The merged image of the high spatial and the high spectral resolution images could provide more information than each individual image. In this study, the feasibility of using the merged image to extract bambootree mixed forests was examined. To achieve this, the merged image was generated from SPOT PAN and Landsat ETM+ satellite images by applying IBS (Intensity- Hue-Saturation) transform, PCA (Principal Component Analysis), Multiplicative transform, and Brovey transform, respectively. The spectral and spatial information was thus evaluated. Among the four merged methods, IHS transform was proved the most effective to present useful information from original images. Besides, using the created merged images with all four merging methods, the land cover type was further classified by the maximum likelihood classifier method, overall extraction accuracy for the bamboo forest was shown to be 80.9%, 75.3%, 74.2% and 78.7%, respectively. Since the classified result displayed the presence of the bamboo forest in mixtures with different tree crown, the results were further interpreted with the aid of the aerial photographs. As a result, bamboo in the mixtures were determined and the occupation percentages of bamboo crown were grouped into four categories : above 90%, 70-90%, 50-70% and 30-50%. For all four categories, reclassification was performed and the extraction accuracy for bamboo forest were obtained as 80.0%, 71.1%, 62.2%, and 53.3%. The corresponding bamboo area ratio in each category against the whole bamboo area occupied by this type, mixed vegetation were 32.4%, 31.3%, 20.5% and 15.7%, respectively. Therefore, it is clear that using the merged image, a great part of the bamboo-tree mixed forest could be extracted in addition to pure bamboo forest; and it is also possible for better assessment of the current bamboo distribution situations and future expansion tendencies in local or larger basin areas.
