3. =Weirs on Sandy or Porous Soil.=--If the channel is very soft or sandy the weir may be built on one or more lines of wells. The wells are not so much to support the weir as to form a curtain and prevent streams, due to the hydraulic gradient A E (fig. 33), from forming under the structure and gradually removing the soil. It is assumed in the case represented by the figure that the maximum head occurs when the downstream channel is dry. Any removal of soil from under the weir may cause its destruction. The wells should be as close together as possible, and the spaces between them carefully filled up with brickwork or concrete to as great a depth as possible, and below that by piles. Instead of wells, lines of sheet piling--cast-iron or wood--can be used. A good fit should be made, but it is not necessary that the joints should be absolutely water-tight. The object is to flatten the hydraulic gradient by increasing the length travelled by the water from B E to B L G H E. Of course, no flattening occurs at a point where the curtain is not water-tight, but if only small interstices exist, none but small trickles of water can pass, and the interstices will probably soon be choked up, just as the sand in a filter bed becomes clogged and has to be washed. In any case, no important stream could develop otherwise than round the toe of the curtain. It has been stated that when a curtain is water-tight the water follows the line B L M G H K E, but this requires proof. Another plan is to cover the bed and sides of the channel with a continuous sheet of concrete extending upstream of the weir from B to D--thus flattening the hydraulic gradient from A E to F E. Instead of concrete, clay puddle can be used with pitching over it. The choice between the different methods depends largely on questions of cost and facility of construction. It has been said that a certain amount of leakage occurs under structures such as the Okla weir (_Art. 4_), which nevertheless remains undamaged. There have, however, been cases in which failures of works have occurred, especially when there has been a great difference between the water-levels of the upstream and downstream reaches, from no other apparent cause than the passage of water underneath the works. Weirs in porous soils have been discussed by Bligh (_Engineering News_, 29th December 1910), who gives the following as safe hydraulic gradients (_s_) or ratio of the greatest head A B to the length B E:-- Fine silt and sand as in the Nile 1 in 18 Fine micaceous sand as in Colorado and Himalayan rivers 1 in 15 Ordinary coarse sand 1 in 12 Gravel and sand 1 in 9 Boulders, gravel and sand 1 in 4 to 1 in 6