Sub-pixel Canopy Phenology reveals divergent Spring transition sequences in Tropical Mixed Deciduous Forest

Mainak Mukhopadhyay , M D Behera

Vegetation phenology regulates terrestrial carbon uptake and land–atmosphere interactions, and is commonly inferred from satellite-derived vegetation indices such as the Normalised Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI). However, these observations represent spatially aggregated signals that mask phenological variability among individual tree canopies in heterogeneous forests. Here, we investigated tree-level spring phenological dynamics in a tropical moist mixed deciduous forest using multispectral PhenoCAM imagery, combining Optical(RGB) and Near-InfraRed(NIR) indices like Green Chromatic Coordinate (Gcc), Red Chromatic Coordinate (Rcc), Normalised Difference Greenness Index (NDGI), and Total Brightness (RGBsum). Three canopy archetypes—a) Early Leaf Flush, (b) Early Flowering-led-Leaf Emergence, and (c) Early Defoliation-led-Delayed Green-up —coexisting within a single satellite pixel, exhibited divergent pheno-phase sequences and timing offsets of several weeks, despite identical environmental forcing. While NDGI consistently captured structural canopy development, Gcc and Rcc reflected plant tissue pigment variation associated with chlorophyll absorption and brown biomass exposure, and RGBsum is strongly influenced by canopy structure and radiative effects. Time-series analysis revealed substantial spectral decoupling across the canopy types indicating asynchronous pheno-transitions due to biological variability and phase-space analysis showed that similar levels of canopy greenness corresponded to distinct structural and optical states across neighbouring canopies, with type-specific nonlinear trajectories and saturation behaviour. Our results demonstrated that spatial aggregation in satellite observations obscures tree-level phenological diversity, potentially biasing estimates of spring onset and ecosystem productivity, indicating the need for multi-metric and scale-aware approaches to phenological characterisation to improve the interpretation of remotely sensed vegetation dynamics in tropical forest ecosystems.