Continuous mapping of forest canopy height using ICESat-2 data and a weighted kernel integration of multi-temporal...

Area of Science:

• Forestry
• Remote Sensing
• Geospatial Analysis

Background:

• Forest Canopy Height (FCH) is vital for understanding forest structure.
• Spaceborne LiDAR provides accurate but discontinuous FCH data.
• Existing methods struggle to integrate point-based LiDAR with continuous remote sensing data.

Purpose of the Study:

• To generate continuous Forest Canopy Height (FCH) maps for northern Iran.
• To develop a novel Weighted Kernel (WK) approach for integrating ICESat-2 ATL08 data with multi-source remote sensing (RS) datasets.
• To improve the spatial continuity and accuracy of FCH estimations.

Main Methods:

• Integrated ICESat-2 ATL08 FCH data with Sentinel-1/2, ALOS-2 PALSAR-2, and FABDEM datasets using Google Earth Engine.
• Developed and applied a novel Weighted Kernel (WK) approach for resampling RS data to match ATL08 segments.
• Utilized machine learning models, including Random Forests, for FCH mapping.

Main Results:

• The WK approach significantly improved the representation of RS data within ATL08 segments, enhancing feature correlation by approximately 6%.
• Random Forest models achieved a high coefficient of determination (R² = 0.71) and reduced errors (RMSE = 4.92 m, MAPE = 29.95%).
• Incorporating spring and autumn Sentinel-1/2 data increased R² by 6% and decreased RMSE by 0.5 m compared to summer-only datasets.

Conclusions:

• The proposed WK approach effectively integrates point-based LiDAR and continuous RS data for accurate FCH mapping.
• The methodology successfully generated continuous FCH maps, addressing limitations of previous studies.
• This research enhances the accuracy of FCH estimations, providing valuable data for forest management and research.