Estimating forest leaf area using cover and fullframe fisheye photography: Thinking inside the circle

This study compares circular fisheye photography and destructive leaf area index (L) estimation with two alternative indirect methods for estimating L in broad-leaved forest: fullframe fisheye photography and cover photography. Fullframe fisheye photography differs from circular fisheye photography in that the images have a reduced field of view; the zenithal range of 0°–90° extends to the corners of the rectangular image, roughly doubling image resolution compared to circular fisheye images. Cover images are obtained by pointing a 70 mm equivalent focal length lens (in 35 mm format) straight upwards. Cover and L were measured in twelve stands of a 17 years old Eucalyptus marginata forest that had been planted at four initial densities: 625, 1250, 2500 and 10,000 trees per hectare. L, from destructive sampling, averaged between 2 and 2.4 for stands planted at between 1250 and 10,000 trees ha−1 but was only 1.3 for the stands planted at 625 trees ha−1. Cover photography provided good indirect estimates of L assuming a spherical leaf distribution, except in the stands with 10,000 trees ha−1. These trees appeared to have a more horizontal leaf angle based on the calculated zenithal extinction coefficient for those stands (∼0.7). Rapid and automated analysis of cover images using WinSCANOPY 2006 yielded similar results to manual image analysis using Adobe Photoshop. Estimates of L from fullframe fisheye photography were better correlated with L from destructive sampling than L estimated from circular fisheye photography, but neither performed as well as cover photography. Photographic methods that use a single threshold to separate sky and foliage appear less sensitive to the camera's gamma function than methods that use a two-value threshold. Higher resolution (>8 megapixel) cameras and better lenses, may further improve L estimation using fullframe fisheye photography. We recommend that cover photography be used for routine L estimation in broadleaf forest until it is demonstrated that fisheye methods can provide similar accuracy.