Using spatiotemporal fusion algorithms to fill in potentially absent landsat images for estimating forest aboveground biomass

Forest aboveground biomass (AGB) estimation is important for forest resource monitoring and carbon stock assessment, but cloud contamination often limits the availability of optical imagery. To address cloud-related gaps in optical imagery, this study developed a forest AGB estimation framework that couples a comparative evaluation of spatiotemporal fusion algorithms with multi-source feature integration. Specifically, STARFM, ESTARFM, and FSDAF were compared to reconstruct cloud-contaminated Landsat imagery using Landsat–MODIS data. Optical features from fused images were combined with SAR, topographic, climatic, and location variables for AGB modelling using RF, XGBoost, and KNN. Results showed that STARFM achieved the best spectral consistency, while the RF model performed best in AGB estimation, with an R2 of 0.643. The final AGB map showed higher biomass in the south and lower biomass in the north. These results indicate the feasibility and potential of integrating spatiotemporal fusion and multi-source remote sensing for AGB estimation in cloud-prone regions.