Divergent responses of temperate tree species to meteorological factors revealed by UAV multispectral vegetation indices in the Qinling Mountains

Monitoring vegetation dynamics using advanced remote sensing technologies is essential for understanding ecosystem responses to environmental changes and informing sustainable management practices. Spectral vegetation indices are widely used for tracking vegetation activity, yet significant uncertainties remain in mixed temperate forests due to structural and functional differences between evergreen and deciduous species. This study leverages high-resolution unmanned aerial vehicle (UAV) multispectral imagery with five spectral bands centered at 475 nm, 560 nm, 670 nm, 720 nm, and 840 nm at a spatial resolution of 15 cm per pixel and machine learning to analyze the annual dynamics of six vegetation indices including Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Green Normalized Difference Vegetation Index (GNDVI), Red-Edge Normalized Difference Vegetation Index (NDVIre), Triangular Vegetation Index (TVI), and Soil-Adjusted Vegetation Index (SAVI), across multiple tree species in the Qinling Mountains, China. We further applied the XGBoost model to identify key climatic drivers, including photosynthetically active radiation (PAR), temperature, precipitation, CO₂, and relative humidity. The results demonstrate that UAV-based multispectral data can accurately distinguish tree genera, achieving a classification accuracy of 59.2% using a support vector machine (SVM) algorithm. Deciduous broadleaf species, such as Celtis sinensis and Ulmus pumila, exhibited strong seasonal fluctuations in vegetation indices, with peak sensitivity to PAR, whereas evergreen conifers like Cedrus deodara maintained stable index values year-round and showed stronger correlations with temperature. The XGBoost analysis revealed that CO₂ concentration was a dominant driver of NDVI dynamics, while PAR was more influential for deciduous species. GNDVI demonstrated higher climate sensitivity than NDVI across functional groups. Therefore, under identical environmental conditions, evergreen and deciduous tree species exhibit distinct phenological strategies and climatic response mechanisms. Unlike previous satellite-based studies that lacked species-level resolution, our UAV approach enables direct comparison of vegetation index dynamics and climate drivers across 20 co-occurring temperate tree species, providing new insights into functional group differentiation in heterogeneous mountain ecosystems. This study underscores the value of UAV-based remote sensing for species-level vegetation monitoring and suggests that future work should integrate hyperspectral and LiDAR sensors to better resolve canopy complexity and improve phenological assessments.