Generating Tree-Lists by Fusing Individual Tree Detection and Nearest Neighbor Imputation Using Airborne LiDAR Data

Abstract

Individual tree detection (ITD) and the area-based approach (ABA) are combined to generate tree-lists using airborne LiDAR data. ITD based on the Canopy Height Model (CHM) was applied for overstory trees, while ABA based on nearest neighbor (NN) imputation was applied for understory trees. Our approach is intended to compensate for the weakness of LiDAR data and ITD in estimating understory trees, keeping the strength of ITD in estimating overstory trees in tree-level. We investigated the effects of three parameters on the performance of our proposed approach: smoothing of CHM, resolution of CHM, and height cutoff (a specific height that classifies trees into overstory and understory). There was no single combination of those parameters that produced the best performance for estimating stems per ha, mean tree height, basal area, diameter distribution and height distribution. The trees in the lowest LiDAR height class yielded the largest relative bias and relative root mean squared error. Although ITD and ABA showed limited explanatory powers to estimate stems per hectare and basal area, there could be improvements from methods such as using LiDAR data with higher density, applying better algorithms for ITD and decreasing distortion of the structure of LiDAR data. Automating the procedure of finding optimal combinations of those parameters is essential to expedite forest management decisions across forest landscapes using remote sensing data.

Fusing diameter distributions predicted by an area-based approach and individual-tree detection in coniferous-dominated forests

“…In cases where the ErrorIndex value did not decrease, the F.BEST resulted in a similar distribution than seen in the distributions predicted by ABA or ITD. The results showed that the fusion parameters depend on the characteristics of the target forests, as has been reported by Shin and Temesgen (2018). The optimization suggested an ABA diameter distribution for 37 % of the plots and ITD diameter distribution for 11 %.…”

“…The filtering sizes are in line with those used in . Moreover, the CHM cell size and the height cut-off related to the ALS height measurements have an effect on the detection of trees (Shin and Temesgen 2018). Predicted stem number was always under-estimated with ITD (Table 4).…”

“…The approach improved the root mean squared error (RMSE) values associated with volume and stem number prediction, although the prediction of the lower canopy layer was solely based on a Weibull distribution for which the parameters were also derived from ITD information. More recently, Lindberg et al (2010), Xu et al (2014), Hou et al (2016), and Shin and Temesgen (2018) have proposed methods for the fusion of ITD and NN-based ABA predictions. Lindberg et al (2010) predicted diameter distributions using ITD, and the predicted diameter distributions were afterwards calibrated to match the plot-level forest attributes predicted with ABA.…”

“…The results of Xu et al (2014) and Hou et al (2016) indicated that a calibration framework based on the cut point improves the RMSE values associated with predicted timber assortment volumes. Shin and Temesgen (2018) divided diameter distributions into ABA and ITD parts according to the different height percentiles derived from ALS data. Thus, the idea was more or less similar to the approach proposed by Xu et al (2014) and Hou et al (2016).…”

“…Thus, the idea was more or less similar to the approach proposed by Xu et al (2014) and Hou et al (2016). Shin and Temesgen (2018) examined the effect of the parametrization of ITD algorithms and the height cut point on the performance of the fused diameter distributions predicted by ITD and ABA. The main outcome of their study was that parameter combinations affect the predictive performance of diameter distributions, but none of the parameter combinations are explicitly superior.…”