LiDAR and Weibull modeling of diameter and basal area

Abstract

This study investigates the ability to predict forest diameter distributions from light detection and ranging (LiDAR) data using Weibull modelling for forest stands in central Ontario. Results suggest that the unimodal 2-parameter Weibull model is a promising technique for the prediction of diameter class distributions, with strong relationships evident for several subgroups (at 95% confidence, r 2 adj =0.83, 0.78, 0.88, 0.80, 0.83, and 0.65, with validation RMSE of 4.09 m 2 /ha, 0.61 stems/ha, 6.05, 0.64, 4.73, and 0.09 for basal area, stem density, and the Weibull a and b parameters for basal area and stem density, respectively). The unimodal models were found to be least effective for the irregularly shaped diameter distributions, particularly for low-density coniferous plots that have undergone shelterwood treatment. A significant improvement in results for these irregular plots was found with a finite mixture modelling approach, suggesting that finite mixture models may extend our ability to predict diameter distributions over large portions of the landscape. Key words: LiDAR, Weibull, finite mixture modeling, diameter class distributions, multiple linear regression