Acoustic Doppler current profilers (ADCPs) are the standard tool for measuring tidal currents at tidal stream energy sites; they are used to estimate several parameters, including turbulent kinetic energy (TKE). However, estimates of TKE from ADCPs are often swamped by wave action. We surmise that this bias can be detected as a data mode: to test this, we present an empirical orthogonal function (EOF) analysis of two months of TKE estimates from ADCP measurements at a tidal energy site with significant wave activity. The results of the analysis were compared with linear wave theory, using data from a wave buoy. The first data mode identified from EOF analysis agrees well with the wave bias predicted by linear theory, and the resulting decomposition of the data set into wave and turbulent components appears realistic. This decomposition is possible from ADCP data alone, and therefore offers a novel and widely applicable analysis technique for simultaneous assessment of turbulence and waves at highly-energetic tidal sites. The method can also be applied retrospectively to historical data sets. We also show that the decomposition can be improved by including higher EOF modes, but this requires an independent measurement of waves to determine the optimum number of modes.