Assessing species-level biases in tree heights estimated from terrain-optimized leaf-off airborne laser scanner (ALS) data

Abstract

Canopy height is an important metric in forest research and management with uses that include estimating stand volume, scheduling silvicultural treatments, and inferring site quality. In recent years, airborne laser scanner (ALS) data have been frequently used to model canopy height continuously and remotely across large areas. A number of studies have demonstrated that ALS is effective in this regard when collected during leaf-on conditions; however, relatively few studies have investigated the accuracy of leaf-off ALS in modelling canopy height. In this article, we assessed species-level biases in heights estimated from terrain-optimized leaf-off ALS data (1.5 points m–2). We focused on several deciduous and coniferous species common to the forests of the northeastern USA. Our study area included 13 sites located in the temperate deciduous forests of eastern Connecticut. Tree heights were measured in the field for 1192 trees which included 17 deciduous and 2 coniferous species. For one site, terrestrial laser scanner (TLS) data were collected and used to estimate tree heights. The ALS data were used to create a 1 m resolution canopy height model (CHM)ALS in which cell values corresponded to the heights of the highest returns. The (CHM)ALS underestimated tree heights with a median difference of approximately 1.3 m when compared to field-based measurements. Height biases ranged from approximately 0.1 to 2.1 m with the smallest bias for black cherry, red maple, shagbark hickory, and black oak and the largest for white ash, red oak, and white oak. We found no significant differences in bias corresponding to species’ leaf-types (i.e. simple, compound, needle). Biases in tree height estimates increased substantially as the (CHM)ALS cell size increased above 1 m. Our study suggests that leaf-off ALS data with a density > 1 point/m2 can be used to estimate tree heights with relatively small bias regardless of the species type.