Assessing UAV multispectral imagery applications in peatland restoration

Restoration of peatland ecosystems is imperative for restoring and retaining the carbon absorption capacity of these systems. Unmanned aerial vehicles (UAVs) have been identified as an effective tool for ecological restoration efforts, including peatland restoration. High-resolution, UAV-collected imagery can be used to classify peatland landscape characteristics including vegetative cover type, moisture levels, and landform types. However, there remains a gap in the research related to classifying UAV imagery in order to identify peatland hags. Hags are steep, bare peat faces created by erosion. Re- profiling and revegetating these eroded landform edges are one of the primary goals of current peatland restoration efforts. This study explored the use of UAV multispectral imagery to identify hags in a degraded peatland through a supervised machine learning classification. A two-scenario classification approach was used to determine the affect the incorporation of terrain data derived from a UAV digital elevation model (DEM) had on classification model performance. The incorporation of terrain characteristics had a major effect on model accuracies and was found to be the larger factor in the enhancement of model performance compared to training and validation data created from varying portions of hag faces. Utilising lower hag faces in the training and validation data returned the most accurate classification, but the differences in accuracies obtained by all three variations were minor for classification iterations utilising terrain input data. These results demonstrate the importance of including terrain data when using UAV multispectral imagery to identify peatland hags. Additionally, the second scenario classification approach presents a method that can be used to increase efficiency in degraded peatland surveying and help progress peatland restoration efforts at landscape scales.