Digital Elevation Models (DEMs) are indispensable tools in geospatial analysis, especially for applications in forested areas. The main aim of this research is to assess the applicability, accuracy, and usability of DEMs derived from Sentinel-1 Synthetic Aperture Radar (SAR) for forest studies, in comparison to the commercial high-resolution DEMs. Using the interferometric SAR (InSAR) technique, two DEMs were generated in 2018 using descending-orbit Sentinel-1 SAR (C-band) data for each SAR polarisation band (VH and VV). Elevation errors (LiDAR − DEM) were analysed using multivariate ordinary least squares regression with heteroscedasticity-robust (HC3) standard errors (n = 5228). Both Sentinel-1 VV and VH models explained a modest proportion of error variability (R² = 0.056, p < 0.001). For VV, elevation error increased significantly with slope (β = 2.72 m deg⁻¹, p < 0.001) and aspect (β = 18.78 m, p < 0.001), while crown density reduced error magnitude (β = −0.40 m %⁻¹, p < 0.001). The VH model showed similar explanatory power (R² = 0.056), but larger absolute deviations, with slope (β = 2.93 m deg⁻¹, p = 0.001) and aspect (β = 18.10 m, p < 0.001) again acting as dominant predictors. Tree species categories were not statistically significant in contributing to elevation errors. WorldDEM elevation errors showed stronger systematic relationships with terrain predictors, resulting in higher explained variance (R² = 0.124, p < 0.001) and greater statistical stability, with weaker but significant relationships to slope (β = 0.16 m deg⁻¹, p = 0.017) and crown density (β = −0.11 m %⁻¹, p < 0.001). Overall, Sentinel-1-derived DEMs showed stronger sensitivity to terrain geometry and canopy structure and substantially higher variability than WorldDEM. Despite the observed variations in Sentinel-1-derived DEMs within the forest application, Sentinel-1 SAR data remains a valuable data resource due to its free and globally available data, updated every six days (6–12 days depending on constellation availability), with an extensive archive spanning nearly a decade. This accessibility makes it a useful option for large-scale studies and regions where commercial SAR datasets are not feasible. The results of this study can enhance our understanding of errors associated with each Sentinel-1 polarisation when generating DEMs in forested areas. Further research should focus on post-processing of Sentinel-1-derived DEMs, such as combining DEMs derived from ascending and descending flight directions. Overall, this study provides an in-depth assessment of Sentinel-1-derived DEMs and highlights trade-offs between cost, accessibility, and accuracy when selecting DEMs for forest applications.
Avoiani et al. (Sat,) studied this question.