Canopy height is important in determining growing stock volume, aboveground biomass, and carbon stocks and evaluating the effectiveness of certain forest management programs. The semi-arid ecosystems of Africa are understudied, leaving questions relating to their role in the overall global carbon cycle unanswered. This study assessed the performance of Gradient Boosting Machine (GBM) and Random Forest (RF) methods to predict woody canopy height in the Hamoye State Forest in Namibia, using multisource satellite remote sensing data from Global Ecosystem Dynamics Investigation (GEDI), L2A, Sentinel-1 and Sentinel-2. Furthermore, the study compared different GEDI relative height metrics (i.e. RH100, RH98, RH95) to determine the most useful one for the study area. Results show that, while GBM outperformed RF across GEDI metrics, GBM’s overall coefficient of determination (R2) values were higher than RFs in relation to RH100 (0.416 versus 0.412), RH98 (0.412 versus 0.374), and RH95 (0.345 versus 0.304). In addition, GBM reported lower errors than RF based on the relative root mean squared error (RMSE) in the range of 1.984 to 2.327 m and bias in the range of 0.168 to 0.363 m. Among the GEDI-derived relative height (RH) metrics, RH98 was identified as the most dependable, offering an optimal balance of accuracy across the study area. These findings suggest that accurate canopy height estimation in semi-arid ecosystems can be achieved by integrating GEDI RH98 metrics and national forest inventory (NFI) data. This synergy between ground measurements and spaceborne LiDAR data improves the reliability of TCH assessments, even in regions with limited GEDI coverage.