The deployment of tidal turbines requires a precise hydrodynamic characterisation of the production site. Acoustic Doppler Current Profilers (ADCP), usually employed for measuring the time-mean characteristics of environmental flows, could also be used for assessing the main features of turbulence. ADCP measurements are sensitive to many sources of uncertainties associated mainly with the spreading of the beams or the assumptions made on flow homogeneity. The ability of ADCPs to accurately measure the hydrodynamic parameters of a given flow can be tested on a synthetic dataset. However, it is difficult to generate a dataset representative of a real environmental flow. In this work, large-eddy simulation of a high Reynolds flow over a rough seabed is performed and used to assess the accuracy of two, coupled, 4-beam ADCP systems forming an 8-beam arrangement. The study confirms the relevance and efficiency of the tested 8-beam configuration for the characterisation of turbulence. The results near the seabed are of a lower quality, with up to 50 % error on the Reynolds stresses for elevations under twice the roughness height, which questions the interpretation of ADCP measurements in the lower part of the water column. Also, the spatial averaging over ADCP cells leads to an underestimation of the turbulence intensity of 10 % to 20 %.