Estimating tree crown size with spatial information of high resolution optical remotely sensed imagery

Tree crown size is a critical biophysical parameter that influences carbon, water and energy exchanges between forest ecosystems and the atmosphere. This study explores the potential of using spatial information of high resolution optical imagery in estimating mean tree crown diameter on a stand basis with an Ikonos image in the Blackwood Division of Duke Forest and its surrounding areas. The theory is based on the disc scene model that the ratio of image variances at two spatial resolutions is determined by the scene structure only. The mean tree crown diameter of a stand on the ground was estimated with a circular sampling plot made in the middle of the stand. The stands were then delineated in the panchromatic band of the Ikonos image. The relationship between mean tree crown diameter with image variance at a single spatial resolution, the ratio of image variances at two spatial resolutions, and the difference of image variances at two spatial resolutions were studied for conifer and hardwood stands, respectively. It was found that the ratio of image variances at 2 m and 3 m spatial resolutions best estimate conifer tree crown diameter (R 2 = 0.7282). Though the image variance at a single resolution and the difference of image variances at two spatial resolutions are also significantly correlated to conifer tree crown diameter, the R 2 is lower. Due to the continuity in canopy structure, the approach does not work well for hardwood stands.