Estimating stand volume in broad-leaved forest using discrete-return LiDAR: plot-based approach

Quantitative assessment of forests is important at a variety of scales for forest planning and management. This study investigated the use of small-footprint discrete-return lidar for estimating stand volume in broad-leaved forest at plot level. Field measurements were conducted at 20 sample plots in the study area in western Japan, composed of temperate broad-leaved trees. Five height variables and two density variables were derived from the lidar data: 25th, 50th, 75th, and 100th percentiles, and mean of laser canopy heights as height variables (h ₂₅, h ₅₀, h ₇₅, h ₁₀₀, h _mean); and ground fraction and only-and-vegetation fraction (d _GF, d _OVF) as density variables, defined respectively as the proportion of laser returns that reached the ground, and the proportion of only echoes (i.e., single pulse returns for which the first and last pulses returned from the same point) within vegetation points. In addition, the normalized difference vegetation index (NDVI), which is often used as an estimator for leaf area index (LAI) and above-ground biomass, was derived from multispectral digital imagery as an alternative density variable (d _NDVI). Nonlinear least-square regression with cross-validation analysis was performed with single variables and combinations; a total of 23 models were studied. The best prediction was found when h ₇₅ and d _OVF were used as independent variables, resulting in adjusted R ² of 0.755 and root-mean-square error (RMSE) of 41.90 m³ ha⁻¹, corresponding to 16.4% of the mean stand volume, better than or comparable to the prediction models of previous studies.