Yao Dong, Le Yang, Mymuna Islam Moon, Yang Liu, Hongke Hao
Estimation of above-ground biomass in forests of North and South mountains in Xining City using multi-source remote sensing data
Dendrobiology 2026, vol. 95: 30-50
https://doi.org/10.12657/denbio.095.003
Abstract:
This study investigates a hierarchical multi-source remote sensing framework for estimating forest aboveground biomass (AGB) in the ecologically fragile North and South Mountains of Xining City, northeastern Qinghai-Tibet Plateau. The framework integrates Sentinel-2 optical imagery, spaceborne LiDAR (GEDI, ICESat-2), airborne LiDAR, and ground surveys to establish a “plot-local-regional” estimation system. Machine learning, particularly Random Forest, was used to model multi-source data relationships. The framework demonstrates strong applicability across scales: (1) at the local scale, the random forest model based on airborne LiDAR canopy structural features achieved high precision (R² = 0.90 RMSE = 2.14 t · ha−¹); (2) at the regional scale, the fusion of GEDI canopy profile metrics and Sentinel-2 spectral indices significantly enhanced estimation accuracy (R² = 0.82, RMSE = 3.90 t · ha−¹), markedly outperforming single-data-source models; (3) machine learning algorithms, particularly Random Forest, proved effective in handling multi-source data and capturing complex nonlinear relationships; (4) the generated 25-meter resolution AGB distribution map reveals clear spatial patterns, with biomass increasing with elevation and being significantly higher on shaded slopes than on sunlit slopes. The results confirm that the proposed framework is applicable for high-precision AGB estimation in high-altitude arid regions and provides a scalable technical pathway for forest carbon monitoring and spatialized management support.
Keywords: forest aboveground biomass, multi-source remote sensing data, machine learning, collaborative inversion