Healthy Peat: AI assisted recovery of the UK’s carbon-rich lowland landscapes
Interview date: Wednesday 28th May
Peatland landscapes are responsible for storing an estimated 25-30% of global terrestrial carbon in just 3% of global lands. Yet draining peatlands to enable agriculture can turn them from carbon stores to carbon sinks via microbially mediated degradation. The UK has a series of policies to arrest the decline of peatlands and rebuild the degraded peatland landscapes. But progress, is hampered by a lack of (a) robust measurements of gas fluxes and water measurements, (b) understanding of peatland degradation and (c) understanding of the microbial component within the peatland system.
Novel approaches to gas flux monitoring, carbon measurements and DNA are bringing down the cost and applicability to monitor these in field. Environmental DNA (eDNA) techniques is rapidly changing the knowledge of peatland microbiomes – the natural carbon engineers of peatland systems whilst novel using FTIR analysis can measure the oxidation of peatlands. However, these will never be extensive enough to monitor all global lowland peatlands and so remote sensing techniques can be used as proxies for gas emissions, microbial activity and hydrological processes. Developments in AI such as deep learning process such neural networks can be used to analyse spatial and temporal patterns in remote sensing data.
This research aims to develop novel machine learning models that can be trained on field measurements and coupled to remote sensing data for accurate remote monitoring of peatlands.
Impact
This project will enable peatland restoration in data poor environments. This is traditionally highly labour intensive to address. A key challenge is understood whether peats have the capacity for recovery based on their relative oxidation and microbial composition: remaining peat health.
The student will work with colleagues from Lincolnshire Wildlife and Fens East Peat Partnership who aim to identify and carry out restoration of peatlands across the East of England. This project using novel approaches of coupling field-based ground truthing with AI and remote sensing to establish and evidence base on the health of peat across the fens landscape. This extensive survey of the peatlands and will identify areas of potential recovery by providing and evidence base to FEPP.
Delivery
The project is divided into three work packages. The first work package, WP1, focuses on the physical basis. This involves deploying gas monitors, collecting and analysing samples for oxidation of organic carbon using FTIR, carbon, and eDNA, and using geostatistics to create a peatland oxidation and peatland microbial map of the sites.
The second work package, WP2, is centred on the remote sensing and machine learning infrastructure. It involves collating remote sensing datasets for each site, for example LiDAR, thermal imaging, multispectral/hyperspectral imaging, and synthetic aperture radar. A series of neural networks will be trained using the maps created by WP1 and the remote sensing data to create a series of models with the capability to remotely: (i) estimate peatland oxidation, (ii) estimate microbial diversity/activity, (iii) estimate water table level, (iv) gas emissions and finally (v) estimate peatland recovery based on a given forcing (e.g. raising water tables).
The third work package, WP3, involves applying the neural network model across the wider Fens catchment and a “data poor” catchment beyond the UK to demonstrate wider applicability.
Capabilities
We welcome applicants from both a Computer Science and/or Environmental Science background. Candidates should show a willingness to learn about both areas of science. Candidates should demonstrate the following capabilities:
- Technical environmental knowledge (specifically peatlands, wetlands, or related systems etc.)
- Computational skills (e.g. GIS, coding such as R or python).
- Learn new skills and acquire new knowledge.
- Stakeholder engagement (i.e. maintain professional relationships with external parties).
- Writing/ Organisational/ project management skills.
Student Development
The student will develop into an interdisciplinary scientist with expertise in environmental science, specifically low-land peats, and computer science, specifically AI. The supervisory team has been balanced to provide expertise in computers in geochemical sciences (Dr Daniel Magnone, Director of Studies), field and laboratory techniques, inc. eDNA, in peatlands and wetlands (Dr Kristen Beck) and artificial intelligence (Professor Simon Parsons). Formal supervision is held at least monthly, and the student will join the C4RG research group. They will be expected to attend weekly research group meetings facilitated to share project successes and updates with peer early career researchers.
To apply for this project, visit our Apply Now page.