Drone-based classification of peatland habitats in the Klinské rašelinisko peatland (Slovakia) using Natural Numerical Networks

Peatlands are ecologically significant wetland ecosystems that provide critical biodiversity support and act as major global carbon stores, yet they are increasingly threatened by degradation and climate change. This study focuses on the Klinské rašelinisko peatland in northern Slovakia, where recent revitalisation efforts require accurate and detailed habitat monitoring by remote sensing. The presented approach is based on high-resolution drone RGB imagery combined with vegetation height data derived from digital surface and terrain models. For automatic classification of the key Natura 2000 habitats, including 91D0 Bog woodlands, 7120 Degraded raised bogs, 7230 Alkaline fens, and KRO06 Mire willow scrub the Natural Numerical Network (NatNet) algorithm is used. The NatNet training and validation datasets were constructed using field-validated polygons, and validation accuracy achieved 94%. The relevancy maps constructed by NatNet provided accurate discrimination of the studied peatland habitats in the study area. In particular, the detection of 7230 fragments in areas mapped as degraded 7120 habitat is noteworthy from a restoration perspective, as it identifies candidate areas of high-quality fen vegetation that may represent early-stage habitat recovery and therefore warrant future field verification and long-term monitoring. These results demonstrate the applicability of the proposed drone–NatNet framework for detailed habitat mapping of the studied peatland.