The three-dimensional under-ice topography has been mapped regularly during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition in 2019/20 (Nicolaus et al., 2022). We used a high-speed, high-resolution multibeam echosounder system (MBES, Imagenex DT101, Port Coquitlam, BC, Canada) attached to a remotely operated vehicle (ROV, M500, Ocean Modules, Åtvidaberg, Sweden, Katlein et al., 2017). The system was operated directly from the sea ice during most dive days on all legs of the MOSAiC expedition. The area covered had a radius of approximately 280 m. Standard multibeam surveys were performed at a dive depth of 20 m while high-resolution surveys at 10 m depth.The MBES consists of a multibeam sonar, an integrated Sound Velocity Probe (SVP), and a Motion Reference Unit (MRU). The operating frequency was 240 kHz. The sonar was used with 480 beams with a swath width of 120° (across track) x 3° (along track) and an effective beam width of 0.75°. The angular resolution resulting from the chosen number of beams and the sector size was 0.25°. The ping rate was automatically set and was based on the range setting, numbers of beams selected and specification of the computer system used for operation. The profile setup options were enabled profile point detection, maximum return, no along track averaging, no bottom following, and no pipe detect. A constant sound velocity of 1437 m/s was applied. Returns from three points were averaged to derive the distance from the MBES to the ice underside. Subtracting the distance to the ice from the vehicle depth yielded the sea-ice draft. More technical details are available here: https://imagenex.com/products/dt101xi.The software to record operate the MBES and record data is available from the Imagenex website (https://imagenex.com/interior-page/software-dowload). Data were processed using the CARIS HIPS and SIPS software (Teledyne CARIS, Fredericton, NB, Canada) applying swath and subset editors for data cleaning. The appreciation '_HR' in the file names indicates higher lateral resolution processing with resolutions of 0.1 m instead of 0.5 m.

Version 1.0, preliminary release August 2021: This dataset is an early release with QC data, but the data set is not corrected or polished to every aspect. It can be used, but issues and pitfalls remain. Known issues are the following:• Roll calibration of the multibeam mounting is imperfect leading to stripe patterns and surface tilts in some datasets.• Draft values have not been checked for absolute height. Cross-check with drillings or open water patches before using, significant offsets can be possible.• Positioning accuracy of the ROV varied, hence some features can be displaced by meters, while lateral accuracy within individual swaths is good.• Some datasets exhibit bulges over the zenit beams .• Due to sound velocity differences, the acoustic position can be scaled and distorted relative to precise surface measurements (e.g. laser scanner).• Not all edge beams have been cleaned up perfectly.• No sound velocity correction.• Reflections might originate from false bottoms in some cases.• Some data on 01.07, 07.07., and 21.07. are affected by bad reflections likely due to a freshwater lens.• The reference system changes during some of the legs due to adaptations in ROV deployment site or transponder configuration.• Some surfaces have considerably lower quality compared to others.• So far, Leg 1 data could not be processed due to problems with ROV positioning and various other aspects (might be unrecoverable).