Estimating average tree crown size using spatial information from Ikonos and QuickBird images: Across-sensor and across-site comparisons

Abstract

The forest canopy is the medium for energy, mass, and momentum exchanges between the forest ecosystem and the atmosphere. Tree crown size is a critical aspect of canopy structure that significantly influences these biophysical processes in the canopy. Tree crown size is also strongly related to other canopy structural parameters, such as tree height, diameter at breast height and biomass. But information about tree crown sizes is difficult to obtain and rarely available from traditional forest inventory. The study objective was to test the hypothesis that a model previously developed for estimation of tree crown size can be generalized across sensors and sites. Our study sites include the Racoon Ecological Management Area in southeast Ohio, USA and the Duke Forest in North Carolina Piedmont, USA. We sampled a series of circular plots in the summers of 2005 and 2007. We derived average tree crown diameter (CD) for trees with diameter at breast height (DBH) greater than 6.4 cm (2.5 in) for each sampling plot. We developed statistical models using image spatial information from Ikonos and QuickBird images as the independent variable and CD for stands in Ohio as the dependent variable. The models provide an explanation of tree crown size for the hardwood stands comparable to other approaches ( R 2 = ∼ 0.5 and RMSE = 0.83 m). Moreover, the models that estimate tree crown size using the ratio of image variances at two spatial resolutions can be applied across sensors and sites, i.e. the statistical models developed with Ikonos images can be applied directly to estimate tree crown size with QuickBird image, and the statistical models developed in Ohio can be applied directly to estimate tree crown size with images in North Carolina. These results indicate that the model developed based on image variance ratio at two spatial resolutions can be used to take advantage of existing sampling plot data and images to estimate CD with more recent images, enhancing the efficiency of forest resources inventory and monitoring.