Estimating soil loss from measurements of sediment movement in streams and rivers faces several problems. Taking the measurements is time consuming and expensive; the accuracy of the measurements is likely to be poor; and even if there are good data on the movement in a stream it is not known where the soil came from and when. Some of the technical problems are discussed in Dickinson and Bolton (1992). However, it may be useful to be able to make comparisons of the movement in different streams, or at different times of year, or from watersheds under different land uses. It was explained in Chapter 1 why reliable quantitative data require calibrated paired watersheds, and why 'before and after' treatments should be avoided.
Sediment movement in streams and rivers takes two forms. Suspended sediment is the finer particles which are held in suspension by the eddy currents in the flowing stream, and which only settle out when the stream velocity decreases, such as when the streambed becomes flatter, or the stream discharges into a pond or lake. Larger solid particles are rolled along the streambed and called the bedload. There is an intermediate type of movement where particles move downstream in a series of bounces or jumps, sometimes touching the bed and sometimes carried along in suspension until they fall back to the bed. This is called movement in saltation, and is a very important part of the process of transport by wind, but in liquid flow the height of the bounces is so low that they are not readily distinguished from rolling bedload.
The relative quantities moved in suspension and as bedload vary greatly. At one extreme, where the sediment is coming from a fine-grained soil such as a wind-deposited loess, or an alluvial clay, the sediment may be almost entirely in suspension. On the other hand, a fast-flowing clear mountain stream may have negligible amounts of suspended matter and almost all the movement by rolling gravel, pebbles and stones on the streambed. Very high concentrations of sediment, as occur in some rivers such as the Yellow River in China and the Mississippi in the USA, may cause significant changes in the rheological properties of the water. The viscosity is higher and the particle settling velocity much lower, so that the threshold between suspended sediment and bedload becomes blurred.
The estimation of suspended load by sampling is relatively simple, but taking a representative sample of bedload is difficult. Both types of sampling are briefly reviewed, also estimation of the total sediment movement, and estimates based on measuring the amount of deposition in ponds or lakes.
There are several sources of error associated with trying to correlate the amount of sediment measured in streams with the extent of erosion within the watershed.
Firstly, there may be significant amounts of erosion taking place which do not contribute to sediment in the stream because the eroded material is deposited before it reaches the stream. The proportion of sediment which does reach the stream compared with the gross sediment movement within the watershed is called the delivery ratio. It could be as low as 1% if there are depressions or heavily vegetated areas where most of soil would be deposited. In a countrywide study of 105 agricultural production areas in the USA Wade and Heady (1978) found delivery ratios varying between 0.1% and 37.8% of the gross erosion.
A second possible source of error is the time factor. In a larger watershed sediment may be eroded and deposited, then eroded again and redeposited, and this process could be repeated a number of times before the sediment reaches the stream. A sample of sediment could include material which had been originally eroded several years previously.
The third difficulty is that the sediment in the stream includes material which has come from several different sources with widely different delivery ratios. Sediment caused by the collapse of gully sides or river banks passes immediately into the streamflow, whereas soil lost from a small cultivated area within a dominantly forested watershed might have high local rates of erosion but contribute little to the total sediment load.
Estimates of the total sediment discharge in streams may be made by combining estimates of the sediment concentration with velocity of flow. The section Velocity/area methods in Chapter 4 explained how velocity varies at different places in the stream,and how an average velocity may be calculated from a series of measurements (Figure 22). The sediment concentration also varies, being usually greater near the bottom, and so the amount of total sediment discharge is the product of these two variables, as in Figure 37.
The simplest way of taking a sample of suspended sediment is to dip a bucket or other container into the stream, preferably at a point where it will be well mixed, such as downstream from a weir or rock bar. The sediment contained in a measured volume of water is filtered, dried and weighed. This gives a measure of the concentration of sediment and when combined with the rate of flow gives the rate of sediment discharge.
A study of alternative sampling techniques in South Africa showed that dip sampling in bottles generally gives concentrations about 25% lower than results obtained from more sophisticated techniques (Rooseboom and Annandale 1981). For single samples taken by scooping a sample, a depth of 300 mm below the surface is recommended as better than sampling at the surface. If the single sample can be taken at any chosen depth, half the depth of flow is recommended as giving the best estimate of average sediment concentration. Where the sampling programme consists of samples on vertical sections at several points across the stream, the recommended pattern is to use six equally spaced sections as shown in Figure 38.
One can allow for variations in sediment concentration at different points in the stream by using an integrating sampler, that is one which gives a single sample combined from small sub-samples taken from different points. A typical sampler is illustrated in Figure 39 and consists of a glass bottle inserted in a fish-shaped frame mounted on a rod when gauging small streams or suspended on a cable for larger streams. For the bottle to fill smoothly and evenly when below the surface it is necessary to have one nozzle or orifice for entry of water, and a second pipe through which the displaced air is ejected. The entry nozzle is usually designed with a slightly expanding cross-section behind the point of entry in order to reduce the risk of back pressure which could interfere with the flow into the bottle. In operation, the sampler is moved from the surface down to the bed and back up to the surface while sampling continuously. A few trial runs will establish how long is required for the bottle to fill during this double journey. It is undesirable for any type of bottle sampler to continue to accept more inflow after the bottle is full as this can lead to an accumulation of sediment in the bottle. In some depth-integrating samplers the bottle is lifted out of the flow when or just before it is filled; other types of sampler may have some device to stop further sampling once the bottle is full.
The point-integrating sampler remains at a fixed point in the stream and samples continuously during the time it takes for the bottle to fill. Opening and closing the valves of the sampler are controlled from the surface electrically or by cables. Samples should be taken at a number of depths at each of several vertical sections, as described in the section Velocity/area methods in Chapter 4, for the gauging of streams by the current meter method, so these two operations are often carried out at the same time.
Another method of obtaining samples at various depths of flow is to use automatic samplers which will take a sample at a predetermined depth of flow. A typical example is shown in Figure 40 using a milk bottle and two lengths of bent tubing. Factory-made versions usually use copper tubing accurately bent to shape, but a simple field version can be made using plastic tubing which is tied to a rigid framework to hold it in place. The bottle starts to fill when the depth of flow reaches point A and starts syphon flow into the bottle, and ceases when the depth of flow rises to point B which is the outlet of the air exhaust pipe. The range of sampling is controlled by adjusting the distance between points A and B. In its simplest form the intake and exhaust pipes are bent into a simple U shape, but this means that flow into the sampler is at right angles to the streamflow, and this may interfere with the sediment concentration, so a more sophisticated version has the two pipes with a second bend to point upstream into the flow, as in Figure 42.
An ingenious home-made automatic sampler was used Kenya and is illustrated in Figure 41. The operation is described as follows.
"The surface samples are collected through a small-bore brass tube which projects upstream to avoid the turbulence caused by the assembly. The tube delivers the water into a standard one-pint milk bottle inside which there is a flexible rubber ball (standard squash ball) that floats up and seals off the bottle internally when full.
"The sub-surface sample is also taken in a standard milk bottle that it initially sealed by a rubber ball attached to a rod projecting from an inverted milk bottle as shown in the diagram. The length of the rod is such that the upper bottle begins to float, thus unsealing the sampling bottle, just as the water level has reached the orifice of the tube in the companion pipe. The surface and sub-surface samples are therefore taken at more or less the same time in the flood. The sub-surface sampling bottle is sealed internally when full with a floating squash ball as before. A further set of pipes is designed to sample the flood at a higher stage, as shown in the diagram." (Pereira and Hosegood 1962).
If it is required to take a series of samples as the river rises, then an automatic sampler can be set up as shown in Figures 42 and 43. The concentration of suspended sediment is usually higher when the river is rising than when it is falling.
Samples can be manually pumped from a stream as shown in Figure 44. However, for nearly twenty years automatic samplers have been available which can pump a small sample into a series of bottles, either at predetermined times and intervals, or as triggered by predetermined flow conditions, usually depth (USDA-ARS 1976). In the early years these devices were usually rather large and cumbersome and subject to malfunctions of the many electrical or mechanical components. However, solid-state electronics have led to a new generation of automatic samplers which are smaller, more reliable and less expensive. Two examples are illustrated in Plates 32 and 33. Modern programmers and timers linked to a float switch or a pressure transducer in the stream allow an almost infinite variety of sampling programmes. Suction lifts of up to six metres are possible, so the sampler can be sited above flood level. Plate 34 shows the Wallingford sampler with an HL flume in the Philippines.
The pattern of rising and falling rates of flow, and the variations of sediment concentration at different flow rates, can to some extent be anticipated from observations, but even sophisticated automatic samplers cannot predict what is going to happen next. An automatic pumping sampler may perform exactly as programmed and take a series of samples as required on a rising and falling flood, but then there is another extreme rainstorm in the catchment which produces an even bigger flood which cannot be sampled because the bottles are all full. The only way this possibility can be avoided is to have a system of continuous measurement of sediment concentration, and two methods are possible.
An optical turbidimeter passes a beam of light through sediment-laden water from a source on one side of a channel to a sensor at the other side. The sensor may either measure the extent to which light is absorbed by the sediment particles, and this is called the attenuation principle, or measure the extent to which the light is scattered by the suspended particles, called the scatter principle. Like pumping samplers, they can be programmed to react to combinations of depth of flow and time, but since the data are recorded in a Read-Only-Memory (ROM) device, any amount of data can be stored. Successful examples of turbidimeters have been developed in South Africa (Grobler and Weaver 1981) and in Indonesia (Brabben 1981).
A similar principle is used in nuclear gauges which measure either the absorption or the scattering of gamma radiation instead of light. The successful use of such instruments in many countries is reported by Walling (1988) who suggests, "However, the instruments are more sophisticated than optical turbidity meters and have been restricted mostly to specialised experimental measurements rather than routine monitoring".
The simplest way to estimate bedload is to dig a hole in the streambed as in Figure 45 and remove and weigh the material that drops into it. The basin upstream of a weir or flume can similarly act as a sediment trap, but it may not be known whether all the bedload has been trapped. Where heavy loads occur this process can be very time consuming and laborious.
Estimates of bedload may be obtained from samples caught in a device which is lowered to the streambed for a measured time then brought up for weighing the catch. Many such devices have been used, and the variety demonstrates the difficulty of taking an accurate and representative sample. The problems with bedload samplers are:
· The sampler disturbs the flow and changes the hydraulic conditions at the entry into the sampler.
· The sampler has to be resting on the streambed and tends to dig in as scour occurs round it.
· To remain stable on the bed it has to be heavy, and this restricts the use to lowering from bridges or purpose-built gantries.
· A sampler needs to rest on a reasonably smooth bed and not perch on large stones or boulders.
The simplest form is a wire basket with a stabilizing tail fin as illustrated in Figure 46. The catch of such devices is low because they interfere with the flow and some material is deflected round the sampler, increasingly as the basket fills up. This is described by saying that back pressure reduces the flow into the sampler, and this description conveys the right image without going into the mechanics of fluid flow. Some samplers have a diverging section behind the orifice, which allows entry to the sampler at the surrounding stream velocity. These are called pressure-difference samplers and an example is illustrated in Figure 47).
Maddock's classification for estimation of the bedload (Maddock 1975)
|Suspended sediment concentration
parts per million
|River bed material||Suspended elements texture||Bedload discharge expressed as % of suspended sediment discharge|
|less than 1000||sand||similar to the river bed||25-150|
|less than 1000||gravel, rocks, hard clay||low sand content||5-12|
|1000 - 7500||sand||similar to the river bed||10-35|
|1000 - 7500||gravel, rocks, hard clay||25% sand or less||5-12|
|more than 7500||sand||similar to the river bed||5-15|
|more than 7500||gravel, rocks, hard clay||25% sand or less||2-8|
A number of studies report the use of radio-active tracers to monitor the bedload movement. The technique is to insert into the stream a radio-active tracer in a form similar to the bedload, that is it should have the same shape, size and weight as the natural sediment. The movement downstream can then be monitored using portable detectors. Alternatively, the tracer can be applied to the surface of naturally-occurring sediment, or it can be incorporated into artificial materials which can be made radio-active by irradiation (Tazioli 1981).
The difficulty of obtaining reliable measurements of bedload has led to some attempts to calculate it from more easily measured parameters but these are not widely used. A very simple method based on knowing the suspended sediment concentration and the texture of both suspended material and bed material is given in Table 10. A sophisticated approach was developed by Einstein (1950) and has been later modified and improved. There are many other theoretical formulas, and much debate on their accuracy and reliability.
A method of avoiding separate estimates of suspended load and bedload is to mix up all the moving sediment and then take a single sample of the mixture. A turbulence flume is a purpose-built structure with obstructions in the streambed to create as much turbulence as possible before the flow passes over a weir where samples can be taken. The same effect may be found at a rock bar or perhaps where the stream flows through a restricted opening such as a bridge or culvert. Large bedload material will settle to the bottom again so the sample has to be taken quickly by scooping a bucket full of the mixture.
Larger streams or rivers may be sampled at a turbulence flume using a slot sampler as illustrated in Figure 48 (Barnes and Johnson 1956). An application of this method is described by Brown, Hansen and Champagne (1970). The sampling slot is narrow with sharp edges, and the water and sediment drop through into a hollow pipe or channel leading to a container. The slot must not be too small or it may be blocked by trash and also unable to catch larger particles, but a slot width of 5 mm set in a weir of 5 m width will take a 1000th sample of the flow. If this is still too large to handle comfortably, further subdivisions may be used, either another slot splitter or a sampling wheel as described in Chapter 3. Some of the difficulties associated with this method are:
· the splitter may affect the flow at entry into the slot;
· the slot may be blocked by floating trash;
· the concentration of sediment may not be equal across the width of the weir;
· coarse bedload may not be evenly mixed or not caught.
At a very much smaller scale, a device in Australia takes a sample of the total load in very small channels such as rills or furrows on cultivated land using a plastic jug as shown in Plate 35. A low suction is applied to the container to avoid entry problems.
Measuring the total amount of sediment deposited in ponds or reservoirs avoids the issue of the sediment delivery ratio, but unless the reservoir is large enough to contain the whole of the runoff, some of the sediment will be carried out over the spillway of the reservoir. The proportion of sediment caught is called trap efficiency and depends on:
· the size of the trap compared with the amount of inflow, that is the proportion of the flow which is caught in the reservoir, and
· the speed of flow through the reservoir which, combined with the particle size of suspended material, determines how much time there is for sediment to settle out in the reservoir. Trap efficiency may be estimated from Figure 49 (Brune 1953).
Accurate surveys of the reservoir or pond basin are required at intervals so that the accumulation can be calculated. In a reservoir which dries up completely this is a relatively simple surveying exercise. When the reservoir is partly flooded, the survey must be carried out from boats, and the level of the sediment may be determined either by a probe or an echo-sounder. It is also necessary to take samples and determine the density of the sediment in order to calculate the weight from the measurement by volume.
The ideal situation is to carry out a survey after the reservoir is completed and before it starts to fill, but existing reservoirs can be used by comparing successive surveys over a period of time. In all cases it is most important to establish a permanent base line so that the same transects are used in successive surveys. A good example of this technique is the work of Rapp in Tanzania (1977).
Computer programs are now available for calculating the storage volume from the relationship of surface area to depth of water (called the stage/area curve), and the total weight of sediment can be calculated from the sediment volume and density.