In performing physical model studies researchers generally want to establish fullydeveloped uniform flow. If the flow is not fully-developed then results from the study cannot be generalized. This study aims to develop a relationship for predicting the channel length required for a rough turbulent flow to become fully developed uniform flow (L unif ) in a laboratory flume. The flume in which we conducted the tests is about 6.1 m in length, 0.46 m wide, and has a variable slope. The channel has smooth walls and the bed was lined with gravel (median particle size, d 50 = 4.6 mm or 8.8 mm). Flow depths and mean velocities ranged, respectively, from 183 mm to 235 mm and 16.3 cm/s to 46.3 cm/s. Six tests were conducted. For each test, longitudinal point velocity measurements (u) were made along the center of the channel, at five elevations (z), and at eleven longitudinal stations (x). We used an acoustic Doppler velocimeter (ADV) to measure u. We found L unif by identifying the value of x at which u was constant for all values for z. Our results indicate that the product of the Reynold's number and Froude number yield the best relationship for predicting L unif (r 2 = 0.88). In addition, with regards to rough turbulent open-channel flow in laboratory flumes, this study questions the concept of a boundary layer that monotonically increases in thickness.