Introduction to Unsteady Flow

1
Introduction to Unsteady Flow
2
Introduction to Unsteady Flow
3
Introduction to Unsteady Flow
4
Introduction to Unsteady Flow
5
 In general, routing techniques may be classified into two categories: hydraulic
routing and hydrologic routing. Hydraulic routing techniques are based on the
solution of the partial differential equations of unsteady open channel flow. These
equations are often referred to as the St. Venant equations or the dynamic wave
equations. Hydrologic routing employs the continuity equation and either an
analytical or an empirical relationship between storage within the reach and
discharge at the outlet as a replacement for the momentum equation.

Introduction to Unsteady Flow

6
 The energy equation is derived for gradually varied flow situations. There are
several instances when the transition from subcritical to supercritical and
supercritical to subcritical flow can occur. These include significant changes
in channel slope, bridge constrictions, drop structures and weirs, and stream
junctions. In some of these instances empirical equations can be used (such
as at drop structures and weirs), while at others it is necessary to apply the
momentum equation in order to obtain a better answer.

The momentum equation is derived from Newton’s second law of motion.

Introduction to Unsteady Flow

7
 Applying Newton's second law of motion to a body of water enclosed by two
cross sections at locations 1 and 2, the expression for the change in
momentum over a unit time can be written as shown above.

Introduction to Unsteady Flow

8
The assumption of a hydrostatic pressure distribution is only valid for slopes less than
1:10. The cos  for a slope of 1:10 (approximately 6 degrees) is equal to 0.995. Because
the slope of ordinary channels is far less than 1:10, the cos  correction for depth can be
set equal to 1.0 (Chow, 1959).

The weight force is calculated as the unit weight of water times the volume. To get the
force along the channel bed it is multiplied by the sin  , which is the bed slope S0.
Estimating the average bed slope for natural channels is very difficult. A bad estimate of
this slope can produce wild estimates of the weight force. HEC-RAS computes a
hydraulic depth at each cross section and then subtracts that from the water surface to
get a mean bed elevation. The mean bed elevations are then used in estimating the
average bed slope between two sections.

Friction losses in the momentum equation are based on boundary friction only. Friction
losses are computed as shear stress times the average wetted perimeter, times the flow
weighted length between the cross sections.

The mass times the acceleration is computed as the discharge times the density of water,
times the change in velocity.

Introduction to Unsteady Flow

9
 Where: Q = Flow (water discharge)
 = velocity distribution coefficient
A = cross-sectional flow area
t = time
x = distance in the flow direction
h = depth
So = bed slope
Sf = friction slope, from Manning’s equation:
n2 v v
Sf 
2.21R 4 / 3

v = flow velocity

Introduction to Unsteady Flow

10
 Where: Q = Flow (water discharge)
 = velocity distribution coefficient
A = cross-sectional flow area
x = distance in the flow direction
h = depth
So = bed slope
Sf = friction slope, from Manning’s equation
n 2v 2
v = flow velocity
Sf 
2.21R 4 / 3

Introduction to Unsteady Flow

11
 Water surface profiles are computed from one cross section to the next by
solving the Energy equation with an iterative procedure called the standard
step method. The Energy equation is written above.

where: Y1 Y2 = depth of water at cross sections

Z1, Z2 = elevation of the main channel inverts

V1, V2 = average velocities (total discharge/ total flow area)

1,  2 = velocity weighting coefficients

g = gravitational acceleration

he = energy head loss

Introduction to Unsteady Flow

12
 To answer the above questions requires additional Information regarding the
purpose of the model. Each river system will have site specific information that
must be considered in order to answer the questions of Steady versus Unsteady flow
and 1D versus 2D model. The following is a partial list of some of the things that I
typically consider when trying to make a modeling approach decision:

Physical description of the river channels, floodplain areas, bridges/culverts, other

hydraulic structures, levees, roads, etc. that the model will be applied to.

What is the typical size, length, and complexity of the systems that these models
will be applied to? Is it a 1 mile, 10, 50, 100, 500, or 1000 mile river system

Will this model be used for Planning type studies, or will it be used for real time
modeling and mapping?

What type of events (hydrology and boundary conditions) will the models be used
to analyze?

What is the typical duration of a flood event on this river system? (1/2 day, 1-day, 3-
days, 1- week, 1-month, or 6-months)

Introduction to Unsteady Flow

13
 Are there unique aspects of the system that will significantly affect the computed
results? Such as: is the river tidally influenced; do wind speed and directions affect
the water surface elevations; is the river affected by floating ice or ice jams; does
there tend to be debris issues during flood, and does the debris tend to pile up at
hydraulic structures; will levee overtopping, breaching, and interior flow routing
need to be addressed, Are there any significant bridges and culverts that will cause
water to backup behind them during significant flood events, etc…?

Introduction to Unsteady Flow

14
 What is the level of accuracy of the terrain data and hydraulic structure data?

What is the general level of accuracy of the hydrology used to drive the models?

What are the required outputs from the model (water surface elevations, water
depths, arrival times, average velocities, detailed velocities in two dimensions at
specific point locations, etc…)?

What is the model purpose and expected level of accuracy required?

Introduction to Unsteady Flow

15
 Steady flow models (1D or 2D) should generally not be used when the following
situations exist in the river system being analyzed (this is not an exhaustive list):

The river is tidally influenced, and the tide has a significant effect on the stage for
the area of interest.

The events being modeled are very dynamic with respect to time (i.e. Dambreak
flood waves; flash floods; river systems in which the peak flow comes up very
quickly, stays high for a very short time, and then recedes quickly).

Flow reversals occur during the event.

Dynamic events such as dam and/or levee overtopping and breaching occur during
the event.

Extremely flat river systems, where gravity is not necessarily the only significant
driving force of the flow.

Introduction to Unsteady Flow

16
Introduction to Unsteady Flow
17
Introduction to Unsteady Flow
18
Introduction to Unsteady Flow
19
Introduction to Unsteady Flow
20
Introduction to Unsteady Flow
21
Introduction to Unsteady Flow
22
 The question of 1D versus 2D hydraulic modeling is a much tougher question than steady versus
unsteady flow. There are definitely some areas where 2D modeling can produce better results than
1D modeling, and there are also situations in which 1D modeling can produce as good as or better
results than 2D models… with less effort and computational requirements. Unfortunately, there is a
very large range of situations that fall into a gray area, and one could list the positive and negative
aspects of both methodologies for specific applications. Here are some areas where I think 2D
modeling can give better results than 1D modeling:
When modeling an area behind a leveed system, and the levee will be overtopped and/or breached,
the water can go in many directions. If that interior area has a slope to it, water will travel overland
in potentially many directions before it finds its way to the lowest point of the protected area, and
then it will begin to pond and potentially overtop and/or breach the levee on the lower end of the
system. However, if a protected area is small, and ultimately the whole area will fill to a level pool,
then 1D model is fine for predicting the final water surface and extent of the inundation.
Bays and estuaries in which the flow will continuously go in multiple directions due to tidal
fluctuations and river flows coming into the bay/estuary at multiple locations and times.
Highly braided streams
Alluvial fans – however, this is very debatable that any numerical model can capture a flood event
accurately on an alluvial fan, due to the episodic nature of flow evolutions that can change the whole
direction of the channels during the event.
Flow around abrupt bends in which a significant amount of super elevation will occur during the
event.
Very wide and flat flood plains, such that when the flows goes out into the overbank area, the water
will take multiple flow paths and have varying water surface elevations and velocities in multiple
directions.
Applications where it is very important to obtain detailed velocities for the hydraulics of flow around
an object, such as a bridge abutment or bridge piers, etc…

Introduction to Unsteady Flow

23
Introduction to Unsteady Flow
24
Introduction to Unsteady Flow
25
Introduction to Unsteady Flow
26
 The following are areas in which I think 1D modeling will produce as good as or better
results than 2D modeling for real time flood forecasting applications, with less effort (both
from a model development, calibration, and application viewpoint, as well as a
computational time viewpoint):
Rivers and floodplains in which the dominant flow directions and forces follow the general
river flow path. This covers a lot of river systems in my opinion, but it is obviously
debatable as to the significance that lateral and vertical velocities and forces impact the
computed water surface elevations and the resulting flood inundation boundary.
Steep streams that are highly gravity driven and have small overbank areas.
River systems that contain a lot of bridges/culvert crossings, weirs, dams and other gated
structures, levees, pump stations, etc…. and these structures impact the computed stages and
flows within the river system. I have not seen any 2D model yet that has a comprehensive
set of hydraulic structure modules/capabilities that can handle the full range of hydraulic
flow situations that can come up on many of our river systems. This is an area that the
current state of the art in 1D models is far ahead of the 2D models. This statement does not
mean that these capabilities cannot be incorporated into a 2D model, It just means that I
have not seen a widely used 2D model that has such a comprehensive set of capabilities.
Medium to large river systems, where we are modeling a large portion of the system (50 or
more miles), and it is necessary to run longer time period forecasts (i.e. 2 week to 6 month
forecasts). Even with the tremendous advancements in multi-processor computing, and
GPU (Graphics Processor Units) computing, there are still significant spatial and simulation
time limitations on what we can effectively use 2D models for in the real time forecasting
domain. This will obviously be changing over time.
Areas in which the basic data does not support the potential gain of using a 2D model. If
you do not have detailed overbank and channel bathymetry, or you only have detailed cross
sections at representative distances apart, many of the benefits of the 2D model will not be
realized due to the poor accuracy of the terrain data.

Introduction to Unsteady Flow

27
Introduction to Unsteady Flow
28
Introduction to Unsteady Flow
29
Introduction to Unsteady Flow
30
 There is also a difference in our institutional experience with steady vs. unsteady
model applications. We have a much larger experience, knowledge and data base
with use of the steady flow approach. Keep in mind that finely tuned steady flow
data sets may not be optimal for unsteady flow modeling of the same physical
setting. Indeed, some of the hydraulics that were folded into that fine tuning may
now be modeled directly with the unsteady flow equations.

Introduction to Unsteady Flow

31
 For steady flow the discharges are usually peak flows derived from statistical
analysis. The development and routing of these flows throughout the system is done
externally to the steady flow model. When using unsteady flow, developing a
complete hydrologic scenario becomes an important part of the application.

Introduction to Unsteady Flow

32
Introduction to Unsteady Flow
33
Introduction to Unsteady Flow
34
Introduction to Unsteady Flow
35
 Step-backwater programs compute the cross section properties exactly from the GR
points for every trial water surface elevation. For computational efficiency, all
unsteady flow models represent cross sections as tables of hydraulic properties vs.
elevation that are pre-processed from the GR data. In RAS (UNET) these tables are
comprised of elevation, conveyance, and storage. The hydraulic values used at any
time at any cross section during the solution are interpolated from the table values.
Rating curves for hydraulic structures have always been pre-computed in HEC-
RAS; what is new in unsteady RAS is the development of these tables for cross
sections. The structure of these tables affects both the accuracy and stability of the
unsteady flow solution.

Introduction to Unsteady Flow

36
 “Theoretically speaking, however, the two equations not only use different velocity
distribution coefficients, although these are nearly equal, but also involve different
meanings of the frictional losses. In the energy equation, the item hf measures the
internal energy dissipated in the whole mass of water in the reach, whereas the item
hf in the momentum equation measures the losses due to external forces exerted on
the water by the walls of the channel. …. The inherent distinction between the two
principles lies in the fact that energy is a scalar quantity whereas momentum is a
vector quantity; also, the energy equation contains a term for internal losses,
whereas the momentum equation contains a term for external resistance.”, Chow,
V.T., Open Channel Hydraulics, McGraw-Hill, 1959, p. 51.
Use and interpretation of Manning’s n as a boundary roughness coefficient is,
therefore, consistent with the momentum formulation.

Introduction to Unsteady Flow

37
Introduction to Unsteady Flow
38
Introduction to Unsteady Flow
39
 In this rather extreme example – cross sections spaced at x = 5000 m in a prismatic
trapezoidal channel with bed slope, So, of 0.001; bottom width of 20 m; top width of
60 m; Manning’s n is 0.033; and the overall length is 100,000 m. – the average
conveyance method shows instabilities, particularly on the falling limb of the
hydrograph. This cross section is near the downstream end of the reach, but
upstream of the influence of the downstream boundary condition. The average
friction slope method solution remains stable and accurate. In most circumstances,
however, this will not be an issue.

Introduction to Unsteady Flow

40
Introduction to Unsteady Flow
41
Introduction to Unsteady Flow
42
 The momentum approach integrates forces acting over the surfaces and ends of a
control volume; therefore, impacts of flow expansions/contractions are captured in
the forces on the upstream and downstream ends of that control volume. Of course,
proper selection of the flow areas needs to be done here. The energy approach
integrates work/energy for the control volume; so, these coefficients are needed to
describe the losses associated with the turbulent energy expenditure associated with
flow expansion/contraction. It was found that using a model calibrated for steady
flow with HEC-RAS unsteady flow can result in lower computed water surfaces due
to missing the complete losses from expansion/contraction turbulence. RAS 4.1 and
later releases have allowed for the use of expansion/contraction losses in unsteady
flow.

Introduction to Unsteady Flow

43
Introduction to Unsteady Flow
44
Introduction to Unsteady Flow
45
Introduction to Unsteady Flow
46
Introduction to Unsteady Flow
47
 In this hypothetical circumstance, if the observed outflow hydrograph looked like
#1, calibration via adjustments to Manning’s n and perhaps adjustment of the cross
section properties would work for the discharge, and associated stage, hydrograph.
If we have circumstance #2 however, we need to look first at the hydrology in the
reach – i.e. ungaged lateral flows (tributary and distributed).

Introduction to Unsteady Flow

48
Introduction to Unsteady Flow
49
Introduction to Unsteady Flow
50
Introduction to Unsteady Flow
51
Introduction to Unsteady Flow
52