Scientific background and motivation: Prescribed burning of peatland vegetation is a controversial topic for which enhanced scientific evidence is required to inform policy . Prescribed burning of peatlands may be conducted for agricultural clearance, wildfire protection, or in the UK, to support the upland economy associated with game birds. Here, burn patch sizes tend to be small and so either not detected by existing Earth Observation fire products which are more suited to larger wildfires, or even when using higher-resolution data they can be difficult to distinguish from other vegetation features such as managed mowing or peatland restoration regrowth. There are also key research gaps (globally) in understanding how the understory vegetation influences change detection. A range of negative impacts of prescribed burning have been reported (e.g. [2-4]) but to understand how these upscale from (perhaps tens of thousands of) individual burn patches each year and to provide regulators (such as CASE partner Natural England) with tools to support compliance and detection of changes in burn management (patch shape, size, distance from water courses, timing, agreements with landholders) an automated method for burn patch detection is required. Peatland wildfires have been associated with negative consequences for air quality and human health (e.g. [5, 6]). However, there have been no observations of air quality impacts associated with prescribed patch burning which can occur across large areas in a region on given days during the burn season. The project also has potential for upscaling to wider environments, such as Boreal systems where patch-scale changes in ecology driven by management or climate change may be important for soil functioning and carbon release .
Aim: This project will develop a novel automated moorland burn patch detection technique providing users with scientific and regulatory outputs (e.g. distance from water courses, patch size distributions, timing, data related to types of landscape/soil/conservation status), and which also detects and models associated air quality impacts of prescribed moorland burning for the first time.
- To use Sentinel 2, other Earth Observation products, and machine learning techniques to accurately map and distinguish high resolution prescribed burn patches from other moorland features.
- To develop an automatically-updating (~real time) Earth Observation-driven burn detection tool that produces maps and statistical outputs related to key variables.
- To use Sentinel data, the WRF-Chem model and ground truthing to determine air quality impacts of prescribed moorland burning.
Methods: The project will be supervised by a multi-disciplinary and diverse team with a pedigree of high quality outputs, impacts and learned society awards, providing the candidate with an excellent support environment to develop their project. For Objective 1 high resolution Sentinel-2 NDVI approaches will be combined with AI feature recognition techniques to differentiate burn patches from features such as mowing and to determine the role of vegetation understory in influencing change detection. The CASE partner may support ground truthing surveys to aid analysis (also providing training for the student). The CASE partner are also keen to understand longer-term change and so coarser-scale LANDSAT data will also be interrogated providing important context. Objective 2 will require advanced computing techniques to automate the Earth Observation outputs (in real time for each Sentinel 2 pass) and provide them in useful scientific and policy-relevant formats for interrogation or alerts (e.g. policy/legal compliance). Spatial statistical analysis will seek to answer questions about changing land manager behaviour (patch distributions in space/time/form) and relationships to other soil, management (e.g. driven by the CASE partner’s needs), climatological, topographical, hydrological and other variables. The student will investigate the potential for wider upscaling of the automated tool to other global regions (e.g. tundra) where patch changes in low height mossy/shrubby vegetation might be crucial for understanding climate feedbacks. Objective 3 will utilise the time and spatial series burn mapping from Objectives 2 and 3, combined with high resolution Earth Observation atmospheric composition measurements (e.g. carbon monoxide (CO) and nitrogen dioxide (NO2) from Sentinel 5 – Precursor (S5P)), existing EO fire emission datasets (e.g. Global Fire Emissions Database (GFED)) and regional atmospheric chemistry modelling (e.g. the WRF-Chem model ). Here, model sensitivity simulations, in combination with observations, can be used to assess surface pollution levels and the associated contribution from prescribed moorland burning. These outputs can inform policy.
Location: The successful candidate will be based in the School of Geography at the University of Leeds, but with strong collaborative links via the supervisors in the School of Earth & Environment (Leeds) and the School of GeoSciences at the University of Edinburgh. The successful candidate will also spend time with CASE partner Natural England.
References:  Brown LE & Holden J 2020, J. Applied Ecol. 57, 2121-31;  Noble A et al 2018., J. Applied Ecol. 55, 559-69;  Holden J et al 2015, Water Resources Res. 51, 6472-84;  Brown LE et al 2015, Freshwater Sci. 34, 1406-25;  Graham AM et al 2020 I, Env. Research Communications. 2;  Kiely L et al 2019, Atmos. Chem. & Phy. 19, 11105–21;  Kropp H et al 2020, Env. Res. Letters 16, 015001;  Graham AM et al 2020 Env. Res. Letters 15, 074018.