Spatial data consistency is very important in geographical research. In recent years, the widespread availability of user-class unmanned aerial vehicles (UAVs) has enabled the acquisition of aerial imagery data and the creation of high spatial resolution digital surface models (DSMs) of local coverage. However, these works may not always be accompanied by accurate geodetic reference, which worsens the positional accuracy of the materials. For example, the models obtained from multi-temporal surveys may be displaced relative to each other by a noticeable amount (up to tens of meters vertically and/or horizontally). Many co-registration techniques are used to align such DSMs, but the accuracy of these transformations has not been sufficiently assessed to date. This paper deals with the accuracy estimation of the co-registration applied to DSMs derived from unmanned aerial imagery data without precise georeferencing. Sets of aerial photographs from 2021–2024 for four sites with different terrain and vegetation characteristics were used. The DSMs acquired for these sites were co-registered using three different techniques: Nuth and Kääb 2011, ICP, ILEM. The result was evaluated by comparing the differences of the co-registered DSMs and those obtained by mutual photogrammetric processing of the original aerial imagery data. It is shown that the ICP co-registration method gives satisfactory outcomes in bare ground surfaces. In forested areas, none of the said techniques allow to obtain acceptable results