Ground-penetrating radar (GPR) reflection imaging is a popular geophysical tool to explore subsurface structures in a non-invasive manner. In terms of GPR, reflective interfaces are defined by contrasts in dielectric permittivity, which result from, for example, variations in soil moisture or ice content. GPR is very suitable for electrically high resistive environments, such as frozen ground (typically > 10,000 m). Here, GPR can be used to explore structural targets at depths up to tens of meters. Furthermore, GPR can be employed to explore more shallow environments where detailed information on the decimeter scale is required. In consequence, 2D GPR reflection profiling is used on different spatial scales in permafrost applications such as active layer characterization and imaging of pingos. However, a 3D strategy might be essential for obtaining a reliable image of subsurface structures, if the geometry of such targets is complex (e.g., structures vary in three dimensions). Additionally, 3D data allow to identify out-of-plane reflection events which might interfere with reflections from target structures. This advantage is especially interesting for the application of GPR in cold environments, where out-of-plane reflections are favored due to a broadened radiation characteristic of GPR antennas on frozen ground compared to unfrozen ground. Here, we present a carefully designed 3D GPR acquisition and processing strategy (Schennen et al., 2016) and employ it to an exemplary data set. Our field site covers an area of approximately 20 m×70 m and is located on top of a Yedoma hill on Bol’shoy Lyakhovsky Island, Northern Siberia. Nearby borehole information provides cryostratigraphic details (up to a depth of approximately 30 m) interpreted in terms of three major stratigraphic units. These comprise two ice complex strata, which enclose a unit of floodplain deposits. Additional ground-truth is available from a 18 m high outcrop of the upper ice complex next to our survey area. Here, we observe large (up to 10 m wide) ice-wedges segmenting the ice- and organic-rich, loess-like sediments. In our unmigrated 3D GPR data, time slices show distinct circular diffraction features. As we move on to succeeding slices, we observe that these features originate from locations below thermokarst mounds, expand with a velocity of 0.17 m/ns, and interfere with each other at later traveltimes (Fig. 1a-d). They result in a complex 3D distribution of diffracted energy evident in the entire data cube. Thus, a 3D migration approach (e.g., Allroggen et al., 2014) is essential to correctly image subsurface structures. Thereby we consider also topographic variations and possible subsurface velocity variations. In our migration result, we observe two distinct horizontal features at depths larger than 20 m. Taking borehole data into account, we interpret these features as the base of the upper ice complex unit and the underlying floodplain deposits, respectively. Furthermore, we are able to trace both interfaces in our data cube and compile our interpretation into a 3D cryostratigraphic model (Fig. 1e), which can be used to scale up borehole and outcrop information. In a concluding 2D vs. 3D comparison, we extract exemplary 2D profiles from our unprocessed 3D data to simulate a 2D GPR acquisition and processing strategy on the same field site. Thus, we are able to investigate the impact of data reduction on each processing step. Comparing results of our 2D and 3D processing strategies demonstrate, that a 3D GPR surveying and processing strategy is critical in complex permafrost settings. Allroggen N, Tronicke J, Delock M, Böniger U. 2014. Topographic migration of 2D and 3D groundpenetrating radar data considering variable velocities. Near Surface Geophysics 13: 253-259.DOI: 10.3997/1873-0604.2014037 Schennen S, Tronicke J, Wetterich S, Allroggen N, Schwamborn G, Schirrmeister L. 2016. 3D GPR imaging of ice complex deposits in northern East Siberia. Geophysics, 81: 1-9.DOI: 10.1190/GEO2015-0129.1