1. INTRODUCTION The subject of freesurface instability has interested many researchers and practitioners, beginning with Vedernikov (1945) who developed the criterion bearing his name. The Vedernikov number V = 1 describes the condition of neutral stability, which separates stable (V < 1) from unstable (V > 1) flow (Chow 1959). Although stable flow attenuates disturbances, unstable flow does not. Thus, unstable flow may promote the development of roll waves in steep channels (Ponce and Maisner, 1993). In addition, under highVedemikovnumber flows, flood waves may steepen to the point at which they become kinematic shock waves (Lighthill and Whitham, 1955; Ponce md Windingland 1985; Porras 1994). Craya (1952) enhanced the Vedernikov criterion by interpreting it as the ratio of relative kinematic wave celerity to relative dynamic wave celerity. Later, Liggett (1975) developed the differential equation of the stable channel, for which V = 0 for all Froude numbers (F ≤ ∞). This condition is too broad, because the maximum Froude number that can be achieved in practice is likely to be much less than ∞ (usually less than 20). Instead, a stable channel may be designed by choosing a neutralstability Froude number F_{ns} that is much less than ∞, such that V ≤ 1 for Froude numbers in the range F ≤ F_{ns}. This proposition is examined here. 2. BACKGROUND The relative kinematic wave celerity, or relative Seddon speed (Seddon 1990), is the following:
in which v = mean velocity; and β is the exponent of the normal flow rating:
where Q = discharge; and A = flow area. The smallamplitude relative dynamic wave celerity, or Lagrange celerity (Chow 1959), is the following:
in which g = gravitational acceleration; and D = hydraulic depth, defined as follows:
where T = top width. The Froude number is as follows (Chow 1959):
Using Eqs. 1, 3, and 5, the Vedernikov number is shown to be the following:
Equation 6 underscores the physical significance of β, or rather of β  1, since by definition, the dimensionless relative kinematic wave celerity is the following:
The Manning equation in SI inits is the following (Chow 1959):
in which v = mean velocity (v = Q/A); n = Manning roughness coefficient; R = hydraulic radius Assuming a crosssectional shape function P = k A^{δ}, Eqs. 2, 7 and 8 lead to the following:
in which c_{drk M} = dimensionless relative kinematic wave celerity assuming Manning friction.
The crosssectional shape is characterized by the parameter δ, which varies typically in the range Case 1: The hydraulically wide channel, for which δ = 0. In this case, the wetted perimeter is constant (P = k) and independent of the flow area (dP/dA = 0). This theoretical cross section is the asymptotic limit to all wide channels (Chow 1959). Case 2: The triangular channel, for which δ = 1/2. In this case, dP/dA = (1/2)/R. The actual shape of a triangular channel is a function of the side slope z (z horizontal : 1 vertical). Case 3: The inherently stable channel, for which δ = 1 (Ponce 1991). In this case, the hydraulic radius is a constant (R = R_{o} = 1/k) and the wetted perimeter is proportional to the flow area: P = k A. This theoretical cross section is the asymptotic limit to all stable channels, as defined here. From these cases, it is inferred that:
The conventional definition of the Vedernikov number (Vedernikov 1945; Chow 1959; Jolly and Yevjevich 1970) is as follows:
in which:
where b = exponent of the Reynolds number R in the frictional power law f = α R^{b}; and γ = a crosssectional shape factor defined as follows:
For Manning friction, b = 1/5, and therefore x = 2/3. Given Eqs. 7, 9, and 10, the equivalence of the two definitions of the Vedemikov number (Eqs. 6 and 11) is confirmed. Furthermore, from Eqs. 6 and 7, the dimensionless relative kinematic wave celerity is recast as follows:
For V = 1, F = F_{ns}, in which F_{ns} is the neutralstability Froude number. Therefore:
3. STABLE CHANNEL DESIGN Liggett (1975) derived the differential equation for the stable channel, defined as that for which δ is a constant and equal to 1. We shall extend the definition of stable channel to the cases where δ is a constant but is less than 1. Thus, two types of stable channels can be formulated. Type 1: An unconditionally stable channel, for which δ = 1 (Liggett 1975). This is the inherently stable channel, which features c_{drk} = 0 and F_{ns} = ∞, and is absolutely stable (V = 0) for all Froude numbers (F ≤ ∞) (Ponce 1991). The hydraulic radius is a constant (R_{o}). Type 2: A conditionally stable channel, for which δ < 1. This channel features c_{drk} > 0 and F_{ns} < ∞, and it is stable (V ≤ 1) for Froude numbers in the range F ≤ F_{ns}. The hydraulic radius varies with flow depth. Because of the usually symmetry of channel sections, a halfcrosssection analysis is appropriate. Hereafter, the asterisk (_{*}) is used as a subscript to refer to half values; e.g., T_{*} is the half top width. The differential wetted perimeter of the stable cross section is defined as follows:
in which h = flow depth. Dividing by dh, and since dA_{*} = T_{*} dh, then:
With Eq. 10, Eq. 17 becomes the following:
The unconditionally stable channel has δ = 1 and hydraulic radius R = R_{o}. Therefore:
subject to T_{*} > R_{o} . The design of a stable channel requires that the hydraulic radius (R_{o}) be specified at the start. To achieve this, the actual channel crosssection must be comprised of two subsections (Liggett 1975): (1) a lower subsection of rectangular, trapezoidal, or triangular shape, and (2) an upper subsection, of stable shape (see Fig. 1).
The lower subsection (of bottom width B, depth h_{o}, and side slope z) defines the hydraulic radius R_{o}, on which the upper subsection is based. The parameter R_{o} is defined at h = h_{o} as follows:
In addition to defining R_{o}, the lower subsection serves the purpose of conveying the low flows. In practice, highvelocity flows may occur at relatively small flow depths. This should be taken into account in the design of the lower subsection. The differential equation 19 represents a family of unconditionally stable channels, with parameter R_{o}. A particular solution for T_{*} = T_{*}_{o} at h = h_{o} is the following:
wich reduces to Liggett's solution (1975) for the special case of T_{*}_{o} = R_{o} as follows:
Since friction has a lower bound and cannot realistically decrease to zero, it follows that there is an upper limit to the Froude number that may be achieved in practice. Thus, a channel of constant δ may be designed to remain stable (V ≤ 1) within a specific Froudenumber range F ≤ F_{ns}, where the chosen F is not likely to be exceeded for a given design condition. For example, according to Chow (1959), a practical upper bound to the Froude number may be F = 20. In this case, the corresponding crosssection parameters are c_{drk} = 0.05 and δ = 0.925, from Eqs. 9 and 15. The extension of Eq. 18 to the conditionally stable channel, for which δ < 1 and R varies with flow depth, leads to the following:
subject to δ T_{*} > R.
Unlike Eq. 19, Eq. 23 cannot be integrated analytically. The shape of the upper subsection T_{*} = The numerical integration proceeds by first selecting the shape of the lower subsection and the total channel depth h_{f} to include lower and upper subsections. In the lower subsection, the flow depth varies in the range 0 ≤ h ≤ h_{o}; in the upper subsection it varies in the range h_{o} ≤ h ≤ h_{f}. In the case of a rectangular lower subsection (z = 0), the flow depth h_{o} and half top width T_{*}_{o} = B_{*} are selected. A suitable choice of F_{ns} enables the calculation of c_{drk M} and δ from Eqs. 9 and 15. The initial values are h = h_{o}; A_{*} = B_{*} h_{o}; P_{*} = B_{*} + h_{o}; T_{*} = T_{*}_{o}; and R = A_{*} / P_{*}. Equation 23 is solved from h = h_{o} to h = h_{f} at increments of Δh. For increased accuracy, a very small Δh is chosen, such as Δh = 0.0001 m. Values of ΔT_{*}, ΔP_{*}, ΔA_{*}, P_{*}, A_{*}, and R are calculated at every increment, in which:
and
The values of h and T_{*} are updated before each subsequent increment.
Figure 1 shows examples of stable channels calculated using the described algorithm. The bottom subsections are rectangular, or half width B_{*} = 2.5 m. Three cross sections are shown: Fig. 1 (a) The following conclusions can be drawn from Fig. 1: (1) For a given neutralstability Froude number F_{ns}, the larger the initial hydraulic radius R_{o}, the narrower the resulting stable cross section; and (2) For a given initial hydraulic radius R_{o}, the smaller the choice of F_{ns}, the narrower the resulting stable cross section. These findings have significant practical implications. Given a suitable choice of R_{o} and F_{ns}, a stable yet relatively narrow channel cross section may be designed following this methodology. Such a channel should be largely free from roll waves and kinematic shock waves, provided the Froude number remains in the range F ≤ F_{ns}.
4. SUMMARY The effect of crosssectional shape on freesurface instability was shown to be characterized by the dimensionless relative kinematic wave celerity c_{drk} = β  1, in which β is the exponent of the normal dischargeflow area rating. Three crosssections of constant c are: (1) hydraulically wide, with c = 2/3; (2) triangular, with c = 1/3; and (3) inherently stable, with c = 0.. The value of c is a function of the crosssectional parameter δ, the exponent of the wetted perimeterflow area relation (P = k A^{δ}).
A stable channel was defined as one featuring constant δ and c_{drk}. Two types of stable channels are formulated: (1) unconditionally stable, for which δ = 1, c_{drk} = 0, and V = 0 for all Froude numbers Results show that the larger the initial hydraulic radius R_{o} for a given F_{ns}, the narrower the resulting stable cross section. Likewise, the smaller the choice of F_{ns} for a given R_{o}, the narrower the resulting stable cross section. Thus, given a suitable choice of R_{o} and F_{ns}, a stable yet relatively narrow channel cross section may be designed. Such a channel should be largely free from roll waves and kinematic shock waves, provided the Froude number remains in the range F ≤ F_{ns}. Additional research is necessary to experimentally verify the findings of this study. APPENDIX. REFERENCES
Chow, V. T. (1959). Openchannel hydraulics. McGrawHill Book Co. Inc,. New York. N.Y. Craya, A. (1952). "The criterion of the possibility o roll wave formation." Gravity waves, Circular No. 521, National Bureau of Standards, Washington, D. C., 141151. Jolly, J. P., and Yevjevich, V. (1970), "Amplification criterion of gradually varied, single peaked waves." Hydro. Paper No. 51, Colorado State University, Fort Collins, Colo. Ligget, J. A. (1975). "Stability  Chapter 6," Unsteady flow in open channels, Vol 1. K. Mahmood and V. Yevjevich, eds., Water Resources Publications, Fort Collins, Colo., 259281. Lighthill, M. J. and Whitham, G. B. (1995). "On kinematic waves I: Flood movement in long rivers.: Proc. Roy. Soc. of London, Vol. A229, London, England, (May), 281316. Ponce, V. M. (1991). "New perspective on the Vedernikov number," Water Resour. Res., 27(7), 17771779. Ponce, V. M., and Maisner, M. P. (1993). "Verification of theory of roll wave formation." J. Hydr. Engrg., ASCE, 119(6), 768773. Ponce, V. M., and Windingland, D. (1985). "Kinematic shock: Sensitivity analysis," J. Hydr. Engrg., ASCE, 114(4), 600611. Porras, P. J. (1994), "Flood wave propagation in inherently stable channels: Theory and applications," MS thesis, Dept. of Civil Engrg., San Diego State University, CA. Seddon, J. (1900), "River hydraulics," ASCE Trans., Vol. 43, 179229. Vedernikov, V. V. (1945). "Conditions at the front of a translation wave disturbing a steady motion of a real fluid," U.S.S.R. Acad. Sciences Comptes Rendus (Doklady), Moskow, U.S.S.R., 48(4), 239424.

160715 12:00 
Documents in Portable Document Format (PDF) require Adobe Acrobat Reader 5.0 or higher to view; download Adobe Acrobat Reader. 