A Preliminary Study on the Method for Extracting Bamboo Groves in Chiba Prefecture, Japan Using Alos/avnir-2 Data

Well-tended bamboo groves, once an integral element in the satoyama, Japan’s traditional countryside mosaic, have over the past half century been abandoned. These abandoned groves are vigorous and are now expanding, causing various problems, including loss of diversity and damage to cropland. To control and manage these groves, an effective yet efficient system for documenting and monitoring their distribution over a wide region is required. Previous researches, using either aerial photograph or satellite images, have been implemented, but these studies suffer from problems with statistical overlearning, or require too much labor and are thus not practical for wide area application. In this study, Google Earth image and field research were used for setting training data and ALOS/AVNIR-2 satellite images were used to classify land cover in a typical countryside area that includes bamboo groves. Both maximum likelihood and decision tree classification methods were employed, with the accuracy test area completely separated from the area used for acquiring training data, thus preventing the classifiers from overlearning and improving their practicability. The accuracy was assessed by comparing the classification results with existing land cover maps. Both classification methods produced almost the same Kappa coefficient (decision tree: 0.41, maximum likelihood: 0.48). Future research can be expected to improve accuracy with higher quality training data and images. The results of this study show that ALOS/AVNIR-2 images can be a useful tool for effective yet cost-efficient mapping and monitoring of bamboo groves at the region or prefecture level. * Corresponding author. This is useful to know for communication with the appropriate person in cases with more than one author.