Airborne Laser Scanning (ALS), also known as Light Detection and Ranging (LiDAR) enables an accurate three-dimensional characterization of vertical forest structure. ALS has proven to be an information-rich asset for forest managers, enabling the generation of highly detailed bare earth digital elevation models (DEMs) as well as estimation of a range of forest inventory attributes (including height, basal area, and volume). Recently, there has been increasing interest in the advanced processing of high spatial resolution digital airborne imagery to generate image-based point clouds, from which vertical information with similarities to ALS can be produced. Digital airborne imagery is typically less costly to acquire than ALS, is well understood by inventory practitioners, and in addition to enabling the derivation of height information, allows for visual interpretation of attributes that are currently problematic to estimate from ALS (such as species, health status, and maturity). At present, there are two limiting factors associated with the use of image-based point clouds. First, a DEM is required to normalize the image-based point cloud heights to aboveground heights; however DEMs with sufficient spatial resolution and vertical accuracy, particularly in forested areas, are usually only available from ALS data. The use of image-based point clouds may therefore be limited to those forest areas that already have an ALS-derived DEM. Second, image-based point clouds primarily characterize the outer envelope of the forest canopy, whereas ALS pulses penetrate the canopy and provide information on sub-canopy forest structure. The impact of these limiting factors on the estimation of forest inventory attributes has not been extensively researched and is not yet well understood. In this paper, we review the key similarities and differences between ALS data and image-based point clouds, summarize the results of current research related to the comparative use of these data for forest inventory attribute estimation, and highlight some outstanding research questions that should be addressed before any definitive recommendation can be made regarding the use of image-based point clouds for this application.

Over the past two decades there has been an abundance of research demonstrating the utility of airborne light detection and ranging (LiDAR) for predicting forest biophysical/inventory variables at the plot and stand levels. However, to date there has been little effort to develop a set of protocols for data acquisition and processing that would move governments or the forest industry towards cost-effective implementation of this technology for strategic and tactical (i.e., operational) forest resource inventories. The goal of this paper is to initiate this process by examining the significance of LiDAR data acquisition (i.e., point density) for modeling forest inventory variables for the range of species and stand conditions representing much of Ontario, Canada. Field data for approximately 200 plots, sampling a broad range of forest types and conditions across Ontario, were collected for three study sites. Airborne LiDAR data, characterized by a mean density of 3.2 pulses m⁻² were systematically decimated to produce additional datasets with densities of approximately 1.6 and 0.5 pulses m⁻². Stepwise regression models, incorporating LiDAR height and density metrics, were developed for each of the three LiDAR datasets across a range of forest types to estimate the following forest inventory variables: (1) average height (R²(adj) = 0.75–0.95); (2) top height (R²(adj) = 0.74–0.98); (3) quadratic mean diameter (R²(adj) = 0.55–0.85); (4) basal area (R²(adj) = 0.22–0.93); (5) gross total volume (R²(adj) = 0.42–0.94); (6) gross merchantable volume (R²(adj) = 0.35–0.93); (7) total aboveground biomass (R²(adj) = 0.23–0.93); and (8) stem density (R²(adj) = 0.17–0.86). Aside from a few cases (i.e., average height and density for some stand types), no decimation effect was observed with respect to the precision of the prediction of the majority of forest variables, which suggests that a mean density of 0.5 pulses m⁻² is sufficient for plot and stand level modeling under these diverse forest conditions across Ontario.

Airborne Laser Scanning (ALS) metrics have been used to develop area-based forest inventories; these metrics generally include estimates of stand-level, per hectare values and mean tree attributes. Tree-based ALS inventories contain desirable information on individual tree dimensions and how much they vary within a stand. Adding size class distribution information to area-based inventories helps to bridge the gap between area- and tree-based inventories. This study examines the potential of ALS and stereo-imagery point clouds to predict size class distributions in a boreal forest. With an accurate digital terrain model, both ALS and imagery point clouds can be used to estimate size class distributions with comparable accuracy. Nonparametric imputations were generally superior to parametric imputations; this may be related to the limitation of using a unimodal Weibull function on a relatively small prediction unit (e.g., 400 m²).

Over the last decade, spatially-explicit modeling of landscape-scale forest attributes for forest inventories has greatly benefitted from airborne laser scanning (ALS) and the area-based approach (ABA) to derive wall-to-wall maps of these forest attributes. Which ALS-derived metrics to include when modeling forest inventory attributes, and how prediction accuracies vary over forest types depends largely on the structural complexity of the forest(s) being studied. Hence, the purpose of this study was to (i) examine the usefulness of adding texture and intensity metrics to height-based ALS metrics for the prediction of several forest resource inventory (FRI) attributes in one boreal and two Great Lakes, St. Lawrence (GLSL) forest region sites in Ontario and (ii) quantify and compare the site and forest type variability within the context of the FRI prediction accuracies. Basal area (BA), quadratic mean diameter-at-breast height (QMD), and stem density (S) were predicted using the ABA and a nonparametric Random Forests (RF) regression model. At the site level, prediction accuracies (i.e., expressed as RMSE (Root Mean Square Error), bias, and R²) improved at the three sites when texture and intensity metrics were included in the predictor set, even though no significant differences (p > 0.05) could be detected using the nonparametric RMANOVA test. Stem density benefitted the most from the inclusion of texture and intensity, particularly in the GLSL sites (% RMSE improved up to 6%). Combining site and forest type results indicated that improvements in site level predictions, due to the addition of texture and intensity metrics to the ALS predictor set, were the result of changes in prediction accuracy in some but not all forest types present at a site and that these changes in prediction accuracy were site and FRI attribute specific. The nonparametric Kruskal–Wallis test indicated that prediction errors between the different forest types were significantly different (p ≤ 0.01). In the boreal site, prediction accuracies for conifer forest types were higher than for deciduous and mixedwoods. Such patterns in prediction accuracy among forest types and FRI attributes could not be observed in the GLSL sites. In the Petawawa Research Forest (PRF), we did detect the impact of silvicultural treatments especially on QMD and S predictions.

Our objective was to model the average wood density in black spruce trees in representative stands across a boreal forest landscape based on relationships with predictor variables extracted from airborne light detection and ranging (LiDAR) point cloud data. Increment core samples were collected from dominant or co-dominant black spruce trees in a network of 400 m² plots distributed among forest stands representing the full range of species composition and stand development across a 1,231,707 ha forest management unit in northeastern Ontario, Canada. Wood quality data were generated from optical microscopy, image analysis, X-ray densitometry and diffractometry as employed in SilviScan™. Each increment core was associated with a set of field measurements at the plot level as well as a suite of LiDAR-derived variables calculated on a 20 × 20 m raster from a wall-to-wall coverage at a resolution of ~1 point m⁻². We used a multiple linear regression approach to identify important predictor variables and describe relationships between stand structure and wood density for average black spruce trees in the stands we observed. A hierarchical classification model was then fitted using random forests to make spatial predictions of mean wood density for average trees in black spruce stands. The model explained 39 percent of the variance in the response variable, with an estimated root mean square error of 38.8 (kg·m⁻³). Among the predictor variables, P20 (second decile LiDAR height in m) and quadratic mean diameter were most important. Other predictors describing canopy depth and cover were of secondary importance and differed according to the modeling approach. LiDAR-derived variables appear to capture differences in stand structure that reflect different constraints on growth rates, determining the proportion of thin-walled earlywood cells in black spruce stems, and ultimately influencing the pattern of variation in important wood quality attributes such as wood density. A spatial characterization of variation in a desirable wood quality attribute, such as density, enhances the possibility for value chain optimization, which could allow the forest industry to be more competitive through efficient planning for black spruce management by including an indication of suitability for specific products as a modeled variable derived from standard inventory data.

In this study, we demonstrate the potential of using high spatial resolution airborne imagery to characterize the structural development stages of forest canopies. Four forest succession stages were adopted: stand initiation, young multistory, understory reinitiation, and old growth. Remote sensing metrics describing the spatial patterns of forest structures were derived and a Random Forest learning algorithm was used to classify forest succession stages. These metrics included texture variables from Gray Level Co-occurrence Measures (GLCM), range and sill from the semi-variogram, and the fraction of shadow and its spatial distribution. Among all the derived variables, shadow fractions and the GLCM variables of contrast, mean, and dissimilarity were the most important for characterizing the forest succession stages (classification accuracy of 89%). In addition, a LiDAR (Light Detection and Ranging) derived forest structural index (predicted Lorey’s height) was employed to validate the classification result. The classification using imagery spatial variables was shown to be consistent with the LiDAR derived variable (R² = 0.68 and Root Mean Square Error (RMSE) = 2.39). This study demonstrates that high spatial resolution imagery was able to characterize forest succession stages with promising accuracy and may be considered an alternative to LiDAR data for this kind of application. Also, the results of stand development stages build a framework for future wildlife habitat mapping.