You are on page 1of 10

WATER RESOURCES BULLETIN

VOL. 11, NO. 5 AMERICAN WATER RESOURCES ASSOCIATION OCTOBER 1975

SEDIMENT ROUTING FOR AGRICULTURAL WATERSHEDS'

J. R. Williams'

ABSTRACT: A sediment routing technique was developed to route sediment yield from small
watersheds through streams and valleys to the outlet of large watersheds. The technique is
based o n the modifEd universal sol loss equation and a first order decay function of travel time
and particle size. Deposition is dependent upon settling velocities of sediment particles, travel
time, and the amount of sediment in suspension. Sediment routing increases sediment yield
prediction accuracy and allows determination of subwatershed contributions to the total
sediment yield. Also, the locations and amounts of floodplain scour and deposition can be
predicted. Another advantage of sediment routing is that measured sediment yield data are not
required. The procedure performed satisfactorily in test routings on two Texas blackland
watersheds Sediment routing will be useful in flood control evaluation, reservoir and channel
design, water quality calculations, environmental impact assessment, and land-use planning.
(KEY TERMS: sediment; routing; universal equation; deposition; floodplain scour.)

INTRODUCTION

Sediment from agricultural watersheds depletes reservoir storage, pollutes streams and
reservoirs, and transports other pollutants. Agricultural sediment sources are sheet, rill,
and gully erosion of the land. Individual sediment-source contributions t o the total
sediment yield depend upon climatic factors, soil characteristics, land slope, crop
management, erosion control practices, and watershed hydraulics. Since these factors vary
considerably and are not usually uniformly distributed on large agricultural watersheds
(area < 1,000 square miles), estimates of sediment yield may be inaccurate if these
factors are averaged and sediment yield is predicted for the entire watershed.
Sediment yield can be more accurately estimated if a large watershed is divided into
subwatersheds (area < 10 square miles) t o compensate for nonuniformly distributed
sediment sources. Also, the effect of watershed hydraulics and sediment particle size can
be included by routing the sediment yield from the subwatersheds to the large watershed

'Paper No. 75065 of the Water Resources Bulletin. Discussions are open until June 1, 1976.
Contribution from the Southern Region, Agricultural Research Service, USDA, and in cooperation
with the Texas Agricultural Experiment Station, Texas A&M University.
'Hydraulic Engineer, Southern Region, Agricultural Research Service, USDA, Temple, Texas
76501.

965
Sediment Routing 966

outlet. Sediment routing not only increases prediction accuracy, but also determines
individual subwatershed contributions to the total sediment yield. Another advantage to
routing is that location and amount of floodplain deposition and scour can be
determined.
The universal soil loss equation (Wischmeier and Smith, 1965) was designed to predict
soil loss from field-size areas. Williams and Berndt (1972) predicted sediment yield from
small watersheds with the universal equation and a delivery ratio. T o increase tce
prediction accuracy and eliminate the need for a delivery ratio, the univeral equation was
modified by replacing the rainfall energy factor with a runoff factor (Williams, 1975,
ARS-S-40). The modified universal equation developed with data from 1 8 small
watersheds at Riesel, Texas, and Hasting, Nebraska, explained about 92 percent of the
variation in sediment yield for 778 individual storms on these watersheds. In tests with
other data, the equation explained about 90 percent of the variation in sediment yield for
529 individual storms on 25 watersheds located at Treynor, Iowa; Oxford, Mississippi;
Chickasha, Oklahoma; and Riesel, Texas.
The modified universal equation can be applied to large watersheds if sediment sources
are uniformly distributed over the watershed and if major watershed tributaries are
hydraulically similar. However, neither of these conditions are found on most large
agricultural watersheds. Therefore, a sediment routing procedure is needed. Unfortu-
nately, .procedures for routing sediment on agricultural watersheds have not been
developed and tested. Boyce (1975) presented a conceptual model for sediment routing
based on delivery ratios. Kling and Olson (1974) described a conceptual model for
estimating sediment transport within a watershed, based on the sequential relationship of
subarea slopes. Sediment transport procedures, like the modified Einstein methid
(Einstein, 1950; Colby and Hembree, 1955), can be used t o compute sediment-flow
relationships at a particular channel section, but no routing mechanism is included, i.e.,
sediment discharge downstream is not related to that upstream. These procedures were
designed for rivers with extremely large drainage areas so that the sediment load is not
greatly affected by agricultural management. Thus, no provision is included for varying
sediment input from subareas or for determining subarea contributions to total sediment
yield.
A sediment routing procedure, based on the modified universal soil loss equation and a
first order decay function of travel time and particle size, was developed and is reported
here. The procedure determines the spatial sediment distribution on a watershed for an
individual storm, but does not consider sediment graphs (temporal sediment distribution).

DEVELOPMENT OF PROCEDURE

The modifEd universal soil loss equation is:

Y = 95 (Q X qp)0.56 K LS C P

where:
Y = sediment yield for an individual storm in tons,
Q =runoff volume in acre-feet,
= peak flow rate in cfs,
qp
9 67 Williams

K = soils factor,
Ls = slope length and steepness factor,
C = crop management factor, and
P = erosion-control practice factor.
Equation (1) can be used to predict sediment yield from large watersheds, if values of K ,
LS,C, and P are uniformly distributed over the watershed and if the major tributaries of
the watershed are hydraulically similar. Since these conditions are not normal for most
large agricultural watersheds, a routing function was developed to predict sediment yield
for any distribution of K, LS, C, and P over a watershed and for any variation in tributary
hydraulics. The procedure is based on the assumption that sediment deposition is
dependent upon settling velocities of the sediment particles, length of travel time, and the
amount of sediment in suspension. For turbulent flow the settling velocity is proportional
to the square root of the particle diameter (Einstein, 1964). These assumptions can be
expressed by the equation:

dY =
- -B Y fi
dt

where:
Y = sediment yield at a particular channel section,
t =time,
B = decay constant, also called routing coefficient, and
D = particle diameter.
Integrating Equation (2) and solving for Y yields the equation:

Y = Yoe- BT G
(3)

where:
Yo = sediment yield at an upstream section, and
T =travel time between the two sections.
To determine the total sediment delivered to the watershed outlet, the subwatershed
contributions are summed in the equation:

RY = Z Yie- B T ~ ~ E C C
(4)
i=1

where:
RY = sediment yield for the entire watershed in tons,

Yi = sediment yield for subwatershed (i) in tons,


B = routing coefficient,
Ti = travel time from subwatershed (i) t o the watershed outlet in hours,
Sediment Routing 968

D50i = median particle diameter of sediment in mm for subwatersheds (i), and


n = number of subwatersheds.
The median particle diameter, D50i, was chosen arbitrarily t o represent the particle
diameter, D. Although other values, like D25 or D75 might be more suitable, D50 was
selected for preliminary tests.
Yi can be predicted fairly accurately with Equation (1) because the subwatersheds are
delineated so that K, LS, C, and P are as uniformly distributed as possible over each
subwatershed. Ti can be computed by flood routing, and D50i can be estimated from
soils information. Thus, only the total sediment yield, RY,and the routing coefficient, B,
are unknown, RY can be predicted fairly accurately with Equation (1) if K, LS, C, P, and
D50i are uniformly distributed over the entire watershed. T o determine B for an
individual storm on a particular watershed uniform distributions of K, LS, C, P, and D50i
are assumed. Thus, Y computed in Equation (1) is equal to RY computed with Equation
(4). Setting the right-hand sides of Equations (1) and (4) equal yields the equation:

If K, LS, C, and P are equal for all subareas, they cancel in Equation (9,thus producing
the equation for determining B.

Where D50, is the minimum value of D50 (about 0.001 mm). A constant value of D50,
is used in determining the routing coefficient, B, for all watersheds so that B is a function
of watershed hydraulics only. Thus, two watersheds that are similar hydraulically will
have similar B values. D50, was selected as a minimum value of D50 t o assure a realistic
sediment delivery for the entire scale of D50 values. Equation (6) is an expression of
runoff rates and volumes, travel times, and two constants (B and D 5 h ) . Thus, B can be
determined by an iterative solution of Equation (6).
Once B has been determined, sediment yield for the watershed can be predicted with
Equation (4) using the actual values of K, LS, C, P, and D50 for each subwatershed. When
the actual values of D50 are used in Equation (4), two watersheds with similar hydraulic
characteristics (similar routing coefficients) will have extremely different delivery ratios,
if their D50 values are greatly different. If the actual values of D50 were used to
determine B, all hydraulically similar watersheds would not have similar routing
coefficients but would have similar delivery ratios, even if the actual values of D50
differed greatly. This illustrates the importance of using a constant D50, in Equation
(6). D50m’s function as a scaling factor will be discussed further after example routings
are presented.
If sheet erosion is the only sediment source, the job is complete. However, if
floodplain and channel scour and gully erosion are involved, their contributions must be
included t o predict sediment yield accurately. Gully erosion is usually estimated on an
annual basis (Soil Conservation Service, 1966). Gully erosion for an individual storm can
be estimated by multiplying the annual amount by the ratio of the storm runoff to
average annual runoff. The gully erosion can be added to the sheet erosion for. the
969 Williams

subwatershed and routed downstream, if the D50 of the two sources is similar. If not,
they should be routed separately.
If floodplain scour is a problem (which it usually is’on large agricultural watersheds),
the amount of scour must be predicted for each routing reach and routed downstream.
Equation (1) can be used to estimate floodplain scour when written in the form:

YFP = 23.6 (A X R X qp)0.56 K LS C P (7)

where:
YFP = sediment yield from floodplain scour of a routing reach in tons,
A = area flooded in acres, and
R = runoff from watershed above the reach in inches.
Extremely high peak flow rates provide a tremendous potential for scour. Also, the P
factor is usually 1.0 because there are no erosion control practices in the floodplain.
However, the LS factor is quite low because slopes are flat. The crop factor, C, is the
dominant factor in floodplain scour. Scour is very minor on a well-covered floodplain but
can be severe on poorly covered floodplains.
Equation (7) can also be used to estimate channel scour. The area flooded is computed
by multiplying channel length by width. The peak flow rate that is contained in the
channel for a particular storm is q . The LS factor is usually high for channels because
the slope is the resultant of the si& slopes and the channel slope. The crop factor, C, is
usually high because most natural channels are not covered with vegetation. As in the
floodplain, the P factor is 1 .O because there are no erosion control practices. The most
critical factor in channel scour is the soils factor, K. If the channel is stable, K must be
very low because all the other factors are high. Research is needed to determine K factors
for channels.
Equation (7) is proposed as a first approximation of a method for estimating channel
and floodplain SCOUT. Research is needed to develop and test procedures for accurately
computing these components. Users of the sediment routing procedure may prefer to use
other methods for estimating channel and floodplain scour, since Equation (7)has not
been tested adequately.
This completes the development of the procedure. Examples are presented to
demonstrate sediment routing,
T A B E 1. Determination of Routing Coefficient for Flood of March 29, 1965.

Subwatershed Area Runoff Volume Peak Flow Rate Travel Time


Number Square Miles Aa-FWt CfS Hours

1.734 456 2,360 2.03


.837 184 1,174 1.46
1.108 244 1,425 0.86
.807 177 1,151 0.34
1.875 3 84 1,882 0.00
0.483 87 638 0.25
6.840 1,53 2 4,368 0.00
B = 6.1
Sediment Routing 970

EXAMPLE ROUTINGS

The first example routing was performed on Watershed G near Riesel, Texas. The
watershed area is 6.84 square miles, slopes are 2 to 3 percent, and soils are mostly
blackland with high clay content. Average annual rainfall is about 34 inches and average
annual runoff is about 6 inches. Gully erosion is negligible and floodplain scour is minor.
The watershed was divided into six subwatersheds and five reaches for flood routing.
HYMO (Williams and Hann, 1972) was used to compute subwatershed hydrographs and
to route the floods. Sediment was routed for five floods and compared with measured
amounts.
Equation (5) ulas used to determine the routing coefficient, B, for each flood. Table 1
shows the inputs for Equation (5) for the storm of March 29, 1965, and the B determined
(6.1).
Once B was determined, Equations( 1) and (4) were used to predict the sediment yield
for Watershed G. Table 2 shows routed and measured sediment yields for the five storms
and the percent of gross erosion delivered (delivery ratio) to the watershed outlet.

TABLE 2. Comparison of Routed and Measured Sediment Yield


for Five Storms on Watershed C.

Sediment Yield eons) Delivery Routing


Date Measured Routed Ratio Coefficient

3-29-65 4,088 4,44 8 46 6.1


2-9-66 1,648 1,533 42 8.7
5-10-68 759 848 44 4.9
8-1 2-66 1,332 1,067 46 4.4
5-1065 1,890 1,470 45 6.4

T o compare routed and measured average annual sediment yield, five floods of
different frequencies were routed through Watershed G to obtain a sediment yield
frequency cuIve. The average annual sediment yield is computed by integrating the
frequency curve (Williams, 1974, IAHS Publication No. 113). Table 3 shows the routed
sediment yields for the f n e floods.

TABLE 3. Routed Sediment Yield-Frequency Relationship for Watershed C.

Return Period Routed Sediment Yield Routing


Wears) WnS) OonslAm) Coefficient

100 4,773 1.09 6.3


25 3,420 0.78 5.8
5 1,975 0.45 5.3
1 76 2 0.17 5.1
0.25 24 7 0.06 4.9
97 1 Williams

TABLE 4. Land Management Systems


and Routed Average Annual Sediment Yields for Elm Creek.

Erosion Control Total Sediment Floodplain Yield


Land Use Ractice Yield (Tons/Acre) (Tons/Acre)

Cultivation None 6.01 10.83


Cultivation except on floodplain None 4.81 0.20
Cultivation except on floodplain Terraces and waterways 0.74 0.20
Permanent pasture None 0.13 0.20

The average annual sediment yield computed by integrating the frequency curve is
0.77 tons/acre. During the 14-year period of record (1960 to 1973) the average annual
sediment yield measured was 0.71 tons/acre.
Elm Creek, near Temple, Texas, was selected to demonstrate the sediment routing
procedure on a large blackland watershed. The watershed area is 167 square miles, slopes
are 2 to 4 percent and soils are mostly blackland with high clay content. Average annual
rainfall is about 34 inches and average annual runoff is about 5.5 inches. Gully erosion
was assumed negligible. The watershed was divided into 90 subwatersheds and 10 reaches
for routing. Flood routing was done by the Soil Conservation Service using their TR-20
program (Soil Conservation Service, 1%9). Sediment has not been measured on Elm
Creek, so routed sediment yield cannot be compared with measured amounts. However,
the watershed can be used to demonstrate the applicability of the sediment routing
procedure to large watersheds and to show that results are reasonable.
Four different frequency floods were routed through the Elm Creek Watershed and
sediment frequency curves were obtained for various land uses and erosion control
practices. Table 4 shows the management systems and the average annual routed sediment
yield for each system.

TABLE 5. Effect of Sediment Particle Size


on Sediment Yield for Elm Creek Watershed.

D50 Particle Average A M U Sediment


~ Delivery
Size (mm) Yield (Tons/Acre) Ratio

0.001 4.87 37.0


.01 2.73 21.o
.1 0.77 6.1
.4 0.33 2.5

To demonstrate the usefulness of sediment routing in evaluating the effect of channel


improvement on sediment yield, channel improvement was assumed for all routing
reaches on Elm Creek. Flood routing was not performed for the improved channel, but
travel times were shortened by 40 percent to account for a large straight channel. The
only purpose of this routing is to show an application of sediment routing, thus, accurate
flood routing is not necessary. With the watershed cultivated, except on the floodplain,
Sediment Routing 972

the improved channel increased the average annual sediment yield from 4.87 to 5.48
tons/acre. The channel was assumed stable, thus the increased sediment yield was caused
by decreasing the travel time.
The effect of sediment particle size on sediment yield can be shown by assuming
various particle sizes and routing. Table 5 shows the results of routing through the Elm
Creek Watershed, assuming the watershed is covered by soil of the particle size shown.
The final example is a routing through Elm Creek with a variety of assumed particle
sizes. This example is presented to demonstrate the applicability of sediment routing to
watersheds that contain major tributaries with extreme differences in particle size. Table
6 shows the results for selected subwatersheds.

TABLE 6. Effect of Extreme Differences in Particle Size


on Major Tributaries of Elm Creek Watershed.

Subwatershed DSO Particle Travel Time Sediment Yield Delivered


Number Size (mm) to Outlet (Hours) to Outlet (Percent)

26 0.005 10.14 45.9


64 0.005 0.76 94.3
1 0.020 14.94 10.1
25 0.020 7.41 32.1
65 0.400 7.48 0.6
88 0.400 0.76 59.3
19 0.250 11.11 0.2

DISCUSSION

Although more testing is needed to determine the accuracy of the sediment routing
procedure, it has performed satisfactorily on two blackland watersheds. Reasonable
results were also obtained when other soils were considered (different particle size). The
average delivery ratio computed by dividing routed sediment yield by gross erosion was
about 0.45 for Watershed G. This compares closely with the value of 0.41 computed from
Maner’s Delivery Ratio Equation (Maner and Barnes, 1953). Assuming Elm Creek has a
D50 = 0.001 mm, the delivery ratio obtained from routing was 0.37. This is considerably
higher than the value of 0.26 computed with Maner’s Equation. However, D50 was
probably greater than 0.001 mm for the watersheds used to develop Maner’s Equation. If
D50 for Elm Creek is assumed to be 0.005 mm, the routed delivery ratio is 0.26. N o D50
data were available to check the delivery ratios shown in Table 5 , but the lower bound
can be examined. Delivery ratios as low as 4 and 6 percent were reported for comparable
size watersheds in the Piedmont (Sedimentation Committee, 1970) and Red Hills (Maner,
1958) physiographic areas, respectively. Thus, the routed delivery ratios are scaled in the
proper range. This is caused by selecting D50, as the minimum value. If D50, was
chosen as 0.4, the delivery ratio scale would range from 0.37 to 0.49.
At this point the reader may have two questions: (1) Why compare results of sediment
routing to delivery ratios? and (2) If delivery ratios are reliable why is sediment routing
needed? The answer to the first question is that delivery ratios are reliable for estimating
average annual sediment yield in areas when data are sufficient to develop relationships
973 Williams

like Maner’s Equation. Thus, the comparison is made to show that sediment routing gives
comparable results. The answer to the second question is that relationships like Maner’s
Equation have been developed for only a few regions because of limited data. Sediment
routing should work in any region without collecting sediment yield data. Even more
important, if delivery ratio equations were available universally, they d o not enable the
determination of subwatershed contributions or the effect of hydraulic changes, such as
channel improvement.
The average annual routed sediment yields shown in Table 4 compared well with
measured data from other watersheds. A plotting of Texas blackland watershed data
reported by Dendy and Champion (1973) indicated an average annual sediment yield of
2.4 tons/acre is appropriate for Elm Creek. This value lies between the routed yield for
cultivation, except in the floodplain, with no conservation practices and the routed yield
with a good conservation program. This condition is quite common for Texas blackland
watersheds (a large portion of cultivated land with some good conservation farming and
some straight-row farming). During a 9-year period at Riesel, Texas, a nonconservation
cultivated watershed yielded about 10 tonslacrelyear. However, this was a small
watershed (176 acres) and the period produced more than normal runoff (about 9
inches/year as compared with the normal of 6 inches/year). Thus, the upper value of 6
tons/acre/year seems reasonable for Elm Creek. Also, at Riesel a watershed with good
conservation practices produced about 0.5 tons/acre/year during a normal runoff period
of 13 years. However, the watershed was only about 60 percent cultivated. If it had been
cultivated, except in the floodplain, as assumed in the example, the routed value of 0.74
tons/acre/year would agree closely.
Table 6 shows how sediment routing can be used to determine individual subwatershed
contributions. Soil type and location on the watershed are quite important in determining
the contributions of subwatersheds. By examining output similar to the information
shown in Table 6, watershed land-use decisions can be made. For example, Subwatersheds
1, 19, and 65 can be cultivated without contributing significantly to the watershed
sediment yield.

SUMMARY AND CONCLUSIONS

A sediment routing technique was developed to route sediment yield from small
agricultural watersheds (-<1 0 square miles) through streams and valleys to the outlet of
large watersheds (< 1,000 square miles). The procedure is based on the modified universal
soil loss equation and a first order decay function of travel time and particle size.
Deposition is dependent upon settling velocities of the sediment particles, travel time, and
the amount of sediment in suspension.
Sediment routing increases sediment yield prediction accuracy and allows determina-
tion of subwatershed contributions to the total sediment yield. Also the locations and
amounts of floodplain scour and deposition can be predicted. Another advantage of
sediment routing is that measured sediment yield data are not required.
The procedure performed satisfactorily in test routings on two Texas blackland
watersheds. More tests are planned with data from other physiographic areas.
Sediment routing will be useful in evaluation of flood control projects, reservoir and
channel design, water quality calculations, environmental impact assessment, and land-use
planning.
Sediment Routing 974

LITERATURE CITED

Boyce, R. C., 1975. Sediment Routing with Sediment-Delivery Ratios. In: Present and Prospective
Technology for Predicting Sediment Yield and Sources, Proceedings of the Sediment-Yield
Workshop, USDA Sedimentation Lab., Oxford, Mississippi, November 28-30, 1972, USDA,
ARSS-40.
Colby, B. R. and C H. Hembree, 1955. Computations of Total Sediment Discharge, Niobrara River
near Cody, Nebraska. U.S. Geol. Survey Water Supply Paper 1357.
Dendy, F.E. and W.A. Champion, 1973. Summary of Reservoir Deposition Surveys Made in the
United States Through 1970. USDA Miscellaneous Publication No. 1266, p. 82.
Einstein, H. A,, 1950. The Bedload Function for Sediment Transportation in Open (hannel Flows.
USDA Tech. Bull. 1026, p. 70.
Einstein, H.A., 1964. Handbook of Applied Hydrology, Section 17-11, Sedimentation, Part 11, River
Sedimentation Ven Te Chow (Editor-irk(hief'). McGraw-Hill Book Company, pp. 1743567).
Kling, G. F. and G. W. Olson, 1974. The Sediment Transport Computer Model. Cornell Agronomy
Mimeo 74-11, Cornell University, Ithaca, New York, p. 26.
Maner, S. B., 1958. Factors Affecting Sediment Delivery Rates in the Red Hills Physiographic Area.
Transactions, Amer. Geophysical Union, 39, pp. 669475.
Maner, S. B. and L H. Barnes, 1953. Suggested Criteria for Estimating Gross Sheet Erosion and
Sediment Delivery Rates for the Blackland Prairies Problem Area in Soil Conservation USDA-SCS
Publication.
Soil Conservation Service, 1966. Procedures for Determining Rates of Land Damage, Land
Depreciation, and Volume of Sediment Produced by Gully Erosion USDA Tech. Release No. 32,
Geology.
Soil Conservation Service, 1969. Computer Program for Project Formulation Hydrology. Tech.
Release No. 20.
Task Committee on Preparation of Manual of Sedimentation, Vito A. Vanoni, (hairman, 1970.
Sedimentation Engineering, Chapter IV: Sediment Sources and Sediment Yields. Journal of the
Hydraulics Div., ASCE 96. No. HY6, Roc. Paper 7337, pp. 1283-1329.
Williams, J. R., 1974. Predicting Sediment Yield Frequency for Rural Basins to Determine Man's
Effect on bng-Term Sedimentation. In: Effects of Man on the Interface of the Hydrobgical Cycle
with the Physical Environment, Symposium Proceedings, IAHS Publication No. 113, pp. 105-108.
Williams, J. R., 1975. Sediment Yield Prediction with Universal Equation Using Runoff Energy
Factor. In: Present and Prospective Technology for Predicting Sediment Yield and Sources,
Proceedings of the Sediment-Yield Workshop, USDA Sedimentation Lab., Oxford, Mississippi,
November 2&30, 1972, USDA, ARSS-40.
Williams, J. R. and H. D. Berndt, 1972. Sediment Yield Computed with Universal Equation. ASCE
Hydraulics Div. Journal, 98, No. HYl2, pp. 2087-2098.
Williams, J. R. and R. W. Hann, 1972. HYMO, a Roblem-Oriented Computer Language for Building
Hydrologic Models. Water Resources Research, 8, pp. 79-86.
Wischmeier, W. H. and D. D. Smith, 1965. Predicting Rainfall-Erosion Losses from Cropland East of
the Rocky Mountains. USDA Agricultural Handbook 282, p. 47.

You might also like