Zhu Xinsheng, Huang Tianbao, Liu Ziyang, Bai Lang, Yang Yongfeng, Ye Jinsheng, Wang Qiulai, Sharma Ram P, Fu Liyong
Research Institute of Forest Resource Information Techniques, Chinese Academy of Forestry, Beijing, China.
Hanan Sanya Urban Ecosystem Observation and Research, Staton, Academy of Inventory and Planning, National Forestry and Grassland Administration, Beijing, China.
Front Plant Sci. 2025 May 8;16:1532138. doi: 10.3389/fpls.2025.1532138. eCollection 2025.
Chinese fir is a crucial afforestation and timber species in southern China. Accurate estimation of its stand biomass is vital for forest resource assessment, ecological industry development, and ecosystem management. However, traditional biomass prediction methods often face limitations in terms of accuracy and efficiency, highlighting the need for more robust modeling approaches.
This study utilized data from 154 forest stands in Guangdong Province to develop biomass regression models that incorporate random effects and dummy variables. The models were based on airborne LiDAR-derived metrics. Among 41 highly correlated LiDAR variables, only two-5% cumulative height percentile and leaf area index-were retained in the final model.
The results revealed that the logistic mixed-effects model was the most effective for estimating leaf biomass, while the empirical mixed-effects model was better suited for other biomass components. Nonlinear models outperformed linear models, with the nonlinear mixed-effects model (incorporating stand age as a random effect) achieving the highest predictive accuracy. Furthermore, machine learning techniques further improved model performance (R² = 0.855 to 0.939). Validation with independent test samples confirmed the robustness and reliability of the nonlinear mixed-effects model.
This study highlights the effectiveness of airborne LiDAR data in providing efficient and precise estimates of stand biomass. It also emphasizes the significant role of stand developmental stages in biomass modeling. The findings contribute to the development of a rigorous and scalable framework for large-scale artificial forest biomass estimation, which has important implications for forest resource monitoring, ecological industry development, and ecosystem management strategies.
杉木是中国南方重要的造林和用材树种。准确估算其林分生物量对于森林资源评估、生态产业发展和生态系统管理至关重要。然而，传统的生物量预测方法在准确性和效率方面往往存在局限性，这凸显了需要更强大的建模方法。
本研究利用广东省154个林分的数据，开发了包含随机效应和虚拟变量的生物量回归模型。这些模型基于机载激光雷达衍生的指标。在41个高度相关的激光雷达变量中，最终模型仅保留了两个变量——5%累积高度百分位数和叶面积指数。
结果表明，逻辑混合效应模型在估算叶片生物量方面最有效，而经验混合效应模型更适合其他生物量组分。非线性模型优于线性模型，其中非线性混合效应模型（将林分年龄作为随机效应）具有最高的预测精度。此外，机器学习技术进一步提高了模型性能（R² = 0.855至0.939）。用独立测试样本进行验证证实了非线性混合效应模型的稳健性和可靠性。
本研究突出了机载激光雷达数据在高效精确估算林分生物量方面的有效性。它还强调了林分发育阶段在生物量建模中的重要作用。这些发现有助于建立一个用于大规模人工林生物量估算的严谨且可扩展的框架，这对森林资源监测、生态产业发展和生态系统管理策略具有重要意义。
