CHAPTER XII
DRAINAGE AND FLOODS
1. =Preliminary Remarks.=--_Arts. 2_ and _3_ of this Chapter deal with the calculation of flood discharges, _Art. 2_ dealing with small streams, in which the water has to be got rid of, and _Art. 3_ with large streams. The remaining articles discuss the methods of predicting floods and of preventing them from doing damage. When the discharge figures have been arrived at in any case, the necessary masonry works can be designed in accordance with the principles described in CHAPS. X. and XI. For remarks regarding the design of channels and banks, see CHAP. IX., _Art. 4_, and also _Art. 6_ of the present Chapter.
In England, land near a stream or flooded area is said to be “awash” when the flood water rises to within 3 feet of the surface of the ground. The drainage of such land is apt to be unsatisfactory. If land is flooded or awash, it may be desirable to shift the outfall of a branch drain to a point lower down in the main outfall.
2. =Small Streams.=--In dealing with small streams, such as branch drains or natural streams not far from their sources, the engineer is concerned only with their maximum discharges. He has to design culverts, bridges or syphons to pass the streams under roads or other works, or to design channels or waste weirs for them. In a settled country there may be already some works in existence on the same stream, and these may form a guide, or it may be possible to obtain local information as to the height or volume of floods. Even in such a case rainfall figures will be most useful. In districts where there is no settled population, and in any case where the stream is ill-defined, and the flow fitful, the rainfall figures may afford the only, or at least far the best, means of estimating the discharge.
The rainfall to be considered in all these cases is the maximum likely to fall in a short period of time. The catchment areas dealt with are small, say 5 square miles or less. It must be assumed that the fall of rain extends to all parts of the catchment area, and that its duration is sufficient for the water from all parts of it to reach the site of the work. The different valleys or divisions of the catchment area should be considered separately, and regard must be had, not only to the area of each division, but to its length and declivity measured along the course of the stream which drains it. On these two factors depend the time taken by the rain water to reach the site of the work. The rate at which the rain water flows over the ground into rills or small subsidiary streams may be taken to be ¼ mile per hour in flat land, and 1 mile per hour on steep hill sides. The velocity of the current in the rills and larger streams is generally 2 to 4 miles per hour. It can, when necessary, be calculated roughly from the size and slope of the stream. To be on the safe side, the highest probable figure can be taken.
The time taken by the water to flow from the furthest points of the catchment area to the site of the work having been arrived at as above, the next thing is to estimate the probable maximum intensity of the rainfall during that time over the whole catchment area. The only figures immediately available will be the mean annual rainfall, or perhaps the maximum fall in twenty-four hours, but it has been shown (CHAP. II., _Art. 5_) how the probable maximum fall over a shorter period may be estimated.
The next thing to be calculated is the “run-off,” _i.e._ the probable proportion of the rainfall which will at once run off. This may be less than the proportion which will eventually become “available,” because some of it may go to feed the underground supply from which springs are fed. The proportion running off a small area in a short time would, under most circumstances, be rather difficult to estimate, but in the case under consideration, only the probable maximum figure is required. This occurs when the ground is saturated. Under these circumstances the ratio of the run-off to the total fall may be somewhat as follows:--
Steep rocky hillsides ·70 to ·90 Ordinary hills ·50 ” ·75 Undulating country ·40 ” ·50 Flat country ·30 ” ·35
The figures can be increased when the surface is specially hard or frozen, and decreased when it is soft, sandy, covered with woods or vegetation, or cultivated.
Whether or not the above procedure is necessary in its entirety depends chiefly on the size of the proposed work and on the degree of inconvenience likely to arise from any wrong estimation of the discharge.
In designing syphons to carry torrents across the Upper Jhelum Canal in the Punjab, the discharge from a catchment area of ·79 square miles was found to be about 4000 cubic feet per second. This is at the rate of about 5000 cubic feet per second per square mile, and is equivalent to a run-off of 7·8 inches in an hour. The catchment area was among low hills, not far from the Himalayas, and the declivities of the rills were very steep. The superintending engineer, Mr R. E. Purves, states[18] that the discharge observations were reliable, and that falls of rain of an inch in ten minutes occur not infrequently, even though the fall in twenty-four hours might not exceed 2 or 3 inches. In order to account for the discharge in the case under consideration, it would at first seem to be necessary to assume not only that a fall at the rate of 7·8 inches per hour had occurred, but that the whole of it had run off. It is not, however, necessary to assume quite so much. The ground being saturated, the rain falling in a period of five minutes might be reaching the discharge site with little loss. A suddenly increased fall at the rate of 6 inches per hour might then occur, and the water travelling more quickly and with hardly any loss, would overtake that already passing the site. This case seems to show that for a very small catchment area the whole of the fall, and more, must be allowed for.
The Chief Engineer of the Punjab did not accept the above figures.[19] He remarked that observations taken under great difficulties as to time and place are liable to error, and he considered that an allowance of rainfall at the rate of 4·8 inches per hour--a rate which had been observed elsewhere--and a run-off of ·75 of the fall would be sufficient. He accepted a discharge of 2000 cubic feet per second for catchment areas of less than 5 square miles, assuming the run-off to be ·75 of the fall, but afterwards increased the figure to 2400 cubic feet per second. The chief engineer did not overlook the fact that in the designs for the drainage aqueducts a free-board of 5 feet had been allowed, and perhaps this led to an acceptance of an estimated discharge less than would otherwise have been accepted. It does not seem to be at all certain that the figure put forward by Mr Purves was far wrong. When the original project estimate for the Upper Jhelum Canal was framed, the irrigation engineers had had no experience of small and steep catchments, and no one had suspected that the discharge per square mile would be anything like the above. The sums of money provided for works for the passages of torrents had to be increased in ratios varying from 2·5 to 1 to 6 to 1.
The following statement shows the figures for other small catchment areas in the neighbourhood of the Upper Jhelum Canal:--
+----------+------------------+---------+ |Catchment | Discharge per | Run-off.| | Area. | square mile. | | +----------+------------------+---------+ |Sq. miles.| Cub. ft. per sec.| Inches. | | | | | | .79 | 5000 | 7·8 | | 1·47 | 3825 | 5·82 | | 2·96 | 2214 | 3·46 | +----------+------------------+---------+
In the south-east of New South Wales flood discharges of 135 and 84 cubic feet per second have been found for catchment areas of ·91 and 2·5 square miles respectively in broken country.
3. =Rivers.=--It is possible to apply the methods of the preceding article to large catchment areas, but the results would be quite unreliable. If the calculations were made so as to err on the side of safety, the resulting discharges would often be enormous. The following table shows some figures based on actual flood discharges. None of the localities have excessive rainfalls, though most are liable to occasional very heavy falls. In mountainous districts in the North of England and in Scotland the flood discharges per square mile of catchment area have been found to vary from 64 to 320 cubic feet per second.
+-----------+---------+-----------+---------+---------------+----------+ | | | | |Flood Discharge| | | Reference |Country. | Locality. |Catchment| per sq. mile | Remarks. | | | Number. | | | Area. | of Catchment | | | | | | | Area. | | +-----------+---------+-----------+---------+---------------+----------+ | | | |Sq. miles| Cub. ft. | | | | | | | per sec. | | | 1 | India. | Upper | 5 to 10 | 1613 | | | | | Jhelum. | | | | | 2 | ” | Nagpur. | 6·6 | 480 | | | 3 | South |Near Cape | 34·5 | 78 | | | | Africa. | Town. | | | | | | | | | | | | 4 | ” |Near Port | 35 | 640 |Estimated.| | | |Elizabeth | | | | | 5 |New South|South-East | 49 | 37 | | | | Wales. | District. | | | | | 6 | India. | Upper | 56 | 1000 | | | | | Jhelum. | | | | | 7 | ” | ” | 174 | 550 | | | 8 |New South|South-East | 418 | 11·2 | | | | Wales. |District. | | | | | 9 | India. |Kali Nadi | 2593 | 51 |Estimated | | | | Stream. | | or more |roughly. | +-----------+---------+-----------+---------+---------------+----------+
The tendency of the figures in column 5 of the table is to decrease as the catchment area increases. This tendency has long been known, and attempts have been made to found on it formulæ for calculating flood discharges. One such formula is Q = _c_ M^{3/4} where Q is the flood discharge in cubic feet per second and M is the area of the catchment in square miles. The formula is roughly correct, _c_ being a constant for catchment areas of not dissimilar characters and with rainfalls not differing much. But for other cases there is no knowing how _c_ may vary, and this renders the formula practically useless. The author of another such formula quotes cases Nos. 5 and 8 in the above table, and the two cases mentioned at the end of _Art. 2_ as agreeing fairly well with the result of his formula. The tendency just mentioned is due to the fact that every river is composed of tributaries which have their own small catchment areas but are, when measured to the general outlet or point where the discharge is under consideration, of very different lengths, to the improbability of heavy rainfall occurring over all these small areas at such times as to cause the different flood waves to arrive simultaneously at the outlet, and to the facts that in the case of the longer tributaries the flood waves flatten out (_Hydraulics_, CHAP. IX., _Arts. 3_ and _4_) so as to arrive more gradually, and that, unless rain is also falling all along their courses, these longer tributaries undergo losses from evaporation and absorption. But occasionally it happens that the various flood waves do arrive at the outlet more or less simultaneously, and that the rainfall continues so long and is so widely distributed--though not necessarily of the same intensity as that which caused the flood--that the flood waves do not flatten out and that losses in the channels do not occur. Floods can thus vary to an extraordinary degree in severity, and formulæ are quite useless. This is why floods occur surpassing all previous records, as, for instance, the recent floods in Paris. However severe a flood may be, it can never be said that the maximum has, even probably, been attained unless it can be shown that the rainfall has been so heavy, so long continued, and so distributed that anything worse is not likely to occur.
The best method of estimating the flood discharge of a large perennial stream is to ascertain, by local inquiry, the height to which it is known to have risen, and to take cross-sections of the channel and calculate the discharge (CHAP. III., _Arts. 4_ and _5_). In designing works, allowance can be made for a flood exceeding any known before. This method applies also to a case in which a river is formed by the junction of two or more large tributaries. It is possible that the tributaries have not, within the memory of man, been in high flood simultaneously. If so, the chances of this occurring are no greater and no less than if the stream was composed merely of a number of small affluents. Remarks regarding intermittent streams are given in CHAP. III., _Art. 7_.
Since an acre contains 43,560 square feet, and a twelfth of this is 3630, it follows that a fall of 4 inches of rain, of which 1 inch runs off, in an hour, gives a discharge of 3630 cubic feet per hour, or about 1 cubic foot per second. This is 640 cubic feet per second for a square mile. The figures in column 5 of the above table show that the run-off was, in the cases quoted, generally far less than 1 inch. In case No. 4 it was 1 inch, and in case No. 2 it was 3/4 inch.
In the case of the Kali Nadi (No. 9 in the table) an aqueduct to carry the Lower Ganges Canal over the stream was being designed. The flood discharge, estimated from the supposed flood-level and cross-section of the stream was (_Min. Proc. Inst. C.E._, vol. xcv.) 26,352 cubic feet per second. The discharge, estimated by assuming a fall of 6 inches of rain in twenty-four hours over the catchment area--then believed to be 3025 square miles--and a run-off of ·25 of the fall, was 114,950 cubic feet per second. This figure was rejected on the ground that the rainfall would not be continuous over so large an area as 3025 square miles. An allowance of 7 cubic feet per second per square mile was made and, a fresh survey having shown that the catchment area was only 2593 square miles, a discharge of 18,000 cubic feet per second was allowed for. The aqueduct was built, about the year 1875, with five arched spans of 35 feet each, the total area of the waterway being about 3000 square feet. The length of the piers and abutments was 212 feet, the width of the canal carried over the aqueduct being 192 feet. In 1884 the aqueduct was partly destroyed by a flood whose discharge was about 44,000 cubic feet per second. In July 1885 it was wholly destroyed by a flood whose discharge was estimated at 132,475 cubic feet per second, but was probably more. The discharge must have been more than 51 cubic feet per second per square mile. The aqueduct was rebuilt with a waterway of about 15,000 square feet. Below the aqueduct there was a bridge which had been standing for a hundred years. Its waterway was only 1146 square feet. It was not much damaged by the flood of 1884, but much of the water passed round it, breaking through the embanked roadway or pouring over it. It is understood that the bridge was destroyed by the flood of 1885.
This case shows the necessity for making every possible allowance in calculating flood discharges for important works. The smallness of the discharge, as calculated from the cross-section of the stream, was probably owing to its being dry when the survey was made, so that the velocity could not be observed, but it is probable that such a discharge as wrecked the aqueduct had never before passed down the stream.
4. =Prediction of Floods.=--At any place high up on the course of a stream, the occurrence of a flood can often be predicted when rain storms--often accompanied in the tropics by lightning--can be seen to be occurring towards the sources of the stream. For any station lower down the stream and for precise information in any case, the readings of gauges higher up the stream can be telegraphed. If the station is at a great distance from the gauge and if there is railway communication, the readings can be sent by post.
In order to be able to predict the time of the arrival of a flood at the lower station the reading of a gauge there, and also of that at the upper station, should be taken at frequent intervals. In the case of large rivers and distances of hundreds of miles, the interval may be six or even twelve hours, but in other cases it should be much less. If the readings are plotted, as in fig. 56, oblique lines can be drawn to connect the saliences and depressions, and the time taken by each change can thus be readily seen. When the upper part of the stream is formed by two or more important tributaries there should be a gauge in each.
As to what constitutes a flood, the gauge diagram of a river (fig. 56) is generally such that a line can be sketched as shown dotted. The rises above this line are floods. The maximum flood discharge of a Northern Indian river is estimated roughly as being 100 times the low-water discharge. Leslie’s rule for floods in the British Isles is that if all the daily discharges of a stream during the year are ranged in order of magnitude, the discharges of the upper quarter are considered to be floods.
In India it is sometimes arranged that a telegram shall, in the low-water stage of the river, be sent from the upper station when a rise of 2 feet occurs in twenty-four hours or any less period, with a further telegram for any such subsequent rise. The telegram states the exact reading on the gauge and whether the water is rising steady or falling. This is given as indicating the procedure that may be followed where the telegraph has to be used, but when long and frequent telegrams are not desirable.
The advancing end of a flood wave may, while the wave is rising and being formed, travel rapidly, but when the wave has been formed it travels at the ordinary rate of flow of the risen stream. The advancing end of a trough may, while it is being formed, travel rapidly, but after formation it travels at the ordinary rate of the fallen stream (_Hydraulics_, CHAP. IX., _Arts. 3_ and _4_). Thus the rate at which a change in water-level travels down a stream depends at first on the amount of the rise or fall, but afterwards on the water-level of the risen or fallen stream.
By taking the above facts into consideration and noting the actual times obtained from the diagram, it will be possible to arrive at the probable time that will be taken by any change. It will also be possible to predict the height of the flood. If it is worth while, an empirical formula can be got out. If there are tributaries, each with a gauge, the matter will be more difficult. Probably the floods in the tributaries will arrive at different times, but even in such cases empirical formulæ have been arrived at, especially in France, and are mentioned in various volumes of the Proceedings of the Institution of Civil Engineers.
In all cases predictions are liable to be more or less upset if rain falls in the tract between the upper and lower gauges. In very dry weather the speed of a flood wave may be somewhat reduced, and the height to which it rises will almost certainly be reduced.
The full effect of a change will not be felt at the lower station unless the change at the upper station is maintained for a sufficiently long period. A short wave or trough flattens out. Thus in any empirical formula or system of prediction, the time over which the change extends at the upper gauge must be taken into account, or else there must be several upper gauges and the readings of all of them be taken into account.
In mountainous districts landslips sometimes occur and block the valley of a stream which then forms a lake. The water gradually rises and eventually flows over the dam and sweeps it away causing a flood, which is of great suddenness and height but decreases very quickly in height as it travels down the valley. In a case which occurred in the Himalayas in 1888 the inhabitants of the valleys, from the dam to the point where the river debouches from the hills, were compelled by Government to vacate all habitations below the probable level of the flood, and no loss of life occurred. Similar floods, but on a smaller scale, may be caused by the bursting of ordinary reservoir dams. In some continental rivers ice may obstruct the stream and cause floods.
5. =Prevention of Floods.=--The extended use of field drains has, in recent years, done much to increase the severity of floods in England and other countries. One method of mitigating or preventing floods is the construction of reservoirs for storing the water. Reservoirs locally known as “washes,” formed by setting back the embankments, exist on the Fen rivers. One wash, on the Nene, below Peterborough, is 12 miles long and half a mile wide and is filled, in floods, to a depth of 7 feet and holds 1 inch of rainfall over the river basin, and this is found to be sufficient. Reservoir construction is, however, in most cases, impracticable owing to the expense. To store the water which is given by 1 inch of rain in the basin of the Thames, a reservoir would be needed 50 feet deep and covering about 7 square miles. It might cost £7,000,000.
The afforestation or reforestation of river basins (CHAP. II., _Art. 4_) is also occasionally undertaken, but is not generally practicable.[20]
The most practicable methods for preventing flooding are lowering the water-level of the stream and constructing embankments along it. These will be considered in the next two articles.
6. =Lowering the Water-Level.=--The water-level of a given length of stream can be lowered by lowering the bed, widening the channel or straightening the channel. The efficiency of these processes is in the order named. As stated in CHAP. I., _Art. 4_, the alteration to the channel must in any case be continued to some point downstream of the reach under consideration. Let the channel be supposed to be of “shallow” section with sloping sides. Let W be the mean width, D the depth, and S the slope. Let it be required to lower the water-level by an amount equal to D/5. This can be effected by lowering the bed by about 25 per cent. of D, or by increasing the width by about 50 per cent., or by increasing the slope by about 100 per cent. If the bed is lowered, V is not affected, and the mean width is reduced. Increase in W reduces D, and therefore reduces the hydraulic radius and the velocity. Hence the large amount of widening necessary. When S is increased the velocity, if R remains the same, is affected only as √S (_Hydraulics_, CHAP. VI., _Art. 2_), but the depth of water is reduced and R therefore reduced. Dressing the sides of a channel, so as to make it smoother, produces the same effect as a slight widening.
It does not, of course, follow that lowering the bed is always the best plan and straightening the worst. Any one of the processes may be more or less impracticable because, for instance, of the hardness of the material to be removed, or the expense--including compensation--of removing obstructions.
A particular kind of widening consists in digging a new channel and keeping both the new and the old channel open.
If a channel contains a weir, or a local raised portion of bed forming a kind of submerged weir, or a contracted place or narrow bridge, the upstream water-level can be lowered by simply removing or reducing the obstruction. The lowering of the water-level will be greatest at the site of the obstruction, and will be zero at some point far upstream (_Hydraulics_, CHAP. VII., _Art. 5_). If the raised portion forms a long shoal, its removal--supposing its height above the general bed to be the same--will have more effect than if it were short. If the height of the raised portion is small compared to the depth of water, or the amount of contraction small compared to the width of the stream, the removal may have much less effect than might appear (CHAP. I., _Art. 4_).
In soft soils one advantage of the straightening system for lowering the water-level is that short-cuts can be dug to a small section, and left to enlarge themselves (CHAP. VII., _Art. 1_).
Another advantage is that after any diversions have enlarged themselves to the size of the rest of the channel--or have originally been so excavated--the whole channel may scour, and the water-level continue to fall. This, of course, should be allowed for if likely to occur.
The same thing may occur in the case of the removal of a weir, shoal, or contracted piece of channel. The scour will act at first close to the site of the obstruction, but it may work upstream.
In widening or deepening a channel for the purpose of mitigating floods, it is a good plan to begin work at the downstream end, because the lowering of the water-level will extend upstream beyond the reach in which work is done, and this may facilitate work further upstream. As regards any tendency for a widened reach to silt up again, any such silting is not likely to be great in a short period of time, and need not prevent the carrying on of work in various reaches, if this is convenient.
7. =Flood Embankments.=--A flood embankment may be close to the edge of the river or it may be set back. If set back it need not follow all the windings of the stream. The setting back of an embankment gives an increased waterway to the stream during floods, and therefore a lower flood-level, but the effect of this is trifling in cases where the depth of the water on the flooded land is small, especially if such land is covered with vegetation, or is otherwise much obstructed. Setting back is generally necessary in cases where the stream is liable to erode the banks to any considerable extent. In such a case the embankment should not be so near to the river as to be in much danger from erosion, but the ground, as already stated (CHAP. IV., _Art. 9_), generally falls, in going away from the river, so that when an embankment is set well back it is in lower ground, more expensive and more liable to breach. The most suitable alignment is a matter of judgment, and depends largely on the extent to which the river is likely to shift.
Embankments should, where possible, be made in straight or properly curved reaches. A flood embankment, at least at its upstream end, should terminate in ground which is above flood-level. The top of an embankment should be, in the case of a large river, 2 or 3 feet above the high flood-level of the river. It should, of course, be graded parallel to the general high flood-level, but neither the gradient nor the height of the flood is usually known with accuracy (CHAP. II., _Arts. 1_ and _2_). There is generally a record or mark of some high flood, and this is taken provisionally as the flood-level. Or the level is calculated approximately from the flood readings on the nearest river gauge. If experience shows that the embankment is too low, it is raised. The cross-section of an embankment depends on the soil, on the extent of damage which results if a breach occurs, on the funds available, and on the value of the land which the embankment has to occupy.
Where an affluent enters the river it will probably be necessary to run out branch embankments. Sometimes cross embankments are run from the main embankment to high land. Their object is to localise the damage if a breach occurs. Along the back of the embankment there may be a drain and it can be made to discharge its water, when the river is not in flood, through the embankment by means of sluices or by pipes closed by flap valves which will not allow flood water from the river to pass through. There may be sluices in the embankment for the purpose of irrigating the land at the back.
The immediate effect of the construction of flood embankments along a river is to raise the water-level, because the floods can no longer spread out over the country, but this effect will not be great if the sectional area of the flood water was small or its velocity low. The river may or may not tend, after the construction of flood embankments, to raise or lower its bed. It has already been remarked that questions of silting or scouring cannot be answered in a general manner. In the case, however, of floods spilling over a piece of country, the depth of the flood water is generally small and the country more or less obstructed. Some deposit of silt generally occurs. The construction of an embankment reduces the area of the flood water, and thus generally reduces the silting and leaves more silt in the river proper. The depth and velocity in the river are increased. Everything depends on which is increased most. Most likely the stream is of shallow section and the velocity is increased most (CHAP. IV., _Art. 6_, par. 6), and the increased silt-supporting power may make up for the increased charge of silt.
Sometimes when a main embankment is set far back, a subsidiary embankment of smaller section is constructed closer to the stream. This is often objectionable. The smaller embankment is liable to breach, and the water then rises suddenly instead of gradually against the main embankment, which is thus endangered to some extent, especially as it is dry instead of being soaked.
It is often said that one effect of embanking a reach of a river is to increase the severity of floods further downstream. The importance of this is generally exaggerated. The narrowing of the flood stream in the embanked portion causes the flood to travel more quickly and rise higher in that particular reach. At a place further downstream the same effect is produced, but in a less degree and only because of the increased velocity and consequent reduction in the flattening out of the flood wave, especially when the rise is soon succeeded by a fall. When there is a gradual rise lasting for a considerable time--and this is most likely to cause a high flood--there is no rise of the flood-level downstream of the embanked reach, except such as is due to the increase in the discharge of the stream consequent on the absorption and evaporation being less than before, owing to the reduced area of flooding in the embanked reach. In the case of a long-continued rise, such as that just mentioned, it is the reach immediately upstream of the embanked reach which will, to some extent, share in the increased height of the floods.
An embankment may suitably have side slopes of 4 to 1 on the river side and 3 to 1 on the land side, with a top 10 feet wide and 3 feet above high flood-level. On the Irrawaddy the top width is generally 8 feet. For very high and very low embankments it is 10 feet and 3 feet respectively. In Holland 1 foot above high flood-level was at one time supposed to be the rule, but in practice it was usually 4 feet. With sandy soil the riverward slope prescribed was 6 to 1. Such flat slopes are not necessary if fascining or stiff soil is used as a protection. On the Rhine the top width of embankments consisting of gravel and sand has been made about 15 feet, but the side slopes were 1½ to 1 and 1 to 1. The embankments had spurs to keep off the current.
Sand, protected as above, makes a good embankment, and rats do not burrow into it. Of course, if a breach occurs in an embankment consisting mainly of sand, it will enlarge very quickly. In some cases an embankment has a core wall of sand or of clay puddle. In Holland, on sandy soil, a trench 8 feet wide is made and taken down to the clay.
Embankments require to be made with great care. The earth should be deposited in layers. In Holland, horses are driven up and down over each layer. In some parts of India the earth for embankments is brought from the borrow pits by scoops drawn by bullocks. The earthwork is of so excellent a character, owing to the earth being trodden down, that no settlement has to be allowed for. Where the soil is sand the top and faces of the embankment should be of good stiff soil, if it can be obtained, for a thickness of 9 inches or a foot, or else the face next the river should be protected by fascining (CHAP. VI., _Art. 3_) for 2 feet above, and several feet below high flood-level. Such protection may be necessary in any case where waves are liable to occur. In Holland embankments are turfed, and trees and shrubs are not allowed to grow. In the Punjab the growth of all kinds of jungle is encouraged. It binds the soil together and protects it from the wash of waves and from winds which blow away sand and dust, and so wear the embankment slowly away.
In embanking a long reach of a river it is convenient to begin from the upstream end, because otherwise floods may get behind the finished part of the embankment and, becoming impounded in a “pocket” formed by the embankment and high land, rise to an abnormal height and, unless gaps in the embankment have been left or are subsequently made, cause breaches.
During high floods pegs should be driven in at frequent intervals, to mark the high flood-levels. If a higher flood occurs, the peg is shifted. The levels of the pegs can be observed at leisure.
When a breach occurs in an embankment, the first thing to do is to protect the ends so that the breach shall not lengthen. If the water passing through a breach becomes pocketed, the embankment may have to be cut to let it out.
Regarding the stoppage of leakages, see CHAP. IX., _Art. 1._ Regarding the closure of breaches, see CHAP. VII., _Art. 2_.
For a description of flood embankments along the great shifting rivers of Northern India, see _Punjab Rivers and Works_.
_Note to Art. 5._--Floods can sometimes be mitigated by sinking pits in the flooded area so that the flood water comes in contact with permeable strata and is absorbed by them.