In the case of railway bridges across the great shifting rivers of India, protection used at one time to be afforded by various systems of spurs. This has now been abandoned in favour of Bell’s guide banks (fig. 54), which are found to be far more satisfactory. These guide banks are discussed in the paper by Spring quoted above (_Art. 1_). The spaces behind the guide banks become filled with water, at least during floods, and are meant to be silted up. An opening in the railway embankment should be provided at A, and another on the opposite side of the river, to ensure a constant flow of water (CHAP. V., _Art. 3_), but they should not be large enough to cause high velocity. The chief danger to which a guide bank is liable is outflanking when the stream assumes the position shown. To guard against this danger it is necessary to have very strong and massive heads to the guide banks. When the bank of the eroding stream, downstream of the guide bank head, becomes a semicircle or thereabouts, the stream takes a short-cut across the sandbank, and to encourage this an artificial cut can be dug, at the season of low water, on any suitable line. [Illustration: FIG. 54.] If the guide banks were made with an increased width of opening at the upper end, this would reduce the chance of outflanking but would increase the danger from a direct attack such as indicated, in the figure, on the left bank. It has been suggested that the width at the upstream end should be less than at the bridge, but this seems undesirable. Probably the form shown in fig. 54 is the proper one. The length of the guide bank upstream of the bridge is made about equal to the span of the bridge between the two guide banks. If made less than this, the river might cut into the line of railway. The length of guide bank downstream of the bridge is generally 300 to 500 feet, being greater as the velocity of the river is greater and the sand of its bed finer. [Illustration: FIG. 55.]