Aim: Land‐use change is a major threat to biodiversity globally. Roads cause direct mortality and limitation of individual movements, which may isolate populations and affect their viability in the long term. Here we provide the first comprehensive global assessment of the exposure of terrestrial mammalian carnivores to roads using an integrated modelling framework.Location: Global.Methods: We estimated critical road densities and critical patch sizes for each species based on a spatially explicit model and life‐history traits. We calculated the distribution of landscape fragment sizes for each carnivore species by intersecting global road density with each species range. The proportion of a species’ geographical range with fragments below the critical patch size is used as an index of the vulnerability to roads.Results: We found that the carnivores expected to be most exposed to roads belong to families Felidae, Ursidae, Mustelidae, Canidae and Procyonidae. Approximately one‐third of the species most affected have not been identified by the IUCN as threatened by roads. Our model projects time to extinction that may be as low as one century for some species, such as the endangered Iberian lynx. Species are expected to be more exposed in areas with medium to high road density but, surprisingly, also in areas where road density is relatively low. Hotspots of the number of species locally endangered by roads occur in North America and Asia.Main conclusions: Our results suggest the need to reassess the status and threats of those species that have not been previously recognized as strongly affected by roads. Our framework can be applied at different spatial scales, to assess the effects of the development of the road network and inform prioritization schemes for road building, and to identify areas for conservation, and species requiring particular mitigation and restoration measures.

Opportunistic citizen science databases are becoming an important way of gathering information on species distributions. These data are temporally and spatially dispersed and could have limitations regarding biases in the distribution of the observations in space and/or time. In this work, we test the influence of landscape variables in the distribution of citizen science observations for eight taxonomic groups. We use data collected through a Portuguese citizen science database (biodiversity4all.org). We use a zero-inflated negative binomial regression to model the distribution of observations as a function of a set of variables representing the landscape features plausibly influencing the spatial distribution of the records. Results suggest that the density of paths is the most important variable, having a statistically significant positive relationship with number of observations for seven of the eight taxa considered. Wetland coverage was also identified as having a significant, positive relationship, for birds, amphibians and reptiles, and mammals. Our results highlight that the distribution of species observations, in citizen science projects, is spatially biased. Higher frequency of observations is driven largely by accessibility and by the presence of water bodies. We conclude that efforts are required to increase the spatial evenness of sampling effort from volunteers.

The scarcity of long-term observational data has limited the use of statistical or machine-learning techniques for predicting intraannual ecological variation. However, time-stamped citizen-science observation records, supported by media data such as photographs, are increasingly available. In the present article, we present a novel framework based on the concept of relative phenological niche, using machine-learning algorithms to model observation records as a temporal sample of environmental conditions in which the represented ecological phenomenon occurs. Our approach accurately predicts the temporal dynamics of ecological events across large geographical scales and is robust to temporal bias in recording effort. These results highlight the vast potential of citizen-science observation data to predict ecological phenomena across space, including in near real time. The framework is also easily applicable for ecologists and practitioners already using machine-learning and statistics-based predictive approaches.

Human-driven land use change and intensification is a major threat to global biodiversity. High levels of land management intensity may reduce species diversity, change the composition and structure of plant and animal communities and disrupt ecological processes. However, there is still scarce information on the impacts of land management intensity on island pollinator communities and their interactions with plants. Here, we aim to assess how different land use types (natural vegetation, semi-natural pastures, and intensive pastures), representing a gradient of grazing intensification, influence pollinator diversity and plant–pollinator interactions on Terceira Island (Azores). We surveyed 30 sites (10 per land use) and recorded 1453 visits by 41 pollinator species. Alpha diversity did not differ among land uses, but grazing intensification reduced the abundance of several native species while favoring some exotics, such as the honeybee. Network analyses showed changes in structural properties and declines in interactions between native species with increasing grazing disturbance. Introduced species, particularly the honeybee, dominated interactions in intensively managed habitats, replacing native species from key ecological roles. Our findings highlight the vulnerability of island ecosystems to grazing intensification and emphasize the need for conservation measures in the Azores, namely the reduction in grazing intensity, restoration of habitat connectivity, and implementation of pollinator-friendly agri-environmental schemes to enhance native biodiversity and sustain ecosystem services.