Multispectral detection of European frog-bit in the South Nation River using Quickbird imagery

Abstract

European frog-bit (EFB) is an invasive species of concern in Ontario, Quebec, and northern New York (USA). The ability to manage and control EFB in these jurisdictions is limited. Improved means to rapidly detect new or emerging EFB colonies could significantly enhance management success and help minimize ecological damage from this invader. This study investigated the feasibility of using high-resolution multispectral Quickbird imagery to detect EFB for a 6 km section of the South Nation River in Ontario, Canada. The objective of this study was to determine if the spectral signature of EFB stands are separable from other wetland vegetation in situ, for a typically colonized wetland. A three-stage supervised fuzzy classification methodology based on fuzzy segmentation, feature analysis, and defuzzification was used to conduct a land cover classification involving a species-level class for EFB. Validation of classifier performance was assessed using ground truth of the percentage canopy cover per species collected through a field survey and high-resolution aerial photography. Classification results indicated this methodology produced a good approximation of EFB's spatial distribution. The land cover classification had a Kappa coefficient of 84.3%, with 81.0% and 77.9% user's and producer's accuracies, respectively, for the EFB class. The results of this study demonstrated the feasibility of using high-resolution multispectral satellite imagery for EFB detection.