The threshold of coarse sediment transport has been examined in natural streambeds in an upland Pennine (U.K.) area. Threshold values of the total boundary shear stress (T0) (for a given grain size), in a narrow natural stream (W/D < 11) are considerably higher than values of T0 in a broad stream (W/D > 11). Efficiency in the entrainment process is related not only to the overall channel geometry, but also varies as a function of discharge in channels characterized by compound roughness.
Empirical curves relating T0 and a mean grain size (d5) are presented, but are limited in application to streams of similar physical and hydraulic characteristics as the ones examined in this investigation. Considerable divergence is noted between these empirical functions and a summary empirical function for general application obtained from a published source. The reasons for this divergence are discussed.
The influence of grain shape was found not to be important in the initiation of motion criterion. This conclusion may reflect the limited range of natural grain shapes in the study streams, but might reasonably apply to other field investigations of similar streams.
Modifications of the Shields' and Yalin diagrams are suggested for practical applications in shallow streams with poorly-graded bed material. The Shields' parameter may be regarded as an inverse function of the relative protrusion of individual grains in the shallow flow depth (d5/D). The increased importance of augmented drag forces, in the entrainment process in shallow flows, is suggested as the physical explanation for the reduced values of the Shields' parameter. However, the relationships presented should not be applied to laboratory experiments concerned with well-graded sediments (therefore beds with little deviation in level), in which the Shields' parameter may be regarded as constant at high Reynolds' grain numbers.
Consistent estimated field values of ϵ, a threshold sediment transport parameter, might be used to compare field data to threshold values derived from statistical arguments and laboratory experiments reported in the literature.