Damage assessment is a crucial aspect in many different fields, especially in the humanitarian context. Often the estimation is in addition time critical as conflict events urge a quick response from NGOs. The presented approach is part of a currently ongoing research project analyzing the potential of cross-stereo satellite imagery, i.e. images from different dates of the same or even different satellite sensors, to allow the generation of pre and post-event 3D information also in situations with difficult data constellations. This 3D information may be used for change detection, damage assessment, and even population estimation based on building height derivation. The study area is located in the city of Mosul, Iraq. The city has been attacked several times and has been seriously damaged, in particular in June and July 2017. A regular Pléiades-tri-stereo optical satellite image of May 6, 2015 was available as pre-event dataset, and a mixed/cross stereo pair of two (mono) Pléiades images of August 21, 2017 and September 4, 2017 as post-event datasets. This reflects quite well a situation where dedicated stereo imagery acquired explicitly for digital surface model (DSM) generation is not available directly after an event. The two DSMs were created using DLR's in-house photogrammetry suite CATENA. Comparison of the pre- and post-event DMSs, to assess the severity of damage, was conducted using object-based image analysis. The outcome is a categorization of identified areas with regard to certainty of correctly classified areas: Areas that contain a high degree of certainty being damaged, areas that are most likely damaged and areas that need further investigation, as there is only a moderate certainty in representing damaged areas. The latter ones are areas that for example have quality issues in the DSM derivation due to shading and therefore hinder a detailed identification of damaged areas by automated means. The results are validated against detailed visual damage assessments available for the old town of Mosul, performed by UNOSAT. The proposed approach is a contribution to assess the degree of damage in conflict areas in addition to typical assessments on optical image data only. The quality of the derived DMS and the damage assessment depends on the available imagery (time between acquisitions, viewing geometry, sensor mix), but is in any case an additional valuable information source. The usage of mixed-date stereo models increases the likeliness to find suited data shortly after such an event and therefore it strongly supports time-critical damage assessments based on satellite data.