The Amazon rainforest contains 40% of all remaining tropical rainforest in the world, but has seen rapid deforestation since the 1960s, and as much as 40% of the Brazilian Amazon could be deforested by 2050. Land-use change is an important man-made driver of climate change. We know that deforestation will generally make the atmosphere both warmer and drier, but how these changes will affect rainfall is more complex. Climate models mostly predict that deforestation will reduce rainfall, but the amount varies from 0 to 60% across different studies. Climate models use grid boxes of 10s to 100s km, which are much larger than a typical cloud. While cloud properties can be estimated from the conditions in the grid box, calculating the amount of rainfall is very uncertain, especially in the tropics. One solution is to run a model with much smaller grid boxes, but focusing on a small region, so that clouds and the detailed deforestation patterns found in the Amazon can be represented explicitly. These studies show that the surface patterns alter local weather patterns, increasing rainfall over the deforested patches, which contradicts climate models. However, because these studies focus on smaller regions, we do not know if these local effects are important for the water cycle of the entire Amazon. This project will combine both approaches, using the increased computing power now available to simulate, for the first time, the entire Amazon basin while also explicitly representing clouds. This is a crucial improvement, because past studies have shown that resolving clouds leads to a complete change in model behaviour, greatly improving how tropical rainfall is represented, including climate extremes like flooding and droughts which have the most impact on local populations. We will use these simulations to investigate how increasing deforestation will affect rainfall over the Amazon, and how these changes compare to those caused by global climate change driven by increasing carbon dioxide levels. The project is particularly exciting because it will provide a comprehensive understanding of how deforestation affects rainfall, simulating both changes in regional climate and the local weather patterns within it which directly affect people. Tropical rainfall is a key area of research in climate modelling, because although it is the most important climatic parameter to end users, it is also the most uncertain. For example rainfall drives a number of economic sectors such as agriculture and hydroelectric power, and while deforestation is used to clear land for agriculture, reductions in rainfall could reduce the yield per hectare, negating any economic gain from increasing the agricultural area. Patterns of deforestation can also affect where it rains, which could help planners identify ways to mitigate some of the negative effects on the remaining forest. This project will engage with stakeholders in the region through workshops to improve our physical understanding in a targeted way to address global challenges which have direct relevance to many people.

Despite an 82% decline in deforestation rates in Amazonia, fires are still on the rise. 2010 has been a year of severe drought and fire in Amazonia. Over the last months (July and August) the number of fire counts has reached 80% of the 2005 values, which was characterized as the drought of the century. Through the beginning of September, fire outbreaks have intensified in southwest Amazonia, including Brazil, Peru and Bolivia. This drought has been associated with warmer than average sea surface temperature in the tropical North Atlantic Ocean, in a scenario similar to the 2005 event. Amazon river levels near Iquitos and the Rio Negro near Manaus reached their lowest levels in the last 40 years and since the records began, respectively. This drought is perhaps the strongest ever recorded in this region. Droughts of this magnitude currently occur with a low frequency; however, their intensity and frequency are likely to increase in the 21st century, increasing the risk of severe wildfires in this fire-sensitive system. Understanding the impacts of forest fires on the carbon stocks and ecophysiology following the 2010 drought event is critical because these events may be common in the future climate of Amazonia. Moreover, we are still unable to predict the occurrence and extent of these droughts, we poorly understand how they affect forest fire patterns and how these fires impact the functioning of Amazonian forests. We therefore aim to quantify the impacts of drought-mediated fires on forest carbon stocks and functioning by investigating both the extent of the 2010 drought, and its influence on forest fires. In this project we will use this drought as a proxy for future climatic conditions in the region, which is likely to increase the probability of understorey forest fires. The Amazon is the largest tropical forest and most biodiverse ecosystem on the planet, storing around 86 billion tons of carbon in its biomass (currently similar to 10-years of fossil fuel emissions). Historically, fires in Amazonia have been reported to be rare, and it is unlikely that this biome is adapted to frequent fires. The increased trend in fire outbreaks in the last decades, associated to human activities, poses a growing risk to the stability of carbon stocks, functioning and diversity of Amazonian forests. Studying the effects of drought-induced fires on closed-canopy tropical forests can provide valuable insights regarding the responses of this ecosystem to future changes in the climate and environment. With our South American and American partners, this team is uniquely positioned to evaluate the effects of forest fires in Amazonia. This proposal brings together a multi-disciplinary group of local ecologists (which are currently tracking this drought on-the-ground), fire ecologists, climatologists as well as forest carbon and remote sensing experts in order to provide an integrative analysis of the climatology of this drought, the extent of forest fires and the carbon losses associated with this event. Working as a team, we will measure the different facets of this drought in multiple scales to provide a comprehensive assessment of its impacts. We will initially quantify the spatial extent of the drought and associated forest fires based on a combination of climate and remote sensing data. The team will then implement an extensive field survey to quantify the impacts of fires on carbon stocks and the functioning of the fire-affected forests. Finally, we will generate the first basin-wide map of the 2010 drought-induced fire impact on the Amazonian carbon stocks by integrating the previous data. Due to our close interaction with local governments and communities, we anticipate that this project will not only provide scientific information to help understand and diagnose the impacts of future events, but will also provide support for the development of public policies in order to mitigate climate change impacts in this region.

This project will propose an urban grammar to describe urban form and will develop artificial intelligence (AI) techniques to learn such a grammar from satellite imagery. Urban form has critical implications for economic productivity, social (in)equality, and the sustainability of both local finances and the environment. Yet, current approaches to measuring the morphology of cities are fragmented and coarse, impeding their appropriate use in decision making and planning. This project will aim to: 1) conceptualise an urban grammar to describe urban form as a combination of "spatial signatures", computable classes describing a unique spatial pattern of urban development (e.g. "fragmented low density", "compact organic", "regular dense"); 2) develop a data-driven typology of spatial signatures as building blocks; 3) create AI techniques that can learn signatures from satellite imagery; and 4) build a computable urban grammar of the UK from high-resolution trajectories of spatial signatures that helps us understand its future evolution. This project proposes to make the conceptual urban grammar computable by leveraging satellite data sources and state-of-the-art machine learning and AI techniques. Satellite technology is undergoing a revolution that is making more and better data available to study societal challenges. However, the potential of satellite data can only be unlocked through the application of refined machine learning and AI algorithms. In this context, we will combine geodemographics, deep learning, transfer learning, sequence analysis, and recurrent neural networks. These approaches expand and complement traditional techniques used in the social sciences by allowing to extract insight from highly unstructured data such as images. In doing so, the methodological aspect of the project will develop methods that will set the foundations of other applications in the social sciences. The framework of the project unfolds in four main stages, or work packages (WPs): 1) Data acquisition - two large sets of data will be brought together and spatially aligned in a consistent database: attributes of urban form, and satellite imagery. 2) Development of a typology of spatial signatures - Using the urban form attributes, geodemographics will be used to build a typology of spatial signatures for the UK at high spatial resolution. 3) Satellite imagery + AI - The typology will be used to train deep learning and transfer learning algorithms to identify spatial signatures automatically and in a scalable way from medium resolution satellite imagery, which will allow us to back cast this approach to imagery from the last three decades. 4) Trajectory analysis - Using sequences of spatial signatures generated in the previous package, we will use machine learning to identify an urban grammar by studying the evolution of urban form in the UK over the last three decades. Academic outputs include journal articles, open source software, and open data products in an effort to reach as wide of an academic audience as possible, and to diversify the delivery channel so that outputs provide value in a range of contexts. The impact strategy is structured around two main areas: establishing constant communication with stakeholders through bi-directional dissemination; and data insights broadcast, which will ensure the data and evidence generated reach their intended users.

SUMMARY The Amazon is the most important biome of South America, harbouring extraordinarily high levels of biodiversity and providing important ecosystems services. This biome is particularly notable for evolving independently from fire and in a moist, warm climate. In recent decades, altered fire regimes and an increasingly hotter and drier climate has pushed this key biome towards ecological thresholds that will likely lead to major losses in biodiversity and ecosystem services. Similarly, the ecotonal forests at the Amazon-Cerrado transition are unique ecosystems in terms of form and function, but they may be the first to suffer large-scale tree mortality and species loss due to the combined effects of increased anthropogenic disturbance, altered fire regimes and a drier climate. Vulnerability of fire and droughts are closely intertwined in Amazonian and transitional forests because fires in this region only occur when there is water stress and a human ignition source. Thus, drought increases vulnerability to fire, but we do not yet understand the magnitude and spatial variation of these vulnerabilities. Once a forest burns there is immediate tree mortality, but recent evidence also shows a significant time-lagged mortality that can last for decades, becoming an important carbon source. However, the mechanistic processes that lead to time-lagged tree mortality in this myriad of forest ecosystems encompassing the Amazon biome and the Amazon-Cerrado transition are still poorly understood. We also lack knowledge on how these processes might vary spatially across the biome and its transition. A better understanding of the mechanisms that lead to tree mortality after fires and droughts is needed to design future policies that emphasise nature-based solutions including restoration and natural regeneration. This proposal presents a multi-level approach that aims at deciphering the mechanisms that underly vulnerability to fire and time-lagged post-fire mortality across the tropical forests in Amazon and Amazon-Cerrado transition. To achieve this aim, we will quantify fire vulnerability at three different scales and link them through an upscaling approach. First, we will identify the ecological mechanisms, reflected through functional traits, that explain why individuals and species die after fires occur. For this, we will focus on poorly understood traits that can be related to fire and/or hydraulic functioning. Second, at the community scale, we will examine how vegetation structure, community traits and microclimate affect the probability to burn, through an intensive characterisation of different vegetation types with multispectral and light detection and ranging (LIDAR) imagery. Third, we will use our our unique ground-dataset on functional traits, vegetation structure and moisture dynamics, and the latest state-of-art remotely sensed information on structure and water stress to predict the vulnerability of the Amazon forests and Amazon-Cerrado transitional forests. This information will be directly applicable for the detection of sensitive hotspots (areas particularly vulnerable to fire) through satellite products. We will deliver quantifiable early-warning metrics of ecosystem vulnerability to fire that can be mapped and incorporated into fire management policies. This is a revised version of a NERC proposal that was rejected with a score of 7 by the NERC Panel in July 2020, and we have carefully addressed the Panel's comments. Specifically, we have clarified the methodology and we have reformulated the hypotheses, so they address vulnerability to fire and not drought fire-interactions.

The importance of the greenhouse gases CO2 and CH4 for climate is well established. There is broad scientific consensus that human activities are the main driver for increasing concentrations of these greenhouse gases (GHGs), particularly over the past century. Based on accurate surface measurements we know that approximately 45% of the CO2 emitted by human activities remain in the atmosphere. The net balance is apparently being taken up by global oceans, terrestrial vegetation and soils. However, there are substantial uncertainties associated with the nature, location and strength of these natural components of the carbon cycle. The Amazon region is one of the largest forested regions in the world, representing the largest reservoir of above ground organic carbon. Amazonia is not only subject to changes in climate but also to rapid environmental change due to fast population growth and economic development causing extensive deforestation and urbanisation. Such external drivers can lead to further shifts in the carbon balance resulting in release of carbon stored in the biomass and soil to the atmosphere, with implications for accelerating the growth of atmospheric GHG concentrations and climate change. Despite its important role for the global carbon cycle, current understanding of the Amazonian, and more broadly the tropical, carbon cycle is poorly constrained by observations. These knowledge gaps result in large uncertainties in the fate of the Amazonian carbon budget under a warming climate, and consequently hamper any predictive skill of carbon-climate models. Since 2009, the Amazon region has been the focus of major UK and Brazilian research projects that aim at improving our knowledge of the Amazonian carbon cycle using detailed, but localized aircraft observations of CO2 and CH4 at a number of sites. These measurements are a great advance but they remain highly localized in space and time. Space-borne measurements have the ability to fill these observational gaps by providing observations with dense spatial and temporal coverage in regions poorly sampled by surface networks. It is essential, however, that such space-based observations are properly tied to the World Meteorological Organization (WMO) reference standard to ensure acceptance of space-based datasets by the carbon cycle community and to prevent misleading results on regional carbon budgets. The central aim of this proposal is to link the in-situ measurements with remotely sensed satellite data to establish an integrated Amazonian Carbon Observatory where satellite data complements the in situ data by filling the gaps between the in situ sites and by extending the coverage over the whole Amazon region. Satellite observations of GHGs are now available from a dedicated instrument on board the Japanese GOSAT satellite and results look very promising. However, satellite retrievals over the Amazon (and the Tropics) are intrinsically difficult and the accuracy of such GHG retrievals has not been established for this region which is a major obstacle for the exploitation of space-based data to constrain carbon fluxes over the Amazon. We propose to establish a network of Brazilian and UK researchers to bridge the gap between in-situ and remote sensing observations and communities and to evaluate the feasibility of remote sensing of GHG concentrations for the purpose of GHG flux monitoring over Amazonia to improve our understanding of the Amazonian carbon cycle and to increase our ability for observing tropical carbon fluxes. The proposed network will bring together world-class expertise to address highly relevant and timely scientific questions that will advance our understanding of the carbon cycle of the Amazon. It will strongly strengthen and expand UK and Brazilian relationships and it will help further strengthen the leading role of UK researchers in many areas relevant to this proposal.