Researchers from the Aerospace Information Research Institute of the Chinese Academy of Sciences (CAS), in collaboration with Beijing Normal University and Fujian Normal University, have developed a 10-meter resolution dataset mapping the vertical structure of forests in Nanping City, Fujian Province, China. The dataset covers the years 2022 and 2023.

The study was recently published in Scientific Data.

To address limitations in existing spaceborne LiDAR-based canopy height products, the research team developed the dataset using a novel Bias Calibration Model combined with a Random Forest regression framework. This approach overcomes key issues including systematic measurement biases in complex terrain, spatial discontinuity, and insufficient accuracy in mountainous forest areas.

The team also incorporated SHapley Additive exPlanations (SHAP) analysis, which enhances understanding of the factors influencing canopy height predictions—such as topography, climate, and spectral features.

The dataset integrates multi-source data, including GEDI spaceborne LiDAR, Sentinel-1/2 imagery, UAV LiDAR field observations, ASTER GDEM topographic data, MODIS land surface temperature, and CHIRPS precipitation data. The result is a comprehensive, high-resolution map of forest canopy height that supports forest monitoring and management in Nanping, an ecologically critical region.

Independent validation against UAV LiDAR plot data confirmed the dataset's high accuracy: R² stood at 0.62 for both 2022 and 2023, with root mean square errors (RMSE) of 2.88 meters and 3.09 meters, respectively. The bias calibration model improved accuracy, reducing RMSE from 11.80 meters to 1.70 meters and nearly eliminating bias entirely. This improvement also boosted R² from 0.29 to 0.80, demonstrating the model's effectiveness in mitigating systematic errors caused by topographic effects on steep slopes and in complex terrain.

The study holds broad applications in forest disturbance monitoring, carbon emission assessment, and sustainable forest management planning.

"The development of this dataset marks a major step forward in forest structure monitoring, offering accurate, high-resolution data that can refine carbon stock assessments and support sustainable forest management strategies," said Prof. YAO Xiaojing, a corresponding author of the study.

This research was supported by the National Key Research and Development Plan of China and the CAS Science and Technology Service Network Initiative.