Forest canopy height mapping considering the seasonal rhythm of different dominant tree species

The accurate estimation of forest canopy height (FCH) is significant for research on carbon storage and climate change. Extrapolating FCH samples at the plot scale to the global or regional level using spatially contiguous remote sensing images is currently a common method for large-scale vegetation attribute mapping. However, the extrapolation process results in a lower FCH estimation accuracy due to signal saturation, a lack of horizontal stand distribution, and incomplete seasonal information expression. Here, a new FCH remote sensing estimation model was created using the seasonal rhythms of dominant tree species (pine, oak, walnut, and other species) in the eastern part of Dali Bai Autonomous Prefecture, Yunnan Province, China. First, a spectral curve model was used to extract seasonal information from the integrated Landsat and Sentinel-2 imagery. Following this, the seasonal rhythm and support vector machine (SVM) were adopted to identify the forest dominant tree species. Integrating active and passive remote sensing features and seasonal rhythm variables, multiple random forest-based FCH estimation models of different forest dominant tree species were constructed to reveal the distribution of FCH. Our results determined a reliable overall classification accuracy of dominant tree species at 0.92, indicating that the classification could be used for subsequent FCH modeling. The validation R² of the FCH model using active and passive remote sensing variables with the random forest algorithm was only 0.37, while the R² value increased to 0.57 when the seasonal rhythm information was included. After the FCH model was established for each forest dominant tree species, the validation R² further increased to 0.70. These results revealed that the seasonal rhythm information and forest tree species composition contribute to reliable FCH distribution mapping. This study provides a robust framework for improving FCH estimation, offering valuable insights for forest management strategies and enhancing carbon storage assessment.