Hyperspectral and LiDAR space-borne data for assessing mountain forest volume and biomass

• Satellite hyperspectral data outperform multispectral data for mixed forest stand volume retrieval. • Aboveground biomass is retrieved with less accuracy than stand volume in mixed forests. • Forest canopy height from satellite LiDAR is a key input to increase retrieval accuracy. • Generated maps match environmental gradients and logging history. Accurate assessment and monitoring of stand volume (SV) and above-ground biomass (AGB) in mixed mountain forests is crucial for sustainable forestry, ecosystem service assessment, and climate change mitigation. While airborne multi/hyper-spectral and LiDAR sensors have been proven effective for SV and AGB retrieval, the potential of spaceborne systems remains understudied. This study evaluates the capability of NASA’s Earth Surface Mineral Dust Source Investigation (EMIT) hyperspectral data, combined with canopy height metrics derived from the Global Ecosystem Dynamics Investigation (GEDI) LiDAR data, to retrieve SV and AGB in two heterogeneous mountain forests in Italy. We compared EMIT with Sentinel-2 (S2) multispectral data as model inputs, with and without GEDI data integration, using five Machine Learning (ML) algorithms: Partial Least Squares Regression (PLSR), Boosted Regression Trees (BRT), Support Vector Machines (SVM), Artificial Neural Networks (ANN), and Gaussian Process Regression (GPR). We then applied the top-performing models to generate spatially explicit SV and AGB maps. Results demonstrated that EMIT-GEDI integration enhanced SV estimation accuracy (R 2 = 0.75 RMSE = 75.48 m 3 ha −1 , GPR model) compared to S2-GEDI (R 2 = 0.69 RMSE = 84.48 m 3 ha −1 , ANN model). AGB was retrieved with significantly lower accuracy than SV, and S2-GEDI models outperformed EMIT-GEDI ones, likely because of the higher S2 spatial resolution better capturing AGB variability associated to different tree species. GEDI LiDAR proved to be a necessary input for accurate SV and AGB retrieval, and GPR was the best-performing ML algorithm. The resulting spatial maps were artifact-free and successfully delineated ecological gradients and management patterns. This study underscores the promise of spaceborne hyperspectral-LiDAR data integration for SV and AGB mapping in mixed mountain forest ecosystems, However, it also emphasizes trade-offs between sensor spectral, spatial and temporal resolutions, thus the importance of upcoming hyperspectral missions, such as CHIME, combining hyperspectral capabilities with high spatial resolution and regular data acquisitions at global scale.