Guth et al. (2025) Benchmarking Elevation Plus Land Surface Parameters Finds FathomDEM and Copernicus DEM Win as Best Global DEMs

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-12-03
• Authors: Peter L. Guth, Sebastiano Trevisani, Carlos Henrique Grohmann, John B. Lindsay, H.I. Reuter
• DOI: 10.3390/rs17233919

Research Groups

• Independent Researcher, Annapolis, MD, USA (Peter L. Guth)
• Dipartimento di Culture del Progetto, University Iuav of Venice, Italy (Sebastiano Trevisani)
• Institute of Astronomy, Geophysics and Atmospheric Sciences, Universidade de São Paulo, Brazil (Carlos H. Grohmann)
• Department of Geography, Environment & Geomatics, The University of Guelph, Guelph, ON, Canada (John B. Lindsay)
• Independent Researcher, Trier, Germany (Hannes I. Reuter)

Short Summary

This study benchmarks six global 1-arc-second Digital Elevation Models (DEMs) against high-resolution lidar-derived reference Digital Terrain Models (DTMs) across 1510 test tiles, evaluating both elevation and derived Land Surface Parameters (LSPs). It finds FathomDEM to be the best overall, with Copernicus DEM as the top unrestricted option, and highlights significant variability in LSP accuracy depending on terrain characteristics and derivative order.

Objective

• To evaluate the performance of six global digital elevation models (DEMs) at 1-arc-second resolution (CopDEM, AW3D30, FABDEM, FathomDEM, EDTM, and GEDTM) by comparing their elevation and derived Land Surface Parameters (LSPs) against high-resolution lidar-derived reference Digital Terrain Models (DTMs).

Study Configuration

• Spatial Scale: Global (1-arc-second resolution, approximately 30 meters at the equator), tested across 1510 approximately 10 km × 10 km tiles from the United States and Western Europe.
• Temporal Scale: The study was published in 2025, evaluating global DEMs which are products of data collected over various periods (e.g., GEDTM data range 2006-2015, ESA World Cover 2021). The reference DTMs are from national mapping agencies, with varying acquisition dates.

Methodology and Data

• Models used:
  • Test DEMs: Copernicus DEM v2023-1 (DSM), ALOS AW3D30 v3.2 (DSM), FABDEM v1.2 (DTM), FathomDEM v1 (DTM), GEDTM v1.2 (DTM).
  • Software for analysis: MICRODEM v2025.11.27, Whitebox v2.4.0, SAGA v9.3.0_x64, lsp-calculator v1.1.1, GDAL.
• Data sources:
  • Reference DTMs: Lidar-derived DTMs with 1-2 meter resolution from national mapping agencies (e.g., USGS, Denmark, Estonia, Finland, Germany, Mexico, Netherlands, Spain, Switzerland, UK).
  • Landcover: ESA World Cover 2021 (10 meter resolution).
  • Global DEMs: Copernicus DEM, ALOS World 3D (AW3D30), FABDEM, FathomDEM, GEDTM.

Main Results

• FathomDEM consistently emerged as the best performing global 1-arc-second DTM, but it has a restrictive license.
• Copernicus DEM (CopDEM) was identified as the best overall among the DEMs with unrestricted licenses.
• ALOS AW3D30 performed best specifically in rugged and steep mountainous areas.
• GEDTM v1.2 performed poorly across most evaluations, particularly for derived Land Surface Parameters (LSPs).
• The accuracy of global DEMs in representing LSPs (e.g., slope, curvatures, roughness) is highly dependent on terrain characteristics, especially average slope and barrenness.
• LSPs computed using first-order partial derivatives (e.g., slope, hillshade) showed the highest correspondence with reference DTMs.
• LSPs requiring second-order partial derivatives (e.g., many curvature measures) exhibited variable and generally lower correspondence, indicating a reduced signal-to-noise ratio.
• LSPs requiring third-order partial derivatives showed extremely low correspondence, suggesting they are predominantly noise.
• Filtering DEMs before LSP computation did not consistently improve results and could worsen them depending on the terrain and specific LSP.

Contributions

• Provides a comprehensive and updated benchmarking of recent 1-arc-second global DEMs, including newly released DTMs like FathomDEM and GEDTM.
• Emphasizes the critical importance of evaluating derived Land Surface Parameters (LSPs) in addition to raw elevation data for a holistic assessment of DEM quality.
• Introduces an improved, more efficient, and geographically diverse UTM-based tiling system for creating reference DTMs, enhancing the robustness of comparisons.
• Offers detailed insights into how terrain characteristics (slope, barrenness, forest cover) and the order of partial derivatives used in LSP computation affect DEM performance.
• Provides practical recommendations for creators of DTMs regarding optimal starting DSMs (CopDEM, ALOS AW3D30) and the necessity of multi-LSP validation.
• Critically assesses the performance and documentation of GEDTM, advising caution for users due to quality control issues and poor LSP generation.

Funding

• The São Paulo Research Foundation (FAPESP—grant #2023/11197-1)
• CNPq (grant #300033/2025-7)
• Natural Sciences and Engineering Research Council of Canada (NSERC) grant number 401107

Citation

@article{Guth2025Benchmarking,
  author = {Guth, Peter L. and Trevisani, Sebastiano and Grohmann, Carlos Henrique and Lindsay, John B. and Reuter, H.I.},
  title = {Benchmarking Elevation Plus Land Surface Parameters Finds FathomDEM and Copernicus DEM Win as Best Global DEMs},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17233919},
  url = {https://doi.org/10.3390/rs17233919}
}