Photogrammetric digital elevation models (DEMs) are often used to derive and monitor surfaces in inaccessible areas. They have been used to monitor the spatial and temporal change of glacier surfaces in order to assess glacier response to climate change. However, deriving photogrammetric DEMs of steep mountainous topography where the surface is often obscured by regions of deep shadow and snow is particularly difficult. Assessing the quality of the derived surface can also be problematic, as high-accuracy ground-control points may be limited and poorly distributed throughout the modelled area. We present a method of assessing the quality of a derived surface through a detailed sensitivity analysis of the DEM collection parameters through a multiple input failure warning model (MIFWM). The variance of a DEM cell elevation is taken as an indicator of surface reliability allowing potentially unreliable areas to be excluded from further analysis. This analysis allows the user to place greater confidence in the remaining DEM. An example of this method is presented for a small mountain glacier in Svalbard, and the MIFWM is shown to label as unreliable more DEM cells over the entire DEM area, but fewer over the glacier surface, than other methods of data quality assessment. The MIFWM is shown to be an effective and easily used method for assessing DEM surface quality.