What question did this study set out to answer?

To evaluate the vertical accuracy of global Digital Elevation Models using ICESat-2 and their impacts on flood modeling.

May 13, 2026Open Access

Evaluating the Vertical Accuracy of Global DEMs Using ICESat-2 and Its Cascading Impact on HAND-Based Flood Modeling in a Low-Gradient Coastal Plain

Key Points

To evaluate the vertical accuracy of global Digital Elevation Models using ICESat-2 and their impacts on flood modeling.
Used ICESat-2 ATL08 laser altimetry as reference for assessing eight open-source DEMs in Lianyungang, China.
Analyzed the effects of slope, aspect, and land cover on DEM errors.
Employed the HAND model to evaluate DEM accuracy on flood inundation simulation.
FABDEM exhibited highest accuracy with RMSE of 1.24 m and NMAD of 0.49 m.
Copernicus DEM GLO-30 followed with RMSE of 1.56 m.
ASTER GDEM had the lowest accuracy with RMSE of 5.36 m.
DEM errors increased with slope and were greater over water bodies compared to cropland.
Flood inundation area varied significantly with the DEMs used, especially at low HAND thresholds.

Abstract

Driven by climate change and population growth, coastal flood risk is rising, making high-precision Digital Elevation Models (DEMs) essential for inundation simulation and risk assessment. Although global open-source DEMs are increasingly available, their regional applicability and uncertainty still require quantitative evaluation. Taking Lianyungang, a coastal city in eastern China, as the study area, this study used ICESat-2 ATL08 laser altimetry as the reference to assess the vertical accuracy of eight mainstream open-source DEMs: the ASTER GDEM, FABDEM, AW3D30 DEM, SRTM DEM, MERIT DEM, NASA DEM, Copernicus DEM, and TanDEM-X DEM. The effects of slope, aspect, and land cover on DEM errors were analyzed, and the Height Above Nearest Drainage (HAND) model was used to evaluate how DEM vertical accuracy and spatial resolution affect flood inundation simulation. The results show that the FABDEM has the highest accuracy (RMSE = 1.24 m; NMAD = 0.49 m), followed by the Copernicus DEM GLO-30 (RMSE = 1.56 m; NMAD = 0.65 m), whereas the ASTER GDEM performs worst (RMSE = 5.36 m; NMAD = 3.69 m). The SRTM DEM systematically underestimates ICESat-2 elevations, with mean and median errors of −1.85 m and −1.80 m, mainly due to acquisition time differences and land-use changes in Lianyungang. DEM errors generally increase with slope, are higher on west-facing slopes, and are larger over water bodies than over cropland and impervious surfaces. HAND simulations show that DEM-derived inundation differences are greatest under low-threshold conditions. At the 1 m HAND threshold, the MERIT DEM produces the largest inundation area (4370.28 km2), while the ASTER GDEM produces the smallest area (3330.53 km2); these differences decrease as the threshold increases. Overall, the FABDEM provides the most accurate flood inundation representation in Lianyungang, while the Copernicus DEM GLO-30 is a reliable alternative.

Evaluating the Vertical Accuracy of Global DEMs Using ICESat-2 and Its Cascading Impact on HAND-Based Flood Modeling in a Low-Gradient Coastal Plain

Key Points

Abstract

Cite This Study

Also Consider

Also Consider