A hybrid method toward accurate mapping of mangroves in a marginal habitat from SPOT multispectral data

Three bands of SPOT multispectral imagery are manually stratified and subsequently analysed using the ERDAS Imagine software to map mangroves in the Waitemata Harbour of Auckland, New Zealand, a climatically marginal habitat of mangroves. The methodology in generating the mangrove distribution maps is described first. The mapped mangroves and their mapping accuracy are presented and discussed next. The mapping accuracy is achieved at 76% in the one-tiered mangrove classification scheme. If mapped as two categories based on density, the thriving, tall mangroves are more accurately mapped (72.5%) than the sporadically-canopied ones (63.3%) due to their heavy confusion with urban residential. The merging of the two classes of mangroves after classification results in an accuracy of 81.4%. In addition to mangrove density, patch size of mangrove habitats, tidal height and topographic shade are the other identified factors degrading the mapping accuracy.