Past human influences on Amazonian forest: The marks of prehistoric human societies on tropical forests can still be detected today. Levis et al. performed a basin-wide comparison of plant distributions, archaeological sites, and environmental data. Plants domesticated by pre-Columbian peoples are much more likely to be dominant in Amazonian forests than other species. Furthermore, forests close to archaeological sites often have a higher abundance and richness of domesticated species. Thus, modern-day Amazonian tree communities across the basin remain largely structured by historical human use. Science , this issue p. 925

Thermal sensitivity of tropical trees: A key uncertainty in climate change models is the thermal sensitivity of tropical forests and how this value might influence carbon fluxes. Sullivan et al. measured carbon stocks and fluxes in permanent forest plots distributed globally. This synthesis of plot networks across climatic and biogeographic gradients shows that forest thermal sensitivity is dominated by high daytime temperatures. This extreme condition depresses growth rates and shortens the time that carbon resides in the ecosystem by killing trees under hot, dry conditions. The effect of temperature is worse above 32°C, and a greater magnitude of climate change thus risks greater loss of tropical forest carbon stocks. Nevertheless, forest carbon stocks are likely to remain higher under moderate climate change if they are protected from direct impacts such as clearance, logging, or fires. Science , this issue p. 869

Aim: Spatial and temporal biases in species‐occurrence data can compromise broad‐scale biogeographical research and conservation planning. Although spatial biases have been frequently scrutinized, temporal biases and the overall quality of species‐occurrence data have received far less attention. This study aims to answer three questions: (1) How reliable are species‐occurrence data for flowering plants in Africa? (2) Where and when did botanical sampling occur in the past 300 years? (3) How complete are plant inventories for Africa?Location: Africa.Methods: By filtering a publicly available dataset containing 3.5 million records of flowering plants, we obtained 934,676 herbarium specimens with complete information regarding species name, date and location of collection. Based on these specimens, we estimated inventory completeness for sampling units (SUs) of 25 km × 25 km. We then tested whether the spatial distribution of well‐sampled SUs was correlated with temporal parameters of botanical sampling. Finally, we determined whether inventory completeness in individual countries was related to old or recently collected specimens.Results: Thirty‐one per cent of SUs contained at least one specimen, whereas only 2.4% of SUs contained a sufficient number of specimens to reliably estimate inventory completeness. We found that the location of poorly sampled areas remained almost unchanged for half a century. Moreover, there was pronounced temporal bias towards old specimens in South Africa, the country that holds half of the available data for the continent. There, high inventory completeness stems from specimens collected several decades ago.Main conclusions: Despite the increasing availability of species occurrence data for Africa, broad‐scale biogeographical research is still compromised by the uncertain quality and spatial and temporal biases of such data. To avoid erroneous inferences, the quality and biases in species‐occurrence data should be critically evaluated and quantified prior to use. To this end, we propose a quantification method based on inventory completeness using easily accessible species‐occurrence data.

The remarkable biodiversity of the Brazilian Amazon is poorly documented and threatened by deforestation. When undocumented areas become deforested, in addition to losing the fauna and flora, we lose the opportunity to know which unique species had occupied a habitat. Here we quantify such knowledge loss by calculating how much of the Brazilian Amazon has been deforested and will likely be deforested until 2050 without having its tree flora sufficiently documented. To this end, we analysed 399 147 digital specimens of nearly 6000 tree species in relation to official deforestation statistics and future deforestation scenarios. We find that by 2017, 30% of all the localities where tree specimens had been collected were mostly deforested. Some 300 000 km ² (12%; 485 25 × 25 km grid cells) of the Brazilian Amazon had been deforested by 2017, without having a single tree specimen recorded. An additional 250 000–900 000 km ² of severely under‐collected rainforest will likely become deforested by 2050. If future tree sampling is to cover this area, sampling effort has to increase two‐ to six‐fold. Nearly 255 000 km ² or 7% of rainforest in the Brazilian Amazon is easily accessible but does yet but remain under‐collected. Our study highlights how progressing deforestation increases the risk of losing undocumented species of a hyper‐diverse tree flora.

Species are the currency of most biodiversity studies. However, many shortfalls and biases remain in our biodiversity estimates, preventing a comprehensive understanding of the eco-evolutionary processes that have shaped the biodiversity currently available on Earth. Biased biodiversity estimates also jeopardize the effective implementation of data-driven conservation strategies, ultimately leading to biodiversity loss. Here, we delve into the concept of the Latitudinal Taxonomy Gradient (LTG) and show how this new idea provides an interesting conceptual link between the Linnean (i.e., our ignorance of how many species there are on Earth), Darwinian (i.e., our ignorance of species evolutionary relationships), and Wallacean (i.e., our ignorance on species distribution) shortfalls. More specifically, we contribute to an improved understanding of LTGs and establish the basis for the development of new methods that allow us to: (i) better account for the integration between different shortfalls and, (ii) estimate how these interactions may affect our understanding about the evolutionary components of richness gradients at macroecological scales.