Concerted political attention has focused on reducing deforestation, and this remains the cornerstone of most biodiversity conservation strategies. However, maintaining forest cover may not reduce anthropogenic forest disturbances, which are rarely considered in conservation programmes. These disturbances occur both within forests, including selective logging and wildfires, and at the landscape level, through edge, area and isolation effects. Until now, the combined effect of anthropogenic disturbance on the conservation value of remnant primary forests has remained unknown, making it impossible to assess the relative importance of forest disturbance and forest loss. Here we address these knowledge gaps using a large data set of plants, birds and dung beetles (1,538, 460 and 156 species, respectively) sampled in 36 catchments in the Brazilian state of Pará. Catchments retaining more than 69–80% forest cover lost more conservation value from disturbance than from forest loss. For example, a 20% loss of primary forest, the maximum level of deforestation allowed on Amazonian properties under Brazil’s Forest Code, resulted in a 39–54% loss of conservation value: 96–171% more than expected without considering disturbance effects. We extrapolated the disturbance-mediated loss of conservation value throughout Pará, which covers 25% of the Brazilian Amazon. Although disturbed forests retained considerable conservation value compared with deforested areas, the toll of disturbance outside Pará’s strictly protected areas is equivalent to the loss of 92,000–139,000 km2 of primary forest. Even this lowest estimate is greater than the area deforested across the entire Brazilian Amazon between 2006 and 2015 (ref. 10). Species distribution models showed that both landscape and within-forest disturbances contributed to biodiversity loss, with the greatest negative effects on species of high conservation and functional value. These results demonstrate an urgent need for policy interventions that go beyond the maintenance of forest cover to safeguard the hyper-diversity of tropical forest ecosystems.

Aboveground tropical tree biomass and carbon storage estimates commonly ignore tree height (<i>H</i>). We estimate the effect of incorporating <i>H</i> on tropics-wide forest biomass estimates in 327 plots across four continents using 42 656 <i>H</i> and diameter measurements and harvested trees from 20 sites to answer the following questions: <br><br> 1. What is the best <i>H</i>-model form and geographic unit to include in biomass models to minimise site-level uncertainty in estimates of destructive biomass? <br><br> 2. To what extent does including <i>H</i> estimates derived in (1) reduce uncertainty in biomass estimates across all 327 plots? <br><br> 3. What effect does accounting for <i>H</i> have on plot- and continental-scale forest biomass estimates? <br><br> The mean relative error in biomass estimates of destructively harvested trees when including <i>H</i> (mean 0.06), was half that when excluding <i>H</i> (mean 0.13). Power- and Weibull-<i>H</i> models provided the greatest reduction in uncertainty, with regional Weibull-<i>H</i> models preferred because they reduce uncertainty in smaller-diameter classes (&le;40 cm <i>D</i>) that store about one-third of biomass per hectare in most forests. Propagating the relationships from destructively harvested tree biomass to each of the 327 plots from across the tropics shows that including <i>H</i> reduces errors from 41.8 Mg ha⁻¹ (range 6.6 to 112.4) to 8.0 Mg ha⁻¹ (−2.5 to 23.0). For all plots, aboveground live biomass was −52.2 Mg ha⁻¹ (−82.0 to −20.3 bootstrapped 95% CI), or 13%, lower when including <i>H</i> estimates, with the greatest relative reductions in estimated biomass in forests of the Brazilian Shield, east Africa, and Australia, and relatively little change in the Guiana Shield, central Africa and southeast Asia. Appreciably different stand structure was observed among regions across the tropical continents, with some storing significantly more biomass in small diameter stems, which affects selection of the best height models to reduce uncertainty and biomass reductions due to <i>H</i>. After accounting for variation in <i>H</i>, total biomass per hectare is greatest in Australia, the Guiana Shield, Asia, central and east Africa, and lowest in east-central Amazonia, W. Africa, W. Amazonia, and the Brazilian Shield (descending order). Thus, if tropical forests span 1668 million km² and store 285 Pg C (estimate including <i>H</i>), then applying our regional relationships implies that carbon storage is overestimated by 35 Pg C (31–39 bootstrapped 95% CI) if <i>H</i> is ignored, assuming that the sampled plots are an unbiased statistical representation of all tropical forest in terms of biomass and height ...

Understanding the processes that determine above-ground biomass (AGB) in Amazonian forests is important for predicting the sensitivity of these ecosystems to environmental change and for designing and evaluating dynamic global vegetation models (DGVMs). AGB is determined by inputs from woody productivity [woody net primary productivity (NPP)] and the rate at which carbon is lost through tree mortality. Here, we test whether two direct metrics of tree mortality (the absolute rate of woody biomass loss and the rate of stem mortality) and/or woody NPP, control variation in AGB among 167 plots in intact forest across Amazonia. We then compare these relationships and the observed variation in AGB and woody NPP with the predictions of four DGVMs. The observations show that stem mortality rates, rather than absolute rates of woody biomass loss, are the most important predictor of AGB, which is consistent with the importance of stand size structure for determining spatial variation in AGB. The relationship between stem mortality rates and AGB varies among different regions of Amazonia, indicating that variation in wood density and height/diameter relationships also influences AGB. In contrast to previous findings, we find that woody NPP is not correlated with stem mortality rates and is weakly positively correlated with AGB. Across the four models, basinwide average AGB is similar to the mean of the observations. However, the models consistently overestimate woody NPP and poorly represent the spatial patterns of both AGB and woody NPP estimated using plot data. In marked contrast to the observations, DGVMs typically show strong positive relationships between woody NPP and AGB. Resolving these differences will require incorporating forest size structure, mechanistic models of stem mortality and variation in functional composition in DGVMs.

Mapping forest types and tree species at regional scales to provide information for ecologists and forest managers is a new challenge for the remote sensing community. Here, we assess the potential of a U‐net convolutional network, a recent deep learning algorithm, to identify and segment (1) natural forests and eucalyptus plantations, and (2) an indicator of forest disturbance, the tree species Cecropia hololeuca, in very high resolution images (0.3 m) from the WorldView‐3 satellite in the Brazilian Atlantic rainforest region. The networks for forest types and Cecropia trees were trained with 7611 and 1568 red‐green‐blue (RGB) images, respectively, and their dense labeled masks. Eighty per cent of the images were used for training and 20% for validation. The U‐net network segmented forest types with an overall accuracy >95% and an intersection over union (IoU) of 0.96. For C. hololeuca, the overall accuracy was 97% and the IoU was 0.86. The predictions were produced over a 1600 km2 region using WorldView‐3 RGB bands pan‐sharpened at 0.3 m. Natural and eucalyptus forests compose 79 and 21% of the region's total forest cover (82 250 ha). Cecropia crowns covered 1% of the natural forest canopy. An index to describe the level of disturbance of the natural forest fragments based on the spatial distribution of Cecropia trees was developed. Our work demonstrates how a deep learning algorithm can support applications such as vegetation, tree species distributions and disturbance mapping on a regional scale.