An extended log-fit method is developed to estimate mean seabed roughness length z 0 from a number of current profiles measured near the seabed in the absence of surface waves. With this new method, the mean seabed roughness length z 0 is estimated by fitting the logarithmic distribution, k(z)=Aln(z/z0), to a time-independent dimensionless mean velocity profile k(z) that is derived from the number of current profiles, where A is a time-independent constant. The field data were collected at a mean water depth of 16m in a tidal channel. An Acoustic Doppler Profiler (ADP) was used to continuously measure tidal current profiles at 16 levels above the seabed for 10min, every 10min over a period of 13days. On the basis of collected field data, it is found that the bed roughness lengths estimated from individual velocity profiles are extremely fluctuating and profile-dependent. In contrast, the mean roughness lengths estimated with this extended log-fit method are shown much less fluctuating and proven more accurate. This study also investigates the sensitivity of the current shear velocity u¯* to different bed roughness lengths.