allowed to proceed with U left equal to U^, though in a few cases U 

 was increased to establish Uj,. Early ripple development with the 0.18- 

 and 0.21 -millimeter sands differed somewhat from that with the 0.55- 

 millimeter sand. 



With the 0.55-millimeter sand, the scow proved effective and the 

 ends of the bed were the last areas to become rippled. At U»Uc, 

 ripples formed either spontaneously or first along the walls. Slight 

 increases in U beyond Uj,, made in experiments 2, 29, and 57, were 

 found to cause the rate of spreading to increase rapidly. The 

 difference between Uj. and U^ appeared to be very slight, probably less 

 than 5 percent and within the uncertainty of U^ itself. No ripples 

 resembling the rolling-grain type described by Bagnold C1946) were 

 observed to form in advance of the spreading vortex ripples. 



With the 0.18- and 0.21-millimeter sands, the scow was ineffective, 

 and the vortex ripples originated along the end of the sand bed at the 

 top of the curved ramps and particularly near the wall in one corner; 

 presumably, this was due to an irregularity. Conditions for spontaneous 

 ripples and initiation of grain motion were distinct, and in three 

 experiments (81, 82, and 84) with the 0. 18 -millimeter sand and in two 

 experiments (101 and 103) with the 0.21 -millimeter sand, U was 

 increased beyond Uq to determine U^. Average ratios of Uj./U^ were 

 1.182 with the 0.18-millimeter sand and 1.156 with the 0.21-millimeter 

 sand. These ratios are fairly close to the 1.16 found by Carstens, 

 Neilson, and Albinbilek (1969) , and much less than the factor of about 

 2 cited by Bagnold (1946) . 



3. Rolling-Grain Ripples . 



With the 0.18- and 0.21-millimeter sands, unlike with the 0.55- 

 millimeter sand, a distinct bed form resembling the rolling-grain 

 ripples of Bagnold (1946) did appear. At U = U , soon after grain 

 motion began, rolling grains associated in bands normal to the flow. 

 These bands were not very straight; the length along their crests and 

 the spacing between them were rather uneven, and they tended to be 

 distributed in irregular patches over the sand bed. Their overall 

 appearance was more like a "rash" than a regular ripple pattern (see 

 Fig. 14). The figure shows the irregular bands covering the left half 

 of the bed while vortex ripples are advancing from the right. These 

 bands were quite stable and, even as U was increased above U^, showed 

 little tendency to become more regular, or to intensify or spread. Sur- 

 prisingly, the rolling-grain bands were observed to take no part in 

 vortex ripple formation. A band more than a few ripple lengths ahead 

 of an advancing pattern of vortex ripples would never independently 

 grow into a vortex crest and would remain unaffected by the advancing 

 pattern until the foremost crest had approached within a ripple length 

 or two. At this point the band would aline itself to merge with the 

 crest. This passive behavior is quite different from the more active 

 role described by Bagnold (1946) and Manohar (1955), and it might be 

 doubted that the bands of this study are rolling-grain ripples as 



45 



