Author: Petra Pracná
Mapping the Earth’s terrain beneath forests and across rugged mountains has always been a challenge. Traditional lidar surveys can achieve high accuracy but are limited in spatial coverage and costly to repeat. With the launch of NASA’s ICESat-2 (Ice, Cloud and land Elevation Satellite-2) and GEDI (Global Ecosystem Dynamics Investigation) missions, we now have a unique opportunity to build terrain models at global scales using spaceborne lidar.
In our study, we explored how these two missions can be combined to generate 90 m resolution digital terrain models (DTMs). We evaluated their accuracy, sampling intensity, and comparability with existing global datasets such as the Copernicus GLO-90 DEM. The research was carried out across diverse and challenging landscapes in Switzerland, New Zealand, and California, using airborne lidar data as reference.
What we found
- ICESat-2 consistently outperformed GEDI in terrain elevation accuracy under various conditions (terrain slope, land cover, beam strength, and time of acquisition).
- Combining ICESat-2 and GEDI significantly improved sampling intensity — more than 60% of 90 m grid cells contained at least one observation, enabling interpolation of 90 m DTMs.
- The spaceborne lidar-derived DTMs achieved RMSEs between 9.9–14.7 m, comparable to the global Copernicus DEM (9.9–15.6 m).
In areas with sufficient observation density (≥4–6 points per cell), spaceborne lidar DTMs even outperformed Copernicus DEMs, achieving errors as low as 3.7 m in forests.
Why it matters
Global digital elevation models (GDEMs) such as Copernicus GLO-90 or SRTM provide near-global coverage but often include vegetation and building offsets, particularly in forested or urban areas. This study demonstrates that combining ICESat-2 and GEDI data can produce terrain models of comparable or even superior accuracy, especially in complex mountainous and forested regions. When sufficient high-quality observations were available (4–6 per 90 m cell), the spaceborne lidar DTMs outperformed the Copernicus GLO-90 DEM, achieving root mean square errors (RMSEs) as low as 3.7 m in forests, compared to 11.2 m for Copernicus. This confirms that, under the right conditions, satellite lidar can not only complement but also improve existing global elevation datasets.
Such an approach opens new opportunities for global-scale, high-resolution terrain mapping based solely on spaceborne lidar data. This next generation of DTMs could become a key tool for environmental monitoring, hydrology, and Earth surface process modeling, advancing our ability to understand and manage the planet’s changing landscapes.
For more details please see: DOI: 10.1016/j.srs.2025.100293