out three vaccination sessions per year in spring, summer and autumn. During our study period, we determined whether each wild boar hunted in our study area was immunized or not against CSF. We used a Bayesian approach to model the changes in the proportion of vaccinated animals in the population of young animals (i.e., <12 months old). This approach allowed to disentangle the effects of the birth peaks (leading to a decrease) and of both the vaccination sessions and natural infection (leading to an increase) on this proportion. We thus inferred, at the individual level, the probability that a non-immunized animal became vaccinated after a particular session. There was a high between-year variability in the effectiveness of the vaccination: the observed patterns were similar in 2008 and 2010, but 2009 was characterized by an overall greater effectiveness of the vaccination. Within a particular year, the spring vaccination session was more effective than the autumn session, probably because of the higher food availability in autumn that render the vaccination places less attractive to the animals. The vaccination effectiveness was rather low in summer, except in 2009, probably because of higher age identification error this year. This model also highlighted an immunisation of animals occurring outside vaccination periods, which suggests either the presence of the CSF virus in our study area, or the consumption of the vaccine outside the vaccination sessions. Finally, we observed a high spatial variability of the probability of vaccination. The effectiveness of the vaccination was indeed strongly related to both the distribution of the forests and the distribution of the vaccination places in our study area. This study highlights an optimal vaccination effort of 1.25 places per km(2) to maximize the proportion of immune wild boar in that area.

Wildlife tracking data management: a new vision.

To date, the processing of wildlife location data has relied on a diversity of software and file formats. Data management and the following spatial and statistical analyses were undertaken in multiple steps, involving many time-consuming importing/exporting phases. Recent technological advancements in tracking systems have made large, continuous, high-frequency datasets of wildlife behavioural data available, such as those derived from the global positioning system (GPS) and other animal-attached sensor devices. These data can be further complemented by a wide range of other information about the animals' environment. Management of these large and diverse datasets for modelling animal behaviour and ecology can prove challenging, slowing down analysis and increasing the probability of mistakes in data handling. We address these issues by critically evaluating the requirements for good management of GPS data for wildlife biology. We highlight that dedicated data management tools and expertise are needed. We explore current research in wildlife data management. We suggest a general direction of development, based on a modular software architecture with a spatial database at its core, where interoperability, data model design and integration with remote-sensing data sources play an important role in successful GPS data handling.

A general framework for the statistical exploration of the ecological niche.

We propose a new statistical framework for the exploratory analysis of the ecological niche, the "General niche-environment system factor analysis" (GNESFA). The data required for this analysis are (i) a table giving the values of the environmental variables in each environment unit (EU, e.g., the patches of habitat on a vector map), (ii) a set of weights measuring the availability of the EUs to the species (e.g., the proportion of the study area covered by a given patch), and (iii) a set of utilization weights describing the use of the EUs by the focal species (e.g., the proportion of detections of the species in each patch). Each row of the table corresponds to a point in the multidimensional space defined by the environmental variables, and each point is associated with two weights. The GNESFA searches the directions in this space where the two weight distributions differ the most, choosing one distribution as the reference, and the other one as the focus. The choice of the utilization as the reference corresponds to the MADIFA (Mahalanobis distances factor analysis), which identifies the directions on which the available EUs are in average the furthest from the optimum of the niche, allowing habitat suitability modelling. The choice of the availability as the reference corresponds to the FANTER (Factor analysis of the niche, taking the environment as the reference), which identifies the directions on which the niche is the furthest from the average environment (marginality) and those on which the niche is the narrowest compared with the environment (specialization). The commonly used ENFA (Ecological niche factor analysis) is at the middle point between the MADIFA and the FANTER, considering both distributions as the reference and the focus simultaneously. When used concurrently, these three analyses allow an extensive exploration of the system.