The construction of major hydroelectric dams is one of the most important current drivers of habitat loss in lowland tropical forests, where the ratio of megawatts of hydropower produced per unit of flooded area is notoriously low. At least 662,000 ha of primary forests were inundated by the nine mega-hydroelectric dams constructed to date across the Brazilian Amazon, and 10 additional major dams will be built by 2022. The hydroelectric energy sector promotes widespread erosion of forest fauna and flora due to conversion of large tracts of forest into islands embedded within a unsuitable freshwater matrix and high deforestation rates throughout the neighbouring reservoir areas. Given escalating investments in hydropower, assessing the effects of mega-dams on forest biodiversity persistence has become a high research priority in tropical forest conservation. The environmental impact of the Balbina Hydroelectric Dam (BHD) in the Central Amazon has been widely considered to be disastrous; <50% of the estimated power supply at the time of construction (1986) is now generated at the expense of 236,000ha of continuous forests that were reduced to an archipelago of ~3,500 islands. However, this experimental landscape provides a unique opportunity to examine biotic responses to habitat fragmentation and isolation. In addition to the long-term relaxation time, the Balbina Dam presents several advantages compared to other fragmented landscapes including a large number of replicate islands, a homogeneous habitat matrix, effective protection from logging and hunting, and partial logistical support from the Uatumã Biological Reserve which manages the reservoir area. Here, we propose to examine how both terrestrial and arboreal vertebrate populations (mammals, birds and reptiles) respond to drastic post-isolation alteration in landscape structure in the Balbina reservoir, and the synergistic interaction of forest disturbance and forest isolation. Quantitative surveys will be conducted at 32 sites using a combination of seven sampling techniques: line-transect censuses, point-counts, camera trapping, track-surveys, enclosed track stations, armadillo burrow counts, and automated digital recordings of the diurnal and nocturnal fauna. Patterns of species persistence and community structure will be quantified and related to habitat structure and composition (forest basal area, canopy gap fraction, canopy height, understorey density, density of live/dead trees and floristic diversity) and different patch and landscape metrics (island size, shape, isolation, land cover). Forest canopy fracture will be assessed using digital hemispherical photographs coupled with high resolution satellite images. This study will document the patterns of local extinction in vertebrate assemblages within a true lacustrine island system and predict species richness and composition across the entire Balbina archipelago using modified species-area relationships. Using an 'analytical toolkit', results from this study will also inform pre-construction environmental impact assessments and licensing standards of planned hydroelectric dams projected for other Amazonian river basins, provided that the dam location and maximum water-level are known and digital elevation (DE) data for the upstream flooded area can be made available. This will allow the development of a predictive framework with which the tradeoffs between hydropower generation and biodiversity erosion can be evaluated for a range of proposed hydroelectric dam project sites.

Tropical forests hold more species of plant and animal than any other kind of terrestrial environment, and store large amounts of greenhouse gases in their trees and soils. Yet most of us are aware that they are also highly threatened by human activities, with media attention often focussing on deforestation - when forests are replaced with alternative land-uses, such as agriculture and cattle ranching. However, forests are also being modified in other ways, when trees are felled for the commercial extraction of timber, or when forest burn in abnormally dry years. These events are known as forest degradation, and affect a larger area of land than deforestation alone. The widespread nature of forest degradation means it is very important to understand whether these human-modified forests are performing similar roles as intact primary forests. How much carbon and nitrogen do they hold, and are these nutrients cycled between the leaves and the forest floor at similar rates as in primary forests? Can these ecosystem processes by predicted by characteristics of the vegetation itself (such as leaf shape and format, and the rate it carries out photosynthesis). And crucially, what are the implications of these changes for the future of these forests - are they able to resist additional modification? This project will answer these questions in two separate Brazilian biomes, the Atlantic Forests of Sao Paulo and the Amazon forests near the city of Santarem. The data we collect in two years of fieldwork will be used to improve our understanding of forest functioning, and can help develop computer simulations of forests. These simulations can then be used to examine how forests may respond to changes in climate, or other human impacts such as logging or fire. These forests are also crucial for biodiversity conservation, as many rare and endemic species are only found in landscapes where forests have already been heavily modified by humans. It is important to assess to what extent they help conserve these species, and what factors could be managed to improve their conservation value. Tropical forests hold a bewildering number of species, and so many of these species are yet to be described. It is therefore important to focus on groups of species which are well known, making birds and plants are two ideal species groups. The detailed work on forest functioning will take place in a limited number of forest plots, as we are limited by the many precise measures that need to be taken over time. In contrast, biodiversity is much quicker to sample, allowing us to examine much larger areas of around one million hectares in the Amazon and in the Atlantic Forest. As well as examining biodiversity in these landscapes, this project will also assess changes in species traits, which are characteristics that link species to the many tasks they perform in nature. By doing so, we will be able to examine the extent to which human-modified forests are losing key ecosystem processes, such as pollination from long-beaked hummingbirds, or the ability of trees to assimilate and store large quantities of carbon. This will provide us with a much better idea of how the many different kinds of human activity are affecting biodiversity, which is important if we are to design landscapes that help protect the many species of conservation concern. For too long, important scientific knowledge has remained locked away in learned journals, and has failed to inform and influence policies. We are determined not to let this happen with our research, as we believe it will produce important insights that can help us preserve the ecological stability of tropical forests and the biodiversity they contain. To facilitate these impacts, we will make every effort to disseminate our findings. These activities include producing a series of short films for YouTube, linking with local schools, and writing policy briefs.

Terrestrial ecosystems currently absorb one quarter of the carbon dioxide that Humankind releases into the atmosphere, thus reducing the rate of climate change. In this context, Amazon rainforest is extremely important, absorbing more than half a billion tonnes of carbon per year. This represents more than the combined emissions from the USA and China. However, we have limited understanding of how the productivity of Amazon forests is controlled, and this reduces our ability to predict what will happen in the future as atmospheric CO2 concentrations continue to rise and the climate changes. One of the main paradigms in ecology is that the productivity of tropical ecosystems, which occur on old, highly-weathered soils, is limited by the availability of phosphorus. This contrasts with more temperate ecosystems whose productivity has been shown to be limited by nitrogen availability. However, the phosphorus paradigm has not been tested in detail as there have been very few nutrient manipulation studies in tropical forests, and no large-scale study has been carried out in Amazon forest. This is a major issue because soil nutrient availability in most of Amazonia is substantially lower than in Panama, the location of the only ongoing fertilisation experiment in tropical lowland rainforest. Thus, the Panama findings may not be representative of large areas of Amazonia, and, therefore, our understanding of the role soil fertility plays in controlling tropical forest productivity is incomplete. Testing the phosphorus paradigm in Amazonia is critical for two reasons. Firstly, eastern and central Amazonia, the area which contains the lowest fertility soils, is considered to be at risk from the adverse effects of climate change, with widespread dieback predicted by some scientists. The resilience of these forests is considered to be highly dependent on whether trees are able to increase their growth in response to rising atmospheric CO2 concentrations, and this ability is likely to depend on the extent to which their growth is currently limited by soil nutrient availability. Secondly, there is growing evidence that the response of ecosystems to global change may differ depending on which nutrient limits their productivity. Therefore, establishing the first large-scale nutrient manipulation study in Amazonia should represent one of greatest priorities for ecosystem and climate change research. We will do just that, manipulating nitrogen, phosphorus and cation availability in central Amazon forest, at a site representative of the most common soil type in the Basin, and will quantify the response of key forest processes. We will determine the impacts on photosynthesis, plant respiration, biomass production and turnover, and decomposition, ultimately allowing us to take a full-ecosystem approach to establish how carbon storage has been affected. The new knowledge and understanding which we generate will be used to improve Amazon process representation in the Joint UK Land Environment Simulator (JULES). This will be the first time that multi-nutrient control of tropical forest function has been included in a dynamic global vegetation model, allowing for more realistic simulation of the response of the Amazon carbon cycle to environmental change. This will improve our ability to predict how the Amazon rainforest will change during the 21st century and what the implications will be for rates of regional and global climate change. In summary, our project will address a fundamental ecological question and will improve greatly our understanding of an issue that contributes substantially to uncertainty in predictions of rates of 21st century climate change; namely, how the productivity of one of the most important natural carbon sinks on the planet, the Amazon rainforest, is controlled.

Wildfires are becoming the new normal across Amazonia. Deforestation is transforming the region at a rate of around 10,000 square km/year (half the area of Wales), and now the area degraded annually -forest logged and burned but not cut down-is greater than the area deforested. Fire has historically been rare in Amazonia, meaning that the forests are not adapted to fire and the trees often die from fires - releasing carbon (C) back to the atmosphere and amplifying global climate change. Burning of tropical forests is already releasing more climate-warming carbon dioxide than fossil fuel burning in the whole of Europe. Trees in Amazonia contain around 7x more C than humans are releasing every year, and soils contain the same amount again, so it is vital to understand what is happening to this C and minimize emissions. As vegetation sheds its leaves, branches, and roots, or dies, some of the C released remains in the soil, and some is later decomposed and released back to the atmosphere. Carbon exists in the soil in many different forms, from new inputs from decomposing plant material to ancient C formed over millennia. Burning adds pyrogenic carbon (PyC) to the soil, a partially burnt form of C that is resistant to decomposition and could make the soil more fertile. Because soil C takes a long time to form, its conservation is particularly important. Despite the widespread increase in fire in Amazonia, there have been few measurements of soil C fractions and dynamics in burned areas - most have focussed on natural forests. Burned forests will have different composition, forest structure, and C dynamics. Understanding how different soil C fractions are formed and lost is crucial to understand how fire and climate change affect C storage. We propose to make major advances in understanding fire impacts, including the processes that affect the type and quantifies of soil C formed, and how C gains/losses vary over time, with soil type, and climate. We will combine new measurements with innovative modelling to inform land management strategies and C budgets. We have already collected data from across Amazonia in intact forests that have not recently burned. Crucially our project will collect a new, comprehensive dataset from human-modified forests, including logged, burned and abandoned land. We will use an approach known as a chronosequence, where we take samples at sites that were burnt at different times in the past, so we can see how the soil C has changed after e.g. 1 year, 2 years, or up to 20 years after a fire. This will then be used to develop a state-of-the-art land surface model, JULES, which forms part of the UK Earth System Model. At our sample sites, we will evaluate how different burn severities affect soil C, both in surface and deep soils, and how these change over time post-burning and with soil, climate, and land-use such as logging. At 3 focal sites, we will take detailed measurements of the decomposition rate of the C over 4 years, comparing measurements with different land-use, burn severity and wet vs dry seasons. Knowing what forms C takes after a fire and how fast it decomposes under different conditions will enable us to build these processes into the JULES model. We will model PyC globally for the first time and make projections of land C changes in Amazonia over the next ~40-60 years under different management practices. As well as transforming scientific understanding of post-fire soil C and its resilience to climate and management, our project will inform socio-environmental planning for sustainable resource use to conserve soil C. We will work with regional partners, fire managers, state and national policymakers to integrate our findings into decision-making to minimise negative fire impacts. Due to the Amazon Basin-scale of our work, these strategies are a crucial step to limit the risk of large-scale loss of soil C.