Integrating tree species composition with height-driven allometry for enhanced aboveground biomass mapping

Abstract

Monitoring the spatial pattern of aboveground biomass (AGB) and identifying its environmental controls are crucial for carbon budget assessments, climate mitigation actions, and sustainable forest management. While remotely sensed AGB estimates have been largely reliant on multi-variable empirical models, key ecological information, such as tree height allometry and continuous tree species composition, has rarely been incorporated into large-scale AGB predictions, resulting in oversimplified landscape averages. In this paper, we developed an integrated framework for AGB mapping that combines species composition with height-based allometric models by leveraging Sentinel-2 imagery, topography, forest height, and vegetation zones. Our methodology involved identifying tree species from multi-temporal imagery and building species-specific allometric models. Species-level AGB was calculated using forest height and then mosaicked to achieve mapping of the study area. Furthermore, the dominant factors affecting AGB spatial variation were investigated with extreme gradient boosting and Shapley additive explanations This approach identified ten tree species with an overall accuracy of 87.93% and yielded an AGB estimation accuracy of r = 0.75 and rRMSE = 32.16% in Northeast China. AGB estimates varied between 69.65 t/ha and 143.26 t/ha across tree species, with a regional average of 88.86 t/ha. Our AGB map performed superiorly to existing products (r: 0.5–0.56, rRMSE: 49.83%–8.83%), while it slightly underestimated the AGB relative to the provincial-level inventory statistics. SHAP analysis revealed that average precipitation in growing season shapes AGB variation (39.15%) across the entire region, while terrain and soil organic carbon contribute to explaining the regional AGB. Our study provides a new approach for enhanced AGB mapping in temperate forests and emphasizes the effects of environmental factors and human activities on AGB spatial patterns.