Classification and ecohydrological modelling of Irish post-extraction peatlands using Plant Functional Types: scenarios for rehabilitation

Ireland’s peatlands comprise up to 23.3% of the country’s area and up to 90% of these peatlands have been degraded [1]. It is crucial to understand these through modelling, because flooding, water quality, and carbon emission risks can be mitigated through the management of these ecosystems. Formerly extracted and degraded peatlands especially pose these risks due to increased uncertainty about the manner of their influence on biodiversity, hydrology, and water chemistry regardless of if they are restored, managed, or left alone. Ireland’s post-extraction peatlands are novel habitats, which will require more explicit parametrisation of vegetation types and quantities for adequate modelling.

This research carries out remote sensing and machine learning methods to identify habitats, subdivided into ranges of Plant Functional Types, in two rewetted Irish peatlands, which had formerly been extracted for fuel: Ballycon and All Saints, Co. Offaly. It attempts to link these habitats to calibrate the process-based model PVN in an Irish context.

Six habitat classes were generated for both sites using drone imagery and ArcGIS Pro’s Deep Learning library; in parallel, the method was applied to PlanetScope imagery at Ballycon using a Python Random Forest algorithm. Results yielded 72-77% accuracy for the different products, though this is highly dependent upon scale.

(ONGOING RESEARCH BELOW)

From this, locations were chosen to scale down to a single-dimension case at each site by ‘translating’ the habitat class for a given area into a range of areal cover per Plant Functional Type. This, alongside hydrological and geotechnical data collected in the field, can be used to develop scenarios for plant growth and carbon emissions, with potential for scaling up again to develop early-stage estimates about the whole site's carbon balance.

 

Paper on remote sensing approaches in prep. Publications relating to developing/calibrating PVN planned, but this work is ongoing.

[1] Gilet et al., 2025: https://doi.org/10.1016/j.landusepol.2025.107792