Influence of Fusing Lidar and Multispectral Imagery on Remotely Sensed Estimates of Stand Density and Mean Tree Height in a Managed Loblolly Pine Plantation

Abstract

Abstract Stereo aerial photography has long been used to measure tree density and height photogrammetrically. Recent attempts have been made to locate and measure trees automatically in high-resolution digital imagery. This study used small-footprint lidar (1.057 µm, 1 mrad divergence, 0.67 m footprint) and high-resolution (0.61 m) multispectral (550, 675, 700, and 800 nm) data sets to estimate stem counts and tree heights in 15-yr-old loblolly pine stands. A data fusion process was used to combine the datasets. Tree identification accuracy and mean height estimation derived from the separate and fused data sets were compared against field data. Tree identification was more accurate using spectral data (78.6% and 92.4%) than lidar data (64.7% and 87.3%) within the two planting densities, respectively. The fused dataset improved accuracy of tree identification over the single-dataset approaches (83.5% and 94.8%). Plot-level mean height of lidar-located trees provided the best estimates of mean field height (average difference = 0.15 m). Missed trees for all methods were shorter than mean field height by up to 3.03 m (fused data). These results indicate fusion of spectral and lidar data will likely improve estimates of mean tree height and stem density. Increased lidar posting density is identified as a key factor to improve tree recognition and measurement. FOR. SCI.49(3):457–466.