2013 Volume 33 Issue 4 Pages 308-318
Tree height is an important parameter in forestry for evaluating stem volume, selling trees, and scheduling tree felling. Current wood prices are low, and some forests are not managed, even in the traditional “Yoshino Forestry” area in Nara Prefecture. Recently, a new approach has been introduced to reduce cutting costs. To achieve cost savings, it is important to determine the distribution of tree heights. The Advanced Land Observing Satellite (ALOS) with the onboard Panchromatic Remote-sensing Instrument for Stereo Mapping (PRISM) sensor was launched on January 24, 2006. PRISM observes land surfaces from three different directions, at a spatial resolution of 2.5 m and with a spectral range from 0.55 to 0.77μm. Digital surface models (DSMs) can be produced from PRISM data. For mountainous areas, the difference in height between the DSM and the ground height can be considered to be the crown height. This study estimated height differences using a DSM and a digital terrain model (DTM) for the Nara Prefecture area of Japan. Abnormal differences in height were recorded, including values lower than -10 m and higher than 70 m. The areas with negative height values corresponded to a lake formed by a dam, a river, and mountain slopes. Areas where heights were considered too high were affected by cloud. In Yoshino District, crown height was validated using two tree-height datasets, one consisting of tree heights measured in a field survey and the other composed of light-detection and ranging (LiDAR) data. Both datasets were compared with the forest stand database for Nara Prefecture. The crown height estimates were separated into two categories : areas with successful and failed estimates. Areas with successful estimates were often on sout-facing slopes. We hypothesize that brightness or texture differences between nadir and backward images resulted in mismatching when the DSM was calculated. For areas with successful estimates, there were differences of several meters between estimated crown heights and average tree heights in the smallest management areas. These results indicate that the estimated crown heights were not sufficiently accurate for estimating average tree heights in the smallest forest management areas. However, the crown height estimates can be used to extract areas of large-diameter trees within the cutting cycle and to note inconsistencies in tree heights compared with the forest management database. The pre-processing method needs to be improved to allow its use in areas where estimates failed.