Fire is the most important disturbance agent worldwide in terms of area and variety of biomes affected, a major mechanism by which carbon is transferred from from the land to the atmosphere, and a globally significant source of aerosols and many trace gas species. Forecasting of fire risk is undertaken in many fire-prone environments to aid dry season pre-planning, and appropriate consideration of fire is also required within dynamic vegetation models that aim to examine vegetation-climate interactions in the past, present and future. Current methods of mapping fire 'risk', 'susceptabilty' or 'danger' use empirical fire danger indexes calibrated against past weather conditions and fire events. As such, they provide little information on process, are appropriate to deal only with current climate, land use and land cover change (LULCC), and are limited in their ability to be tested and constrained by EO products or other observational data (e.g. ignition 'hotspots', burned area, pyrogenic C release etc). The objective of FireMAFS is to resolve these limitations by developing a robust method to forecast fire activity (fire danger indices, ignition probabilities, burnt area, fire intensity etc) via a process-based model of fire-vegetation interactions, tested, improved, and constrained using state-of-the-art EO data products and driven by seasonal weather forecasts issued with many months lead-time. Specific aims are to: (i) develop the methodology for using EO and other observational data on vegetation (fuel) condition, fire activity and fire effects to test, improve and constrain sub-components and end-to-end predictions of a forward model of fire-vegetation interactions and to inform, test and restrict the model when used in forecast mode to ensure it is nudged along the optimum trajectory, and is furthermore reset when the observation period catches up with the prior period of prediction; (ii) drive the improved forward model by seasonal weather forecast ensembles, predicting spatio-temporal variability in fire 'danger' indices, fire occurrence and a range of subsequent fire behaviour and fire effects (intensity, rate of spread, burned area, above/below ground C stock change, and trace gas/aerosol emissions) and evaluate their usefulness for seasonal fire prediction at 1 / 6 months lead time and for prognostic studies run under future projected climate and LULCC scenarios.

Fire is the most important disturbance agent worldwide in terms of area and variety of biomes affected, a major mechanism by which carbon is transferred from from the land to the atmosphere, and a globally significant source of aerosols and many trace gas species. Forecasting of fire risk is undertaken in many fire-prone environments to aid dry season pre-planning, and appropriate consideration of fire is also required within dynamic vegetation models that aim to examine vegetation-climate interactions in the past, present and future. Current methods of mapping fire 'risk', 'susceptabilty' or 'danger' use empirical fire danger indexes calibrated against past weather conditions and fire events. As such, they provide little information on process, are appropriate to deal only with current climate, land use and land cover change (LULCC), and are limited in their ability to be tested and constrained by EO products or other observational data (e.g. ignition 'hotspots', burned area, pyrogenic C release etc). The objective of FireMAFS is to resolve these limitations by developing a robust method to forecast fire activity (fire danger indices, ignition probabilities, burnt area, fire intensity etc) via a process-based model of fire-vegetation interactions, tested, improved, and constrained using state-of-the-art EO data products and driven by seasonal weather forecasts issued with many months lead-time. Specific aims are to: (i) develop the methodology for using EO and other observational data on vegetation (fuel) condition, fire activity and fire effects to test, improve and constrain sub-components and end-to-end predictions of a forward model of fire-vegetation interactions and to inform, test and restrict the model when used in forecast mode to ensure it is nudged along the optimum trajectory, and is furthermore reset when the observation period catches up with the prior period of prediction; (ii) drive the improved forward model by seasonal weather forecast ensembles, predicting spatio-temporal variability in fire 'danger' indices, fire occurrence and a range of subsequent fire behaviour and fire effects (intensity, rate of spread, burned area, above/below ground C stock change, and trace gas/aerosol emissions) and evaluate their usefulness for seasonal fire prediction at 1 / 6 months lead time and for prognostic studies run under future projected climate and LULCC scenarios.

Fire is the most important disturbance agent worldwide in terms of area and variety of biomes affected, a major mechanism by which carbon is transferred from from the land to the atmosphere, and a globally significant source of aerosols and many trace gas species. Forecasting of fire risk is undertaken in many fire-prone environments to aid dry season pre-planning, and appropriate consideration of fire is also required within dynamic vegetation models that aim to examine vegetation-climate interactions in the past, present and future. Current methods of mapping fire 'risk', 'susceptabilty' or 'danger' use empirical fire danger indexes calibrated against past weather conditions and fire events. As such, they provide little information on process, are appropriate to deal only with current climate, land use and land cover change (LULCC), and are limited in their ability to be tested and constrained by EO products or other observational data (e.g. ignition 'hotspots', burned area, pyrogenic C release etc). The objective of FireMAFS is to resolve these limitations by developing a robust method to forecast fire activity (fire danger indices, ignition probabilities, burnt area, fire intensity etc) via a process-based model of fire-vegetation interactions, tested, improved, and constrained using state-of-the-art EO data products and driven by seasonal weather forecasts issued with many months lead-time. Specific aims are to: (i) develop the methodology for using EO and other observational data on vegetation (fuel) condition, fire activity and fire effects to test, improve and constrain sub-components and end-to-end predictions of a forward model of fire-vegetation interactions and to inform, test and restrict the model when used in forecast mode to ensure it is nudged along the optimum trajectory, and is furthermore reset when the observation period catches up with the prior period of prediction; (ii) drive the improved forward model by seasonal weather forecast ensembles, predicting spatio-temporal variability in fire 'danger' indices, fire occurrence and a range of subsequent fire behaviour and fire effects (intensity, rate of spread, burned area, above/below ground C stock change, and trace gas/aerosol emissions) and evaluate their usefulness for seasonal fire prediction at 1 / 6 months lead time and for prognostic studies run under future projected climate and LULCC scenarios.

Fire is the most important disturbance agent worldwide in terms of area and variety of biomes affected, a major mechanism by which carbon is transferred from from the land to the atmosphere, and a globally significant source of aerosols and many trace gas species. Forecasting of fire risk is undertaken in many fire-prone environments to aid dry season pre-planning, and appropriate consideration of fire is also required within dynamic vegetation models that aim to examine vegetation-climate interactions in the past, present and future. Current methods of mapping fire 'risk', 'susceptabilty' or 'danger' use empirical fire danger indexes calibrated against past weather conditions and fire events. As such, they provide little information on process, are appropriate to deal only with current climate, land use and land cover change (LULCC), and are limited in their ability to be tested and constrained by EO products or other observational data (e.g. ignition 'hotspots', burned area, pyrogenic C release etc). The objective of FireMAFS is to resolve these limitations by developing a robust method to forecast fire activity (fire danger indices, ignition probabilities, burnt area, fire intensity etc) via a process-based model of fire-vegetation interactions, tested, improved, and constrained using state-of-the-art EO data products and driven by seasonal weather forecasts issued with many months lead-time. Specific aims are to: (i) develop the methodology for using EO and other observational data on vegetation (fuel) condition, fire activity and fire effects to test, improve and constrain sub-components and end-to-end predictions of a forward model of fire-vegetation interactions and to inform, test and restrict the model when used in forecast mode to ensure it is nudged along the optimum trajectory, and is furthermore reset when the observation period catches up with the prior period of prediction; (ii) drive the improved forward model by seasonal weather forecast ensembles, predicting spatio-temporal variability in fire 'danger' indices, fire occurrence and a range of subsequent fire behaviour and fire effects (intensity, rate of spread, burned area, above/below ground C stock change, and trace gas/aerosol emissions) and evaluate their usefulness for seasonal fire prediction at 1 / 6 months lead time and for prognostic studies run under future projected climate and LULCC scenarios.

Insects are the little things that run the world (E.O. Wilson). With increasing recognition of the importance of insects as the dominant component of almost all ecosystems, there are growing concerns that insect biodiversity has declined globally, with serious consequences for the ecosystem services on which we all depend. Major gaps in knowledge limit progress in understanding the magnitude and direction of change, and hamper the design of solutions. Information about insects trends is highly fragmented, and time-series data is restricted and unrepresentative, both between different groups of insects (e.g. lepidoptera vs beetles vs flies) and between different regions. Critically, we lack primary data from the most biodiverse parts of the world. For example, insects help sustain tropical ecosystems that play a major role in regulating the global climate system and the hydrological cycle that delivers drinking water to millions of people. To date, progress in insect monitoring has been hampered by many technical challenges. Insects are estimated to comprise around 80% of all described species, making it impossible to sample their populations in a consistent way across regions and ecosystems. Automated sensors, deep learning and computer vision offer the best practical and cost-effective solution for more standardised monitoring of insects across the globe. Inter-disciplinary research teams are needed to meet this challenge. Our project is timely to help UK researchers to develop new international partnerships and networks to underpin the development of long-term and sustainable collaborations for this exciting, yet nascent, research field that spans engineering, computing and biology. There is a pressing need for new research networks and partnerships to maximize potential to revolutionise the scope and capacity for insect monitoring worldwide. We will open up this research field through four main activities: (a) interactive, online and face-to-face engagement between academic and practitioner stakeholders, including key policy-makers, via online webinars and at focused knowledge exchange and grant-writing workshops in Canada and Europe; (b) a knowledge exchange mission between the UK and North America, to share practical experience of building and deploying sensors, develop deep learning and computer vision for insects, and to build data analysis pipelines to support research applications; (c) a proof-of-concept field trial spanning the UK, Denmark, The Netherlands, Canada, USA and Panama. Testing automated sensors against traditional approaches in a range of situation; (d) dissemination of shared learning throughout this project and wider initiatives, building a new community of practice with a shared vision for automated insect monitoring technology to meet its worldwide transformational potential. Together, these activities will make a significant contribution to the broader, long-term goal of delivering the urgent need for a practical solution to monitor insects anywhere in the world, to ultimately support a more comprehensive assessment of the patterns and consequences of insect declines, and impact of interventions. By building international partnerships and research networks we will develop sustainable collaborations to address how to quantify the complexities of insect dynamics and trends in response to multiple drivers, and evaluate the ecological and human-linked causes and consequences of the changes. Crucially, this project is a vital stepping-stone to help identify solutions for addressing the global biodiversity crisis as well as research to understand the biological impacts of climate change and to design solutions for sustainable agriculture. Effective insect monitoring underpins the evaluation of future socio-economic, land-use and climate mitigation policies.