Abstract
Blanket peatlands are globally rare, and many have been severely eroded. Natural recovery and revegetation (‘self-restoration’) of bare peat surfaces are often observed but are poorly understood, thus hampering the ability to reliably predict how these ecosystems may respond to climatic change. We hypothesised that morphometric/topographic-related microclimatic variables may be key controls on successional pathways and vegetation patterning in self-restoring blanket peatlands. We predicted the occurrence probability of four common peatland plant species (Calluna vulgaris, Eriophorum vaginatum, Eriophorum angustifolium and Sphagnum spp.) using a digital surface model (DSM) generated from drone imagery at a pixel size of 20 cm, a suite of variables derived from the DSM, and an ensemble learning method (random forests). All four species models provided accurate fine-scale predictions of habitat suitability (accuracy > 90%, area under curve (AUC) > 0.9, recall and precision > 0.8). Mean elevation (within a 1 m radius) was often the most influential variable. Topographic position, wind exposure and the heterogeneity or ruggedness of the surrounding surface were also important for all models, whilst light-related variables and a wetness index were important in the Sphagnum model. Our approach can be used to improve prediction of future responses and sensitivities of peatland recovery to climatic changes and as a tool to identify areas of blanket peatlands that may self-restore successfully without management intervention.
Original language | English |
---|---|
Journal | Ecosystems |
Early online date | 18 Dec 2018 |
DOIs | |
Publication status | Published - 2018 |
Keywords
- natural revegetation
- DSM
- eroded peatland
- restoration
- drone
- topography
- random forest