Newly discovered species are often threatened with extinction but in many cases have received limited conservation effort. To guide future conservation, it is important to determine the extinction risk of newly described species. Here, we test how time since formal description of a species is linked to its threat status to obtain a better insight into the possible threat status of newly described species and as yet undescribed species. We compiled IUCN Red List data for 53,808 species from five vertebrate groups described since 1758. Extinction risk for more recently described species has increased significantly over time; the proportion of threatened species among newly described species has increased from 11.9% for species described between 1758 and 1767 to 30.0% for those described between 2011 and 2020. Based on projections from our analysis, this could further increase to 47.1% by 2050. The pattern is consistent across vertebrate taxonomic groups and biomes. Current species extinction rates estimated from data of all known species are therefore highly likely to be underestimated. Intensive fieldwork to boost discovery of new species and immediate conservation action for newly described species, especially in tropical areas, is urgently required.

Threshold nonlinearities in the relationship between island area and species richness can result in dramatic declines in richness with a seemingly small decline in area near the threshold. What is not known, is whether threshold declines in richness are also accompanied by nonlinear changes in functional trait space and non‐random shifts of trait group composition in response to declining area. Plant species richness was recorded, and four functional diversity (FD) indices calculated based on 12 traits for 76 tree species in 5,082 plots (5 × 5 m) on 29 islands in the Thousand Island Lake region, Zhejiang, China. We tested for threshold nonlinearities in richness and FD relationships using segmented regression, with randomized resampling using a null model to account for potential sampling artefacts. We clustered tree species into trait groupings based on trait dissimilarity, and used multinomial segmented models to test whether threshold area effects on FD were associated with shifts in trait group composition. Species richness and three of the four FD indices showed significant threshold nonlinearities with declining island area below a threshold of ca 1.16–6.88 ha. K‐means clustering identified two trait groups, reflecting species at the ‘fast’ versus ’slow’ ends of the trait spectrum. Significant area thresholds in trait group composition were driven by increasing relative richness, but decreasing relative abundance, of species with resource‐conservation traits on islands below the threshold. Synthesis. We found a threshold collapse in richness on islands below c. 1 ha, resulting in a significant decrease in functional trait space (functional richness) and an increase in the degree of niche differentiation among species in the community (functional divergence) on small islands. Threshold effects were associated with non‐random re‐assembly of trait group composition, favouring the relative richness of species at the ‘slow’ resource‐conservation end of the trait spectrum, but the relative abundance of a few species with ‘fast’ resource‐acquisition traits. These results suggest that disturbance‐driven dynamics potentially drive a functional shift in the ecosystem state at a critical threshold. We show that a minimum critical habitat area is required to maintain functional diversity in woody plant communities.

Altered microclimatic conditions and higher disturbance at forest edges create environmental stress and modify resource gradients from edge to interior, changing the selection pressures acting on individuals. Although community‐weighted trait‐mean (CWM) shifts along edge gradients have been widely documented at the species level, it is unclear how edge effects act at the individual level, and whether the direction of intraspecific trait shifts mirrors that of CWM shifts in response to edge effects. On 20 islands in the Thousand Island Lake, China, we established 484 plots (2 × 2 m) in a stratified random design across distances of 0–128 m from the forest edge. Within each plot, we sampled leaves (n = 34 768) from within and among all 2993 individuals of 68 species and measured five leaf traits (leaf area, LA; specific leaf area, SLA; leaf dry matter content, LDMC; thickness, LT; chlorophyll content, LCC). Using generalized linear mixed models, we found that different leaf traits exhibited contrasting shifts in inter‐ versus intraspecific trait variation in response to edge effects. For SLA, LT and LCC, negative covariance between inter‐ and intraspecific trait shifts resulted in dampening of community‐wide trends compared to CWM response to edge effects. In contrast, the community‐wide trend for LDMC was reinforced due to positive covariance between inter‐ and intraspecific trait shifts, while for LA the direction of covariance shifted from negative to positive on small versus large islands. Together, edge effects alter selection regimes in reassembling plant communities. Predicting the community‐wide consequences depends on the degree to which there is negative versus positive covariance between species sorting and within‐species adaptation. The widely‐used CWM approach can mask contrasting trait selection pressures acting on individuals within local populations. Individual‐level trait variation can improve understanding of community re‐assembly trajectories in response to global environmental change.