(8) At such a bifurcation it may be necessary, during times of low supply, to head up the water in the main channel and some silt may temporarily be deposited in it. When the heading up ceases, the silt is scoured away but it mostly goes into the branch whose bed level is the lower. It is best to design such bifurcations so that the sill levels of the two branches are equal and, if possible, so that their bed levels are equal.[11] Otherwise the channel which is likely to get most silt should have the steeper gradient. (9) Any existing well established régime should not be tampered with. [11] Appendix A in _River and Canal Engineering_ deals with some instances of fallacies in questions concerning flow in open streams. An extract from it describing a remarkable divide wall recently constructed at the head of the Gagera branch, Lower Chenab Canal, is given in Appendix A of this book. Experience shows that in designing Irrigation Channels in the plains of India in accordance with Kennedy’s figures, the maximum ratio of bed width to depth of water is as follows:-- Discharge, c. ft. per second 10 25 100 200 500 1,000 Ratio 3·5 4 4·5 5 6 6 The actual gradients of the canals generally range from about 1 in 8,000 for a main canal to 1 in 2,000 for the tail of a distributary, but near the head of a canal where the bed is of boulders and shingle, the gradient may be as steep as 1 in 1,000.[12] The velocity in this last case may be 5 feet per second but generally it is not more than 3 or 4 feet per second in canals and branches, and 1 to 2 feet per second in distributaries. [12] On the Upper Jhelum Canal, 1 in 970. In designing the channels, N, in Kutter’s co-efficient, may be taken as ·0225 or ·020, according to judgment. For new and smooth channels ·020 is generally correct. A channel generally becomes rougher by use but sometimes it becomes smoother. Cases have occurred in which N has been found to be ·016. This question is discussed in _Hydraulics_, Chap. VI. The bed width of a canal is reduced, where a distributary takes off, in such a way that when the canal and distributary are both running full, the depth of water in the canal continues to be uniform and the flow to be uniform. When the distributary is closed there is heading up in the canal upstream of the off-take, but not enough to make any appreciable difference unless the capacity of the distributary is a large fraction of that of the canal and even then no harm is likely to result.