- This thesis mainly consists of two parts: (1) comparing statistical modeling methods based on the area-based approach (ABA) for predicting forest inventory attributes using airborne light detection and ranging (LiDAR) data (Chapter 2), and (2) suggesting a new methodology fusing the individual tree detection (ITD) approach and the ABA for generating tree-lists using airborne LiDAR data (Chapter 3).
Chapter 2 compared selected modeling methods used to predict five forest attributes, basal area (BA), stem volume (VOL), Lorey’s height (LOR), quadratic mean diameter (QMD), and tree density (DEN), from airborne LiDAR metrics in southwestern Oregon, USA. The selected methods included most similar neighbor (MSN) imputation, gradient nearest neighbor (GNN) imputation, Random Forest (RF) based imputation, BestNN imputation, Ordinary least square (OLS) regression, spatial linear model (SLM), and geographically weighted regression (GWR). Several performances of each method were assessed by 500 simulations with different numbers of training data. No modeling methods was always superior to the others in prediction of the forest attributes. The best method varied according to response variable, prediction type, and performance measures, even though there was a leading group (SLM, OLS, BestNN, and GWR) that always outperformed the other methods in root mean squared prediction error (RMSPE). Model’s performance was quite affected when a small number of training data was used in modeling procedure. The optimal sizes of training data were 100-150 for point prediction and 200-250 for total prediction. SLM showed its applicability to wider conditions in that it produced better performance in most cases. RF imputation produced poorer performances than the other methods, particularly with lower prediction interval coverage. This might be because RF imputation had some bias and smaller prediction standard error; RF’s poor performance did not stem from the smaller number of predictor variables.
In Chapter 3, a new approach, combining ITD and ABA, was proposed 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. The 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.