The Evaluation of High Resolution Aerial Imagery for Monitoring of Bracken Fern

Abstract

The Royal Natal National Park and the Rugged Glen Nature Reserve are part of the uKhahlamba Drakensberg Park (UDP) World Heritage Site and have infestations of bracken fern (Pteridium aquilinum [L.] Kuhn). Prior image classification research on bracken fern were constrained by low resolution satellite imagery and the inability of hard classifiers to account for mixed pixels. Currently there are differing views on which season is best for mapping of bracken fern. To overcome these constraints high resolution aerial imagery of 0.5m spatial resolution and a soft classifier, fuzzy classification, were employed to identify bracken fern infestations. This study compared imagery captured in winter and spring to determine which season was better suited for the image classification of bracken fern. The winter and spring classified images produced overall accuracies of 81.4% and 94.4% with Kappa coefficients of 0.63 and 0.89 respectively. These results show that high resolution imagery in conjunction with fuzzy classification can be used to identify bracken fern and that spring is more suitable for monitoring of bracken fern as compared to winter.