Vegetation mapping of Kushiro wetland in northeast Hokkaido, Japan: application of SPOT images, aerial balloon photographs and airborne color near infrared (CNIR) images for classification

Abstract

Our study was designed to evaluate the potential of the utility for various spectral and spatial resolutions to classify the wetland vegetation into the species level. We investigated which combination is the most suitable for delineating and mapping specific vegetation types. The mission we employed had three main stages, (1) Making digital processed vegetation maps by manual interpretation of mosaicking aerial balloon photos with high resolution (15 cm/pixel) and discussing about the efficiency for the classification of wetland vegetation including extensive ground truth in the summer of 1998 and 2001. (2) Classification of vegetation using airborne CNIR video images with high resolution (30 cm/pixel) based on the vegetation maps composed by balloon photos as the training data. (3) SPOT (XS and PAN) images were evaluated for mapping the whole wetland vegetation using these tools. We obtained twenty-seven categories of individual vegetation and ten typical types of vegetation community with higher accuracy classification. In addition, we could extract conservation wetland plants, twenty-two genuses and thirty-nine species and they were mapped. Each cover type was delineated from aerial balloon photos and airborne CNIR images using manual interpretation techniques. It was obviously effective to classify temperate wetland vegetation into genus and species level, especially small shrub mixed with herbaceous plants, moss bog with pools and dwarf shrubs with sedge, moss and alpine plants with the information of the status of the leaves with environmental conditions of vitality and phenology in Carex. spp. and Phragmites communis. This result is important because this level of detail types could not be retrieved without these utilizations from any of the satellite image data sets directly. We concluded that availability of high resolution training data such as balloon mosaic photos and CNIR images were obviously powerful tools to classify typical wetland vegetation into genus and species level. It was apparently suitable for delineating and mapping the specific vegetation types precisely. These high resolution training data were useful to apply to the whole wetland vegetation of SPOT images to classify typical wetland vegetation types. This paper provides the capabilities for monitoring the typical type of temperate wetland vegetation.