Accurate Estimation of Maize Above-Ground Biomass Using Integrated Multispectral and LiDAR Data
WU Qiang1, YANG Mo-han1, DUAN Feng-hui1, WANG Zan-pu2, KANG Jia-kun1, YANG Hao3, YANG Gui-jun3, ZHANG Zhi-yong1, MA Xin-ming1, CHENG Jin-peng1*
1. College of Agronomy, Henan Agricultural University, Zhengzhou 450046, China
2. College of Information and Management Science, Henan Agricultural University, Zhengzhou 450046, China
3. Information Technology Research Center, Beijing Academy of Agriculture and Forestry Sciences; Key Laboratory for Agricultural Remote Sensing Mechanism and Quantitative Remote Sensing, Ministry of Agriculture and Rural Affairs, Beijing 100097, China
Abstract：Accurate estimation of maize Above Ground Biomass (AGB) is a core task in precision agriculture management. Spectral remote sensing technology, by capturing the reflectance characteristics of crop canopies across different wavelengths, can effectively reflect the physiological state of maize but is susceptible to interference from complex canopy structures. In contrast, Light Detection and Ranging (LiDAR) technology can acquire high-precision three-dimensional structural information of maize. Still, it has difficulty revealing the physiological characteristics of crops, leading to limitations when using single data sources for biomass estimation. Therefore, this study developed a method for estimating maize above-ground biomass that integrates multispectral and LiDAR data. The experiment was conducted from 2021 to 2022 at the Beijing Xiaotangshan Precision Agriculture Demonstration Base, collecting data from 140 sample plots covering 7 maize varieties. A P4M multispectral UAV was used to acquire canopy reflectance spectral data during key growth stages. In contrast, an M600 UAV equipped with a Riegl VUX-1 LiDAR was used to obtain three-dimensional point cloud data. Above-ground leaf biomass (AGLB), above-ground stem biomass (AGSB), and total above-ground biomass (AGB) were measured, respectively. Twelve commonly used vegetation indices including NDVI and OSAVI were extracted from multispectral data. At the same time, nine structural features including maximum height (HMax) and average height (AspAvg) were calculated from LiDAR point clouds based on Triangulated Irregular Networks (TIN). Random forest algorithms were employed to construct biomass estimation models, and feature importance was evaluated using SHapley Additive exPlanations (SHAP) method. Results showed that compared to spectral estimation models, the fusion model significantly improved biomass estimation accuracy. The coefficient of determination (R2) of the fusion model reached 0.80 (AGLB), 0.78 (AGSB), and 0.73 (AGB), representing improvements of 5.2%, 27.8%, and 12.3% respectively; root mean square error (RMSE) decreased to 61.67, 248.61, and 356.78 g·m-2, representing reductions of 8.2%, 43.6%, and 15.4% respectively. Both spectral indices and structural features showed significant correlations with biomass (r=0.52~0.83), with HMax, AspAvg, RVI, and OSAVI being key variables. SHAP analysis revealed that structural features contributed most significantly to stem biomass, while spectral indices had a greater impact on leaf biomass. This study demonstrates the complementarity and synergistic effects of the two technologies in biomass estimation, providing reliable methodological support for crop growth monitoring and above-ground biomass estimation in precision agriculture, and promoting the digitalization and intelligent development of agricultural management.
吴 强，杨默含，段锋辉，王赞朴，康家坤，杨 浩，杨贵军，张志勇，马新明，程金鹏. 融合多光谱与LiDAR的玉米地上生物量精准估算研究[J]. 光谱学与光谱分析, 2025, 45(10): 2906-2914.
WU Qiang, YANG Mo-han, DUAN Feng-hui, WANG Zan-pu, KANG Jia-kun, YANG Hao, YANG Gui-jun, ZHANG Zhi-yong, MA Xin-ming, CHENG Jin-peng. Accurate Estimation of Maize Above-Ground Biomass Using Integrated Multispectral and LiDAR Data. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2025, 45(10): 2906-2914.
