---
title: Channel Flow and Backwater
description: "# A drain discharging to a fixed river level Our [MiTS fixed-outfall example](https://mes100.com/docs/fix-water-depth-as-the-end-point-of-drain/) follows a storm through a single drain that discharges to a river held at a fixed level. The drain is 100 m long and 1.2 m wide,…"
url: "https://mes100.com/docs/channel-flow-and-backwater/"
updated: "2026-11-17"
category: Exposition / Drain
---

# Channel Flow and Backwater

# A drain discharging to a fixed river level

Our [MiTS fixed-outfall example](https://mes100.com/docs/fix-water-depth-as-the-end-point-of-drain/) follows a storm through a single drain that discharges to a river held at a fixed level. The drain is 100 m long and 1.2 m wide, its invert falls from RL 29.00 m to RL 28.00 m, and the river stands at RL 28.50 m. MiTS reports three phases. The flow rises to about 1.5 m³/s, holds there from 30 to 60 minutes, and then recedes. After 90 minutes the flow is almost zero, yet the drain does not empty: water stays in its lower half. A second model in the same article shows the outlet flow reversing, with river water flowing back in.

That article explains the last phase by hand: once the flow stops, the water surface lies level with the river. It leaves the other phases open. Does the river also set the drain level at the peak? If not, what does? And what governs the water while the flow is rising, falling or reversing? [Figure 1](#fig-channel-river-drain) shows the two steady states we can draw straight away.

Figure 1. A drain held by a fixed river level: the surface at rest and at peak flow

In [Figure 1](#fig-channel-river-drain), the dashed blue line is the surface after the storm, level with the river. It meets the bed at chainage 50 m, so the upstream half of the drain is dry.

The solid blue line is the surface at the 1.5 m³/s peak. It is not level: it runs about 0.415 m above the bed and falls with it. Only still water lies level. Flowing water needs a falling surface to push it along against the friction of the bed and walls, and on this drain the bed slope provides that fall. At the outlet this surface is still below the river level, so at the peak the river has no effect on the drain.

From [Figure 1](#fig-channel-river-drain) and the MiTS results we can therefore see three behaviours of the drain water surface. A drain that lies wholly below the river level at rest, or one of milder slope, can show two more:

1. Level with the river, when there is no flow: the dashed line in [Figure 1](#fig-channel-river-drain).
2. At its own normal depth, falling with the bed and unaffected by the river, for a large steady flow in a steep drain: the solid line in [Figure 1](#fig-channel-river-drain).
3. Changing with time, on neither line, while the flow rises, falls or reverses, as before and after the peak in the MiTS results.
4. Slightly above the river level, for a small steady flow.
5. Held up by the river into a backwater profile, for a larger steady flow in a drain of mild slope.

[Figure 2](#fig-channel-five-states) sketches the five cases side by side, with the river held at a fixed level in every panel. Panels 1 and 2 repeat the two surfaces of [Figure 1](#fig-channel-river-drain).

Figure 2. Five behaviours of a drain discharging to a river at a fixed level

In this article we derive all five from conservation of mass and Newton’s second law, applied to a short reach of drain. We then simplify these general equations, one assumption at a time, into the three ways MSMA 3 allows the water surface to be calculated: uniform flow, a standard-step backwater profile and a full time-dependent simulation. For each we show which assumptions give MSMA 3’s printed equations. MSMA 3 leaves the time-dependent calculation to software, so we solve two simple cases by hand: a sudden inflow and a slowly rising one. Finally, we return to the MiTS example and reproduce its results after the storm and at the peak.

Our earlier [tailwater and backwater article](https://mes100.com/docs/tailwater-backwater-from-bernoulli/) covers the steady case only. It starts from Bernoulli’s equation and works the standard-step iteration in detail, including the MSMA 3 Appendix 6.5 example. This article starts one step further back, so that the same energy balance appears as a special case, and goes on to flow that changes with time. We do not repeat the trial calculations here.

# Governing equations for a channel reach

## The channel reach and its section quantities

A reach is a length of channel between two cross-sections that we choose for the analysis. It may be a whole drain or only part of one; it is the stretch of water whose mass and momentum we balance. We take a straight, fixed, prismatic channel: its cross-sectional shape does not change along the reach. The bed slope is small, the water is incompressible, and the surface is open to atmospheric pressure. The water surface is approximately level across each section. Depth changes gradually enough that vertical acceleration is small compared with gravity. This last assumption allows a hydrostatic pressure distribution even though the water is moving downstream.

Figure 3. A short reach and its channel cross-section

In [Figure 3](#fig-channel-volume), distance x  increases downstream and t  is time. The bed elevation z_b  and water-surface elevation z_s  use the same fixed datum. The vertical water depth is y=z_s-z_b . For the small slopes used here, we neglect the difference between vertical and bed-normal depth. The bed slope is S_0=-dz_b/dx , positive for a falling bed.

The flowing section area is A , the free-surface width is B , and the wetted perimeter is P . All are evaluated at the local depth. The hydraulic radius is R=A/P . Discharge Q  is positive downstream, and the section mean speed is V=Q/A  for the forward-flow cases considered below. Areas have units m², lengths m, discharge m³/s and speed m/s.

## Continuity equation

Take the short reach of length \Delta x  in [Figure 3](#fig-channel-volume). Let q_l  be net lateral inflow per unit channel length, in m²/s, positive into the channel. Runoff arriving along a drain can supply this inflow; a side withdrawal gives a negative value. With constant density, we can balance water volume directly:

 ddt∫xx+ΔxA(ξ,t) dξ=Q(x,t)−Q(x+Δx,t)+∫xx+Δxql(ξ,t) dξ.\frac{d}{dt}\int_x^{x+\Delta x} A(\xi,t)\,d\xi =Q(x,t)-Q(x+\Delta x,t)+\int_x^{x+\Delta x}q_l(\xi,t)\,d\xi.(1)

Here \xi  is a distance variable along the reach. Divide by \Delta x  and let the reach length tend to zero. The inflow-minus-outflow difference becomes the negative downstream derivative of discharge:

 ∂A∂t+∂Q∂x=ql.\frac{\partial A}{\partial t}+\frac{\partial Q}{\partial x}=q_l.(2)

Each term has units m²/s. If the section stores water, \partial A/\partial t  is positive. If flow is steady, [Equation (2)](#eq-channel-continuity) gives dQ/dx=q_l . Thus steady discharge is constant along the reach only when there is no net lateral inflow or withdrawal.

Figure 4. A 100 m reach receiving lateral runoff

For example, steady inflow of 4.000 m³/s with uniform lateral inflow of 0.002 m²/s over 100 m gives Q_{out}=4.000+0.002(100)=4.200\;\text{m}^3/\text{s} . Using one discharge for both ends would omit 0.200 m³/s of arriving water.

[Figure 4](#fig-channel-lateral) shows the numerical balance. In an unsteady reach, some of that arriving water can remain in storage, and the outlet need not immediately carry 4.200 m³/s. Integrating [Equation (2)](#eq-channel-continuity) over an entire facility gives the storage balance used in [level-pool routing](https://mes100.com/docs/from-conservation-of-mass-to-level-pool-routing/), which further assumes that the stored water surface stays level.

## Hydrostatic pressure and the downstream force balance

We measure gauge pressure p  relative to atmosphere. Let \eta  be height above the local bed, and \rho  water density. Negligible vertical acceleration gives \partial p/\partial\eta=-\rho g , with p=0  at \eta=y . Integrating from that surface gives

 p(η)=ρg(y−η).p(\eta)=\rho g(y-\eta).(3)

Figure 5. Hydrostatic pressure at one section

The orange triangle in [Figure 5](#fig-channel-pressure) gives the pressure distribution. For a rectangle of width B , integrate pressure over horizontal strips of area B\,d\eta . The total pressure force F_p  on the section is

 Fp=ρgB∫0y(y−η) dη=ρgB[yη−η22]0y=ρgBy22.F_p=\rho gB\int_0^y(y-\eta)\,d\eta =\rho gB\left[y\eta-\frac{\eta^2}{2}\right]_0^y =\frac{\rho gBy^2}{2}.(4)

We can also keep the prismatic shape general. Let b(\eta)  be the section width at height \eta  above the bed and define the pressure integral I_1  in m³:

 I1(y)=∫0y(y−η)b(η) dη,A(y)=∫0yb(η) dη.I_1(y)=\int_0^y(y-\eta)b(\eta)\,d\eta, \qquad A(y)=\int_0^y b(\eta)\,d\eta.(5)

Differentiating I_1  with respect to depth gives dI_1/dy=A : the upper-limit term is zero because y-\eta=0  at the surface, and differentiating the integrand leaves b(\eta) . Therefore the net pressure force on the short reach is, to first order, -\rho gA(\partial y/\partial x)\Delta x . A deeper downstream section pushes upstream more strongly.

Figure 6. The three forces along the drain on the water in a short reach

[Figure 6](#fig-channel-force-balance) shows the three forces along the drain on the water between the two cuts. The downstream component of weight is approximately \rho gAS_0\Delta x . Boundary shear acts against forward motion. Let \overline{\tau}  be the mean resisting shear stress over the wetted perimeter. Define the dimensionless friction slope by \overline{\tau}P=\rho gAS_f , so the resisting force is \rho gAS_f\Delta x . Adding these three forces gives \sum F_x , the net force on the water in the reach in the downstream x  direction, in N. We divide it by \rho\Delta x , the mass of water per unit section area, so that each term has units m³/s², the units of the momentum equation below:

 ∑FxρΔx=−gA∂y∂x+gA(S0−Sf).\frac{\sum F_x}{\rho\Delta x} =-gA\frac{\partial y}{\partial x}+gA(S_0-S_f).(6)

Equivalently, the pressure and gravity terms combine to -gA\,\partial z_s/\partial x , since z_s=z_b+y  and dz_b/dx=-S_0 . A level free surface has no net hydrostatic pressure-and-gravity drive along the reach. In a non-prismatic channel, changing wall geometry adds pressure-force contributions; differentiating I_1  as though it depended only on depth would omit them.

## Momentum and energy correction factors

Newton’s second law also needs the rate at which water carries momentum through each cut. If every particle crossed a section at the mean speed V , that rate would be \rho QV . In a real drain the speed is lower near the bed and walls and higher in the middle, and the faster water carries more than its share. Later, the energy balance meets the same problem with velocity head. We correct both with a factor that depends on the shape of the velocity distribution.

Figure 7. Two equal-area strips with different local velocities

In [Figure 7](#fig-channel-velocity-strips), the two bands have equal cross-sectional areas but different local downstream speeds. Their combined discharge gives the section mean V . We will use this stated distribution to illustrate the two correction factors below.

The local downstream velocity component u  varies over the area. The discharge is Q=\int_A u\,dA , but the rate at which downstream momentum crosses a section is \rho\int_Au^2\,dA . Squaring the mean speed generally gives a different result. Define the momentum correction factor \beta  by

 β=∫Au2 dAAV2,ρ∫Au2 dA=ρβQ2A.\beta=\frac{\int_Au^2\,dA}{AV^2}, \qquad \rho\int_Au^2\,dA=\rho\beta\frac{Q^2}{A}.(7)

The kinetic-energy flux involves a third power. For predominantly downstream flow, the kinetic-energy correction factor \alpha  is

 α=∫Au3 dAAV3,ρ2∫Au3 dA=ρQαV22.\alpha=\frac{\int_Au^3\,dA}{AV^3}, \qquad \frac{\rho}{2}\int_Au^3\,dA=\frac{\rho Q\alpha V^2}{2}.(8)

Both coefficients are dimensionless and equal one for a uniform axial speed. For two equal-area strips carrying speeds 0.5V  and 1.5V , the mean remains V , but \beta=(0.5^2+1.5^2)/2=1.25  and \alpha=(0.5^3+1.5^3)/2=1.75 . We use \beta  in momentum flux and \alpha  in velocity head. From the uniform-flow section onwards we take \alpha=\beta=1 , which is MSMA 3’s assumption of a uniform velocity distribution.

## Section momentum equation with lateral inflow

Figure 8. Water and momentum crossing the boundaries of a short reach

In [Figure 8](#fig-channel-momentum-reach), the dashed cuts bound the water whose momentum we balance. The upper panel shows the downstream forces and momentum carried through the cuts. The lower plan detail separates the amount of side inflow from its velocity component along the drain.

The downstream momentum held in the short reach is approximately \rho Q\Delta x . Its rate of accumulation is \rho(\partial Q/\partial t)\Delta x . The difference between outgoing and incoming momentum flux is \rho\,\partial(\beta Q^2/A)/\partial x\,\Delta x .

Lateral inflow also carries momentum. Let V_l  be its velocity component along the channel, in m/s, evaluated where it crosses the control-volume boundary. Its downstream momentum supply per unit reach length is \rho q_lV_l . Water entering perpendicular to the drain can have appreciable lateral speed while V_l  is approximately zero. For withdrawal, use the downstream component of the withdrawn water’s velocity with negative q_l .

Newton’s second law sets momentum accumulation plus net outgoing flux equal to the forces in [Equation (6)](#eq-channel-force-sum) plus the lateral supply. Dividing by \rho\Delta x  gives

 ∂Q∂t+∂∂x(βQ2A)+gA∂y∂x=gA(S0−Sf)+qlVl.\frac{\partial Q}{\partial t} +\frac{\partial}{\partial x}\left(\beta\frac{Q^2}{A}\right) +gA\frac{\partial y}{\partial x} =gA(S_0-S_f)+q_lV_l.(9)

Together, [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum) are the section-averaged shallow-water equations for our stated model. “Shallow” means that the vertical acceleration and pressure assumptions are appropriate over the longitudinal scale of interest; the water need not be shallow in metres. The equations retain changes with time and acceleration along the reach.

To see the effect of lateral water on the mean speed, take \beta=1  and substitute Q=AV . Applying the product rule to the first two terms gives

 ∂(AV)∂t+∂(AV2)∂x=A(∂V∂t+V∂V∂x)+V[∂A∂t+∂(AV)∂x].\frac{\partial(AV)}{\partial t}+\frac{\partial(AV^2)}{\partial x} =A\left(\frac{\partial V}{\partial t}+V\frac{\partial V}{\partial x}\right) +V\left[\frac{\partial A}{\partial t}+\frac{\partial(AV)}{\partial x}\right].(10)

The bracket is q_l  by [Equation (2)](#eq-channel-continuity). Subtract Vq_l  and divide by A>0 :

 ∂V∂t+V∂V∂x+g∂y∂x=g(S0−Sf)+qlA(Vl−V).\frac{\partial V}{\partial t}+V\frac{\partial V}{\partial x} +g\frac{\partial y}{\partial x} =g(S_0-S_f)+\frac{q_l}{A}(V_l-V).(11)

Perpendicular inflow with V_l=0  supplies water that the main stream must accelerate downstream, giving the negative term -q_lV/A . Setting V_l=0  therefore removes the lateral term from [Equation (9)](#eq-channel-momentum) but not from [Equation (11)](#eq-channel-velocity-form). If no water is added or removed along the reach, set q_l=0  in both balances. [USACE’s lateral-momentum guidance](https://www.hec.usace.army.mil/confluence/rasdocs/ras1dtechref/6.4/theoretical-basis-for-one-dimensional-and-two-dimensional-hydrodynamic-calculations/1d-unsteady-flow-hydrodynamics/implicit-finite-difference-scheme/lateral-influx-of-momentum) discusses the same acceleration of incoming water.

## MSMA 3 continuity and Saint-Venant equations as special cases

MSMA 3 writes these balances in two places in its Part 1. Section and equation numbers from MSMA 3 throughout this article refer to revision DFR2-Rev2. In each case we reduce [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum) to the printed form and state the assumption each step needs. MSMA 3 writes the section mean velocity as U  in §4.5.3 and V  in §4.5.8, and the flow depth as h  in §4.5.3 and y  in §4.5.8; we use V  and y  throughout.

First, take a channel with no water added or removed along it, so q_l=0 . Our two balances become

 ∂A∂t+∂Q∂x=0,∂Q∂t+∂∂x(βQ2A)=−gA∂y∂x+gA(S0−Sf).\frac{\partial A}{\partial t}+\frac{\partial Q}{\partial x}=0, \qquad \frac{\partial Q}{\partial t}+\frac{\partial}{\partial x}\left(\beta\frac{Q^2}{A}\right) =-gA\frac{\partial y}{\partial x}+gA(S_0-S_f).(12)

MSMA 3 §4.5.1 and §4.5.3 print

 ∂A∂t+∂Q∂x=0(Eq. 4.1),∂∂t(UA)+∂∂x(βU2A)=−gA∂h∂x+gA(So−Sf)(Eq. 4.11),\begin{aligned} &\frac{\partial A}{\partial t}+\frac{\partial Q}{\partial x}=0 &&\text{(Eq. 4.1)},\\ &\frac{\partial}{\partial t}(UA)+\frac{\partial}{\partial x}(\beta U^2A) =-gA\frac{\partial h}{\partial x}+gA(S_o-S_f) &&\text{(Eq. 4.11)}, \end{aligned}

and call the pair the Saint-Venant equations. Substituting Q=UA  and h=y  makes them identical to [Equation (12)](#eq-channel-saint-venant). For steady flow, \partial A/\partial t=0  and the continuity equation leaves \partial Q/\partial x=0 : the discharge is constant along the channel, Eq. 4.3 of MSMA 3.

Figure 9. Water supplied along a rectangular drain: rain, infiltration and bank inflow

Second, §4.5.8 keeps water arriving along the channel. Its equations describe a rectangular channel of bottom width b . Rain of intensity i  less infiltration at rate f , both in m/s, falls on the water surface, and lateral inflow q_L , in m²/s, enters from each of the two banks. [Figure 9](#fig-channel-msma-supply-rect) shows these supplies on one cross-section. Our net lateral inflow per unit length is then

 ql=(i−f)b+2qL.q_l=(i-f)b+2q_L.(13)

With A=by , Q=Vby  and constant b , we divide [Equation (2)](#eq-channel-continuity) by b  and expand the derivative of Vy :

 ∂y∂t+V∂y∂x+y∂V∂x=(i−f)+2qLb.\frac{\partial y}{\partial t}+V\frac{\partial y}{\partial x}+y\frac{\partial V}{\partial x} =(i-f)+\frac{2q_L}{b}.(14)

This is Eq. 4.46 of MSMA 3. For momentum we start from the velocity form, [Equation (11)](#eq-channel-velocity-form), which already takes \beta=1 ; MSMA 3 lists a uniform velocity distribution among its assumptions. Rain and bank inflow arrive with no velocity component along the drain, so V_l=0  and the lateral term is -q_lV/A . The friction term is gS_f=\overline{\tau}P/(\rho A)  from the definition of the friction slope, and for the rectangle P/A=(b+2y)/(by) . Writing \tau_0  for the bed shear stress \overline{\tau}  and \theta\approx S_0  for the small bed slope gives

 ∂V∂t+V∂V∂x+g∂y∂x=−[(i−f)+2qLb]Vy−(1+2yb)τ0ρy+gθ.\frac{\partial V}{\partial t}+V\frac{\partial V}{\partial x}+g\frac{\partial y}{\partial x} =-\left[(i-f)+\frac{2q_L}{b}\right]\frac{V}{y} -\left(1+\frac{2y}{b}\right)\frac{\tau_0}{\rho y}+g\theta.(15)

Eq. 4.47 of MSMA 3 has the same acceleration, pressure, friction and slope terms, but prints the lateral-supply term with a plus sign, +\{(i-f)+2q_L/b\}V/y . With that sign, water arriving with no downstream velocity would speed the stream up. The momentum balance requires the opposite: the stream shares its downstream momentum with the arriving water, so this term slows it. We use the minus sign of [Equation (15)](#eq-channel-msma-momentum-rect). MSMA 3 lists its assumptions as shallow water, a small bed slope and a uniform velocity distribution, which are ours.

## Uniform, gradually varied and unsteady flow

Before simplifying the equations, we note how MSMA 3 uses them in design. Its design procedure, Part 1 §2.6.9.3, sizes a drain by the conventional capacity calculation and then revises the calculation to include tailwater, friction loss and junction loss. Its item vi allows three ways to calculate the resulting water-surface profile, listed in [Table 1](#tbl-channel-msma-calculations).

**Table 1.** The three MSMA 3 water-surface calculations and the equations that give them

| Calculation (MSMA 3 name) | What it calculates | Equations in this article |
| --- | --- | --- |
| Normal depth (“fully simplified”) | Uniform flow in each drain segment, the Manning capacity calculation used to size a drain for its design peak flow | Normal depth, [Equation (17)](#eq-channel-normal) |
| Backwater profile (“partially simplified”) | Non-uniform profile by the standard-step method, starting from the tailwater level | GVF, [Equation (26)](#eq-channel-gvf), and standard step, [Equation (28)](#eq-channel-standard-step) |
| Time-dependent simulation (“computer simulation”) | The full dynamic wave equations | [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum) with their time derivatives |

Part 2 §6.2.3 adds that an unsteady flow model needs computer-based simulation and is not commonly used in typical stormwater drain design. MSMA 2 (2012) sizes drains by the first approach, with Manning capacity and V=Q/A  in Chapter 14, §14.2.5 and Appendix 14.A.

Each approach drops terms from the full equations, and each treats the river level differently. Uniform flow ignores it, the standard step starts from it, and the full equations follow how the drain approaches it with time. The following sections make each reduction and state the conditions it needs.

None of the three calculations answers the simplest question a fixed river raises: when may we take the drain level as the river level? To answer it we use one technique twice. We take water at rest at the river level as the base state, treat the flow as a small departure from it, and keep only the largest term in that departure. In the steady part of the article the small quantity is the discharge. The departure then begins at C_uQ^2 , a result of the GVF equation. In the unsteady part the small quantity is the amplitude of a disturbance that changes with time. Its leading terms are linear, giving the linearised equations, which describe waves. A slowly changing inflow then joins the two results: the steady C_uQ^2  rise plus a head that accelerates the water.

[Figure 10](#fig-channel-path) maps both threads. Solid arrows lead to MSMA 3’s three calculations. Dashed arrows lead to the shaded perturbation results, which this article derives in addition to MSMA 3. Each section states the full assumptions behind its arrow.

Figure 10. From the general channel equations to MSMA 3's three water-surface calculations and the perturbation results about the resting river level

# Steady uniform flow

## Assumptions and reduction from the governing equations

We retain the straight, fixed, prismatic, small-slope channel and hydrostatic pressure assumptions of the governing equations. For steady uniform flow, we also take no water added or removed along the reach and a uniform section velocity distribution:

 ql=0,β=1,∂A∂t=∂Q∂t=0,dydx=0.q_l=0,\qquad \beta=1,\qquad \frac{\partial A}{\partial t}=\frac{\partial Q}{\partial t}=0, \qquad \frac{dy}{dx}=0.

Continuity, [Equation (2)](#eq-channel-continuity), then gives dQ/dx=0 . Since the section is prismatic and depth is constant, dA/dx=0  and dV/dx=0 . Thus d(Q^2/A)/dx=0 , and [Equation (9)](#eq-channel-momentum) reduces to 0=gA(S_0-S_f) . For a wet section, A>0 , this gives S_f=S_0 . We use Manning resistance below to find the depth that satisfies this balance.

## Manning resistance and normal depth

Momentum gives us the relation between boundary shear and friction slope. We still need a resistance relationship to calculate that slope. We use Manning’s empirical relationship in SI units for forward flow:

 V=1nR2/3Sf,Sf=n2Q2A2R4/3.V=\frac{1}{n}R^{2/3}\sqrt{S_f}, \qquad S_f=\frac{n^2Q^2}{A^2R^{4/3}}.(16)

The roughness coefficient n  has SI dimensions s/m^{1/3} . Its value represents the chosen lining and roughness condition; we do not derive it from Newton’s law. For reversing flow, resistance must oppose the actual velocity, so in a fixed downstream coordinate S_f  must be written proportional to Q|Q| .

Figure 11. Uniform flow has equal depths and parallel bed and water surface

In [Figure 11](#fig-channel-uniform), equal depth at the two cuts makes the water surface parallel to the bed. The two grey flow arrows have the same mean speed, and the same discharge passes both cuts. Equal depths and section shapes give equal pressure forces at the cuts, so their net contribution vanishes. The purple downstream weight balances the red boundary resistance. Substituting Manning’s relation, [Equation (16)](#eq-channel-manning), into S_f=S_0  gives

 Sf=S0,Q=1nA(yn)R(yn)2/3S0.S_f=S_0,\qquad Q=\frac{1}{n}A(y_n)R(y_n)^{2/3}\sqrt{S_0}.(17)

The depth y_n  satisfying [Equation (17)](#eq-channel-normal) is the normal depth. It describes the equilibrium between downstream weight and resistance for the specified discharge, section, roughness and positive bed slope. A downstream structure or river can impose another depth.

This is the Manning capacity equation that MSMA 2 and MSMA 3 use to size a drain for its design peak flow, and the simplest of the three MSMA 3 water-surface calculations. The MSMA 3 normal-depth formula, Part 1 §4.5.5.2, is for a wide rectangular channel. Substituting A=By  and the wide-channel approximation R\approx y  into [Equation (17)](#eq-channel-normal) gives Q=(1/n)By^{5/3}\sqrt{S_0} , so

 yn≈(nQBS0)3/5.y_n\approx\left(\frac{nQ}{B\sqrt{S_0}}\right)^{3/5}.(18)

MSMA 3 prints the same expression as Eq. 4.18, h_o=\left(nQ/BS_o^{1/2}\right)^{3/5} , and states B>10h  for it, where h  is the flow depth.

That condition limits the error of the wide-channel approximation. For a rectangle the exact hydraulic radius is R=By/(B+2y)=y/(1+2y/B) ; the 2y  is the wetted height of the two walls, which R\approx y  ignores. Both [Equation (17)](#eq-channel-normal) and [Equation (18)](#eq-channel-wide-normal) must carry the same discharge, so By_n^{5/3}(1+2y_n/B)^{-2/3}=By_{wide}^{5/3} , which gives

 ywide=yn(1+2ynB)−2/5.y_{wide}=y_n\left(1+\frac{2y_n}{B}\right)^{-2/5}.(19)

Here y_n  is the true normal depth and y_{wide}  the depth from [Equation (18)](#eq-channel-wide-normal). Because the wide-channel formula leaves out the wall friction, it always gives too small a depth. At MSMA 3’s limit, B/y_n=10 , it is 7% low.

Narrower channels fare worse. At B/y_n=4  the formula is 15% low, and the MiTS drain at peak flow, with B/y_n=1.2/0.415=2.9 , would be 19% low. Our worked examples therefore keep the walls through R=A/P . The one exception is the MSMA 3 worksheet we reproduce later, which uses R\approx y  even though its B/h  is only about 4.

Normal depth also has a time-dependent counterpart, the kinematic wave of MSMA 3 §4.5.8.1. It keeps the time derivative in continuity but neglects the inflow, surface-slope and inertia terms in momentum compared with bed slope and friction. For a wide rectangular channel, let q=Vy  be the discharge per unit width, in m²/s. Continuity is [Equation (14)](#eq-channel-msma-continuity-rect) with V\,\partial y/\partial x+y\,\partial V/\partial x=\partial q/\partial x . Momentum is reduced to S_f=S_0  at every instant, so [Equation (18)](#eq-channel-wide-normal) holds for the current discharge:

 ∂y∂t+∂q∂x=(i−f)+2qLb,q=S0n y5/3.\frac{\partial y}{\partial t}+\frac{\partial q}{\partial x}=(i-f)+\frac{2q_L}{b}, \qquad q=\frac{\sqrt{S_0}}{n}\,y^{5/3}.(20)

MSMA 3 prints these as Eq. 4.48 and Eq. 4.49, q=\alpha y^m , where its rating coefficient \alpha  and exponent m  depend on whether the flow is laminar or turbulent. That \alpha  is unrelated to our energy correction factor. Manning resistance gives the turbulent values: the rating coefficient is \sqrt{S_0}/n  and m=5/3 . A kinematic wave carries changes in discharge and depth downstream only, so it cannot carry a river level up the drain.

# Steady gradually varied flow

## Specific energy and critical depth

Before deriving the backwater profile we need one more depth, the critical depth, because it decides whether the river can affect the drain at all. Flow deeper than critical depth is slow enough for a change at the river to travel upstream into the drain; flow shallower than critical depth is too fast, and the river cannot reach in. Comparing critical depth with the normal depth of [Equation (17)](#eq-channel-normal) therefore tells us whether the river controls the drain. We use this comparison for the 300 m example drain and again for the MiTS drain at its peak.

We consider steady flow with no water added or removed along the reach. Continuity, [Equation (2)](#eq-channel-continuity), then gives \partial Q/\partial x=0 , so the discharge Q  is the same at every section. We also take the energy correction factor of [Equation (8)](#eq-channel-alpha) as \alpha=1 , which treats the velocity as uniform over the section. The specific energy E  is the total head at a section measured above the local bed: the depth plus the velocity head. At a fixed discharge it changes with depth, and we call the depth at which it is smallest the critical depth y_c .

Figure 12. Depth and velocity head at fixed discharge on either side of critical flow

In [Figure 12](#fig-channel-critical-states), the blue column represents water depth and the orange column represents velocity head. A shallow flow needs a larger velocity to carry the same discharge through the smaller area. A deep flow has less velocity head but a larger depth. At the critical depth their sum is smallest. The three panel values follow by substituting A=By  into [Equation (21)](#eq-channel-specific-energy) and using the critical condition below; the panels are alternative flow states, not three consecutive sections of one channel.

For unit energy correction factor, the specific energy is

 E(y)=y+V22g=y+Q22gA(y)2.E(y)=y+\frac{V^2}{2g}=y+\frac{Q^2}{2gA(y)^2}.(21)

Increasing depth adds pressure-and-elevation head above the bed but enlarges the flowing area and reduces velocity head. Define hydraulic depth D_h=A/B  and Froude number Fr=V/\sqrt{gD_h} . In the rectangular panels, D_h=y . Since dA/dy=B , differentiation at fixed discharge gives

 dEdy=1−Q2BgA3=1−Fr2,Fr=VgDh,Dh=AB.\frac{dE}{dy}=1-\frac{Q^2B}{gA^3}=1-Fr^2, \qquad Fr=\frac{V}{\sqrt{gD_h}},\quad D_h=\frac{A}{B}.(22)

The minimum occurs where this derivative is zero, at Fr=1 . That first condition below applies to any prismatic section; for a rectangle, substituting A(y_c)=By_c  gives the second expression:

 Q2BgA(yc)3=1,yc=(Q2gB2)1/3.\frac{Q^2B}{gA(y_c)^3}=1, \qquad y_c=\left(\frac{Q^2}{gB^2}\right)^{1/3}.(23)

For a rectangle with Q>0 , the second derivative is 3Q^2/(gB^2y^4)>0 , confirming the minimum. Critical depth depends only on discharge and geometry; normal depth also depends on resistance and bed slope. That difference is why comparing the two depths tells us which one controls the drain.

## Gradually varied flow equation

### Assumptions and reduction from the governing equations

The river-controlled reach in [Figure 13](#fig-channel-gvf-profile) carries the same discharge through both cuts, while its depth varies along the drain.

Figure 13. Backwater depth increases downstream while surface elevation falls

We keep the prismatic, small-slope, hydrostatic channel model, but allow depth to vary along the reach. The additional conditions are

 ql=0,β=1,∂A∂t=∂Q∂t=0.q_l=0,\qquad \beta=1,\qquad \frac{\partial A}{\partial t}=\frac{\partial Q}{\partial t}=0.

Continuity gives dQ/dx=0 , so V=Q/A(y)  varies only through depth. We retain dy/dx  and the convective acceleration V\,dV/dx  in momentum. These are the terms that distinguish a backwater profile from uniform flow. The divided GVF equation below also requires Fr\ne1 ; the river-controlled profile illustrated here uses the subcritical branch, Fr<1 , with its downstream level specified.

### Derivation of the depth gradient

For steady flow with no water added or removed along the reach, [Equation (11)](#eq-channel-velocity-form) reduces to

 VdVdx+gdydx=g(S0−Sf).V\frac{dV}{dx}+g\frac{dy}{dx}=g(S_0-S_f).(24)

With constant discharge, V=Q/A , so

 dVdx=−QA2dAdx=−VBAdydx.\frac{dV}{dx}=-\frac{Q}{A^2}\frac{dA}{dx} =-\frac{VB}{A}\frac{dy}{dx}.(25)

Substituting [Equation (25)](#eq-channel-speed-gradient) into [Equation (24)](#eq-channel-steady-momentum) and dividing by gravity gives

 (1−V2BgA)dydx=S0−Sf,dydx=S0−Sf1−Fr2.\left(1-\frac{V^2B}{gA}\right)\frac{dy}{dx}=S_0-S_f, \qquad \frac{dy}{dx}=\frac{S_0-S_f}{1-Fr^2}.(26)

This is the gradually varied flow (GVF) equation. The numerator measures the difference between bed slope and resistance; the denominator accounts for the changing velocity head as the depth changes.

For a mild rectangular channel, y_n>y_c . If a downstream river imposes y>y_n , the larger area and hydraulic radius reduce S_f  below S_0 . Flow remains subcritical, so both numerator and denominator in [Equation (26)](#eq-channel-gvf) are positive: depth increases downstream. However, the bed falls downstream. The water-surface elevation can therefore fall downstream while the depth increases. In [Figure 13](#fig-channel-gvf-profile), the solid blue surface falls towards the river while its distance above the brown bed increases.

For subcritical flow, small surface disturbances can travel upstream as well as downstream: their ideal shallow-water speeds are V-\sqrt{gD_h}  and V+\sqrt{gD_h} . We derive the wave speed \sqrt{gD_h}  from the unsteady equations in [Equation (49)](#eq-channel-rest-wave); the Froude number is the ratio of the water speed to it. A change in river level can therefore travel up the drain only when Fr<1 . We start a steady subcritical profile from a justified downstream level and calculate upstream. A wholly supercritical reach needs its upstream control; a downstream high level may instead create a hydraulic jump and change the flow arrangement. Near critical flow, a large gradient or a jump requires a transition or momentum calculation suited to that region; we do not march the divided GVF equation through its zero denominator.

This profile is the basis of MSMA 3’s backwater calculation. MSMA 3 §4.5.6 prints

 dhdx=So−Sf1−Fr2(Eq. 4.20),hc=(Q2gB2)1/3(Eq. 4.21),Q2T(h)g[A(h)]3=1(Eq. 4.22).\begin{aligned} &\frac{dh}{dx}=\frac{S_o-S_f}{1-Fr^2} &&\text{(Eq. 4.20)},\\ &h_c=\left(\frac{Q^2}{gB^2}\right)^{1/3} &&\text{(Eq. 4.21)},\\ &\frac{Q^2T(h)}{g[A(h)]^3}=1 &&\text{(Eq. 4.22)}. \end{aligned}

With h=y  and the top width T=B , these are [Equation (26)](#eq-channel-gvf) and the two conditions of [Equation (23)](#eq-channel-critical). They carry the assumptions of our reduction: steady flow, no water added or removed along the reach, a prismatic channel, hydrostatic pressure and unit correction factors. MSMA 3 measures h  perpendicular to the bed; for the small slopes here that equals our vertical depth. In the MSMA 3 Froude number the relevant depth is A/T , our D_h , while in its GVF equation the differentiated depth is the local water depth y . We keep D_h  and y  separate because they differ outside a rectangle.

## Standard-step calculation between two sections

Standard step solves a finite-reach energy balance. From [Equation (24)](#eq-channel-steady-momentum),

 ddx(y+V22g)=S0−Sf.\frac{d}{dx}\left(y+\frac{V^2}{2g}\right)=S_0-S_f.(27)

Add dz_b/dx=-S_0 . The total head H=z_b+y+V^2/(2g)  then satisfies dH/dx=-S_f . Integrating between upstream section u  and downstream section d , separated by \Delta x>0 , gives the head loss h_L=\int_{x_u}^{x_d}S_f\,dx . The endpoint heads and their components are marked in [Figure 14](#fig-channel-section-heads).

Figure 14. Depth, bed elevation, velocity head and head loss between two sections

In [Figure 14](#fig-channel-section-heads), the green line is the energy grade line and the blue line is the water surface. The orange brackets measure the velocity heads between them. The total-head drop h_L=H_u-H_d  is measured between the two green elevations, with both heads referred to the same datum. The energy balance becomes

 zb,u+yu+αuVu22g=zb,d+yd+αdVd22g+hL,hL≈Sf,u+Sf,d2Δx.\begin{aligned} z_{b,u}+y_u+\frac{\alpha_uV_u^2}{2g} &=z_{b,d}+y_d+\frac{\alpha_dV_d^2}{2g}+h_L,\\ h_L&\approx\frac{S_{f,u}+S_{f,d}}{2}\Delta x. \end{aligned}(28)

The derivation and numerical examples here take both energy correction factors as one. The endpoint mean friction slope is a numerical approximation to the friction integral. If a reach contains a distinct junction or transition loss, add its justified head loss separately and use its specified velocity. Our straight uniform-width drain has no such local loss.

The MSMA 3 standard-step method, Part 1 §4.5.6.3, prints

 z1+h1+α1V122g=z2+h2+α2V222g+hL(Eq. 4.24),hL≈SˉfΔx+∑hminor,Sˉf=12(Sf1+Sf2)(Eq. 4.25),Sf=(nQAR2/3)2(Eq. 4.26).\begin{aligned} z_1+h_1+\frac{\alpha_1V_1^2}{2g}&=z_2+h_2+\frac{\alpha_2V_2^2}{2g}+h_L\quad\text{(Eq. 4.24)},\\ h_L&\approx\bar{S}_f\Delta x+\sum h_{minor},\qquad \bar{S}_f=\tfrac12(S_{f1}+S_{f2})\quad\text{(Eq. 4.25)},\\ S_f&=\left(\frac{nQ}{AR^{2/3}}\right)^2\quad\text{(Eq. 4.26)}. \end{aligned}

With section 1 upstream, section 2 downstream, z=z_b , h=y  and no minor losses, these are [Equation (28)](#eq-channel-standard-step) and the Manning friction slope of [Equation (16)](#eq-channel-manning). Part 2 §6.4.7, Eqs. 6.21–6.22, applies the same balance to drains. For subcritical flow MSMA 3 marches upstream from a known section 1, which is then the downstream section, and the head loss belongs on its side of Eq. 4.24. We label the sections u  and d  by position so that the head loss is always added to the downstream total head.

Given downstream depth, we solve [Equation (28)](#eq-channel-standard-step) for the upstream depth on the appropriate flow branch. The [existing backwater article](https://mes100.com/docs/tailwater-backwater-from-bernoulli/#msma-3-worked-example-appendix-65) shows the trial calculations. At the end of this section we apply it to the same MSMA 3 example and compare one long step with a resolved profile.

## Level surface at zero discharge

The GVF equation, [Equation (26)](#eq-channel-gvf), can be integrated numerically for any flow, and we do so later in this section. A numerical profile, however, gives the depths and levels only for the one flow we put in. Engineers want to know: when may we simply take the drain water level as the river level, and how quickly does that approximation fail as the flow increases?

We find the answers by perturbing the GVF profile about water at rest, the technique outlined with [Figure 10](#fig-channel-path). First, with no flow the drain level is exactly the river level. Second, for a small flow we obtain a closed-form estimate of how far the surface rises above that level, which gives the largest flow for which the river level is still acceptable. Both this small-flow rise and the later time-dependent changes are measured from the level surface derived here.

### Assumptions and reduction from the governing equations

[Figure 15](#fig-channel-rest-connection) locates the wet reach connected to the river and the water edge where the wet-section equations stop.

Figure 15. The river fixes the connected resting surface up to the dry edge

We consider the part of the drain connected to the river and allow it to settle to rest. No rainfall or runoff arrives along the reach and no water is withdrawn or lost through the bed, so q_l=0 . The water has stopped accelerating, and with no flow there is no resistance. In a wet section, the conditions are

 Q=V=0,ql=0,Sf=0,∂A∂t=∂Q∂t=0.Q=V=0,\qquad q_l=0,\qquad S_f=0,\qquad \frac{\partial A}{\partial t}=\frac{\partial Q}{\partial t}=0.

The momentum-flux term vanishes with Q . With A>0  and dz_b/dx=-S_0 , the remaining terms of [Equation (9)](#eq-channel-momentum) give

 gAdydx=gAS0,dzsdx=dzbdx+dydx=−S0+S0=0.gA\frac{dy}{dx}=gAS_0, \qquad \frac{dz_s}{dx}=\frac{dz_b}{dx}+\frac{dy}{dx}=-S_0+S_0=0.(29)

The connected river boundary, z_s(x_d)=Z_r , fixes the constant surface elevation. This wet-section reduction stops at the water edge; we treat the dry bed separately below.

### Resting surface and water depth

Figure 16. Why still water lies level

[Figure 16](#fig-channel-rest-balance) shows what the two equations of [Equation (29)](#eq-channel-rest-surface) mean. The first is a balance of two forces on still water. The weight of the water pulls it down the sloping bed, the term gAS_0 , shown in purple. Because the water is deeper downstream, the pressure there is higher and pushes it back, the term gA\,dy/dx , shown in orange. The two balance when the depth grows downstream at exactly the rate the bed falls. The second equation shows what that means for the surface: the depth gained equals the bed lost, so the surface elevation does not change along the drain. Still water lies level, as in a bathtub or a pond. The GVF equation, [Equation (26)](#eq-channel-gvf), says the same with Q=0 : with no flow there is no friction, S_f=0 , and no velocity, Fr=0 , which leaves dy/dx=S_0 .

The river decides which level. The drain and the river form one connected body of still water, so wherever the drain is wet its surface stands at the river elevation Z_r . The depth at any point is then simply the distance from the bed up to that level:

 y0(x)=Zr−zb(x)where the bed is submerged.y_0(x)=Z_r-z_b(x)\quad\text{where the bed is submerged}.(30)

The subscript zero marks this resting state. In the MiTS drain the river stands at Z_r=  RL 28.50 m. The outlet, with invert RL 28.00 m, holds 0.50 m of water; the depth shrinks steadily upstream and reaches zero halfway along, where the invert is RL 28.50 m. That is the water edge in [Figure 1](#fig-channel-river-drain).

Upstream of the water edge the bed is above the river level and the drain is dry. [Equation (30)](#eq-channel-rest-depth) would give a negative depth there, so the resting depth along the drain is y_0=\max(0,Z_r-z_b) .

## Small-discharge approximation to the GVF equation

### Assumptions and reduction from the governing equations

In [Figure 17](#fig-channel-small-flow), the dashed resting surface stays above the bed throughout the reach. The solid blue surface shows the small steady-flow departure from the same fixed river level.

Figure 17. Resting water and a small steady flow against the same river

We use the steady GVF assumptions, with q_l=0 , dQ/dx=0  and \alpha=\beta=1 . The river level is fixed, and the whole reach must be wet at rest, with smooth section geometry away from a dry edge or critical flow:

 zs(xd)=Zr,y0(x)=Zr−zb(x)>0,Fr2≪1,∣y−y0∣≪y0.z_s(x_d)=Z_r,\qquad y_0(x)=Z_r-z_b(x)>0,\qquad Fr^2\ll1,\qquad |y-y_0|\ll y_0.

For forward flow we use Manning resistance, S_f=n^2Q^2/(A^2R^{4/3}) . Friction and velocity head therefore both contain Q^2 . Keeping only the leading surface change lets us replace A  and R  in their coefficients by the resting values A_0  and R_0 . The derivation below gives \delta z_u\approx C_uQ^2 . The illustrated rise above the river requires C_u>0 ; its magnitude must also be small enough for the resting-geometry approximation to remain accurate.

### Leading surface change from the resting state

Now let a small steady discharge Q  flow down the drain into the river, with the river level, bed and roughness unchanged. Here Q  is an input: the steady flow the drain must pass, such as a design peak from the Rational Method, or the link discharge a model reports. [Figure 17](#fig-channel-small-flow) shows what happens. The river still fixes the surface at the outlet, and the surface upstream rises slightly to push the flow through. How far does it rise? We will find that for a small flow the rise is proportional to Q^2 : halving the flow quarters the rise, so a small flow barely moves the level the river has set.

Here \delta z_u=z_{s,u}-Z_r  is the upstream surface change from rest. The steady profile obeys the GVF equation. Because we want the water surface rather than the depth, we write [Equation (26)](#eq-channel-gvf) with z_s=z_b+y  and dz_b/dx=-S_0 :

 dzsdx=Fr2S0−Sf1−Fr2.\frac{dz_s}{dx}=\frac{Fr^2S_0-S_f}{1-Fr^2}.(31)

At rest both Fr^2  and S_f  are zero, and the surface is level. With flow, the surface slopes down towards the river mainly because friction S_f  must be overcome. Rather than integrate [Equation (31)](#eq-channel-gvf-surface) directly, we can use its integrated form between the two ends, the energy balance [Equation (28)](#eq-channel-standard-step), which comes from the same steady momentum equation. Taking \alpha_u=\alpha_d=1 , and retaining the friction integral before approximating it, gives

 δzu=∫xuxdSf dx+Q22g(1Ad2−1Au2).\delta z_u =\int_{x_u}^{x_d}S_f\,dx+\frac{Q^2}{2g} \left(\frac{1}{A_d^2}-\frac{1}{A_u^2}\right).(32)

This is an identity within our steady energy model: the areas are still the unknown flowing areas. Manning’s relation makes S_f=n^2Q^2/(A^2R^{4/3}) . Both terms on the right are proportional to Q^2  when we hold the geometry at its resting value.

To make the consequence precise, we write the upstream surface change as a power series in the discharge itself:

 δzu=a1Q+a2Q2+⋯ .\delta z_u=a_1Q+a_2Q^2+\cdots.(33)

The coefficients a_1  (s/m²) and a_2  (s²/m⁵) depend on the channel and the river level, but not on Q . Compare this with [Equation (32)](#eq-channel-steady-surface-change). Every term on its right-hand side carries a factor Q^2 , so nothing there can match a_1Q . Thus a_1=0 : the surface does not start to rise in proportion to the flow. The first change is proportional to Q^2 . [Figure 18](#fig-channel-expansion-terms) shows the difference: the curve leaves zero flat, whereas a linear term would leave it at a slope.

Figure 18. Upstream surface change against discharge: no linear term, only a quadratic one

The flowing areas in [Equation (32)](#eq-channel-steady-surface-change) differ from their resting values because the surface has risen. For smooth section geometry away from dry or critical conditions, that rise is itself proportional to Q^2 , so the area and hydraulic-radius changes are too. Substituting them into a term already multiplied by Q^2  changes it only at order Q^4 . To find a_2  we may therefore use the resting geometry.

We evaluate the section at its resting depth y_0(x) . Let A_0(x)=A[y_0(x)]  be that area and P_0(x)=P[y_0(x)]  the length of bed and walls touching the water. The hydraulic radius is the ratio R_0(x)=A_0(x)/P_0(x) , measured in metres. [Figure 19](#fig-channel-resting-geometry) shows these quantities at one section.

Figure 19. Resting cross-section: area, wetted perimeter and hydraulic radius

For the rectangular section in [Figure 19](#fig-channel-resting-geometry), A_0=By_0 , P_0=B+2y_0  and R_0=By_0/(B+2y_0) . The free-surface width B_0  equals B . Hydraulic depth uses the surface width, D_0=A_0/B_0 , whereas hydraulic radius uses the wetted perimeter. We now calculate the leading coefficient, a_2 , which we call C_u , using that resting geometry:

 δzu≈Q2Cu,Cu=∫xuxdn2A0(x)2R0(x)4/3 dx+12g[1A0,d2−1A0,u2].\begin{aligned} \delta z_u&\approx Q^2 C_u,\\ C_u&=\int_{x_u}^{x_d}\frac{n^2}{A_0(x)^2R_0(x)^{4/3}}\,dx +\frac{1}{2g}\left[\frac{1}{A_{0,d}^2}-\frac{1}{A_{0,u}^2}\right]. \end{aligned}(34)

We obtain the same result by simplifying [Equation (31)](#eq-channel-gvf-surface) directly. Both Fr^2  and S_f  are proportional to Q^2 , so for a small flow the denominator 1-Fr^2  may be replaced by one, and both may be evaluated at the resting geometry. Integrating -S_f  from the upstream end to the river gives the friction integral in C_u . At rest the depth increases downstream at the rate S_0 , so Fr^2S_0  is the rate at which velocity head Q^2/(2gA_0^2)  falls along the reach; integrating it gives the endpoint velocity-head term. [Equation (34)](#eq-channel-small-steady-correction) is therefore the GVF profile with only its leading term in Q  retained.

The coefficient C_u  has units s²/m^5 , making Q^2C_u  a length. The neglected term is of order Q^4 ; we check its size numerically below. Friction gives a positive contribution; the endpoint velocity-head difference can have either sign. A shallower upstream section accelerates the water more strongly and can reduce the upstream surface rise needed to pass the flow. Thus a surface rise is not guaranteed by the existence of a small forward discharge alone.

For a horizontal prismatic channel of resting depth y_0  and length L , both resting areas are equal. The endpoint term cancels and the integrand is constant. With V_1=Q/A_0  denoting the leading mean velocity, [Equation (34)](#eq-channel-small-steady-correction) becomes

 δzu≈n2LA02R04/3 Q2=n2LR04/3V12.\delta z_u\approx\frac{n^2L}{A_0^2R_0^{4/3}}\,Q^2=\frac{n^2L}{R_0^{4/3}}V_1^2.(35)

The surface rise depends on the through-discharge Q , entering upstream, through V_1 . It is the friction head needed to push that discharge along length L  at the resting area.

## Applicability of the small-discharge result

[Equation (34)](#eq-channel-small-steady-correction) and [Equation (35)](#eq-channel-horizontal-perturbation) say four things about a drain whose outlet level is held by a river:

1. **The river still sets the level.** At the outlet the rise is zero. Moving upstream, the surface climbs gradually above the river level, and the largest rise is at the upstream end.
2. **The rise grows with the square of the flow.** Doubling Q  multiplies \delta z_u  by about four, and halving it divides by about four. For a small flow, the drain level is therefore very close to the river level.
3. **Friction causes the rise.** The surface must slope to drive the water against bed and wall resistance, so a longer, rougher or shallower drain rises more for the same flow. The velocity-head term adds or subtracts a smaller amount, depending on how the flowing area changes between the two ends.
4. **We can tell when “drain level = river level” is good enough.** If the design can accept a level error \Delta  (m), the flow may be treated as small while C_uQ^2\le\Delta .

Rearranging the last condition gives the largest flow for which the river level may be used directly:

 Q≤ΔCu.Q\le\sqrt{\frac{\Delta}{C_u}}.(36)

Above that flow we must calculate the backwater profile instead. This limit assumes the leading estimate is accurate, which the numerical check below confirms for flows of that size.

A small Froude number alone does not guarantee a small rise, because friction head accumulates over the length of a long reach. The quadratic conclusion also depends on the resistance law. Manning resistance is proportional to V^2 ; if resistance were proportional to velocity, as in laminar flow, the leading rise would instead be proportional to Q .

## Numerical check against the full GVF profile

We reuse the illustrative 300 m rectangular drain, 3.000 m wide, with n=0.015\;\text{s/m}^{1/3} , S_0=0.001  and g=9.81\;\text{m/s}^2 . Its upstream and downstream inverts are RL 10.000 m and RL 9.700 m. The fixed river is RL 10.900 m. RL denotes elevation above one survey datum. At rest, depth is 0.900 m upstream and 1.200 m downstream, so the whole reach is wet.

Figure 20. A 0.500 m³/s flow raises the upstream surface by about 2.59 mm

For distance 0\le x\le300\;\text{m}  measured downstream, the resting depth is y_0(x)=0.900+0.001x . The finite-width geometry is A_0=3y_0 , P_0=3+2y_0  and R_0=A_0/P_0 . At the upstream end these give A_0=2.700\;\text{m}^2 , P_0=4.800\;\text{m}  and R_0=0.562500\;\text{m} . At the downstream end they give A_0=3.600\;\text{m}^2 , P_0=5.400\;\text{m}  and R_0=0.666667\;\text{m} . Substituting in [Equation (34)](#eq-channel-small-steady-correction) gives a friction contribution of 0.0134256275\;\text{s}^2/\text{m}^5  and an endpoint velocity contribution of -0.0030588031\;\text{s}^2/\text{m}^5 . Thus

 Cu=0.0103668244  s2/m5,δzu≈0.0103668244 Q2C_u=0.0103668244\;\text{s}^2/\text{m}^5, \qquad \delta z_u\approx0.0103668244\,Q^2(37)

when discharge is in m³/s and the surface change is in metres. In [Table 2](#tbl-channel-small-flow), the leading surface change is this formula, C_uQ^2  from [Equation (37)](#eq-channel-numerical-perturbation), for each discharge. We compare it with the full nonlinear GVF equation, [Equation (26)](#eq-channel-gvf), integrated upstream from the same 1.200 m outlet depth. [Figure 20](#fig-channel-perturbation-values) shows the 0.500 m³/s case.

**Table 2.** Upstream surface change for the 300 m drain: leading term C_uQ^2  against the full GVF profile

| Steady discharge Q  (m³/s) | Leading surface change C_uQ^2  (mm) | Full GVF surface change (mm) | Difference relative to full GVF change |
| --- | --- | --- | --- |
| 0 | 0 | 0 | — |
| 0.25 | 0.647927 | 0.647772 | 0.024% |
| 0.50 | 2.591706 | 2.589230 | 0.096% |
| 1.00 | 10.366824 | 10.326886 | 0.387% |
| 2.00 | 41.467297 | 40.811131 | 1.608% |
| 4.00 | 165.869190 | 154.846822 | 7.118% |

Doubling discharge from 0.25 to 0.50 m³/s almost quadruples the small surface change. The approximation is within about 0.002 mm of the full calculation at 0.50 m³/s, but overestimates the change by about 11 mm at 4 m³/s. The extra digits in [Table 2](#tbl-channel-small-flow) expose the numerical comparison; field roughness and survey accuracy are separate limits.

At 4 m³/s, normal depth is 0.919800 m and critical depth is 0.565895 m. The river-controlled upstream depth is 1.054847 m and upstream surface RL 11.054847 m. Extending the river level horizontally would give depth 0.900 m there, about 155 mm too low for this flow. Normal depth, the simplest MSMA 3 calculation, gives upstream surface RL 10.920 m, still 135 mm too low, because it imposes S_f=S_0  and takes no account of the river. These calculations answer different boundary and flow assumptions.

For this drain, the practical reading is as follows. Suppose a 10 mm level error is acceptable. [Equation (36)](#eq-channel-small-flow-limit) gives Q\le\sqrt{0.010/0.0103668}\approx0.98\;\text{m}^3/\text{s} . Up to about 1 m³/s, we may take the drain water level as the river level, RL 10.900 m, throughout. At 4 m³/s the upstream surface is about 155 mm above the river, so we must calculate the backwater profile with [Equation (26)](#eq-channel-gvf) or the standard step.

## How the full GVF values were calculated

The full gradually varied flow (GVF) calculation uses [Equation (26)](#eq-channel-gvf) with the current depth at every position. It is nonlinear because changing depth changes A , R , V  and Fr . The leading approximation holds those geometric quantities at their resting values in the correction coefficient; the full calculation updates them as it follows the profile.

We start at the river and work upstream. Let s=300-x  be distance upstream from the outlet, in metres, as marked in [Figure 20](#fig-channel-perturbation-values). Thus s=0  at the river, s=300  at the upstream end, and dx/ds=-1 . Substituting B=3 , S_0=0.001  and the finite-width hydraulic radius into [Equation (26)](#eq-channel-gvf) gives

 dyds=f(y)=Sf(y)−0.0011−Q2/[9.81(3)2y3],Sf(y)=[0.015Q(3y)R(y)2/3]2,R(y)=3y3+2y,y(s=0)=1.200  m.\begin{aligned} \frac{dy}{ds}=f(y) &=\frac{S_f(y)-0.001}{1-Q^2/[9.81(3)^2y^3]},\\ S_f(y)&=\left[\frac{0.015Q}{(3y)R(y)^{2/3}}\right]^2, \qquad R(y)=\frac{3y}{3+2y},\\ y(s=0)&=1.200\;\text{m}. \end{aligned}(38)

Here f(y)  is the upstream depth gradient, in metres per metre, for one fixed discharge Q . We integrate this initial-value equation in distance with the fourth-order Runge–Kutta method. For distance increment \Delta s=0.25\;\text{m} , let Y_j  be the calculated depth after j  increments, at s_j=j\Delta s . Uppercase Y_j  denotes the numerical depth at a station; lowercase y_0(x)  remains the resting-depth function. The four intermediate gradient estimates k_1  to k_4  and the next depth are

 k1=f(Yj),k2=f(Yj+12Δs k1),k3=f(Yj+12Δs k2),k4=f(Yj+Δs k3),Yj+1=Yj+Δs6(k1+2k2+2k3+k4).\begin{aligned} k_1&=f(Y_j),\\ k_2&=f(Y_j+\tfrac12\Delta s\,k_1),\\ k_3&=f(Y_j+\tfrac12\Delta s\,k_2),\\ k_4&=f(Y_j+\Delta s\,k_3),\\ Y_{j+1}&=Y_j+\frac{\Delta s}{6}(k_1+2k_2+2k_3+k_4). \end{aligned}(39)

These are distance steps along a steady profile, not timesteps in a storm simulation. The intermediate depths allow the gradient to change within each increment. We repeat 1,200 increments to cover the 300 m reach.

For Q=0.500\;\text{m}^3/\text{s} , the first increment starts at Y_0=1.200000\;\text{m} . The first four gradients are approximately -0.000994176535 , -0.000994174923 , -0.000994174923  and -0.000994173310 . Substitution into [Equation (39)](#eq-channel-gvf-distance-step) gives depth 1.199751456\;\text{m}  at s=0.25\;\text{m} . [Table 3](#tbl-channel-gvf-trace) lists the depth at selected distances along the reach.

**Table 3.** Full GVF profile for Q=0.500\;\text{m}^3/\text{s} , integrated upstream from the river

| Distance upstream s (m) | Position x (m) | Calculated depth (m) | Water-surface RL (m) |
| --- | --- | --- | --- |
| 0 | 300 | 1.200000000 | 10.900000000 |
| 0.25 | 299.75 | 1.199751456 | 10.900001456 |
| 50 | 250 | 1.150308184 | 10.900308184 |
| 150 | 150 | 1.051045564 | 10.901045564 |
| 300 | 0 | 0.902589230 | 10.902589230 |

The bed elevation at distance s  is RL 9.700+0.001s  m, so each water-surface elevation is that bed elevation plus the calculated depth. At the upstream end, the surface rise above the river is therefore

 δzu=(10.000+0.9025892295)−10.900=0.0025892295  m=2.5892295  mm.\begin{aligned} \delta z_u&=(10.000+0.9025892295)-10.900\\ &=0.0025892295\;\text{m}=2.5892295\;\text{mm}. \end{aligned}(40)

That is the 0.50 m³/s entry in the full-GVF column of [Table 2](#tbl-channel-small-flow). Repeating the calculation for each listed discharge gives the other entries. Reducing the distance increment from 0.25 m to 0.125 m changes the upstream depth by less than 10^{-9}\;\text{m}  for these cases. This refinement checks the numerical distance integration; the stated channel geometry, roughness and fixed downstream level remain the physical assumptions.

## MSMA 3 worked example: one standard step and the resolved profile

We finish the steady case with MSMA 3’s own example, Part 2 Appendix 6.5. It solves one standard step over the whole reach, which we can now compare with the full GVF integration of the previous subsection. Its inputs are Q=20\;\text{m}^3/\text{s} , rectangular width 10 m, n=0.015\;\text{s/m}^{1/3} , length 500 m, bed slope 1:2500 and downstream depth 2.500 m. The upstream bed is 0.200 m above the downstream bed. The worksheet’s friction arithmetic uses the wide-channel approximation R\approx y ; we first retain it to compare the same model. Both energy factors are one and only friction loss is included.

Figure 21. MSMA 3 Appendix 6.5 inputs and the resolved steady result

[Figure 21](#fig-channel-msma) labels the common inputs and the integrated wide-channel result. To reproduce the single-step calculation, set the downstream bed elevation to zero and the upstream bed elevation to 0.200 m in [Equation (28)](#eq-channel-standard-step):

 0.200+yu+(2/yu)219.62=2.500+0.800219.62+250[Sf(yu)+Sf(2.500)],Sf(y)=0.0152(2/y)2y4/3.\begin{aligned} 0.200+y_u+\frac{(2/y_u)^2}{19.62} &=2.500+\frac{0.800^2}{19.62}\\ &\quad+250[S_f(y_u)+S_f(2.500)],\\ S_f(y)&=\frac{0.015^2(2/y)^2}{y^{4/3}}. \end{aligned}(41)

The [existing article’s trial rounds](https://mes100.com/docs/tailwater-backwater-from-bernoulli/#msma-3-worked-example-appendix-65) give the balanced subcritical root y_u=2.318949\;\text{m} . MSMA 3 prints 2.73 m; that value does not balance the energy equation with its stated inputs. The existing article explains the source arithmetic and bed-elevation discrepancy. Here we use the balanced equation to compare step sizes, in [Table 4](#tbl-channel-msma-steps).

**Table 4.** MSMA 3 Appendix 6.5: one standard step against finer steps and direct GVF integration

| Calculation with the same wide-channel approximation | Upstream depth (m) |
| --- | --- |
| One 500 m standard step | 2.318949172 |
| Five 100 m steps | 2.318803005 |
| Ten 50 m steps | 2.318798418 |
| Twenty 25 m steps | 2.318797271 |
| One hundred 5 m steps | 2.318796904 |
| Direct GVF integration, 0.25 m increments | 2.318796889 |

The one-step result is about 0.152 mm above the resolved profile. Refinement changes the numerical integration of friction while preserving the hydraulic assumptions. Restoring the finite rectangular hydraulic radius R=10y/(10+2y)  instead gives integrated upstream depth 2.335502\;\text{m} , about 16.7 mm higher than the integrated wide-channel result. That larger difference comes from the wide-channel approximation, which this channel, at B/h\approx4 , does not satisfy; we report it separately from step-size error. The single wide-channel step, refined steps and finite-width calculation all remain subcritical for these inputs. This example demonstrates the steady reduction of our equations; it is not a comparison with a MiTS run.

# Unsteady flow

The third option in MSMA 3 is to run drainage software that solves the full dynamic wave equations, which keep every term of [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum), including the time derivatives. MSMA 3 itself only writes the equations down: §4.5.8.2 states them for channel flow when the inflow terms are negligible and the channel is wide. Setting i=f=0  and q_L=0  in [Equation (14)](#eq-channel-msma-continuity-rect) and [Equation (15)](#eq-channel-msma-momentum-rect), and 2y/b\to0  for a wide channel, gives

 ∂y∂t+V∂y∂x+y∂V∂x=0,∂V∂t+V∂V∂x+g∂y∂x=−τ0ρy+gθ.\frac{\partial y}{\partial t}+V\frac{\partial y}{\partial x}+y\frac{\partial V}{\partial x}=0, \qquad \frac{\partial V}{\partial t}+V\frac{\partial V}{\partial x}+g\frac{\partial y}{\partial x} =-\frac{\tau_0}{\rho y}+g\theta.(42)

These are Eqs. 4.50 and 4.51 of MSMA 3. The printed Eq. 4.51 also contains a stray arrow symbol after V\,\partial V/\partial x , which we read as a typesetting error. For a small stream receiving rain and overland flow along its length, MSMA 3 keeps the supply in continuity:

 ∂A∂t+∂Q∂x=(i−f)(b+2δy)+2qL(Eq. 4.52).\frac{\partial A}{\partial t}+\frac{\partial Q}{\partial x}=(i-f)(b+2\delta y)+2q_L\quad\text{(Eq. 4.52)}.

Figure 22. Trapezoidal drain in MSMA 3 Eq. 4.52: surface width and side slope

This is [Equation (2)](#eq-channel-continuity) for the trapezoidal drain of [Figure 22](#fig-channel-msma-supply-trap). Each side slopes \delta  horizontal to 1 vertical, so at depth y  each side adds \delta y  to the bottom width and the surface width is B=b+2\delta y . Rain less infiltration over that surface, plus inflow from both banks, gives q_l=(i-f)B+2q_L . MSMA 3 defines \delta y  only as “slope”; we read \delta  as this side slope.

MSMA 3 gives no hand solution of these equations; its design procedure leaves them to drainage software such as MiTS. In this section we take two cases simple enough to solve by hand and find what the time-dependent terms add to the steady backwater profile.

## Linearised equations for a small disturbance

For steady flow we perturbed the GVF profile about water at rest, with depth y_0=Z_r-z_b  and area A_0 , and found the settled rise \delta z_u\approx C_uQ^2  in [Equation (34)](#eq-channel-small-steady-correction). We now apply the same perturbation to the full continuity and momentum equations, keeping their time derivatives. The base state, resting geometry and resistance law stay the same; what changes is that the disturbance may now vary with time. The time derivatives describe water going into storage and the acceleration needed to start or change the flow. We will find that they are first order in the disturbance, whereas friction and convective acceleration, which produced C_uQ^2 , are second order. The leading equations are therefore linear.

### Assumptions and reduction from the governing equations

[Figure 23](#fig-channel-transient-setup) shows the resting and disturbed surfaces at the same position, and the small strip of area added when the surface rises.

Figure 23. A changing surface and discharge measured from the resting state

We keep the earlier fully wet, hydrostatic channel model, no side supply (q_l=0 ), unit momentum factor (\beta=1 ) and fixed river level. Write \delta z=z_s-Z_r  for the surface departure from rest. Before making a small-amplitude approximation, [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum) give

 B∂δz∂t+∂Q∂x=0,∂Q∂t+gA∂δz∂x=−∂∂x(Q2A)−gASf.\begin{aligned} B\frac{\partial\delta z}{\partial t} +\frac{\partial Q}{\partial x}&=0,\\ \frac{\partial Q}{\partial t} +gA\frac{\partial\delta z}{\partial x} &=-\frac{\partial}{\partial x}\left(\frac{Q^2}{A}\right)-gAS_f. \end{aligned}(43)

Here A  and B  are still evaluated at the current depth. Setting the time derivatives to zero gives constant discharge and the steady momentum balance used to derive [Equation (32)](#eq-channel-steady-surface-change). We now keep those derivatives and ask which terms survive when the changing flow is small. For a small dimensionless amplitude \epsilon , write

 zs=Zr+ϵz^,Q=ϵQ^,0<ϵ≪1,z^(xd,t)=0.z_s=Z_r+\epsilon\hat z,\qquad Q=\epsilon\hat Q, \qquad 0<\epsilon\ll1,\qquad \hat z(x_d,t)=0.

Here \hat z  and \hat Q  are the leading surface and discharge disturbances, in metres and m³/s. On a fixed time scale, local acceleration and the surface-gradient force are first order in \epsilon . Convective momentum flux and signed Manning resistance, proportional to Q|Q| , are second order. Keeping only the first-order terms gives equations that are linear in \hat z  and \hat Q , the linearised equations. This approximation requires disturbances small enough to keep the reach wet and the area change small compared with its resting area.

### First-order continuity and momentum

In [Figure 23](#fig-channel-transient-setup), z_s(x,t)  is the current surface elevation, and \delta z=z_s-Z_r  is its change from the resting river level. The green strip in the cross-section is the area change \delta A=A-A_0 . All changes are evaluated at the same position x  and time t . The fixed boundary requires \delta z=0  where the drain meets the river.

For the resting surface width B_0  marked in the cross-section, the same depth-to-area relation used in the steady approximation gives A=A_0+\epsilon B_0\hat{z}+O(\epsilon^2) . Here O(\epsilon^2)  denotes terms at least quadratic in amplitude. Substituting into the continuity balance in [Equation (43)](#eq-channel-small-unsteady-balance), dividing by \epsilon  and retaining first-order terms gives

 B0∂z^∂t+∂Q^∂x=0.B_0\frac{\partial\hat{z}}{\partial t} +\frac{\partial\hat{Q}}{\partial x}=0.(44)

In the momentum balance of [Equation (43)](#eq-channel-small-unsteady-balance), the two terms on the right are quadratic in amplitude. These are the same convective and resistance terms that produce the earlier steady Q^2  correction. The two terms on the left can be first order when the discharge changes on a fixed time scale. Dividing by \epsilon  and keeping first-order terms gives

 ∂Q^∂t+gA0∂z^∂x=0.\frac{\partial\hat{Q}}{\partial t} +gA_0\frac{\partial\hat{z}}{\partial x}=0.(45)

This equation permits a sloping surface even when the discharge is small: the slope accelerates the water. Small discharge therefore allows us to neglect the quadratic terms at first order, while acceleration can remain.

The steady limit of these equations shows why the steady expansion, [Equation (33)](#eq-channel-small-flow-series), had no linear term. If the time derivatives in [Equation (44)](#eq-channel-linear-continuity) and [Equation (45)](#eq-channel-linear-momentum) vanish, then

 ∂Q^∂x=0,∂z^∂x=0,z^(xd)=0 ⟹ z^(x)=0.\frac{\partial\hat Q}{\partial x}=0,\qquad \frac{\partial\hat z}{\partial x}=0, \qquad \hat z(x_d)=0\ \Longrightarrow\ \hat z(x)=0.

This gives a level surface at first order, which is a_1=0 . To find the smaller steady departure, we must restore the terms that the linearisation discarded. Take a settled forward flow, so \partial Q/\partial t=0  and continuity gives dQ/dx=0 . Divide the momentum balance in [Equation (43)](#eq-channel-small-unsteady-balance) by gA  and substitute Manning resistance:

 dδzdx=−Q2gAddx(1A)−n2Q2A2R4/3.\frac{d\delta z}{dx} =-\frac{Q^2}{gA}\frac{d}{dx}\left(\frac{1}{A}\right) -\frac{n^2Q^2}{A^2R^{4/3}}.(46)

Both remaining terms carry Q^2 . For their leading contribution we use the same resting geometry as in the small-discharge calculation, A\approx A_0  and R\approx R_0 . Since (1/A_0)\,d(1/A_0)/dx=\tfrac12\,d(1/A_0^2)/dx , the equation becomes

 dδzdx≈−Q2[12gddx(1A02)+n2A02R04/3].\frac{d\delta z}{dx}\approx -Q^2\left[ \frac{1}{2g}\frac{d}{dx}\left(\frac{1}{A_0^2}\right) +\frac{n^2}{A_0^2R_0^{4/3}}\right].(47)

Integrate from the upstream section x_u  to the river at x_d . The fixed river gives \delta z(x_d)=0 , so the left-hand integral is 0-\delta z_u . Reversing the sign gives

 δzu≈Q2{12g[1A0,d2−1A0,u2]+∫xuxdn2A02R04/3 dx}=CuQ2.\begin{aligned} \delta z_u&\approx Q^2\left\{ \frac{1}{2g}\left[\frac{1}{A_{0,d}^2}-\frac{1}{A_{0,u}^2}\right] \right.\\ &\qquad\left.{}+\int_{x_u}^{x_d} \frac{n^2}{A_0^2R_0^{4/3}}\,dx\right\} =C_uQ^2. \end{aligned}(48)

The expression in braces is exactly C_u  in [Equation (34)](#eq-channel-small-steady-correction). We have recovered the steady result from the time-dependent balances by first letting the flow settle, then retaining the quadratic terms. A settled flow thus raises the surface only at second order, while a changing flow acts at first order through storage and acceleration. The rest of this section follows the first-order effects; [Equation (58)](#eq-channel-quasi-small-flow-connection) later adds the two together for a slowly rising inflow. Throughout, the river level stays fixed; a changing river level would add its own first-order boundary term.

## Wave speed in a locally uniform reach

To calculate the response to a sudden inflow, we need to know how quickly its disturbance reaches the river and how much the surface rises for the imposed discharge. We derive the wave speed c  and the discharge-to-level relation here. The next subsection uses them to calculate the travel time L/c , follow reflections at both ends and obtain the oscillation period 4L/c  and surface amplitude Q_{in}/(B_0c) .

For a local horizontal prismatic reach with constant A_0  and B_0 , differentiate [Equation (44)](#eq-channel-linear-continuity) with time and [Equation (45)](#eq-channel-linear-momentum) with distance. Eliminating the mixed discharge derivative gives

 ∂2z^∂t2=c2∂2z^∂x2,c=gA0B0=gD0,D0=A0B0.\frac{\partial^2\hat{z}}{\partial t^2} =c^2\frac{\partial^2\hat{z}}{\partial x^2}, \qquad c=\sqrt{\frac{gA_0}{B_0}}=\sqrt{gD_0}, \qquad D_0=\frac{A_0}{B_0}.(49)

The small disturbance travels at speed c  relative to the resting water; D_0  is resting hydraulic depth. This c  is the \sqrt{gD_h}  in the Froude number of [Equation (22)](#eq-channel-energy-derivative). In flowing water the disturbance is also carried downstream at V , which gives the two speeds V\pm c  that decided downstream control in the GVF section.

For one downstream-travelling disturbance with shape F(x-ct) , let \hat{z}=F(x-ct) . Substitution into [Equation (44)](#eq-channel-linear-continuity) gives \hat{Q}=cB_0\hat{z}  when there is no underlying uniform discharge offset. The physical perturbation velocity is \delta V=\epsilon\hat{Q}/A_0  to first order, and \delta z=\epsilon\hat{z} . Therefore

 δz=D0cδV=cgδVfor one downstream-travelling linear disturbance.\delta z=\frac{D_0}{c}\delta V=\frac{c}{g}\delta V \quad\text{for one downstream-travelling linear disturbance}.(50)

The sign reverses for one upstream-travelling disturbance. Where incident and reflected waves overlap, their contributions add; at the fixed river level their surface changes must sum to zero.

Figure 24. A small travelling disturbance over water initially 1 m deep

In [Figure 24](#fig-channel-wave), D_0=1.000\;\text{m}  gives c=3.13209\;\text{m/s} . A downstream-travelling disturbance with \delta V=0.010\;\text{m/s}  has \delta z=0.003193\;\text{m} , or about 3.19 mm. The corresponding velocity head is only \delta V^2/(2g)=0.00000510\;\text{m} , about 0.0051 mm. The wave’s surface displacement is the head that accelerates the water, and it is first order in \delta V ; velocity head is second order. Treating the rise as velocity head would miss almost all of it.

## Response to a sudden inflow

### Assumptions and reduction from the governing equations

[Figure 25](#fig-channel-step-boundaries) marks the imposed inflow, fixed river level and first travelling wave. The river boundary will reflect the wave when it arrives.

Figure 25. A sudden inflow launches a wave towards a fixed river level

We use the linearised equations on a horizontal prismatic reach, so the resting area A_0  and surface width B_0  are constant along its length L . The initial state and boundary conditions for a small positive inflow are

 δz(x,0)=0,Q(x,0)=0(0<x<L),Q(0,t)=Qin(t>0),δz(L,t)=0(t>0).\begin{aligned} \delta z(x,0)&=0,\quad Q(x,0)=0 &&(0<x<L),\\ Q(0,t)&=Q_{in} &&(t>0),\\ \delta z(L,t)&=0 &&(t>0). \end{aligned}

Here \delta z=z_s-Z_r . We idealise the inflow as an instantaneous step, hold it constant afterwards, and require Q_{in}/(B_0c)\ll D_0 , where c=\sqrt{gA_0/B_0}  and D_0=A_0/B_0 . At this order we neglect Manning friction and convective acceleration. The resulting wave solution therefore retains reflections but has no damping; the later numerical comparison restores the nonlinear terms.

### Wave reflections and upstream level

We now follow the waves launched by the inflow as they reflect between the two ends, working directly with the physical changes \delta z  and Q  instead of the scaled hats. [Figure 26](#fig-channel-reflection-stages) shows four snapshots during the first cycle.

Figure 26. Wave reflections during the first sudden-inflow cycle

From [Equation (50)](#eq-channel-travelling-wave-amplitude), a downstream-travelling wave carries Q=cB_0\,\delta z . For an upstream-travelling wave the same substitution into [Equation (44)](#eq-channel-linear-continuity) gives Q=-cB_0\,\delta z . We follow the waves one leg at a time. Write

 h0=QinB0c.h_0=\frac{Q_{in}}{B_0c}.(51)

1. **0<t<L/c .** The upstream boundary launches a downstream wave carrying Q_{in} . Behind its front, \delta z=h_0  and Q=Q_{in} ; ahead of it the water is still at rest.
2. **L/c<t<2L/c .** At the river the level must stay at Z_r , so the wave reflects with \delta z=-h_0 . Travelling upstream, that reflection carries Q=-cB_0(-h_0)=+Q_{in} . Behind it, \delta z=h_0-h_0=0  and Q=2Q_{in} : the drain empties into the river at twice the inflow.
3. **2L/c<t<3L/c .** At the upstream end the discharge must remain Q_{in} , not 2Q_{in} . The boundary therefore launches a downstream wave carrying -Q_{in} , with \delta z=-h_0 . Behind it, \delta z=-h_0  and Q=Q_{in} : the surface is now below the river level.
4. **3L/c<t<4L/c .** The river reflects this wave as an upstream wave with \delta z=+h_0  and Q=-Q_{in} . Behind it, \delta z=0  and Q=0 : the drain is momentarily back at rest, the state at t=0 .

The cycle then repeats. At the upstream end, the level is a square wave:

 δzu(t)={+h0,0<t<2L/c,−h0,2L/c<t<4L/c,Tosc=4Lc.\delta z_u(t)= \begin{cases} +h_0, & 0<t<2L/c,\\ -h_0, & 2L/c<t<4L/c, \end{cases} \qquad T_{osc}=\frac{4L}{c}.(52)

The outlet discharge alternates between 0  and 2Q_{in}  in the same rhythm. Averaged over one period T_{osc} , the level change is zero and the outlet discharge is Q_{in} : the steady answer at this order. The drain behaves like an organ pipe that is open at the river end, where the level is fixed, and closed at the upstream end, where the discharge is fixed. Such a pipe resonates with a period of four travel times.

For a 300 m drain, 3 m wide, with resting depth 1.000 m and n=0.015 , a sudden inflow of 0.3 m³/s gives c=3.132\;\text{m/s} , h_0=0.3/(3\times3.132)=0.0319\;\text{m}  and T_{osc}=4(300)/3.132=383\;\text{s} , about 6.4 minutes. By comparison, the steady rise for the same flow from [Equation (35)](#eq-channel-horizontal-perturbation) is only 1.3 mm. A sudden start lifts the upstream level about 24 times as much as the flow eventually needs.

Figure 27. Upstream level after a sudden 0.3 m³/s inflow: linear square wave and full-equation solution

[Figure 27](#fig-channel-step-response) compares [Equation (52)](#eq-channel-step-response) with a numerical solution of the full equations, [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum), including Manning friction and the convective term. The first plateau is 31.8 mm against the linear 31.9 mm, and the period is 384 s against 383 s. The square wave then decays: successive rises are 31.8, 23.8 and 18.0 mm. Friction is one cause. The other is that a 32 mm wave on 1 m of water is not quite linear: its fronts steepen, travel at slightly different speeds and lose energy. For a step ten times smaller the decay falls to about 3% per period. Linear theory therefore gives the amplitude and period of the response, but not how quickly it dies away.

## Slowly rising inflow: storage and acceleration terms

In a storm the inflow rarely starts suddenly; it rises over tens of minutes. If it changes slowly enough, the drain has time to adjust, and at each moment its profile is close to the steady backwater profile for the current inflow. This is a different perturbation from the linearised one. The base state is no longer water at rest but the steady GVF profile for the current inflow, and the small quantity is no longer the size of the disturbance but its rate of change. The discharge itself need not be small, so the friction and convective terms stay in full. We now make that idea precise and find the first correction.

Take the drain of the small-discharge numerical check, with no side inflow and \beta=1 . Since the bed is fixed, \partial A/\partial t=B\,\partial z_s/\partial t , and [Equation (2)](#eq-channel-continuity) becomes

 ∂Q∂x=−B∂zs∂t.\frac{\partial Q}{\partial x}=-B\frac{\partial z_s}{\partial t}.(53)

Dividing [Equation (9)](#eq-channel-momentum) by gA  and using \partial y/\partial x-S_0=\partial z_s/\partial x  gives the surface slope:

 ∂zs∂x=−Sf−1gA∂∂x(Q2A)−1gA∂Q∂t.\frac{\partial z_s}{\partial x}=-S_f-\frac{1}{gA}\frac{\partial}{\partial x}\left(\frac{Q^2}{A}\right)-\frac{1}{gA}\frac{\partial Q}{\partial t}.(54)

The first two terms on the right are the steady terms that produce the GVF profile. The last is the acceleration term. [Equation (53)](#eq-channel-storage-continuity) says that while the surface is rising, some of the arriving water goes into storage, so the discharge falls slightly along the drain.

## Quasi-steady approximation and its first correction

### Assumptions and reduction from the governing equations

[Figure 28](#fig-channel-quasi-assumptions) compares the current steady profile with the extra head needed when dQ_{in}/dt>0 . This correction is proportional to the rate of change of discharge, dQ_{in}/dt , through [Equation (57)](#eq-channel-quasi-steady). A nonzero discharge Q  held constant has dQ_{in}/dt=0 : its steady friction and velocity-head effects are already included in the GVF profile.

Figure 28. A slow rise adds acceleration head to the current steady profile

We retain the no-side-inflow conditions q_l=0  and \beta=1  used in [Equation (53)](#eq-channel-storage-continuity) and [Equation (54)](#eq-channel-surface-slope-unsteady). The river level stays fixed, and the inflow changes slowly enough for a nearby steady subcritical profile to exist. To leading order we neglect storage, then estimate local acceleration from the imposed inflow:

 ∂Q∂x≈0,Q(x,t)≈Qin(t),∂Q∂t≈dQindt.\frac{\partial Q}{\partial x}\approx0,\qquad Q(x,t)\approx Q_{in}(t),\qquad \frac{\partial Q}{\partial t}\approx\frac{dQ_{in}}{dt}.

We evaluate the steady terms and the area in the acceleration integral on that current steady profile. Thus the correction below includes acceleration head while neglecting the discharge variation caused by filling or emptying the reach. The validity subsection gives the acceleration and storage time scales against which we must check the inflow variation. A further small-discharge approximation is needed only when we replace the steady GVF level by Z_r+C_uQ_{in}^2 .

### Upstream level with acceleration head

We now use the steady profile already derived from [Equation (26)](#eq-channel-gvf) as the starting approximation at each time. Set its discharge to the current inflow Q_{in}(t)  and keep the same river boundary. Write that profile as z_s^{GVF}(x;Q_{in}(t)) ; its upstream value is z_u^{GVF}(Q_{in}(t)) . For a small forward discharge, [Equation (48)](#eq-channel-recovered-steady-rise) gives this value directly as Z_r+C_uQ_{in}(t)^2 . For a larger discharge we calculate the steady GVF profile numerically instead.

To find what slow acceleration adds, we substitute the assumptions above into [Equation (54)](#eq-channel-surface-slope-unsteady). The friction and convective terms then give the slope of the current GVF profile, and the acceleration term remains:

 ∂zs∂x≈dzsGVFdx−1gAdQindt.\frac{\partial z_s}{\partial x}\approx \frac{d z_s^{GVF}}{dx} -\frac{1}{gA}\frac{dQ_{in}}{dt}.(55)

Here A  is the area along the current steady profile. Integrating between the two ends makes the connection explicit. Both profiles end at the same river elevation Z_r , and dQ_{in}/dt  is independent of position, so

 Zr−zu(t)≈Zr−zuGVF(Qin(t))−dQindt∫xuxddxgA.\begin{aligned} Z_r-z_u(t)&\approx Z_r-z_u^{GVF}(Q_{in}(t))\\ &\qquad-\frac{dQ_{in}}{dt}\int_{x_u}^{x_d}\frac{dx}{gA}. \end{aligned}(56)

Cancel Z_r  and rearrange:

 zu(t)≈zuGVF(Qin(t))+dQindt∫xuxddxgA.z_u(t)\approx z_u^{GVF}\bigl(Q_{in}(t)\bigr)+\frac{dQ_{in}}{dt}\int_{x_u}^{x_d}\frac{dx}{gA}.(57)

For the small-discharge case, substitute the earlier result z_u^{GVF}\approx Z_r+C_uQ_{in}^2  and use the resting area A_0  in the acceleration integral:

 zu(t)−Zr≈CuQin(t)2+dQindt∫xuxddxgA0.z_u(t)-Z_r\approx C_uQ_{in}(t)^2 +\frac{dQ_{in}}{dt}\int_{x_u}^{x_d}\frac{dx}{gA_0}.(58)

The two perturbations meet in this equation. The first term is the second-order steady rise from [Equation (34)](#eq-channel-small-steady-correction). The second is the first-order acceleration head of the linearised momentum equation, [Equation (45)](#eq-channel-linear-momentum), integrated along the drain. When the inflow is held constant and the flow has settled, dQ_{in}/dt=0  and we recover z_u-Z_r\approx C_uQ_{in}^2 . Setting the time derivative to zero at one instant during a transient does not remove waves already travelling along the drain; this approximation assumes the slow adjustment described here.

The integral, in s/m², measures the inertia of the water column: the head needed to change its discharge at a given rate. While the inflow rises, the upstream surface stands higher than the steady profile, because part of the head is accelerating the water. While it falls, the surface stands lower.

## Validity of the quasi-steady approximation

The drain adjusts on two time scales. Let \Delta z  be the steady rise above the river and \bar{A}  a typical flowing area.

- **Acceleration time**, T_a\approx QL/(g\bar{A}\Delta z) : the time for the head \Delta z  to accelerate the water column to discharge Q .
- **Storage time**, T_s\approx BL\Delta z/Q : the time for the discharge Q  to fill the extra volume above the river level.

[Equation (57)](#eq-channel-quasi-steady) applies when the inflow changes little over both times. Their product is T_aT_s=BL^2/(g\bar{A})=(L/c)^2 , the square of the wave travel time, so a hydrograph slow compared with both is also slow compared with the waves.

For the 300 m drain of the small-discharge numerical check, at 4 m³/s the steady rise is 0.155 m, giving T_a\approx230\;\text{s}  and T_s\approx35\;\text{s} . At 1 m³/s the rise is only 10.3 mm and T_a\approx940\;\text{s} , about 16 minutes. Because the steady rise grows as Q^2 , the head available to accelerate the water is tiny at small flows. The acceleration term is therefore most important early in a rising hydrograph, when the steady rise is still small.

## Worked example: one-hour rise to 4 m³/s

[Figure 29](#fig-channel-hour-rise-setup) shows the geometry, resting depths and imposed inflow for this calculation.

Figure 29. Geometry and inflow for the one-hour rise in the 300 m drain

Use the same 300 m drain: 3 m wide, n=0.015 , inverts RL 10.000 m and RL 9.700 m, river fixed at RL 10.900 m. The inflow rises smoothly from zero to 4 m³/s over T=60  minutes and is then held:

 Qin(t)=2[1−cos⁡(πtT)] m3/s(0≤t≤T),Qin(t)=4 m3/s(t>T).\begin{aligned} Q_{in}(t)&=2\left[1-\cos\left(\frac{\pi t}{T}\right)\right]\ \text{m}^3/\text{s} && (0\le t\le T),\\ Q_{in}(t)&=4\ \text{m}^3/\text{s} && (t>T). \end{aligned}(59)

At rest the area varies linearly from 2.7 m² to 3.6 m², so

 ∫0300dxgA0=19.81×3×0.001ln⁡1.2000.900=9.78  s/m2.\int_0^{300}\frac{dx}{gA_0}=\frac{1}{9.81\times3\times0.001}\ln\frac{1.200}{0.900}=9.78\;\text{s/m}^2.

For each time we calculate z_u^{GVF}  from [Equation (26)](#eq-channel-gvf), as in the small-flow numerical check, and add the acceleration term using the integral along the current steady profile. A closed form needs no profile calculation at all. It uses [Equation (58)](#eq-channel-quasi-small-flow-connection), with the small-flow result for the steady part and the resting-area integral for the acceleration part:

 zu(t)≈Zr+CuQin(t)2+9.78 dQindt,Cu=0.01037  s2/m5.z_u(t)\approx Z_r+C_uQ_{in}(t)^2+9.78\,\frac{dQ_{in}}{dt}, \qquad C_u=0.01037\;\text{s}^2/\text{m}^5.(60)

[Table 5](#tbl-channel-quasi-steady) compares these estimates with a numerical solution of the full equations.

**Table 5.** Upstream level of the 300 m drain during the one-hour rise to 4 m³/s

| Time (min) | Inflow (m³/s) | Closed form (RL m) | Steady GVF level (RL m) | GVF with acceleration term (RL m) | Full equations (RL m) |
| --- | --- | --- | --- | --- | --- |
| 10 | 0.268 | 10.9093 | 10.9007 | 10.9093 | 10.9096 |
| 20 | 1.000 | 10.9251 | 10.9103 | 10.9250 | 10.9239 |
| 30 | 2.000 | 10.9585 | 10.9408 | 10.9576 | 10.9552 |
| 40 | 3.000 | 11.0081 | 10.9899 | 11.0041 | 11.0002 |
| 50 | 3.732 | 11.0529 | 11.0361 | 11.0441 | 11.0412 |
| 60 | 4.000 | 11.0659 | 11.0548 | 11.0548 | 11.0556 |
| 80 | 4.000 | 11.0659 | 11.0548 | 11.0548 | 11.0549 |

[Figure 30](#fig-channel-quasi-steady) plots the steady, corrected and full-equation levels through the rise.

Figure 30. Upstream level during a one-hour rise to 4 m³/s: steady profile, quasi-steady estimate and full equations

During the rise the steady profile alone is up to 15 mm too low. [Equation (57)](#eq-channel-quasi-steady) brings the estimate within 4 mm. The closed form [Equation (60)](#eq-channel-closed-form-example) is within about 3 mm up to 2 m³/s. At 4 m³/s it is about 11 mm high, the same overestimate that the small-flow numerical check found for steady flow, because C_uQ^2  keeps the resting geometry. The remaining difference comes from the neglected storage: water filling the rising surface slightly reduces the discharge further downstream, and with it the friction. At 10 minutes the steady rise is only 0.7 mm, while the actual rise is 9.6 mm; almost all of it is acceleration head. Once the inflow is held, the level settles to the steady backwater value within about a millimetre.

# Hand calculations for the existing MiTS fixed-outfall results

The [MiTS fixed-outfall article](https://mes100.com/docs/fix-water-depth-as-the-end-point-of-drain/) models one drain discharging to a river held at a fixed level. Its inputs are listed in [Table 6](#tbl-channel-mits-inputs).

**Table 6.** Inputs of the MiTS fixed-outfall drain

| Quantity | Value |
| --- | --- |
| Drain section | Rectangular, 1.2 m wide, 0.9 m high |
| Length and Manning roughness | 100 m; n=0.013\;\text{s/m}^{1/3} |
| Inverts | 29.00 m upstream, 28.00 m downstream, so S_0=0.01 |
| Fixed outfall (river) level | 28.50 m |

The linked article reports an existing MiTS run with Dynamic Wave routing, which solves the time-dependent equations, [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum). [Table 7](#tbl-channel-mits-agreement) compares our hand calculations with its plotted link depths. Both calculations agree with the values shown to two decimal places. The subsections that follow show how we obtain them.

**Table 7.** Manual mean end depths compared with the existing MiTS link-depth plots

| Flow state | Manual end-depth calculation (m) | Mean end depth (m) | MiTS plotted link depth (m) | Numerical comparison |
| --- | --- | --- | --- | --- |
| At rest after the storm | Upstream 0; downstream 0.500 | 0.250 | About 0.25 | Same value at the displayed precision |
| Steady peak, 1.5 m³/s | Calculated normal depth 0.415; downstream 0.500 | 0.457 | About 0.46 | Approximately 0.003 m apart; both round to 0.46 m |

We interpret the link depth as the arithmetic mean of the two end depths. The linked article plots only that link depth, not the two node depths, so 0.415 m is our calculated normal depth rather than a software result. Since the software values are read approximately from plots, the 3 mm difference is not a measured simulation error. No new MiTS run or reproduction of the changing-flow curves is reported here.

## Phase 1: level surface after the flow stops

After about 90 minutes the flow is nearly zero. The surface is level at the river elevation, so [Equation (30)](#eq-channel-rest-depth) applies:

- Outlet: y_d=28.50-28.00=0.50\;\text{m} .
- Upstream end: the invert, 29.00 m, is above the river, so the bed is dry and y_u=0 .
- Mean of the end depths: (0+0.50)/2=0.25\;\text{m} . MiTS plots about 0.25 m.

The mean is a reporting quantity, not a depth found along the drain. Halfway along, the invert is at 28.50 m, exactly where the water meets the bed, so the local depth there is zero.

## Phase 2: normal depth at peak flow

From 30 to 60 minutes the discharge is steady at about Q=1.5\;\text{m}^3/\text{s}  and MiTS plots a link depth of about 0.46 m. We need two depths for this discharge.

Critical depth from [Equation (23)](#eq-channel-critical):

 yc=(1.529.81×1.22)1/3=0.542  m.y_c=\left(\frac{1.5^2}{9.81\times1.2^2}\right)^{1/3}=0.542\;\text{m}.

Normal depth from [Equation (17)](#eq-channel-normal), with A=1.2y  and R=1.2y/(1.2+2y) , solved by trial:

 1.5=10.013(1.2yn)(1.2yn1.2+2yn)2/30.01⇒yn=0.415  m.1.5=\frac{1}{0.013}(1.2y_n)\left(\frac{1.2y_n}{1.2+2y_n}\right)^{2/3}\sqrt{0.01} \quad\Rightarrow\quad y_n=0.415\;\text{m}.

Normal depth is below critical depth, so at this flow the drain is steep and the water runs supercritical, at about 3.0 m/s with Froude number 1.49. Supercritical flow is controlled from upstream: a disturbance cannot travel against it. The river can only reach into the drain through a hydraulic jump, and from the momentum balance a jump from 0.415 m needs a downstream depth of at least 0.693 m. The river gives only 0.50 m. The drain therefore flows at about normal depth all the way to the outlet, and the fixed river level does not raise it.

Figure 31. At 1.5 m³/s the MiTS drain runs at normal depth below the river level

[Figure 31](#fig-channel-mits-peak) compares the two depths. With the upstream end at normal depth and the outlet held at the river’s 0.50 m, the mean end depth is

 0.415+0.502=0.457  m,\frac{0.415+0.50}{2}=0.457\;\text{m},

against the plotted link depth of about 0.46 m.

The physical conclusion differs from the one the fixed outfall suggests at first sight. In this drain the river level matters after the storm, when it holds water in the lower half. At peak flow the drain is steep enough that the flow ignores the river. The backwater calculations earlier in this article apply when the drain is mild (y_n>y_c ), or when the river level is high enough to drown the outlet.

## Phase 3: rising flow, falling flow and backflow

Between those phases the discharge changes with time, from zero to 1.5 m³/s and back. MiTS’s second example goes further: a storage empties through an outlet whose invert is at 27.00 m into a river fixed at 28.00 m, and for a time the outlet flow is negative. Water flows from the river back into the storage.

No steady equation reproduces these curves. They need [Equation (2)](#eq-channel-continuity) and [Equation (9)](#eq-channel-momentum) with their time derivatives, stepped forward in time from the starting levels with the inflow hydrograph, as MiTS does. Our analysis explains the mechanism but does not reproduce the curve.

# Conclusion

Starting from conservation of mass and Newton’s second law, we derived the channel-flow balances underlying MSMA 3’s Saint-Venant equations. With the stated assumptions and Manning’s empirical resistance law, we obtained MSMA 3’s normal-depth, gradually varied flow and standard-step equations. Retaining the time derivatives gives the dynamic-wave equations. These calculations share the same physical basis; their different assumptions determine which terms remain.

A fixed river level sets the connected surface at rest. Perturbing about that resting surface, a small steady flow raises it by C_uQ^2 , and the drain level may be taken as the river level while Q\le\sqrt{\Delta/C_u} . The same perturbation applied to the time-dependent equations gives linear waves: a sudden inflow lifts the upstream level by Q_{in}/(B_0c)  and makes it oscillate with period 4L/c . For a slow change, we add head proportional to dQ_{in}/dt  to the steady profile. River control also depends on the flow regime and tailwater, so normal depth alone cannot decide it.

MiTS agrees with the manual calculations based on these MSMA equations in the two fixed-outfall states checked here: the calculated mean end depths of 0.250 m at rest and 0.457 m at the steady peak match the plotted link depths of about 0.25 m and 0.46 m, using the mean-end-depth interpretation. The river-controlled backwater profiles and changing-flow curves remain to be compared with MiTS.

### *Related*

 [![](https://i0.wp.com/mes100.com/wp-content/uploads/2024/10/me-small.jpg?resize=100%2C100&ssl=1)](https://mes100.com/author/admin/)[Ngu Soon Hui](https://mes100.com/author/admin/)

I’m the [Benevolent Dictator for Life](https://en.wikipedia.org/wiki/Benevolent_dictator_for_life) for MiTS Software cum Editor of this website. Read more [here](/about-ngu-soon-hui/).

You can also contact me at soonhui@mes100.com
