Measuring individual tree height using a combination of stereophotogrammetry and lidar Benoît St-Onge, Julien Jumelet, Mario Cobello, and Cédric Véga Abstract: Photogrammetric methods using parallaxes can be employed to measure tree heights on aerial photographs. Because it is often impossible to measure ground elevation near trees growing in dense forests, such height measure- ments remain prone to error. Our objective was to solve this problem by combining a stereomodel and a digital terrain model (DTM) produced by an airborne-scanning system that uses light detection and ranging (lidar). A stereopair of scanned aerial photographs was first registered to a lidar DTM. The elevation of the apex of 202 Thuja occidentalis (L.) individuals was measured by an observer on a digital photogrammetric workstation. The tree base elevations were read from the lidar DTM and subtracted from the corresponding apex elevations to calculate individual tree heights. These were then compared with the heights measured in the field. The average photo-lidar bias was 0.59 m, and the average deviation of 1.01 m decreased to 0.88 m when the bias was removed. It was demonstrated that the photo- graphic clearness of the tree apices influences the height error, while the density of the lidar echoes under the forest canopy does not. Using this method, retrospective studies of changes in tree height become feasible by using archived aerial photographs and recent lidar DTMs. Résumé : Des méthodes photogrammétriques utilisant la parallaxe peuvent être employées afin de mesurer la hauteur des arbres sur des photographies aériennes. Parce qu’il est souvent impossible de mesurer l’altitude du terrain près des arbres en forêt dense, de telles mesures de hauteur demeurent toutefois sujettes à erreur. Notre objectif était de ré- soudre ce problème en combinant un stéréo-modèle et un modèle numérique de terrain (MNT) produit par un système aéroporté à balayage qui utilise le lidar (détection et télémétrie par la lumière). Un couple de photographies aériennes stéréoscopiques scannées a d’abord été calé sur un MNT lidar. L’altitude de l’extrémité de la cime de 202 Thuja occi- dentalis (L.) a été mesurée par un observateur en utilisant un affichage stéréoscopique et des lunettes à obturation par cristaux liquides. L’altitude de la base des arbres a été lue sur le MNT lidar et soustraite de l’altitude des cimes corres- pondantes afin de calculer la hauteur de chaque arbre. Celle-ci a été comparé à la hauteur mesurée sur le terrain. Le biais moyen de cette méthode était de 0,59 m et l’écart moyen a été réduit de 1,01 à 0,88 m apres avoir éliminé le biais. Il a été démontré que la netteté photographique du sommet des arbres influence la précision de la méthode, alors que la densité des échos lidar au sol sous le couvert est sans effet. Avec cette méthode, des études rétrospectives sur les changements dans la hauteur des arbres deviennent possibles en utilisant des photographies aériennes d’archive et des MNT lidar récents. St-Onge et al. 2130 Introduction Tree height is an attribute of great importance in forestry and forest ecology, but its measurement is costly and time consuming. For this reason, alternatives to field measure- ments, such as the parallax method applied to aerial photo- graphs, were developed as early as 1936 (Andrews 1936). The classic parallax method consists of calculating the ele- vation difference between the apex and the base of trees ap- pearing on stereopairs of aerial photographs based on photogrammetric observations. Its accuracy depends on a number of factors, such as the resolution of the aerial photo- graphs and the precision of the photogrammetric instruments. The principal difficulty, however, is to correctly assess the ground-level elevation at, or near, the tree base. On aerial photographs of closed forest canopies, the ground is rarely visible in the immediate vicinity of most trees. Tree base el- evation is therefore replaced by ground elevation at the clos- est opening in the canopy. The topographic variations over the distance separating measured trees and ground-level ob- servations translate into tree height errors that can reach sev- eral metres. Moreover, the observer may mistake the surface of dense shrubs for bare ground and can thus underestimate tree height (Spurr 1960, p. 364). The advent of digital photo- grammetry (see, for example, Gagnon et al. 1993) has im- proved the precision of stereoscopic measurements but did not solve the ground visibility problem. Scanning laser altimetry, hereafter referred as “lidar” (LIght Detection And Ranging), can produce dense and ac- curate ground elevation measurements under forest canopies, owing to the capacity of small footprint laser pulses emitted by airborne lidars to propagate through small canopy open- Can. J. For. Res. 34: 2122–2130 (2004) doi: 10.1139/X04-093 © 2004 NRC Canada 2122 Received 20 January 2004. Accepted 11 May 2004. Published on the NRC Research Press Web site at http://cjfr.nrc.ca on 9 November 2004. B. St-Onge, 1 J. Jumelet, M. Cobello, and C. Véga. Department of Geography, Université de Québec à Montréal, P.O. Box 8888, Centre-Ville, Montréal, QC H3C 3P8, Canada. 1 Corresponding author (e-mail: St-Onge.Benoit@uqam.ca).