Related Experiment Video
Updated: Jul 16, 2025
00:09
Leaf Area Index Estimation Using Three Distinct Methods in Pure Deciduous Stands
Published on: August 29, 2019
13.6K
Estimation of forest canopy closure in northwest Yunnan based on multi-source remote sensing data colla-boration
Wen-Wu Zhou1, Qing-Tai Shu1, Shu-Wei Wang1
1College of Forestry, Southwest Forestry University, Kunming 650224, China.
Ying Yong Sheng Tai Xue Bao = the Journal of Applied Ecology
|September 11, 2023
Summary
Forest canopy closure (FCC) estimation is crucial for forest resources and biodiversity assessment. This study successfully used ICESat-2/ATLAS data and machine learning models to accurately estimate FCC at both footprint and regional scales.
Area of Science:
- Forestry and Remote Sensing
- Geospatial Analysis
- Ecology and Biodiversity
Background:
- Forest canopy closure (FCC) is a key indicator for forest resource management and biodiversity assessment.
- Accurate and cost-effective methods for regional FCC estimation are in high demand.
- Multi-source remote sensing data integration offers a promising approach for FCC inversion.
Purpose of the Study:
- To estimate forest canopy closure (FCC) at the footprint scale using ICESat-2/ATLAS data and various machine learning models.
- To develop a regional-scale FCC estimation model by integrating footprint-scale data with multi-source remote sensing and terrain factors.
- To compare the performance of different models for FCC estimation and identify optimal parameters.
Main Methods:
- Utilized ICESat-2/ATLAS lidar data and ground-truth measurements from 54 plots for footprint-scale FCC estimation.
- Employed Bayesian optimization (BO) to improve Random Forest (RF), K-nearest neighbor (KNN), and Gradient Boosting Regression Tree (GBRT) models.
- Integrated footprint-scale FCC data with Sentinel-1/2 imagery and terrain factors for regional FCC estimation using a BO-optimized Deep Neural Network (DNN).
Main Results:
- Six key parameters extracted from ATLAS lidar footprints were identified as highly contributive for FCC estimation.
- The BO-optimized GBRT model achieved the best footprint-scale FCC estimation accuracy (R=0.65, RMSE=0.10).
- The BO-DNN model provided a regional FCC estimation accuracy of R²=0.47, outperforming ordinary Kriging (R²=0.26).
Conclusions:
- ICESat-2/ATLAS data, combined with advanced machine learning, enables accurate footprint-scale FCC estimation.
- The developed regional-scale FCC model using BO-DNN offers a low-cost, high-precision solution for mountainous areas.
- This approach provides a valuable reference for extrapolating footprint-scale estimations to regional levels for forest monitoring.
