Evaluation and improvement of the vertical accuracy of the global open DEM under forest environment

Existing multi-source Digital Elevation Models (DEMs) still have uncertainties in estimating understory terrain. The study aims to use Goddard’s LiDAR, Hyperspectral & Thermal Imager (G-LiHT) as validation data to investigate the accuracy of L2A-level Products of the Global Ecosystem Dynamics Investigation (GEDI02_A) and six open-source DEMs in estimating understory terrain across six research areas. This study will quantify the influence of canopy height, canopy coverage, and leaf area index on estimation accuracy, and improve the accuracy of DEMs. The research findings indicate that the accuracy of GEDI02_A is the highest, with Root Mean Square Error (RMSE)=6.20m. Next, the Federal Railway Authority of Germany’s DEM (FABDEM) with RMSE = 8.46 m. Canopy height exhibits a higher correlation with the estimated accuracy of forest understory terrain. Finally, optimizing FABDEM based on the mathematical interpolation method using GEDI02_A reduces the RMSE from 8.46 m to 6.83 m. Slope ranging from 0-3% demonstrates the most significant improvement.