Abstract. Clouds are assumed to play an important role in the Arctic amplification process. This motivated a detailed investigation of cloud processes, including radiative and turbulent fluxes. Data from the aircraft campaign ACLOUD were analyzed with a focus on the mean and turbulent structure of the cloudy boundary layer over the Fram Strait marginal sea ice zone in late spring and early summer 2017. Vertical profiles of turbulence moments are presented from contrasting atmospheric boundary layers (ABLs) from 4 d. They differ by the magnitude of wind speed, boundary-layer height, stability, the strength of the cloud-top radiative cooling and the number of cloud layers. Turbulence statistics up to third-order moments are presented, which were obtained from horizontal-level flights and from slanted profiles. It is shown that both of these flight patterns complement each other and form a data set that resolves the vertical structure of the ABL turbulence well. The comparison of the 4 d shows that especially during weak wind, even in shallow Arctic ABLs with mixing ratios below 3 g kg-1, cloud-top cooling can serve as a main source of turbulent kinetic energy (TKE).Well-mixed ABLs are generated where TKE is increased and vertical velocity variance shows pronounced maxima in the cloud layer. Negative vertical velocity skewness points then to upside-down convection. Turbulent heat fluxes are directed upward in the cloud layer as a result of cold downdrafts. In two cases with single-layer stratocumulus, turbulent transport of heat flux and of temperature variance are both negative in the cloud layer, suggesting an important role of large eddies. In contrast, in a case with weak cloud-top cooling, these quantities are positive in the ABL due to the heating from the surface. Based on observations and results of a mixed-layer model it is shown that the maxima of turbulent fluxes are, however, smaller than the jump of the net terrestrial radiation flux across the upper part of a cloud due to the (i) shallowness of the mixed layer and (ii) the presence of a downward entrainment heat flux. The mixed-layer model also shows that the buoyancy production of TKE is substantially smaller in stratocumulus over the Arctic sea ice compared to subtropics due to a smaller surface moisture flux and smaller decrease in specific humidity (or even humidity inversions) right above the cloud top. In a case of strong wind, wind shear shapes the ABL turbulent structure, especially over rough sea ice, despite the presence of a strong cloud-top cooling. In the presence of mid-level clouds, cloud-top radiative cooling and thus also TKE in the lowermost cloud layer are strongly reduced, and the ABL turbulent structure becomes governed by stability, i.e., by the surface–air temperature difference and wind speed. A comparison of slightly unstable and weakly stable cases shows a strong reduction of TKE due to increased stability even though the absolute value of wind speed was similar. In summary, the presented study documents vertical profiles of the ABL turbulence with a high resolution in a wide range of conditions. It can serve as a basis for turbulence closure evaluation and process studies in Arctic clouds.