Revealing forest phenomenological heterogeneity in Yunnan using ICESat-2-derived canopy density and MODIS time series

Canopy structure (canopy density, canopy height, etc.) regulates phenological processes both directly and indirectly by influencing microenvironmental conditions such as light availability, temperature, and moisture. However, the nonlinear mechanisms through which canopy density affects phenology remain unclear due to the complexity of forest climatic conditions and interactions among multiple environmental factors. Here, we explore forest phenological heterogeneity in Yunnan Province by integrating ICESat-2 LiDAR data and MODIS time series data. Firstly, we developed a multi-source data fusion framework to enhanced LAI retrieval. Secondly, we used a random forest model to retrieval seasonal mean (SM-LAI) and amplitude (SA-LAI) LAI. Based on the spatial heterogeneity levels of SM-LAI and SA-LAI, start (SOS), end (EOS), and length (LOS) of growing season were extracted from MODIS EVI2 for the period 2001–2022. Finally, we analyzed the coupling relationships between canopy density and phenological indicators using statistical and modeling techniques. The results show that: (1) regions with lower SM-LAI and higher SA-LAI (corresponding to sparse and unstable canopy structures) exhibit earlier SOS, EOS, and shorter LOS; while regions with higher SM-LAI and lower SA-LAI (denser and more stable canopies) show delayed SOS, EOS, and longer LOS; (2) phenological indicators in lower SA-LAI (structurally sparse) areas show greater interannual variability, while structurally dense regions exhibit more synchronized phenological transitions; (3) phenological indicators show significant binomial relationships with SM-LAI and SA-LAI. This study contributes a novel approach to understanding forest phenological heterogeneity in complex plateau terrain, emphasizing how variations in canopy density influence spatially variable phenological dynamics in forest ecosystems.