Our new report shows how Earth observation imagery can be used to monitor threatened peatland habitats.
Peatland covers over 10% of the UK’s land surface, and 80% of this is thought to be in damaged condition, due to threats such as drainage, extraction, burning and overgrazing (Bain et al. 2011; IUCN 2018). Peatlands are vital ecosystems for regulating our climate and drinking water, contributing to flood prevention, and providing habitat for nationally and internationally important species (FAO 2020). Monitoring of peatland condition has traditionally relied on ground-based surveys; however, Earth observation may provide a cost-effective dynamic method for large-scale assessment.
Our analysts have been exploring how this type of assessment can be best carried out at different scales, building upon previous work funded by the EO Centre of Excellence and Scottish Government, and carried out by the UK Centre for Ecology & Hydrology and the James Hutton institute. Our new report explores the use of different types of earth observation data, the best machine learning algorithms, and how much sampling data to use in models. Using very-high resolution (VHR) imagery across the Pennines, we created fine-scale maps of bare peat in the landscape, indicative of poor peatland condition. We then scaled this up to predict bare peat across a wider region by combining with Sentinel-2 satellite imagery, using regression modelling and machine learning techniques. Our findings highlight how this type of analysis can indicate where areas of peatland are in poor condition, whilst noting factors that influence the success of models, such as ensuring training data for the models comes from habitats in both poor and good condition. The report also highlights a need for further ground data and investigation to explore the limitations of such predictions.
Now, we are partnering with Natural England, Welsh Government and Forest Research under a Copernicus User Uptake project as part of the Caroline Herschel Framework Partnership Agreement, to apply this assessment to upland and lowland sites across England, Scotland and Wales. Here, we are developing this methodology further to detect changes in peatland condition over time, using a time series of Sentinel-2 imagery from 2015 to 2020. The aim is to assess how reliably this can detect changes and make further recommendations for how to operationalise this type of approach on a wider scale.
Having a better understanding of peatland condition and how it is changing is important to help us assess the overall status of the habitat and the ecosystem services it provides. It can also help us understand where to target restoration measures, and to assess their effectiveness, leading to better, more sustainable land management decisions.