Predicting LiDAR-derived biomass distributions by Weibull models in a subtropical forest

Accurate information on aboveground biomass (AGB) distributions of individual trees is critical to support sustainable forest management, maintain regional carbon cycle and mitigate climate change. Light Detection and Ranging (LiDAR) is a promising active remote sensing technology can provide reliable estimates of forest parameters. Area-based approach (ABA) is appropriate for wall-to-wall estimation of these parameters. In this study, we employed an ABA estimates of AGB by predicting individual tree AGB distributions over a subtropical forest study site. The total plot-level AGB was firstly predicted and the prediction of individual tree AGB distributions was generated by a two-parameter Weibull function. Then the fitted Weibull parameters were further estimated by LiDAR metrics. In addition, all models were assessed by regressed against LiDAR metrics in coniferous forest models. Finally, the stem density for each plot was evaluated by the parameter retrieval method with predicted total AGB and mean tree AGB derived from predicted Weibull parameters of individual tree AGB distribution. The results showed that the AGB and two Weibull parameters were generally predicted well (R
²
=0.79-0.92, rRMSE=8.46%-20.80%). For stem density estimation, the regressed model explained 76% of variability in field stem density. The relationship between predicted and reference AGB distributions when the predicted frequencies of the AGB distributions were scaled to ground-truth stem number (mean Reynolds error index eR=30.83) was relatively stronger than when predicted frequencies were scaled to stem number predicted from LiDAR data (mean eR=33.67). This study demonstrated the distributions of individual forest structural parameters can potentially contribute to enrich ABA forest attributes inventory for airborne LiDAR.