Evaluating the sensitivity of vegetation indices to leaf area index variability at individual tree level using multispectral drone acquisitions

Abstract

Vegetation indices (VIs) are widely applied to estimate leaf area index (LAI) for monitoring vegetation vigor and growth dynamics. However, the saturation issues in VIs caused by crown closure during the growing season pose significant challenges to the application of VIs in LAI estimation, particularly at the individual tree level. To address this, the feasibility of common VIs for LAI estimation at the individual tree level throughout the growing season was analyzed using data from digital hemispherical photography (DHP) and Unmanned Aerial Vehicle (UAV) acquisition. Additionally, the physical mechanisms underlying a generic VI-based estimation model were explored using the PROSAIL model and Global Sensitivity Analysis (GSA). Furthermore, the relationships between observed LAI derived from DHP and UAV-based VIs across different phenological development phases throughout the growing season were analyzed. The results suggested that the normalized difference vegetation index (NDVI) and its faster substitute infrared percentage vegetation index (IPVI) exhibited the best capabilities for LAI estimation (R² = 0.55 and RMSE = 0.77 for both) across the entire growing season. The LAI-VI relationship varied seasonally due to the saturation issues on VIs, with R² values increasing from the leaf budburst to the growing stage, decreasing during maturation, and rising again in the senescence stage. This indicated that seasonal effects induced by phenological changes should be considered when estimating LAI using VIs. Additionally, the saturation of VIs was influenced by soil background, leaf properties (especially leaf chlorophyll content [C_ab] and dry matter content [C_m]), and canopy structures (especially average leaf inclination angle, ALA). Compared to satellites, UAV-based sensors were more effective at mitigating spectral saturation at fine-scale due to their finer spatial resolution and narrower bandwidth. The drone-based VIs used in this study provided reliable estimates and effectively described temporal variability in LAI, contributing to a better understanding of VI saturation effects.