Multi-scale estimation of photosynthetic capacity in larch forests using UAV hyperspectral data: From leaf to canopy

Understanding forest photosynthetic capacity is essential for monitoring carbon dynamics under global change. UAV-based imaging spectroscopy is a powerful tool for assessing canopy leaf traits, but the extension of spectral-trait relationships to the canopy scale remains unclear. This study uses UAV-based hyperspectral imaging data to evaluate the photosynthetic characteristics of larch forests across different climate zones in China. We investigate UAV-derived imaging spectroscopy for mapping canopy-level leaf physiological traits, including chlorophyll content, leaf nitrogen, and photosynthetic capacity (Vc, max and Jmax) across three distinct climate zones. High-resolution UAV imaging spectral data and ground-based leaf trait measurements, including biochemical (chlorophyll, leaf nitrogen), morphological (leaf mass per area, LMA), and physiological traits (Vc, max and Jmax), were collected from 150 tree crowns at all sites. We developed and validated models for estimating physiological traits from canopy spectra using Partial Least Squares Regression (PLSR), focusing on the transferability of leaf-level models to the canopy scale. The results show that UAV-based canopy spectra can effectively estimate canopy-level Vc, max25 (R 2 = 0.56, RMSE = 9.57 μmol CO 2 m −2 s −1 , nRMSE = 17.7 %) and Jmax25 (R 2 = 0.38, RMSE = 34.8, nRMSE = 18.6 %). Additionally, other leaf traits across all climate zones were accurately predicted, including leaf mass per area (LMA), leaf water content (LWC), chlorophyll content (Chl), nitrogen content (Narea), and phosphorus content (Parea), with R 2 values ranging from 0.30 to 0.44 and nRMSE between 18.8 % and 24.4 %. Significant differences in canopy trait variability were observed, with Vc, max25 and Jmax25 values driven by climate variability. The range of Vc, max25 (40.5–70.6 μmol CO 2 m −2 s −1 ) and Jmax25 (80.6–120.4 μmol CO 2 m −2 s −1 ) was wider at the ES site compared to the FS and TS sites, indicating that species differences have a greater impact on photosynthetic capacity. These models demonstrated good transferability, showing robust performance across forests in different climate zones with only slight differences in predictive accuracy. However, canopy structure significantly influenced spectral-trait relationships, particularly for Vc, max and Jmax. While canopy structure had a moderate impact on accuracy, canopy-scale models performed slightly lower than leaf-level models in some cases. This study offers new insights into UAV-based imaging spectroscopy for mapping canopy leaf physiological traits and emphasizes the need to understand different physiological mechanisms at the canopy scale when expanding spectral-trait relationships. • First demonstration of UAV hyperspectral imaging revealing canopy structure's impact on spectral-trait relationships. • UAV hyperspectral imaging combined with LiDAR bridges gap in canopy-scale trait estimation. • Innovative approach shows canopy structure alters spectral-physiological trait relationships (Vc,max25, Jmax25). • New methodology clarifies spectral-trait linkages across leaf to canopy scales in diverse climates. • Breakthrough in understanding canopy structure's role in remote sensing of forest ecosystem traits.