If an embankment to keep out floods is to be made along a river which is so large that its flood-level cannot be appreciably affected by the construction of the work, it may be necessary to obtain information only as to the actual flood-levels, and as to the extent to which the stream is liable to erode its bank. If a length of the bank of a stream has to be protected against scour, it is necessary to know of what materials the bed and bank are composed, and whether the channel is liable to changes and to what extent. It is also desirable to know to what extent the water transports solids, if any. In some kinds of protective work these solids are utilised. But in cases where the stream is to be much interfered with, it is necessary to have full information concerning it, not only as regards water-levels, changes in the channel, and transport of solids, but as regards the longitudinal profile and cross-sections, and the discharges corresponding to different water-levels. The collection of some of this information, particularly as to the water-levels and discharges at different times of the year and in floods, may occupy a considerable time. Methods of ascertaining the quantity of silt carried in the water of a stream are described in CHAP. IV., _Art. 4_. Remarks regarding the other kinds of information required--the stream being still supposed to be large and perennial--are given in _Arts. 2_ to _5_ of this chapter. The degree of accuracy required in the information depends, however, on the importance of the work, and sometimes the procedure can be simplified. Detailed remarks on gauges and on the instruments used and methods adopted for observing discharges and surface slopes, are given in _Hydraulics_, CHAP. VIII. and Appendix H. 2. =Stream Gauges.=--Unless the stream being dealt with is an artificial one, it is unlikely that the flow in the reach with which the work is concerned will be uniform. The rise and fall of the water at one place cannot therefore be correctly inferred from those at another. It will be desirable to have two gauges, either read daily or else automatically, recording the water-level, one near each end of the reach concerned, with intermediate gauges if the reach is very long. If, in or near the reach, there is already a gauge which has been regularly read, it may be sufficient to set up only one new gauge, and to read it only for such a period of time as will give a good range of water-level, and to compare the readings with those of the old gauge. The readings of the new gauge for water-levels outside the range of those observed can then be inferred, but if the stream is very irregular this may involve some trouble (_Art. 4_).