We propose a method for classifying and monitoring the wetland vegetation in a wide area using satellite remote sensing and a geographical information system (GIS). The study area is in Sarobetsu Mire, northern Hokkaido, Japan. The method is expected to contribute to wetland conservation from its ability to provide spatial maps of basic information. It classifies wetland vegetation by using multivariate analysis of satellite images based on groundtruth data, spectroscopic data and flora data.

more »

... used stepwise discriminant analysis to classify satellite images from Landsat (7 bands) and ASTER (9 bands) into wetland vegetation of 7 classes: sphagnum, sedge, dwarf bamboo, reed and three mixed classes. The classification maps drafted from the satellite images were found to accurately reproduce the actual vegetation. An increase in dwarf bamboo vegetation and decline in sphagnum vegetation were quantitatively identified from comparison between the 1991 and 2000 Landsat data. A monitoring method that can analyze vegetation changes in greater detail was proposed. This method sets monitoring sites and monitoring lines by superimposing polygonal and linear data on a satellite image.