Documente Academic
Documente Profesional
Documente Cultură
Abstract
Existing resistance formulas produce a wide range of friction-factor estimates for gravel bed streams. The purpose of
this paper is to develop a reliable resistance formula in terms of the Darcy-Weisbach friction factor f. Published data
were screened and used to establish the formula. The existing formulas have considered that f is a function of relative
roughness D84/R only, where R is the hydraulic radius and D84 is the particle size referred to the intermediate diameter
that equals or exceeds that of 84 percent of bed sediments. In this paper, f is considered as a function of Froude number
in addition to the relative roughness. f for D84/R>1 displays a different trend than that for D84/R<1 perhaps due to the
invalid assumption of a logarithmic velocity distribution for D84/R>1. An f formula for D84/R<1 has been established.
Key Words: Flow resistance, Darcy-Weisbach friction factor, Relative roughness, Relative submergence, Gravel bed
channels, Open channels
1 Introduction
Accurate estimation of flow resistance is important for the hydraulic analysis of streams. Resistance coefficients can
be better estimated for lowland alluvial channels than upland gravel bed streams using existing formulas. For gravel bed
streams, existing formulas generally produce a wide range of estimates. It is the purpose of this paper to develop a
reliable resistance formula in terms of the Darcy-Weisbach friction factor for gravel bed streams. In doing so, published
data and formulas were first reviewed and screened. A new formula was then established based on the screened data and
compared with existing formulas. At least six different equations are in the literature which are based on the
Darcy-Weisbach formula. In order to set the stage for the new formula, the several present formulas are presented. Then
the screening criterion is described and used to explain some of the problems with previous analyses.
where u * is the shear velocity defined as u* W o/ U with W o and U being boundary shear stress and the density of
water, respectively. The Darcy-Weisbach friction factor f may be related to the commonly used Manning’s roughness
coefficient n as follows:
1 1.486 §¨ R1 / 6 ·¸
(3)
f n ¨© 8 g ¸
¹
where R is the hydraulic radius in ft and g = 32.2 ft sec-2.
The Darcy-Weisbach friction factor was experimentally investigated for pipe flows by Nikuradse (1933). The
1
President, Multech Engineering Consultants, Inc. 1650 Zanker Road, Suite 210, San Jose, CA 95112,
E-mail: wang@multecheng.com
2
Chief Hydrologist, Multech Engineering Consultants, Inc. 1650 Zanker Road, Suite 210, San Jose, CA 95112,
E-mail: dawdy@multecheng.com
Note: The original manuscript of this paper was received in Mar. 2012. The revised version was received in Jun. 2013. Discussion
open until Mar. 2015.
- 126 - International Journal of Sediment Research, Vol. 29, No. 1, 2014, pp. 126–132
experiments were conducted on smooth pipes and on rough pipes internally coated with sand grains of uniform size.
The results were presented in graphical form in the well-known Moody diagram. Prandtl (1926) and von Karman (1930)
established the following resistance formula for turbulent flows in fully rough pipes:
1 § k ·
2 log¨ s ¸ (4)
f © 3 .7 d ¹
The friction factor determined from the Moody diagram can be applied to turbulent open channel flows with uniform
bed sediments, provided that the channels are without bed forms and bank vegetation and the effect of the free surface is
negligible. The application can be made by substituting the pipe diameter d with the hydraulic radius R (namely, d = 4R)
and taking a characteristic grain diameter as ks in Eq. 4 as follows:
1 §k ·
2.34 2 log¨ s ¸ (5)
f ©R¹
International Journal of Sediment Research, Vol. 29, No. 1, 2014, pp. 126–132 - 127 -
where D50 is the particle size that equals or exceeds that of 50 percent of bed sediments.
F. Bathurst (1985) used data collected from 16 sites in upland rivers in Britain to establish a relationship for the
friction factor. The data collected were limited to flows within well-defined channels with uniform bed slopes and no
significant bank vegetation. Sediment transport involved only coarse materials (gravel, cobbles, and boulders) rather
than fine materials (sand). The sites were either riffles between pools or channel reaches without pools. The bed
material size distribution of each site was obtained from the surface layer according to the Wolman pebble count
procedure (Wolman, 1954). The relationship established is as follows:
8 § d ·
4 5.62 log¨¨ ¸
¸ (14)
f © D84 ¹
The above formula can be rearranged as follows:
1 §D ·
1.41 1.98 log¨ 84 ¸ (15)
f © d ¹
It is generally recognized that large bed particles have a more significant effect than small particles on flow resistance.
Consequently, D84 is more suitable than D50 to serve as the representative height of roughness elements. On this basis,
the formulas using D84 as the representative height of roughness elements are compared in terms of 1 / f in Fig. 1.
The comparison includes formulas proposed by Leopold and Wolman (1957), Limerinos (1970), Hey (1979), and
Bathurst (1985). In addition, the extension of the Prandtl formula for turbulent flows in fully rough pipes, Eq. 5 with ks
= D84, was used to serve as a baseline for screening. It can be seen from Fig. 1 that the formulas are practically parallel
with significant differences in intercept. This is due to the narrow range of 1.98 - 2.03 for the coefficient of the
logarithmic term (as the slope of the equation) and a rather wide range of 1 - 1.41 for the constant (as the intercept of
the equation). It also can be seen from Fig. 1 that the values of 1 / f predicted by Prandtl’s formula are the highest,
followed by those by Bathurst, Limoerinos, Hey, and Leopold. The large discrepancies indicate a need to improve the
resistance relationship.
5 Data screening
Discrepancies among existing resistance formulas are partly due to varying quality of the data on which they were
based. It is well recognized that data measurement of steep streams involving coarse bed materials (gravel, cobbles, and
boulders) are prone to errors due to sensitivity of the flow field to disturbance caused by operation of data measurement
and due to difficulties in conducting accurate bed material size analyses The resulting errors may include cross sectional
geometry, flow depth, velocity, discharge, water surface slope, energy slope, bed material sizes, etc. For instance,
Bathurst (1985) indicated errors of up to r5% for depth, r8% for water surface slope, r15% for bed material size,
and r1% for distance in the data from the upland British rivers. This resulted in errors of up to r12% for f
and r16% for d/D84. Also, for steeper streams, step and pool conditions may exist, in which energy is dissipated in the
falls into the pools. Therefore, data should be screened for accuracy before they are used to establish a resistance
formula.
In the present study, published data were screened before being used to establish the flow resistance relationship. The
data included those reported by Limerinos (1970), Hey (1979), Bray (1979), Jarrett (1984), Bathurst (1985), and Thorne
and Zevenbergen (1985). According to Hey (1979), the representative height of roughness elements is 3.5 times D84.
This finding may be used in conjunction with the Prandtl resistance formula, Eq. 5, to serve as the basis of screening to
discard data having unrealistically small values for friction factor. To be conservative, however, ks was taken as D84
instead of 3.5 times D84 for a rough boundary turbulent flow, which is consistent with 3 times D84 suggested by Charlton
et al. (1978). The value calculated from Eq. 5 is considered the highest possible value of 1 / f (namely, the lowest
possible value of f) for a rough boundary turbulent flow with the given representative relative roughness k s / R . This is
because energy loss associated with eddies is minimal when the roughness elements are of uniform size as in the
experiments by Nikuradse (1933). In the present study, any observed data with values of 1 / f greater than 95% of
those calculated from Eq. 5 were considered inaccurate and, therefore, discarded in the present analysis. The screening
results are shown in Figs. 2, 3, 4, 5, 6, and 7 for the data reported by Limerinos (1970), Hey (1979), Bray (1979), Jarrett
(1984), Bathurst (1985), and Thorne and Zevenbergen (1985), respectively.
Fig. 4 Screening results of observed data by Bray (1979) Fig. 5 Screening results of observed data by Jarrett (1984)
Fig. 6 Screening results of observed data Fig. 7 Screening results of observed data
by Bathurst (1985) by Thorne and Zevenbergen (1985)
(defined as the observed value of 1 / f minus the predicted value) and D84/R shows that E varies with D84/R. As
shown in Fig. 9, Eq. 22 tends to under-predict 1 / f when D84/R>1. To improve prediction accuracy, only the data
with D84/R< 1 were used for further regression analysis. The resulting formula is as follows:
1 §D ·
1.80 1.89 log( F ) 2.03 log¨ 84 ¸ for D84/R< 1 (23)
f © R ¹
Equation 23 shows that 1 / f increases with increasing Froude number. This implies that f decreases with increasing
Froude number. This tendency seems consistent with the experimental results for supercritical flows by Habibzadeh and
Omid (2009). One possible explanation for this tendency is that, for mountain gravel bed streams, while Froude number
is less than one based on average values of velocity and hydraulic depth over the entire cross section, local Froude
numbers near coarse bed materials such as cobbles and boulders may be greater than one, resulting in local in-phase
relations between bed features and surface waves and thereby reducing the portion of the head loss caused by formation
of eddies. A comparison of predicted and observed values of 1 / f for D84/R< 1 are shown in Fig. 10. The standard
error of estimate and correlation coefficient for Eq. 23 are 0.47 and 0.89, respectively. It is noted that, without including
Froude number, the regressed constant C’ in Eq. 20 for the data with D84/R< 1 is 1.18 and the corresponding standard
- 130 - International Journal of Sediment Research, Vol. 29, No. 1, 2014, pp. 126–132
error of estimate and correlation coefficient are 0.53 and 0.85, respectively. The limitation of D84/R< 1 is perhaps
consistent with the requirement that the velocity profile must follow that of Eq. 17. It can be reasoned that shallow
flows with protruding coarse bed materials (such as gravel, cobbles, and boulders) with D84/R> 1 are not capable of
developing a well-defined logarithmic velocity profile due to intense flow impingement on coarse bed particles and the
adoption of 2.03 as the coefficient for the logarithmic term in the resistance formula is no longer valid. Under such
conditions, the general functional forms of Eq. 20 and Eq. 21 are not suitable for use to develop a resistance formula. It
is noted that the reciprocal of D84/R is termed relative submergence by Bathurst (1985) and flows with D84/R >1 were
classified to be in the region of large-scale roughness.
Fig. 8 Predicted (using Eq. 22) versus observed values for 1 / f Fig. 9 Predicted residual E from Eq. 22 versus D84/R
Fig. 10 Comparison of predicted (using Eq. 23) versus observed values of 1 / f for D84/R<1
Table 1 Standard errors of estimate and correlation coefficients of 1 / f for different formulas for D84/R<1
Formula Standard Error of Estimate Correlation Coefficient
Present Study 0.47 0.89
Leopold and Wolman (1957) 0.53 0.85
Limerinos (1970) 0.53 0.85
Hey (1979) 0.53 0.85
Bary (1979) 0.66 0.82
Griffths (1981) 0.58 0.83
Bathurst (1985) 0.53 0.85
International Journal of Sediment Research, Vol. 29, No. 1, 2014, pp. 126–132 - 131 -
Fig. 11 Comparison of observed and predicted values of 1 / f for D84/R<1 (using Eq. 23) versus Leopold and Wolman’s formula
References
Bathurst J. C. 1985, Flow resistance estimation in mountain rivers. Journal of Hydraulic Engineering, ASCE, Vol. 111, No. 4.
Bray D. I. 1979, Estimating average velocity in gravel-bed rivers. Journal of Hydraulic Engineering, ASCE, Vol. 105, No. 9.
Charlton F. G., Brown P. M., and Benson R. W. 1978, The hydraulic geometry of some gravel rivers in Britain. HR Wallingford
Report. IT 180.
Griffiths G. A. 1981, Flow resistance in coarse gravel bed rivers. Journal of Hydraulic Engineering, ASCE, Vol. 107, No. 7.
Habibzadeh A. and Omid M. H. 2009, Bedload resistance in supercritical flow. International Journal of Sediment Research, Vol. 24,
No. 4.
Hey R. D. 1979, Flow resistance in gravel bed rivers. Journal of Hydraulic Engineering, ASCE, Vol. 105, No. 4.
Jarret R. D. 1984, Hydraulics of high gradient streams. Journal of Hydraulic Engineering, ASCE, Vol. 110, No. 11.
Keulegan G. H. 1938, Laws of turbulent flow in open channels. U.S. National Bureau of Standards, Journal of Research, Vol. 21.
Leopold L. B. and Wolman M. G. 1957, River channel patterns; braided, meandering, and straight. U. S. Geol. Survey Professional
Paper 282-B.
Limerinos J. T. 1970, Determination of the Manning coefficient from measured bed roughness in natural channels. Water Supply
Paper 1898-B, U.S. Geological Survey, Washington.
Manning R. 1991, On the flow of water in open channels and pipes. Institution of Civil Engineers of Ireland, Transactions, Vol. 20.
Nikuradse J. 1933, Stromungsgesetze in rauhen Rohrne (Laws of flow in rough pipes).Ver. deutscher Ingenieure, Forschungsheft, No.
361, Berlin.
Prandtl L. 1926, Uber die ausgebildeterturbulenz (On fully developed turbulence). Proceedings, the Second International Congress of
Applied Mechanics, Zurich.
Soto, A. U. and Madrid-Aris M. 1994, Roughness coefficient in mountain rivers. Hydraulic Engineering 1994 – Volume 1, Edited by
G. Cotroneo and R. Rumer, ASCE.
Thorne C. and Zevenbergen L. 1985, Estimating mean velocity in mountain rivers. Journal of Hydraulic Engineering, ASCE, Vol.
111, No. 4.
von Karman T. 1930, Mechanische Aehnlichkeit und turbulenz (Mechanical similarity and turbulence). The Third International
Congress of Applied Mechanics, Stockholm, Proceedings v. 1.
Wolman M. G. 1954, A method of sampling coarse river-bed material. American Geophysical Union, Transaction, Vol. 35, No. 6.
- 132 - International Journal of Sediment Research, Vol. 29, No. 1, 2014, pp. 126–132