Digital Elevation Models (DEMs) are a valuable source of knowledge about relief and terrain characteristics. This paper presents an analysis of the accuracy of open-access DEMs in Spain: SRTM, GDEM and the recently released national DEM (known as PNOA DEM). The models were tested by computing the Root Mean Square Error (RMSE) from elevation points gathered from the digital version of the 1:10,000 National Topographical Maps, and fitting the models using the classical linear regression analysis. In addition, the magnitude and pattern of errors in primary and secondary derived terrain attributes were explored by using a DEM generated ad hoc from elevation points and contour lines of the MTN (scale 1:10,000). The results showed the RMSE ranging from 6.1 m for the SRTM to 33.0 m for the GDEM. Surprisingly, a RMSE of 31.2 m was estimated for the recently available PNOA DEM in Spain. Results also suggest that GDEM slightly underestimates altitudes in the study area, while this trend was not found in SRTM or PNOA DEM. The highest values of RMSE showed by the PNOA DEM were particularly located in the roughest areas of steep slopes and high aspect variability. Finally, the exploration of the errors accounted by generating primary and secondary terrain attributes suggests that the utility of these models is strictly limited to description, visualization and representation of relief.