3 April 2024 in Proceedings Of The Royal Society B-Biological Sciences
Protected area (PA) networks are a pivotal tool to fight biodiversity loss, yet they often need to balance the mission of nature conservation with the socio-economic need of giving opportunity for outdoor recreation. Recreation in natural areas is important for human health in an urbanized society, but can prompt behavioural modifications in wild animals. Rarely, however, have these responses being studied across multiple PAs and using standardized methods. We deployed a systematic camera trapping protocol at over 200 sites to sample medium and large mammals in four PAs within the European Natura 2000 network to assess their spatio-temporal responses to human frequentation, proximity to towns, amount of open habitat and topographical variables. By applying multi-species and single-species models for the number of diurnal, crepuscular and nocturnal detections and a multi-species model for nocturnality index, we estimated both species-specific- and meta-community-level effects, finding that increased nocturnality appeared the main strategy that the mammal meta-community used to cope with human disturbance. However, responses in the diurnal, crepuscular and nocturnal site use were mediated by species' body mass, with larger species exhibiting avoidance of humans and smaller species more opportunistic behaviours. Our results show the effectiveness of standardized sampling and provide insights for planning the expansion of PA networks as foreseen by the Kunming–Montreal biodiversity agreement.
12 December 2024 in Ecology and Evolution
Camera trapping has become increasingly common in ecological studies, but is hindered by analyzing large datasets. Recently, artificial intelligence (deep learning models in particular) has emerged as a promising solution. However, applying deep learning for images processing is complex and often requires programming skills in Python, reducing its accessibility. Some authors addressed this issue with user‐friendly software, and a further progress was the transposition of deep learning to R, a statistical language frequently used by ecologists, enhancing flexibility and customization of deep learning models without advanced computer expertise. We aimed to develop a user‐friendly workflow based on R scripts to streamline the entire process, from selecting to classifying camera trap images. Our workflow integrates the MegaDetector object detector for labelling images and custom training of the state‐of‐the‐art YOLOv8 model, together with potential for offline image augmentation to manage imbalanced datasets. Inference results are stored in a database compatible with Timelapse for quality checking of model predictions. We tested our workflow on images collected within a project targeting medium and large mammals of Central Italy, and obtained an overall precision of 0.962, a recall of 0.945, and a mean average precision of 0.913 for a training set of only 1000 pictures per species. Furthermore, the custom model achieved 91.8% of correct species‐level classifications on a set of unclassified images, reaching 97.1% for those classified with > 90% confidence. YOLO, a fast and light deep learning architecture, enables application of the workflow even on resource‐limited machines, and integration with image augmentation makes it useful even during early stages of data collection. All R scripts and pretrained models are available to enable adaptation of the workflow to other contexts, plus further development.
1 January 2025 in Canadian Journal of Zoology
Wolves ( Canis lupus Linnaeus, 1758) have recolonized much of Europe and recently returned to the Apuan Alps (Central Italy), a partially isolated mountain chain unoccupied by wolves between 1850s and 2000s, after extirpation in early 1800s. Early stages of large carnivore recolonization often go undetected, particularly when standardized surveys are lacking. In the Apuan Alps, this left the spatiotemporal dynamics of wolf return unresolved. To reconstruct this process, we combined opportunistic playback-howling surveys (2007–2020) with camera-trap records (2011–2020), and fit a Bayesian occupancy model accounting for preferential sampling on camera-trap data. Playback surveys first confirmed reproduction in 2014; breeding units increased to four by 2018. Camera-trap occupancy rose 15-fold (2011–2018), despite early site‐selection bias. Systematic sampling (2019–2020) validated these trends, estimating occupancy at ∼0.60 and detection probabilities threefold higher than opportunistic surveys. Opportunistic data could not fully detail early recolonization and spatial mechanisms due to likely misdetection of initial reproduction. Nonetheless, combined evidence suggested a two-phase pattern consistent with other wolf recolonizations: an initial lag period followed by rapid expansion via short-distance diffusion and jump-dispersal. Although prone to biases, opportunistic records can yield insights into cryptic recolonizations, guide early management and conservation, and inform future monitoring of recolonizing large carnivores.
20 August 2025 in Animal Behaviour