Stratified AGB inversion driven by DGTHI: quantifying topographic controls on biomass prediction across tree species

Abstract Accurate forest aboveground biomass (AGB) estimation is crucial for global carbon cycle research. While existing studies have utilized topographic factors in remote sensing, they often fail to systematically quantify multi-dimensional heterogeneity or address species-specific responses. This study pioneers the application of the Digital Elevation Model (DEM) Grid Topographic Heterogeneity Index (DGTHI) to enhance AGB inversion models. The DGTHI is a composite metric integrating elevation variability, relief, surface roughness, and mean slope. Using airborne Light Detection and Ranging (LiDAR) data and 8,804 field-measured trees from Mengyin County, Shandong Province, China, we developed a DGTHI-stratified modeling framework. This framework dissects how topographic heterogeneity governs species-level AGB estimation accuracy at the county scale. Results demonstrate that: (1) DGTHI outperformed conventional single-factor topographic corrections, with heterogeneity effects on feature selection following a species hierarchy: acacia > pine > cypress > poplar; (2) The DGTHI-driven stratification led to a significant improvement in model accuracy, with R 2 increasing by 0.08 to 0.17 compared to the unstratified models; (3) Spatial AGB patterns (27–217 t/ha in May 2023) revealed southwest–northeast highs and northwest–southeast lows, directly modulated by DGTHI-mapped heterogeneity. As the first integration of DGTHI into species-specific AGB inversion, this work provides a transferable paradigm for precision carbon mapping in topographically complex forests.