Protected areas are the cornerstones of global biodiversity conservation efforts 1,2 , but to fulfil this role they must be effective at conserving the ecosystems and species that occur within their boundaries. This is particularly imperative in tropical forest hotspots, regions that concentrate a major fraction of the world's biodiversity while also being under intense human pressure 3-5 . But these areas strongly lack adequate monitoring datasets enabling to contrast biodiversity in protected areas with comparable unprotected sites 6,7 .Here we take advantage of the world's largest citizen science biodiversity dataset -eBird 8 -to quantify the extent to which protected areas in eight tropical forest biodiversity hotspots are effective at retaining bird diversity, and to understand the underlying mechanisms. We found generally positive effects of protection on the diversity of bird species that are forest-dependent, endemic to the hotspots, or threatened or Near Threatened, but not on overall bird species richness. Furthermore, we show that in most of the hotspots examined this is driven by protected areas preventing both forest loss and degradation.Our results support calls for increasing the extent and strengthening the management efforts within protected areas to reduce global biodiversity loss 9-11 .

Recently, a number of studies have reported somewhat contradictory patterns of temporal trends in arthropod abundance, from decline to increase. Arthropods often exhibit non-monotonous variation in abundance over time, making it important to account for temporal coverage in interpretation of abundance trends, which is often overlooked in statistical analysis. Combining four recently analysed datasets that led to contrasting outcomes, we first show that temporal abundance variations of arthropods are nonmonotonous. Using simulations, we show non-monotony is likely to bias estimated linear abundance trends. Finally, analysing empirical data, we show that heterogeneity in estimated abundance trends is significantly related to the variation in temporal baseline of analysed time series. Once differences in baseline years, habitats and continents are accounted for, we do not find any statistical difference in estimated linear abundance trends among the four datasets. We also show that short time series produce more stochastic abundance trends than long series, making the dearth of old and long-term time series a strong limitation in the assessment of temporal trends in arthropod abundance. The lack of time series with a baseline year before global change acceleration is likely to lead to an underestimation of global change effects on biodiversity.

No abstract