Indirect monitoring of heterogeneous tropical agroforestry systems using active and passive remote sensing

Monitoring agroforestry systems remains challenging due to canopy heterogeneity and the coexistence of species with contrasting dynamics. While field-based methods offer high accuracy, they are inefficient for rapid and multitemporal structural assessments. This study integrated LiDAR and multispectral data collected using a Matrice 350 RTK equipped with a Zenmuse L2 sensor and a RedEdge-P camera. Raw LiDAR data were processed in DJI Terra v4.1 and subsequently pre-processed and corrected in TerraSolid v23.011, whereas multispectral products were generated in Agisoft Metashape Professional v2.2.1. The derived metrics indicated greater growth in System A, driven by fast-growing species, whereas System B showed an overall reduction with slight increases in the upper percentiles. In addition, MSAVI and MTVI2 were sensitive to canopy structure, while GNDVI and NDRE responded to foliage content. The agreement analysis revealed a slight bias (0.09 m) toward height overestimation by LiDAR compared to the hypsometer, with no apparent proportional error. This approach provides a replicable framework for multitemporal monitoring of structural and physiological changes in tropical vegetation, with potential for regional scaling and application in sustainable forest system management. • LiDAR and multispectral data were integrated to monitor agroforestry systems • Structural and physiological metrics captured contrasting dynamics in two systems • MSAVI and MTVI2 reflected canopy structure, while GNDVI and NDRE tracked foliage • LiDAR heights showed a slight bias (+0.09 m) compared with hypsometer field data