Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
Ramírez
1
CHAPTER 11
1
PREDICTION AND MODELING OF FLOOD HYDROLOGY AND HYDRAULICS
JORGE A. RAMÍREZ
Water Resources, Hydrologic and Environmental Sciences
Civil Engineering Department
Colorado State University
Fort Collins, Colorado 805231372, USA
ABSTRACT. The basic principles underlying the most commonly used physicallybased
models of the rainfallrunoff transformation process are reviewed. A thorough knowledge
of these principles is a prerequisite for flood hazard studies and, thus, this chapter
reviews several physicallybased methods to determine flood discharges, flow depths, and
other flood characteristics. The chapter starts with a thorough review of linear system
theory applied to the solution of hydrologic flood routing problems in a spatially
aggregated manner  Unit Hydrograph approaches. The chapter then proceeds to a review
of distributed flood routing approaches, in particular the kinematic wave and dynamic
wave approaches. The chapter concludes with a brief discussion about distributed
watershed models, including single event models in which flow characteristics are
estimated only during the flood, and continuous event models in which flow
characteristics are determined continuously during wet periods and dry periods.
INTRODUCTION
Flood prediction and modeling refer to the processes of transformation of rainfall
into a flood hydrograph and to the translation of that hydrograph throughout a watershed
or any other hydrologic system. Flood prediction and modeling generally involve
approximate descriptions of the rainfallrunoff transformation processes. These
descriptions are based on either empirical, or physicallybased, or combined conceptual
physicallybased descriptions of the physical processes involved. Although, in general,
the conceptualizations may neglect or simplify some of the underlying hydrologic
transport processes, the resulting models are quite useful in practice because they are
simple and provide adequate estimates of flood hydrographs.
In modeling single floods, the effects of evapotranspiration, as well as the
interaction between the aquifer and the streams, are ignored. Evapotranspiration may be
ignored because its magnitude during the time period in which the flood develops is
negligible when compared to other fluxes such as infiltration. Likewise, the effect of the
streamaquifer interaction is generally ignored because the response time of the
subsurface soil system is much longer than the response time of the surface or direct
runoff process. In addition, effects of other hydrologic processes such as interception and
depression storage are also neglected. Eventbased modeling generally involves the
following aspects:
1
Ramírez, J. A., 2000: Prediction and Modeling of Flood Hydrology and Hydraulics. Chapter 11 of Inland Flood Hazards: Human,
Riparian and Aquatic Communities Eds. Ellen Wohl; Cambridge University Press.
Jorge A. Ramírez
2
a) evaluation of the rainfall flux over the watershed I(x, t) as a function of space and
time;
b) evaluation of the rainfall excess or effective rainfall flux as a function of space and
time, I
e
(x, t). Effective rainfall is the rainfall available for runoff after infiltration and
other abstractions have been accounted for; and
c) routing of the rainfall excess to the watershed outlet in order to determine the
corresponding flood hydrograph, Q(t).
Hydrologic flood prediction models may be categorized into physical models and
mathematical models. Mathematical models describe the system behavior in terms of
mathematical equations representing the relationships between system state, input and
output. Mathematical models can, in turn, be categorized as either purely conceptual
models or physicallybased models. Depending on whether the functions relating input,
output and system state are functions of space and time, these models may be further
categorized as lumped models or distributed models. Lumped models do not account
explicitly for the spatial variability of hydrologic processes, whereas distributed models
do. Lumped models use averages to represent spatially distributed function and
properties.
UNIT HYDROGRAPH ANALYSIS
Sherman (1932) first proposed the unit hydrograph concept. The Unit Hydrograph
(UH) of a watershed is defined as the direct runoff hydrograph resulting from a unit
volume of excess rainfall of constant intensity and uniformly distributed over the drainage
area. The duration of the unit volume of excess or effective rainfall, sometimes referred to
as the effective duration, defines and labels the particular unit hydrograph. The unit
volume is usually considered to be associated with 1 cm (1 inch) of effective rainfall
distributed uniformly over the basin area.
The fundamental assumptions implicit in the use of unit hydrographs for modeling
hydrologic systems are:
a) Watersheds respond as linear systems. On the one hand, this implies that the
proportionality principle applies so that effective rainfall intensities (volumes) of different
magnitude produce watershed responses that are scaled accordingly. On the other hand, it
implies that the superposition principle applies so that responses of several different
storms can be superimposed to obtain the composite response of the catchment.
b) The effective rainfall intensity is uniformly distributed over the entire river basin.
c) The rainfall excess is of constant intensity throughout the rainfall duration.
d) The duration of the direct runoff hydrograph, that is, its time base, is independent of
the effective rainfall intensity and depends only on the effective rainfall duration.
Jorge A. Ramírez
3
When the effective rainfall is given as a hyetograph, that is, as a sequence of M
rainfall pulses of the same duration ∆t, the corresponding direct runoff hydrograph can be
expressed as the discrete convolution of the rainfall hyetograph and the Unit Hydrograph
as,
Q
n
= P
m
U
n −m+ 1
m=1
m
*
∑
, m
*
= min(n, M) (1a)
Q
n
= Q(n∆t ), n = 1,L, N (1b)
P
m
= I
e
(τ )dτ ,
(m−1)∆t
m∆t
∫
m = 1,L, M (1c)
where P
m
is the volume of the m
th
effective rainfall pulse, Q
n
is the direct runoff, and U
n
m+1
are the Unit Hydrograph ordinates.
Although the above assumptions lead to acceptable results, watersheds are indeed
nonlinear systems. For example, unit hydrographs derived from different rainfallrunoff
events, under the assumption of linearity, are usually different, thereby invalidating the
linearity assumption.
The determination of unit hydrographs for particular basins can be carried out
either using the theoretical developments of linear system theory; or using empirical
techniques. For either case, simultaneous observations of both precipitation and
streamflow must be available. These two approaches are presented in more detail in later
sections.
Hydrograph Components
Total streamflow during a precipitation event includes the baseflow existing in the
basin prior to the storm and the runoff due to the given storm precipitation. Total
streamflow hydrographs are usually conceptualized as being composed of:
a) Direct Runoff, which is composed of contributions from surface runoff and quick
interflow. Unit hydrograph analysis refers only to direct runoff.
b) Baseflow, which is composed of contributions from delayed interflow and
groundwater runoff.
Surface runoff includes all overland flow as well as all precipitation falling
directly onto stream channels. Surface runoff is the main contributor to the peak
discharge.
Interflow is the portion of the streamflow contributed by infiltrated water that
moves laterally in the subsurface until it reaches a channel. Interflow is a slower process
Jorge A. Ramírez
4
than surface runoff. Components of interflow are quick interflow, which contributes to
direct runoff, and delayed interflow, which contributes to baseflow (e.g., Chow, 1964.)
Groundwater runoff is the flow component contributed to the channel by
groundwater. This process is extremely slow as compared to surface runoff.
Schematically in Figure 11.1, the streamflow hydrograph is subdivided into a)
Rising Limb: rising portion of the hydrograph, composed mostly of surface runoff. b)
Crest: zone of the hydrograph around peak discharge. c) Falling (or Recession) Limb:
Portion of the hydrograph after the peak discharge, composed mostly of water released
from storage in the basin. The lower part of this recession corresponds to groundwater
flow contributions.
The main factors affecting hydrograph shape are:
1) Drainage characteristics: basin area, basin shape, basin slope, soil type and land use,
drainage density, and drainage network topology. Most changes in land use tend to
increase the amount of runoff for a given storm (e.g., Chow et al., 1988; Singh, 1989;
Bras, .1990).
2) Rainfall characteristics: rainfall intensity, duration, and their spatial and temporal
distribution; and storm motion, as storms moving in the general downstream direction
tend to produce larger peak flows than storms moving upstream (e.g., Chow et al.,
1988; Singh, 1989; Bras, .1990).
Hydrographs are also described in terms of the following time characteristics (see
Figure 11.1):
Time to Peak, t
p
: Time from the beginning of the rising limb to the occurrence of the peak
discharge.
The time to peak is largely determined by drainage characteristics such as
drainage density, slope, channel roughness, and soil infiltration characteristics. Rainfall
distribution in space also affects the time to peak.
Time of Concentration, t
c
: Time required for water to travel from the most hydraulically
remote point in the basin to the basin outlet. For rainfall events of very long duration, the
time of concentration is associated with the time required for the system to achieve the
maximum or equilibrium discharge. Kibler (1982) and Chow et al. (1988) summarize
several of many empirical and physicallybased equations for t
c
that have been developed.
The drainage characteristics of length and slope, together with the hydraulic
characteristics of the flow paths, determine the time of concentration.
Lag Time, t
l
: Time between the center of mass of the effective rainfall hyetograph and the
center of mass of the direct runoff hydrograph.
Jorge A. Ramírez
5
The basin lag is an important concept in linear modeling of basin response. The
lag time is a parameter that appears often in theoretical and conceptual models of basin
behavior. However, it is sometimes difficult to measure in real world situations. Many
empirical equations have been proposed in the literature. The simplest of these equations
computes the basin lag as a power function of the basin area.
Time Base, t
b
: Duration of the direct runoff hydrograph.
Baseflow separation.
As the Unit Hydrograph concept applies only to direct runoff, the direct runoff
must be separated from the baseflow. Baseflow separation or hydrograph analysis is the
process of separating the direct runoff (surface runoff and quick interflow) from the
baseflow. This separation is somewhat arbitrary, but corresponds to theoretical concepts
of basin response.
Subjective methods. Several subjective methods are shown in Figure 11.1. The
simplest one consists in arbitrarily selecting the discharge marking the beginning of the
rising limb as the value of the baseflow and assuming that this baseflow discharge
remains constant throughout the storm duration. A second method consists in arbitrarily
selecting the beginning of the groundwater recession on the falling limb of the
hydrograph (usually assumed to occur at a theoretical inflection point) and connecting
this point by a straight line to the beginning of the rising limb. A third example of
subjective methods consists in extending the recession prior to the storm by a line from
the beginning of the rising limb to a point directly beneath the peak discharge and then
connecting this point to the beginning of the groundwater recession on the falling limb.
Area method. The area method of baseflow separation consists in determining the
beginning of the baseflow on the falling limb with the following empirical equation,
2 . 0
bA N = (2)
relating the time in days from the peak discharge, N, to the basin area, A. When A is in
square miles, b equals 1. When A is in square kilometers, b equals 0.8. This equation is
unsuitable for smaller watersheds and should be checked for a number of hydrographs
before using.
The master recession curve method. This method consists in modeling the
response of the groundwater aquifer as a linear reservoir of parameter k. This assumption
leads to the following equation for the groundwater recession hydrograph,
Q(t) = Q(t
o
)e
−(t −t
o
) / k
(3)
where Q(t) is the baseflow at time t; Q(t
o
) is a reference baseflow discharge at time t
o
, and
k is the recession constant for baseflow. This method is based on a linear reservoir model
of unforced basin response (that is, response from storage) and it can be used to separate
Jorge A. Ramírez
6
the contributions to the recession flow from surface storage, subsurface storage, and
groundwater aquifer storage. It involves the determination of several recession constants.
t
l
t
p
t
b
Effective precipitation
Streamflow hydrograph
Baseflow hydrographs
Figure 11.1: Schematic Description of Hydrograph
UNIT HYDROGRAPHS: EMPIRICAL DERIVATION
The following are essential steps in deriving a unit hydrograph from a single
storm:
1) Separate the baseflow and obtain the direct runoff hydrograph (DRH).
2) Compute the total volume of direct runoff and convert this volume into
equivalent depth of effective rainfall (in centimeters or in inches) over the entire basin.
3) Normalize the direct runoff hydrograph by dividing each ordinate by the
equivalent volume (in or cm) of direct runoff (or effective rainfall).
Jorge A. Ramírez
7
4) Determine the effective duration of excess rainfall. To do this, obtain the
effective rainfall hyetograph (e.g., use the φindex, the Horton, Green and Ampt, or Philip
equations, or some other method to determine infiltration losses) and its associated
duration. This duration is the duration associated with the unit hydrograph.
Unit hydrographs are fundamentally linked to the duration of the effective rainfall
event producing them. They can only be used to predict direct runoff from storms of the
same duration as that associated with the UH, or from storms which can be described as a
sequence of pulses, each of the same duration as that associated with the UH (see
Equation 1).
Unit Hydrographs for Different Effective Duration
A unit hydrograph for a particular watershed is developed for a specific duration
of effective rainfall. When dealing with a rainfall of different duration a new unit
hydrograph must be derived for the new duration. The linearity property implicit in the
UH analysis can be used to generate UH’s associated with larger or smaller effective
rainfall pulse duration. This procedure is sometimes referred to as the Scurve
Hydrograph method.
SCurve Hydrograph Method
An Shydrograph represents the response of the basin to an effective rainfall event
of infinite duration. Assume that an UH of duration D is known and that an UH for the
same basin but of duration D’ is desired. The first step is to determine the Scurve
hydrograph by adding a series of (known) UH’s of duration D, each lagged by a time
interval D. The resulting superposition represents the runoff resulting from a continuous
rainfall excess of intensity 1/D.
Lagging the Scurve in time by an amount D’ and subtracting its ordinates from
the original unmodified Scurve yields a hydrograph corresponding to a rainfall event of
intensity 1/D and of duration D’. Consequently, to convert this hydrograph whose volume
is D’/D into a unit hydrograph of duration D’, its ordinates must be normalized by
multiplying them by D/D’. The resulting ordinates represent a unit hydrograph associated
with an effective rainfall of duration D’.
UNIT HYDROGRAPHS: LINEAR SYSTEM THEORY
A hydrologic system (a basin) is said to be a linear system if the relationship
between storage, inflow, and outflow is such that it leads to a linear differential equation.
The hydrologic response of such systems can be expressed in terms of an impulse
response function (IRF) through a socalled Convolution Equation. Linear systems
possess the properties of additivity and proportionality, which are implicit in the
convolution equation. Linear reservoirs, for example, are special cases of a general
hydrologic system model in which the storage is linearly related to the output by a
constant k.
Impulse Response Function
Jorge A. Ramírez
8
The IRF of a linear system represents the response of the system to an
instantaneous impulse (excitation) of unit volume applied at the origin in time (t=0). The
response of continuous linear systems can be expressed, in the time domain, in terms of
the impulse response function via the convolution integral as follows,
Q(t) = I
e
(τ )u(t −τ )dτ
0
t
∫
(4)
where u(t) is the impulse response function of the system.
In hydrology, it has been customary to assume that watersheds behave as linear
systems. When dealing with hydrologic systems, u(t) represents the instantaneous unit
hydrograph (IUH), and Q(t) and I
e
(t) represent direct runoff and excess or effective
precipitation, respectively. Thus, an Instantaneous Unit Hydrograph represents the
response of a watershed (discharge at its outlet as a function of time) to a unit volume of
precipitation uniformly distributed over the basin and occurring instantaneously at time t
= 0.
Unit Step Response Function
The unit step response function (SRF) is the theoretical counterpart to the Scurve
hydrograph concept presented earlier in the empirical UH analysis section. It represents
the runoff hydrograph from a continuous effective rainfall of unit intensity. As can be
seen from its definition, it is the convolution of 1 and u(t), and obtained as,
∫
=
t
dt t u t g
0
) ( ) ( (5)
Unit Pulse Response Function
The unit pulse response function (PRF) is the theoretical counterpart to the UH
concept presented earlier. It represents the runoff hydrograph from a constant effective
rainfall of intensity 1/∆t and of duration ∆t.
h(t) =
1
∆t
[g(t) − g(t − ∆t)] =
1
∆t
u(τ )dτ
t −∆t
t
∫
(6)
From its definition, the PRF can be seen as the normalized difference between two lagged
SRF’s (Scurve hydrographs), lagged by an amount ∆t. This is analogous to the procedure
presented earlier in the section on Shydrograph analysis.
Discrete Convolution Equation
When the effective rainfall is given as a hyetograph, that is, as a sequence of
rainfall pulses of the same duration ∆t, the corresponding direct runoff hydrograph can be
expressed as the discrete convolution of the rainfall hyetograph and a Unit Hydrograph,
Jorge A. Ramírez
9
Q
n
= P
m
U
n −m+ 1
m=1
m
*
∑
, m
*
= min(n, M) (1a)
Q
n
= Q(n∆t ), n = 1,L, N (1b)
P
m
= I
e
(τ )dτ ,
(m−1)∆t
m∆t
∫
m = 1,L, M (1c)
where P
m
is the volume of the m
th
effective rainfall pulse and the Unit Hydrograph
ordinates are given by,
∫
∆ + −
∆ −
+ −
∆
= ∆ + − =
t m n
t m n
m n
d u
t
t m n h U
) 1 (
) (
1
) (
1
] ) 1 [( τ τ (7)
The UH ordinates correspond to the area under the IUH between two consecutive time
intervals.
The discrete convolution can be expressed alternatively as a matrix equation as,
P [ ]U [ ]= Q [ ] (8a)
P
1
P
2
P
1
P
2
L
P
M
L P
1
P
M
P
2
L
P
M
U
1
U
2
M
U
N−M+1
=
Q
1
Q
2
Q
n
Q
N
(8b)
where M is the number of effective rainfall pulses; NM+1 is the number of UH
ordinates; and N is the number of direct runoff ordinates. Matrix P is of dimensions (N x
(NM+1)). Recasting the discrete convolution equation in this manner allows for an
objective (and optimal) mathematical derivation of the UH ordinates. The estimation of
the Unit Hydrograph from simultaneous observations of effective precipitation (P
m
) and
direct runoff (Q
n
) can be seen as a linear static estimation problem for which a method
like leastsquares, linear programming, or others can be used.
The leastsquares approach tries to obtain a set of UH ordinates that minimizes the
sum of squares of the errors, and leads to the following solution for the UH,
[U] = [[ P]
T
[P]]
−1
[P]
T
[Q] (9)
Jorge A. Ramírez
10
where [[P]
T
[P]]
1
is the inverse of matrix [[P]
T
[P]], and the superscript T stands for
transpose. This approach is adequate when the data (precipitation and discharge) are
relatively error free or the errors can be expected to be small. When the data can be
expected to contain large errors a different approach may be more adequate. Such
approach would minimize the sum of the absolute value of the errors. Linear
programming would then be the appropriate solution procedure.
CONCEPTUAL (SYNTHETIC) UNIT HYDROGRAPHS
As indicated earlier, sets of concurrent observations of effective rainfall and direct
runoff are required for the derivation of unit hydrographs. Thus, the resultant UH is
specific to the particular watershed defined by the point on the stream where the direct
runoff observations were made. When no direct observations are available, or when UH’s
for other locations on the stream in the same watershed or for nearby watersheds of
similar characteristics are required, Synthetic Unit Hydrograph procedures must be used.
Synthetic Unit Hydrograph procedures can be categorized as (e.g., Chow et al.,
1988): 1) those based on models of watershed storage (e.g., Nash, 1957, 1958, 1959;
Dooge, 1959; etc.); 2) those relating hydrograph characteristics (time to peak, peak flow,
etc.) to watershed characteristics (e.g., Snyder, 1938; Geomorphologic Instantaneous Unit
Hydrograph); and 3) those based on a dimensionless unit hydrograph (e.g., Soil
Conservation Service, 1972).
Conceptual UH’s Based on Models of Watershed Storage
The Nash and Dooge Models.
Linear Reservoirs  A linear reservoir is characterized by a linear relationship between the
storage and the output as,
S(t) = kQ(t) (10)
The impulse response function of such linear reservoir is,
u(t ) =
1
k
e
−t / k
(11)
Nash (1957,1958, 1959, 1960) proposed a cascade of n equal linear reservoirs as a
model on which to base the derivation of IUH’s for natural watersheds. The Nash model
is one of the most widely used models in applied hydrology. Using the convolution
equation (eq. 4) and the impulse response function for a single linear reservoir (eq. 11),
the IUH corresponding to the Nash Model can be easily obtained as follows,
u
n
(t) =
1
kΓ(n)
(
t
k
)
n−1
e
−t / k
(12)
Jorge A. Ramírez
11
This equation is a two parameter gamma distribution function, where n is the shape
parameter, and k is the scale parameter. When rainfall and runoff data are available, the
shape parameter can be estimated as the inverse of the second nondimensional moment
of the IUH about the centroid (i.e., the inverse of the square of the coefficient of
variation); and the scale parameter can be estimated as the ratio of the first order moment
of the IUH about the origin to n. A general form for the peak discharge, Q
max
, due to a
finite pulse of rainfall excess with duration t
r
and uniform intensity I
e
was suggested by
Adom et al. (1989) in the form of a Weibull cumulative distribution function (CDF), i.e.
=


.

\

−
β
α
l
r
t
t
e
AI Q e  1
max (13)
where A is the contributing area, and t
l
is the basin lagtime, which is given by t
l
= nk.
The values of a and b, which are parameters depending on n, i.e. the number of reservoirs
in the cascade, are given in Table 1 (Ramírez, et al., 1994).
TABLE 1. Parameter values for the peak discharge equation of the linear
reservoir cascade model
n 1 2 3 4
a 1 1.017 1.175 1.348
b 1 1.220 1.313 1.370
Linear Channels  A linear channel is characterized by continuity and momentum
equations given by,
∂Q
∂x
+
∂A
∂t
= 0
(14a)
A = C(x)Q (14b)
For conditions of no lateral inflow, the unit impulse response function of such linear
channel is,
u(t ) = δ (t − T) (15)
which implies that the downstream outflow is equal to the upstream inflow delayed by a
lagtime T representing the travel time through the system.
The cascade of linear reservoirs model can be modified in order to account for
translation by using combinations of linear channels and linear reservoirs. Dooge (1959)
developed a general unit hydrograph theory under the assumption that basin response can
Jorge A. Ramírez
12
be represented by a cascade of linear channels and linear reservoirs in series. For this
model the drainage area of the watershed is divided into sub areas using the isochrones. A
linear channel in series with a linear reservoir represents each sub area. If this is done
assuming a system of n equal linear reservoirs of parameter k and n equal linear channels
of lagtime T, the IUH or impulse response function is,
u
n
(t) =
1
kΓ(n)
(
t − nT
k
)
n−1
e
−(t −nT )/ k
(16)
As a general rule, linear channels and linear reservoirs can be combined in series and/or
in parallel to model complex hydrologic systems.
Conceptual UH’s Based on Relationships Between Hydrologic Response and
Watershed Characteristics
Flow hydrographs travel through a stream network after runoff is generated on the
hillslopes. The drainage network develops so that a certain load of water and sediment (as
a result of erosive processes) can be evacuated in an optimal manner (least energy
expenditure (e.g., Molnár and Ramírez, 1998a, 1998b)). Therefore, the characteristics of
the outflow hydrograph and the characteristics of the hillslopes and drainage network
must be intimately linked as they shape and modify each other.
Geomorphologic Unit Hydrographs. Geomorphologic parameters have been widely used
to characterize and model hydrologic response. In particular, geomorphologic parameters
have been used in synthesizing unit hydrographs for gauged and ungauged watersheds; as
well as for estimating flood potential indices and flood frequency distributions. Recent
efforts have concentrated on establishing a theoretical link between hydrologic response
on the one hand, and climatic and geomorphologic characteristics on the other. Even
though climate and geologic processes are continually changing and affecting hydrologic
response, at the time scales of interest, climate and geomorphology may be considered
constant boundary conditions, which uniquely define watershed behavior. The theoretical
link was developed in the late seventies and early eighties and was based on the ideas of
the topologically random model for drainage networks. The fundamental postulates of the
random topological model can be summarized as follows (e.g., Shreve, 1967; Gupta and
Waymire, 1983; Bras, 1990):
a) in the absence of natural controls, natural drainage networks are topologically random;
and,
b) interior and exterior link lengths, as well as their associated areas, are independent
random variables.
The GIUH. RodríguezIturbe and Valdés (1979) linked the hydrologic response of
watersheds in terms of the IUH to the their geomorphologic parameters, and to a dynamic
parameter. The IUH is interpreted as the probability density function of the travel times to
the basin outlet of water droplets randomly and uniformly distributed over the watershed.
The resulting IUH or impulse response function for the basin is a function of the
Jorge A. Ramírez
13
probability density function of the travel times of droplets in streams of a given order and
of the transition probabilities of water droplets going from a given state (stream segment)
to another of higher order. In this model, streams of a given order define the states of the
system. The travel times in the streams are assumed to be exponentially distributed and
independent of one another. The initial probabilities of a drop landing anywhere on the
basin, as well as the transition probabilities which are required in order to define the
probability function of a particular path that a drop may take to the outlet, can be defined
as functions of only geomorphologic and geometric parameters. The initial probability of
a drop falling in an area of order i is equal to the percent contributing area for the given
order. The initial probabilities are thus related to the average ratio of the average area of
subbasins of a given order to the average area of subbasins of the next higher order (i.e.,
the area ratio, RA). Similarly, the transition probabilities from state i (stream order) to state
j, where j represents a stream of higher order, are functions of the number of streams of
order i draining into streams of order j, divided by the total number of streams of order i.
Thus, the transition probabilities are related to the average ratio of the number of streams
of a given order to the number of streams of the next order (i.e., bifurcation ratio, RB).
These ratios are the parameters of the socalled Horton Laws summarized below.
The Horton law of stream numbers states that there exists a geometric relationship
between the number of streams of a given order N
ω
and the corresponding order, ω. The
parameter of this geometric relationship is the Bifurcation Ratio, RB. This ratio has been
found in nature to be between 3 and 5.
N
ω
= R
B
Ω−ω
(17)
The Horton law of stream lengths states that there exists a geometric relationship
between the average length of streams of a given order and the corresponding order, ω.
The parameter of this relationship is the socalled Length Ratio, R
L
.
L
ω
= L
1
R
L
ω−1
(18)
The Horton law of stream areas states that there exists a geometric relationship
between the average area drained by streams of a given order and the corresponding order
ω. The parameter of this relationship is the socalled Area Ratio, R
A
. This ratio has been
observed in nature to vary within a narrow range of 3 to 5.
A
ω
= A
1
R
A
ω−1
(19)
In the equations above, Ω is the order of the basin, and the overbar indicates the average
value of the corresponding variable.
RodríguezIturbe and Valdés (1979) suggested that the probability distribution
function of travel time in streams of a given order is an exponential distribution with
parameter λ representing the inverse of the mean travel time. They suggest estimating λ as
the ratio of a characteristic velocity, V, and a characteristic length scale given by the mean
length of streams of the given order. Valdés et al. (1979) suggest using the peak velocity
Jorge A. Ramírez
14
for the characteristic velocity. RodríguezIturbe and Valdés (1979) obtain expressions for
the time to peak, t
p
, and peak discharge, q
p
, of the IUH which are functions of
geomorphology, and of the dynamic parameter represented by the peak velocity. Their
results, obtained through numerical integration of the full GIUH, and multiple regression
analysis, are simple, elegant, and easy to use. They are,
q
p
=
1. 31
L
Ω
R
L
0.43
V
(20)
and,
t
p
=
0. 44L
Ω
V
(
R
B
R
A
)
0.55
R
L
−0.38
(21)
where L
Ω
, is the length of the highest order stream in kilometers, V is in meters per
second, and the units of q
p
and t
p
are the customary units of inverse time in hours, and
time in hours, respectively.
The dependence of the GIUH on the dynamic parameter V can be used to address
nonlinear watershed response issues. Nonlinear effects in the basin response manifest
themselves in the characteristic discharge velocity. The theoretical framework of the
GIUH above was verified through numerical experiments with a detailed physically
based watershed model on several basins in Venezuela and Puerto Rico with excellent
results (Valdés et al., 1979). They indicate that the nonlinear characteristics of the
response function of a basin can be modeled with a linear scheme such as the GIUH but
with a characteristic velocity that is representative of the discharge velocity for the given
event. RodríguezIturbe et al. (1979) show that the dynamic parameter of the GIUH can
be taken as the spacetime average flow velocity for a given rainfallrunoff event in a
basin. This is based on the hypothesis of spatial uniformity of the flow velocity
distribution throughout the river network (Pilgrim, 1976 and 1977). However, Agnese et
al. (1988) showed that the estimation of the dynamic parameter of the GIUH could also
be performed for those basins where flow velocity distribution depends on stream order.
The parameters of the Nash model and the geomorphologic descriptors of
drainage networks can be related by relating approximate measures of volume of the IUH
for a cascade of linear reservoirs and the volume of the GIUH (Rosso, 1984). Rosso
equated the product of the time to peak and peak flow of the GIUH with the product of
the time to peak and peak flow of the Nash model IUH. It is simple to show that for the
GIUH this product equals,
05 . 0 55 . 0
) ( 58 . 0 ) (
L
B
A
GIUH p p
R
R
R
q t = (22)
and for the Nash model IUH, this product equals,
Jorge A. Ramírez
15
) ( / ) 1 ( ) (
) 1 (
α α
α α
Γ − =
−
e q t
Nash p p
(23)
These equations allow estimation of the shape parameter. Using numerical algorithms to
solve the resulting equation Rosso obtained,
07 . 0 78 . 0
) ( 29 . 3
L
A
B
R
R
R
= α (24)
Ω
−
= L V R R R k
L B A
1 48 . 0
)] /( [ 70 . 0 (25)
where α and k are the shape parameter and the scale parameter of the Nash model IUH,
respectively. When the parameters of the Nash model are estimated using these equations,
and results of their application are compared to other empirical parameter estimation
procedures, the adequacy of the geomorphologic connection is clearly demonstrated.
These results improve the prediction capability of the Nash model for regions where no
hydrologic records exist. The shape parameter of the IUH is shown to be a function only
of Horton ratios; whereas the scale parameter is shown to depend both on geomorphology
and streamflow velocity.
Geomorphologic IUH's have also been proposed that are based on characterizing
drainage networks as a function of network links as opposed to streams (Gupta et al.,
1986; Mesa and Mifflin, 1986; Troutman and Karlinger, 1984, 1985, 1986; and Gupta
and Mesa, 1988.) In that case, the GIUH is expressed in terms of the width function as,
∫
∞
=
0
*
) , ( ) , ( ) ( dx t x N t x g t u (26)
where g(x, t) is the hydrologic response function of a single channel a distance x from the
basin outlet; and N
*
(x) is the normalized width function which represents the probability
density function of the number of links at a distance x from the outlet. N
*
(x, t) is given by,
∫
∞
=
0
*
) , ( / ) , ( ) , ( dx t x N t x N t x N (27)
where N(x, t) is the width function, which measures the number of links at a distance x
from the basin outlet. The linkbased GIUH above has been approximated by its
conditional expected value, conditional on a topological parameter vector (Karlinger and
Troutman, 1985). When the topological parameter is the magnitude M of the network,
Karlinger and Troutman have shown that, asymptotically for large M and for g(x, t)
representing a simple translation, the expected conditional value of u(t) is,
0 ),
4
exp(
) / ( 2
] / ) ( [
2
2 2
2
>
−
= t
M
t V
V M
t
M t u E
i i
l l
(28)
Jorge A. Ramírez
16
where M is the magnitude of the basin and equal to the number of first order streams,
l
i
is
the mean length of the interior links, and V is the velocity of translation. Troutman and
Karlinger show that this result is always true regardless of the form of g(x, t).
Finally, Rinaldo et al. (1991) studied hydrologic response by decomposing the
process of river runoff into two distinct contributions, one accounting for the travel time
within individual reaches (hydrodynamic dispersion), and the other accounting for river
network composition (geomorphologic dispersion). Because the analysis showed the
latter one to play the major role in determining basin response, models based on the
accurate specification of the geometry and the topology of the network and simplified
dynamics are theoretically validated irrespective of the choice of the travel time
probability density function.
Channel Losses in the GIUH. The GIUH has undergone several modifications and
enhancements in order to make it more physicallybased. One of those modifications
consists in incorporating channel infiltration losses. DíazGranados et al. (1983a), using
previous results by Kirshen and Bras (1983), derived expressions for the probability
distribution function of travel times in streams of any given order taking into account
channel infiltration losses. The channel response to an instantaneous input anywhere
along the channel is interpreted as the conditional probability distribution function of the
travel time of a drop traveling a given distance along the channel. This response is
obtained as the solution of the linearized equations of motion for unsteady flow in a wide
rectangular channel in which the infiltration losses are assumed to be proportional to the
instantaneous discharge at any point along the channel. The proportionality coefficient is
known as the infiltration parameter. Results indicate that the linear reservoir assumption,
implicit in the GIUH (exponential distribution of stream travel times), is in fact adequate.
Studies on the geomorphologic response of river basins have increasingly stressed
the importance of providing an accurate description of the quantitative properties of river
network systems. Within this context, for example, La Barbera and Rosso (1987, 1989)
first indicated the fractal nature of river networks and its relation to quantitative
geomorphology as initiated by Horton's studies on river network composition. In addition,
general theories of the evolution of drainage systems have been developed. For example,
energy dissipation concepts have been used to link the local and global properties of a
river network to the hydrologic conditions of its watershed (e.g., Howard, 1990;
RodríguezIturbe et al., 1992; Rigon et al., 1993; Molnár and Ramírez, 1998a, 1998b).
These general theories allow for a fundamental description of global network properties
and thus should be incorporated into improved geomorphologic models of basin response.
The Climatic GIUH  The GcIUH. The instantaneous unit hydrograph is derived under the
assumption that it is a random function of climatic and geomorphologic characteristics
and that it varies with the characteristics of the rainfall excess. Accordingly, Rodríguez
Iturbe et al. (1982a) obtain probability density functions of the peak discharge and the
time to peak as functions of the rainfall intensity, i, and duration, t
r
, as well as
geomorphology. Rainfall is characterized by a constant intensity over the storm duration,
Jorge A. Ramírez
17
or a rectangular pulse process. These two characteristics, together with geomorphologic
parameters for a first order basin, define the dynamic parameter of the GIUH.
Using the kinematic wave approximation for flow routing along streams of first
order, and a derived distribution approach, RodríguezIturbe et al. (1982a) obtained
analytical expressions for the probability density functions of the time to peak, t
p
, and the
peak discharge, qp, of the GIUH that depend on the mean rainfall intensity; hence the
name GcIUH. These probability distribution functions assume that the time of
concentration of first order streams is much smaller than the duration of the rainfall. They
are,
f (q
p
) = 3. 534Πq
p
1.5
exp(−1. 412Πq
p
2.5
)
f (t
p
) =
0.656Π
t
p
3.5
exp(
−0. 262Π
t
p
2.5
)
(29)
where,
5 . 1
5 . 2
Ω Ω
Ω
= Π
α
L
R A i
L
(30)
The first moments of peak discharge and time to peak of the GcIUH can be
obtained as,
E(q
p
) =
0. 774
Π
0.4
σq
p
=
0. 327
Π
0. 4
E(t
p
) = 0.858Π
0.4
σt
p
= 0. 915Π
0.4
(31)
where L
Ω
, is the length of the highest order stream in kilometers, A
Ω
, is the basin area in
square kilometers, i is the mean rainfall intensity in centimeters per hour, and α
Ω
is in s

1
m
1/3
. The units of q
p
and t
p
are the customary units of inverse time in hours, and time in
hours, respectively.
For individual storm events, the above equations can be manipulated to obtain
expressions for the peak discharge and time to peak as functions of the particular storm
intensity and duration. These expressions lead to,
4 . 0
871 . 0
Π
=
p
q (32a)
4 . 0
585 . 0 Π =
p
t (32b)
5 . 1
5 . 2
Ω Ω
Ω
= Π
α
L
i
R iA
L
(32c)
Jorge A. Ramírez
18
The GcIUH establishes a link between climate, watershed geomorphology, and
the hydrologic response of the basin. Furthermore, the GcIUH allows estimation of the
IUH for a given particular rainfall input, so that problems associated with nonlinear basin
behavior are avoided. For example, runoff from a given rainfall event can be computed
using an IUH derived from the given storm event characteristics, thereby avoiding the
wellknown errors incurred when using an IUH based on a different event.
RodríguezIturbe et al. (1982b) compare results by the GcIUH theory with results
obtained from IUH's derived from simulated rainfallrunoff events for one basin and three
different climates. The theoretical distributions of the time to peak and peak discharge
compare very well with the experiments. Variations in q
p
and t
p
resulting from nonlinear
basin behavior induce large uncertainty in the predicted peak discharge and time to peak
for individual events, when using traditional UH approaches (e.g., Caroni et al. 1986,
Caroni and Rosso 1986, and Rosso and Caroni 1986.) These problems are avoided if the
GcIUH is used, due to its functional dependence on the rainfall characteristics.
Snyder’s Synthetic Unit Hydrograph
The synthetic unit hydrograph of Snyder (1938) is based on relationships found
between three characteristics of a standard unit hydrograph and descriptors of basin
morphology. These relationships are based on a study of 20 watersheds located in the
Appalachian Highlands and varying in size from 10 to 10,000 square miles. The
hydrograph characteristics are the effective rainfall duration, t
r
, the peak direct runoff
rate, q
p
, and the basin lag time, t
l
. From these relationships, five characteristics of a
required unit hydrograph for a given effective rainfall duration may be calculated (e.g.,
Chow et al., 1988; Bras, 1990): the peak discharge per unit of watershed area, q
pR
, the
basin lag, t
lR
, the base time, t
b
, and the widths, W (in time units) of the unit hydrograph at
50 and 75 percent of the peak discharge.
Standard unit hydrograph. A standard unit hydrograph is associated with a
specific effective rainfall duration, t
r
, defined by the following relationship with basin lag,
t
l
,
t
l
= 5.5t
r
(33)
For a standard unit hydrograph the basin lag, t
l
, and the peak discharge, q
p
, are given by,
3 . 0
1
) (
c t l
LL C C t = (34)
q
p
=
C
2
C
p
A
t
l
(35)
The basin lag time of the standard unit hydrograph (equation 34) is in hours, L is
the length of the main stream in kilometers (miles) from the outlet to the upstream divide,
L
c
is the distance in kilometers (miles) from the outlet to a point on the stream nearest the
centroid of the watershed area, and C
1
= 0.75 (1.0 for English units). The product LL
c
is a
Jorge A. Ramírez
19
measure of watershed shape. C
t
is a coefficient derived from gauged watersheds in the
same region, and represents variations in watershed slopes and storage characteristics.
The peak discharge of the standard unit hydrograph (equation 35) is in m
3
/s (cfs), A is the
basin area in km
2
(mi
2
), and C
2
= 2.75 (640 for English units). As C
t,
C
p
is a coefficient
derived from gauged watersheds in the area, and represents the effects of retention and
storage.
To compute C
t
and C
p
for a gauged watershed, the values of L and L
c
are
measured. From a derived unit hydrograph of the watershed, values of its associated
effective duration t
R
in hours, its basin lag t
lR
in hours, and its peak discharge q
pR
in m
3
/s
are obtained. If t
lR
= 5.5t
R
, then the derived unit hydrograph is a standard unit hydrograph
and t
r
= t
R
, t
l
= t
lR
, and q
p
= q
pR
, and C
t
and C
p
are computed by the equations for t
l
and q
p
given above, corresponding to the standard unit hydrograph.
If t
lR
is quite different from 5.5t
R
, the standard basin lag is computed using:
t
l
= t
lR
+
t
r
− t
R
4
(36)
This equation must be solved simultaneously with the equation for the standard unit
hydrograph lag time, t
l
= 5.5t
r
, in order to obtain t
r
and t
l
. The value of C
t
is then obtained
using the equation for t
l
corresponding to the standard unit hydrograph. The value of C
p
is
obtained using the expression for q
p
corresponding to the standard unit hydrograph, but
using q
p
= q
pR
and t
l
= t
lR
.
When an ungauged watershed appears to be similar to a gauged watershed, the
coefficients C
t
and C
p
for the gauged watershed can be used in the above equations to
derive the required synthetic unit hydrograph for the ungauged watershed.
Required unit hydrograph. The peak discharges of the standard and required
UH’s are related as follows,
q
pR
=
q
p
t
l
t
lR
(37)
Assuming a triangular shape for the UH, and given that the UH represents a direct
runoff volume of 1 cm (1 in), the base time of the required UH may be estimated by,
t
b
=
C
3
A
q
pR
(38)
where C
3
is 5.56 (1290 for the English system).
As an aid in drawing an adequate UH, the U.S. Army Corps of Engineers
developed relationships for the widths of the UH at values of 50% (W
50
) and 75% (W
75
)
Jorge A. Ramírez
20
of q
pR
. The width in hours of the UH at a discharge equal to a certain percent of the peak
discharge q
pR
is given by Chow et al. (1988) as,
08 . 1
%
−

.

\

=
A
q
C W
lR
w
(39)
where the constant C
w
is 1.22 (440 for English units) for the 75% width and equal to 2.14
(770 for English units) for the 50% width. Usually, onethird of this width is distributed
before the peak time and twothirds after the peak time, as recommended by the U.S.
Army Corps of Engineers. However, several other authors have recommended different
distribution ratios. For example, Hudlow and Clark (1969) recommend a partition of 4/10
and 6/10, respectively.
Figure 11.2 illustrates the form of Snyder’s synthetic UH. Note that the time lag is
not the same as the time to peak. Also, note that the widths of the hydrograph at 50% and
75% of the peak flow are distributed such that the longer time is to the right of the time to
peak.
0
0.25
0.5
0.75
1
1.25
Time Ratio
0 1 2 3 4
Discharge
Ratio
t
b
t
l
t
r
t
p
W
75
W
50
Figure 11.2: Snyder’s Synthetic Unit Hydrograph
Conceptual UH’s Based on Dimensionless Hydrographs
Soil Conservation Service Dimensionless Hydrograph. The dimensionless unit
hydrograph developed by the Soil Conservation Service has been obtained from the UH’s
for a great number of watersheds of different sizes and for many different locations. The
SCS dimensionless hydrograph is a synthetic UH in which the discharge is expressed as a
ratio of discharge, q, to peak discharge, q
p
and the time by the ratio of time, t, to time to
peak of the UH, t
p
. Given the peak discharge and the lag time for the duration of the
excess rainfall, the UH can be estimated from the synthetic dimensionless hydrograph for
the given basin.
Jorge A. Ramírez
21
0
0.2
0.4
0.6
0.8
1
1.2
Discharge
Ratio
0 1 2 3 4 5
Time Ratio
Figure 11.3: SCS Dimensionless Unit Hydrograph
The SCS suggests that the dimensionless UH can be described in terms of an
equivalent triangular hydrograph. The values of q
p
and t
p
can then be estimated using this
simplified triangular unit hydrograph whose height is equal to q
p
and whose time base, t
b,
is equal to 2.67 t
p
. The time is usually expressed in hours (SCS), and the discharge in
m
3
/s/cm (or cfs/in). After analysis of a great number of UH’s, the SCS recommends a
recession duration of 1.67 t
p
. Because the volume of direct runoff must equal 1 cm, it can
be shown that q
p
=C A/ t
p
where C = 2.08 (483.4 in the English system) and A is the
drainage area in square kilometers (square miles). From a study of many large and small
rural watersheds the basin lag is t
l
= 0.6t
c
, where t
c
is the time of concentration of the
watershed. The time to peak, t
p,
is then equal to t
r
/2+t
l
(e.g., U.S. Soil Conservation
Service, 1985).
Flood Frequency Distributions and Climatic GIUH. Eagleson (1972) presents a general
framework for the development of flood frequency distributions, F(Q
p
)(see also
Stedinger, Chapter 12, this volume). Based on Eagleson's framework (see Figure 11.4),
several authors have developed flood frequency distributions that include various
assumptions about the characteristics of the rainfall process in terms of its probability
density function, f
I
(i), and the watershed processes transforming rainfall into runoff
encoded in the function g(i,t;θ). Either using numerical or analytical procedures these
authors have arrived at different flood frequency distributions. One of the most recent and
most interesting developments has been the use of the GIUH concept in order to quantify
and analyze the effect of geomorphologic characteristics in the determination of regional
flood frequency.
Jorge A. Ramírez
22
Figure 11.4. General framework for flood frequency analysis (adapted from Eagleson,
1972.)
Hebson and Wood (1982) derived a flood frequency distribution function from
assumed climatic distributions for the rainfall parameters and using the GIUH as the
transformation function of rainfall into runoff. These authors assume a model of
rectangular pulses for storms whose intensity and duration are independent, and
exponentially distributed random variables. In terms of the equivalent recurrence interval,
T
E
, Hebson and Wood obtain,
lnT
E
= −ln[λ exp(−λt
re
0
t
k
∫
−β
*
(Q
p
−Q
b
) / A
c
g(t
re
))dt
re
+
exp(−λt
k
− β
*
(Q
p
− Q
b
) / A
c
)] − lnn
(40)
where n is the average number of annual direct runoff events, Q
p
is the flood peak under
consideration, Q
b
is the base flow, A
c
is the contributing area, g(t) is the area under the
GIUH as a function of time, t
re
is the duration of excess rainfall, t
k
is the duration of the
GIUH or kernel length, λ is the inverse mean duration of rainfall events, and β
*
is a
modified areal inverse mean intensity of rainfall events. They applied this concept to two
thirdorder basins, namely the Bald Eagle Creek in Pennsylvania, and the Davidson River
in North Carolina. Comparisons of their flood frequency distributions with the one
Rainfall Model
Runoff Model
Derivation of Peak Flow
Probability Distribution
Function
∫∫
=
R
I p Q
di i f t i g Q F ) ( ) ; , ( ) ( θ
r
) (
p Q
Q F
) ; , ( θ
r
t i g Q
p
=
) (i f
I
Climate Parameters, β
r
Catchment Parameters, θ
r
Jorge A. Ramírez
23
derived by Eagleson (1972), which is based on kinematic wave concepts for overland
flow, indicate that the bestfit GIUH produces better representation of the observed flood
frequency distributions.
DíazGranados et al. (1983, 1984) derived a flood frequency distribution based on
the GcIUH. Effective rainfall intensity, i
e
, and effective rainfall duration, t
e
, were obtained
by assuming precipitation to be a process of rectangular pulses, and runoff to be generated
only by the Horton infiltration excess mechanism. Runoff is then generated only when the
precipitation intensity exceeds the soil infiltration capacity. Infiltration capacity was
defined by using Philip's solution to the onedimensional, concentrationdependent
equation of the diffusion process in unsaturated porous media. Storm intensity and
duration were assumed to be independent and exponentially distributed random variables.
They used a derived distribution approach in order to link the joint probability
distribution function of storm duration and storm intensity and a physicallybased
representation of the infiltration process in order to obtain the joint probability density
function of effective storm duration and effective storm intensity. This derived
distribution was used together with a triangular representation of the GcIUH in order to
obtain a flood frequency distribution. The flood frequency distributions obtained by this
method depend on climatic and geomorphologic parameters of the watershed as well as
on soil properties governing the infiltration process. In terms of the equivalent recurrence
interval, their derived flood frequency distribution is,
)] ( 1 [
1
P Q E
Q F n T − =
−
(41)
] [ ) 1 ( ) 2 exp( 1 ) (
4
1
∑
=
−
+ + Γ − − =
i
i Q Q
J I a Q F
σ
σ σ σ β δ (42)
where I, and J
i
are integral equations that cannot be solved analytically, and which depend
on the mean intensity and the mean duration of precipitation events, as well as on the
infiltration sorptivity of the soil, the discharge under consideration, and geomorphologic
parameters of the watershed (see DíazGranados et al., 1984).
Given that the above distribution depends on the elusive initial soil moisture,
DíazGranados et al. (1984) resort to using arguments of ecological optimality for water
limited natural systems in order to define a long term average soil moisture, s
o
. Using this
approach, they verified their flood frequency distribution against observed distributions
from two very different climatic regimes. Their results indicate good agreement for both
wet and arid climates, although the agreement is better for the arid climate of Santa Paula
Creek in California than for the wet climate of the Nashua River basin in New
Hampshire.
However, obtaining the analytical form of the flood frequency distribution from
climate and geomorphology often results in a cumbersome derivation, sometimes yielding
implicit equations to be solved numerically. To overcome this problem, Adom et al.
(1989) introduced the method of approximate moments by Taylor series to estimate
Jorge A. Ramírez
24
second order statistics of both flood peak and volume. For this purpose, they used the Soil
Conservation Service's Curve Number method to describe basin rainfall excess, and the
GIUH to model surface runoff response. After a given type of frequency distribution has
been identified at the regional scale by analyzing annual flood series, this method can be
used to estimate the flood frequency distribution for ungauged catchments.
FLOOD ROUTING
The term flood routing refers to procedures to determine the outflow hydrograph
at a point downstream in a river (or reservoir) as a function of the inflow hydrograph at a
point upstream.
As flood waves travel downstream they are attenuated and delayed. That is, the
peak flow of the hydrograph decreases and the time base of the hydrograph increases. The
shape of the outflow hydrograph depends upon the channel geometry and roughness, bed
slope, length of channel reach, and initial and boundary flow conditions.
The propagation of flood waves in a channel is a gradually varied unsteady flow
process, which is governed by conservation of mass and momentum equations. The
solution of these equations in a distributed manner is referred to as distributed routing of
flood waves. When no spatial variability is taken into account and when the channel reach
or reservoir is considered as a black box, the corresponding routing procedure is referred
to as lumped routing.
Lumped Flood Routing
Lumped or hydrologic routing is governed by the continuity equation and a
storagedischarge functional relationship. The continuity equation is,
dS(t)
dt
= I(t) − O(t) (43)
In this equation, S(t) is the storage within the system (channel reach or reservoir), I(t) is
the inflow hydrograph at the upstream end of the reach, and O(t) is the outflow
hydrograph at the downstream end of the reach. The continuity equation can be integrated
over a given ∆t to obtain,
S(t
i + 1
) − S(t
i
) = dS(t) = I(t)dt
ti
t
i+1
∫
S(ti )
S(t
i+ 1
)
∫
− O(t)dt
ti
t
i +1
∫
(44)
where the subscripts i and i+1 refer to the beginning of two consecutive time intervals.
Assuming a linear variation of input and output fluxes during the ∆t leads to,
S(t
i + 1
) − S(t
i
) =
∆t
2
[I(t
i + 1
) + I(t
i
)] −
∆t
2
[O(t
i +1
) + O(t
i
)] (45)
Jorge A. Ramírez
25
In equation (46) there are two unknown quantities, S(t
i+1
) and O(t
i+1
). Thus, a
second equation relating S(t) and O(t) is required. The nature of this relationship may be
linear, or nonlinear, or one to one, or hysteretic. The complexity of the routing procedure
depends on the nature of this relationship.
Reservoir or level pool routing refers to flood routing for systems whose storage
and outflow are related by a function of the type S(t) = f[O(t)] which is onetoone (i.e.,
unique, nonhysteretic). This characteristic of their S vs. O relationships implies that for a
given set of conditions (e.g., stage or storage) the outflow is unique, independent of how
that storage is achieved. Such systems have a pool that is wide and deep compared to its
length in the direction of flow, low flow velocities, and horizontal water surfaces or
negligible backwater effects. Reservoirs or systems with horizontal water surfaces have
relationships of the onetoone type.
The solution procedure involves rearranging the continuity equation such that all
unknown quantities are on the left hand side of the equation,
2S(t
i + 1
)
∆t
+ O(t
i + 1
) = [I(t
i +1
) + I(t
i
)] + [
2S(t
i
)
∆t
− O(t
i
)] (46)
For level pool systems, the storage and the outflow are onetoone functions of elevation
(see Figures 11.5 and 11.6).
0
50
100
150
200
250
300
350
400
450
500
0.E+00 1.E+07 2.E+07 3.E+07 4.E+07 5.E+07 6.E+07 7.E+07 8.E+07 9.E+07
Storage (m
3
)
E
l
e
v
a
t
i
o
n
(
m
)
Figure 11.5: Storage vs. Elevation Relationship
Jorge A. Ramírez
26
0
100
200
300
400
500
600
700
800
0 50 100 150 200 250 300 350 400 450 500
Elevation (m)
D
i
s
c
h
a
r
g
e
(
m
/
s
)
Figure 11.6: Elevation vs. Discharge Relationship
Thus, the left hand side of the equation above is a unique function of water
surface elevation in the system, only. Usually, the storageelevation relationship is
available from topographic surveys, and the outflowelevation relationship is available
from hydraulic considerations with respect to the outlet structures (e.g., spillways, etc.).
The solution involves the development of the function 2S/∆t + O = f[O(h)] (see
Figure 11.7) and then using the function sequentially to solve for O(t
i+1
) for every time
step. In this equation, h is water surface elevation in the reservoir.
0
100
200
300
400
500
600
700
800
0 1000 2000 3000 4000 5000 6000 7000 8000 9000
2S/∆t+O (m
3
/s)
D
i
s
c
h
a
r
g
e
(
m
3
/
s
)
Figure 11.7: StorageIndication Function
As observed in Figure 11.8, the effect of storage is to redistribute the flow
hydrograph by shifting its centroid from the position of that of the inflow hydrograph to
that of the outflow hydrograph. For these systems, the peak outflow occurs when the
Jorge A. Ramírez
27
outflow hydrograph intersects the inflow hydrograph, at which time the maximum storage
in the system is also achieved (see Figure 11.8).
0
50
100
150
200
250
0 50 100 150 200 250 300 350 400 450 500
Time (h)
D
i
s
c
h
a
r
g
e
(
m
3
/
s
)
Figure 11.8: Level Pool Routing: Inflow and Outflow Hydrographs
Channel Routing The storagedischarge relationship in the case of channel routing is
not onetoone, but hysteretic. In other words, the outflow from a given reach is not a
single valued function of the storage in the reach. In its most simple interpretation, the
magnitude of the outflow discharge depends on whether the outflow is occurring during
the rising limb or the recession limb of the inflow hydrograph. This behavior is a result of
backwater effects, which are important in long and narrow flow routing conditions such
as occur in channels.
The Muskingum method is one of the most widely used methods of lumped,
channel flow routing. In addition to the lumped form of the continuity equation,
dS(t)
dt
= I(t) − O(t) (43)
this method assumes a linear storage discharge relationship as given below,
S(t) = k[xI(t) + (1− x)O(t)] (47)
This relationship specifies that the storage is proportional to a convex linear combination
of inflow and outflow rates. The Muskingum method has two parameters, k and x.
Parameter k has units of time and represents the average travel time of a kinematic wave
through the reach, and it is known as the storage time constant. Parameter x is
dimensionless and represents the relative importance of the effects of inflow rates on
defining the storage in the reach. Theoretically this parameter varies between 0 and 1.
Combining the continuity and storage equations, and integrating over a short time
interval, the outflow at the end of a given time interval, O(t
i+1
) can be expressed as a
linear combination of the inflow rates at the beginning and end of the given interval, I(t
i
)
and I(t
i+1
), as well as the outflow rate at the beginning of the interval, O(t
i
) as,
Jorge A. Ramírez
28
O(t
i +1
) = C
o
I(t
i+ 1
) + C
1
I(t
i
) + C
2
O(t
i
) (48)
The coefficients or weights of this linear combination can be computed as a function of k,
x, and ∆t as indicated below,
C
o
= (−kx + 0.5∆t) /C
3
(49a)
C
1
= (kx + 0.5∆t) /C
3
(49b)
C
2
= (k − kx − 0.5∆t) / C
3
(49c)
C
3
= k − kx + 0.5∆t
(49d)
It is simple to show that the Muskingum routing solution can be expressed also as a
discrete convolution equation as,
Q
n
= I
n−m+ 1
m= 1
n
∑
U
m
(50)
where U
m
is the UH of the Muskingum method whose ordinates are given by the
following equations,
U
1
= C
o
U
2
= C
o
C
2
+ C
1
U
m
= U
m− 1
C
2
, m> 2
(51)
The Muskingum parameters k and x can be easily estimated graphically, or by
using the leastsquares method. The graphical method consists in choosing x such that the
loop resulting from the plot of S vs. (xI + (1x)O) becomes as close to a straight line as
possible (see Figure 11.9ad). The slope of the resulting straight line is an estimate of the
parameter k (see equation 47). In order to ensure positivity of the unit hydrograph
ordinates, ∆t must be selected such that x < 0.5 ∆t/k < 1x.
This method results in a trial and error procedure which, when properly carried
out, constitutes a graphical implementation of the leastsquares method. Or in other
words, the leastsquares method is a numerical expression of the graphical method.
Observe that routing in systems with hysteretic S vs. O relationships leads to additional
modifications of the inflow hydrograph represented in a delay of the response of the
system and in asynchronous occurrences of the maximum storage and maximum outflow,
the latter occurring after the former (see Figure 11.10).
Jorge A. Ramírez
29
0.E+00
1.E+06
2.E+06
3.E+06
4.E+06
5.E+06
6.E+06
7.E+06
8.E+06
9.E+06
1.E+07
0 10 20 30 40 50
Weighted Average Influx (m
3
/s)
S
t
o
r
a
g
e
(
m
3
)
x = 0.05
Figure 11.9a: Weighted Average Influx vs. Storage Function  x = 0.05.
0.E+00
1.E+06
2.E+06
3.E+06
4.E+06
5.E+06
6.E+06
7.E+06
8.E+06
9.E+06
1.E+07
0 10 20 30 40 50
Weighted Average Influx (m
3
/s)
S
t
o
r
a
g
e
(
m
3
)
x = 0.08
Figure 11.9b: Weighted Average Influx vs. Storage Function  x = 0.08.
0.E+00
1.E+06
2.E+06
3.E+06
4.E+06
5.E+06
6.E+06
7.E+06
8.E+06
9.E+06
1.E+07
0 10 20 30 40 50
Weighted Average Input (m
3
/s)
V
o
l
u
m
e
(
m
3
)
x = 0.09
Figure 11.9c Weighted Average Influx vs. Storage Function  x = 0.09.
Jorge A. Ramírez
30
0.E+00
1.E+06
2.E+06
3.E+06
4.E+06
5.E+06
6.E+06
7.E+06
8.E+06
9.E+06
1.E+07
0 10 20 30 40 50
Weighted Average Influx (m
3
/s)
S
t
o
r
a
g
e
(
m
3
)
x = 0.10
Figure 11.9d Weighted Average Influx vs. Storage Function  x = 0.10
0
20
40
60
80
100
120
0 20 40 60 80 100 120 140
Time (h)
D
i
s
c
h
a
r
g
e
(
m
3
/
s
)
Figure 11.10: Muskingum Routing: Inflow and Outflow Hydrographs
Using a hydraulic analogy and applying the Muskingum equations over a short ∆x,
the Muskingum routing equation constitutes an approximate solution to the kinematic
wave equations (see Distributed Flood Routing section below) (Cunge, 1969). Expressing
the solution in a spacetime domain, the Muskingum routing equations can be rewritten
as,
Q
i +1
j +1
= C
o
Q
i
j +1
+ C
1
Q
i
j
+ C
2
Q
i +1
j
(52)
where the flow discharge
j
i
Q refers to position i in space and j in time. In this solution,
the parameters k, and x, required in order to obtain C
o
, C
1
, C
2
, and C
3
are estimated as,
k =
∆x
c
k
(53a)
Jorge A. Ramírez
31
x =
1
2
(1−
Q
c
k
BS
o
∆x
) (53b)
In addition, Cunge also demonstrated that this solution constitutes an approximate
solution of a modified diffusion equation if the parameters k and x are estimated as
expressed above. In those equations, c
k
is the celerity of the kinematic wave
corresponding to Q and B, where B is the width of the water surface, Q is the discharge, A
is the cross sectional area of flow, S
o
is the bottom slope, and ∆x is the increment in space.
Based on this analysis, it can be shown that x should be between 0 and 0.5.
Distributed Flood Routing
The characteristics of hydrologic systems are extremely variable in space and
time. The excitation of the system (i.e., precipitation), the initial conditions, and the
boundary conditions are, in general, functions of space and time. Therefore, the response
of the system, that is, the flow of water through the soil and in the channels, is a
distributed process in which the characteristics of the flow change both in time and space.
Flow characteristics such as flow discharge, flow depth, and flow velocity can be
obtained as a function of space and time using a dynamic, distributed flow routing
procedure.
The governing equations for onedimensional, unsteady flow in an open channel
are,
∂Q
∂x
+
∂A
∂t
= q (54)
∂Q
∂t
+
∂(βQ
2
/ A)
∂x
+ gA(
∂y
∂x
− S
o
+ S
f
) − βqv
x
= 0
(55)
where Q is discharge, A is the cross sectional flow area, y is flow depth, S
o
is the channel
bottom slope, S
f
is the friction slope, β is the momentum correction coefficient for non
uniform flow velocity distribution, q is lateral inflow per unit length of flow, and v
x
is the
magnitude of the lateral inflow velocity component in the direction of flow.
These two equations are known as the de SaintVenant equations. The main
assumptions encoded in these equations are: 1) the flow is onedimensional; 2) the
vertical accelerations are negligible so that the vertical pressure distribution is hydrostatic
(gradually varied flow); 3) the slope of the channel bottom is small and the channel
bottom is fixed; 4) resistance coefficients for steady uniform turbulent flow are
applicable; and 5) the fluid is incompressible and of constant density.
The momentum equation includes terms for changes in momentum due to: a)
changes in velocity over time (local acceleration); b) changes in velocity in space along
the channel (convective acceleration); c) water level differences along the channel and
consequent imbalance of pressure forces; d) gravitational acceleration; e) frictional
forces; and f) lateral inflow. Local and convective accelerations represent the effect of
Jorge A. Ramírez
32
inertial forces in the flow. Depending on the flow conditions, some of these terms may be
negligible as compared to others in the equation giving rise to either kinematic, or
diffusive, or full dynamic wave situations. This sometimes leads to simplifications of the
flood routing problem (e.g., Chow et al, 1988; Singh, 1989; Bras, 1990).
Kinematic Waves. Kinematic waves govern the flow when inertial and pressure forces
are not important. In natural flood wave conditions both kinematic and dynamic waves
are present. However, for situations where the Froude number is less than 2, dynamic
waves tend to attenuate more rapidly than kinematic waves (e.g., Eagleson, 1970).
Although there is no single set of criteria to define when a kinematic wave solution is
acceptable, the magnitude of the socalled kinematic parameter is useful. The kinematic
parameter is defined as,
K =
S
o
L
y
o
F
2
(56)
where S
o
is the bottom slope, L is the length of (overland or channel) flow, y
o
is the
normal flow depth at the downstream end of the flow length, and F is Froude number for
normal flow at the downstream end of the flow length. Woolhiser and Liggett (1967)
have shown that for K values greater than 10, the flow dynamics tend to be dominated by
kinematic waves. In overland flow situations the value of K usually exceeds this
threshold, (K~O(10
2
 10
3
)), warranting a kinematic wave approach for such flow routing
problems.
For a kinematic wave, the momentum equation reduces to,
f o
S S = (57)
implying that the energy grade line is parallel to the channel bottom, and that the flow is
steady and uniform. The above momentum equation can be shown to be equivalent to the
following stagedischarge relationship,
β
αA Q = (58)
which, for example, can be satisfied by Manning equation,
A R
n
S
Q
f
3 / 2
= (59)
Together with the continuity equation,
q
t
A
x
Q
= +
∂
∂
∂
∂
(54)
these equations represent the routing equations for the kinematic wave flow routing
approach.
Jorge A. Ramírez
33
If an observer moves with the kinematic wave at a speed equal to the kinematic
wave celerity, the observer would see the flow rate increase at a rate equal to the lateral
inflow rate, q, as shown below,
q
t
Q
c x
Q
dx
dt
t
Q
x
Q
dx
dQ
k
= + = + =
∂
∂
∂
∂
∂
∂
∂
∂ 1
(60)
It can be seen that for conditions of no lateral inflow, that is, for q = 0, dQ/dt = 0. Thus,
kinematic waves do not attenuate; they simply translate downstream without dissipation.
Given that at any cross section Q and A are functionally related as Q = αA
β
, the continuity
equation can be rewritten as,
∂Q
∂x
+
dA
dQ
∂Q
∂t
= q (61)
and,
∂A
∂t
=
dA
dQ
∂Q
∂t
=
1
αβA
β −1
(
∂Q
∂t
) (62)
∂Q
∂x
+
1
αβA
β −1
(
∂Q
∂t
) =
∂Q
∂x
+
Q
1/ β − 1
α
1/ β
β
∂Q
∂t
= q
(63)
From equations 60 and 63, it is easy to obtain the following expression for the kinematic
wave celerity,
c
k
=
dQ
dA
= αβA
β − 1
(64)
Assuming that overland flow situations can be described as flow over planes, the
kinematic wave approximation has been widely applied to construct models of the
rainfallrunoff process. A basin is described as a collection of such overland flow planes
representing the hillslope processes feeding into a network of collector channels which
feed into a main channel. In this application, the lateral inflow is taken to be equal to the
rainfall excess, and the channel inflow is taken to be flow per unit width of plane. The
kinematic wave model offers the advantage over the unit hydrograph method that it is a
onedimensional, spatiallydistributed solution of the physical equations governing the
surface flow. Furthermore, for overland flow waves in the case of homogeneous planes,
the kinematic wave equations have explicit analytical solutions. However, more complex
representations of the flow process must sometimes be used  twodimensional and fully
dynamic.
In general, the SaintVenant equations do not have explicit analytical solutions,
except in a few particular cases (e.g., kinematic waves for homogeneous planes). Thus,
numerical solution procedures must be used, like finitedifferences methods, or the
Jorge A. Ramírez
34
method of characteristics. Finite difference methods solve the partial differential
equations on a grid placed over the xt plane. For the kinematic wave problem presented
above, these solutions can be linear or nonlinear.
A simple implementation of the linear and nonlinear solutions suggested by
Chow et al., (1988) is adapted below. For the linear solution, the equation is linearized by
substituting an average of known solutions for the coefficient of the nonlinear term. This
leads to the following solution,
Q
i +1
j +1
− Q
i
j +1
∆x
+
1
α
1/ β
β
(
Q
i +1
j
+ Q
i
j + 1
2
)
1 / β −1
(
Q
i+ 1
j +1
− Q
i+ 1
j
∆t
) = (
q
i +1
j + 1
+ q
i+1
j
2
) (65)
Q
i +1
j +1
=
[
∆t
∆x
Q
i
j +1
+
1
α
1/ β
β
Q
i + 1
j
(
Q
i +1
j
+ Q
i
j +1
2
)
1/ β− 1
+ ∆t(
q
i +1
j +1
+ q
i +1
j
2
)]
[
∆t
∆x
+
1
α
1/ β
β
(
Q
i + 1
j
+ Q
i
j + 1
2
)
1 / β −1
]
(66)
In this linear solution, the subscript refers to the space coordinate and the superscript
refers to the time coordinate. The solution advances on a time line from upstream to
downstream.
For the nonlinear kinematic wave solution, the partial differential equations are
recast in finite difference form as shown below. The same subscript and superscript
convention applies as for the linearized solution.
Q
i +1
j +1
−Q
i
j +1
∆x
+
A
i +1
j +1
− A
i +1
j
∆t
=
q
i+1
j +1
+ q
i +1
j
2
(67)
∆t
∆x
Q
i+ 1
j +1
+α
−1/ β
(Q
i +1
j + 1
)
1 / β
=
∆t
∆x
Q
i
j +1
+ α
−1/ β
(Q
i +1
j
)
1/ β
+ ∆t(
q
i +1
j +1
+ q
i +1
j
2
)
(68)
C =
∆t
∆x
Q
i
j + 1
+ α
−1/ β
(Q
i +1
j
)
1/ β
+ ∆t(
q
i +1
j +1
+ q
i +1
j
2
)
(69)
An iterative solution based on Newton’s method is shown below.
f (Q
i+1
j +1
) =
∆t
∆x
Q
i + 1
j+ 1
+α
−1/ β
(Q
i +1
j + 1
)
1 / β
− C
(71)
f ' (Q
i + 1
j +1
) =
∆t
∆x
+
1
α
1/ β
β
(Q
i +1
j + 1
)
1 / β −1
(72)
Jorge A. Ramírez
35
(Q
i +1
j + 1
)
k +1
= (Q
i+ 1
j +1
)
k
−
∆t
∆x
(Q
i+1
j +1
)
k
+ α
−1/ β
[(Q
i + 1
j +1
)
k
]
1/ β
− C
∆t
∆x
+
1
α
1/ β
β
[(Q
i+ 1
j +1
)
k
]
1 / β −1
(73)
Newton’s method is an iterative numerical solution procedure, which
progressively determines improvements to a trial solution as a function of the gradient of
the function at the trial solution. The method converges relatively quickly. A convergence
criterion can be defined in advance, so that a solution is reached when this criterion is met
(e.g., the magnitude of the improvement is less than a threshold).
Dynamic Waves. Dynamic waves govern the flow whenever inertial and pressure
forces are significant when compared to the gravitational and frictional forces. This
condition occurs for unsteady, nonuniform flows for which backwater or tidal effects are
important as, for example, during the movement of a flood wave in a system of channels
of shallow slopes, and may be induced by downstream reservoirs or other downstream
controls, by tributaries, by tides, by storm surges (e.g., caused by hurricanes), or by large
discontinuities in the flow characteristics as a result of dam breaks or large reservoir
releases. As indicated above for the case of hydraulic river routing, backwater effects
induce rating curves that are hysteretic, that is, rating curves for which the discharge is
not a single valued function of stage (i.e., storage per unit length of channel) in the
section.
Full dynamic wave flood routing models can be both steadystate and unsteady
state models. The Hydrologic Engineering Center's River Analysis System (HECRAS)
model is an example of a steadystate flow model used for the computation of water
surface profiles associated with one dimensional, gradually varied, steady flows. HEC
RAS is designed to perform onedimensional hydraulic simulations for a network of
channels, and computes surface water profiles associated with subcritical, supercritical,
and mixed flow regimes. Surface water profiles are solved sequentially by solving the
energy equation using an iterative algorithm called the standard step method. The energy
equation is,
e
h
g
v
Z Y
g
v
Z Y + + + = + +
2 2
1
1 1
1 1
2
2 2
2 2
α α
(74)
where h
e
is the energy head loss, Y
i
is the surface water depth at section i, Z
i
is the
elevation of the channel bottom at section i, v
i
is the average velocity at cross section i,
and g is the acceleration of gravity. The energy head loss between two cross sections
accounts for friction losses and for losses induced by contractions or expansions and is
evaluated as,
g
v
g
v
C LS h
f e
2 2
2
2 2
2
1 1
α α
− + = (75)
Jorge A. Ramírez
36
where L is the reach length weighted by discharge, S
f
is a representative friction
slope between the two sections, and C is a coefficient for expansion/contraction losses.
For situations where flow regimes change abruptly from supercritical to subcritical or
vice versa, the HECRAS model uses the steady state momentum equation instead, as the
energy equation is not applicable under those conditions. The momentum equation is
written in this case as,
x f x
v Q F W P P ∆ = − + − ρ
1 2
(76)
where P
i
is the hydrostatic pressure force at section i, W
x
is the component of the
weight of the water in the section in the direction of flow, F
f
is the force due to friction in
the reach between sections, Q is the discharge, ρ is the density of water, and ∆v
x
is the
change in average velocity within the reach. In addition, the model allows for the
specification of variable cross sectional characteristics, both at a section and along the
channel. Common applications of the HECRAS model include flood plain delineation
and analysis for flood protection and insurance purposes, modeling of single and/or
multiple culverts and bridges, modeling of gated spillways and weirs, modeling of
floodway encroachment, and of icecovered rivers.
In the context of fully dynamic, unsteady flows, the models of the National
Weather Service developed by Fread and collaborators are well documented and
commonly used (e.g., Fread, 1974, 1976, 1978, 1988; Fread and Lewis, 1988). Among
these models are the DWOPER model (Dynamic Wave Operational Model) (Fread,
1978), the DAMBRK model for dam failure studies (Fread, 1988), and the FLDWAV
model (Fread and Lewis, 1988) which synthesizes DWOPER and DAMBRK. These and
similar models solve the fully unsteady dynamic wave equations using finite differences.
However, the treatment and presentation of the details of the solution is beyond the scope
of this chapter. For those details the reader is kindly referred to the above references and
those given therein.
DISTRIBUTED WATERSHED MODELS
Distributed and semidistributed watershed models are generally categorized into
singleevent simulation models and continuous simulation models. The former models are
concerned with simulating a single flood as a result of the occurrence of an individual
precipitation or floodinducing event. The latter models are concerned with the simulation
of watershed processes during a period of time that encompasses more than one
precipitation or floodcausing event. Thus, continuous models are also concerned with
evaporation and subsurface soil moisture redistribution during the interstorm periods.
Most of these models include elements or components that are described using the
concepts of previous sections in this chapter. However, it is implied that watershed
models deal with the transformation of precipitation into runoff, whereas some of the
approaches presented earlier in this chapter are concerned simply with the translation of a
flood wave, not necessarily with the complete precipitationrunoff transformation. These
models are then referred to as flood hydraulics models (e.g., HECRAS.)
Jorge A. Ramírez
37
SingleEvent Simulation Models
There exists a great variety of single event watershed models. Singh (1989)
provides a summary table listing fifteen such models and reviews some of them. The
following sections are based on the review presented in Ramírez et al. (1994). Reference
is made to two models developed by the U.S. Geological Survey and one developed by
the Hydrologic Engineering Center of the U.S. Corps of Engineers. However, many other
alternative models are available such as the TR20 (U.S. Soil Conservation Service,
1973), SWMM (Metcalf & Eddy Inc., 1971), IHM (Morris, 1980), and FLEA (Ranzi and
Rosso, 1991a).
The USGS rainfallrunoff model (Dawdy and O'Donnell, 1965) was one of the
first single event models developed to predict flood volume and peak rates of runoff for
small drainage areas. The model has been utilized by the U.S. Geological Survey in over
twenty states to develop peak flows and hydrographs for state highway agencies. The
process is triggered by the precipitation input P, which augments surface storage R, which
in turn is depleted by infiltration F and evaporation E
R
. The channel outflow Q begins
when the threshold value R
*
is exceeded. The infiltration is evaluated by a Horton type
equation. The infiltration F and capillary rise C augment the soil storage M, which is
depleted by transpiration E
M
. Deep percolation D occurs when the soil moisture threshold
M
*
is exceeded, augmenting groundwater storage G, which in turn is depleted by capillary
rise C and base flow B. If G exceeds the threshold value G
*
, M is absorbed into G, then C
and D disappear and E
M
and F operate on G. The channel storage S and groundwater
storage G are assumed to be linear reservoirs with storage constants K
S
and K
G
,
respectively. The channel flow Q is routed through a linear reservoir S yielding the flow
Q
S,
which added to the base flow B gives the total flow Q.
An alternative USGS model is the distributed routing rainfallrunoff model
developed by Dawdy et al. (1972). This model uses a modified version of the Philip's
infiltration equation and the infiltration capacity concept of Crawford and Linsley (1966)
to determine the excess rainfall. A kinematic wave model component is used for routing.
The model was specifically designed for routing urban flood discharges through a
branched system of pipes or natural channels using rainfall as input. The drainage basin
is presented as a set of segments (overland flow, channel reservoir and nodal) which
jointly describe all subbasins of the entire basin.
A single event watershed model which has been widely used for estimation of
surface runoff and river/reservoir flow in a basin during floods is the HEC1 Flood
Hydrograph Package. It was originally developed in 1967 by the staff of the Hydrologic
Engineering Center at Sacramento, California (now at Davis, California). The current
version of the HEC1 package includes computational capabilities of dam breaks, project
optimization and kinematic wave programs (HEC, 1981).
The hydrologic simulation capabilities of HEC1 include several techniques to
input and distribute the precipitation, treat the precipitation as rainfall or snowfall,
Jorge A. Ramírez
38
compute rainfall and snowmelt losses and excess, and determine subbasin outflow
hydrographs by various hydrologic routing techniques. The model may be used to
simulate a simple singlebasin watershed or a very complex basin with practically
unlimited number of subbasins and river reaches. The HEC1 model can account for
temporal and spatial variability of the precipitation runoff process in a semidistributed
sense. That is, within a subbasin, HEC1 uses spatially and temporally lumped
parameters to simulate the precipitationrunoff process. The precipitation hyetograph
(rain and/or snow) is input over the sub basin, and the losses are computed, leaving an
excess precipitation hyetograph which in turn is transformed into surface runoff
hydrograph through a specified unit hydrograph. The subsurface runoff hydrograph is
computed separately and added to the surface runoff hydrograph to yield the total
subbasin runoff hydrograph. In addition to being capable of hydrologic simulation, the
HEC1 package also has a provision for evaluating reservoir and channel development
plans for flood control purposes by performing the economic analyses of flood damages
for existing and postdevelopment conditions. An additional application of the calibrated
model is for impact assessment studies of basin modifications and channel improvements.
This can be accomplished by modifying the physical parameters of the subbasin or
routing reach in which the change is expected or proposed and by reanalyzing the
response of the watershed.
Continuous Simulation Models
Three continuous watershed models developed in the United States of America
are summarized in this section. The review is based on that of Ramírez et al. (1994). The
models are: the Stanford Watershed Model (SWM), the National Weather Service Model
(NWS) and the Precipitation Runoff Modeling System (PRMS) of the US Geological
Survey. However, many alternative models exist such as the SSARR (U.S. Army Corps
of Engineers, 1972) for daily streamflow simulations and forecasting for large
watersheds, the SHE model developed in Europe (Abbott, 1986a, 1986b), or the LBRM
developed by the U.S. Great Lakes Environmental Research Laboratory (Croley, 1985)
for simulation of daily runoff in the Great Lakes basins.
The Stanford Watershed Model (SWM) was originally developed by Crawford
and Linsley (1966) for the continuous simulation of the hydrologic processes in a
watershed at hourly time scales. Since then, the original model has undergone a number
of revisions and expansions to satisfy specific needs and requirements (e.g., Claborn and
Moore, 1970; Liou, 1970; Anderson, 1971; Ricca, 1972). Several models such as the
Agricultural Runoff Management Model, ARM, (Donigian and Crawford, 1979) and the
Hydrologic Simulation Program, HSPF, of EPA (Johanson et al., 1980) employ the basic
structure of the original SWM for runoff simulation but in addition simulate water quality
and sediment erosion.
The SWM model partitions the watershed into land segments whose boundaries
are established according to hydrologic characteristics and user's needs. The segmentation
of land is usually based on meteorologic, soil, topographic, and drainage considerations.
Thus, each land segment is assumed to be homogeneous with respect to hydrologic
Jorge A. Ramírez
39
characteristics, and is assumed to produce a homogeneous hydrologic response.
Streamflow may be computed at several locations in the stream network.
Two storage zones control overland flow, infiltration, interflow and inflow to the
groundwater storage. The upper zone storage may be physically interpreted as consisting
primarily of depression storage whereas the lower zone represents the soil moisture
storage above the capillary zone. The role of the upper zone is to simulate the initial
response to rainfall, and therefore it is of importance for small storms, and for the first
few hours of major storms. The lower zone plays a major role during large storms in
controlling temporal variation of infiltration rates. Infiltrated water moves to the lower
zone and groundwater storage. Additional infiltration may occur from depression storage
or from runoff, giving rise to delayed infiltration. A fraction of total infiltrated water
reaches the groundwater storage, which then contributes to baseflow. Moisture may be
lost from the system as evapotranspiration from interception, upper zone, lower zone, and
groundwater storage, and stream surfaces. Direct inflows from the impervious areas,
overland flow, interflow, and groundwater flow are combined to become total inflow into
the channel network which contributes to the streamflow hydrograph.
The U.S. National Weather Service (NWS) developed a computerized system of
operational river forecast procedures including data acquisition and processing,
computation of mean basin precipitation, snow accumulation and ablation, explicit soil
moisture accounting, channel routing, parameter optimization and verification, and
operational forecasting. This set of models, known as the National Weather Service
River Forecast System, (NWSRFS), has been extensively described (e.g., Monro and
Anderson, 1974; Peck, 1976). The soil moisture accounting model of the NWSRFS has
been developed by the NWS Sacramento, California River Forecast Center (e.g., Burnash
et al., 1973). The NWS model can be partitioned into two major components, namely:
the Sacramento soil moisture accounting model, and the flow routing model. The
Sacramento soil moisture accounting model is a continuous, conceptual, lumped input
and lumped parameter model which transforms the rainfall input in the land phase of the
hydrologic cycle into channel inflows. The flow routing model is responsible for routing
the channel inflows to the basin or subbasin outlet using unit hydrograph and Muskingum
routing methods.
The Sacramento soil moisture accounting (SMA) model assumes that a catchment
consists of two reservoirs, the upper zone and the lower zone. This conceptualization is
intended to provide a simple but effective representation of the hydrologic process of the
catchment starting with precipitation, the subsequent vertical and horizontal movement of
water through and over the soil, and finally the production of runoff. The upper zone
represents the upper soil layer and interception storage, whereas the lower zone represents
the bulk of the soil moisture and groundwater storage. Each zone stores water in two
forms, tension water and free water. Tension water is only depleted by
evapotranspiration. Depletion of free water occurs vertically as percolation and
horizontally as evapotranspiration and interflow. In the lower zone, there are two types of
free water storage: primary storage, which is slow draining, and provides baseflow over
long periods of time; and supplementary storage, which is fast draining, and supplements
Jorge A. Ramírez
40
baseflow after a period of relatively recent rainfall. Movement of water from upper zone
to lower zone is achieved through percolation, which is a nonlinear function of upper
zone free available water and lower zone moisture deficiency.
Two levels of watershed partitioning are required in the model. The first level
partitions the watershed into soil moisture accounting areas where each area is a
homogeneous unit in terms of the SMA model parameters. Rainfall and
evapotranspiration input data are assumed uniform over each SMA area. The second
level of partitioning divides an SMA area into smaller homogeneous units representing
individual flow planes (overland and channel). Each flow plane is assumed to have
homogeneous unit graphMuskingum model parameters. Although there are two levels
of watershed partitioning, one can assume a flow plane to coincide with the SMA area.
Streamflow is computed at the outlet or flow points of each flow plane.
The model is set up to simulate basin hydrology on hourly or longer time
intervals. Rainfall input data can only be given on an hourly basis whereas
evapotranspiration and streamflow input data can be given on a longer time interval as an
integer multiple of the time interval of rainfall. In such cases, the model uniformly
distributes the evapotranspiration data and computes streamflow over the time basis of
rainfall data and over the time interval specified for streamflows.
Finally, the model generates five components of flow: 1) direct runoff from
permanent and temporary impervious areas, 2) surface runoff which occurs when the
upper zone free water storage is full and precipitation intensity exceeds the rate of
percolation and interflow, 3) interflow resulting from lateral drainage of the upper zone
free water storage, 4) supplemental baseflow, and 5) primary baseflow. The first three
runoff components represent the total channel inflow whereas the latter two are the total
baseflow. In the NWS model, the total channel inflow constitutes the surface runoff
contribution to the streamflow hydrograph routed via unit hydrograph/Muskingum
routing methods, and a portion of the total baseflow is the subsurface runoff contribution
to streamflow. This subsurface flow contribution is added to the routed streamflow at the
basin or subbasin outlet using a linear decay weighting function which is similar to the
hydrograph routing method.
The U.S. Geological Survey (USGS) developed a PrecipitationRunoff Modeling
System (PRMS) to simulate surface water runoff, sediment yields and general basin
hydrology under various combinations of precipitation, climate and land use (Leavesley et
al., 1983). The system is physically based such that each component of the hydrologic
cycle is expressed as functional relationships of known physical laws or empirical
knowledge of the hydrologic process. The PRMS model can be either a lumped or
distributed parameter model, capable of simulating mean daily flows and storm flow
hydrographs.
The watershed is partitioned into hydrologic response units (HRU's) which may or
may not be associated with individual sub watersheds. HRU's are defined based on
characteristics such as slope, orientation of hillslope or aspect, soil type, vegetation type,
Jorge A. Ramírez
41
elevation and precipitation distribution (Leavesley et al., 1983), which imply similar
hydrologic response behavior. Water and energy balances are determined daily for each
HRU. The inputs are daily precipitation, maximum and minimum daily air temperature,
daily solar radiation and daily streamflow (for calibration). Precipitation is reduced by
interception. The energy inputs of air temperature and solar radiation drive the processes
of evaporation, transpiration, sublimation and snowmelt.
The model is represented by a series of storage (reservoirs). Each HRU has a soil
zone reservoir. This represents the soil mantle that can lose water by evaporation and
transpiration. Average root depth of the predominant vegetation on the HRU defines the
depth of the soilzone reservoir. Water storage in this zone is increased by infiltration
and is decreased by evapotranspiration. The soil zone is represented by two layers. The
upper layer (recharge layer) loses water by evaporation and transpiration, whereas the
lower layer loses water only by transpiration.
When the soil zone reservoir reaches field capacity, excess water percolates to the
subsurface and groundwater reservoirs. The subsurface reservoir responds faster than the
groundwater reservoir. The subsurface reservoir can be linear or nonlinear, whereas the
groundwater reservoir responds linearly. Subsurface and groundwater reservoirs can be
defined for each HRU or can be defined for larger areas comprising several HRU's. The
snow component of PRMS simulates the accumulation and depletion of a snowpack on
each HRU.
The USGS model can simulate basin hydrology on a daily and on a shorter time
interval for flood simulation. The daily mode simulates hydrologic components on daily
averages or total values. Shorter time simulation, with time steps in the order of minutes,
computes water discharge and sediment yield from selected rainfall events. Normally the
model operates on a daily mode until it reaches dates of storm events. In these cases, the
model shifts to the shorter time interval mode until the storm period terminates. The
model then returns to the daily mode. Leavesley and Stannard (1990) used ARC/INFO
with PRMS to simulate hydrographs in the East Fork Carson River, and Luellwitz (1991)
used ARC/INFO with PRMS to simulate hydrographs in a sub alpine watershed in
Colorado. This model has been incorporated into the Modular Modeling System of the
U.S.G.S., which is briefly described in the next section.
Current Developments
Flood hydrology has developed very rapidly after the works of Sherman and
Horton in the 1930's and 40's. This has been the result of improvements in our
understanding of the fundamental laws governing hydrologic process behavior and of
improvements in computational power and observational capabilities. Parallel to the
development of increasingly more powerful computer hardware has been the
development of computer programming languages, database systems, and graphics and
image processing tools which have made easier the use of large amounts of input data and
have enhanced the analysis and interpretation of the output.
Jorge A. Ramírez
42
The development of digital terrain models (DTM) and data base systems for
spatial data, socalled Geographic Information Systems (GIS), has had a tremendous
impact on our ability to describe and understand the effects of highly heterogeneous
boundary conditions on hydrologic response. GIS makes it possible to integrate efficiently
not only the topography, but the geomorphology, soil type, vegetation and land use
characteristics of the basin, with physicallybased hydrologic models of watersheds.
These geomorphoclimatic and hydraulic characteristics, whose extreme variability in
space and time has profound nonlinear effects on hydrologic response, have become
increasingly available (e.g., NEXRAD radar precipitation fields) with the advent and
rapid development of remote sensing. GRASS of the U.S. Corps of Engineers and
ARC/INFO of Environmental Systems Research Institute of California, USA, are two
examples of GIS packages widely used in conjunction with DTM's, remotely sensed data
(i.e., precipitation), and hydrologic modeling.
The Modular Modeling System (MMS) (Leavesley et al, 1996) is an integrated
system of computer software that has been developed with financial support from the
U.S.G.S. to provide the research and operational framework needed to support
development, testing, and evaluation of physicalprocess algorithms and to facilitate
integration of userselected sets of algorithms into operational physicalprocess models.
As such, the MMS integrates all of the above referred to improvements in computer
software and hardware within a modeling framework aimed at facilitating and improving
hydrologic modeling. MMS uses a module library that contains modules for simulating a
variety of water, energy, and biogeochemical processes (e.g., it includes the PRMS
model). A model is created by selectively coupling the most appropriate modules from
the library to create a suitable model for the desired application. Where existing modules
do not provide appropriate process algorithms, new modules can be developed. The MMS
modeling system also provides installation capabilities for new module development, and
model development and application using the MMS graphical user interface. The MMS
enhances the ability of the modeler to address the specific conditions under consideration,
as it allows building of different model configurations, either from a bank of hydrologic
simulation modules, or by an explicit coding of new modules. As an example, Epstein
and Ramírez (1993, 1994) used the GRASS GIS package with PRMS in an earlier
version of the MMS to simulate streamflow hydrographs in the Rio Grande basin. Their
work addresses issues of climatic variability and its impact on hydrologic response.
During the last decade, recognition of the significant impact of spatial variability
on hydrologic response has prompted renewed interest in the development of spatially
distributed hydrologic models for flood routing and for flood risk analyses. In addition to
describing the complex interactions of local hillslope processes, variable soil and
vegetation characteristics, and channel network, spatially distributed modeling has the
advantage of providing predictions at any location throughout the river network, not only
at the outlet. This is an important advantage in planning and design problems where
predictions are required at a specific river cross section, at which perhaps no historical
observations are available.
Jorge A. Ramírez
43
Mancini and Rosso (1989) used a GIS to investigate the spatial variability of SCS
Curve Number within the basin, thus assessing the effects of such variability on basin
scale estimates. Brath et al. (1989a and 1989b) analyzed scale effects in distributed
rainfallrunoff modeling, Pilotti and Rosso (1990) presented SHELL (i.e., a general
framework for modeling the distributed response of drainage basins to storm rainfall), and
Ranzi and Rosso (1991b) reported a physically based approach to modeling distributed
snowmelt in alpine catchments.
Burlando et al. (1994) developed a methodology for the evaluation of flood risk in
a distributed modeling framework. Their flood risk analyzer (FLORA) combines a
geomorphologic model of flood frequency with a distributed description of the basin
given in terms of soil types, land use, and topography. Runoff potential and infiltration
characteristics are described as a function of the SCS Curve Number model. The
geomorphologic model of flood frequency is based on the ideas of Eagleson (1972)
introduced earlier in this chapter. In their application, Burlando et al. use the approach of
Bacchi and Rosso (1988) to derive the second order properties of the frequency
distribution of flood peaks for any point of the basin as a function of the stochastic
properties of precipitation, and of the soil type, land use, and geomorphologic
characteristics of the section in question. Assuming an Independent Poisson Marks model
for precipitation (Eagleson, 1972), and the Soil Conservation Service Curve Number
model for infiltration losses, they show that the frequency distribution of the annual
maxima is an Extreme Value Type II distribution, and obtain expressions for its second
order moments as a function of climate and geomorphology. They applied this
methodology successfully to the Sansobbia river basin in Northern Italy.
Physically based distributed modeling of watershed processes has become
increasingly feasible in recent years. In addition to the development of improved
computational capabilities, DTM's and GIS tools, this has been brought about by
advances in our understanding of the fundamental physical processes underlying the
hydrologic cycle and of the solution of the mathematical equations representing those
processes. Generally, the various hydrologic processes governing the transformation of
rainfall into runoff are extremely complex, nonlinear, and spatially and temporally
variable, giving rise to complex sets of strongly coupled nonlinear differential equations.
Analytical solutions for these equations have been found for the simplest cases only, and
one dimensional and two dimensional numerical solution schemes have been developed
and applied. For example, Saghafian (1992), Ogden (1992), Ogden and Julien (1993),
Ogden et al. (1995), and Saghafian et al. (1995) developed two dimensional numerical
schemes for routing the excess rainfall through the overland phase of a watershed. The
CASC2D model (e.g., Saghafian et al., 1995), is a distributedparameter, two
dimensional, unsteady diffusion wave model that routes overland flow using a diffusion
wave approximation for the flow dynamics. The equations of motion are solved using a
finite difference scheme. The model includes continuous soilmoisture accounting,
rainfall interception, infiltration, surface and channel runoff routing, soil erosion and
sediment transport. Over each pixel of the domain, flow is allowed to take place only one
dimensionally and in the direction of the steepest slope of either of 2 orthogonal
directions in an orthogonal cascade of planes. Input to each pixel is in the form of
Jorge A. Ramírez
44
precipitation and upstream inflow from all directions. Infiltration losses are diagnosed at
each pixel using the GreenAmpt infiltration equations. CASC2D is one of the surface
water hydrologic models support by the Watershed Modeling System (WMS) under
development at Brigham Young University. More recently, Fiedler and Ramírez (1998)
developed a fully unsteady, fully dynamic, 2dimensional hydrodynamic wave model for
the simulation of the rainfallrunoff process that allows for fully interactive infiltration. In
addition to the explicit consideration of fully interactive infiltration, Fiedler and Ramírez
developed a numerical scheme capable of describing the flow dynamics as affected by
microtopographic conditions and other rapidly varying soil and vegetation characteristics.
As such, this capability allows for consideration of issues of scale and takes into account
the impact of microchannel formation (e.g., rills, etc.) on runoff production and flood
dynamics.
Jorge A. Ramírez
45
REFERENCES
Abbott, M. B. (1986a). An Introduction to the European Hydrological System – Système
Hydrologique Europèen, SHE – 1: History and Philosophy of a PhysicallyBased,
Distributed Modelling System. Journal of Hydrology, 87, 4559.
Abbott, M. B. (1986b). An Introduction to the European Hydrological System – Système
Hydrologique Europèen, SHE – 2: Structure of a PhysicallyBased, Distributed
Modelling System. Journal of Hydrology, 87, 6177.
Adom, D. N., Bacchi, B., Brath, A., & Rosso, R., (1989). On the geomorphoclimatic
derivation of flood frequency (peak and volume) at the basin and regional scale. In
New Directions for Surface Water Modeling, eds. M.L. Kavvas, pp.165176, IAHS
Publ. no.181.
Agnese, C., D'Asaro, F. & Giordano, C., (1988). Estimation of the time scale of the
geomorphologic instantaneous unit hydrograph from effective streamflow velocity.
Water Resources Research, 24(7), 969978.
Anderson, E. A., (1971). FORTRANIV Program for Stanford Watershed Model IV.
NOAA, NWS, Office of Hydrology, Silver Springs, MD., February.
Bacchi, B., & Rosso, R., (1988). Analisi Geomorfoclimatica dei Modeli di
Regionalizzazione della Frequenza delle Piene. In Proc. XXI Ital. Conf. On
Hydraulics and Water Engineering, L'Aquila, September 58, 1, pp. 1527.
Bras, R. L., (1990). Hydrology, an Introduction to Hydrologic Science. Addison Wesley.
Brath, A., La Barbera, P., Mancini, M. & Rosso, R., (1989a). Analysis of Scale Effects in
Distributed RainfallRunoff Modeling. In Modeling and Simulation, eds. W.G.
Vogt & M.H. Mickle, Vol.20, part 4, pp. 15211526.
Brath, A., La Barbera, P., Mancini, M. & Rosso, R., (1989b). The use of distributed
rainfallrunoff models based on GIS at different scales of information. Proceedings
3rd National Conference on Hydraulic Engineering, pp.448453, Am. Soc. Civ.
Engnrs., New Orleans, August 1418.
Burnash, R. J. C., Ferral, R. L., and McGuire, R. A., (1973). A Generalized Streamflow
Simulation System: Conceptual Modeling for Digital Computers. U.S. Department
of Commerce, National Weather Service and State of California, Department of
Water Resources, Sacramento, California.
Burlando, P., Mancini, M., & Rosso, R., (1994). FLORA: A Distributed Flood Risk
Analyzer. In Computer Support for Environmental Impact Assessment eds. G.
Guariso and B. Page, Elsevier Science B. V. North Holland, Amsterdam, pp. 91
102.
Jorge A. Ramírez
46
Caroni, E. & Rosso, R., (1986). A Comparison between direct and indirect estimation of
the Nash model of catchment response. Proceedings International Conference on
Hydrological Processes in the Catchment, Cracow, May 8 11, Vol.2, pp.2734.
Caroni, E., Rosso, R., & Siccardi, F., (1986). Nonlinearity and time variance of the
hydrologic response of a small mountain creek. In Scale Problems in Hydrology, eds.
V.K. Gupta, I. RodriguezIturbe & E.F. Wood, pp.19 38, Boston, Mass: D.Reidel
Publ. Co.
Chow V. T., (ed.) (1964). Handbook of Applied Hydrology. New York: McGraw Hill.
Chow, V.T., Maidment, D., and Mays, L. W., (1988). Applied Hydrology. McGraw Hill.
Claborn, B.J. and Moore, W. (1970). Numerical Simulation in Watershed Hydrology.
Hydraulic Engineering Laboratory, University of Texas, Rep. No. HYD 147001.
Clark, C.O. (1945). Storage and The Unit Hydrograph. Proc. Am. Soc. Civ. Eng., vol. 9,
pp13331360.
Crawford, N.H. and Linsley, R.K. Jr. (1966). Digital Simulation in Hydrology. Stanford
Watershed Model IV. Dept. of Civil Engineering, Stanford University, Stanford,
Calif., Tech. Rep. No. 39, July.
Croley, T. E. II. (1985). Forecasting Great Salt Lake levels. Proceedings, Problems and
Prospects for Predicting Great Salt Lake Levels. (eds.) P.A. Kay and H.F. Diaz, pp.
179188. University of Utah, Salt Lake City, UT, March 2628. Center for Public
Affairs and Administration.
Cunge, J.A. (1969). “On the Subject of a Flood Propagation Method (Muskingum
Method).” Journal of Hydraulic Research, International Association of Hydraulic
Research, 7(2), 205230.
Dawdy, D.R. and O'Donnell, T. (1965). Mathematical models of catch water behavior.
Proc. of the ASCE, Journal of the Hydraulics Div. Vol. 91, N. HY4, p. 123137.
Dawdy, D.R., Lichty, R.W. and Bergman, J.M. (1972). A rainfallrunoff simulation
model for estimation of flood peaks for small drainage basins. U.S. Geological Survey
Professional Paper 506B, 28 pp.
DíazGranados, M. A., Bras, R. L., and Valdés, J. B. (1983a). Incorporation of Channel
Losses in the Geomorphologic IUH. R. M. Parsons Laboratory T. R. No: 293, MIT.
DíazGranados, M. A., Valdés, J. B., and Bras, R. L. (1983). A Derived Flood Frequency
Distribution Based on the Geomorphoclimatic IUH and the Density Function of
Rainfall Excess. R. M. Parsons Laboratory T. R. No. 292, MIT.
Jorge A. Ramírez
47
DíazGranados, M. A., Valdés, J. B., and Bras, R. L. (1984). A PhysicallyBased Flood
Frequency Distribution. Water Resources Research, 20(7), 9951002.
Donigian, A. S., Jr., and Crawford, N. H., (1979). User's Manual for the Nonpoint Source
(NPS) Model. Environ. Res. Info. Center, U. S. Environmental Protection Agency,
Cincinnati, Ohio.
Dooge, J. C. I. (1959). A General Theory of the Unit Hydrograph. Journal of Geophysical
Research, 64(2): 241256.
Eagleson, P. S. (1970). Dynamic Hydrology. McGraw Hill.
Eagleson, P. S. (1972). Dynamics of Flood Frequency. Water Resources Research, 8(4),
878898.
Epstein, D. and Ramírez, J. A. (1993). A Daily Spatial Disaggregation Approach and its
Application in Hydrologic Sensitivity Analysis of the Upper Rio Grande to Climate
Variability. American Geophysical Union Front Range Meeting, Boulder, February
8  10.
Epstein, D., and Ramírez, J. A. (1994). Spatial Disaggregation for Studies of Climatic
Hydrologic Sensitivity. ASCE Journal of the Hydraulics Division 120(12): 1449
1467.
Fread, D.L. (1974). Numerical Properties of Implicit FourPoint Finite Difference
Equations of Unsteady Flow, HRL45, NOAA Tech. Memo NWS HYDRO18,
Hydrologic Research Laboratory, National Weather Service, Silver Spring, Maryland.
Fread, D.L. (1976). Theoretical Development of Implicit Dynamic Routing Model, HRL
113, Hydrologic Research Laboratory, National Weather Service, Silver Spring,
Maryland.
Fread, D.L. (1978). NWS Operational Dynamic Wave Model, in Verification of
Mathematical and Physical Models, Proceedings of 26
th
Annual Hydr. Div. Specialty
Conf., ASCE, College Park, Md., pp. 455464.
Fread, D.L. (1988). The NWS DAMBRK Model: Theoretical Background/User
Documentation, HRL256, National Weather Service, Silver Spring, Maryland.
Fread, D.L., and Lewis, J. M. (1988). FLDWAV: A Generalized Flood Routing Model,
Proc. National Conference on Hydraulic Engineering, ASCE, Colorado Springs,
Colorado, pp. 668673.
Fiedler, F. R., and Ramírez, J. A. 1998. A Numerical Method for Hydrodynamic
Modeling of Overland Flow. Intl. Jour. for Numerical Methods in Fluids. (Accepted
for publication).
Jorge A. Ramírez
48
Gupta, V. K. and Mesa, O. J. (1988). Runoff Generation and Hydrologic Response via
Channel Network Geomorphology: Recent Progress and Open Problems. Journal of
Hydrology, 102(14), 3  28.
Gupta, V. K., aand Waymire, E. (1983). On the Formulation of an Analytical Approach to
Hydrologic Response and Similarity at the Basin Scale. Journal of Hydrology, 65, 95
123.
Gupta, V. K., Waymire, E., and RodríguezIturbe, I. (1986). On Scales, Gravity, and
Network Structure in Basin Runoff. In: V. K. Gupta, I. RodríguezIturbe, and E. F.
Wood, eds. Scale Problems in Hydrology. Dordrecht, Holland: D. Reidel.
Hebson, C. and Wood, E. F. (1982). A derived Flood Frequency Distribution Using
Horton Order Ratios. Water Resources Research, 18(5), 15091518.
Howard, A. D. (1990). Theoretical Model of Optimal Drainage Networks. Water
Resources Research, 26(9), 21072117.
Hudlow, M.D. and Clark, R. A. (1969). Hydrologic Synthesis by Digital Computer.
Journal of the Hydraulics Division, Proceedings of the American Society of Civil
Engineers, 95(HY3), pp. 839860.
HEC  Hydrologic Engineering Center, (1981). The New HEC1 Flood Hydrograph
Package. Tech. Paper No. 82, Army Corps of Engineers, Calif.
Johanson, R. C., Imhoff, J. C., and Davis, H. H. (1980). User's Manual for Hydrologic
Simulation ProgramFORTRAN (HSPF). EPA600/980015, U.S. EPA, Athens,
Georgia.
Karlinger, M. R. and Troutman, B. M. (1985). Assessment of the Instantaneous Unit
Hydrograph Derived from the Theory of Topologically Random Networks. Water
Resources Research, 21(11), 16931702.
Kibler, D. F. (1982). Desktop Methods for Urban Stormwater Calculation. Chapter 4. In
D. F. Kibler (eds.) Urbam Stormwater Hydrology, Water Resources Monograph 7,
American Geophysical Union, Washington, DC.
Kirshen, D. M. and Bras, R. L. (1983). The Linear Channel and its Effect on the
Geomorphologic IUH. Journal of Hydrology, 65, 175208.
La Barbera, P. & Rosso, R. (1987). Fractal geometry of river networks (abstract), EOS
Trans. AGU, 68(44), 1276.
La Barbera, P. & Rosso, R. (1989). On the fractal dimension of stream networks, Water
Resources Research, 25(4), 735741.
Jorge A. Ramírez
49
Leavesley, G.H., Lichty, R.W., Troutman, B.M. and Saindon, L.G. (1983). Precipitation
Runoff Modeling System: User's Manual, U.S. Geological Survey WaterResources
Investigations Report, 834328, 207 pp.
Leavesley, G.H., Restrepo, P. J., Markstrom, S. L., Dixon, M., and Stannard, L. G. (1996)
The Modular Modeling System (MMS): User's Manual, (Updated March 1998). U.S.
Geological SurveyOpen File Report 96151.
Leavesley, G. H. and Stannard, L. G. 1990. Application of Remotely Sensed Data in a
Distributed Parameter Watershed Model U.S. Geological Survey Water Resources
Investigations.
Liou, E. Y. (1970). Opset: Program for Computerized Selection of Watershed Parameter
Values of the Stanford Watershed Model. Research Dept. No. 34, Water Resour. Inst.,
University of Kentucky, Lexington.
Luellwitz, T. (1991). The Application of a Deterministic Hydrologic Model and GIS to
Simulate Water Yield Increase due to Timber Harvest in a Subalpine Watershed.
Master of Science Report, Department of Earth Resources, Colorado State University,
Fort Collins.
Mancini, M. & Rosso, R. (1989). Using GIS to assess spatial variability of SCS Curve
Number at the basin scale, in: New Directions for Surface Water Modeling, edited by
M.L. Kavvas, IAHS Publ. no.181, pp.435444.
Mesa, O. J. and Mifflin, E. R. (1986). On the Relative Role of Hillslope and Network
Geometry in Hydrologic Response. In: V. K. Gupta, I. RodríguezIturbe, and E. F.
Wood, eds. Scale Problems in Hydrology. Dordrecht, Holland: D. Reidel.
Metcalf and Eddy, (1971). Storm Water Management Model. Vol.1. Environmental
Protection Agency, Water Resources Engineers, University of Florida, Gainesville,
Florida.
Molnár, P. and Ramírez, J. A. (1998a). Energy Dissipation Theories and Optimal Channel
Characteristics of River Networks. Water Resources Research. 37(7), 18091818.
Molnár, P. and Ramírez, J. A. (1998b). An Analysis of Energy Expenditure and
Downstream Hydraulic Geometry at Goodwin Creek. Water Resources Research.
37(7), 18191829.
Monro, John C., And Anderson, Eric A., "National Weather Service River Forecasting
System, " (Proc. Paper 10549) Journal Of The Hydraulics Division, Vol. 100, No.
Hy5, Asce, May 1974, Pp. 621630, May 1, 1974.
Morris, E. M., (1980). Forecasting Flood Flows in Grassy and Forested Basins Using a
Deterministic Distributed Mathematical Model. International Association of Scientific
Hydrology Publication 129, 247255.
Jorge A. Ramírez
50
Nash, J. E. (1957). The Form of the Instantaneous Unit Hydrograph. International
Association of Scientific Hydrology Publication, 45(3), 114121.
Nash, J. E. (1958). Determining Runoff from Rainfall. Proceedings of the Institution of
Civil Engineers, 10, 163184.
Nash, J. E. (1959). Systematic Determination of Unit Hydrograph Parameters. Journal of
Geophysical Research, 64(1), 111115.
Nash, J. E. (1960). A Unit Hydrograph Study, with Particular Reference to British
Catchments. Proceedings of the Institution of Civil Engineers, 17, 249282.
Ogden, F.L. (1992). Twodimensional runoff modeling with weather radar data, Ph.D.
Dissertation, Dept. of Civil Engineering, Colorado State University, Fort Collins, CO
80523, 208 pp.
Ogden, F.L., and Julien, P. Y. (1993). Runoff sensitivity to temporal and spatial rainfall
variability at runoff plane and small basin scales. Water Resources Research, 29(8),
25892597.
Ogden, F.L., Richardson, J. R. and Julien, P. Y. (1995). Similarity in catchment response
2. Moving rainstorms. Water Resources Research, 31(6), 15431547.
Peck, E. L. (1976). Catchment Modeling and Initial Parameter Estimation for the
National Weather Service River Forecasting System. NOAA Technical Memorandum
NWS Hydro31, Washington, DC.
Pilgrim, D.H. (1976). Travel times and non linearity of flood runoff from tracer
measurements on a small watershed, Water Resources Research, 12(4), 487496.
Pilgrim, D.H. (1977). Isochrons of travel time and distribution of flood storage from a
tracer study on a small watershed. Water Resources Research, 13(3), 587595.
Pilotti, M. & Rosso, R. (1990). Shell: A general framework for modeling the distributed
response of a drainage basin, in: Computational Methods in Surface Hydrology,
edited by G. Gambolati, A. Rinaldo, C. Brebbia, W.G. Gray & G.F. Pinder, Springer
Verlag, Berlin, pp.517522.
Ramírez, J. A., Salas, J. S. and R. Rosso, R. 1994: Determination of Flood Characteristics
by PhysicallyBased Methods. Chapter 6 NATO Advanced Study Institute on Coping
with Floods. Eds. Giuseppe Rossi, Nilgun Hamancioglu and Vujica Yevjevich;
Kluwer Academic Publishers; pp. 77110.
Ranzi, R. & Rosso, R. (1991a) : Flea: Flood Event Analyzer, in: Guida al Software
Ambientale, edited by G. Guariso, pp.121126, Patron, Bologna, (in italian).
Jorge A. Ramírez
51
Ranzi, R. & Rosso, R. (1991b). A physically based approach to modeling distributed
snowmelt in a small alpine catchment, in: Snow Hydrology and Forests in High
Alpine Areas, edited by H. Bergmann, H. Lang, W. Frey, D. Issler & B. Salm, IAHS
Publ. no.205, pp.141150.
Ramírez, J. A., 2000: Prediction and Modeling of Flood Hydrology and Hydraulics.
Chapter 11 of Inland Flood Hazards: Human, Riparian and Aquatic Communities
Eds. Ellen Wohl; Cambridge University Press.
Ricca V.T. (1972). The Ohio State University Version of the Stanford Streamflow
Simulations Model, Part I  Technical Aspects, Ohio State University, Columbus,
May.
Rigon, R., Rinaldo A., RodríguezIturbe, I., Bras, R.L., and IjjászVásquez, E.J. (1993)
“Optimal Channel Networks: A Framework for the Study of River Basin
Morphology,” Water Resources Research, 29(6), 16351646.
Rinaldo, A., Marani, A. & Rigon, R.(1991). Geomorphologic dispersion, Water
Resources Research, 27(44), 513525.
RodríguezIturbe, I., Rinaldo, A., Rigon, R., Bras, R.L., Marani, A., and IjjászVásquez,
E.J. (1992). Energy Dissipation, Runoff Production, and the ThreeDimensional
Structure of River Basins. Water Resources Research, 28(4), 10951103.
RodríguezIturbe, I., González Sanabria, M. and Bras, R. L.(1982a). A
Geomorphoclimatic Theory of the Instantaneous Unit Hydrograph. Water Resources
Research, 18(4), 877886.
RodríguezIturbe, I., González Sanabria, M. and Caamaño, G. (1982b). On the Climatic
Dependence of the IUH: A RainfallRunoff Analysis of the Nash Model and the
Geomorphoclimatic Theory. Water Resources Research, 18(4), 887903.
RodríguezIturbe, I., Devoto, G. and Valdés, J. B. (1979). Discharge Response Analysis
and Hydrologic Similarity: The Interrelation Between the Geomorphologic IHU and
the Storm Characteristics. Water Resources Research, 15(6), 14351444.
RodríguezIturbe, I., and Valdés, J. B. (1979). The Geomorphologic Structure of
Hydrologic Response. Water Resources Research, 15(6), 14091420.
Rosso, R. (1984). Nash Model Relation to Horton Order Ratios. Water Resources
Research, 20(7), 914920.
Rosso, R. & Caroni, E.(1986). Analysis, estimation and prediction of the hydrological
response from catchment geomorphology, Excerpta, 1, 93108.
Jorge A. Ramírez
52
Saghafian, B. (1992). The Effect of Spatially Varied Characteristics on Watershed
Response: A TwoDimensional Distributed Approach. Ph. D. Dissertation, Civil
Engineering Department, Colorado State University, Fort Collins, CO.
Saghafian, B., Julien, P. Y. and Ogden, F. L.1995, Similarity in catchment response 1.
Stationary rainstorms. Water Resources Research, 31(6), 15331541.
Sherman, L. K. (1932). Streamflow from Rainfall by the Unit Graph Method. Eng. News
Rec., 108, 501505.
Shreve, R. L. (1967). Infinite Topologically Random Channel Networks. J. Geol., 75,
178186.
Singh, V. P., (1989). Hydrologic Systems. Watershed Modeling. Volume II. Prentice Hall,
Englewood Cliffs, New Jersey.
Snyder, F.F. (1938). Synthetic Unit Graphs. Trans. Am. Geophys. Union, 19, 447454.
Soil Conservation Service, Hydrology, (1972). Sec. 4 of National Engineering Handbook,
Soil Conservation Service, U.S. Department of Agriculture, Washington, D.C.
Troutman, B. M. and Karlinger, M. R. (1984). On the Expected Width Function for
Topologically Random Channel Networks. Journal of Applied Probability, 21, 836
884.
Troutman, B. M. and Karlinger, M. R. (1985). Unit Hydrograph Approximation
Assuming Linear Flow Through Topologically Random Channel Networks. Water
Resources Research, 21(5), 743754.
Troutman, B. M. and Karlinger, M. R. (1986). Averaging Properties of Channel Networks
Using Methods in Stochastic Branching Theory. In: V. K. Gupta, I. RodríguezIturbe,
and E. F. Wood, eds. Scale Problems in Hydrology. Dordrecht, Holland: D. Reidel.
Valdés, J. B., Fiallo, Y., and RodríguezIturbe, I. (1979). A RainfallRunoff Analysis of
the Geomorphologic IUH. Water Resources Research, 15(6), 14211434.
U. S. Army Corps of Engineers, (1972). Program Description and User's Manual for
SSARR Model Streamflow Synthesisi and Reservoir Regulation. Program 724K5
G0010.
U. S. Soil Conservation Service, (1973). Computer Program for Project Formulation
Hydrology. Technical Release No. 20, U.S. Department of Agriculture, Washington,
D.C.
U. S. Soil Conservation Service, (1985), National Engineering Handbook, Sec. 4,
Hydrology, U.S. Department of Agriculture, Washington, D.C.
Jorge A. Ramírez
53
Woolhiser, D.A. and Liggett, J. A. (1967). Unsteady OneDimensional Flow over a
Plane: The Rising Hydrograph. Water Resources Research, 3(3), 753771.
Jorge A. Ramírez
a) evaluation of the rainfall flux over the watershed I(x, t) as a function of space and time; b) evaluation of the rainfall excess or effective rainfall flux as a function of space and time, Ie(x, t). Effective rainfall is the rainfall available for runoff after infiltration and other abstractions have been accounted for; and c) routing of the rainfall excess to the watershed outlet in order to determine the corresponding flood hydrograph, Q(t). Hydrologic flood prediction models may be categorized into physical models and mathematical models. Mathematical models describe the system behavior in terms of mathematical equations representing the relationships between system state, input and output. Mathematical models can, in turn, be categorized as either purely conceptual models or physicallybased models. Depending on whether the functions relating input, output and system state are functions of space and time, these models may be further categorized as lumped models or distributed models. Lumped models do not account explicitly for the spatial variability of hydrologic processes, whereas distributed models do. Lumped models use averages to represent spatially distributed function and properties. UNIT HYDROGRAPH ANALYSIS Sherman (1932) first proposed the unit hydrograph concept. The Unit Hydrograph (UH) of a watershed is defined as the direct runoff hydrograph resulting from a unit volume of excess rainfall of constant intensity and uniformly distributed over the drainage area. The duration of the unit volume of excess or effective rainfall, sometimes referred to as the effective duration, defines and labels the particular unit hydrograph. The unit volume is usually considered to be associated with 1 cm (1 inch) of effective rainfall distributed uniformly over the basin area. The fundamental assumptions implicit in the use of unit hydrographs for modeling hydrologic systems are: a) Watersheds respond as linear systems. On the one hand, this implies that the proportionality principle applies so that effective rainfall intensities (volumes) of different magnitude produce watershed responses that are scaled accordingly. On the other hand, it implies that the superposition principle applies so that responses of several different storms can be superimposed to obtain the composite response of the catchment. b) The effective rainfall intensity is uniformly distributed over the entire river basin. c) The rainfall excess is of constant intensity throughout the rainfall duration. d) The duration of the direct runoff hydrograph, that is, its time base, is independent of the effective rainfall intensity and depends only on the effective rainfall duration.
2
Jorge A. Ramírez
When the effective rainfall is given as a hyetograph, that is, as a sequence of M rainfall pulses of the same duration ∆t, the corresponding direct runoff hydrograph can be expressed as the discrete convolution of the rainfall hyetograph and the Unit Hydrograph as,
Qn = ∑ PmUn −m + 1 , m = min(n, M)
* m= 1 m*
(1a) (1b) (1c)
Qn = Q(n∆t ),
m ∆t
n = 1,L, N m = 1,L, M
P = m
e (m −1) ∆t
∫ I (τ )dτ ,
where Pm is the volume of the mth effective rainfall pulse, Qn is the direct runoff, and Unm+1 are the Unit Hydrograph ordinates. Although the above assumptions lead to acceptable results, watersheds are indeed nonlinear systems. For example, unit hydrographs derived from different rainfallrunoff events, under the assumption of linearity, are usually different, thereby invalidating the linearity assumption. The determination of unit hydrographs for particular basins can be carried out either using the theoretical developments of linear system theory; or using empirical techniques. For either case, simultaneous observations of both precipitation and streamflow must be available. These two approaches are presented in more detail in later sections. Hydrograph Components Total streamflow during a precipitation event includes the baseflow existing in the basin prior to the storm and the runoff due to the given storm precipitation. Total streamflow hydrographs are usually conceptualized as being composed of: a) Direct Runoff, which is composed of contributions from surface runoff and quick interflow. Unit hydrograph analysis refers only to direct runoff. b) Baseflow, which is composed of contributions from delayed interflow and groundwater runoff. Surface runoff includes all overland flow as well as all precipitation falling directly onto stream channels. Surface runoff is the main contributor to the peak discharge. Interflow is the portion of the streamflow contributed by infiltrated water that moves laterally in the subsurface until it reaches a channel. Interflow is a slower process
3
Jorge A. Ramírez
than surface runoff. Components of interflow are quick interflow, which contributes to direct runoff, and delayed interflow, which contributes to baseflow (e.g., Chow, 1964.) Groundwater runoff is the flow component contributed to the channel by groundwater. This process is extremely slow as compared to surface runoff. Schematically in Figure 11.1, the streamflow hydrograph is subdivided into a) Rising Limb: rising portion of the hydrograph, composed mostly of surface runoff. b) Crest: zone of the hydrograph around peak discharge. c) Falling (or Recession) Limb: Portion of the hydrograph after the peak discharge, composed mostly of water released from storage in the basin. The lower part of this recession corresponds to groundwater flow contributions. The main factors affecting hydrograph shape are: 1) Drainage characteristics: basin area, basin shape, basin slope, soil type and land use, drainage density, and drainage network topology. Most changes in land use tend to increase the amount of runoff for a given storm (e.g., Chow et al., 1988; Singh, 1989; Bras, .1990). 2) Rainfall characteristics: rainfall intensity, duration, and their spatial and temporal distribution; and storm motion, as storms moving in the general downstream direction tend to produce larger peak flows than storms moving upstream (e.g., Chow et al., 1988; Singh, 1989; Bras, .1990). Hydrographs are also described in terms of the following time characteristics (see Figure 11.1): Time to Peak, tp: Time from the beginning of the rising limb to the occurrence of the peak discharge. The time to peak is largely determined by drainage characteristics such as drainage density, slope, channel roughness, and soil infiltration characteristics. Rainfall distribution in space also affects the time to peak. Time of Concentration, tc: Time required for water to travel from the most hydraulically remote point in the basin to the basin outlet. For rainfall events of very long duration, the time of concentration is associated with the time required for the system to achieve the maximum or equilibrium discharge. Kibler (1982) and Chow et al. (1988) summarize several of many empirical and physicallybased equations for tc that have been developed. The drainage characteristics of length and slope, together with the hydraulic characteristics of the flow paths, determine the time of concentration. Lag Time, tl: Time between the center of mass of the effective rainfall hyetograph and the center of mass of the direct runoff hydrograph.
4
Jorge A. Ramírez
The basin lag is an important concept in linear modeling of basin response. The lag time is a parameter that appears often in theoretical and conceptual models of basin behavior. However, it is sometimes difficult to measure in real world situations. Many empirical equations have been proposed in the literature. The simplest of these equations computes the basin lag as a power function of the basin area. Time Base, tb: Duration of the direct runoff hydrograph. Baseflow separation. As the Unit Hydrograph concept applies only to direct runoff, the direct runoff must be separated from the baseflow. Baseflow separation or hydrograph analysis is the process of separating the direct runoff (surface runoff and quick interflow) from the baseflow. This separation is somewhat arbitrary, but corresponds to theoretical concepts of basin response. Subjective methods. Several subjective methods are shown in Figure 11.1. The simplest one consists in arbitrarily selecting the discharge marking the beginning of the rising limb as the value of the baseflow and assuming that this baseflow discharge remains constant throughout the storm duration. A second method consists in arbitrarily selecting the beginning of the groundwater recession on the falling limb of the hydrograph (usually assumed to occur at a theoretical inflection point) and connecting this point by a straight line to the beginning of the rising limb. A third example of subjective methods consists in extending the recession prior to the storm by a line from the beginning of the rising limb to a point directly beneath the peak discharge and then connecting this point to the beginning of the groundwater recession on the falling limb. Area method. The area method of baseflow separation consists in determining the beginning of the baseflow on the falling limb with the following empirical equation,
N = bA 0.2
(2)
relating the time in days from the peak discharge, N, to the basin area, A. When A is in square miles, b equals 1. When A is in square kilometers, b equals 0.8. This equation is unsuitable for smaller watersheds and should be checked for a number of hydrographs before using. The master recession curve method. This method consists in modeling the response of the groundwater aquifer as a linear reservoir of parameter k. This assumption leads to the following equation for the groundwater recession hydrograph,
Q(t) = Q(to )e −(t −t o ) / k
(3)
where Q(t) is the baseflow at time t; Q(to) is a reference baseflow discharge at time to, and k is the recession constant for baseflow. This method is based on a linear reservoir model of unforced basin response (that is, response from storage) and it can be used to separate
5
It involves the determination of several recession constants. 2) Compute the total volume of direct runoff and convert this volume into equivalent depth of effective rainfall (in centimeters or in inches) over the entire basin.1: Schematic Description of Hydrograph UNIT HYDROGRAPHS: EMPIRICAL DERIVATION The following are essential steps in deriving a unit hydrograph from a single storm: 1) Separate the baseflow and obtain the direct runoff hydrograph (DRH). 3) Normalize the direct runoff hydrograph by dividing each ordinate by the equivalent volume (in or cm) of direct runoff (or effective rainfall).Jorge A. and groundwater aquifer storage. 6 . Effective precipitation Streamflow hydrograph Baseflow hydrographs tl tp tb Figure 11. Ramírez the contributions to the recession flow from surface storage. subsurface storage.
The first step is to determine the Scurve hydrograph by adding a series of (known) UH’s of duration D. obtain the effective rainfall hyetograph (e. The hydrologic response of such systems can be expressed in terms of an impulse response function (IRF) through a socalled Convolution Equation. use the φindex. inflow. Linear systems possess the properties of additivity and proportionality. SCurve Hydrograph Method An Shydrograph represents the response of the basin to an effective rainfall event of infinite duration. This duration is the duration associated with the unit hydrograph. each lagged by a time interval D. The resulting superposition represents the runoff resulting from a continuous rainfall excess of intensity 1/D. To do this. Green and Ampt. the Horton. which are implicit in the convolution equation. or from storms which can be described as a sequence of pulses. or some other method to determine infiltration losses) and its associated duration. Lagging the Scurve in time by an amount D’ and subtracting its ordinates from the original unmodified Scurve yields a hydrograph corresponding to a rainfall event of intensity 1/D and of duration D’. Consequently. When dealing with a rainfall of different duration a new unit hydrograph must be derived for the new duration. This procedure is sometimes referred to as the Scurve Hydrograph method. They can only be used to predict direct runoff from storms of the same duration as that associated with the UH. Ramírez 4) Determine the effective duration of excess rainfall. Assume that an UH of duration D is known and that an UH for the same basin but of duration D’ is desired. UNIT HYDROGRAPHS: LINEAR SYSTEM THEORY A hydrologic system (a basin) is said to be a linear system if the relationship between storage. or Philip equations. Unit Hydrographs for Different Effective Duration A unit hydrograph for a particular watershed is developed for a specific duration of effective rainfall.. and outflow is such that it leads to a linear differential equation. The linearity property implicit in the UH analysis can be used to generate UH’s associated with larger or smaller effective rainfall pulse duration. each of the same duration as that associated with the UH (see Equation 1). Linear reservoirs. The resulting ordinates represent a unit hydrograph associated with an effective rainfall of duration D’. Unit hydrographs are fundamentally linked to the duration of the effective rainfall event producing them. for example. are special cases of a general hydrologic system model in which the storage is linearly related to the output by a constant k.Jorge A. to convert this hydrograph whose volume is D’/D into a unit hydrograph of duration D’. Impulse Response Function 7 . its ordinates must be normalized by multiplying them by D/D’.g.
It represents the runoff hydrograph from a continuous effective rainfall of unit intensity. g (t ) = ∫ u (t )dt 0 t (5) Unit Pulse Response Function The unit pulse response function (PRF) is the theoretical counterpart to the UH concept presented earlier. and Q(t) and Ie(t) represent direct runoff and excess or effective precipitation. This is analogous to the procedure presented earlier in the section on Shydrograph analysis. it has been customary to assume that watersheds behave as linear systems. an Instantaneous Unit Hydrograph represents the response of a watershed (discharge at its outlet as a function of time) to a unit volume of precipitation uniformly distributed over the basin and occurring instantaneously at time t = 0. Unit Step Response Function The unit step response function (SRF) is the theoretical counterpart to the Scurve hydrograph concept presented earlier in the empirical UH analysis section. as a sequence of rainfall pulses of the same duration ∆t. lagged by an amount ∆t. Thus. Ramírez The IRF of a linear system represents the response of the system to an instantaneous impulse (excitation) of unit volume applied at the origin in time (t=0). The response of continuous linear systems can be expressed. As can be seen from its definition. It represents the runoff hydrograph from a constant effective rainfall of intensity 1/∆t and of duration ∆t. 1 1 h(t) = [g(t) − g(t − ∆t)] = ∫ u(τ )dτ ∆t ∆ t t −∆ t t (6) From its definition. and obtained as. in the time domain. the corresponding direct runoff hydrograph can be expressed as the discrete convolution of the rainfall hyetograph and a Unit Hydrograph. Discrete Convolution Equation When the effective rainfall is given as a hyetograph. In hydrology. it is the convolution of 1 and u(t). 8 . Q(t) = ∫ Ie (τ )u(t − τ )dτ 0 t (4) where u(t) is the impulse response function of the system.Jorge A. respectively. that is. When dealing with hydrologic systems. in terms of the impulse response function via the convolution integral as follows. u(t) represents the instantaneous unit hydrograph (IUH). the PRF can be seen as the normalized difference between two lagged SRF’s (Scurve hydrographs).
m = min(n. The leastsquares approach tries to obtain a set of UH ordinates that minimizes the sum of squares of the errors. M) * m= 1 m* (1a) (1b) (1c) Qn = Q(n∆t ). NM+1 is the number of UH ordinates. Matrix P is of dimensions (N x (NM+1)). N m = 1. and N is the number of direct runoff ordinates.L. or others can be used. U n−m+1 1 = h[(n − m + 1)∆t ] = ∆t ( n − m +1) ∆t ( n − m ) ∆t ∫ u (τ )dτ (7) The UH ordinates correspond to the area under the IUH between two consecutive time intervals. [U ] = [[ P] [ P]] [ P] [ Q] T T −1 (9) 9 . and leads to the following solution for the UH. The estimation of the Unit Hydrograph from simultaneous observations of effective precipitation (Pm) and direct runoff (Qn) can be seen as a linear static estimation problem for which a method like leastsquares. The discrete convolution can be expressed alternatively as a matrix equation as. M P = m e (m −1) ∆t ∫ I (τ )dτ .L. where Pm is the volume of the mth effective rainfall pulse and the Unit Hydrograph ordinates are given by. Recasting the discrete convolution equation in this manner allows for an objective (and optimal) mathematical derivation of the UH ordinates. [P ][U ] = [Q ] P1 P2 P M P1 P2 PM Q1 Q2 U1 L U2 L P = Qn 1 M P 2 UN − M +1 L PM QN (8a) (8b) where M is the number of effective rainfall pulses. linear programming.Jorge A. m ∆t n = 1. Ramírez Qn = ∑ PmUn −m + 1 .
peak flow. 1957. Synthetic Unit Hydrograph procedures can be categorized as (e. 1972). etc. un (t) = 1 t ( ) n−1 e −t / k k Γ (n) k (12) 10 . 1959.g. 1958. or when UH’s for other locations on the stream in the same watershed or for nearby watersheds of similar characteristics are required. Thus. 1959. When the data can be expected to contain large errors a different approach may be more adequate. 4) and the impulse response function for a single linear reservoir (eq. Geomorphologic Instantaneous Unit Hydrograph). Synthetic Unit Hydrograph procedures must be used.g. Linear programming would then be the appropriate solution procedure. Soil Conservation Service.Jorge A. and 3) those based on a dimensionless unit hydrograph (e. Dooge. Ramírez where [[P] T[P]] 1 is the inverse of matrix [[P] T[P]]. Linear Reservoirs . etc. When no direct observations are available.. CONCEPTUAL (SYNTHETIC) UNIT HYDROGRAPHS As indicated earlier.g. 1959. 11). Conceptual UH’s Based on Models of Watershed Storage The Nash and Dooge Models. Using the convolution equation (eq. 1960) proposed a cascade of n equal linear reservoirs as a model on which to base the derivation of IUH’s for natural watersheds.A linear reservoir is characterized by a linear relationship between the storage and the output as. 1988): 1) those based on models of watershed storage (e.1958.g. the resultant UH is specific to the particular watershed defined by the point on the stream where the direct runoff observations were made.) to watershed characteristics (e. u(t ) = 1 −t / k e k (10) (11) Nash (1957. the IUH corresponding to the Nash Model can be easily obtained as follows. S(t) = kQ(t) The impulse response function of such linear reservoir is.). Chow et al. 2) those relating hydrograph characteristics (time to peak.. Such approach would minimize the sum of the absolute value of the errors.. and the superscript T stands for transpose.. The Nash model is one of the most widely used models in applied hydrology. Snyder.. This approach is adequate when the data (precipitation and discharge) are relatively error free or the errors can be expected to be small. Nash. 1938. sets of concurrent observations of effective rainfall and direct runoff are required for the derivation of unit hydrographs.
220 3 1. TABLE 1. Parameter values for the peak discharge equation of the linear reservoir cascade model n a b 1 1 1 2 1.e..A linear channel is characterized by continuity and momentum equations given by.. i. due to a finite pulse of rainfall excess with duration tr and uniform intensity Ie was suggested by Adom et al. the number of reservoirs in the cascade. Dooge (1959) developed a general unit hydrograph theory under the assumption that basin response can 11 . which is given by tl = nk. t −α r 1 . ∂Q ∂A + =0 ∂x ∂t A = C(x)Q (14a) (14b) For conditions of no lateral inflow. i. When rainfall and runoff data are available. which are parameters depending on n.e tl = AI e β Qmax (13) where A is the contributing area. The values of a and b.Jorge A.348 1. the unit impulse response function of such linear channel is. The cascade of linear reservoirs model can be modified in order to account for translation by using combinations of linear channels and linear reservoirs.017 1. Ramírez This equation is a two parameter gamma distribution function. the shape parameter can be estimated as the inverse of the second nondimensional moment of the IUH about the centroid (i.313 4 1. u(t ) = δ (t − T ) (15) which implies that the downstream outflow is equal to the upstream inflow delayed by a lagtime T representing the travel time through the system.e. et al. Qmax. are given in Table 1 (Ramírez. the inverse of the square of the coefficient of variation). (1989) in the form of a Weibull cumulative distribution function (CDF). where n is the shape parameter. A general form for the peak discharge.370 Linear Channels . and the scale parameter can be estimated as the ratio of the first order moment of the IUH about the origin to n. and k is the scale parameter.e. and tl is the basin lagtime. 1994).175 1.
The IUH is interpreted as the probability density function of the travel times to the basin outlet of water droplets randomly and uniformly distributed over the watershed. and. the characteristics of the outflow hydrograph and the characteristics of the hillslopes and drainage network must be intimately linked as they shape and modify each other. The fundamental postulates of the random topological model can be summarized as follows (e.g. Bras. and climatic and geomorphologic characteristics on the other. Gupta and Waymire. un (t) = 1 t − nT n −1 −(t −nT )/ k ( ) e k k Γ (n) (16) As a general rule. which uniquely define watershed behavior. natural drainage networks are topologically random. Shreve. b) interior and exterior link lengths. Conceptual UH’s Based on Relationships Between Hydrologic Response and Watershed Characteristics Flow hydrographs travel through a stream network after runoff is generated on the hillslopes. linear channels and linear reservoirs can be combined in series and/or in parallel to model complex hydrologic systems. Molnár and Ramírez. and to a dynamic parameter. Recent efforts have concentrated on establishing a theoretical link between hydrologic response on the one hand. as well as for estimating flood potential indices and flood frequency distributions. RodríguezIturbe and Valdés (1979) linked the hydrologic response of watersheds in terms of the IUH to the their geomorphologic parameters. at the time scales of interest. geomorphologic parameters have been used in synthesizing unit hydrographs for gauged and ungauged watersheds. For this model the drainage area of the watershed is divided into sub areas using the isochrones. 1983. 1998b)). In particular. The theoretical link was developed in the late seventies and early eighties and was based on the ideas of the topologically random model for drainage networks. as well as their associated areas. The GIUH.Jorge A. 1967. Geomorphologic parameters have been widely used to characterize and model hydrologic response. Therefore.g... Even though climate and geologic processes are continually changing and affecting hydrologic response. climate and geomorphology may be considered constant boundary conditions. Ramírez be represented by a cascade of linear channels and linear reservoirs in series. The resulting IUH or impulse response function for the basin is a function of the 12 . 1990): a) in the absence of natural controls. 1998a. the IUH or impulse response function is. are independent random variables. The drainage network develops so that a certain load of water and sediment (as a result of erosive processes) can be evacuated in an optimal manner (least energy expenditure (e. A linear channel in series with a linear reservoir represents each sub area. Geomorphologic Unit Hydrographs. If this is done assuming a system of n equal linear reservoirs of parameter k and n equal linear channels of lagtime T.
RA. divided by the total number of streams of order i.. The parameter of this relationship is the socalled Area Ratio. This ratio has been found in nature to be between 3 and 5. (1979) suggest using the peak velocity 13 . RL. the area ratio. Valdés et al. In this model.e. The parameter of this geometric relationship is the Bifurcation Ratio.e. The travel times in the streams are assumed to be exponentially distributed and independent of one another. Thus. R . and a characteristic length scale given by the mean length of streams of the given order. where j represents a stream of higher order. The initial probability of a drop falling in an area of order i is equal to the percent contributing area for the given order. They suggest estimating λ as the ratio of a characteristic velocity.. ω Lω = L1 RL −1 (18) The Horton law of stream areas states that there exists a geometric relationship between the average area drained by streams of a given order and the corresponding order ω. The initial probabilities of a drop landing anywhere on the basin. These ratios are the parameters of the socalled Horton Laws summarized below. Similarly. the transition probabilities from state i (stream order) to state j. streams of a given order define the states of the system.Jorge A. bifurcation ratio. as well as the transition probabilities which are required in order to define the probability function of a particular path that a drop may take to the outlet. can be defined as functions of only geomorphologic and geometric parameters. the transition probabilities are related to the average ratio of the number of streams of a given order to the number of streams of the next order (i. The initial probabilities are thus related to the average ratio of the average area of subbasins of a given order to the average area of subbasins of the next higher order (i. R ). R ). are functions of the number of streams of order i draining into streams of order j. ω. This ratio has been observed in nature to vary within a narrow range of 3 to 5. The parameter of this relationship is the socalled Length Ratio. A B The Horton law of stream numbers states that there exists a geometric relationship between the number of streams of a given order Nω and the corresponding order. B Nω = RΩ −ω B (17) The Horton law of stream lengths states that there exists a geometric relationship between the average length of streams of a given order and the corresponding order. and the overbar indicates the average value of the corresponding variable. ω Aω = A1 RA −1 (19) In the equations above. Ramírez probability density function of the travel times of droplets in streams of a given order and of the transition probabilities of water droplets going from a given state (stream segment) to another of higher order. ω. V. RodríguezIturbe and Valdés (1979) suggested that the probability distribution function of travel time in streams of a given order is an exponential distribution with parameter λ representing the inverse of the mean travel time. Ω is the order of the basin.
elegant. V is in meters per second.44 L Ω RB (R ) V A 0. obtained through numerical integration of the full GIUH. 14 . 1976 and 1977). This is based on the hypothesis of spatial uniformity of the flow velocity distribution throughout the river network (Pilgrim. The theoretical framework of the GIUH above was verified through numerical experiments with a detailed physicallybased watershed model on several basins in Venezuela and Puerto Rico with excellent results (Valdés et al. are simple. Nonlinear effects in the basin response manifest themselves in the characteristic discharge velocity. and multiple regression analysis.58( R A 0. They indicate that the nonlinear characteristics of the response function of a basin can be modeled with a linear scheme such as the GIUH but with a characteristic velocity that is representative of the discharge velocity for the given event. respectively. Ramírez for the characteristic velocity.Jorge A. They are. 1984). Rosso equated the product of the time to peak and peak flow of the GIUH with the product of the time to peak and peak flow of the Nash model IUH. this product equals. The parameters of the Nash model and the geomorphologic descriptors of drainage networks can be related by relating approximate measures of volume of the IUH for a cascade of linear reservoirs and the volume of the GIUH (Rosso. qp.55 −0.38 RL (21) where LΩ.. of the IUH which are functions of geomorphology. RodríguezIturbe and Valdés (1979) obtain expressions for the time to peak. 1979). (1988) showed that the estimation of the dynamic parameter of the GIUH could also be performed for those basins where flow velocity distribution depends on stream order. Agnese et al. and easy to use.31 0.43 RL V LΩ (20) and. RodríguezIturbe et al.55 0. The dependence of the GIUH on the dynamic parameter V can be used to address nonlinear watershed response issues. (t p q p ) GIUH = 0.05 ) RL RB (22) and for the Nash model IUH. tp = 0. qp = 1. Their results. and of the dynamic parameter represented by the peak velocity. tp. and peak discharge. and the units of qp and tp are the customary units of inverse time in hours. is the length of the highest order stream in kilometers. (1979) show that the dynamic parameter of the GIUH can be taken as the spacetime average flow velocity for a given rainfallrunoff event in a basin. and time in hours. However. It is simple to show that for the GIUH this product equals.
asymptotically for large M and for g(x. t) representing a simple translation. 1988. whereas the scale parameter is shown to depend both on geomorphology and streamflow velocity. The shape parameter of the IUH is shown to be a function only of Horton ratios. t) is the width function. 1985).78 0. respectively. t ) = N ( x. t ) / ∫ N ( x. Mesa and Mifflin.70[ RA /( RB RL )]0. and Gupta and Mesa. Using numerical algorithms to solve the resulting equation Rosso obtained. t )dx 0 ∞ (26) where g(x. Geomorphologic IUH's have also been proposed that are based on characterizing drainage networks as a function of network links as opposed to streams (Gupta et al. Karlinger and Troutman have shown that. 1984. and results of their application are compared to other empirical parameter estimation procedures. When the topological parameter is the magnitude M of the network. α = 3. These results improve the prediction capability of the Nash model for regions where no hydrologic records exist. u (t ) = ∫ g ( x. 1986.29( RB 0. N ( x. t )dx * 0 ∞ (27) where N(x. the expected conditional value of u(t) is. 1986. t > 0 2M ( li / V ) 2 4M l i2 (28) 15 .07 ) RL RA (24) (25) k = 0.Jorge A. The linkbased GIUH above has been approximated by its conditional expected value. t) is the hydrologic response function of a single channel a distance x from the basin outlet. When the parameters of the Nash model are estimated using these equations. 1986. t) is given by. Troutman and Karlinger. 1985. the GIUH is expressed in terms of the width function as. and N*(x) is the normalized width function which represents the probability density function of the number of links at a distance x from the outlet.48V −1LΩ where α and k are the shape parameter and the scale parameter of the Nash model IUH. conditional on a topological parameter vector (Karlinger and Troutman. N*(x. the adequacy of the geomorphologic connection is clearly demonstrated. Ramírez (t p q p ) Nash = (α − 1)α e (1−α ) / Γ(α ) (23) These equations allow estimation of the shape parameter. t ) N * ( x..) In that case. which measures the number of links at a distance x from the basin outlet. E[u (t ) / M ] = t − V 2t 2 exp( ).
general theories of the evolution of drainage systems have been developed. The Climatic GIUH . and the other accounting for river network composition (geomorphologic dispersion). 1993.The GcIUH. as well as geomorphology. t). Finally. models based on the accurate specification of the geometry and the topology of the network and simplified dynamics are theoretically validated irrespective of the choice of the travel time probability density function. li is the mean length of the interior links. and V is the velocity of translation. tr. This response is obtained as the solution of the linearized equations of motion for unsteady flow in a wide rectangular channel in which the infiltration losses are assumed to be proportional to the instantaneous discharge at any point along the channel. 1992. 1998a. In addition.. implicit in the GIUH (exponential distribution of stream travel times). Ramírez where M is the magnitude of the basin and equal to the number of first order streams.. Rigon et al. Channel Losses in the GIUH. (1982a) obtain probability density functions of the peak discharge and the time to peak as functions of the rainfall intensity. Accordingly. 16 .g. La Barbera and Rosso (1987. Rinaldo et al. Troutman and Karlinger show that this result is always true regardless of the form of g(x. 1998b). The GIUH has undergone several modifications and enhancements in order to make it more physicallybased. Howard. (1983a).Jorge A. One of those modifications consists in incorporating channel infiltration losses. The proportionality coefficient is known as the infiltration parameter. derived expressions for the probability distribution function of travel times in streams of any given order taking into account channel infiltration losses. RodríguezIturbe et al. DíazGranados et al. for example. Because the analysis showed the latter one to play the major role in determining basin response. The channel response to an instantaneous input anywhere along the channel is interpreted as the conditional probability distribution function of the travel time of a drop traveling a given distance along the channel. Within this context. The instantaneous unit hydrograph is derived under the assumption that it is a random function of climatic and geomorphologic characteristics and that it varies with the characteristics of the rainfall excess. is in fact adequate. Molnár and Ramírez. one accounting for the travel time within individual reaches (hydrodynamic dispersion). and duration. energy dissipation concepts have been used to link the local and global properties of a river network to the hydrologic conditions of its watershed (e. Results indicate that the linear reservoir assumption. These general theories allow for a fundamental description of global network properties and thus should be incorporated into improved geomorphologic models of basin response.. For example. 1989) first indicated the fractal nature of river networks and its relation to quantitative geomorphology as initiated by Horton's studies on river network composition. using previous results by Kirshen and Bras (1983). Studies on the geomorphologic response of river basins have increasingly stressed the importance of providing an accurate description of the quantitative properties of river network systems. Rainfall is characterized by a constant intensity over the storm duration. (1991) studied hydrologic response by decomposing the process of river runoff into two distinct contributions. RodríguezIturbe et al. i. 1990.
f (q p ) = 3. define the dynamic parameter of the GIUH.262Π 0.858Π 0. These expressions lead to.4 (32a) (32b) (32c) t p = 0. 4 E(t p ) = 0.412Πq2. RodríguezIturbe et al. qp = 0.585Π 0. and a derived distribution approach.5 Ω 1 iAΩ RLα Ω. of the GIUH that depend on the mean rainfall intensity. tp.871 Π 0. and αΩ is in s1m1/3.5 17 . the above equations can be manipulated to obtain expressions for the peak discharge and time to peak as functions of the particular storm intensity and duration. is the length of the highest order stream in kilometers.5 Ω Π= 1 i AΩ RLα Ω.534Πq1.327 Π0. E(qp ) = 0. L2. They are. (1982a) obtained analytical expressions for the probability density functions of the time to peak. respectively. and time in p p hours.5 2. and the peak discharge. AΩ.5 ) p p f (t p ) = where.4 σt p = 0.5 (30) −0. Ramírez or a rectangular pulse process.774 Π 0. i is the mean rainfall intensity in centimeters per hour. is the basin area in square kilometers.4 (31) where LΩ. Using the kinematic wave approximation for flow routing along streams of first order. These probability distribution functions assume that the time of concentration of first order streams is much smaller than the duration of the rainfall.4 σq p = 0.915Π 0.4 Πi = L2. These two characteristics. For individual storm events.5 tp tp (29) The first moments of peak discharge and time to peak of the GcIUH can be obtained as.Jorge A.5 exp(−1. hence the name GcIUH.656Π exp( ) 3. qp. The units of q and t are the customary units of inverse time in hours. together with geomorphologic parameters for a first order basin.
tlR. From these relationships.. due to its functional dependence on the rainfall characteristics. five characteristics of a required unit hydrograph for a given effective rainfall duration may be calculated (e. are given by.5tr (33) For a standard unit hydrograph the basin lag. runoff from a given rainfall event can be computed using an IUH derived from the given storm event characteristics.. qp.g. tr. W (in time units) of the unit hydrograph at 50 and 75 percent of the peak discharge. 1990): the peak discharge per unit of watershed area. The theoretical distributions of the time to peak and peak discharge compare very well with the experiments.75 (1.000 square miles. For example.3 (34) (35) qp = C2 Cp A tl The basin lag time of the standard unit hydrograph (equation 34) is in hours. The hydrograph characteristics are the effective rainfall duration. (1982b) compare results by the GcIUH theory with results obtained from IUH's derived from simulated rainfallrunoff events for one basin and three different climates. The product LLc is a 18 . Chow et al. the base time.g. watershed geomorphology. qp. These relationships are based on a study of 20 watersheds located in the Appalachian Highlands and varying in size from 10 to 10. t l = C1C t ( LLc ) 0. tb.) These problems are avoided if the GcIUH is used. the GcIUH allows estimation of the IUH for a given particular rainfall input. A standard unit hydrograph is associated with a specific effective rainfall duration. Bras. the basin lag. and Rosso and Caroni 1986. and the peak discharge. the peak direct runoff rate. when using traditional UH approaches (e. and the widths. 1988. Furthermore. 1986. L is the length of the main stream in kilometers (miles) from the outlet to the upstream divide. Lc is the distance in kilometers (miles) from the outlet to a point on the stream nearest the centroid of the watershed area. and the basin lag time. tl. Variations in qp and tp resulting from nonlinear basin behavior induce large uncertainty in the predicted peak discharge and time to peak for individual events. Caroni et al. RodríguezIturbe et al. and C1 = 0. so that problems associated with nonlinear basin behavior are avoided. and the hydrologic response of the basin. qpR. tl. tl .Jorge A. Standard unit hydrograph. Ramírez The GcIUH establishes a link between climate. Snyder’s Synthetic Unit Hydrograph The synthetic unit hydrograph of Snyder (1938) is based on relationships found between three characteristics of a standard unit hydrograph and descriptors of basin morphology.. tl = 5. tr. thereby avoiding the wellknown errors incurred when using an IUH based on a different event.0 for English units). defined by the following relationship with basin lag. Caroni and Rosso 1986.
56 (1290 for the English system). tl = 5. Army Corps of Engineers developed relationships for the widths of the UH at values of 50% (W50) and 75% (W75) 19 . and its peak discharge qpR in m3/s are obtained. and represents variations in watershed slopes and storage characteristics. The value of Cp is obtained using the expression for qp corresponding to the standard unit hydrograph. Required unit hydrograph. As an aid in drawing an adequate UH. To compute Ct and Cp for a gauged watershed. the base time of the required UH may be estimated by. Ramírez measure of watershed shape. and Ct and Cp are computed by the equations for tl and qp given above. its basin lag tlR in hours. As Ct. and C2 = 2. the coefficients Ct and Cp for the gauged watershed can be used in the above equations to derive the required synthetic unit hydrograph for the ungauged watershed. values of its associated effective duration tR in hours. From a derived unit hydrograph of the watershed. When an ungauged watershed appears to be similar to a gauged watershed. tl = tlR. If tlR = 5. in order to obtain tr and tl. The peak discharges of the standard and required UH’s are related as follows. If tlR is quite different from 5.S. q pR = q p tl tlR (37) Assuming a triangular shape for the UH. corresponding to the standard unit hydrograph. The value of Ct is then obtained using the equation for tl corresponding to the standard unit hydrograph.5tR . but using qp = qpR and tl = tlR. A is the basin area in km2 (mi2). the values of L and Lc are measured.Jorge A. then the derived unit hydrograph is a standard unit hydrograph and tr = tR. tb = C3 A qpR (38) where C3 is 5. and given that the UH represents a direct runoff volume of 1 cm (1 in). The peak discharge of the standard unit hydrograph (equation 35) is in m3/s (cfs). Cp is a coefficient derived from gauged watersheds in the area. Ct is a coefficient derived from gauged watersheds in the same region. the U.5tr. the standard basin lag is computed using: tl = tlR + tr − tR 4 (36) This equation must be solved simultaneously with the equation for the standard unit hydrograph lag time.5tR. and represents the effects of retention and storage. and qp = qpR.75 (640 for English units).
(1988) as.2: Snyder’s Synthetic Unit Hydrograph Conceptual UH’s Based on Dimensionless Hydrographs Soil Conservation Service Dimensionless Hydrograph. Ramírez of qpR.22 (440 for English units) for the 75% width and equal to 2.Jorge A. qp and the time by the ratio of time. to time to peak of the UH.25 1 Discharge 0. However.25 0 0 tl tp 1 2 3 4 Time Ratio tb Figure 11. the UH can be estimated from the synthetic dimensionless hydrograph for the given basin. For example. several other authors have recommended different distribution ratios.75 Ratio 0.08 (39) where the constant Cw is 1. onethird of this width is distributed before the peak time and twothirds after the peak time. q W% = C w lR A −1. q. note that the widths of the hydrograph at 50% and 75% of the peak flow are distributed such that the longer time is to the right of the time to peak. as recommended by the U.14 (770 for English units) for the 50% width. The SCS dimensionless hydrograph is a synthetic UH in which the discharge is expressed as a ratio of discharge. t. Hudlow and Clark (1969) recommend a partition of 4/10 and 6/10. 1.2 illustrates the form of Snyder’s synthetic UH.5 tr W 75 W 50 0. The width in hours of the UH at a discharge equal to a certain percent of the peak discharge qpR is given by Chow et al. tp. Also. Army Corps of Engineers. Usually. Note that the time lag is not the same as the time to peak. The dimensionless unit hydrograph developed by the Soil Conservation Service has been obtained from the UH’s for a great number of watersheds of different sizes and for many different locations. respectively. to peak discharge.S. Figure 11. 20 . Given the peak discharge and the lag time for the duration of the excess rainfall.
Eagleson (1972) presents a general framework for the development of flood frequency distributions. fI(i). U. this volume).08 (483.. and the discharge in m3/s/cm (or cfs/in).3: SCS Dimensionless Unit Hydrograph The SCS suggests that the dimensionless UH can be described in terms of an equivalent triangular hydrograph. Because the volume of direct runoff must equal 1 cm. Flood Frequency Distributions and Climatic GIUH. is equal to 2. tb.67 tp.67 tp. The values of qp and tp can then be estimated using this simplified triangular unit hydrograph whose height is equal to qp and whose time base.Jorge A. several authors have developed flood frequency distributions that include various assumptions about the characteristics of the rainfall process in terms of its probability density function. 21 .6 Ratio 0. After analysis of a great number of UH’s. From a study of many large and small rural watersheds the basin lag is tl = 0.4 0.S.2 1 0.2 0 0 1 2 3 4 5 Time Ratio Figure 11.6tc. Ramírez 1.8 Discharge 0.g. Either using numerical or analytical procedures these authors have arrived at different flood frequency distributions. The time to peak.4). and the watershed processes transforming rainfall into runoff encoded in the function g(i. One of the most recent and most interesting developments has been the use of the GIUH concept in order to quantify and analyze the effect of geomorphologic characteristics in the determination of regional flood frequency. 1985). where tc is the time of concentration of the watershed. the SCS recommends a recession duration of 1. Based on Eagleson's framework (see Figure 11. The time is usually expressed in hours (SCS).θ). Chapter 12.4 in the English system) and A is the drainage area in square kilometers (square miles). it can be shown that qp=C A/ tp where C = 2. Soil Conservation Service.t. is then equal to tr/2+tl (e. tp. F(Qp)(see also Stedinger.
λ is the inverse mean duration of rainfall events. β r r Q p = g (i.Jorge A. θ r Figure 11. Comparisons of their flood frequency distributions with the one 22 . Hebson and Wood obtain. tre is the duration of excess rainfall. 1972. Qp is the flood peak under consideration. and β* is a modified areal inverse mean intensity of rainfall events. General framework for flood frequency analysis (adapted from Eagleson. t. tk is the duration of the GIUH or kernel length.4. In terms of the equivalent recurrence interval. and the Davidson River in North Carolina. Ramírez f I (i ) Derivation of Peak Flow Probability Distribution Function r FQ (Q p ) = ∫∫ g (i.) Hebson and Wood (1982) derived a flood frequency distribution function from assumed climatic distributions for the rainfall parameters and using the GIUH as the transformation function of rainfall into runoff. and exponentially distributed random variables. namely the Bald Eagle Creek in Pennsylvania. ln TE = − ln[λ ∫ exp(− λ tre − β * (Qp − Qb ) / Acg(tre ))dtre + tk 0 exp(− λ tk − β * (Qp − Qb ) / Ac )] − ln n (40) where n is the average number of annual direct runoff events. They applied this concept to two thirdorder basins.θ ) Runoff Model Catchment Parameters. t .θ ) f I (i)di R FQ (Q p ) Rainfall Model Climate Parameters. Qb is the base flow. These authors assume a model of rectangular pulses for storms whose intensity and duration are independent. g(t) is the area under the GIUH as a function of time. TE. Ac is the contributing area.
and which depend on the mean intensity and the mean duration of precipitation events. (1984) resort to using arguments of ecological optimality for waterlimited natural systems in order to define a long term average soil moisture. Runoff is then generated only when the precipitation intensity exceeds the soil infiltration capacity. concentrationdependent equation of the diffusion process in unsaturated porous media. obtaining the analytical form of the flood frequency distribution from climate and geomorphology often results in a cumbersome derivation. To overcome this problem. Ramírez derived by Eagleson (1972). which is based on kinematic wave concepts for overland flow. The flood frequency distributions obtained by this method depend on climatic and geomorphologic parameters of the watershed as well as on soil properties governing the infiltration process. Infiltration capacity was defined by using Philip's solution to the onedimensional. In terms of the equivalent recurrence interval. (1983. indicate that the bestfit GIUH produces better representation of the observed flood frequency distributions. TE−1 = n[1 − FQ (QP )] FQ (QQ ) = 1 − δ exp( βa − 2σ )Γ(σ + 1)σ −σ (41) [I + ∑ Ji ] i =1 4 (42) where I. as well as on the infiltration sorptivity of the soil. so. Effective rainfall intensity. 1984). DíazGranados et al. This derived distribution was used together with a triangular representation of the GcIUH in order to obtain a flood frequency distribution. and Ji are integral equations that cannot be solved analytically. They used a derived distribution approach in order to link the joint probability distribution function of storm duration and storm intensity and a physicallybased representation of the infiltration process in order to obtain the joint probability density function of effective storm duration and effective storm intensity. (1989) introduced the method of approximate moments by Taylor series to estimate 23 . ie. te. their derived flood frequency distribution is. Adom et al. and effective rainfall duration. the discharge under consideration. Their results indicate good agreement for both wet and arid climates. and runoff to be generated only by the Horton infiltration excess mechanism. were obtained by assuming precipitation to be a process of rectangular pulses. sometimes yielding implicit equations to be solved numerically.Jorge A. DíazGranados et al. Given that the above distribution depends on the elusive initial soil moisture. However. 1984) derived a flood frequency distribution based on the GcIUH. Storm intensity and duration were assumed to be independent and exponentially distributed random variables. and geomorphologic parameters of the watershed (see DíazGranados et al. Using this approach.. they verified their flood frequency distribution against observed distributions from two very different climatic regimes. although the agreement is better for the arid climate of Santa Paula Creek in California than for the wet climate of the Nashua River basin in New Hampshire.
the corresponding routing procedure is referred to as lumped routing. and O(t) is the outflow hydrograph at the downstream end of the reach. When no spatial variability is taken into account and when the channel reach or reservoir is considered as a black box. dS(t) = I(t) − O(t) dt (43) In this equation. For this purpose. which is governed by conservation of mass and momentum equations. this method can be used to estimate the flood frequency distribution for ungauged catchments. Ramírez second order statistics of both flood peak and volume. bed slope. and the GIUH to model surface runoff response. length of channel reach. After a given type of frequency distribution has been identified at the regional scale by analyzing annual flood series. the peak flow of the hydrograph decreases and the time base of the hydrograph increases. Lumped Flood Routing Lumped or hydrologic routing is governed by the continuity equation and a storagedischarge functional relationship. they used the Soil Conservation Service's Curve Number method to describe basin rainfall excess. Assuming a linear variation of input and output fluxes during the ∆t leads to. The continuity equation can be integrated over a given ∆t to obtain. and initial and boundary flow conditions. S(ti + 1 ) − S(ti ) = ∆t 2 [ I(ti + 1 ) + I(ti )] − ∆t 2 [O(ti +1 ) + O(ti )] (45) 24 .Jorge A. S(t) is the storage within the system (channel reach or reservoir). I(t) is the inflow hydrograph at the upstream end of the reach. The continuity equation is. As flood waves travel downstream they are attenuated and delayed. S(t i+ 1 ) S(ti + 1 ) − S(ti ) = ∫ t i+1 dS(t) = ∫ I(t)dt − ∫ O(t)dt ti ti ti +1 (44) S(ti ) where the subscripts i and i+1 refer to the beginning of two consecutive time intervals. The propagation of flood waves in a channel is a gradually varied unsteady flow process. FLOOD ROUTING The term flood routing refers to procedures to determine the outflow hydrograph at a point downstream in a river (or reservoir) as a function of the inflow hydrograph at a point upstream. The solution of these equations in a distributed manner is referred to as distributed routing of flood waves. The shape of the outflow hydrograph depends upon the channel geometry and roughness. That is.
nonhysteretic).Jorge A.E+07 Storage (m ) Figure 11..E+07 4. stage or storage) the outflow is unique. the storage and the outflow are onetoone functions of elevation (see Figures 11. independent of how that storage is achieved. unique. and horizontal water surfaces or negligible backwater effects. The complexity of the routing procedure depends on the nature of this relationship. a second equation relating S(t) and O(t) is required. Reservoir or level pool routing refers to flood routing for systems whose storage and outflow are related by a function of the type S(t) = f[O(t)] which is onetoone (i. The nature of this relationship may be linear. Thus.E+07 3 6.E+07 7.E+07 3. The solution procedure involves rearranging the continuity equation such that all unknown quantities are on the left hand side of the equation.E+00 1. low flow velocities.5 and 11. 500 450 400 350 Elevation (m) 300 250 200 150 100 50 0 0. 2S(ti + 1 ) 2S(ti ) + O(ti + 1 ) = [ I(ti +1 ) + I(ti )] + [ − O(t i )] ∆t ∆t (46) For level pool systems.E+07 8.e.6). Ramírez In equation (46) there are two unknown quantities. Such systems have a pool that is wide and deep compared to its length in the direction of flow.E+07 5. This characteristic of their S vs.5: Storage vs.E+07 2..E+07 9. Elevation Relationship 25 . or one to one. or hysteretic. or nonlinear. S(ti+1) and O(ti+1). Reservoirs or systems with horizontal water surfaces have relationships of the onetoone type.g. O relationships implies that for a given set of conditions (e.
the peak outflow occurs when the 26 . For these systems.7) and then using the function sequentially to solve for O(ti+1) for every time step. the left hand side of the equation above is a unique function of water surface elevation in the system.g.. Ramírez 800 700 600 Discharge (m/s) 500 400 300 200 100 0 0 50 100 150 200 250 300 350 400 450 500 Elevation (m) Figure 11. only. Discharge Relationship Thus. the storageelevation relationship is available from topographic surveys.Jorge A. The solution involves the development of the function 2S/∆t + O = f[O(h)] (see Figure 11. and the outflowelevation relationship is available from hydraulic considerations with respect to the outlet structures (e.7: StorageIndication Function As observed in Figure 11. the effect of storage is to redistribute the flow hydrograph by shifting its centroid from the position of that of the inflow hydrograph to that of the outflow hydrograph. h is water surface elevation in the reservoir. etc. In this equation.8. 800 700 600 Discharge (m3/s) 500 400 300 200 100 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 2S/∆t+O (m3/s) Figure 11. Usually.6: Elevation vs. spillways.).
but hysteretic.Jorge A. I(ti) and I(ti+1). S(t) = k[ xI(t) + (1− x)O(t)] (47) This relationship specifies that the storage is proportional to a convex linear combination of inflow and outflow rates. k and x. the outflow at the end of a given time interval. at which time the maximum storage in the system is also achieved (see Figure 11. Parameter x is dimensionless and represents the relative importance of the effects of inflow rates on defining the storage in the reach. as well as the outflow rate at the beginning of the interval. the outflow from a given reach is not a single valued function of the storage in the reach. In addition to the lumped form of the continuity equation. Ramírez outflow hydrograph intersects the inflow hydrograph. dS(t) = I(t) − O(t) dt (43) this method assumes a linear storage discharge relationship as given below. Theoretically this parameter varies between 0 and 1. channel flow routing. the magnitude of the outflow discharge depends on whether the outflow is occurring during the rising limb or the recession limb of the inflow hydrograph. 250 200 Discharge (m3/s) 150 100 50 0 0 50 100 150 200 250 300 350 400 450 500 Time (h) Figure 11.8: Level Pool Routing: Inflow and Outflow Hydrographs Channel Routing The storagedischarge relationship in the case of channel routing is not onetoone.8). which are important in long and narrow flow routing conditions such as occur in channels. The Muskingum method has two parameters. 27 . O(ti+1) can be expressed as a linear combination of the inflow rates at the beginning and end of the given interval. Parameter k has units of time and represents the average travel time of a kinematic wave through the reach. and it is known as the storage time constant. In its most simple interpretation. Combining the continuity and storage equations. and integrating over a short time interval. In other words. The Muskingum method is one of the most widely used methods of lumped. O(ti) as. This behavior is a result of backwater effects.
Observe that routing in systems with hysteretic S vs. This method results in a trial and error procedure which. Ramírez O(ti +1 ) = Co I(t i+ 1 ) + C1 I(t i ) + C2 O(ti ) (48) The coefficients or weights of this linear combination can be computed as a function of k. 28 . Co = (−kx + 0. Qn = ∑ I n −m + 1Um m= 1 n (50) where Um is the UH of the Muskingum method whose ordinates are given by the following equations. U1 = Co U2 = Co C2 + C1 Um = Um− 1C2 . and ∆t as indicated below. the leastsquares method is a numerical expression of the graphical method.9ad). when properly carried out. or by using the leastsquares method.5 ∆t/k < 1x.5∆t) / C3 C3 = k − kx + 0.5∆t (49a) (49b) (49c) (49d) It is simple to show that the Muskingum routing solution can be expressed also as a discrete convolution equation as.5 ∆t) /C3 C1 = (kx + 0. x.Jorge A.10). O relationships leads to additional modifications of the inflow hydrograph represented in a delay of the response of the system and in asynchronous occurrences of the maximum storage and maximum outflow. The graphical method consists in choosing x such that the loop resulting from the plot of S vs. The slope of the resulting straight line is an estimate of the parameter k (see equation 47).5∆t) /C3 C2 = (k − kx − 0. constitutes a graphical implementation of the leastsquares method. the latter occurring after the former (see Figure 11. ∆t must be selected such that x < 0. In order to ensure positivity of the unit hydrograph ordinates. Or in other words. m > 2 (51) The Muskingum parameters k and x can be easily estimated graphically. (xI + (1x)O) becomes as close to a straight line as possible (see Figure 11.
E+06 Volume(m3) 6.E+06 Storage (m3) 6.E+06 5.E+07 9. 1.09 8.E+00 0 10 20 30 40 50 Weighted Average Influx (m3/s) Figure 11. 1.E+06 5.E+06 2.E+06 x = 0.E+06 7.E+06 0.08.E+06 4. Storage Function .05. 29 .05 8.E+06 4.E+06 1.E+00 0 10 20 30 40 50 Weighted Average Influx (m3/s) Figure 11.E+06 Storage (m3) 6.E+06 3.E+06 4. Ramírez 1. Storage Function .E+06 0.Jorge A. Storage Function .E+06 3.9c Weighted Average Influx vs.9a: Weighted Average Influx vs.x = 0.E+07 9.E+00 0 10 20 30 3 40 50 Weighted Average Input (m /s) Figure 11.x = 0.E+06 7.9b: Weighted Average Influx vs.E+06 5.E+06 1.E+06 x = 0.E+06 7.E+06 2.E+06 2.09.x = 0.E+06 0.E+06 x = 0.E+06 1.08 8.E+07 9.E+06 3.
the Muskingum routing equations can be rewritten as.E+06 1. and x.E+06 3. Storage Function .E+06 x = 0.Jorge A. 1969). Qij++11 = Co Qij +1 + C1 Qij + C2 Qij+1 (52) where the flow discharge Qi j refers to position i in space and j in time.E+06 Storage (m3) 6.E+06 2.E+00 0 10 20 30 40 50 Weighted Average Influx (m3/s) Figure 11.9d Weighted Average Influx vs. k= ∆x ck (53a) 30 . the parameters k.x = 0. C2.E+06 5.10 120 100 Discharge (m3/s) 80 60 40 20 0 0 20 40 60 80 100 120 140 Time (h) Figure 11. required in order to obtain Co. C1. and C3 are estimated as. the Muskingum routing equation constitutes an approximate solution to the kinematic wave equations (see Distributed Flood Routing section below) (Cunge. Expressing the solution in a spacetime domain.E+06 7.10 8.E+06 0. Ramírez 1.E+06 4.E+07 9. In this solution.10: Muskingum Routing: Inflow and Outflow Hydrographs Using a hydraulic analogy and applying the Muskingum equations over a short ∆x.
and vx is the magnitude of the lateral inflow velocity component in the direction of flow. d) gravitational acceleration. functions of space and time. Sf is the friction slope. ∂Q ∂A + =q ∂x ∂t ∂Q ∂ (βQ2 / A) ∂y + + gA( − So + S f ) − β qvx = 0 ∂t ∂x ∂x (54) (55) where Q is discharge. in general.e. A is the cross sectional flow area. 2) the vertical accelerations are negligible so that the vertical pressure distribution is hydrostatic (gradually varied flow). 3) the slope of the channel bottom is small and the channel bottom is fixed. Local and convective accelerations represent the effect of 31 . Q is the discharge. it can be shown that x should be between 0 and 0.. the flow of water through the soil and in the channels. ck is the celerity of the kinematic wave corresponding to Q and B. The excitation of the system (i.Jorge A. β is the momentum correction coefficient for nonuniform flow velocity distribution. where B is the width of the water surface. and f) lateral inflow. the response of the system. Cunge also demonstrated that this solution constitutes an approximate solution of a modified diffusion equation if the parameters k and x are estimated as expressed above. Distributed Flood Routing The characteristics of hydrologic systems are extremely variable in space and time. unsteady flow in an open channel are. and the boundary conditions are. flow depth. distributed flow routing procedure. 4) resistance coefficients for steady uniform turbulent flow are applicable. In those equations. The governing equations for onedimensional.5. c) water level differences along the channel and consequent imbalance of pressure forces. and flow velocity can be obtained as a function of space and time using a dynamic. Flow characteristics such as flow discharge. Ramírez 1 Q x = (1 − ) ck BSo ∆x 2 (53b) In addition. These two equations are known as the de SaintVenant equations. that is. e) frictional forces. the initial conditions. q is lateral inflow per unit length of flow. So is the bottom slope. A is the cross sectional area of flow. The momentum equation includes terms for changes in momentum due to: a) changes in velocity over time (local acceleration). Based on this analysis. and 5) the fluid is incompressible and of constant density. Therefore. The main assumptions encoded in these equations are: 1) the flow is onedimensional. precipitation). b) changes in velocity in space along the channel (convective acceleration). So is the channel bottom slope. y is flow depth. and ∆x is the increment in space. is a distributed process in which the characteristics of the flow change both in time and space.
Depending on the flow conditions. Woolhiser and Liggett (1967) have shown that for K values greater than 10. However. Bras. For a kinematic wave. Q= Sf n R2/3 A (59) Together with the continuity equation. Chow et al. dynamic waves tend to attenuate more rapidly than kinematic waves (e. or full dynamic wave situations. K= So L yo F 2 (56) where So is the bottom slope. 1988. Q = αA β (58) which. Kinematic waves govern the flow when inertial and pressure forces are not important. or diffusive. 1990). and that the flow is steady and uniform. and F is Froude number for normal flow at the downstream end of the flow length. warranting a kinematic wave approach for such flow routing problems. yo is the normal flow depth at the downstream end of the flow length. 1989. In natural flood wave conditions both kinematic and dynamic waves are present.Jorge A. Singh.103)). So = S f (57) implying that the energy grade line is parallel to the channel bottom. Eagleson. The above momentum equation can be shown to be equivalent to the following stagedischarge relationship. some of these terms may be negligible as compared to others in the equation giving rise to either kinematic.. The kinematic parameter is defined as. the flow dynamics tend to be dominated by kinematic waves. can be satisfied by Manning equation. for situations where the Froude number is less than 2. This sometimes leads to simplifications of the flood routing problem (e. for example. Kinematic Waves.g.. the magnitude of the socalled kinematic parameter is useful. L is the length of (overland or channel) flow. Although there is no single set of criteria to define when a kinematic wave solution is acceptable. ∂Q ∂A + =q ∂x ∂ t (54) these equations represent the routing equations for the kinematic wave flow routing approach. 32 . Ramírez inertial forces in the flow. the momentum equation reduces to. In overland flow situations the value of K usually exceeds this threshold.g. (K~O(102 . 1970).
like finitedifferences methods. Thus.twodimensional and fully dynamic. or the 33 . and the channel inflow is taken to be flow per unit width of plane.Jorge A. the continuity equation can be rewritten as. 1 ∂A dA ∂Q ∂Q = = ) β−1 ( ∂t dQ ∂t αβ A ∂t 1 ∂Q ∂Q ∂Q Q1/ β − 1 ∂Q + ( )= + =q ∂x αβA β − 1 ∂t ∂x α 1/ β β ∂t (61) (62) (63) From equations 60 and 63. Thus. Furthermore. Given that at any cross section Q and A are functionally related as Q = αAβ. In this application. as shown below. the observer would see the flow rate increase at a rate equal to the lateral inflow rate. However. more complex representations of the flow process must sometimes be used . for q = 0. A basin is described as a collection of such overland flow planes representing the hillslope processes feeding into a network of collector channels which feed into a main channel.. the kinematic wave equations have explicit analytical solutions. they simply translate downstream without dissipation. except in a few particular cases (e. that is.g. kinematic waves do not attenuate. kinematic waves for homogeneous planes). spatiallydistributed solution of the physical equations governing the surface flow. q. The kinematic wave model offers the advantage over the unit hydrograph method that it is a onedimensional. the SaintVenant equations do not have explicit analytical solutions. the lateral inflow is taken to be equal to the rainfall excess. In general. the kinematic wave approximation has been widely applied to construct models of the rainfallrunoff process. dQ ∂Q ∂Q dt ∂Q 1 ∂Q = + = + =q dx ∂x ∂t dx ∂x c k ∂t (60) It can be seen that for conditions of no lateral inflow. for overland flow waves in the case of homogeneous planes. dQ/dt = 0. ∂Q dA ∂Q + =q ∂x dQ ∂t and. ck = dQ = αβA β − 1 dA (64) Assuming that overland flow situations can be described as flow over planes. numerical solution procedures must be used. Ramírez If an observer moves with the kinematic wave at a speed equal to the kinematic wave celerity. it is easy to obtain the following expression for the kinematic wave celerity.
Ramírez method of characteristics. For the kinematic wave problem presented above. j f (Qi++1 ) = 1 ∆ t j+ 1 + Q + α −1/ β (Qij+11 )1 / β − C ∆x i + 1 1 ∆t + + 1/ β (Qij+11 )1 / β −1 ∆x α β (71) f ' (Qij++11 ) = (72) 34 . For the nonlinear kinematic wave solution. For the linear solution. The same subscript and superscript convention applies as for the linearized solution. (1988) is adapted below.Jorge A. A simple implementation of the linear and nonlinear solutions suggested by Chow et al. This leads to the following solution.. the equation is linearized by substituting an average of known solutions for the coefficient of the nonlinear term. these solutions can be linear or nonlinear. the subscript refers to the space coordinate and the superscript refers to the time coordinate. j Qij++11 − Qij +1 Aij++11 − Aij+1 qi++11 + qij+1 + = ∆x ∆t 2 (67) q j +1 + qij+1 ∆ t j +1 ∆t j + 1 + Qi+ 1 + α −1/ β (Qij+11 )1 / β = Qi + α −1/ β (Qij+1 )1/ β + ∆t( i +1 ) ∆x ∆x 2 C= q j +1 + qij+1 ∆t j + 1 Qi + α −1/ β (Qij+1 )1/ β + ∆t( i +1 ) ∆x 2 (68) (69) An iterative solution based on Newton’s method is shown below. 1 Q j + Qij + 1 1 / β −1 Qij++11 − Qij+ 1 q j + 1 + qij+ 1 Qij++11 − Qij +1 + 1/ β ( i + 1 ) ( ) = ( i +1 ) ∆x α β 2 ∆t 2 j j +1 j +1 j 1 Q + Qi 1/ β − 1 q + qi +1 ∆t j +1 Qi + 1/ β Qij+ 1 ( i +1 ) + ∆t( i + 1 )] ∆x α β 2 2 = j j+1 1 Q + Qi 1 / β −1 ∆t [ + 1/ β ( i + 1 ) ] ∆x α β 2 (65) [ Qi +1 j +1 (66) In this linear solution. The solution advances on a time line from upstream to downstream. the partial differential equations are recast in finite difference form as shown below. Finite difference methods solve the partial differential equations on a grid placed over the xt plane.
and computes surface water profiles associated with subcritical. during the movement of a flood wave in a system of channels of shallow slopes. Ramírez ∆ t j +1 −1/ β 1/ β (Qi+ 1 )k + α [(Qij++11 )k ] − C j+1 j +1 (Qi +1 )k +1 = (Qi+ 1 )k − ∆ x ∆t 1 j + 1/ β [(Qi++1 )k ]1 / β −1 1 ∆x α β (73) Newton’s method is an iterative numerical solution procedure. supercritical. Dynamic waves govern the flow whenever inertial and pressure forces are significant when compared to the gravitational and frictional forces. which progressively determines improvements to a trial solution as a function of the gradient of the function at the trial solution. and mixed flow regimes. or by large discontinuities in the flow characteristics as a result of dam breaks or large reservoir releases. the magnitude of the improvement is less than a threshold). for example. and g is the acceleration of gravity. nonuniform flows for which backwater or tidal effects are important as. The energy equation is. Yi is the surface water depth at section i.. The energy head loss between two cross sections accounts for friction losses and for losses induced by contractions or expansions and is evaluated as. caused by hurricanes). This condition occurs for unsteady.g. backwater effects induce rating curves that are hysteretic. Y2 + Z 2 + 2 α 2 v2 2g = Y1 + Z1 + 1 α1v1 2g + he (74) where he is the energy head loss. Zi is the elevation of the channel bottom at section i. The method converges relatively quickly. by tides. by tributaries. by storm surges (e. vi is the average velocity at cross section i.. gradually varied.g. steady flows. Surface water profiles are solved sequentially by solving the energy equation using an iterative algorithm called the standard step method. so that a solution is reached when this criterion is met (e. he = LS f + C α1v12 2g − 2 α 2 v2 2g (75) 35 . storage per unit length of channel) in the section.. Full dynamic wave flood routing models can be both steadystate and unsteadystate models. As indicated above for the case of hydraulic river routing.e. Dynamic Waves. that is.Jorge A. A convergence criterion can be defined in advance. HECRAS is designed to perform onedimensional hydraulic simulations for a network of channels. rating curves for which the discharge is not a single valued function of stage (i. The Hydrologic Engineering Center's River Analysis System (HECRAS) model is an example of a steadystate flow model used for the computation of water surface profiles associated with one dimensional. and may be induced by downstream reservoirs or other downstream controls.
not necessarily with the complete precipitationrunoff transformation. Ramírez where L is the reach length weighted by discharge. 1988). whereas some of the approaches presented earlier in this chapter are concerned simply with the translation of a flood wave. The latter models are concerned with the simulation of watershed processes during a period of time that encompasses more than one precipitation or floodcausing event. 1988). 1988) which synthesizes DWOPER and DAMBRK. continuous models are also concerned with evaporation and subsurface soil moisture redistribution during the interstorm periods. In addition.g. the models of the National Weather Service developed by Fread and collaborators are well documented and commonly used (e. HECRAS. and C is a coefficient for expansion/contraction losses. 1988. Thus. 1978. Common applications of the HECRAS model include flood plain delineation and analysis for flood protection and insurance purposes. Fread.) 36 . Sf is a representative friction slope between the two sections. unsteady flows. the HECRAS model uses the steady state momentum equation instead. and of icecovered rivers. P2 − P + Wx − F f = ρQ∆v x 1 (76) where Pi is the hydrostatic pressure force at section i..Jorge A. the treatment and presentation of the details of the solution is beyond the scope of this chapter. it is implied that watershed models deal with the transformation of precipitation into runoff. However. Q is the discharge. Ff is the force due to friction in the reach between sections. For situations where flow regimes change abruptly from supercritical to subcritical or vice versa. The former models are concerned with simulating a single flood as a result of the occurrence of an individual precipitation or floodinducing event. modeling of single and/or multiple culverts and bridges. ρ is the density of water. The momentum equation is written in this case as. Among these models are the DWOPER model (Dynamic Wave Operational Model) (Fread. and ∆vx is the change in average velocity within the reach. These and similar models solve the fully unsteady dynamic wave equations using finite differences. Fread and Lewis. 1976.. Wx is the component of the weight of the water in the section in the direction of flow. 1974. Most of these models include elements or components that are described using the concepts of previous sections in this chapter. modeling of floodway encroachment. and the FLDWAV model (Fread and Lewis.g. In the context of fully dynamic. the model allows for the specification of variable cross sectional characteristics. as the energy equation is not applicable under those conditions. For those details the reader is kindly referred to the above references and those given therein. modeling of gated spillways and weirs. However. These models are then referred to as flood hydraulics models (e. both at a section and along the channel. 1978). DISTRIBUTED WATERSHED MODELS Distributed and semidistributed watershed models are generally categorized into singleevent simulation models and continuous simulation models. the DAMBRK model for dam failure studies (Fread.
which in turn is depleted by capillary rise C and base flow B. If G exceeds the threshold value G*. A kinematic wave model component is used for routing. then C and D disappear and EM and F operate on G. The drainage basin is presented as a set of segments (overland flow. and FLEA (Ranzi and Rosso. IHM (Morris. Ramírez SingleEvent Simulation Models There exists a great variety of single event watershed models. which augments surface storage R. 1965) was one of the first single event models developed to predict flood volume and peak rates of runoff for small drainage areas. 1981).S. The channel outflow Q begins when the threshold value R* is exceeded. Deep percolation D occurs when the soil moisture threshold M* is exceeded. California (now at Davis. The infiltration F and capillary rise C augment the soil storage M. The current version of the HEC1 package includes computational capabilities of dam breaks. An alternative USGS model is the distributed routing rainfallrunoff model developed by Dawdy et al. 1991a). 37 . 1980). The channel storage S and groundwater storage G are assumed to be linear reservoirs with storage constants KS and KG.S.S. treat the precipitation as rainfall or snowfall. which is depleted by transpiration EM. However. This model uses a modified version of the Philip's infiltration equation and the infiltration capacity concept of Crawford and Linsley (1966) to determine the excess rainfall. Reference is made to two models developed by the U. Soil Conservation Service. M is absorbed into G. Geological Survey in over twenty states to develop peak flows and hydrographs for state highway agencies. The USGS rainfallrunoff model (Dawdy and O'Donnell. The following sections are based on the review presented in Ramírez et al. Corps of Engineers. Singh (1989) provides a summary table listing fifteen such models and reviews some of them. SWMM (Metcalf & Eddy Inc. Geological Survey and one developed by the Hydrologic Engineering Center of the U. 1971). It was originally developed in 1967 by the staff of the Hydrologic Engineering Center at Sacramento. The infiltration is evaluated by a Horton type equation. The hydrologic simulation capabilities of HEC1 include several techniques to input and distribute the precipitation. The channel flow Q is routed through a linear reservoir S yielding the flow QS. (1994).. project optimization and kinematic wave programs (HEC.S. many other alternative models are available such as the TR20 (U. The model has been utilized by the U. (1972). A single event watershed model which has been widely used for estimation of surface runoff and river/reservoir flow in a basin during floods is the HEC1 Flood Hydrograph Package. The process is triggered by the precipitation input P. 1973). The model was specifically designed for routing urban flood discharges through a branched system of pipes or natural channels using rainfall as input. California).Jorge A. which in turn is depleted by infiltration F and evaporation ER. channel reservoir and nodal) which jointly describe all subbasins of the entire basin. which added to the base flow B gives the total flow Q. augmenting groundwater storage G. respectively.
Ramírez compute rainfall and snowmelt losses and excess. topographic. leaving an excess precipitation hyetograph which in turn is transformed into surface runoff hydrograph through a specified unit hydrograph. the HEC1 package also has a provision for evaluating reservoir and channel development plans for flood control purposes by performing the economic analyses of flood damages for existing and postdevelopment conditions. and the losses are computed. Since then.Jorge A. (1994). 1985) for simulation of daily runoff in the Great Lakes basins. soil. each land segment is assumed to be homogeneous with respect to hydrologic 38 . 1972). 1970.S. HSPF. This can be accomplished by modifying the physical parameters of the subbasin or routing reach in which the change is expected or proposed and by reanalyzing the response of the watershed. The subsurface runoff hydrograph is computed separately and added to the surface runoff hydrograph to yield the total subbasin runoff hydrograph..S. An additional application of the calibrated model is for impact assessment studies of basin modifications and channel improvements. The HEC1 model can account for temporal and spatial variability of the precipitation runoff process in a semidistributed sense. 1980) employ the basic structure of the original SWM for runoff simulation but in addition simulate water quality and sediment erosion. Claborn and Moore.g. 1970. the SHE model developed in Europe (Abbott. (Donigian and Crawford. Liou. and drainage considerations. The SWM model partitions the watershed into land segments whose boundaries are established according to hydrologic characteristics and user's needs. ARM. 1971. The precipitation hyetograph (rain and/or snow) is input over the sub basin. The review is based on that of Ramírez et al. 1979) and the Hydrologic Simulation Program. HEC1 uses spatially and temporally lumped parameters to simulate the precipitationrunoff process. the original model has undergone a number of revisions and expansions to satisfy specific needs and requirements (e. The model may be used to simulate a simple singlebasin watershed or a very complex basin with practically unlimited number of subbasins and river reaches. the National Weather Service Model (NWS) and the Precipitation Runoff Modeling System (PRMS) of the US Geological Survey. 1986b). Army Corps of Engineers. The segmentation of land is usually based on meteorologic. Thus. The models are: the Stanford Watershed Model (SWM). Ricca. However. and determine subbasin outflow hydrographs by various hydrologic routing techniques. of EPA (Johanson et al. many alternative models exist such as the SSARR (U. 1972) for daily streamflow simulations and forecasting for large watersheds. or the LBRM developed by the U. Several models such as the Agricultural Runoff Management Model. That is. 1986a. Great Lakes Environmental Research Laboratory (Croley. The Stanford Watershed Model (SWM) was originally developed by Crawford and Linsley (1966) for the continuous simulation of the hydrologic processes in a watershed at hourly time scales. In addition to being capable of hydrologic simulation. Continuous Simulation Models Three continuous watershed models developed in the United States of America are summarized in this section.. Anderson. within a subbasin.
The U. interflow. overland flow. and groundwater storage. Burnash et al. infiltration. Infiltrated water moves to the lower zone and groundwater storage. The role of the upper zone is to simulate the initial response to rainfall. Depletion of free water occurs vertically as percolation and horizontally as evapotranspiration and interflow. The soil moisture accounting model of the NWSRFS has been developed by the NWS Sacramento. and stream surfaces. the upper zone and the lower zone.Jorge A. there are two types of free water storage: primary storage. explicit soil moisture accounting. and for the first few hours of major storms. computation of mean basin precipitation. known as the National Weather Service River Forecast System.g. which is slow draining. California River Forecast Center (e. Each zone stores water in two forms.g. National Weather Service (NWS) developed a computerized system of operational river forecast procedures including data acquisition and processing. and therefore it is of importance for small storms. 1974. Monro and Anderson. Direct inflows from the impervious areas. The lower zone plays a major role during large storms in controlling temporal variation of infiltration rates. upper zone. has been extensively described (e.. The Sacramento soil moisture accounting (SMA) model assumes that a catchment consists of two reservoirs. The flow routing model is responsible for routing the channel inflows to the basin or subbasin outlet using unit hydrograph and Muskingum routing methods. conceptual. This set of models. interflow and inflow to the groundwater storage. lower zone. Moisture may be lost from the system as evapotranspiration from interception. 1973). namely: the Sacramento soil moisture accounting model. and operational forecasting. and provides baseflow over long periods of time. channel routing. The upper zone storage may be physically interpreted as consisting primarily of depression storage whereas the lower zone represents the soil moisture storage above the capillary zone. The upper zone represents the upper soil layer and interception storage.. which then contributes to baseflow. tension water and free water. and finally the production of runoff. A fraction of total infiltrated water reaches the groundwater storage. Two storage zones control overland flow. Ramírez characteristics. 1976). giving rise to delayed infiltration. Peck. lumped input and lumped parameter model which transforms the rainfall input in the land phase of the hydrologic cycle into channel inflows.S. Streamflow may be computed at several locations in the stream network.. and the flow routing model. and groundwater flow are combined to become total inflow into the channel network which contributes to the streamflow hydrograph. In the lower zone. and supplementary storage. and is assumed to produce a homogeneous hydrologic response. Additional infiltration may occur from depression storage or from runoff. (NWSRFS). The Sacramento soil moisture accounting model is a continuous. whereas the lower zone represents the bulk of the soil moisture and groundwater storage. The NWS model can be partitioned into two major components. This conceptualization is intended to provide a simple but effective representation of the hydrologic process of the catchment starting with precipitation. Tension water is only depleted by evapotranspiration. parameter optimization and verification. and supplements 39 . snow accumulation and ablation. the subsequent vertical and horizontal movement of water through and over the soil. which is fast draining.
sediment yields and general basin hydrology under various combinations of precipitation. The model is set up to simulate basin hydrology on hourly or longer time intervals. orientation of hillslope or aspect.. 1983). soil type. 3) interflow resulting from lateral drainage of the upper zone free water storage. the model generates five components of flow: 1) direct runoff from permanent and temporary impervious areas. 2) surface runoff which occurs when the upper zone free water storage is full and precipitation intensity exceeds the rate of percolation and interflow. which is a nonlinear function of upper zone free available water and lower zone moisture deficiency. The first three runoff components represent the total channel inflow whereas the latter two are the total baseflow. The PRMS model can be either a lumped or distributed parameter model. the model uniformly distributes the evapotranspiration data and computes streamflow over the time basis of rainfall data and over the time interval specified for streamflows.S. HRU's are defined based on characteristics such as slope. Streamflow is computed at the outlet or flow points of each flow plane. The system is physically based such that each component of the hydrologic cycle is expressed as functional relationships of known physical laws or empirical knowledge of the hydrologic process. Movement of water from upper zone to lower zone is achieved through percolation. climate and land use (Leavesley et al. The watershed is partitioned into hydrologic response units (HRU's) which may or may not be associated with individual sub watersheds. Rainfall and evapotranspiration input data are assumed uniform over each SMA area. The first level partitions the watershed into soil moisture accounting areas where each area is a homogeneous unit in terms of the SMA model parameters. Although there are two levels of watershed partitioning. Finally. the total channel inflow constitutes the surface runoff contribution to the streamflow hydrograph routed via unit hydrograph/Muskingum routing methods. Geological Survey (USGS) developed a PrecipitationRunoff Modeling System (PRMS) to simulate surface water runoff. In the NWS model. 40 . Rainfall input data can only be given on an hourly basis whereas evapotranspiration and streamflow input data can be given on a longer time interval as an integer multiple of the time interval of rainfall. 4) supplemental baseflow. capable of simulating mean daily flows and storm flow hydrographs. and a portion of the total baseflow is the subsurface runoff contribution to streamflow. and 5) primary baseflow. Each flow plane is assumed to have homogeneous unit graphMuskingum model parameters. This subsurface flow contribution is added to the routed streamflow at the basin or subbasin outlet using a linear decay weighting function which is similar to the hydrograph routing method. Two levels of watershed partitioning are required in the model. Ramírez baseflow after a period of relatively recent rainfall. vegetation type.Jorge A. The second level of partitioning divides an SMA area into smaller homogeneous units representing individual flow planes (overland and channel). The U. one can assume a flow plane to coincide with the SMA area. In such cases.
Water and energy balances are determined daily for each HRU. Each HRU has a soilzone reservoir. Average root depth of the predominant vegetation on the HRU defines the depth of the soilzone reservoir. The snow component of PRMS simulates the accumulation and depletion of a snowpack on each HRU. daily solar radiation and daily streamflow (for calibration). This model has been incorporated into the Modular Modeling System of the U.S.Jorge A. The inputs are daily precipitation. with time steps in the order of minutes. The energy inputs of air temperature and solar radiation drive the processes of evaporation. maximum and minimum daily air temperature.S. The daily mode simulates hydrologic components on daily averages or total values. The soil zone is represented by two layers. The subsurface reservoir responds faster than the groundwater reservoir. The model is represented by a series of storage (reservoirs). Precipitation is reduced by interception. The model then returns to the daily mode. 41 . Leavesley and Stannard (1990) used ARC/INFO with PRMS to simulate hydrographs in the East Fork Carson River. In these cases. This represents the soil mantle that can lose water by evaporation and transpiration. whereas the lower layer loses water only by transpiration. Ramírez elevation and precipitation distribution (Leavesley et al.. transpiration. Water storage in this zone is increased by infiltration and is decreased by evapotranspiration. This has been the result of improvements in our understanding of the fundamental laws governing hydrologic process behavior and of improvements in computational power and observational capabilities. sublimation and snowmelt. Current Developments Flood hydrology has developed very rapidly after the works of Sherman and Horton in the 1930's and 40's.G. Shorter time simulation. When the soil zone reservoir reaches field capacity. The upper layer (recharge layer) loses water by evaporation and transpiration. the model shifts to the shorter time interval mode until the storm period terminates. excess water percolates to the subsurface and groundwater reservoirs. The USGS model can simulate basin hydrology on a daily and on a shorter time interval for flood simulation. and graphics and image processing tools which have made easier the use of large amounts of input data and have enhanced the analysis and interpretation of the output. 1983). Parallel to the development of increasingly more powerful computer hardware has been the development of computer programming languages. database systems.. which imply similar hydrologic response behavior. which is briefly described in the next section. and Luellwitz (1991) used ARC/INFO with PRMS to simulate hydrographs in a sub alpine watershed in Colorado. whereas the groundwater reservoir responds linearly. Normally the model operates on a daily mode until it reaches dates of storm events. Subsurface and groundwater reservoirs can be defined for each HRU or can be defined for larger areas comprising several HRU's. The subsurface reservoir can be linear or nonlinear. computes water discharge and sediment yield from selected rainfall events.
USA. A model is created by selectively coupling the most appropriate modules from the library to create a suitable model for the desired application. not only at the outlet.S. recognition of the significant impact of spatial variability on hydrologic response has prompted renewed interest in the development of spatially distributed hydrologic models for flood routing and for flood risk analyses.S. This is an important advantage in planning and design problems where predictions are required at a specific river cross section. variable soil and vegetation characteristics. During the last decade. and channel network.g. testing.. Corps of Engineers and ARC/INFO of Environmental Systems Research Institute of California. spatially distributed modeling has the advantage of providing predictions at any location throughout the river network. socalled Geographic Information Systems (GIS). As an example. NEXRAD radar precipitation fields) with the advent and rapid development of remote sensing. Where existing modules do not provide appropriate process algorithms. The MMS modeling system also provides installation capabilities for new module development.S. with physicallybased hydrologic models of watersheds. either from a bank of hydrologic simulation modules. it includes the PRMS model).g. new modules can be developed.e. Their work addresses issues of climatic variability and its impact on hydrologic response. and biogeochemical processes (e. energy. or by an explicit coding of new modules.G. The Modular Modeling System (MMS) (Leavesley et al. at which perhaps no historical observations are available. Ramírez The development of digital terrain models (DTM) and data base systems for spatial data. and evaluation of physicalprocess algorithms and to facilitate integration of userselected sets of algorithms into operational physicalprocess models. as it allows building of different model configurations. These geomorphoclimatic and hydraulic characteristics.. the MMS integrates all of the above referred to improvements in computer software and hardware within a modeling framework aimed at facilitating and improving hydrologic modeling. vegetation and land use characteristics of the basin. In addition to describing the complex interactions of local hillslope processes. soil type. but the geomorphology. has had a tremendous impact on our ability to describe and understand the effects of highly heterogeneous boundary conditions on hydrologic response. and model development and application using the MMS graphical user interface. and hydrologic modeling. precipitation). whose extreme variability in space and time has profound nonlinear effects on hydrologic response. GRASS of the U. 1994) used the GRASS GIS package with PRMS in an earlier version of the MMS to simulate streamflow hydrographs in the Rio Grande basin. 42 . GIS makes it possible to integrate efficiently not only the topography.. 1996) is an integrated system of computer software that has been developed with financial support from the U. Epstein and Ramírez (1993. The MMS enhances the ability of the modeler to address the specific conditions under consideration. to provide the research and operational framework needed to support development. are two examples of GIS packages widely used in conjunction with DTM's. remotely sensed data (i. MMS uses a module library that contains modules for simulating a variety of water.Jorge A. As such. have become increasingly available (e.
infiltration. 1995). Burlando et al. twodimensional. In their application. soil erosion and sediment transport. (1989a and 1989b) analyzed scale effects in distributed rainfallrunoff modeling. and topography..e.Jorge A.. (1994) developed a methodology for the evaluation of flood risk in a distributed modeling framework. Ogden et al. giving rise to complex sets of strongly coupled nonlinear differential equations. For example. They applied this methodology successfully to the Sansobbia river basin in Northern Italy. DTM's and GIS tools. they show that the frequency distribution of the annual maxima is an Extreme Value Type II distribution. this has been brought about by advances in our understanding of the fundamental physical processes underlying the hydrologic cycle and of the solution of the mathematical equations representing those processes. and obtain expressions for its second order moments as a function of climate and geomorphology. Brath et al. Over each pixel of the domain. The geomorphologic model of flood frequency is based on the ideas of Eagleson (1972) introduced earlier in this chapter.. land use. Ogden and Julien (1993). is a distributedparameter. and spatially and temporally variable. Pilotti and Rosso (1990) presented SHELL (i. rainfall interception. Ogden (1992). Their flood risk analyzer (FLORA) combines a geomorphologic model of flood frequency with a distributed description of the basin given in terms of soil types. (1995). Analytical solutions for these equations have been found for the simplest cases only. and of the soil type. Burlando et al. flow is allowed to take place only onedimensionally and in the direction of the steepest slope of either of 2 orthogonal directions in an orthogonal cascade of planes. In addition to the development of improved computational capabilities. Saghafian et al. the various hydrologic processes governing the transformation of rainfall into runoff are extremely complex. land use. and the Soil Conservation Service Curve Number model for infiltration losses. unsteady diffusion wave model that routes overland flow using a diffusion wave approximation for the flow dynamics. The model includes continuous soilmoisture accounting. Physically based distributed modeling of watershed processes has become increasingly feasible in recent years. Generally. a general framework for modeling the distributed response of drainage basins to storm rainfall). and one dimensional and two dimensional numerical solution schemes have been developed and applied. and Ranzi and Rosso (1991b) reported a physically based approach to modeling distributed snowmelt in alpine catchments. (1995) developed two dimensional numerical schemes for routing the excess rainfall through the overland phase of a watershed. Input to each pixel is in the form of 43 . Saghafian (1992). thus assessing the effects of such variability on basin scale estimates. use the approach of Bacchi and Rosso (1988) to derive the second order properties of the frequency distribution of flood peaks for any point of the basin as a function of the stochastic properties of precipitation. nonlinear. Assuming an Independent Poisson Marks model for precipitation (Eagleson. and geomorphologic characteristics of the section in question.g. Ramírez Mancini and Rosso (1989) used a GIS to investigate the spatial variability of SCS Curve Number within the basin. The equations of motion are solved using a finite difference scheme. surface and channel runoff routing. The CASC2D model (e. and Saghafian et al. Runoff potential and infiltration characteristics are described as a function of the SCS Curve Number model. 1972).
44 . More recently.) on runoff production and flood dynamics. this capability allows for consideration of issues of scale and takes into account the impact of microchannel formation (e. 2dimensional hydrodynamic wave model for the simulation of the rainfallrunoff process that allows for fully interactive infiltration. Ramírez precipitation and upstream inflow from all directions. etc.Jorge A. Fiedler and Ramírez developed a numerical scheme capable of describing the flow dynamics as affected by microtopographic conditions and other rapidly varying soil and vegetation characteristics. CASC2D is one of the surfacewater hydrologic models support by the Watershed Modeling System (WMS) under development at Brigham Young University. As such.. Infiltration losses are diagnosed at each pixel using the GreenAmpt infiltration equations. Fiedler and Ramírez (1998) developed a fully unsteady. fully dynamic. rills. In addition to the explicit consideration of fully interactive infiltration.g.
Water Resources Research. Guariso and B. Mickle. (1971). M. B. & Rosso.. R. Estimation of the time scale of the geomorphologic instantaneous unit hydrograph from effective streamflow velocity. Addison Wesley... Amsterdam. (1988). P.S. R. A. In Computer Support for Environmental Impact Assessment eds..G. & Rosso. P. C.. and McGuire. P..... pp.. L. La Barbera. & Giordano. Soc. Brath. M. E.181. Vogt & M. (1988). R. A. C. Brath.20.. eds. In Proc. 91102. Ferral. Am. FLORA: A Distributed Flood Risk Analyzer. Agnese. A Generalized Streamflow Simulation System: Conceptual Modeling for Digital Computers. Mancini. Distributed Modelling System. September 58. 87. Ramírez REFERENCES Abbott. Burnash. Analisi Geomorfoclimatica dei Modeli di Regionalizzazione della Frequenza delle Piene. Abbott. F. Department of Water Resources. Page. & Rosso. Journal of Hydrology. February. L. B. In New Directions for Surface Water Modeling. M. California.165176. (1990). (1986a). NOAA. U. MD.. 6177. 1. FORTRANIV Program for Stanford Watershed Model IV. Burlando.448453. part 4. A. On Hydraulics and Water Engineering.L. Mancini. Bacchi. Bras. & Rosso. G. Engnrs. Journal of Hydrology. A. In Modeling and Simulation. W. (1994).. 45 . 969978. (1986b). On the geomorphoclimatic derivation of flood frequency (peak and volume) at the basin and regional scale. Bacchi. R. & Rosso. L'Aquila... R. N. An Introduction to the European Hydrological System – Système Hydrologique Europèen.. La Barbera. 4559. North Holland. An Introduction to the European Hydrological System – Système Hydrologique Europèen. New Orleans.. Analysis of Scale Effects in Distributed RainfallRunoff Modeling. Hydrology. 15211526. Conf. SHE – 2: Structure of a PhysicallyBased. R. Anderson... National Weather Service and State of California. 87. V. Proceedings 3rd National Conference on Hydraulic Engineering. (1989a). A. Distributed Modelling System.Jorge A.. Elsevier Science B. Kavvas. R. XXI Ital. pp. 1527. IAHS Publ. Silver Springs. Department of Commerce. SHE – 1: History and Philosophy of a PhysicallyBased. Adom. J. pp. M. Vol. NWS.. M. Civ. August 1418. M. eds. pp. D.. B.H.. an Introduction to Hydrologic Science. Sacramento. (1973). R. no. Office of Hydrology. C.. 24(7). D'Asaro. (1989b). Mancini. B. Brath. R. The use of distributed rainfallrunoff models based on GIS at different scales of information. (1989). pp.
J. May 8. Digital Simulation in Hydrology... Bras. (1983). pp.J. pp. A rainfallrunoff simulation model for estimation of flood peaks for small drainage basins. and Valdés. and Bergman. pp13331360. R.. E. M. B. J. Stanford Watershed Model IV. Proc. Maidment. M. L. N. Proceedings. U. and Mays. (1986). University of Utah. A. University of Texas. 91. No. & Rosso.. Civ. Cracow. Chow V.. Wood. Proceedings International Conference on Hydrological Processes in the Catchment. Vol. pp. Journal of the Hydraulics Div. Applied Hydrology. Rep. Nonlinearity and time.variance of the hydrologic response of a small mountain creek. International Association of Hydraulic Research. Incorporation of Channel Losses in the Geomorphologic IUH. In Scale Problems in Hydrology. Problems and Prospects for Predicting Great Salt Lake Levels.A. Crawford. MIT. R.. HYD 147001. D. Stanford. 46 . Rosso. R. (1970). Soc. RodriguezIturbe & E. July. 9.Reidel Publ. UT.R.W. (1988). Center for Public Affairs and Administration. March 2628.S. Am. (1986). (1969). No: 293. D.M. Clark. p. M. Co. eds. Forecasting Great Salt Lake levels. Ramírez Caroni. (1972).. T. R.F. 205230. Kay and H. 39. V. R. J. (ed. DíazGranados.. Cunge. Boston. I. A Comparison between direct and indirect estimation of the Nash model of catchment response. Parsons Laboratory T.H.) (1964). Handbook of Applied Hydrology. R. F. Lichty.2734. Chow. R. M.. and Bras. J. Hydraulic Engineering Laboratory. B. L. Stanford University.Jorge A. McGraw Hill.A. Vol. (1965). Geological Survey Professional Paper 506B. (1945).. T. vol. E. B. No. Mass: D. No. of Civil Engineering. W. Storage and The Unit Hydrograph.F. R.. of the ASCE. Dept. and Moore. Jr. N. (1966).2. Gupta. L. Parsons Laboratory T. Claborn. 123137. Valdés. Eng. Caroni. A. R. MIT.T. Numerical Simulation in Watershed Hydrology. Proc. “On the Subject of a Flood Propagation Method (Muskingum Method). Rep. HY4. E. 179188.38. 292. C. DíazGranados. and Linsley.) P. R. (1983a)..” Journal of Hydraulic Research. Dawdy. (eds. II. 7(2).19. Calif. New York: McGraw Hill.11. & Siccardi. T. 28 pp.. Salt Lake City. A Derived Flood Frequency Distribution Based on the Geomorphoclimatic IUH and the Density Function of Rainfall Excess. Tech. Croley. Diaz. (1985). D. W. V.K.O. Mathematical models of catch water behavior.. and O'Donnell.R.K. Dawdy..
(1979). J. and Crawford. Eagleson. D. Maryland. and Bras. and Ramírez. The NWS DAMBRK Model: Theoretical Background/User Documentation. 878898. Eagleson. A. A PhysicallyBased Flood Frequency Distribution. M. FLDWAV: A Generalized Flood Routing Model. S. (1988). 20(7). and Ramírez. D. (1993).. J.. N. Environmental Protection Agency. National Weather Service. Dooge. Intl. ASCE. B. Boulder. February 8 . Div. (Accepted for publication). Res. Water Resources Research. H.. 668673. Fiedler. Dynamic Hydrology. and Ramírez. (1976). Epstein.L.L. D. (1972). 8(4).Jorge A. C. Maryland. (1978). pp. Jr.. Maryland. Colorado Springs.. J. 47 . 9951002. A. Silver Spring. Fread. NOAA Tech. ASCE Journal of the Hydraulics Division 120(12): 14491467. Memo NWS HYDRO18. Epstein. 1998. Numerical Properties of Implicit FourPoint Finite Difference Equations of Unsteady Flow. User's Manual for the Nonpoint Source (NPS) Model. (1970). Hydrologic Research Laboratory. HRL45. R. (1984). 455464.10. Theoretical Development of Implicit Dynamic Routing Model.L. American Geophysical Union Front Range Meeting. Journal of Geophysical Research.. S.L. D. Spatial Disaggregation for Studies of Climatic Hydrologic Sensitivity. Water Resources Research. A Numerical Method for Hydrodynamic Modeling of Overland Flow. Cincinnati. A. (1988). Ohio. Fread. F. NWS Operational Dynamic Wave Model. National Weather Service. A General Theory of the Unit Hydrograph. Silver Spring. P. J. L. McGraw Hill. ASCE. Info. (1994). Donigian. HRL113. A. National Weather Service. Dynamics of Flood Frequency. S.L. and Lewis. D. Fread.. M. Fread. 64(2): 241256.. Valdés. I. P. (1959). Center. for Numerical Methods in Fluids. Specialty Conf. Hydrologic Research Laboratory. Environ. in Verification of Mathematical and Physical Models. Jour. Md. (1974). D. A. Proc.. R. Ramírez DíazGranados. Silver Spring. pp. Proceedings of 26th Annual Hydr. National Conference on Hydraulic Engineering. J.. Fread. HRL256. D. A Daily Spatial Disaggregation Approach and its Application in Hydrologic Sensitivity Analysis of the Upper Rio Grande to Climate Variability. J. College Park. Colorado. S. U.
Waymire. K. R.S. P. 102(14). Athens. Water Resources Research. Johanson. La Barbera. (1980). Hydrologic Synthesis by Digital Computer.28. (1990). M. R. Water Resources Research. 95(HY3). U. On the fractal dimension of stream networks. Hebson. V. pp. R. Fractal geometry of river networks (abstract). Scale Problems in Hydrology. HEC .Jorge A. M. P. & Rosso. EPA600/980015. E. D. Journal of Hydrology.. 16931702. F. (1989). Howard. La Barbera. K. 21(11). and RodríguezIturbe. and Network Structure in Basin Runoff. F. L. RodríguezIturbe. Gupta. Theoretical Model of Optimal Drainage Networks. Kirshen. Gupta. and Davis. H. and Wood. EPA. AGU. D. Water Resources Monograph 7. Wood. Journal of Hydrology. Water Resources Research.. Desktop Methods for Urban Stormwater Calculation. The Linear Channel and its Effect on the Geomorphologic IUH. B. and Clark. Journal of the Hydraulics Division. A. Hudlow. Chapter 4. and Troutman. (1987). and Mesa. Calif. Washington. D. 15091518. C. American Geophysical Union. Imhoff. R. M. Kibler. 82. 18(5). Journal of Hydrology. Dordrecht. Paper No. eds. (1988). Army Corps of Engineers. Gupta. F. (1982). 26(9). 839860. 68(44). C. EOS Trans.D. 48 . & Rosso. (1986). J. C. Assessment of the Instantaneous Unit Hydrograph Derived from the Theory of Topologically Random Networks.) Urbam Stormwater Hydrology. On Scales. On the Formulation of an Analytical Approach to Hydrologic Response and Similarity at the Basin Scale. In D. Ramírez Gupta. Gravity. (1983). M. E. 21072117. Reidel. DC. The New HEC1 Flood Hydrograph Package. Proceedings of the American Society of Civil Engineers. 3 . R. R. V. A derived Flood Frequency Distribution Using Horton Order Ratios. 65. 175208.. (1983). 1276.Hydrologic Engineering Center. I. and Bras. K. 735741.. O. I. (1981). J. User's Manual for Hydrologic Simulation ProgramFORTRAN (HSPF). Holland: D. Runoff Generation and Hydrologic Response via Channel Network Geomorphology: Recent Progress and Open Problems.. (1982). Water Resources Research. A. V. 65. 95123. Karlinger. (1985). K. and E. F. E. Kibler (eds. aand Waymire. 25(4). Tech. H. Georgia. In: V. (1969).
F. Gupta. S. Water Resources Engineers.L. PrecipitationRunoff Modeling System: User's Manual. Florida.. 1990. L. G. and Ramírez. and Mifflin.. No. Opset: Program for Computerized Selection of Watershed Parameter Values of the Stanford Watershed Model. O. On the Relative Role of Hillslope and Network Geometry in Hydrologic Response. Forecasting Flood Flows in Grassy and Forested Basins Using a Deterministic Distributed Mathematical Model. Energy Dissipation Theories and Optimal Channel Characteristics of River Networks. Luellwitz. Restrepo. R. U. Molnár. 37(7). 34. Hy5. E.. Y. Ramírez Leavesley. 1974. G. E. & Rosso. University of Florida. Reidel. 100.H. M. and Ramírez. M. L. Geological Survey Water Resources Investigations. and Stannard. Leavesley. In: V. L. 247255. (1998a). Metcalf and Eddy. (Updated March 1998). Dixon.M. And Anderson.Jorge A. University of Kentucky. no. May 1. Geological Survey WaterResources Investigations Report. Wood. 834328. Mancini. G. (1998b).. edited by M. Scale Problems in Hydrology. 18091818. (1971). (1986). A. Water Resources Research. Storm Water Management Model. J.181. and Stannard. John C. Liou. L. Vol. (1996) The Modular Modeling System (MMS): User's Manual. Holland: D.. E. Geological SurveyOpen File Report 96151. R. RodríguezIturbe. IAHS Publ. 207 pp. M. Dordrecht. pp. No. An Analysis of Energy Expenditure and Downstream Hydraulic Geometry at Goodwin Creek.. International Association of Scientific Hydrology Publication 129..W. Research Dept. in: New Directions for Surface Water Modeling. Water Resour... 49 . B. Environmental Protection Agency. G. Kavvas. and E. Monro. 18191829. R. Colorado State University. Pp. Department of Earth Resources. 37(7). Using GIS to assess spatial variability of SCS Curve Number at the basin scale.S. Leavesley. J. (1983). 621630. J. Asce. J. Paper 10549) Journal Of The Hydraulics Division. Fort Collins. Morris. (1991). Troutman. P. Molnár. Lexington. H. and Saindon. Mesa. Application of Remotely Sensed Data in a Distributed Parameter Watershed Model U. Inst. Lichty. (1980).435444.H.S. Water Resources Research. (1970).G. I. Master of Science Report. Eric A. (1989).S. Markstrom. K. " (Proc. The Application of a Deterministic Hydrologic Model and GIS to Simulate Water Yield Increase due to Timber Harvest in a Subalpine Watershed. "National Weather Service River Forecasting System. Gainesville. U. T. G. May 1974.. A. Vol. eds. P. P.1.
(1993). Shell: A general framework for modeling the distributed response of a drainage basin. Water Resources Research. Nash. J. Y. Fort Collins. R. Rinaldo. J.G. Isochrons of travel time and distribution of flood storage from a tracer study on a small watershed. F. & Rosso. Journal of Geophysical Research. Kluwer Academic Publishers. M. R. and Julien. Similarity in catchment response 2. S. pp. Guariso. Chapter 6 NATO Advanced Study Institute on Coping with Floods. Giuseppe Rossi. Runoff sensitivity to temporal and spatial rainfall variability at runoff plane and small basin scales. F. F. Bologna. (1976). J. DC. 45(3). Ogden. Ogden. Salas. 114121.. Dissertation. Water Resources Research. Washington. Ogden. Springer Verlag. L. NOAA Technical Memorandum NWS Hydro31. in: Guida al Software Ambientale. 15431547. A Unit Hydrograph Study. (1959). D. 25892597. (1957). A. Pinder. pp. E.. Systematic Determination of Unit Hydrograph Parameters.H. Water Resources Research. 249282. 13(3).F. P.L. Gambolati. Rosso. edited by G. Proceedings of the Institution of Civil Engineers. Nash. and R. Ramírez Nash. in: Computational Methods in Surface Hydrology. J. 587595. 29(8). Richardson. R. Gray & G.121126. J. R. 31(6). Berlin. 111115. CO 80523. 10.. pp. Eds. Patron. of Civil Engineering. Y. Proceedings of the Institution of Civil Engineers. (1992). Pilgrim. D. W. The Form of the Instantaneous Unit Hydrograph. Pilgrim. & Rosso. Water Resources Research. E. 163184. Dept. 208 pp. 12(4). Colorado State University. (1991a) : Flea: Flood Event Analyzer. 1994: Determination of Flood Characteristics by PhysicallyBased Methods. Ph. (1960). J. (1976). E. Determining Runoff from Rainfall. Catchment Modeling and Initial Parameter Estimation for the National Weather Service River Forecasting System. R. (in italian). 50 .517522. Moving rainstorms.L.L. 64(1). edited by G. International Association of Scientific Hydrology Publication.D. E. P. (1977). Travel times and non linearity of flood runoff from tracer measurements on a small watershed. Ramírez. J. (1958). (1995). A. Pilotti. Nash. Twodimensional runoff modeling with weather radar data. Nilgun Hamancioglu and Vujica Yevjevich. C. 77110.Jorge A.H. with Particular Reference to British Catchments. 487496. Brebbia. E. Ranzi. and Julien. (1990). Peck. 17.
J. Rigon. Rinaldo.. and Caamaño. and IjjászVásquez. 18(4). R. Marani. Rinaldo. On the Climatic Dependence of the IUH: A RainfallRunoff Analysis of the Nash Model and the Geomorphoclimatic Theory.Jorge A. 513525. Salm. Water Resources Research.141150. 877886. E. Cambridge University Press. E.(1982a). IAHS Publ. (1982b). and Valdés. 14091420. 15(6). D. E. 51 . R. Excerpta. Energy Dissipation. R.L. RodríguezIturbe..J. and Bras.. Ramírez.. W. Ramírez Ranzi.. Ricca V. R. Water Resources Research. 28(4). Part I .T. & Rosso.205.. Frey.J. (1979). 29(6). Water Resources Research... 914920. & Rigon. Rigon. Ohio State University. Bras. I. in: Snow Hydrology and Forests in High Alpine Areas. RodríguezIturbe.” Water Resources Research. H. R. 15(6). 16351646. A. J.. RodríguezIturbe.. González Sanabria. 14351444. Discharge Response Analysis and Hydrologic Similarity: The Interrelation Between the Geomorphologic IHU and the Storm Characteristics. Issler & B. R. (1993) “Optimal Channel Networks: A Framework for the Study of River Basin Morphology. Rosso. M. González Sanabria. Columbus.. (1991b). Bergmann. Water Resources Research. May. R. (1979). 20(7). A. R. 18(4). Riparian and Aquatic Communities Eds. A.L. Lang. and the ThreeDimensional Structure of River Basins. The Ohio State University Version of the Stanford Streamflow Simulations Model. L. A physically based approach to modeling distributed snowmelt in a small alpine catchment. 887903. Analysis. 10951103. & Caroni.(1986). B. (1972). edited by H. G. no. (1992). Bras. A. G. Water Resources Research. and Valdés.Technical Aspects. R. RodríguezIturbe. B. Runoff Production. Geomorphologic dispersion. J. pp. Marani. Rosso. Rinaldo A. I. I. I. 93108. The Geomorphologic Structure of Hydrologic Response. Chapter 11 of Inland Flood Hazards: Human. M. and IjjászVásquez. Water Resources Research.. I. 2000: Prediction and Modeling of Flood Hydrology and Hydraulics.. Water Resources Research.. 1. R. 27(44).(1991). A Geomorphoclimatic Theory of the Instantaneous Unit Hydrograph. Ellen Wohl. Devoto. RodríguezIturbe.. RodríguezIturbe. (1984). estimation and prediction of the hydrological response from catchment geomorphology. I. A. Nash Model Relation to Horton Order Ratios.
836884. and Karlinger. Geophys. 21. 52 . Reidel. Snyder. U. F. L. 178186.. M. Soil Conservation Service. Troutman. Shreve. Hydrology. (1989).. Union. P. In: V. (1984). National Engineering Handbook. (1985). Saghafian. 447454. 4. Unit Hydrograph Approximation Assuming Linear Flow Through Topologically Random Channel Networks. D. Water Resources Research. Volume II. Valdés. Wood. and RodríguezIturbe. 19. Sec. 75. Troutman. Dordrecht. J. Trans.S. M. Department of Agriculture. R. and Karlinger. L. K. Infinite Topologically Random Channel Networks. Y. Watershed Modeling. Soil Conservation Service. Similarity in catchment response 1. 501505.C. Fiallo. (1967). Program Description and User's Manual for SSARR Model Streamflow Synthesisi and Reservoir Regulation. 15331541. U. I. The Effect of Spatially Varied Characteristics on Watershed Response: A TwoDimensional Distributed Approach. Gupta. U. B. U. Water Resources Research. M. Fort Collins. Sec.S. A RainfallRunoff Analysis of the Geomorphologic IUH. Colorado State University.S. Ph. Hydrology. B. Department of Agriculture. 14211434. Eng. Am. CO. News Rec. S. Civil Engineering Department. I. 21(5). and Karlinger.. 108. 4 of National Engineering Handbook. Englewood Cliffs. Holland: D. 15(6). U. B.C.Jorge A. M. Soil Conservation Service. (1938). F. R. Synthetic Unit Graphs.. B. M. Computer Program for Project Formulation Hydrology. R. Prentice Hall. (1986). Army Corps of Engineers. Washington. (1932). B. Troutman. R. and E. Washington. Julien. Dissertation. D. (1972). On the Expected Width Function for Topologically Random Channel Networks. Y. Program 724K5G0010. and Ogden. 31(6). Sherman. (1972). F. L. RodríguezIturbe. Ramírez Saghafian. Streamflow from Rainfall by the Unit Graph Method.C. U.1995. M. 20. Geol. (1973).F. 743754. D. S. K. eds.. V. Soil Conservation Service. Journal of Applied Probability. New Jersey. Hydrologic Systems. Water Resources Research. Singh. B. S. (1985). P. Stationary rainstorms. Averaging Properties of Channel Networks Using Methods in Stochastic Branching Theory. (1992). Scale Problems in Hydrology.. Washington. D. (1979). Technical Release No. Department of Agriculture. J.
53 . Unsteady OneDimensional Flow over a Plane: The Rising Hydrograph. J. A.A. Ramírez Woolhiser. Water Resources Research.Jorge A. (1967). D. and Liggett. 3(3). 753771.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.