【Objective】 The new generation of the space-based altimetry global ecosystem dynamics investigation (GEDI) system is of great significance to forest observation and management. In order to explore the performance of GEDI version 2 data (V2 data) inversion of understory topography, this study uses airborne radar data to verify the accuracy of understory topography inversion, and explores the factors affecting the accuracy.【Method】 Taking the Cibola forest in the United States and the Maoer Mountain forest in China as the research objects, the performances of GEDI V2 data in coniferous forests and mixed coniferous and broad-leaved forests were verified using G-liht and Maoer Mountain high-precision airborne radar data. The effects of different beam intensities, spot times, slopes and vegetation coverage on the accuracy of terrain inversion were analyzed.【Result】 The root mean square error (RMSE) of topographic inversion accuracy in the Cibola taiga area of the United States was 2.33 m, and the average absolute error (MAE) was 1.48 m. The RMSE value of the topographic inversion accuracy in the coniferous and broad-leaved mixed forest area of Maoer Mountain was 4.49 m, and the MAE value was 3.33 m. With the increase in slope and vegetation coverage, the topographic inversion accuracy of the two forest types decreased.【Conclusion】 The GEDI V2 data inversion accuracy of understory topography in coniferous forests was higher than that of mixed coniferous and broad-leaved forests. Strong beams were better than coverage beams, and the accuracy was higher during the daytime in humid areas, and better at night in arid areas. The accuracy of steep areas was reduced, the terrain inversion accuracy was higher in areas with medium and low vegetation coverage, and the performances of terrain determination in areas with high vegetation coverage were decreased.