I imagine that for DTM obtained from Stereoscopic procedures or interpolation from a set of topographic points a comparaison with the original data can help, but what for DTM obtained from Remote Sensing ?
Usually terrain models derived from remote sensors will have their initial accuracy evaluated using blind control methods. For example with LIDAR, the remotely sensed measurements are often compared to survey-grade control points (the blind control) taken on the ground in various land covers (woods, open areas, etc) to assess the penetration and accuracy of the LIDAR dataset. This comparative procedure is expressed as RMSE (root mean square error) values for the horizontal and vertical planes, and is often documented in a QA/QC report / metadata.
After the fact, or if the RMSE report was unavailable, you could do a blind control comparison of your own by comparing your own points collected in the field with the remotely sensed data.
Keep in mind that the accuracy of the original remotely sensed data is not necessarily going to be the same as the DTM / DEM derived from it. Interpolation, point thinning, and the chosen grid size will all contribute to potential dilution of the original data's accuracy.
Some literature on this topic:
- COLOSIMO, G., CRESPI, M., DE VENDICTIS, L., JACOBSEN, K. (2009). Accuracy evaluation of SRTM and ASTER DSMs. In Proceedings of the 29th EARSeL Symposium, Chania, Greece.
- GOROKHOVICH, Y., & VOUSTIANIOUK, A. (2006). Accuracy assessment of the processed SRTM-based elevation data by CGIAR using field data from USA and Thailand and its relation to the terrain characteristics. Remote Sensing of Environment, 104(4), 409-415.