Development of Guidance for Runoff Coefficient Selection and Modified Rational
Unit Hydrograph Method for Hydrologic Design
by
Nirajan Dhakal
A dissertation submitted to the Graduate Faculty of
Auburn University
in partial fulfillment of the
requirements for the Degree of
Doctor of Philosophy
Auburn, Alabama
May 7, 2012
Keywords: Rational Method, Runoff Coefficient, Unit Hydrograph,
Modified Rational Unit Hydrograph
Copyright 2012 by Nirajan Dhakal
Approved by
Xing Fang, Chair, Associate Professor of Civil Engineering
T. Prabhakar Clement, Professor of Civil Engineering
Jose G. Vasconcelos, Assistant Professor of Civil Engineering
Luke J. Marzen, Professor of Geography
ii
Abstract
The rational method is the most widely used method by hydraulic and drainage
engineers to estimate design discharges. The runoff coefficient (C) is a key parameter for
the rational method. Literaturebased C values (C
lit
) are listed for different landuse/land
cover conditions in various design manuals and textbooks, but C
lit
appear not to be
derived from any observed data. In this study, C
lit
values were derived for 90 watersheds
in Texas from two sets of landcover data for 1992 and 2001. C values were also
estimated using observed rainfall and runoff data for more than 1,600 events in the study
watersheds using two different approaches (1) the volumetric approach (C
v
) (2) the rate
based approach (C
rate
).When compared with the C
v
values, about 80 percent of C
lit
values
were greater than C
v
values. This result might indicate that literaturebased C
overestimate peak discharge for drainage design when used with the rational method.
Similarly, when compared with the C
rate
values, about 75 percent of C
lit
values were
greater than C
rate
values, however, for developed watersheds with more impervious cover,
C
lit
values were greater than C
rate
values. Ratebased C were also developed as function of
return period for 36 undeveloped watersheds in Texas using peak discharge frequency
from previously published regional regression equations and rainfall intensity frequency
for return periods of 2, 5, 10, 25, 50, and 100 years. The C values of this study increased
with return period more rapidly than the increase suggested in prior literature.
iii
To use the rational method for hydraulic structures involving storage, the
modified rational method (MRM) was developed. The hydrograph developed using the
MRM can be considered application of a special unit hydrograph (UH) that is termed the
modified rational unit hydrograph (MRUH) in this study. Being a UH, the MRUH can be
applied to nonuniform rainfall distributions and for watersheds with drainage areas
greater than typically used for the rational method (a few hundred acres). The MRUH
was applied to 90 watersheds in Texas using 1,600 rainfallrunoff events. The MRUH
performed as well as other three UH methods (Gamma, ClarkHEC1, and NRCS) when
the same rainfall loss model was used.
iv
Acknowledgments
I would like to express sincere gratitude to my academic advisor, Dr. Xing Fang,
for his valuable guidance, support, encouragement and dynamism throughout the work.
He is an excellent professional researcher as well a very good human being and he will
always remain as an inspiration to me throughout my life.
I would also like to thank Professors T. Prabhakar Clement, Jose G. Vasconcelos
and Luke J. Marzen for serving as my committee members, and for their valuable time to
review my proposal, dissertation and give suggestions that have helped to improve the
research.
The research fund has been supported by the Texas Department of Transportation
(TxDOT). I am grateful to Dr. Theodore G. Cleveland of the Texas Tech University, Dr.
David B. Thompson of R. O. Anderson Engineering, Inc., and Dr. William H. Asquith of
the USGS for their technical guidance in the research.
Finally, I would like to express my deepest gratitude to my family and friends for
their support and encouragement throughout the work.
v
Table of Contents
Abstract ............................................................................................................................... ii
Acknowledgments.............................................................................................................. iv
List of Tables ................................................................................................................... viii
List of Figures ......................................................................................................................x
List of Abbreviations ....................................................................................................... xiv
Chapter 1. Introduction
1.1 Background .......................................................................................................1
1.2 Research Objectives ..........................................................................................5
1.3 Study Area and RainfallRunoff Database .......................................................6
1.4 Organization of Dissertation .............................................................................7
1.5 References .........................................................................................................9
Chapter 2. Estimation of Volumetric Runoff Coefficients for Texas Watersheds
Using LandUse and RainfallRunoff Data
2.1 Abstract ...........................................................................................................12
2.2 Introduction .....................................................................................................13
2.3 Watershed Studied and RainfallRunoff Database .........................................15
2.4 Estimation of Runoff Coefficients Using LULC Data ...................................18
2.5 Estimation of the BackComputed Volumetric Runoff Coefficients
(C
vbc
) Using Observed RainfallRunoff Data .................................................22
2.6 Estimation of Volumetric Runoff Coefficients from the RankOrdered
Pairs of Observed Rainfall and Runoff Depths ..............................................28
vi
2.7 Discussion .......................................................................................................32
2.8 Summary .........................................................................................................34
2.9 Acknowledgements .........................................................................................37
2.10 Notation.........................................................................................................37
2.11 References .....................................................................................................39
Chapter 3. Ratebased Estimation of the Runoff Coefficients for Selected
Watersheds in Texas
3.1 Abstract ...........................................................................................................54
3.2 Introduction .....................................................................................................54
3.3 Study Area and RainfallRunoff Database .....................................................58
3.4 Runoff Coefficients Estimated from Event RainfallRunoff Data .................61
3.5 Runoff Coefficients from the FrequencyMatching Approach .......................65
3.6 Summary .........................................................................................................72
3.7 Acknowledgements .........................................................................................74
3.8 Notation...........................................................................................................74
3.9 References .......................................................................................................76
Chapter 4. Return Period Adjustments for Runoff Coefficients Based on Analysis
in Texas Watersheds
4.1 Abstract ...........................................................................................................89
4.2 Introduction .....................................................................................................89
4.3 Study Watersheds............................................................................................92
4.4 Runoff Coefficients for Different Return Periods ..........................................94
4.5 Discussion .......................................................................................................96
4.6 Summary .........................................................................................................99
4.7 Acknowledgements .......................................................................................100
4.8 Notation.........................................................................................................101
vii
4.9 References .....................................................................................................102
Chapter 5. Modified Rational Unit Hydrograph Method and Applications in Texas
Watersheds
5.1 Abstract .........................................................................................................110
5.2 Introduction ...................................................................................................110
5.3 Applications of MRUH in Texas Watersheds ..............................................117
5.4 Estimated Runoff Hydrographs from Different Unit Hydrograph
Methods ........................................................................................................125
5.5 Sensitivity of the MRUH to unit hydrograph duration .................................128
5.6 Discussion and Summary ..............................................................................129
5.7 Notation.........................................................................................................132
5.8 Appendix A: Statistical Measures to Evaluate Model Performance ............133
5.9 References .....................................................................................................135
Chapter 6. Conclusions and Recommendations
6.1 Conclusions ....................................................................................................153
6.2 Recommendations ..........................................................................................158
6.3 References ......................................................................................................160
viii
List of Tables
Table 2.1 Runoff Coefficients Selected for Various LandCover Classes from
NLCD 2001 ............................................................................................................42
Table 2.2 Statistical Summary of Average C
lit
Using NLCD 1992 and Watershed
Minimum, Average and Maximum C
lit
Using NLCD 2001................................43
Table 2.3 Statistical Summary of Watershedaverage C
lit
Using NLCD 2001 for
Developed and Undeveloped Watersheds .............................................................43
Table 2.4 Statistical Summary of C
vbc
and R
v
from NURP (USEPA 1983) .......................44
Table 2.5 Statistical Summary of Watershedaverage C
vbc
for Developed and
Undeveloped Watersheds .......................................................................................44
Table 2.6 Statistical Summary of C
vr
and Absolute Difference (ABS) of C
vr
with
Watershedaverage C
lit
for 83 Texas Watersheds ..................................................44
Table 2.7 Statistical Summary of C
vr
for Developed and Undeveloped Watersheds
and Absolute Difference (ABS) of C
vr
with Watershedaverage C
vbc
and C
lit
for Developed Watersheds .....................................................................................45
Table 3.1 Statistical Summary of C
rate
Calculated from Observed RainfallRunoff
Event Data ..............................................................................................................80
Table 3.2 Statistical Summary of WatershedMedian, WatershedMean and
Standard Deviation of C
rate
for 80 Texas Watersheds ...........................................80
Table 3.3 Statistical Summary of WatershedMedian C
rate
for Developed and
Undeveloped Watersheds .......................................................................................80
Table 3.4 Statistical Summary of C
r
, and Differences among C
r
, C
rate
and C
lit
for 80
Texas watersheds ...................................................................................................81
Table 3.5 Statistical Summary of C
r
and C
lit
for Developed and Undeveloped
Watersheds .............................................................................................................81
Table 4.1 Statistical Summary of C(T) for Select Return Periods (T) for Texas
Watersheds ...........................................................................................................105
ix
Table 4.2 Statistical Summary of the Frequency Factors C(T)/C(10) for Texas
Watersheds ...........................................................................................................105
Table 4.3 Frequency Factor or Multiplier for Literaturebased Rational Runoff
Coefficient C from Different Sources ..................................................................105
Table 5.1 Quantitative Measures of the Success of the Peak Discharges Modeled
Using MRUH with Back Computed Volumetric Runoff Coefficient (C
vbc
)
and Time of Concentration Estimated Using Four Equations .............................139
Table 5.2 Quantitative Measures of the Success of the Time to Peak Modeled
Using MRUH with BackComputed Volumetric Runoff Coefficient (C
vbc
)
and Time of Concentration Estimated Using Four Equations .............................139
Table 5.3 Quantitative Measures of the Success of the Peak Discharges and Time to
Peak Modeled Using MRUH with Time of Concentration Estimated Using
Kirpich Equation and Runoff Coefficients Estimated Using Two Different
Methods (C
vbc
and C
lit
) .........................................................................................140
Table 5.4 Quantitative Measures of the Success of the Peak Discharges Modeled
Using Four Unit Hydrograph Models for 1,600 RainfallRunoff Events in
90 Texas Watersheds ...........................................................................................140
Table 5.5 Sensitivity of Peak Discharges Modeled using MRUH on Unit
Hydrograph Duration for the Rainfall Event on 07/28/1973 for the
Watershed Associated with the USGS Streamflowgaging Station 08178600
Salado Creek, San Antonio, Texas ......................................................................141
Table 5.6 Sensitivity of Peak Discharges Modeled using MRUH on Unit
Hydrograph Duration for 1,600 rainfall events in 90 Texas Watersheds ............141
x
List of Figures
Fig 2.1. Map showing the U.S. Geological Survey streamflowgaging stations
(dots) associated with the watershed locations in Texas ......................................46
Fig 2.2. Cumulative distributions of C
lit
obtained using NLCD 1992 and NLCD 2001
(top) and C
lit
using NLCD 2001 for developed and undeveloped watersheds
(bottom)..................................................................................................................47
Fig 2.3. Cumulative distributions of C
vbc
for watershedaverage, watershedmedian,
and all rainfallrunoff events (top) and watershedaverage C
vbc
for
developed and undeveloped watersheds (bottom) .................................................48?
Fig 2.4. Volumetric runoff coefficients from different studies versus percent
impervious area including regression lines (top) and runoff coefficients with
one standard deviations for 45 Texas watersheds and 60 NURP watersheds
(bottom)..................................................................................................................49
Fig 2.5. The rankordered pairs of observed runoff and rainfall depths (top) and
runoff coefficients derived from the rankordered pairs of observed runoff
and rainfall depths versus total rainfall depths (bottom). All data presented
are for the USGS gage station 08042650 (North Trinity Basin in Texas) .............50
Fig 2.6. Cumulative distributions of C
vr
and C
lit
(top) and C
vr
for developed and
undeveloped watersheds (bottom) .........................................................................51
Fig 2.7. Runoff coefficients C
vr
, C
vbc
(watershedaverage), and runoff coefficients
R
v
from 60 NURP watersheds versus percent impervious area including
lines for the regression equation (2.6) and the regression equation (2.7) ..............52
Fig 2.8. Runoff coefficients C
lit
(watershedaverage), C
vbc
(watershedaverage), C
vr
,
and R
v
plotted against watershed area (km
2
) ..........................................................52
Fig 2.9. Runoff coefficients C
vbc
(watershedaverage) and C
vr
plotted against C
lit
(watershedaverage) for 83 Texas watersheds .......................................................53
Fig 3.1. Map showing U.S. Geological Survey streamflowgaging stations
representing 80 developed and undeveloped watersheds in Texas. .......................82
xi
Fig 3.2. Average rainfall intensity I and runoff coefficient C
rate
as a function of t
av
for two storm events: (A) on 09/22/1969 in Dry Branch in Fort Worth,
Texas (U.S. Geological Survey [USGS] streamflowgaging station
08048550 Dry Branch at Blandin Street, Fort Worth Texas), and (B) on
04/25/1970 in Honey Creek near Dallas, Texas. (USGS streamflowgaging
station 08058000 Honey Creek subwatershed number 12 near McKinney,
Texas). ....................................................................................................................83
Fig 3.3. Cumulative distributions of runoff coefficient (C
rate
values): (A) for
rainfallrunoff events when the time to peak (T
p
) was less than the time of
concentration (T
c
); for rainfallrunoff events when the time to peak (T
p
) was
greater than or equal to the time of concentration (T
c
); watershedaverage
(mean); and watershedmedian; and (B) watershedmedian C
rate
for
developed and undeveloped Texas watersheds. .....................................................84
Fig 3.4. (A) Runoff coefficient (C
r
) values derived from the rankordered pairs of
observed peak discharge and maximum rainfall intensity during each storm
event in mm/hr at USGS streamflowgaging station 08042650 North Creek
Surface Water Station 28A near Jermyn, Texas, and (B) the rankordered
pairs of the observed peak discharge and the average rainfall intensity for
the same station. .....................................................................................................85
Fig 3.5. Cumulative distributions of: (A) C
r
, watershedmedian C
rate
and C
lit
, and
(B) distributions of C
r
and C
lit
for developed and undeveloped watersheds. ........86
Fig 3.6. Runoff coefficients versus the percentage of impervious area, IMP. ...................87
Fig 3.7. Modeled peak discharges (Q
p
) from rational equation (3.1) using C
r
and
watershedmedian C
rate
for 1,500 rainfallrunoff events in 80 Texas
watersheds against observed peak discharges. .......................................................88
Fig 3.8. Ratebased runoff coefficients C
rate
for all events, watershedmean C
rate
,
and C
r
versus drainage area (km
2
) in 80 Texas watersheds. ..................................88
Fig 4.1. Map showing the locations of U.S. Geological Survey streamflowgaging
stations in Texas associated with the 36 undeveloped watersheds considered
for this study (two stations are very close and overlapped each other). ..............106
Fig 4.2. Estimation of C(T) versus T for three undeveloped Texas watersheds. .............107
Fig 4.3. Box plot for the distribution of runoff coefficients for different return
periods from Texas watersheds, from Hotchkiss and Provaznik (1995) and
from Young et al. (2009). ....................................................................................108
Fig 4.4. Frequency factors for different return periods from Texas watersheds,
Young et al. (2009), French et al. (1974) and FHWA (Jens 1979) and Gupta
(1989). ..................................................................................................................109
xii
Fig 5.1. The modified rational hydrographs for three different cases: (A) Duration
of rainfall (D) is equal to time of concentration (T
c
), (B) Duration of rainfall
is greater than T
c
, and (C) Duration of rainfall is less than T
c
. ............................142
Fig 5.2. The modified rational unit hydrographs (MRUH) developed for: (A) two
lab settings from Yu and McNown (1964) and (B) for the watershed
associated with USGS streamflowgaging station 08157000 Waller Creek,
Austin, Texas. ......................................................................................................143
Fig 5.3. Incremental rainfall hyetograph and observed and modeled runoff
hydrographs using the MRUH for the two lab tests on concrete surfaces:
(A) 152.4 m
?
0.3 m and (B) 76.8 m
?
0.3 m reported by Yu and McNown
(1964). ..................................................................................................................144
Fig 5.4. Map showing the U.S. Geological Survey streamflowgaging stations
(dots) associated with the watershed locations in Texas. ....................................145
Fig 5.5. Incremental rainfall hyetograph and observed and modeled runoff
hydrographs using the MRUH with T
c
estimated by four empirical
equations for the event on 07/08/1973 for the watershed associated with the
USGS streamflowgaging station 08157000 Waller Creek, Austin, Texas. ........146
Fig 5.6. Modeled versus observed peak discharges for 1,600 rainfallrunoff events
in 90 Texas watersheds. Modeled results were developed from MRUH
using event C
vbc
and T
c
estimated using four different methods: (A)
HaktanirSezen equation, (B) JohnstoneCross equation, (C) Williams
equation and (D) Kirpich equation. ...................................................................147
Fig 5.7. Modeled versus observed time to peak for 1,600 rainfallrunoff events in
90 Texas watersheds. Modeled results were developed from MRUH using
event C
vbc
and T
c
estimated using four different methods: (A) Haktanir
Sezen equation, (B) JohnstoneCross equation, (C) Williams equation and
(D) Kirpich equation. ..........................................................................................148
Fig 5.8. Modeled versus observed peak discharges developed from MRUH using
C
vbc
(triangles) and C
lit
(circles) with time of concentration estimated using
Kirpich equation for 90 Texas watersheds. ..........................................................149
Fig 5.9. (A) Modified rational, Gamma, ClarkHEC1, and NRCS unit hydrographs
developed for the watershed associated with USGS streamflowgaging
Station 08048520 Sycamore, Fort Worth, Texas; and (B) Rainfall
hyetograph, observed and modeled runoff hydrographs using the four
different unit hydrographs for the rainfall event on 07/28/1973 for the same
watershed. ............................................................................................................150
Fig 5.10. Modeled versus observed peak discharges using: (A) MRUH, (B) Gamma
UH, (C) ClarkHEC1 UH, and (4) NRCS UH for 1,600 rainfallrunoff
events in 90 Texas watersheds. ............................................................................151
xiii
Fig 5.11. Observed and modeled runoff hydrographs using MRUH with six unit
hydrograph durations (5, 10, 20, 30, 40 and 50 minutes) for the rainfall
event on 07/28/1973 for the watershed associated with the USGS
streamflowgaging station 08178600 Salado Creek, San Antonio, Texas. ..........152
xiv
List of Abbreviations
ASCE American Society of Civil Engineers
DEM Digital Elevation Model
GIS Geographic Information System
IDF IntensityDurationFrequency
IUH Instantaneous Unit Hydrograph
LULC Landuse/landcover
MRM Modified Rational Method
MRUH Modified Rational Unit Hydrograph Method
NLCD National Land Cover Dataset
NRCS Natural Resources Conservation Service
NURP National Urban Runoff Program
RHM Rational Hydrograph Method
SBUH Santa Barbara Unit Hydrograph
TxDOT Texas Department of Transportation
USEPA United States Environmental Protection Agency
USGS U.S. Geological Survey
WPCF Water Pollution Control Federation
1
Chapter1. Introduction
1.1 Background
Rational method and Runoff coefficient
Early storm water or catchment runoff estimation throughout the world was based
on designer?s experience and judgment. Current practice is that the watershed that is to be
drained by a proposed storm sewer system will be generally divided into one or more
subcatchments or subwatersheds that are of reasonable size and are approximately
homogeneous in nature. These urban watersheds may include residential, commercial or
industrial areas, but usually have larger proportions of pavement and the streets and roads
which are the principal surface drainage conveyance, have short time of concentration,
and have welldefined flow paths, typically through gutters, ditches and medians of
streets and roads. Each year, billions of dollars are spent on new construction of drainage
structures. For the safety, design of hydraulic structures is done based on the peak
discharge (Q
p
) as the design flow. Therefore, Q
p
is the major hydrological parameter
required for the hydraulic design purpose. Additional parameters such as volume of
runoff and time of peak flow are required in some cases such as for the design of
facilities that use storage such as detention and retention basins.
The rational method is the most widely used method by hydraulic and drainage
engineers to estimate peak design discharges, which are used to size a variety of drainage
2
structures for small urban (developed) and rural (undeveloped) watersheds (Viessman
and Lewis 2003). The rational method was developed in the United States by Emil
Kuichling (1889) and introduced to Great Britain by LloydDavies (1906). The peak
discharge (Q
p
in m
3
/s in SI units or ft
3
/s in English units) for the method is computed
using:
CIAmQ
p 0
? (1.1)
where C is the runoff coefficient (dimensionless), I is the average rainfall intensity
(mm/hr or in./hr) for a storm with a duration equal to a critical period of time (typically
assumed to be the time of concentration), A is the drainage area (hectares or acres), and
m
o
is a dimensional correction factor (1/360 = 0.00278 in SI units, 1.008 in English
units).
C is the variable of the rational method least amenable to precise determination,
and estimation of C calls for judgment on the part of the engineer (ASCE and WPCF
1960; TxDOT 2002). C can vary substantially depending on watershed conditions.
Therefore, research to document appropriate values of C is needed. Typical C values,
representing the integrated effects of many watershed conditions (C
lit
), are listed for
different landuse/landcover (LULC) conditions in various design manuals and
textbooks (Chow et al. 1988; Viessman and Lewis 2003). Values of C
lit
published by the
joint committee were obtained from a survey, which received ?71 returns of an extensive
questionnaire submitted to 380 public and private organizations throughout the United
States.? The results represented decades of professional practice experience using the
rational method to determine runoff volumes in stormsewer design applications (ASCE
and WPCF 1960). No justification based on observed rainfall and runoff data for the
3
selected C
lit
values was provided in the ASCE and WPCF (1960) manual. In short, we
conclude that C
lit
values appeared heuristically determined, and therefore comparison of
C values derived from observed rainfall and runoff data to the C
lit
values is important.
In this study, we focus on applying different methods to estimate C for 90
watersheds in Texas using observed rainfall and runoff data for 1,600 events. C
lit
values
were derived from two sets of LULC data for 1992 and 2001. Volumetric runoff
coefficients (C
v
) were estimated by event totals of observed rainfall and runoff depths
from more than 1,600 events observed in the watersheds. C
v
values were also estimated
using rankordered pairs of rainfall and runoff depths (frequency matching).
It is important to stress that rational method is the ratebased method (eq. 1.1).
Current runoff coefficients given in textbooks and design manuals are neither volumetric
nor ratebased because they were not derived from observed data but are used for the
ratebased rational method. In this study, the ratebased runoff coefficients, C
rate
, were
estimated for each of 1,600 rainfallrunoff events and the time window used to determine
average rainfall intensity was time of concentration computed using the Kirpich method.
Subsequently, the frequencymatching approach was used to extract a representative
runoff coefficient, C
r
, for each watershed. C values developed from both the volumetric
and ratebased approaches are compared independently with C
lit
values.
A substantial criticism of the rational method arises because observed C values
vary from storm to storm (Schaake et al. 1967; Pilgrim and Cordery 1993). The C has
been considered a function of return period by various researchers (Jens 1979; Pilgrim
and Cordery 1993; Hotchkiss and Provaznik 1995; Titmarsh et al. 1995; Young et al.
2009). Using watershed parameters, such as drainage area, slope, and channel length and
4
regression equations of discharge and rainfall intensity at different return periods T of 2,
5, 10, 25, 50, and 100years, the C(T) were estimated in 36 undeveloped watersheds in
Texas. Subsequently, frequency factors C
f
(T) = C(T)/C(10) were computed. Results of
C(T) and frequency factors C
f
(T) were analyzed and compared with previous studies.
Modified Rational Method (MRM)
Incorporation of detention basins to mitigate effects of urbanization on peak flows
required design methodologies to include the volume of runoff as well as the peak
discharge (Rossmiller 1980). To use the rational method for hydraulic structures
involving storage, the modified rational method (MRM) was developed (Poertner 1974).
While the original rational method is meant to produce only the peak design discharge,
the MRM produces a runoff hydrograph and the runoff volume of the entire watershed.
The MRM, which has found widespread use in the engineering practice since 1970s, is
used to size detention/retention facilities for a specified recurrence interval and
concurrent release rate. The MRM is based on the same assumptions as the conventional
rational method, that is, the rainfall is uniform in space over the drainage area being
considered and the rainfall intensity is uniform throughout the duration of the storm
(Rossmiller 1980).
The MRM was revisited and reevaluated in this study. The hydrograph developed
from application of the MRM is a special case of the unit hydrograph method and will be
termed the modified rational unit hydrograph (MRUH) in this study. Being a unit
hydrograph, the MRUH can be applied to nonuniform rainfall distributions. Furthermore,
the MRUH can be used on watersheds with drainage areas in excess of the typical limit
5
for application of the rational method or the modified rational method (a few hundred
acres). The MRUH was applied to 90 watersheds in Texas using 1,600 rainfallrunoff
events. The Gamma UH, ClarkHEC1 UH, and NRCS dimensionless UH were also used
to predict peak discharges of all events in the database.
1.2 Research Objectives
This research work is a part of TxDOT Project 06070 ?Use of the Rational and
the Modified Rational Methods for TxDOT Hydraulic Design?. The principal objective
of the project is to evaluate appropriate conditions for the use of the rational method and
modified rational methods for designs on small watersheds, evaluate and refine, if
necessary, current tabulated values of the runoff coefficient and construct guidelines for
TxDOT analysts for the selection of appropriate parameter values for Texas conditions.
The specific objectives are:
1. Estimation of areallyweighted literaturebased runoff coefficients (C
lit
) for the study
watersheds using landuse data.
2. Estimation of volumetric runoff coefficients (C
v
) for the study watersheds using
rainfallrunoff data.
3. Estimation of the rate based runoff coefficients (C
rate
) for the study watersheds using
rainfallrunoff data and comparison of C
lit
with both the C
v
and C
rate
.
4. Estimation of the runoff coefficients for different return periods and compare the
current frequency multiplier C(T)/C(10) in the literature with our results.
6
5. Evaluate the applicability of the modified rational unit hydrograph method (MRUH) if
blindly applied to watersheds of sizes greater than originally intended with either the
rational method or the modified rational method (that is, a few hundred acres).
6. Study the effects of runoff coefficient and the timing parameters on predictions of
runoff hydrographs using MRUH.
1.3 Study Area and RainfallRunoff Database
Watershed data from a larger dataset accumulated by researchers from the U.S.
Geological Survey (USGS) Texas Water Science Center, Texas Tech University,
University of Houston, and Lamar University (Asquith et al. 2004) and previously used
in a series of research projects funded by the Texas Department of Transportation
(TxDOT) were used for this study. The data were collected as a part of USGS small
watershed projects and urban watershed studies during 1959?1986 (Asquith et al. 2004).
The original data, available in the form of 220 printed USGS data reports, were
transcribed to digital format manually (Asquith et al. 2004). Incidentally, these data also
are used by Cleveland et al. (2006), Asquith and Roussel (2007), Fang et al. (2007,
2008), and Dhakal et al. (2012).
The dataset comprises 90 USGS streamflowgaging stations in Texas, each
representing a different watershed (Fang et al. 2007, 2008). There are 29, 21, 7, 13
watersheds in Austin, Dallas, Fort Worth, and San Antonio areas, respectively, and
remaining 20 watersheds are small rural watersheds in Texas. The drainage area of study
watersheds ranged from approximately 0.8?440.3 km
2
(0.3?170 mi
2
), with median and
mean values of 17.0 km
2
(6.6 mi
2
) and 41.1 km
2
(15.9 mi
2
), respectively. There are 33,
7
57, and 80 study watersheds with drainage areas less than 13 km
2
(5 mi
2
), 26 km
2
(10
mi
2
), and 65 km
2
(25 mi
2
), respectively. The stream slope of study watersheds ranged
from approximately 0.0022?0.0196, with median and mean values of 0.0075 and 0.081,
respectively. The percentage of impervious area (IMP) of study watersheds ranged from
approximately 0.0?74.0, with median and mean values of 18.0 and 28.4, respectively.
The rainfallrunoff dataset comprised about 1,600 rainfallrunoff events. The
number of events available for each watershed varied; for some watersheds only a few
events were available whereas for some others as many as 50 events were available
(Cleveland et al. 2006). Values of rainfall depths for 1,600 events ranged from 3.56 mm
(0.14 in.) to 489.20 mm (19.26 in.), with median and mean values of 57.15 mm (2.25 in.)
and 66.29 mm (2.61 in.), respectively. Values of maximum rainfall intensities calculated
using time of concentration for 1,600 events ranged from 0.01 mm/min (0.03 in./hr) to
2.54 mm/min (6.01 in./hr), with median and mean values of 0.25 mm/min (0.58 in./hr)
and 0.30 mm/min (0.72 in./hr), respectively.
1.4 Organization of Dissertation
This dissertation is organized into six chapters. Chapter two to five are organized
in journal paper format prepared for ASCE journal publication. Parts of results of the
study were presented in two conference papers:
1. Dhakal, N., Fang, X., Cleveland, T. G., Thompson, D. B., and Marzen, L. J. (2010).
"Estimation of rational runoff coefficients for Texas watersheds." Proceeding (CDROM)
for 2010 World Environmental and Water Resources Congress, Providence, Rhode
Island.
8
2. Nirajan Dhakal, Xing Fang, Theodore G. Cleveland, and David B. Thompson, 2011.
?Revisiting Modified Rational Method?. Proceeding (CDROM) for 2011 World
Environmental and Water Resources Congress, Palm Springs, CA, May 2226, 2011.
Chapter two deals with the estimation of the volumetric runoff coefficients (C
v
)
for the study watersheds using both landuse and rainfallrunoff data. Two regression
equations of C
v
versus percent impervious area were developed and combined into a
single equation which can be used to rapidly estimate C
v
values for similar Texas
watersheds. The work of this chapter has been published in the ASCE Journal of
Irrigation and Drainage Engineering (Nirajan Dhakal, Xing Fang, Theodore G.
Cleveland, David B. Thompson, William H. Asquith, and Luke J. Marzen, 2012
(January). ?Estimation of Volumetric Runoff Coefficients for Texas Watersheds Using
LandUse and RainfallRunoff Data.? ASCE Journal of Irrigation and Drainage
Engineering, 138(1):4354, DOI=10.1061/(ASCE)IR.19434774.0000368).
Chapter three deals with the estimation of the rate based runoff coefficients for
the study watersheds from the rainfallrunoff data. An equation applicable to many Texas
watersheds is proposed to estimate C as a function of impervious area. The work of this
chapter has been revised and resubmitted for publication in the ASCE Journal of
Hydrologic Engineering (Nirajan Dhakal, Xing Fang, William H. Asquith, Theodore G.
Cleveland, and David B. Thompson. ?Ratebased Estimation of the Runoff Coefficients
for Selected Watersheds in Texas?. ASCE Journal of Hydrologic Engineering).
Chapter four deals with the estimation of the runoff coefficients based on the
return period. The work of this chapter has been submitted for review and publication in
the ASCE Journal of Irrigation and Drainage Engineering (Nirajan Dhakal, Xing Fang,
9
William H. Asquith, Theodore G. Cleveland, and David B. Thompson. ?Return Period
Adjustments for Runoff Coefficients Based on Analysis in Texas Watersheds?. ASCE
Journal of Irrigation and Drainage Engineering. In Review).
Chapter five deals with the development and application of the Modified Rational
Unit Hydrograph Method (MRUH) for the study watersheds. The Gamma UH, Clark
HEC1 UH, and NRCS dimensionless UH were also used to predict peak discharges of
all events in the database. The work of this chapter has been submitted for review and
publication in the ASCE Journal of Hydrologic Engineering (Nirajan Dhakal, Xing Fang,
David B. Thompson, and Theodore G. Cleveland. ?Modified Rational Unit Hydrograph
Method and Applications in Texas Watersheds?. ASCE Journal of Hydrologic
Engineering. In Review).
Chapter six summarizes the conclusion of the study and provides some
recommendations for the future study in this area.
1.5 References
ASCE, and WPCF. (1960). Design and construction of sanitary and storm sewers,
American Society of Civil Engineers (ASCE) and Water Pollution Control Federation
(WPCF).
Asquith, W.H., Thompson, D.B., Cleveland, T.G., and Fang, X. (2004). ?Synthesis of
rainfall and runoff data used for Texas department of transportation research projects
0?4193 and 0?4194.? U.S. Geological Survey, OpenFile Report 2004?1035, Austin,
Texas.
Asquith, W.H., and Roussel, M.C. (2007). ?An initialabstraction, constantloss model
for unit hydrograph modeling for applicable watersheds in Texas.? U.S. Geological
Survey Scientific Investigations Report 2007?5243, 82 p.
[http://pubs.usgs.gov/sir/2007/5243].
Chow, V. T., Maidment, D. R., and Mays, L. W. (1988). Applied hydrology, McGraw
Hill, New York.
10
Cleveland, T.G., He, X., Asquith, W.H., Fang, X., and Thompson, D.B. (2006).
?Instantaneous unit hydrograph evaluation for rainfallrunoff modeling of small
watersheds in north and south central Texas.? Journal of Irrigation and Drainage
Engineering, 132(5), pp. 479?485.
Dhakal, N., Fang, X., Cleveland, T.G., Thompson, D.B., Asquith, W.H., and Marzen, L.J.
(2012). ?Estimation of volumetric runoff coefficients for Texas watersheds using
landuse and rainfallrunoff data.? Journal of Irrigation and Drainage Engineering,
138(1):4354.
Fang, X., Thompson, D.B., Cleveland, T.G., and Pradhan, P. (2007). ?Variations of time
of concentration estimates using NRCS velocity method.? Journal of Irrigation and
Drainage Engineering, 133(4), pp. 314?322.
Fang, X., Thompson, D.B., Cleveland, T.G., Pradhan, P., and Malla, R. (2008). ?Time of
concentration estimated using watershed parameters determined by automated and
manual methods.? Journal of Irrigation and Drainage Engineering, 134(2), pp. 202?
211.
Hotchkiss, R. H., and Provaznik, M. K. (1995). ?Observations on the rational method C
value.? Watershed management: planning for the 21st century; proceedings of the
symposium sponsored by the Watershed Management Committee of the Water
Resources, New York.
Jens, S. W. (1979). Design of Urban Highway Drainage, Federal Highway
Administration (FHWA), Washington DC.
Kuichling, E. (1889). "The relation between the rainfall and the discharge of sewers in
populous areas." Transactions, American Society of Civil Engineers 20, 1?56.
LloydDavies, D. E. (1906). "The elimination of storm water from sewerage systems."
Minutes of Proceedings, Institution of Civil Engineers, Great Britain, 164, 41.
Poertner, H. G. (1974). "Practices in detention of urban stormwater runoff: An
investigation of concepts, techniques, applications, costs, problems, legislation, legal
aspects and opinions." American Public Works Association, Chicago, IL.
Pilgrim, D. H., and Cordery, I. (1993). ?Flood runoff.? In Handbook of Hydrology, D. R.
Maidment, ed., McGrawHill, New York, 9.1?9.42.
Rossmiller, R. L. (1980). "Rational formula revisited." International Symposium on
Urban Storm Runoff, University of Kentucky, Lexington, KY., 112.
Roussel, M.C., Thompson, D.B., Fang, D.X., Cleveland, T.G., and Garcia, A.C. (2005).
?Timing parameter estimation for applicable Texas watersheds.? 0?4696?2, Texas
Department of Transportation, Austin, Texas.
11
Schaake, J. C., Geyer, J. C., and Knapp, J. W. (1967). ?Experimental examination of the
rational method.? Journal of the Hydraulics Division, ASCE, 93(6), pp. 353?370.
Titmarsh, G. W., Cordery, I., and Pilgrim, D. H. (1995). ?Calibration Procedures for
Rational and USSCS Design Flood Methods.? Journal of Hydraulic Engineering,
121(1), pp. 61?70.
Viessman, W., and Lewis, G. L. (2003). Introduction to hydrology, 5th Ed., Pearson
Education, Upper Saddle River, N.J., 612.
Young, C. B., McEnroe, B. M., and Rome, A. C. (2009). ?Empirical determination of
rational method runoff coefficients.? Journal of Hydrologic Engineering, 14, 1283 p.
12
Chapter 2. Estimation of Volumetric Runoff Coefficients for Texas Watersheds
Using LandUse and RainfallRunoff Data
2.1 Abstract
The rational method for peak discharge (Q
p
) estimation was introduced in the
1880s. Although the rational method is considered simplistic, it remains an effective
method for estimating peak discharge for small watersheds. The runoff coefficient (C) is
a key parameter for the rational method and there are various ways to estimate C.
Literaturebased C values (C
lit
) are listed for different landuse/landcover (LULC)
conditions in various design manuals and textbooks. However, these C
lit
values were
developed without much basis on observed rainfall and runoff data. C
lit
values were
derived for 90 watersheds in Texas from two sets of LULC data for 1992 and 2001; C
lit
values derived from the 1992 and 2001 LULC datasets were essentially the same. Also
for this study, volumetric runoff coefficients (C
v
) were estimated by event totals of
observed rainfall and runoff depths from more than 1,600 events observed in the
watersheds. Watershedmedian and watershedaverage C
v
values were computed and both
are consistent with the data from the National Urban Runoff Program. C
v
values also
were estimated using rankordered pairs of rainfall and runoff depths (frequency
matching). As anticipated, C values derived by all three methods (literaturebased, event
totals, and frequency matching) consistently have larger values for developed watersheds
than undeveloped watersheds. Two regression equations of C
v
versus percent impervious
13
area were developed and combined into a single equation which can be used to rapidly
estimate C
v
values for similar Texas watersheds.
2.2 Introduction
Estimation of peak discharge and runoff values for use in designing certain
hydraulic structures (e.g., crossroad culverts, drainage ditches, urban storm drainage
systems, and highway bridge crossings) are important and challenging aspects of
engineering hydrology (Viessman and Lewis 2003). Various methods are available to
estimate peak discharges and runoff volumes from urban watersheds (Chow et al. 1988).
The rational method is the most widely used method by hydraulic and drainage engineers
to estimate design discharges, which are used to size a variety of drainage structures for
small urban (developed) and rural (undeveloped) watersheds (Viessman and Lewis
2003). The rational method was developed in the United States by Emil Kuichling (1889)
and introduced to Great Britain by LloydDavies (1906). The peak discharge (Q
p
in m
3
/s
in SI units or ft
3
/s in English units) for the method is computed using:
,AICmQ
op
?
(2.1)
where C is the runoff coefficient (dimensionless), I is the average rainfall intensity
(mm/hr or in./hr) for a storm with a duration equal to a critical period of time (typically
assumed to be the time of concentration), A is the drainage area (hectares or acres), and
m
o
is a dimensional correction factor (1/360 = 0.00278 in SI units, 1.008 in English
units).
The precise definition and subsequent interpretation of C varies. The C of a
watershed can be defined either as the ratio of total depth of runoff to total depth of
14
rainfall or as the ratio of peak rate of runoff to rainfall intensity for the time of
concentration (Wanielista and Yousef 1993). Kuichling (1889) analyzed observed rainfall
and discharge data for developed urban watersheds in Rochester, NY, and computed the
percentage of the rainfall discharged during the period of greatest flow as Q
p
/(IA), which
is equal to runoff coefficient C from equation (2.1). Kuichling concluded that the
percentage of the rainfall discharged for any given watershed studied is nearly equal to
the percentage of impervious surface within the watershed, and this is the original
meaning of C introduced by Kuichling (1889). Using Kuichling?s definition, C = 0 for a
strictly pervious surface and C = 1 for a strictly impervious surface.
C is the variable of the rational method least amenable to precise determination,
and estimation of C calls for judgment on the part of the engineer (ASCE and WPCF
1960; TxDOT 2002). Typical C values, representing the integrated effects of many
watershed conditions, are listed for different landuse/landcover (LULC) conditions in
various design manuals and textbooks (Chow et al. 1988; Viessman and Lewis 2003).
The source of these published C values (literaturebased C, called C
lit
in this paper)
derives from the 1960 sanitary and storm sewer design manual produced by a joint
committee of the American Society of Civil Engineers (ASCE) and the Water Pollution
Control Federation (WPCF). Values of C
lit
published by the joint committee were
obtained from a survey, which received ?71 returns of an extensive questionnaire
submitted to 380 public and private organizations throughout the United States.? The
results represented decades of professional practice experience using the rational method
to determine runoff volumes in stormsewer design applications (ASCE and WPCF
1960). No justification based on observed rainfall and runoff data for the selected C
lit
15
values was provided in the ASCE and WPCF (1960) manual. However, analysis of
observed rainfall and runoff data was presented by Kuichling (1889).
In this paper, three methods were implemented to estimate C for 90 watersheds in
Texas (Figure 2.1). The first method used LULC information for a watershed and
published C
lit
values for various land uses to derive a watershedcomposite C
lit
. The
second method estimated the volumetric runoff coefficient (C
v
) values by the ratio of
total runoff depth to total rainfall depth for individual storm events. About 1,600 rainfall
runoff events measured from 90 Texas watersheds were analyzed to determine event,
watershedmedian, and watershedaverage C
v
values. C
v
determined from storm events is
called the backcomputed volumetric runoff coefficient (C
vbc
) in this paper. The third
method computed probabilistic C
v
values from the rankordered pairs of observed rainfall
and runoff depths of a watershed and extracted a representative C
v
for the watershed from
the plot of C
v
versus rainfall depths. C
v
determined from the rankordered data is called
the rankordered volumetric runoff coefficient (C
vr
) in this paper. The third method is
similar to the procedure used by Schaake et al. (1967). C values estimated by the three
different methods were analyzed and compared. Regression equations of C
vbc
and C
vr
versus percent impervious area are presented.
2.3 Watersheds Studied and RainfallRunoff Database
Watershed data taken from a larger dataset (Asquith et al. 2004) accumulated by
researchers from the U.S. Geological Survey (USGS) Texas Water Science Center, Texas
Tech University, University of Houston, and Lamar University, and previously used in a
series of research projects funded by the Texas Department of Transportation (TxDOT)
16
were used for this study. The dataset comprises 90 USGS streamflowgaging stations in
Texas, each representing a different watershed (Fang et al. 2007, 2008). Location and
distribution of the stations in Texas are shown in Figure 2.1. There are 29, 21, 7, 13
watersheds in Austin, Dallas, Fort Worth, and San Antonio areas, respectively, and
remaining 20 watersheds are small rural watersheds in Texas (Figure 2.1). The drainage
area of study watersheds ranged from approximately 0.8?440.3 km
2
(0.3?170 mi
2
), with
median and mean values of 17.0 km
2
(6.6 mi
2
) and 41.1 km
2
(15.9 mi
2
), respectively.
There are 33, 57, and 80 study watersheds with drainage areas less than 13 km
2
(5 mi
2
),
26 km
2
(10 mi
2
), and 65 km
2
(25 mi
2
), respectively. The stream slope of study
watersheds ranged from approximately 0.0022?0.0196, with median and mean values of
0.0075 and 0.081, respectively. The percentage of impervious area (IMP) of study
watersheds ranged from approximately 0.0?74.0, with median and mean values of 18.0
and 28.4, respectively.
Many would argue that the application of the rational method is not appropriate
for the range of watershed areas presented in this study. For example, watershed
drainage area is a criteria used to select a hydrologic method (Chow et al. 1988) to
compute peak discharge according to TxDOT guidelines for drainage design. The
TxDOT guidelines recommend the use of the rational method for watersheds with
drainage areas less than 0.8 km
2
(200 acres) (TxDOT 2002). However, French et al.
(1974) estimated values of the runoff coefficient in New South Wales, Australia, for 37
rural watersheds, ranging in size up to 250 km
2
(96 mi
2
). Young et al. (2009) determined
runoff coefficients for 72 rural watersheds in Kansas with drainage areas up to 78 km
2
17
(30 mi
2
). ASCE and WPCF (1960) made the following statements when the rational
method was introduced for design and construction of sanitary and storm sewers:
?Although the basic principles of the rational method are applicable to large drainage
areas, reported practice generally limits its use to urban areas of less than 5 sq miles.
Development of data for application of hydrograph methods is usually warranted on
larger areas? (ASCE and WPCF 1960, p. 32).
Chow et al. (1988) and Viessman and Lewis (2003) do not specify an area limit
for application of the rational method. Pilgrim and Cordery (1993) stated that the rational
method is one of the three methods widely used to estimate peak flows for small to
medium sized basins. ?It is not possible to define precisely what is meant by ?small? and
?medium? sized, but upper limits of 25 km
2
(10 mi
2
) and 550 km
2
(200 mi
2
), respectively,
can be considered as general guides? (Pilgrim and Cordery 1993, p. 9.14). Results of this
study will further indicate that there is no demonstrable trend in runoff coefficient with
drainage area.
The rainfallrunoff dataset comprised about 1,600 rainfallrunoff events recorded
during 1959?1986. The number of events available for each watershed varied; for some
watersheds only a few events were available whereas for some others as many as 50
events were available (Cleveland et al. 2006). Values of rainfall depths for 1,600 events
ranged from 3.56 mm (0.14 in.) to 489.20 mm (19.26 in.), with median and mean values
of 57.15 mm (2.25 in.) and 66.29 mm (2.61 in.), respectively. Values of maximum
rainfall intensities calculated using time of concentration for 1,600 events ranged from
0.01 mm/min (0.03 in./hr) to 2.54 mm/min (6.01 in./hr), with median and mean values of
0.25 mm/min (0.58 in./hr) and 0.30 mm/min (0.72 in./hr), respectively.
18
A geospatial database was developed from another TxDOT project (Roussel et al.
2005), containing watershed boundaries for the 90 watersheds, delineated using a 30
meter digital elevation model (DEM). The geospatial database contains watershed
drainage area, longitude and latitude of the USGS streamflowgaging station, which was
treated as the outlet of the watershed, and 42 watershed characteristics (e.g., main
channel length, channel slope, basin width, etc.) of each individual watershed (Roussel et
al. 2005). Each of the 90 watersheds was classified as either developed (urbanized) or
undeveloped (Roussel et al. 2005, Cleveland et al. 2008). Fortyfour developed
watersheds are located in four metropolitan areas in Texas (Austin, Dallas, Fort Worth,
and St. Antonio) and used for USGS urban studies from 1959 to 1986. Thirtysix
undeveloped watersheds include 20 small rural watersheds and 16 watersheds in
suburban of the four metropolitan areas. The classification scheme of developed and
undeveloped watersheds parallels and accommodates the disparate discussion and
conceptualization in more than 220 USGS reports that provided the original data for the
rainfall and runoff database (Asquith et al. 2004). Although this binary classification
seems arbitrary, it was purposeful and reflected the uncertainty in watershed development
condition at the time the rainfallrunoff data were collected (Asquith and Russell 2007).
This binary classification was successfully used to develop regression equations to
estimate the shape parameter and the time to peak for regional Gamma unit hydrographs
for Texas watersheds (Asquith et al. 2006).
2.4 Estimation of Runoff Coefficients Using LULC Data
19
C is strongly dependent on land use, and to a lesser extent, on watershed slope
(Schaake et al. 1967; ASCE 1992). For watersheds with multiple landuse classes, a
composite (areaweighted average) runoff coefficient, C
lit
, can be estimated using:
,
1
1
?
?
?
?
?
n
i
i
n
i
ii
lit
A
AC
C (2.2)
where, i = i
th
subarea with particular landuse type, n = total number of landuse classes
in the watershed, C
i
= literaturebased runoff coefficient for i
th
landuse class, and A
i
=
subarea size for i
th
landuse class in the watershed (TxDOT 2002). In this study,
watershedcomposite C
lit
values were derived for the 90 watersheds in Texas using LULC
information and published C
lit
values from various literatures. A geographic information
system (GIS) was used for subareal extraction of different LULC classes within a
particular watershed (ESRI 2004). The 1992 and 2001 National Land Cover Data
(NLCD) for Texas were obtained from the USGS website http://seamless.usgs.gov/
(accessed on May 30, 2008).
Each watershed has different LULC classes distributed within its boundary. Of 16
LULC classes from the NLCD 2001 data, 15 were used for the 90 watersheds studied;
definitions of NLCD LULC classes are available at
http://www.epa.gov/mrlc/definitions.html (accessed on May 30, 2008). Runoff
coefficients were assigned for the 12NLCD, 2001LULC classes or mixed classes as
listed in Table 2.1, which are based on 15NLCD 2001LULC classes; the table includes
sources and references for the selected C values. From all sources considered, C values
typically were not available for most of 15NLCD LULC classes, but similar landuse
types from literature were identified to match NLCD LULC (Table 2.1). A C value of 1
20
was assigned to open water, woody wetlands and emergent herbaceous wetlands and is
not shown in Table 2.1. For the other LULC classes a range of C values were available
from the mentioned sources under similar LULC types, and the average values (listed in
column 3 of Table 2.1) were taken as literature C values for the study before a sensitivity
analysis of C
lit
on selected C values for different LULC classes from literature was
conducted.
Using NLCD 2001 data and standard published mean C values (Table 2.1),
composite runoff coefficients, C
lit
, for the 90 Texas watersheds were developed using
equation (2.2) (Table 2.2). Values of C
lit
ranged from 0.29 to 0.63, with median and mean
values of 0.50 and 0.47, respectively (Table 2.2). Estimates of C
lit
for a given watershed
may differ, depending on the experience and judgment used in assigning them to LULC
classes and estimating areas for landuse classes. For example, Harle (2002) determined
C
lit
for a subset of 36 watersheds from the 90 Texas watersheds using standard C
lit
tables
published by TxDOT (2002). The average absolute difference between Harle?s estimate
of C
lit
and that presented in this study was 0.06 and the maximum absolute difference was
0.13.
The NLCD 1992 data were used to examine the potential for temporal differences
in composite C
lit
estimates. When NLCD 1992 data were used, 18 out of 21 LULC
classes for the 90 watersheds were used. This difference (from the 15LULC classes
determined using NLCD 2001) occurred because there were more land use (LU) codes
for some landcover classes in NLCD 1992. Summary statistics of the composite runoff
coefficients, C
lit
,
obtained using NLCD 1992 and 2001 data are listed in Table 2.2. The
average values of the runoff coefficients for 90 watersheds derived from two LULC
21
datasets are the same (0.47, see Table 2.2). The median absolute difference of C
lit
derived
from the two LULC datasets is 0.03 with a minimum difference of 0.00 and a maximum
difference of 0.14. The differences between C
lit
obtained using two LULC datasets are
plotted in Figure 2.2 (top). The authors conclude that there is no substantial difference
between C
lit
values derived from the 1992 and 2001 LULC datasets because the paired t
test gives pvalue (Ayyub and McCuen 2003) of 0.88, much larger than the level of
significance 0.05 for the level of confidence 95%.
A sensitivity analysis was conducted to examine effect of selected C values for
different LULC classes from literature on watershed composite runoff coefficient C
lit
.
The minimum and maximum C values for each LULC class from literature (Table 2.1)
were used to derive watershedminimum and watershedmaximum C
lit
values for each
watershed, respectively. The NLCD 2001 data was used for the sensitivity analysis. The
cumulative distributions of watershedminimum, average, and maximum C
lit
values
obtained using NLCD 2001 are shown in Figure 2.2 (top) and summary statistics of these
C
lit
values are listed in Table 2.2. Values of watershedminimum C
lit
ranged from 0.13 to
0.60, with median and mean values of 0.41 and 0.38, respectively (Table 2.2). Values of
watershedmaximum C
lit
ranged from 0.38 to 0.68, with median and mean values of 0.58
and 0.55, respectively (Table 2.2). The differences of watershedmaximum and
watershedminimum C
lit
values for 90 Texas watersheds ranged from 0.04 to 0.34, with
median and mean differences of 0.14 and 0.17, respectively. ASCE and WPCF (1960)
and design manuals (e.g., TxDOT 2002) and textbooks (e.g., Viessman and Lewis 2003)
give a range of C values (not a single value) for different land use types, and the range of
published C values for the same land use is from 0.04 to 0.3, same variations of C
lit
for 90
22
Texas watersheds. This indicates uncertainty and variation of peak discharge estimation
using the rational method.
The amount of developed land in a watershed is a key factor governing the runoff
from the watershed. To study the relation of the composite runoff coefficients to the
development factor of the watersheds, statistical summaries of C
lit
(Table 2.3) from
NLCD 2001 data were obtained separately for the 90 watersheds, which were classified
as developed or undeveloped (Roussel et al. 2005). The corresponding cumulative
frequency distributions are shown in Figure 2.2 (bottom). The median value of C
lit
(watershedaverage) for undeveloped watersheds is 0.37 and for the developed
watersheds is 0.54. The average values of C
lit
for undeveloped and developed watersheds
are 0.39 and 0.54, respectively (Table 2.3). C
lit
values of the developed watersheds are
distinctly greater than those from undeveloped watersheds (pvalue < 0.0001 from the
pooled ttest), as shown in Figure 2.2 (bottom); the combination of LULC data and
published C
lit
values provides representative estimates of C
lit
to reflect landuse
development in a watershed.
2.5 Estimation of the BackComputed Volumetric Runoff Coefficients (C
vbc
) Using
Observed RainfallRunoff Data
The concept of a rainfallrunoff event volumetric runoff coefficient (C
v
) in
hydrology dates to the beginning of the 20
th
century. An example is Sherman (1932), who
used the percent of rainfall when he introduced the unithydrograph method. C
v
is defined
as the portion of rainfall that becomes runoff during an event (Merz et al. 2006).
Estimates of C
v
from an individual event are usually determined by three steps: (1)
23
separation into single events, (2) separation of observed streamflow into base flow and
direct runoff, and (3) estimation of event C
v
as the ratio of direct flow or runoff volume to
event rainfall volume (Merz et al. 2006). C
v
is based on the integrated response of the
watershed, that is, the transformation of rainfall volume to runoff volume.
French et al. (1974) evaluated C
v
for several rural catchments in New South
Wales; Calomino et al. (1997) computed C
v
for 66 events for a urban watershed (91.5%
impervious area and total drainage area of 0.019 km
2
(1.89 ha)). For the urbanized
watershed studied by Calomino et al. (1997), event C
v
ranged from 0.31 to 0.88, and C
v
was strongly correlated to the total rainfall depth (P): C
v
= 0.57 P
0.042
(R
2
= 0.96)
(Calomino et al. 1997). The Water Planning Division of the United States Environmental
Protection Agency (USEPA) operated the National Urban Runoff Program (NURP). This
program had 20 projects throughout the United States to study pollutants from 76 urban
watersheds, with drainage area ranging from 0.004 to 115 km
2
(USEPA 1983). NURP
researchers collected rainfall and runoff data from these watersheds, with the number of
events ranging from 5 to 121. A runoff coefficient, R
v
(USEPA 1983), defined as the ratio
of runoff volume to rainfall volume, was determined for each of the NURPmonitored
storm events. The median value of the runoff coefficients, the coefficient of variation,
and the percent impervious area were reported for all watersheds used in the study
(USEPA 1983).
In this study, estimates of the volumetric runoff coefficient are called the back
computed volumetric runoff coefficient (C
vbc
)
,
and an individualevent C
vbc
was obtained
for kth storm event by the ratio of the total runoff depth, R
k
(mm or in.), to the total
rainfall depth, P
k
(mm or in.), by:
24
k
kk
vbc
Peventtheforrainalltotal
Rrunoffeventtotal
C
,
,
? (2.3)
The study database comprised 1,600 rainfallrunoff events with observed rainfall and
runoff data collected from 90 watersheds in Texas (Fang et al. 2007). Therefore, 1,600
event runoff coefficients C
vbc
were obtained using equation (2.3). Event C
vbc
ranged from
near 0.0 to 1.0, covering the range of possible values. The cumulative distributions of
C
vbc
are presented in Figure 2.3 and summary statistics are listed in Table 2.4. For the 90
study watersheds in Texas, no substantial relation between rainfall depth and C
vbc
was
detected (Pearson?s correlation coefficient r = 0.2 at the 0.1 percent level of significance
because of pvalue less than 0.0001). For example, for 19 events with total rainfall depth
less than 12.7 mm (0.5 in.), computed C
vbc
ranged from 0.050 to 0.844. For 253 events
with total rainfall depth between 76.2 mm (3 in.) and 101.6 mm (4 in.), the computed C
vbc
ranged from 0.006 to 0.982. Based on review of Figure 2.3, C
vbc
is less than 0.1 for 13
percent of events. Furthermore, C
vbc
exceeds 0.9 for one percent of events. The regression
relation between the runoff coefficient C
vbc
and the total runoff depth (R) was C
vbc
=
0.374 R
0.699
. The regression explained about 76 percent of the variance between runoff
depth and runoff coefficient and the regression coefficients were statistically significant
at the 0.1 percent level of significance (pvalue less than 0.0001).
C
vbc
values calculated for all events in the same watershed varied from one event to
another, e.g., depending on antecedent moisture condition before a rainfall event.
Statistical parameters of the range of C
vbc
values as the difference between maximum and
minimum C
vbc
values calculated for all events in the same watershed are given in Table
2.4. The maximum and average values of the range of event C
vbc
in the same watershed
is 0.97 and 0.52 (Table 2.4) for 1,600 rainfallrunoff events in 90 Texas watersheds,
25
respectively. This finding is supported by previous studies by French et al. (1974) and
USEPA (1983). Variations of event C
vbc
in the same watershed determined from
observed rainfall and runoff data are much larger than ranges of published C values for
the same land use type.
Watershedaverage and median values of C
vbc
were calculated from C
vbc
values for
all rainfallrunoff events observed in the same watershed and developed for 83 of the 90
watershed dataset in Texas. Of the 90 watersheds in Texas, 7 were excluded; less than 4
rainfallrunoff events were available for analysis in each of these 7 watersheds.
Computed watershedaverage C
vbc
ranged from about 0.1 to 0.67 and from about 0.06 to
0.76 for the watershedmedian C
vbc
(Table 2.4). These values are similar to values of C
lit
estimated from LULC data, which ranged from 0.29 to 0.68 (Figure 2.2 and Table 2.2).
About 80% of the C
vbc
median and average values were less than 0.5. Watershedmedian
R
v
ranged from 0.02 to 0.93 for 76 watersheds studied in the National Urban Runoff
Program (USEPA 1983). The average values of the watershedaverage and the
watershedmedian C
vbc
are approximately the same, 0.33 and 0.31, respectively (Table
2.4). As shown in Figure 2.3, the cumulative frequency distributions of the watershed
average and watershedmedian of C
vbc
values are similar and the maximum absolute
difference of watershedaverage and median C
vbc
is less than 0.10.
The developed and undeveloped watershed classifications (Roussel et al. 2005)
were used to sort the watershedaverage C
vbc
values for additional statistical analysis. The
results are listed in Table 2.5 and cumulative distributions of C
vbc
are shown in Figure
2.3. The cumulative distributions are distinctly different: developed watersheds have
greater C
vbc
(watershedaverage) in comparison to undeveloped watersheds (pvalue <
26
0.0001 from the pooled ttest). The median values of the watershedaverage C
vbc
for
undeveloped and developed watersheds are 0.19 and 0.37, respectively (Table 2.5).
For this study, the percentage of impervious area (IMP) was computed using 1992
NLCD. Of the 90 study watersheds, 45 have percent impervious area greater than 15
percent. The watershedmedian runoff coefficients C
vbc
and R
v
versus percent impervious
area for the 45 developed watersheds in Texas and the 60 watersheds from NURP are
shown in Figure 2.4 (top). For 76 watersheds among those studied in NURP (USEPA
1983), two separate graphs of watershedmedian runoff coefficient versus percent
impervious area were developed and reported by USEPA (1983): one graph is for the 60
watersheds and another is for 16 watersheds.
?The separate grouping is based on the fact that the relationship for these sites (16
watersheds) is internally consistent and significantly different (much lower) than the bulk
of the project results? (USEPA 1983 p. 660).
Polynomial regression lines were fit to the 60 NURP watershed data and to the
combined watershed data for the combined group of 60 NURP and 45 Texas watersheds
(watershedmedian C
vbc
). The regression lines are displayed in Figure 2.4 (top).
Coefficients of determination R
2
for the two datasets are 0.79 and 0.57, respectively.
The regression equation obtained from the combined 60 NURP watershed data and
the 45 Texas watershed data (watershedmedian C
vbc
) is
,036.0289.1275.2843.1
23
???? IMPIMPIMPC
v
(2.4)
where, C
v
= volumetric runoff coefficient and IMP = percent impervious area expressed
as a fraction (50% = 0.5) of the watershed area. Urbanization alters the land surface and
increases IMP. Although other watershed parameters, e.g., basin development factor
27
(Sauer et al 1983), can be used to quantify the degree of urbanization, IMP was used in
this study to correlate it to C
v
because Kuichling (1889) concluded that runoff coefficient
for any given watershed he studied is nearly equal to the percentage of impervious
surface within the watershed.
For comparison, Urbonas et al. (1989) used watershedmedian runoff coefficients
from the group of 60 NURP watersheds and several runoff coefficients developed for
watersheds in the Denver area to develop a polynomial regression equation between
runoff coefficient and percent impervious area. The Urbonas et al. equation (not repeated
here) currently (2010) is used by the Denver Urban Drainage and Flood Control District
in its Drainage Criteria Manual (from http://www.udfcd.org/downloads/down_
critmanual.htm, accessed on January 10, 2010) to determine C for hydrologic soil group
(HSG) Types C and D. The curve for the Urbonas et al. equation also is shown in Figure
2.4 (top). Although the regression parameters differ between the three regressions shown,
the three regression curves are similar and have a maximum absolute difference of C
v
less
than 0.1.
Values of watershedmedian C
vbc
for the 45 Texas watersheds are generally
consistent with those from the 60 NURP watershed data (Figure 2.4). Standard deviations
from the watershedaverage C
vbc
were calculated for Texas watersheds and shown in
Figure 2.4 (bottom) as solid circles with thick error bars. Standard deviations and
coefficients of variation from watershedaverages C
vbc
ranged from 0.04 to 0.30 (Table
2.4) and 0.15 to 1.34 for 83 Texas watersheds, respectively.
For the NURP data, watershedmedian values and the coefficients of variation
were reported (USEPA 1983), but the watershedaverage runoff coefficients (R
v
) were
28
not reported. In order to examine the variation of runoff coefficient for NURP data,
reported watershedmedian values were used as watershedaverages to estimate the
standard deviations from reported coefficient of variations, and the statistical distribution
parameters of estimated standard deviations for NURP watersheds is listed in Table 2.4.
Estimated maximum standard deviation from NURP watershed data is greater than 1.0,
which is impossible if R
v
ranged from 0.0 to 1.0, possibly because watershedmedian R
v
were used. That NURP watershed has watershedmedian R
v
of 0.17 and the coefficient of
variation of 6.64, which is much larger than 1.34, the maximum coefficient of variation
for 83 Texas watersheds. There are 15 NURP watersheds having estimated standard
deviations greater than 0.3, the maximum standard deviation for 83 Texas watersheds
(Table 2.4). The median standard deviations are approximately equal for both datasets
(Table 2.4). The NURP data (median R
v
for the 60 watersheds) plus and minus estimated
one standard deviation are shown in Figure 2.4 (bottom) as open squares with wide error
bars. Standard deviations from watershedaverage C
vbc
for the 45 Texas watersheds are
consistently less than those from the 60 NURP watersheds (Figure 2.4). The NURP data
cover a greater range of percent impervious area for watersheds (Figure 2.4) that are
useful to develop the regression equation (2.4), which make the regression applicable to a
wider range of watersheds.
2.6 Estimation of Volumetric Runoff Coefficients from the RankOrdered Pairs of
Observed Rainfall and Runoff Depths
29
Schaake et al. (1967) examined the rational method using observed rainfall and
runoff data collected from 20 gaged urban watersheds in Baltimore, MD. The size of
watershed drainage area used by Schaake et al. was 0.6 km
2
(150 acres) or smaller.
Schaake et al. (1967) used a frequencymatching approach to prepare their data for
analysis. The frequencymatching approach was independently sorting observed rainfall
intensity (average intensity over the watershed lag time) and peak runoff rate before
computing the runoff coefficient using the rational method. That is, the rainfall intensity
and peak runoff rate are paired on the rank order and not the event order.
Schaake et al. (1967) concluded that the frequency of occurrence of the computed
design peak runoff rate is the same as the frequency of the rainfall intensity selected by
the designer. Schaake et al. (1967) developed a regression equation to relate ratebased C
(determine from peak discharge and rainfall intensity) to the imperviousness of the
watershed and the main channel slope. Hjelmfelt (1980) and Hawkins (1993) used a
similar frequency matching procedure as Schaake et al. (1967), except they used rank
ordered rainfall and runoff depths for computing actual curve numbers from historical
rainfallrunoff events.
For each of the 90 Texas watersheds in this study, the total rainfall depth and the
total runoff depth were ranked independently from greatest to least. As an example, the
rankordered pairs of total rainfall depth (mm) and total runoff (mm) for 13 events at
USGS gage station 08042650 (North Trinity Basin, TX) are presented in Figure 2.5 (top
panel). The volumetric runoff coefficient C
v
was computed from the rankordered pairs of
total runoff and rainfall depths using:
,
j
j
vrj
P
R
C ?
(2.5)
30
where, C
vrj
is the C
v
corresponding to the total runoff depth R
j
and the total rainfall depth
P
j
of the j
th
order of rainfallrunoff pairs (subscript ?r? for C
vrj
stands for rankordered).
A plot of runoff coefficient (C
vrj
) versus the total rainfall depth was prepared for
each watershed. For example, the plot for USGS gage station 08042650 is presented in
Figure 2.5 (bottom panel). For most of the study watersheds, C
vrj
increases until acquiring
an approximate constant value. This constant value was considered as representative C
vr
for the watershed; for example, watershed representative runoff coefficient C
vr
= 0.17 for
North Trinity Basin watershed associated with USGS gage station 08042650 (Figure 2.5).
In addition to applying Hawkins?s procedure (Hawkins 1993), i.e., asymptotic
determination of C
vr
from C
vrj
versus rainfall depth, watershed C
vr
can be estimated from
the slope of the regression line obtained from the plots of the rankordered total runoff
depth versus the rankordered total rainfall depth as shown in the top panel of Figure 2.5.
For example, for USGS station 08042650, the regression equation developed from rank
ordered runoff and rainfall data is ?Totalrunoff (mm) = 0.167 ? Totalrainfalldepth
(mm)?. Therefore, the watershed representative C
vr
is 0.167 and equal to the slope of the
regression equation. For most of the study watersheds, C
vr
obtained from both
procedures have approximately the same value. Statistical distribution parameters of C
vr
for 83 Texas watersheds (7 out of 90 watersheds were excluded because of the too few
rainfallevent data available) are listed in Table 2.6. The mean and median values of C
vr
are the same and equal to 0.40. The values of C
vr
range from 0.10 to 0.78. The
cumulative distributions of the runoff coefficients C
vr
and C
lit
are shown in Figure 2.6
(top). The median value of the absolute differences C
vr
C
lit
 is 0.14 (Table 2.6).
31
Examining the cumulative frequency distributions of C
lit
and C
vr
for the 90 study
watersheds (Figure 2.6 top), about 80 percent of the C
lit
values exceed the C
vr
values.
Watershedrepresentative runoff coefficients C
vr
were grouped into two categories:
those from developed and those from undeveloped watersheds (Roussel et al. 2005). The
statistical summary of C
vr
for the two groups is listed in Table 2.7 and the corresponding
cumulative frequency distributions are presented in Figure 2.6 (bottom). The median
runoff coefficient C
vr
from undeveloped watersheds is 0.24, and the median value from
developed watersheds is 0.48. Based on this observation, C
vr
derived from rankordered
rainfallrunoff data reflects effects of watershed development, specifically the increase of
percent impervious area (Figure 2.7). A statistical summary of the absolute differences
C
vr
 C
vbc

and C
vr
 C
lit
 for the 45 developed watersheds in Texas is listed in Table 2.7.
Small average and median values of absolute differences C
vr
 C
vbc
 indicate that C
vr
is
similar to C
vbc
because both were derived from observed rainfallrunoff data. Average
and median values of absolute differences C
vr
 C
lit
 are greater than those of C
vr
 C
vbc

(Table 2.7), and indicate that C
vr
derived from rainfallrunoff data differs from C
lit
derived from landuse data and published runoff coefficients (see Figures 2.2 and 2.6).
The watershed representative C
vr
and watershedmedian C
vbc
from 45 developed
Texas watersheds and runoff coefficient R
v
from 60 NURP watersheds were plotted
against percent impervious area (Figure 2.7). Results from these three datasets are
consistent ? overall increasing volumetric runoff coefficient with the increase of percent
impervious area or degree of development. A polynomial regression line was fitted to
combined data from 60 NURP watersheds and watershed representative C
vr
for the 45
32
Texas developed watersheds (Figure 2.7). R
2
for the regression equation (2.6) = 0.57,
which is the same as R
2
for equation (2.4).
043.0315.1940.1469.1
23
???? IMPIMPIMPC
v
(2.6)
The regression equations (2.4) and (2.6) were combined by averaging their coefficients to
get a single equation for general application in Texas watersheds similar to 45 developed
Texas watersheds.
04.030.111.266.1
23
???? IMPIMPIMPC
v
(2.7)
Equation (2.7) can be used to estimate C
v
for developed (urban) watersheds based on
impervious cover. Equation (2.7) is plotted on Figure 2.7, which also includes a curve
for Equation (2.6) and data points of C
vbc
, C
vr
, and R
v
values versus percent impervious
area (IMP). Figure 2.7 and Equation (2.7) indicate that C
v
is not equal to 1.0 when IMP =
100%. This is because R
v
estimated in the NURP study is for watersheds greater than
0.004 km
2
(1 acre) and C
v
estimated in this study for watersheds greater than 0.8 km
2
(200 acres); therefore, Equation (2.7) does not apply to very small 100% impervious
catchment such as a small parking lot. Further study is needed to correlate runoff
coefficients for undeveloped watersheds (Figure 2.6) to soil types and other watersheds
characteristics.
2.7 Discussion
Volumetric runoff coefficients, watershedaverage C
vbc
and C
vr
for 83 Texas
watersheds, watershedaverage C
lit
for 90 Texas watersheds, and watershedmedian R
v
for NURP 60 watersheds (USEPA 1983), were plotted against drainage area A in km
2
(Figure 2.8). Pearson?s correlation coefficients between C
vbc
, C
vr
, C
lit
, R
v
and A (km
2
) are
0.20, 0.12, 0.27, and 0.26 with pvalues of 0.060, 0.256, 0.009, and 0.044,
33
respectively. Therefore, at the 90% confidence level, C
vbc
, C
lit
, and R
v
has no substantial
relation with area (Figure 2.8). C
vr
has no substantial relation with area only at the 70%
confidence level. Above statistical analyses between volumetric C values and drainage
area indicated that there is no demonstrable relation in volumetric C with drainage area as
Young et al. (2009) reported. This finding supports the conclusion by ASCE and WPCF
(1960), Pilgrim and Cordery (1993), and Young et al. (2009) that the rational method
may be applied to much larger drainage areas than typically assumed in some design
manuals, as long as the watershed is unregulated (Young et al. 2009). The authors do not
advocate specifying specific limit that should be imposed on drainage area for application
of the rational method. It is the duty and responsibility of the enduser of the rational
method to apply appropriate engineering judgment and experience in developing designs.
The authors explicitly are not advocating application of the rational method for
larger watersheds because the steadystate assumption of the rational method for design
purposes is questionable. However, extensive data analysis (Asquith 2010) suggests that
inherent relations between runoff coefficient and drainage area are insubstantial if time of
concentration of a watershed is reasonably estimated for determining rainfall intensity.
The authors support the recommendation of ASCE and WPCF in 1960 ?Development of
data for application of hydrograph methods is usually warranted on larger areas? (ASCE
and WPCF 1960, p. 32).
The authors explicitly recognize that volumetric runoff coefficients might not
have direct applicability in use of the rational method for engineering design purposes.
Therefore, the authors did not apply the rational method and use volumetric runoff
coefficients (C
vbc
and C
vr
) to predict peak discharges for 1,600 events and compare
34
predicted and observed peak discharges. To predict peak flows using volumetric runoff
coefficients is inconsistent with the assertion that ratebased values for the runoff
coefficient be used. In a subsequent paper, the authors determined ratebased rational
runoff coefficients for these 90 Texas watersheds, and applied the rational method with
ratebased runoff coefficients to predict peak discharges and compared predicted and
observed peak discharges.
Volumetric runoff coefficients estimated from observed rainfall and runoff data
for 83 Texas watersheds were plotted against literaturebased C
lit
(watershedaverage)
determined for the same watersheds (Figure 2.9). Regression equations between
watershedaverage C
vbc
, C
vr
and C
lit
were developed and shown in Figure 2.9, with
Pearson?s correlation coefficients r = 0.36 and 0.26 at the 95% confidence level (pvalues
of 0.0007 and 0.01), respectively. Therefore, regression analyses indicated volumetric
runoff coefficients determined from rainfall and runoff data are weakly correlated to
literaturebased C
lit
.
2.8 Summary
Volumetric runoff coefficients were estimated for 90 Texas watersheds using
three different methods. The first method is estimation of literaturebased runoff
coefficients (C
lit
) using published values and GIS analysis of LULC classes to construct
areally weighted values over a watershed. C
lit
was obtained independently from 1992 and
2001 NLCD and using minimum, average, and maximum published C values for
different LULC. No substantial difference in the results of watershedaverage C
lit
was
obtained using the 1992 or 2001 versions of the landuse data. For the study watersheds,
35
watershedaverage C
lit
ranged from 0.29 to 0.68 with median and average values about
0.5. The differences of watershedmaximum and watershedminimum C
lit
values for 90
Texas watersheds ranged from 0.04 to 0.34, with median and mean differences of 0.14
and 0.17, respectively. When C
lit
(watershedaverage) is grouped into developed and
undeveloped watersheds, the range of C
lit
for developed watersheds was between 0.37
and 0.63, with a median value of 0.54. The median value of C
lit
for developed watersheds
exceeds that for undeveloped watersheds. This result stems from the fact that published
runoff coefficients, even though they were not developed from observed rainfallrunoff
measurements and instead resulted from a survey on engineering practices in 1950s,
reflect the physical meanings of the original runoff coefficients introduced by Kuichling
in 1889 ? the runoff coefficient is related to the percent impervious area within the
watershed. Therefore, published runoff coefficients remain useful for engineering design
of drainage systems.
The second method is based on use of backcomputed volumetric runoff
coefficients (C
vbc
) from observed rainfallrunoff measurements of more than 1,600 events
by the ratio of total runoff depth to total rainfall depth for individual storm event. Event
volumetric runoff coefficients cover all possible values from 0.0 to 1.0 with 10% of all
values less than 0.08 and 10% of all values greater than 0.63 (Figure 2.3). The maximum
and average values of the range of event C
vbc
in the same watershed is 0.97 and 0.52
(Table 2.4) for 1,600 rainfallrunoff events in 90 Texas watersheds, respectively.
Watershedaverage and watershedmedian values of C
vbc
and estimates of standard
deviations were extracted. The distributions of the watershedaverage and watershed
median C
vbc
are similar. Watershedmedian values of C
vbc
ranged from 0.06 to 0.76 with
36
an average of 0.31. Watershedmedian values of C
vbc
for 45 developed watersheds in
Texas with percent imperviousness greater than 15% are consistent with median values of
runoff coefficient R
v
reported for 60 NURP watersheds by the USEPA.
The third method involved the computation of runoff coefficients by the
frequency matching procedures of observed total rainfallrunoff depths from a watershed.
A single watershedspecific value of the runoff coefficient (C
vr
) was developed from the
plot of rankordered runoff coefficients versus rainfall depths. The values of C
vr
ranged
from 0.10 to 0.78 with the median value 0.40. The C
vr
values for the developed
watersheds are consistently higher than those for the undeveloped watersheds. The
distribution of C
vr
is different from that of C
lit
with about 80 percent of C
lit
value greater
than C
vr
value. This result might indicate that literaturebased runoff coefficients
overestimate peak discharge for drainage design when used with the rational method.
Runoff coefficients derived from observed rainfall and runoff data in 90 Texas
watersheds in this study are volumetric based (ratio of total runoff and rainfall depth) and
are useful in transforming rainfall depth to runoff depth such as is done in the curve
number method (SCS 1963) and for watershed rainfallrunoff modeling, e.g., the
fractional loss model (McCuen 1998, p. 493). Current runoff coefficients given in
textbooks and design manuals are neither volumetric nor ratebased (determined from
peak discharge and rainfall intensity) because they were not derived from observed data
but are used for the ratebased rational method.
Two regression equations (2.4) and (2.6) were developed using watershedmedian
C
vbc
and watershedrepresentative C
vr
data combined with median runoff coefficients R
v
from 60 NURP watersheds. Coefficient of determination R
2
for both equations are 0.57
37
and these equations were combined into a single equation (2.7) which can be used to
estimate volumetric runoff coefficients for developed urban watersheds that are similar to
the 45 developed watersheds in Texas. The published limits on drainage area for
application of the rational method seem to be arbitrary. Results from this study supports
the conclusion by ASCE and WPCF (1960), Pilgrim and Cordery (1993), and Young et
al. (2009) that the rational method may be applied to much larger drainage areas than
typically assumed in some design manuals.
2.9 Acknowledgments
The authors thank TxDOT project director Mr. Chuck Stead, P.E., and project
monitoring advisor members for their guidance and assistance. They also express their
thanks to technical reviewers Meghan Roussel and Glenn Harwell from Texas Water
Science Center in Austin, and Fort Worth, respectively, and to three anonymous
reviewers; the comments and suggestions greatly improved the paper. This study was
partially supported by TxDOT Research Projects 0?6070, 0?4696, 0?4193, and 0?4194.
2.10 Notation
The following symbols are used in this paper:
A = watershed drainage area in hectares or acres;
A
i
= sub area for ith land cover classes in the watershed;
C = runoff coefficient;
C
i
= literaturebased runoff coefficient for ith landcover class (Table 2.1);
38
C
v
= volumetric runoff coefficient, portion of rainfall that becomes runoff, determined
from regression equations;
C
vbc
= watershed average or median backcomputed volumetric runoff coefficient;
C
k
vbc
= backcomputed volumetric runoff coefficient for the kth event;
C
lit
= literature?based runoff coefficient developed from landuse data;
C
vrj
= runoff coefficient estimated from the ratio of jth rankordered runoff and rainfall
data pair;
C
vr
= watershed representative runoff coefficient estimated from the distribution of ratios
of rankordered runoff and rainfall;
I = average rainfall intensity (mm/hr or in./hr) with the duration equal to time of
concentration;
IMP = percent of impervious area expressed as a fraction (50% = 0.5) for a watershed
area;
m
o
= the dimensional correction factor (1.008 in English units, 1/360 = 0.00278 in SI
units);
no = total number of land cover classes in a watershed (Equation 2.2);
P
j
= total rainfall depth of the j
th
order of rank runoff data series;
P
k
= total rainfall depth of the k
th
event;
Q
p
= peak discharge or runoff rate in m
3
/s or ft
3
/s;
R
j
= total runoff depth of the j
th
order of rank rainfall data series;
R
k
= total runoff depth of the k
th
event;
R
v
= runoff coefficient as the ratio of runoff volume to rainfall volume determined by
USEPA for the NURP data.
39
2.11 References
ASCE, and WPCF. (1960). Design and construction of sanitary and storm sewers,
American Society of Civil Engineers (ASCE) and Water Pollution Control Federation
(WPCF).
ASCE. (1992). "Hydrology and introduction to water quality." Design and construction
of urban stormwater management systems, American Society of Civil Engineers,
New York, 6397.
Asquith, W. H., Thompson, D. B., Cleveland, T. G., and Fang, X. (2004). "Synthesis of
rainfall and runoff data used for Texas department of transportation research projects
04193 and 04194." U.S. Geological Survey, OpenFile Report 20041035, Austin,
Texas.
Asquith, W. H., Thompson, D. B., Cleveland, T. G., and Fang, X. (2006). "Unit
hydrograph estimation for applicable Texas watersheds." Texas Tech University,
Center for Multidisciplinary Research in Transportation, Austin, Texas.
Asquith, W. H., and Roussel, M. C. (2007). "An initialabstraction, constantloss model
for unit hydrograph modeling for applicable watersheds in Texas." U.S. Geological
Survey Scientific Investigations Report 20075243, 82 p. [
http://pubs.usgs.gov/sir/2007/5243 ].
Ayyub, B. M., and McCuen, R. H. (2003). Probability, statistics, and reliability for
engineers and scientists, 2nd Ed., Chapman & Hall/CRC, Boca Raton, 640 p.
Calomino, F., Veltri, P., Prio, P., and Niemczynowicz, J. (1997). "Probabilistic analysis
of runoff simulations in a small urban catchment." Water Science and Technology,
36(89), 5156.
Chow, V. T., Maidment, D. R., and Mays, L. W. (1988). Applied hydrology, McGraw
Hill, New York.
Cleveland, T. G., He, X., Asquith, W. H., Fang, X., and Thompson, D. B. (2006).
"Instantaneous unit hydrograph evaluation for rainfallrunoff modeling of small
watersheds in north and south central Texas." Journal of Irrigation and Drainage
Engineering, 132(5), 479485.
Cleveland, T. G., Thompson, D. B., Fang, X., and He, X. (2008). "Synthesis of unit
hydrographs from a digital elevation model." Journal of Irrigation and Drainage
Engineering, 134(2), 212221.
40
Dunne, T., and Leopold, L. B. (1978). Water in Environmental Planning, San Francisco,
California: W. H. Freeman and Company.
ESRI. (2004). "Using ArcGIS spatial analyst." Environmental Systems Research
Institute, Inc. (ESRI), Redlands, California.
Fang, X., Thompson, D. B., Cleveland, T. G., and Pradhan, P. (2007). "Variations of time
of concentration estimates using NRCS velocity method." Journal of Irrigation and
Drainage Engineering, 133(4), 314322.
Fang, X., Thompson, D. B., Cleveland, T. G., Pradhan, P., and Malla, R. (2008). "Time
of concentration estimated using watershed parameters determined by automated and
manual methods." Journal of Irrigation and Drainage Engineering, 134(2), 202211.
French, R., Pilgrim, D. H., and Laurenson, E. M. (1974). "Experimental examination of
the rational method for small rural catchments." Civil Engineering Transactions, The
Institution of Engineers, Australia, CE16(2), 95102.
Harle, H. K. (2002). "Identification of appropriate size limitations for hydrologic
modeling for the state of Texas," Master Thesis, Department of Civil and
Environmental Engineering, Texas Tech University, Lubbock, Texas.
Hawkins, R. H. (1993). "Asymptotic determination of runoff curve numbers from data."
Journal of Irrigation and Drainage Engineering, 119(2), 334345.
Hjelmfelt, A. T. (1980). "Empirical investigation of curve number technique." Journal of
the Hydraulics Division, 106(9), 14711476.
Institute of Hydrology (1976). "Water balance of the headwater catchments of the Wye
and Severen 19701975." Report No. 33, Institute of Hydrology, UK.
Kuichling, E. (1889). "The relation between the rainfall and the discharge of sewers in
populous areas." Transactions, American Society of Civil Engineers 20, 1?56.
Law, F. (1956). "The effect of afforestation upon the yield of water catchment areas."
Technical report, British Association for Advancement of Science, Sheffield,
England.
LloydDavies, D. E. (1906). "The elimination of storm water from sewerage systems."
Minutes of Proceedings, Institution of Civil Engineers, Great Britain, 164, 41.
Merz, R., Bl?schl, G., and Parajka, J. (2006). "Spatiotemporal variability of event runoff
coefficients." Journal of Hydrology, 331(34), 591604.
41
Mulholland, P. J., Wilson, G. V., and Jardine, P. M. (1990). "Hydrogeochemical response
of a forested watershed to storms: Effects of preferential flow along shallow and deep
pathways." Water Resources Research 26(12), 30213036.
Roussel, M. C., Thompson, D. B., Fang, D. X., Cleveland, T. G., and Garcia, A. C.
(2005). "Timing parameter estimation for applicable Texas watersheds." 046962,
Texas Department of Transportation, Austin, Texas.
Sauer, V. B., Thomas, W. O., Stricker, V. A., and Wilson, K. V. (1983). "Flood
characteristics of urban watersheds in the United States." U.S. Geological Survey
WaterSupply Paper 2207, U.S. Government Printing Office, Washington D.C.
Schaake, J. C., Geyer, J. C., and Knapp, J. W. (1967). "Experimental examination of the
rational method." Journal of the Hydraulics Division, ASCE, 93(6), 353?370.
Schwab, G. O., and Frevert, R. K. (1993). Elementary Soil and Water Engineering,
Krieger Publishing Company.
SCS. (1963). "Section 4, Hydrology." National Engineering Handbook, The Soil
Conservation Service (SCS), U.S. Dept. of Agriculture Washington, D.C.
Sherman, L. K. (1932). "Streamflow from rainfall by the unitgraph method." Eng.
NewsRec., 108, 501505.
TxDOT. (2002). Hydraulic design manual, The bridge division of the Texas Department
of Transportation (TxDOT), 125 East 11th street, Austin, Texas 787012483.
Urbonas, B., Guo, C. Y. J., and Tucker, L. S. (1989). "Optimization of stormwater quality
capture volume." Proceeding of an Engineering Foundation Conference "Urban
Stormwater Quality Enhancement  Source Control, Retrofitting, and Combined
Sewer Technology", Devos Platz, Switzerland, ASCE.
USEPA. (1983). "Results of the nationwide urban runoff program, Volume I  Final
Report." National Technical Information Service (NTIS) Accession Number: PB84
185552, Water Planning Division, United States Environmental Protection Agency
(USEPA), Washington D.C.
Viessman, W., and Lewis, G. L. (2003). Introduction to hydrology, 5th Ed., Pearson
Education, Upper Saddle River, N.J., 612.
Wanielista, M. P., and Yousef, Y. A. (1993). Stormwater Management, John Wiley &
Sons, Inc., New York.
Young, C. B., McEnroe, B. M., and Rome, A. C. (2009). "Empirical determination of
rational method runoff coefficients." Journal of Hydrologic Engineering, 14, 1283.
42
Table 2.1 Runoff Coefficients (C) Selected for Various LandCover Classes from NLCD
2001.
NLCD
classification
NLCD classification description C Land use or description in the source
21
Developed, Open Space
0.4
(1),(2)
Residential: single family areas (0.30.5)
22 Developed , Low Intensity 0.55
(3)
50 % of area impervious (0.55)
23 Developed, Medium Intensity 0.65
(3)
70% of area impervious (0.65)
24 Developed , High Intensity 0.83
(2)
Business: downtown areas (0.70.95)
31 Barren Land 0.3
(2),(7)
Sand or sandy loam soil, 05% (0.150.25);
black or loessial soil, 05% (0.180.3);
heavy clay soils; shallow soils over
bedrock: pasture (0.45)
41 Deciduous Forest 0.52
(4)
Deciduous forest (Tennessee) (0.52)
42 Evergreen Forest 0.48
(5),(6)
Forest (UK) (0.280.68); Forest (Germany)
(0.330.59)
43 Mixed Forest 0.48
(5),(6)
Forest (UK) (0.280.68); Forest (Germany)
(0.330.59)
52 Shrub/Scrub 0.3
(7)
Woodland, sandy and gravel soils (0.1);
loam soils (0.3); heavy clay soils (0.4);
shallow soil on rock (0.4)
71 Grassland/Herbaceous 0.22
(3)
Pasture, grazing HSG A (0.1); HSG B
(0.2); HSG C (0.25); HSG D (0.3)
81 Pasture/Hay 0.35
(7)
Pasture, sandy and gravel soils (0.15);
loam soils (0.35); heavy clay soils (0.45);
shallow soil on rock (0.45)
82 Cultivated Crops 0.4
(7)
Cultivated, sandy and gravel soils (0.2);
loam soils (0.4); heavy clay soils (0.5);
shallow soil on rock (0.5)
Sources: 
(1)
ASCE (1992),
(2)
TxDOT (2002),
(3)
Schwab and Frevert (1993),
(4)
Mulholland et al. (1990) ,
(5)
Law (1956),
(6)
Hydrology (1976),
(7)
Dunne and Leopold (1978)
Note Numbers in parenthesis are ranges for runoff coefficients given in the source (literature).
HSG = hydrologic soil group
43
Table 2.2 Statistical Summary of Average C
lit
Using NLCD 1992 and Watershed
Minimum, Average and Maximum C
lit
Using NLCD 2001.
Statistical
distribution
parameters
Watershedaverage
1
C
lit
using NLCD
1992
(1)
C
lit
using NLCD 2001
Absolute
difference
(1) ? (3)
Watershed
minimum
1
(2)
Watershed
average
(3)
Watershed
maximum
1
(4)
Minimum 0.32 0.13
0.29 0.38
0.00
Maximum 0.68 0.60
0.63 0.68
0.14
25% Quartile 0.40 0.24
0.38 0.48
0.02
Median 0.47 0.41
0.50 0.58
0.03
75% Quartile 0.52 0.50
0.55 0.60
0.05
Average 0.47 0.38
0.47 0.55
0.04
Standard deviation 0.09 0.14
0.10 0.09
0.03
1
Watershedaverage, minimum, and ?maximum C
lit
values were derived using mean, minimum, and
maximum C values for each LULC from literature (Table 2.1), respectively.
Table 2.3 Statistical Summary of Watershedaverage C
lit
Using NLCD 2001 for
Developed and Undeveloped Watersheds.
Undeveloped Developed
Minimum 0.29 0.37
Maximum 59 0.63
25% Quartile 0.33 0.52
Median 0.37 0.54
75% Quartile 0.43 0.58
Average 0.39 0.54
Standard deviation 0.07 0.06
44
Table 2.4?Statistical Summary of C
vbc
and R
v
from NURP (USEPA 1983).
C
vbc
All events
Range of
C
vbc
1
Watershed
Median C
vbc
Watershed
average C
vbc
Standard
deviation
C
vbc
Standard
deviation R
v
2
Minimum 0.00
0.02 0.06 0.10 0.04 0.02
Maximum 0.99
0.97 0.76 0.67 0.30 1.13
25% Quartile 0.17
0.37 0.17 0.20 0.12 0.10
Median 0.29
0.53 0.30 0.31 0.16 0.16
75% Quartile 0.47
0.66 0.42 0.42 0.19 0.28
Average 0.33
0.52 0.31 0.33 0.16 0.21
Standard
deviation 0.21
0.22 0.17 0.15 0.05 0.18
1
Range of C
vbc
is difference between maximum and minimum C
vbc
values calculated for all events in the
same watershed.
2
Standard deviations for R
v
(USEPA 1983) were estimated from median values and
coefficients of variation of R
v
for 60 NURP watersheds.
Table 2.5 Statistical Summary of Watershedaverage C
vbc
for Developed and
Undeveloped Watersheds.
Undeveloped Developed
Minimum 0.10 0.17
Maximum 0.56 0.67
25% Quartile 0.15 0.30
Median 0.19 0.37
75% Quartile 0.36 0.48
Average 0.24 0.39
Standard deviation 0.12 0.13
?
?
?
Table 2.6 Statistical Summary of C
vr
and Absolute Difference (ABS) of C
vr
with
Watershedaverage C
lit
for 83 Texas Watersheds.
C
vr
ABS (C
vr
 C
lit
)
Minimum 0.10 0.01
Maximum 0.78 0.40
25% Quartile 0.24 0.07
Median 0.40 0.14
75% Quartile 0.52 0.24
Average 0.40 0.16
Standard deviation 0.18 0.11
?
45
Table 2.7 Statistical Summary of C
vr
for Developed and Undeveloped Watersheds and
Absolute Difference (ABS) of C
vr
with Watershedaverage C
vbc
and C
lit
for Developed
Watersheds.?
Undeveloped Developed ABS (C
vr
 C
vbc
) ABS (C
vr
 C
lit
)
Minimum 0.10 0.20 0.00 0.01
Maximum 0.70 0.74 0.45 0.38
25% Quartile 0.18 0.34 0.02 0.06
Median 0.24 0.48 0.04 0.12
75% Quartile 0.44 0.60 0.12 0.18
Average 0.31 0.46 0.08 0.14
Standard deviation 0.18 0.15 0.10 0.10
?
46
Fig. 2.1 Map showing the U.S. Geological Survey streamflowgaging stations (dots)
associated with the watershed locations in Texas.
47
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
D
i
s
t
r
i
b
u
t
i
o
n
(
%
)
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
D
i
s
t
r
i
b
u
t
i
o
n
(
%
)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Composite Runoff Coefficient,
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Undeveloped
Developed
1992 (watershedaverage)
2001 ( watershedminimum)
2001 (watershedaverage)
2001 (watershedmaximum)
C
lit
Fig. 2.2 Cumulative distributions of C
lit
obtained using NLCD 1992 and NLCD 2001
(top) and C
lit
using NLCD 2001 for developed and undeveloped watersheds (bottom).
48
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
D
i
s
t
r
i
b
u
t
i
o
n
(
%
)
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
D
i
s
t
r
i
b
u
t
i
o
n
(
%
)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Watershedaverage,
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Runoff Coefficient,
Undeveloped
Developed
Watershedaverage
Wat ershedmedian
All rainfallrunoff events
C
vbc
C
vbc
Fig 2.3. Cumulative distributions of C
vbc
for watershedaverage, watershedmedian, and
all rainfallrunoff events (top) and watershedaverage C
vbc
for developed and
undeveloped watersheds (bottom).
49
0.8
0.4
0.0
0.4
0.8
1.2
R
u
n
o
f
f
c
o
e
f
f
i
c
i
e
n
t
0.0
0.2
0.4
0.6
0.8
1.0
R
u
n
o
f
f
c
o
e
f
f
i
c
i
e
n
t
0 10 20 30 40 50 60 70 80 90 100
Percent impervious
0 10 20 30 40 50 60 70 80 90 100
Texas (45 watersheds) NURP (60 watersheds)
Texas (45 watersheds)
NURP (60 watersheds)
Denver (HSG C & D)
NURP
NURP + Texas
Fig. 2.4 Volumetric runoff coefficients from different studies versus percent impervious
area including regression lines (top) and runoff coefficients with one standard deviations
for 45 Texas watersheds and 60 NURP watersheds (bottom).
50
0.00
0.05
0.10
0.15
0.20
0.25
C
v
r
j
0.0
0.2
0.4
0.6
0.8
1.0
1.2
T
o
t
a
l
R
u
n
o
f
f
(
i
n
c
h
e
s
)
0 1 2 3 4 5
Total Rainfall (inches)
0 1 2 3 4 5
C
vr
Runoff = 0.167 x Rainfall
R
2
= 0.81
Fig. 2.5 The rankordered pairs of observed runoff and rainfall depths (top) and runoff
coefficients derived from the rankordered pairs of observed runoff and rainfall depths
versus total rainfall depths (bottom). All data presented are for the USGS gage station
08042650 (North Trinity Basin in Texas).
51
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
D
i
s
t
r
i
b
u
t
i
o
n
(
%
)
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
D
i
s
t
r
i
b
u
t
i
o
n
(
%
)
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Runoff Coefficient,
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Runoff Coefficient
Undeveloped
Developed
C
vr
C
lit
using NLCD 2001
C
vr
Fig. 2.6 Cumulative distributions of C
vr
and C
lit
(top) and C
vr
for developed and
undeveloped watersheds (bottom).
52
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
R
u
n
o
f
f
c
o
e
f
i
c
i
e
n
t
0 10 20 30 40 50 60 70 80 90 100
Percent impervious
C
vr
C
vbc
R
v
(NURP) Equation (2.6) Equation (2.7)
Fig. 2.7 Runoff coefficients C
vr
, C
vbc
(watershedaverage), and runoff coefficients R
v
from 60 NURP watersheds versus percent impervious area including lines for the
regression equation (2.6) and the regression equation (2.7).
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
R
u
n
o
f
f
c
o
e
f
f
i
c
i
e
n
t
0.01 0.1 1 10 100
Watershed area (mi
2
)
C
vbc
C
vr
C
lit
R
v
(NURP)
Fig. 2.8 Runoff coefficients C
lit
(watershedaverage), C
vbc
(watershedaverage), C
vr
, and
R
v
plotted against watershed area (km
2
).
53
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
R
u
n
o
f
f
c
o
e
f
f
i
c
i
e
n
t
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
(average)
Watershedaverage C C
vr
1:1 line
C
lit
C
vbc
= 0.53 C
lit
+ 0.075 (R
2
= 0.13)
C
vr
= 0.49 C
lit
+ 0.161 (R
2
= 0.07)
vbc
Fig. 2.9 Runoff coefficients C
vbc
(watershedaverage) and C
vr
plotted against C
lit
(watershedaverage) for 83 Texas watersheds.
54
Chapter 3. Ratebased Estimation of the Runoff Coefficients for Selected
Watersheds in Texas
3.1 Abstract
The runoff coefficient, C, of the rational method is an expression of rate
proportionality between rainfall intensity and peak discharge. Values of C were derived
for 80 developed and undeveloped watersheds in Texas using two distinct methods. First,
the ratebased runoff coefficients, C
rate
, were estimated for each of 1,500 rainfallrunoff
events. Second, the frequencymatching approach was used to extract a runoff
coefficient, C
r
, for each watershed. Using the 80 Texas watersheds, comparison of the
two methods shows that about 75 percent of literaturebased runoff coefficients are
greater than C
r
and the watershedmedian C
rate
, but for developed watersheds with more
impervious cover, literaturebased runoff coefficients are less than C
r
and C
rate
. An
equation applicable to many Texas watersheds is proposed to estimate C as a function of
impervious area.
3.2 Introduction
The search for a reliable method for estimation of peak discharges for small and
ungaged undeveloped (rural) watersheds has led to various engineeringdesign methods
(French et al. 1974). These various methods often are applicable for developed (suburban
55
to urban) watersheds (Chow et al. 1988). The rational method likely is the most often
applied method used by hydraulic and drainage engineers to estimate design discharges
for small watersheds. These design discharges are used to size a variety of drainage
structures for small undeveloped and developed watersheds throughout the United States
(Viessman and Lewis 2003).
The rational method (Kuichling 1889) computes the peak discharge, Q
p
, (in m
3
/s
in SI units or ft
3
/s in English units) by:
CIAmQp
0
? (3.1)
where C is the runoff coefficient (dimensionless), I is the rainfall intensity (mm/hr or
in/hr) over a critical period of storm time (typically taken as the time of concentration,
T
c
), A is the drainage area (hectares or acres), and m
o
is the dimensional correction factor
(1/360 = 0.00278 in SI units, 1.008 in English units). Steadystate conditions are needed
for the application of the rational method (French et. al 1974). From inspection of the
equation, it is evident that C is an expression of rate proportionality between rainfall
intensity and peak discharge (flow rate). The theoretical range of values for C is between
0 and 1. The typical ?whole watershed? C values, that is, C values representing the
integrated effects of various surfaces in the watershed and other watershed properties, are
listed for different general landuse conditions in various design manuals and textbooks.
Examples of textbooks that include tables of C values are Chow et al. (1988) and
Viessman and Lewis (2003). Published C values, C
lit
, were sourced from American
Society of Civil Engineers (ASCE) and the Water Pollution Control Federation (WPCF)
in 1960 (ASCE and WPCF 1960). The C
lit
values were obtained from a response survey,
which received ?71 returns of an extensive questionnaire submitted to 380 public and
56
private organizations throughout the United States.? No justification based on analyses of
observed rainfall intensity and peak discharge data for the C
lit
values is apparent in ASCE
and WPCF (1960)?a few analyses of observed relations between rainfall intensity and
peak discharge were considered by Kuichling (1889). In short, the authors of this paper
conclude that C
lit
values appear to be heuristically determined, and therefore a
comparison of ratebased C values derived from observed rainfall and runoff data to the
C
lit
values is made in this study.
Estimation of reliable values of C presents a substantial difficulty in the rational
method and a major source of uncertainty in many small watershed projects (Pilgrim and
Cordery 1993). Furthermore, the concept of ?runoff coefficient? for a watershed is a term
fraught with ambiguities. The volumetric runoff coefficient, C
v
, is the ratio of total runoff
to rainfall (Merz et al. 2006; Dhakal et al. 2012). Because C as an expression of rate
proportionality (equation 3.1), such a coefficient is termed a ratebased runoff coefficient,
C
rate
. It is important to stress that although the C
lit
found in ASCE and WPCF (1960) are
intended for use in the ratebased rational method, C
lit
appear not to be derived from any
observed data.
When observed rainfall and runoff data are available, the C
rate
is computed through the
rational method (Pilgrim and Cordery 1993) as:
AtIm
Q
C
av
p
rate
)(
0
? , (3.2)
where I(t
av
) is the average rainfall intensity over a t
av
time period, which is the rainfall
intensity averaging time period (a critical time period) rather than the entire rainfall event
duration (Kuichling 1889; Schaake et al. 1967). The t
av
should be a period of time for a
57
storm that contributes runoff that produces the observed Q
p
. Kuichling (1889) argues
against using the entire rainfall event duration, t
w
, to obtain average rainfall intensity, I
w
,
because I
w
is in general not appropriate, which results from rainfall durations of real
storms often being greater than the characteristic time of small watersheds considered by
Kuichling. Thus, a lasting contribution of Kuichling (1889) was the introduction of the
concept of the time of concentration, T
c
, of a watershed. This time also is termed a
?critical storm duration? because this uses an average rainfall intensity that produces
reliable peak discharge estimates. The T
c
is influenced in part by drainage area, which is a
major criterion used to assess the applicability of the rational method (Chow et al. 1988).
In particular, TxDOT (2002) recommends the use of the rational method for watersheds
with very small drainage areas of < 0.8 km
2
(200 acres).
Other investigators have reported C values derived from analysis of observed rainfall and
runoff data for various watersheds throughout the world. Schaake et al. (1967) examined
the rational method using experimental rainfall and runoff data collected from 20 small
urban watersheds of < 0.6 km
2
(150 acres) in Baltimore, MD. Those authors used
watershed lag time to compute average rainfall intensity and used a frequencymatching
approach.
Hotchkiss and Provaznik (1995) estimated C
rate
for 24 rural watersheds in south
central Nebraska using eventpaired and frequencymatched data. Young et al. (2009)
estimated C
rate
for 72 rural watersheds in Kansas with drainage areas of < 78 km
2
(30
mi
2
) for different return periods. The peak discharge for each return period was estimated
using annual peak frequency analysis of the gaged peak discharges and rainfall intensity
obtained from rainfall intensitydurationfrequency tables (Young et al. 2009).
58
In this study, two methods were used to estimate C
rate
for 80 selected watersheds
in Texas. Both methods rely on analysis of observed rainfall and runoff data. First, C
rate
was estimated using equation (3.2); the I(t
av
) was computed as the maximum intensity for
a moving time window of duration T
c
before and up to the time to peak, T
p
. The T
c
was
derived for the study watersheds using the KerbyKirpich approach (Roussel et al. 2005;
Fang et al. 2008). A total of about 1,500 rainfallrunoff events from 80 Texas watersheds
were analyzed to determine eventspecific, watershedmedian, and watershedmean C
rate
values. Second, the frequencymatching approach (Schaake et al., 1967) was used to
derive a representative C referred to as C
r
for each of the 80 watersheds. The study also
compares C values from the two different methods and those published in the literature.
Finally, an equation of C as a function of the percentage of impervious area is proposed
for the 80 watersheds.
3.3 Study Area and RainfallRunoff Database
Watershed data from a larger dataset accumulated by researchers from the U.S.
Geological Survey (USGS) Texas Water Science Center, Texas Tech University,
University of Houston, and Lamar University (Asquith et al. 2004) are used for this
study. Ten watersheds out of about 90 represented by USGS streamflowgaging stations
in the source database (Asquith et al. 2004) are not used in this study because less than
four rainfall and runoff events were recorded for each of these 10 watersheds. The
locations of 80 USGS streamflowgaging stations representing 80 watersheds in Texas
are shown in Figure 3.1. Incidentally, these data also are used by Asquith and Roussel
(2007), Cleveland et al. (2006), Fang et al. (2007, 2008), and Dhakal et al. (2012). The
59
rainfallrunoff dataset consists of about 1,500 rainfallrunoff events which occurred
between 1959 and 1986. The number of events available for each watershed varied from
4 to 50 events with median and mean values of 16 and 19 events, respectively. Values of
rainfall depths for about 1,500 events ranged from 3.56 mm (0.14 in.) to 489.20 mm
(19.26 in.), with median and mean values of 57.66 mm (2.27 in.) and 66.8 mm (2.63 in.),
respectively.
The drainage area of study watersheds range from approximately 0.2?320 km
2
(0.1?123.6 mi
2
); the median and mean values are 17.0 km
2
(6.6 mi
2
) and 37.3 km
2
(14.4
mi
2
), respectively. The stream slope of study watersheds range from approximately
0.0022?0.0196 dimensionless; the median and mean values are 0.0076 and 0.0081,
respectively. The percentage of impervious area (IMP) of study watersheds range from
approximately 0 to 73%; the median and mean values are 18.0 and 28.2, respectively.
There has been discussion in the literature concerning the size of watersheds for which
the application of the rational method is appropriate. For application of the rational
formula, Kuichling (1889, pages 40?41) stated: ?For large areas, on the other hand, a
more elaborate analysis becomes necessary in order to find under what condition the
absolute maximum discharge will occur, although the method of procedure above
indicated will remain the same.? Kuichling (1889) did not suggest a specific large area
limit. ASCE and WPCF (1960, p. 32), made the following statement when the rational
method was introduced for design and construction of sanitary and storm sewers:
?Although the basic principles of the rational method are applicable to large drainage
areas, reported practice generally limits its use to urban areas of less than 5 square miles.?
Pilgrim and Cordery (1993, p. 9.14) explained that the rational method is one of the three
60
methods widely used to estimate peak flows for small to medium sized basins, and wrote
?it is not possible to define precisely what is meant by ?small? and ?medium? sized, but
upper limits of 25 km
2
(10 mi
2
) and 550 km
2
(200 mi
2
), respectively, can be considered as
general guides.? Young et al. (2009) stated that the rational method might be applied to
much larger drainage areas than typically assumed in some design manuals, as long as the
watershed is unregulated. Results of this study will further indicate that there is no
demonstrable relation between runoff coefficient and drainage area.
For any watershed (regardless of its size), some of the attributes necessary to
apply the Kuichling method are the time of concentration (T
c
), main channel length (L
c
),
and channel slope (S
c
). For each of the 80 Texas watersheds, a geospatial database was
developed by Roussel et al. (2005) containing L
c
and S
c
for each watershed, along with
drainage area, basin width, longitude, latitude, and 39 other watershed characteristics. For
this paper, L
c
and S
c
are used to estimate time of concentration T
c
by Kirpich (1940) for
channel flow plus travel time for overland flow using Kerby (1959). A combination of
the methods of Kirpich (1940) and Kerby (1959) is discussed by Roussel et al. (2005) and
Fang et al. (2008). The Kirpich equation (1940) was developed from the Soil
Conservation Services (SCS) data for rural watersheds with drainage areas less than 0.45
km
2
and is presented below:
,978.3
385.077.0 ?
?
ccc
SLT (3.3)
where, L
c
is the channel length in km and S
c
is the channel slope in m/m. Fang et al.
(2007, 2008) demonstrated that, for watersheds with relatively large drainage areas (more
than 50 km
2
), the Kirpich equation provides as reliable an estimate of T
c
as the other
empirical equations developed for large watersheds and the SCS velocity method
61
(Viessman and Lewis 2003). The T
c
estimated using the Kirpich equation reasonably
approximate the average T
c
estimated from observed rainfall and runoff data (Fang et al.
2007). The T
c
for the study watersheds ranged from 1.1 hours to 16.7 hours with median
and mean values of 2.8 hours and 3.8 hours respectively.
Each of the 80 Texas watersheds was previously classified as either developed or
undeveloped (Roussel et al. 2005, Cleveland et al. 2008). The classification scheme of
developed and undeveloped watersheds is consistent with the characterization of
watersheds in more than 220 USGS reports of Texas data from which the original data
for the rainfall and runoff database were obtained (Asquith et al. 2004). Although this
binary classification seems arbitrary, it does take into account the uncertainty in
watershed development conditions for the time period of available data (Asquith and
Roussel 2007). This binary classification was used by Asquith et al. (2006) in a
regionalization study of unit hydrographs for the Texas watersheds (Asquith et al., 2004).
Using the binary classification scheme for the 80 Texas watersheds, there are 44
developed watersheds in four metropolitan areas in Texas (Austin, Dallas, Fort Worth,
and San Antonio) and 36 undeveloped watersheds. The 36 undeveloped watersheds
consist of 16 watersheds near these four cities and 20 rural watersheds.
3. 4 Runoff Coefficients Estimated from Event RainfallRunoff Data
Ratebased C Derived for Individual RainfallRunoff Events
62
For this study, the intensity I in equation (3.2) is the maximum rainfall intensity
before the time to peak, T
p
, of a runoff hydrograph and is calculated as the maximum
intensity found by a moving time window of duration t
av
through the 5minute interval
rainfall hyetograph for the storm event. For data processing, only the largest Q
p
for each
storm event (in the case of multiple peaks in the overall hydrograph) was used.
The computation of C
rate
is illustrated by example. In Figure 3.2, the I and C
rate
values are
shown as a function of t
av
for two storm events gaged by the USGS: one on 09/22/1969
at USGS streamflowgaging station 08048550 Dry Branch at Blandin Street, Fort Worth,
Texas (hereinafter Dry Branch) and the second on 04/25/1970 at 08058000 Honey Creek
near McKinney, Texas (hereinafter Honey Creek). As shown in Figure 3.2, as t
av
increases I decreases and C
rate
increases. For example, for the storm event at Dry Branch,
as t
av
increases from 5 minutes to 3.5 hours, I decreases from about 119 mm/hr (4.7 in/hr)
to about 16.3 mm/hr (0.64 in/hr) and C
rate
increases from 0.04 to 0.30. For the Dry
Branch and Honey Creek watersheds, estimated T
c
is 1.8 hours and 1.5 hours,
respectively (Figure 3.2). These T
c
values are derived from the KerbyKirpich method
(Roussel et al. 2005; Fang et al. 2008); the corresponding C
rate
values for the Dry Branch
and Honey Creek watersheds are 0.23 and 1.70, respectively (Figure 3.2). Following the
analysis leading to Figure 3.2, one C
rate
value was determined using I corresponding to a
moving time window T
c
for each of about 1,500 events from the 80 Texas watersheds.
The occurrence of C
rate
> 1 is related to unknown errors in T
c
used to calculate I (see
Figure 3.2), rainfall characteristics, fundamental measurement errors of rainfall and
runoff data, or other unusual hydrologic factors. Several studies (French et. al 1974;
Pilgrim and Cordery 1993; Young et al. 2009) have shown that values of C
rate
greater
63
than 1 are possible when ratebased C was determined from observed peak flow rate and
computed rainfall intensity over a critical period of storm time. If the time of
concentration is exactly correct for the watershed, and if the rainfall were spatially and
temporally homogeneous and isolated (no preceding rainfall), then C
rate
would have to be
less than or equal to 1, but the rainfall normally varies in time and in space. Averaging
temporal and spatial variability of rainfall leads to lower rainfall intensities and
consequently lower predicted peak discharge values. Thus, when using an average
rainfall as a predictor of Q
p
, the C value will necessarily be higher than if the rainfall
were truly uniform in space and time.
Most of the C
v
values derived from about 1,500 rainfall events in Texas
watersheds (Dhakal et al. 2012) are between 0 and 1. Ratebased runoff coefficient C
rate
and volumetricbased runoff coefficient C
v
are defined differently and were determined
using different approaches from observed rainfall and runoff data. The use of ratebased
runoff coefficients is appropriate if one wants to determine peak discharge using the
rational method, and volumetric runoff coefficient can be used to estimate fractional
rainfall loss using the constant fraction method (McCuen 1998) or for hydrologic
modeling and runoff volume design purposes for a stormwater quality control basin
(USEPA 1983; Guo and Urbonas 1996; Mays 2004).
Frequency distributions of C
rate
values computed for about 1,500 events from the
80 Texas watersheds are shown in Figure 3.3, and summary statistics are listed in Table
3.1. Recalling that T
c
was computed using the KerbyKirpich approach, for events where
T
p
< T
c
, the mean value of C
rate
is 0.31. In contrast, for events where T
p
? T
c
, the mean
value of C
rate
is 0.50 (Table 3.1). From inspection of the frequency distributions of C
rate
64
shown in Figure 3.3, estimates of C
rate
are significantly greater (WelchSatterthwaite t
test, pvalue < 0.0001) for storm events when T
p
? T
c
than those from events when T
p
<
T
c
. Therefore, values of C
rate
are dependent on the duration of rainfall event, which
supports the idea proposed by Kuichling (1889). Kuichling?s idea is that as T
c
is reached,
discharge for a watershed becomes a maximum (a peak) because the entire area is
contributing runoff to the outlet. For cases considered in this research, the maximum
value of Crate sometimes exceeded 1 (up to 4.48 when T
p
? T
c
). For 124 of about 1,500
events, the calculated C
rate
is greater than 1.
Watershed Mean and Median Runoff Coefficients
Watershed mean and median values of C
rate
for the 80 Texas watersheds were
calculated for all observed storms in the same watershed (regardless of whether T
p
was
less than or greater than T
c
). Computed watershed means of C
rate
range from 0.07 to 1.79,
and the watershed medians of C
rate
range from 0.07 to 1.73 (Table 3.2). The average
values of the individual watershed mean and median C
rate
are 0.44 and 0.40, respectively
(Table 3.2). Standard deviations from the watershedmeans C
rate
range from 0.03 to 0.87.
Frequency distributions of the watershed mean and median of C
rate
are shown in Figure
3.3; these distributional locations (means) are similar at a significance level of 0.01
(paired ttest, pvalue = 0.04).
The amount of developed land in a watershed influences various runoff
characteristics of a watershed. To study the relation between C
rate
and the binary
watershed development classification, statistical summaries of watershed median C
rate
were computed (Table 3.3). The watershed median C
rate
for developed watersheds range
65
from 0.17 to 1.73, and for undeveloped watersheds, the watershed median C
rate
ranged
from 0.07 to 0.73. Of the developed watersheds, only two had a watershedmedian C
rate
>
1. The corresponding frequency distributions of watershedmedian C
rate
for developed
and undeveloped watersheds are shown in Figure 3.3B. The watershedmedian value of
C
rate
for developed watersheds is 0.40 and for undeveloped watersheds watershedmedian
value of C
rate
is 0.20 (Table 3.3). The C
rate
values of the developed watersheds are
significantly larger than those from the undeveloped watersheds (Figure 3.3) as
anticipated (WelchSatterthwaite ttest, pvalue <0.0001).
3.5 Runoff Coefficients from the FrequencyMatching Approach
The frequencymatching approach assumes return periods of rainfall and runoff
events are the same (Hawkins 1993). Specifically, the Tyear storm produces the Tyear
peak discharge. An alternative viewpoint is that the frequencymatching approach forces
the largest rainfall intensity to produce the largest peak discharge within a given dataset.
The authors observed that this assumption is implicit in circumstances of practical
application of the rational method. Many design engineers assume that the Tyear storm
produces the Tyear discharge. Although not a physical requirement, this assumption
generally is appropriate in small watersheds.
The maximum rainfall intensities and the observed peak discharges were
independently ranked from largest to smallest for each of the 80 Texas watersheds. The
frequencymatched C was computed from the rankordered pairs of the observed peak
discharge and the maximum rainfall intensity for each storm event using:
66
AIm
Q
C
j
pj
rj
0
? , (3.4)
where C
rj
is the runoff coefficient corresponding to the maximum rainfall intensity I
j
, the
observed peak discharge Q
pj
of the j
th
rankorder of I
max
Q
p
data pairs, and drainage area
A. A plot of runoff coefficients, C
rj
, versus the maximum rainfall intensity was prepared
for each watershed. For most of the watersheds, the C
rj
increases until acquiring an
approximate constant value as judged by an analyst. This constant value is referred to as
C
r
. For example, the plot for USGS streamflowgaging station 08042650 North Creek
Surface Water Station 28A near Jermyn, Texas (hereinafter North Creek near Jermyn) is
presented in Figure 3.4A; C
r
= 0.20 for this watershed.
The C
r
also can be estimated from the slope of the regression line obtained from the plots
of the rankordered Q
pj
/(0.00278*A) or Q
pj*
values versus the rankordered I
j
. For
example, the regression equation for North Creek near Jermyn is Q
pj*
(m
3
/s/ha) = 0.19 * I
j
(mm/hr) [Figure 3.4B]. The slope of this (and other similar equations) also is
representative of C
r
. Using the slope of the line, C
r
= 0.19 for the North Creek near
Jermyn. For most of the 80 Texas watersheds, C
r
values obtained using analyst judgment
or the regression slope method have approximately the same value, and C
r
values ranged
from 0.10 to 1.2 (one outlying C
r
value of 1.2 was the only C
r
value > 1). The mean and
medians for C
r
were 0.42 and 0.37, respectively. A statistical summary of C
r
is listed in
Table 3.4.
Comparison C
r
to Watershed Median C
rate
and Literaturebased C
lit
67
For the 80 Texas watersheds, distributions of C
r
and watershed median C
rate
follow the same shape (Figure 3.5A) and are not statistically different at the 0.05
significance level (paired ttest, pvalue = 0.27). The difference between C
r
and
watershed median C
rate
for each watershed was calculated, and a statistical summary of
the differences is listed in Table 3.4. The median value of C
r
minus C
rate
differences is
0.03 (Table 3.4). The minimum and maximum differences are 0.53 and 0.52 (Table 3.4),
respectively, and quartiles of the differences between C
r
and watershedmedian C
rate
are
considered acceptably small (less than 0.06). About 74% of C
r
and watershedmedian
C
rate
values differ less than ?0.1 (Table 3.4).
The frequency distributions of literaturebased C
lit
from landuse data (Dhakal et
al. 2012) for developed and undeveloped watersheds are shown in Figures 3.5A and
3.5B. The differences between C
r
and C
lit
or C
rate
and C
lit
are larger than the differences
between C
r
and C
rate
. When the runoff coefficient is less than 0.55, C
lit
is greater than C
r
,
otherwise C
lit
is smaller than C
r
(Figure 3.5A). About 75% of C
lit
values are greater than
C
r
(Figure 3.5A). For typical applications of the rational method in urban (developed)
watersheds, using the typically smaller C
lit
value for the watershed would underestimate
Q
p
for design purposes. The difference between watershedmedian C
r
and C
lit
or C
rate
and
C
lit
for each watershed was calculated, and statistical summary of the differences is listed
in Table 3.4. The median (50th percentile) of C
r
minus C
lit
and C
rate
minus C
lit
are 0.11
and 0.14, respectively, compared to the smaller mean differences between C
r
and C
lit
and
C
rate
and C
lit
(0.06 or 0.07, respectively) (Table 3.4).
C
r
and C
lit
for Developed and Undeveloped Watersheds
68
C
r
and C
lit
were grouped into two categories of developed and undeveloped
watersheds (Roussel et al. 2005). Statistical summaries of C
r
and C
lit
for developed and
undeveloped watersheds are listed in the Table 3.5; C
r
and C
lit
frequency distributions are
shown in Figure 3.5B. The median value of C
r
for undeveloped watersheds is 0.26 and
the median value for the developed watershed is 0.45. These median values are similar to
those for watershedmedian C
rate
(Table 3.3). The median and mean values of C
lit
are
larger than those of C
r
for both developed and undeveloped watersheds (Table 3.5).
About 68 and 78% of C
lit
are larger than C
r
for developed and undeveloped watersheds,
respectively (Figure 3.5).
C
r
in relation to impervious area
For this study, the percentage of impervious area for each watershed was
computed using 1992 National Land Cover Data for Texas (Vogelmann and others,
2001). C
r
for 45 Texas watersheds with watershed imperviousness (IMP) greater than
10% are plotted in Figure 3.6. Schaake et al. (1967) developed the regression equation
SIMPC 05.065.014.0 ???
(referred to herein as the ?Schaake et al. equation?) for
urban drainage areas in Baltimore, MD to relate C
r
(for a return period of 5 years) to the
relative imperviousness of the drainage area and channel slope of the watershed. C
r
was
calculated using the Schaake et.al equation for the 45 Texas watersheds with watershed
imperviousness greater than 10%. For comparison purposes with C
r
values for 45 Texas
watersheds with watershed imperviousness (IMP) greater than 10%, C
r
values calculated
using the Schaake et. al equation also are plotted on Figure 3.6, along with C values
extracted from Jens (1979), and equation 3.5 from Asquith (2011).
69
The results of these three studies [this study, Schaake et al. (1967), and Jens
(1979)] are consistent?the value of C increases with increasing IMP. Asquith (2011)
proposes a single equation to estimate C for Texas watersheds as a function of IMP. The
equation is
15.085.0 ?? IMPC (3.5)
The equation was used to estimate the runoff coefficient C* (?Cstar?) for the unified
rational method (URAT) developed for a TxDOT research project summarized in
Cleveland et al. (2011). Equation 3.5 is plotted in Figure 3.6 and is consistent with the
general pattern of the data.
Several studies (Jens 1979; Pilgrim and Cordery 1993; Hotchkiss and Provaznik
1995; Titmarsh et al. 1995; Young et al. 2009) have demonstrated that C is highly
dependent on the return period T. In this study, ratebased runoff coefficients were not
derived for any return period because the observed data do not include all events that
would constitute the complete annual series needed for the frequency analysis. Return
period based C(T) values were computed by the authors using regional regression
equations for Q
p
and I for the 36 undeveloped Texas watersheds in the database and
presented as a separate paper (Dhakal et al. 2011).
Correlation between C and Watershed Area
In order to evaluate the C values for the 80 Texas watersheds, C
r
and watershed
median C
rate
were used to estimate the peak discharge rates (Q
p
) for each of about 1,500
rainfallrunoff events using the rational equation (3.1). The observed versus the modeled
70
Q
p
are shown in Figure 3.7. The peak relative error (QB) between the observed and the
modeled peak discharges was estimated to analyze the model results (Cleveland et al.
2006):
,
i
ii
O
OP
QB
?
? (3.6)
where, P
i
are the modeled peak discharge values, O
i
are the observed peak discharge
values. Cleveland et al. (2006) suggested the following range of the QB for the
acceptance of model performance:
25.025.0 ??? QB (3.7)
Median QB values derived using C
r
and watershedmedian C
rate
are 0.11 and 0.00,
respectively. Similarly, use of C
r
resulted in about 56% and use of watershedmedian
C
rate
resulted in about 59% of storms with QB less than ?50%. About 87 percent of the
modeled Q
p
values from both cases are within about half of a log cycle from the equal
value line (Figure 3.7). The differences between the observed and modeled Q
p
are
generally within about a third of a log cycle, which is an uncertainty similar to that
reported for regional regression equations of peak discharge in Texas by Asquith and
Roussel (2009).
The observation that about 87% of the modeled Q
p
values from both cases are
within about half of a log cycle from the equal value line supports the conclusion by
ASCE and WPCF (1960), Pilgrim and Cordery (1993), and Young et al. (2009) that the
rational method may be applied to much larger drainage areas than typically indicated
(assumed) in some design manuals, as long as streamflows in the watershed are
unregulated (Young et al. 2009). ASCE and WPCF (1960) state ?Although the basic
71
principles of the rational method are applicable to large drainage areas, reported practice
generally limits its use to urban areas of less than 5 square miles. Development of data
for application of hydrograph methods is usually warranted on larger areas.? Kuichling
(1889, pages 40?41) made a similar statement for application of the rational formula to
large watershed areas.
C
rate
for all events, watershedmean C
rate
, and C
r
versus watershed area (km
2
) are
displayed on Figure 3.8. Of note on Figure 3.8 is that the runoff coefficients are subject
to substantial variability. That is, based on visual examination, there appears to be no
relation between watershed drainage area and runoff coefficient. To apply a quantitative
test, Pearson correlation coefficients between watershedmean C
rate
and C
r
and watershed
area are 0.27 and 0.26 with pvalues of 0.012 and 0.018, respectively. Therefore, at the
95% confidence level (n = 80 observations), Pearson correlation coefficients between
C
rate
and C
r
and watershed area are statistically significant but correlations are weak
(determination coefficient r
2
? 0.07) and C exhibits high variability (Fig. 3.8). Only
about 7% of the variance is described by the correlation. Although statistically
significant, the contribution of the correlation to description of the variability of C
rate
and
C
r
is not useful in an engineering context.
Of the 80 study watersheds, drainage area exceeds 40 km
2
(15 mile
2
) for 17. The
choice of 40 km
2
is completely arbitrary for the purposes of examining the runoff
coefficient for relatively large watersheds. For this group of largest watersheds, average
values of watershedmean C
rate
and C
r
are 0.27 and 0.29, respectively. Standard
deviations of watershedmean C
rate
and C
r
are 0.11 and 0.12, respectively. In comparison,
literaturebased C
lit
from landuse data (Dhakal et al. 2012) for these watersheds ranged
72
from 0.30 to 0.55 with average value of 0.42 and standard deviation of 0.10. By
inference, literaturebased C
lit
might be too large (Fig. 3.5), so estimates derived from
application of C
lit
to relatively large watersheds might lead to overly conservative
estimates of discharge. Therefore, literaturebased C values, e.g., published by ASCE and
WPCF (1960) and current textbooks and design manuals, should not be used for
watersheds with large drainage area.
Although published values for C are not appropriate for relatively large
watersheds, the rational method can be applied if reasonable estimates of runoff
coefficient can be derived. One source would be observations of runoff coefficient from
hydrologically similar watersheds. Another would be derivation from observations of
rainfall and runoff from the watershed of interest. The published limits [5 square miles,
ASCE and WPCF (1960); 200 acres, TxDOT (2002)] on the maximum drainage area for
application of the rational method seem to be arbitrary.
The authors do not advocate any specific limits that should be imposed on
drainage area for application of the rational method. Therefore, it remains the
responsibility of the enduser to apply appropriate engineering judgment when applying
the rational method and the assumptions associated with the method, such as steadystate
conditions.
3.6 Summary
The runoff coefficient, C, of the rational method is an expression of rate
proportionality between rainfall intensity and peak discharge. Two methods were used to
estimate C. Both methods used about 1,500 observed rainfall and runoff events data from
73
80 Texas watersheds to derive C. For the first method, the ratebased runoff coefficient,
C
rate
, was estimated for each rainfallrunoff event by the ratio of event peak discharge in
a time series to the corresponding largest average rainfall intensity, I, in the same time
series, averaged over the time window length. Time of concentration, T
c
, was used as the
time window length to estimate I. The T
c
values estimated using the KerbyKirpich
method were used for the 80 watersheds studied. The ratebased C is dependent on
rainfall intensity averaging time t
av
used for the study, because based on equation (3.2),
estimates of the runoff coefficient based on observed data cannot be decoupled from the
selection of the timeresponse characteristic. Watershedmean and watershedmedian
values of C
rate
were derived. The distributions of the watershedmean and watershed
median C
rate
are similar. Lastly, the C
rate
values for the developed watersheds are
consistently higher than those for undeveloped watersheds. For the second method, the
frequencymatching approach, similar to the procedure used by Schaake et al. (1967),
was used to sort peak discharges and average rainfall intensities independently and then
to compute the ratebased C from the rational formula. A constant runoff coefficient C
r
for the watershed was derived from the plot of the ratebased C versus I. The C
r
values
for the developed watersheds are consistently greater than those for the undeveloped
watersheds; about 74% of C
r
and watershedmedian C
rate
differ less than ?0.1 (Table 3.4).
The values of C
r
and C
rate
were compared with the literature based runoff coefficients
(C
lit
) developed from landuse data for these study watersheds (Dhakal et al. 2012).
About 75% of C
lit
values are greater than C
r
(Figure 3.5). For typical applications of the
rational method in developed (urban) watersheds, watershed C
lit
is less than C
r
(Figure
3.5); using smaller C
lit
would underestimate Q
p
for design. An equation was proposed to
74
estimate ratebased C as a function of the percentage of impervious area (IMP) for Texas
watersheds, and prediction from the equation is consistent with the results from Schaake
et al. (1967) and Jens (1979).
3.7 Acknowledgments
The authors thank TxDOT project director Mr. Chuck Stead, P.E., and project
monitoring advisor members for their guidance and assistance. They also express thanks
to technical reviewers Jennifer Murphy and Nancy A. Barth from USGS Tennessee and
California Water Science Centers, respectively, and to three anonymous reviewers; the
comments and suggestions greatly improved the paper. This study was partially
supported by TxDOT Research Projects 0?6070, 0?4696, 0?4193, and 0?4194.
3.8 Notation
The following symbols are used in this paper:
A = drainage area in hectares or acres;
C
lit
= literature?based runoff coefficient developed from landuse data;
C
rate
= ratebased runoff coefficient;
C
r
= runoff coefficients from the frequency matching approach;
C
rj
= runoff coefficient estimated from the ratio of jth rankordered peak discharge and
the maximum rainfall intensity data pairs;
C
v
= volumetric runoff coefficient;
75
C* = runoff coefficients as a function of percentage of impervious area from equation
(3.5);
I = average rainfall intensity (mm/hr or in. /hr) with the duration equal to time of
concentration;
I
j
= the maximum rainfall intensity
of the j
th
order;
IMP = percentage of impervious area expressed as a decimal (50% = 0.5) for a watershed
area;
I
w
= average rainfall intensity from the entire rainfall event duration;
j = j
th
term in the sequence of ordered peak discharge and the maximum rainfall intensity
data pairs;
L
c
= channel length in km;
m
o
= the dimensional correction factor (1.008 in English units, 1/360 = 0.00278 in SI
units);
O
i
= observed peak discharge for computing QB;
P
i
= modeled peak discharge for computing QB;
QB = peak relative error between the observed and simulated peak discharges;
Q
p
= peak discharge in m
3
/s or ft
3
/s;
Q
pj
= peak discharge of the j
th
rankorder of maximum rainfall intensity and peak
discharge data pairs in cubic meters per second;
Q
pj*
= Q
pj
divided by 0.0028 times the drainage area in cubic meters per second per
hectare;
76
S = channel slope (Schaake et al. (1967));
S
c
= channel slope in m/m;
t
av
= rainfall intensity averaging time period;
T
c
= time of concentration;
T
p
= time to peak;
t
w
= rainfall event duration;
3.9 References:
ASCE and WPCF. (1960). ?Design and construction of sanitary and storm sewers.?
American Society of Civil Engineers (ASCE) and Water Pollution Control Federation
(WPCF).
Asquith, W.H., Thompson, D.B., Cleveland, T.G., and Fang, X. (2004). ?Synthesis of
rainfall and runoff data used for Texas department of transportation research projects
0?4193 and 0?4194.? U.S. Geological Survey, OpenFile Report 2004?1035, Austin,
Texas.
Asquith, W.H., Thompson, D.B., Cleveland, T.G., and Fang, X. (2006). ?Unit
hydrograph estimation for applicable Texas watersheds.? Texas Tech University,
Center for Multidisciplinary Research in Transportation, Austin, Texas.
Asquith, W.H., and Roussel, M.C. (2007). ?An initialabstraction, constantloss model
for unit hydrograph modeling for applicable watersheds in Texas.? U.S. Geological
Survey Scientific Investigations Report 2007?5243, 82 p.
[http://pubs.usgs.gov/sir/2007/5243].
Asquith, W.H., and Roussel, M.C. (2009). "Regression equations for estimation of annual
peakstreamflow frequency for undeveloped watersheds in Texas using an L
momentbased, PRESSMinimized, residualadjusted approach." U.S. Geological
Survey Scientific Investigations Report 2009?5087, 48 p.
Asquith, W.H. (2011). ?A proposed unified rational method for Texas? in Cleveland,
T.G., Thompson, D.B., Fang, X., (2011), ?Use of the Rational and Modified Rational
Methods for TxDOT Hydraulic Design.? Research Report 0?6070?1, section 3, pp.
18?58, Texas Tech University, Lubbock, TX.
77
Chow, V.T., Maidment, D.R., and Mays, L.W. (1988). Applied hydrology, McGraw Hill
New York.
Cleveland, T.G., He, X., Asquith, W.H., Fang, X., and Thompson, D.B. (2006).
?Instantaneous unit hydrograph evaluation for rainfallrunoff modeling of small
watersheds in north and south central Texas.? Journal of Irrigation and Drainage
Engineering, 132(5), pp. 479?485.
Cleveland, T.G., Thompson, D.B., Fang, X., and He, X. (2008). ?Synthesis of unit
hydrographs from a digital elevation model.? Journal of Irrigation and Drainage
Engineering, 134(2), pp. 212?221.
Cleveland, T.G., Thompson, D.B., and Fang, X. (2011). ?Use of the Rational and
Modified Rational Methods for TxDOT Hydraulic Design.? Research Report 0?
6070?1, 143 p., Texas Tech University, Lubbock, TX.
Chow, V.T., Maidment, D.R., and Mays, L.W. (1988). Applied hydrology, McGraw Hill,
New York.
Dhakal, N., Fang, X., Cleveland, T.G., Thompson, D.B., Asquith, W.H., and Marzen, L.J.
(2012). ?Estimation of volumetric runoff coefficients for Texas watersheds using
landuse and rainfallrunoff data.? Journal of Irrigation and Drainage Engineering,
138(1):4354.
Dhakal, N., Fang, X., Asquith, W.H., Cleveland, T.G., and Thompson, D.B., (2011).
?Return period adjustments for runoff coefficients based on analysis in Texas
watersheds.? Journal of Irrigation and Drainage Engineering (submitted).
Fang, X., Thompson, D.B., Cleveland, T.G., and Pradhan, P. (2007). ?Variations of time
of concentration estimates using NRCS velocity method.? Journal of Irrigation and
Drainage Engineering, 133(4), pp. 314?322.
Fang, X., Thompson, D.B., Cleveland, T.G., Pradhan, P., and Malla, R. (2008). ?Time of
concentration estimated using watershed parameters determined by automated and
manual methods.? Journal of Irrigation and Drainage Engineering, 134(2), pp. 202?
211.
French, R., Pilgrim, D.H., and Laurenson, B.E. (1974). ?Experimental examination of the
rational method for small rural catchments.? Civil Engineering Transactions, The
Institution of Engineers, Australia CE16(2), pp. 95?102.
Guo, J.C.Y., and Urbonas, B. (1996). "Maximized detention volume determined by
runoff capture ratio." Journal of Water Resources Planning and Management, 122(1),
3339.
78
Hawkins, R.H. (1993). ?Asymptotic determination of runoff curve numbers from data.?
Journal of Irrigation and Drainage Engineering, 119(2), pp. 334?345.
Hotchkiss, R.H., and Provaznik, M.K. (1995). ?Observations on the rational method C
value.? Watershed management; planning for the 21st century; proceedings of the
symposium sponsored by the Watershed Management Committee of the Water
Resources, New York.
Jens, S.W. (1979). Design of urban highway drainage, Federal Highway Administration
(FHWA), Washington DC.
Kerby, W.S. (1959). ?Time of concentration for overland flow.? Civil Engineering, 29(3),
174 p.
Kirpich, Z.P. (1940). ?Time of concentration of small agricultural watersheds.? Civil
Engineering, 10(6), 362 p.
Kuichling, E. (1889). ?The relation between the rainfall and the discharge of sewers in
populous areas.? Transactions, American Society of Civil Engineers 20, pp. 1?56.
Mays, L.W. (2004). Urban stormwater management tools, McGrawHill, New York.
McCuen, R.H. (1998). Hydrologic analysis and design, PrenticeHall, Inc., Upper Saddle
River, N.J.
Merz, R., Bl?schl, G., and Parajka, J. (2006). ?Spatiotemporal variability of event runoff
coefficients.? Journal of Hydrology, 331(3?4), pp. 591?604.
Pilgrim, D.H., and Cordery, I. (1993). ?Flood runoff.? Handbook of hydrology, D. R.
Maidment, ed., McGrawHill, New York, 9.1?9.42.
Roussel, M.C., Thompson, D.B., Fang, D.X., Cleveland, T.G., and Garcia, A.C. (2005).
?Timing parameter estimation for applicable Texas watersheds.? 0?4696?2, Texas
Department of Transportation, Austin, Texas.
Schaake, J.C., Geyer, J.C., and Knapp, J.W. (1967). ?Experimental examination of the
rational method.? Journal of the Hydraulics Division, ASCE, 93(6), pp. 353?370.
Titmarsh, G.W., Cordery, I., and Pilgrim, D.H. (1995). ?Calibration procedures for
rational and USSCS design flood methods.? Journal of Hydraulic Engineering,
121(1), pp. 61?70.
TxDOT. (2002). Hydraulic design manual, The Bridge Division of the Texas Department
of Transportation (TxDOT), 125 East 11th street, Austin, Texas 78701.
79
USEPA. (1983). "Results of the nationwide urban runoff program, Volume I  Final
Report." National Technical Information Service (NTIS) Accession Number: PB84
185552, Water Planning Division, United States Environmental Protection Agency
(USEPA), Washington D.C.
Viessman, W., and Lewis, G.L. (2003). Introduction to hydrology, 5th Ed., Pearson
Education, Upper Saddle, N.J., 612 p.
Vogelmann, J.E., S.M. Howard, L. Yang, C. R. Larson, B. K. Wylie, and J. N. Van Driel.
(2001). ?Completion of the 1990s National Land Cover Data Set for the conterminous
United States.? Photogrammetric Engineering and Remote Sensing, 67, pp. 650662.
Young, C.B., McEnroe, B.M., and Rome, A.C. (2009). ?Empirical determination of
rational method runoff coefficients.? Journal of Hydrologic Engineering, 14, 1283 p.
80
Table 3.1 Statistical Summary of C
rate
Calculated from Observed RainfallRunoff Event
Data.
T
p
? T
c
1
T
p
< T
c
2
All events
Minimum 0.01 0.01
0.01
Maximum 4.48 2.68
4.48
25% Quartile 0.21 0.13
0.17
Median 0.38 0.24
0.32
75% Quartile 0.68 0.40
0.56
Mean 0.50 0.31
0.43
Standard deviation 0.42 0.27
0.39
Note:
1
for 952 events, and
2
for 548 events.
Table 3.2 Statistical Summary of WatershedMedian, WatershedMean and Standard
deviation Values of C
rate
for 80 Texas Watersheds.
WatershedMedian C
rate
WatershedMean C
rate
Standard deviation
Minimum 0.07 0.07 0.03
Maximum 1.73 1.79 0.87
25 % Quartile 0.20 0.27 0.16
Median 0.31 0.36 0.22
75 % Quartile 0.55 0.56 0.37
Mean 0.40 0.44 0.27
Standard deviation 0.29 0.27 0.15
Table 3.3 Statistical Summary of WatershedMedian C
rate
for Developed and
Undeveloped Watersheds.
Undeveloped
1
Developed
2
Minimum 0.07 0.17
Maximum 73 1.73
25%Quartile 0.13 0.30
Median 0.20 0.40
75%Quartile 0.28 0.71
Mean 0.24 0.53
Standard deviation 0.16 0.31
Note:
1
for 36 undeveloped watersheds, and
2
for 44 developed watersheds.
81
Table 3.4 Statistical Summary of C
r
, and Differences among C
r
, C
rate
and C
lit
for 80
Texas watersheds.
C
r
C
r
 C
rate
1
C
r
 C
lit
C
rate
1
 C
lit
Minimum 0.09 0.53 0.44 0.46
Maximum 1.20 0.52 0.66 1.19
25 % Quartile 0.25 0.02 0.17 0.24
Median 0.37 0.03 0.11 0.14
75 % Quartile 0.54 0.06 0.02 0.01
Mean 0.42 0.02 0.06 0.07
Standard deviation 0.23 0.14 0.21 0.27
1
is watershedmedian C
rate
.
Table 3.5 Statistical Summary of C
r
and C
lit
for Developed and Undeveloped Watersheds.
C
r
Undeveloped
watersheds
C
r
Developed
watersheds
C
lit
Undeveloped
watersheds
C
lit
Developed
watersheds
Minimum 0.09 0.18 0.29 0.37
Maximum 0.69 1.20 0.59 0.63
25%Quartile 0.20 0.34 0.33 0.52
Median 0.26 0.45 0.37 0.54
75%Quartile 0.42 0.62 0.44 0.58
Mean 0.32 0.50 0.39 0.54
Standard deviation 0.17 0.23 0.08 0.06
82
Fig. 3.1 Map showing U.S. Geological Survey streamflowgaging stations representing
80 developed and undeveloped watersheds in Texas.
83
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
C
r
a
t
e
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
C
r
a
t
e
0
20
40
60
80
100
120
I
(
m
m
/
h
r
)
20
40
60
80
100
120
I
(
m
m
/
h
r
)
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
t
av
(hr)
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
C
rate
I(mm/hr)
U.S. Geological Survey streamflowgaging station
08058000 Honey Creek subwatershed number 12
near McKinney, Texas
(Area = 3.11 km )
U.S. Geological Survey streamflowgaging station
08048550 Dry Branch at Blandin Street, Fort Worth, Texas
(Area = 2.87 km )
T
c
= 1.8 hr., C
rate
= 0.23
T
c
= 1.5 hr., C
rate
= 1.7
(A)
(B)
2
2
Fig 3.2 Average rainfall intensity I and runoff coefficient C
rate
as a function of t
av
for two
storm events: (A) on 09/22/1969 in Dry Branch in Fort Worth, Texas (U.S. Geological
Survey [USGS] streamflowgaging station 08048550 Dry Branch at Blandin Street, Fort
Worth Texas), and (B) on 04/25/1970 in Honey Creek near Dallas, Texas. (USGS
streamflowgaging station 08058000 Honey Creek subwatershed number 12 near
McKinney, Texas).
84
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
P
e
r
c
e
n
t
a
g
e
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
P
e
r
c
e
n
t
a
g
e
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Watershedmedian,
0.01 0.1 1
Runoff Coefficient,
Undeveloped (36 Texas watersheds)
Developed (44 Texas watersheds)
All 80 Texas watersheds
T < T
T >= T
watershedmean
watershedmedian
C
rate
C
rate
(A)
(B)
p
c
p c
Fig. 3.3 Cumulative distributions of runoff coefficient (C
rate
values): (A) for rainfall
runoff events when the time to peak (T
p
) was less than the time of concentration (T
c
); for
rainfallrunoff events when the time to peak (T
p
) was greater than or equal to the time of
concentration (T
c
); watershedaverage (mean); and watershedmedian; and (B)
watershedmedian C
rate
for developed and undeveloped Texas watersheds.
85
0
2
4
6
8
10
12
Q
p
j
*
(
m
3
/
s
/
h
a
)
0.00
0.05
0.10
0.15
0.20
0.25
C
r
j
0 10 20 30 40 50
I
j
(mm/hr)
0 10 20 30 40 50
C
r
Q
pj*
= 0.19 x I
j
R
2
= 0.88
(A)
(B)
Fig. 3.4 (A) Runoff coefficient (C
r
) values derived from the rankordered pairs of
observed peak discharge and maximum rainfall intensity during each storm event in
mm/hr at USGS streamflowgaging station 08042650 North Creek Surface Water Station
28A near Jermyn, Texas, and (B) the rankordered pairs of the observed peak discharge
and the average rainfall intensity for the same station.
86
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
P
e
r
c
e
n
t
a
g
e
0
10
20
30
40
50
60
70
80
90
100
C
u
m
u
l
a
t
i
v
e
P
e
r
c
e
n
t
a
g
e
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
Runoff Coefficient
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
C
r
for 36 undeveloped Texas watersheds
C
r
for 44 developed Texas watersheds
C
lit
for undeveloped watersheds
C
lit
for developed watersheds
C
r
for 80 Texas watersheds
watershedmedian C for 80 Texas watersheds
C
lit
for undeveloped and developed watersheds
(A)
(B)
rate
Fig. 3.5 Cumulative distributions of: (A) C
r
, watershedmedian C
rate
and C
lit
, and (B)
distributions of C
r
and C
lit
for developed and undeveloped watersheds.
87
0.0
0.2
0.4
0.6
0.8
1.0
1.2
R
u
n
o
f
f
C
o
e
f
f
i
c
i
e
n
t
0 10 20 30 40 50 60 70 80 90 100
Percentage of impervious area,
Equation 3.5 (Asquith 2011)
C
r
for 45 Texas watersheds with watershed imperviousness greater than 10%
C values extracted from Jens (1979) for comparison with Texas C values
C
r
IMP
estimated from the equation proposed by Schaake et. al (1967) for 45 Texas watersheds
with watershed imperviousness greater than 10%
EXPLANATION
r
Fig. 3.6 Runoff coefficients versus the percentage of impervious area, IMP.
88
1
10
100
M
o
d
e
l
e
d
Q
p
(
m
3
/
s
)
1 10 100
Observed Q
p
(m
3
/s)
modeled using C
modeled using C
1:1 line
? half of log cycle
800
800
rate
r
Fig. 3.7 Modeled peak discharges (Q
p
) from rational equation (3.1) using C
r
and
watershedmedian C
rate
for 1,500 rainfallrunoff events in 80 Texas watersheds against
observed peak discharges.
0.01
0.1
1
R
u
n
o
f
f
C
o
e
f
f
i
c
i
e
n
t
1 10 100
Area (km
2
)
C
r
watershedmean C
C
rate
(all events)
rate
Fig 3.8. Ratebased runoff coefficients C
rate
for all events, watershedmean C
rate
, and C
r
versus drainage area (km
2
) in 80 Texas watersheds.
89
Chapter 4. Return Period Adjustments for Runoff Coefficients Based on Analysis in
Texas Watersheds
4.1 Abstract
The rational method for peak discharge (Q
p
) estimation was introduced in the
1880s. The runoff coefficient (C) is a key parameter for the rational method, and the C
has been declared a function of return period by various researchers. Ratebased runoff
coefficients as function of return period, C(T), were developed for 36 undeveloped
watersheds in Texas using peak discharge frequency from previously published regional
regression equations and rainfall intensity frequency for return periods T of 2, 5, 10, 25,
50, and 100 years. The C(T) values developed in this study are most applicable to
undeveloped watersheds. The C(T) values of this study increase with T more rapidly than
the increase suggested in prior literature. When the larger frequency factors are applied, if
any resulting C(T) is greater than unity, C(T) is suggested to be set to 1.
4.2 Introduction
The rational method introduced by Kuichling (1889) is typically used to compute
the peak discharge, Q
p
(in m
3
/s in SI units or ft
3
/s in English units) for designing drainage
structures:
,CIAmQ
op
?
(4.1)
where C is the runoff coefficient (dimensionless), I is the rainfall intensity (mm/hr or
in/hr) over a critical period of storm time (typically taken as the time of concentration, T
c
,
90
of the watershed), A is the drainage area (hectares or acres), and m
o
is the dimensional
correction factor (1/360 = 0.00278 in SI units, 1.008 in English units). From inspection of
the equation, it is evident that C is an expression of rate proportionality between I and Q
p
.
Typical ?whole watershed? C values, that is, C values representing the integrated
effects of various surfaces in the watershed and other watershed properties, are listed for
different general landuse conditions in various design manuals and textbooks. Examples
of textbooks that include tables of C values are Chow et al. (1988) and Viessman and
Lewis (2003). Published C values, C
lit
, were sourced from American Society of Civil
Engineers (ASCE) and the Water Pollution Control Federation (WPCF) in 1960 (ASCE
and WPCF 1960). The C
lit
values were obtained from a response survey, which received
?71 returns of an extensive questionnaire submitted to 380 public and private
organizations throughout the United States.? No justification based on observed rainfall
and runoff data for the selected C
lit
values was provided in the ASCE and WPCF (1960)
manual.
A substantial criticism of the rational method arises because observed C values
vary from storm to storm (Schaake et al. 1967; Pilgrim and Cordery 1993). The ASCE
and WPCF (1960) manual, in describing tabulations of rational method C, state ?The
coefficients on these two tabulations (of C values) are applicable for storms having 5 to
10year return periods (0.2 to 0.1 annual exceedance probabilities). Less frequent, higher
intensity storms will require the use of higher coefficients because infiltration and other
losses have a proportionately smaller effect on runoff.? Schaake et al. (1967) found that
the average percentage increase of the coefficient for the 10year return period C(10) was
only 10 percent as compared to the coefficient for the 1year return period C(1), and
91
proposed adoption of a single value of C for design circumstances. The C has been
considered a function of return period by various researchers (Jens 1979; Pilgrim and
Cordery 1993; Hotchkiss and Provaznik 1995; Titmarsh et al. 1995; Young et al. 2009).
Considering C as function of return period T, the rational formula can be expressed as
(Jens 1979; Pilgrim and Cordery 1993):
Q(T) = C(T) I(T) A = C
lit
C
f
(T) I(T) A, (4.2)
where C(T) is C as function of the return period T, I(T) is rainfall intensity as function of
T, C
lit
is literaturebased C as previously defined herein based on T values ranged from 2
to 10 year recurrence intervals from text books (e.g., Viessman and Lewis 2003) or
design manuals (e.g., TxDOT 2002), and C
f
(T) is a frequency factor or multiplier (DRCG
1969; Jens 1979). Equation 4.2 implies a conversion of I(T) to Q
p
(T) where T denotes the
same return period for both I and Q. Equation (4.2) also is the probabilistic interpretation
of the rational formula and commonly used in design practices (French et. al 1974;
Pilgrim and Cordery 1993).
Relatively, few studies have been performed to determine rational C values using
frequencybased analysis of data (Young et al. 2009). An influential paper by Schaake et
al. (1967) examined the rational method using experimental rainfall and runoff data
collected from 20 small urban watersheds [< 0.6 km
2
(0.23 mi
2
)] in Baltimore, MD.
Those authors used watershed lag time to compute average rainfall intensity and used a
frequencymatching approach to compute ratebased C values. Hotchkiss and Provaznik
(1995) estimated C
rate
for 24 rural watersheds in southcentral Nebraska using event
paired and frequencymatched data. Young et al. (2009) estimated C
rate
for 72 rural
watersheds in Kansas with drainage areas less than 78 km
2
(30 mi
2
) for different return
92
periods. The peak discharge Q(T) for each T was estimated using annual peak frequency
analysis of observed streamflow records and rainfall intensity obtained from rainfall
intensitydurationfrequency tables (Young et al. 2009).
For this study, returnperiod based runoff coefficients C(T) were computed using
Q
p
and I calculated, not from statistical analysis of observed pairing of rainfall and runoff
data, but from coupling of regional regression equations (Asquith and Slade 1997) and
rainfall intensity (TxDOT 2002) for 36 undeveloped Texas watersheds. Subsequently,
frequency factors C
f
(T) = C(T)/C(10) were computed. Using C for the 10year return
period as a base value to compute frequency factors C
f
(T) is consistent with literature
(French et. al 1974; Pilgrim and Cordery 1993; Young et al. 2009). Results of C(T) and
frequency factors C
f
(T) were analyzed and compared with previous studies.
4.3 Study Watersheds
The study watersheds comprise 36 undeveloped watersheds in Texas, which have
been used previously by the authors and associates (Asquith et al. 2004). The 36
watersheds consist of 20 rural watersheds and 16 suburban watersheds in near four cities:
Austin, Dallas, Fort Worth, and San Antonio. Locations and geographic distribution of
the streamflowgaging stations associated with these watersheds are shown in Figure 4.1.
The classification scheme of developed and undeveloped watersheds
accommodates the characterization of watersheds in more than 220 USGS reports of
Texas data from which the original data for the rainfall and runoff database were obtained
(Asquith et al. 2004). Although this binary classification seems arbitrary, it was
purposeful and reflects the uncertainty in precise watershed development conditions for
93
the time period of available data (Asquith and Roussel 2007). This same binary
classification was successfully used to prepare regression equations to estimate the shape
parameter and the time to peak for regional gamma unit hydrographs for Texas
watersheds (Asquith et al. 2006).
The drainage area of study watersheds range from approximately 2.3?320 km
2
(0.9?123.6 mi
2
); the median and mean values are 20.7 km
2
(8 mi
2
) and 56.7 km
2
(21.9
mi
2
), respectively. The stream slopes of study watersheds range from approximately
0.0022?0.0196 dimensionless; the median and mean values are both 0.0089, respectively.
Many practitioners could argue that the application of the rational method is not
appropriate for the range of watershed areas presented in this study. ASCE and WPCF
(1960) made the following statements when the rational method was introduced for
design and construction of sanitary and storm sewers: ?Although the basic principles of
the rational method are applicable to large drainage areas, reported practice generally
limits its use to urban areas of less than 5 sq miles.? (ASCE and WPCF 1960, p. 32).
Pilgrim and Cordery (1993) stated that the rational method is one of the three methods
widely used to estimate peak flows for small to medium sized basins. ?It is not possible
to define precisely what is meant by ?small? and ?medium? sized, but upper limits of 25
km
2
(10 mi
2
) and 550 km
2
(200 mi
2
), respectively, can be considered as general guides?
(Pilgrim and Cordery 1993, p. 9.14). Young et al. (2009) stated that the rational method
may be applied to much larger drainage areas than typically assumed in some design
manuals provided that the watershed is unregulated. Thompson (2006) stated that
watershed drainage area does not appear to be an applicable factor for discriminating
between appropriate hydrologic technologies (such as rational method, regional
94
regression equations, and sitespecific flood frequency relations), other methods for
discrimination between procedures for making designdischarge estimates should be
investigated.
A geospatial database of properties for the 36 watersheds was developed by
Roussel et al. (2005). For this paper, basinshape factor, main channel length and channel
slope were used to estimate time of concentration T
c
by Kirpich (1940) for channel flow
plus travel time for overland flow using Kerby (1959). This combination of methods to
compute T
c
is discussed by Roussel et al. (2005) and Fang et al. (2008). The Kirpich
equation (1940) was developed from the Natural Resources Conservation Service
(NRCS) data for rural watersheds with drainage areas less than about 0.45 km
2
.
Fang et
al. (2007, 2008) demonstrated that, for watersheds with large drainage areas, the Kirpich
equation provides as reliable an estimate of T
c
as the other empirical equations developed
for large watersheds and as the NRCS velocity method (Viessman and Lewis 2003). The
T
c
estimated using the Kirpich equation reasonably approximate the average T
c
estimated
from observed rainfall and runoff data (Fang et al. 2007).
4.4 Runoff Coefficients for Different Return Periods
Ratebased C(T) values for the 36 study watersheds in Texas, and corresponding
frequency factors were determined for various return periods using equation (4.3)
(Pilgrim and Cordery 1993):
,
),(
)(
)(
0
ATTIm
TQ
TC
c
?
(4.3)
where Q(T), C(T), and I(T
c
, T) are peak discharge, runoff coefficient, and rainfall
intensity for the recurrence interval T, respectively. In this study, Q(T) for each of the 36
95
undeveloped Texas watersheds was estimated by regional regression equations for Texas
developed by Asquith and Slade (1997) that use contributing drainage area, a basinshape
factor, and main channel slope. The basinshape factor is defined as the ratio of main
channel length squared to contributing drainage area (sq. mi./sq. mi. or km
2
/km
2
) (TxDOT
2002).
The T
c
for each watershed in Texas was developed using the KerbyKirpich
equation (Roussel et al. 2005; Fang et al. 2008). Considering the county in which each
watershed is located, the rainfall intensity, I(T
c
, T), for each return period was estimated
using rainfall intensitydurationfrequency (IDF) relations (TxDOT 2002) with duration
T
c
:
,
)(
),(
g
c
c
fT
e
TTI
?
?
(4.4)
where e, f, and g are coefficients for specific frequencies and Texas counties (TxDOT
2002). With Q(T) from Asquith and Slade (1997) and I(T
c
, T) from TxDOT design
manual (TxDOT 2002), equation (4.3) was used to compute C(T) for each watershed and
for each return period of T = 2, 5, 10, 25, 50 and 100 years.
The C(T) versus T for three undeveloped Texas watersheds are presented as
illustrative examples in Figure 4.2. For these three watersheds, C(T) increases with
increasing T. The value of C(100) is 0.6 for Deep Creek, 1.05 for East Elm Creek, and
1.3 for Escondido Creek. The occurrence of C(T) > 1 could be related to inherent
uncertainties of Q(T) and I(T
c
, T). Several studies (French et. al 1974; Pilgrim and
Cordery 1993; Young et al. 2009) have shown that values of C(T) greater than 1 are
possible when ratebased C was determined from observed peak discharge and rainfall
intensity. So it is not a stretch to see C(T) > 1 for purely statistically independent studies
96
of Q and I such as Asquith and Slade (1997) and TxDOT (2002). Analysis of observed
rainfall and runoff data in 90 Texas watersheds has shown that only the volumetric runoff
coefficient, C
v
, as the ratio of total runoff depth to total rainfall depth, is less than 1 for all
storm events (Dhakal et al. 2012).
Statistical summaries of C(T) are listed in Table 4.1 and corresponding boxplots
of the distribution are shown in Figure 4.3. The median C values by T, as well as the
curves shown in Figure 4.2, show that C(T) increases with the increasing recurrence
interval for undeveloped watersheds in Texas. Ratios of C(T)/C(10) or frequency factors
C
f
(T) are derived for the Texas watersheds and statistical summaries of the ratios are
listed in Table 4.2; the mean and median values are of special importance for
representation of frequency.
4.5 Discussion
Comparing C(T) and C
f
(T) for Texas Watersheds with Other Studies
Young et al. (2009) estimated median C(T) from observed data for 72 rural
watersheds in Kansas and these values are shown in Figure 4.3 for comparison. The
results of C(T) for Texas watersheds reported in Table 4.1 and Figure 4.3 are consistent
with results reported by Young et al. (2009). The mean values of C(T) derived from
observed data for 24 rural watersheds in southcentral Nebraska (Hotchkiss and
Provaznik 1995) are shown in Figure 4.3. Literaturebased C values for the Nebraska
watersheds are 0.35 for T < 10 years (Hotchkiss and Provaznik 1995). The mean C(T)
values reported by Hotchkiss and Provaznik (1995) for the Nebraska watersheds are
97
larger than not only median C(T) determined for the Texas and Kansas watersheds but
also C
lit
.
French et al. (1974) mapped C(10) values in New South Wales, Australia, for 37
rural watersheds, with drainage area up to 250 km
2
(96 mi
2
). The relations between
frequency factors and return period from French et al. (1974), reported by Young et al.
(2009), Jens (1979) and Gupta (1989), and the results for the Texas watersheds are shown
in Figure 4.4. The frequency factors C
f
(T) determined (1) for the Texas watersheds, (2)
by Young et al. (2009), and (3) by French et al. (1974) exceed the textbook values from
Gupta (1989) and Viessman and Lewis (2003), and exceed TxDOT (2002) values when T
> 10 years. The Texas frequency factors C
f
(T) are similar to those determined for Kansas
watersheds by Young et al. (2009). Lastly, the Texas frequency factors C
f
(T) exceed
those from watersheds in New South Wales, Australia (French et al. 1974) by about 15
percent when T > 10 years, and less when T < 10 years.
The frequency factors C
f
(T) specified for Denver watersheds (DRCG 1969; Jens
1979) and later published in other textbooks (e.g., Gupta 1989, Viessman and Lewis
2003) and design manuals (e.g., TxDOT 2002) are listed in Table 4.3. Typically, a
frequency factor C
f
(T) of 1.0 is used when T < 10 years (Table 3). The frequency factors
C
f
(T) extracted from the FHWA curve (Jens 1979) for percent impervious area equal to 0
percent and 65 percent and from the study by French et al. (1974) and Young et al.
(2009) are listed in Table 4.3 for comparison. Of special note is the observation that the
Texas frequency factors, C(2)/C(10) and C(5)/C(10) as well as those from French et al.
(1974) and Young et al. (2009), are not equal to 1.0 as in ASCE and WPCF (1960) [and
later by Gupta (1989) and Viessman and Lewis (2003) and in the TxDOT (2002)] but
98
rather are ratios less than 1.0 (Table 4.2 and 4.3). This means that the C(T) for T < 10
year (more frequent storms) of Texas, Kansas and Australia is less than C
lit
commonly
recommended in the literature.
The frequency factors C
f
(T) values were extracted from the FHWA curve (Jens
1979) for percent impervious areas of 65 percent because they are approximately the
same as frequency factors C
f
(T) values presented in design manuals and textbooks (e.g.,
TxDOT design manual [2002]; Gupta 1989; Viessman and Lewis 2003). The frequency
factors C
f
(T) presented in design manuals and textbooks are seemingly more appropriate
for urban watersheds with relatively large percentage of impervious area.
The larger frequency factors C
f
(T) determined for Texas watersheds and those
determined for Kansas watersheds (Young et al. 2009) are for undeveloped watersheds
with impervious cover less than a few percent. The larger frequency factors C
f
(T) of
Texas are similar to frequency factors C
f
(T) extracted from the FHWA curve (Jens 1979)
for 0 percent impervious areas (Table 4.3). These frequency factors C
f
(T) were proposed
by Bernard (1938). The frequency factors C
f
(T) from the FHWA curve (Jens 1979) for
100 percent impervious area is approximately 1.1 for T of 25, 50, and 100 years. If it is
assumed that C is 1 for 100 percent impervious areas, then frequency factors C
f
(T) should
be 1.0 for 100 percent impervious area for any T. Therefore, variable frequency factors
C
f
(T) as additional function of percent of impervious area (Jens 1979) is a reasonable
conjecture and supported by Young et al. (2009) as well as this study for Texas
watersheds. When frequency factors C
f
(T) is applied and if resulting C(T) is greater than
1, Jens (1979), Gupta (1989), and TxDOT (2002) indicate C(T) should be set equal to 1.
99
C for 100year Return Period
In an adaption of the rational method, Bernard (1938) proposed that C varied in a
functional manner with the Tyear return period when related to the maximum or limiting
C values (called C
max
):
,)100(
max
x
TCC ? (4.5)
where x is the exponent and ranges from 0.15 to 0.23 for undeveloped watersheds
(Bernard 1938). Bernard (1938) assumed the C
max
value corresponds to C(100). In
relation to equation (4.5), Jens (1979) proposed C
max
= C(100) = 1.0 for watersheds with
any percentage of impervious area for application of the equation (4.5) for the FHWA
Manual (Jens 1979). C(100) for the 36 Texas watersheds range from 0.34 to 1.44 with
mean and median values of 0.86 and 0.94 (Table 4.1). C(100) for three Texas watersheds
also are presented as illustrative examples shown in Figure 4.2. Stubchaer (1975) applied
the calibrated Santa Barbara Unit Hydrograph (SBUH) method on a 388acre urban
watershed and developed C(T) using the frequency analysis of rainfall and simulated
runoff from the SBUH. The C(100) value determined for the watershed is 0.65
(Stubchaer 1975). The C(100) values for watersheds with different percentages of
impervious cover from the Denver Manual (DRCG 1969) range from 0.20 to 0.96 and
from Chow et al. (1988) range from 0.36 to 0.97; C(100) are consistently less than 1.
4.6 Summary
The runoff coefficients C(T) for different return periods (T) were developed for
the 36 undeveloped Texas watersheds using previously published regional regression
100
equations of peak discharge and countybased tabulated empirical coefficients for a
model of rainfall intensities at different T. C(T) values increase with T and these increases
are more than previously thought. The frequency factors C
f
(T) = C(T)/C(10) determined
in this study exceed those values in textbooks such as Gupta (1989) and Viessman and
Lewis (2003) and those from TxDOT (2002) when T > 10 years. The frequency factors
C(2)/C(10) and C(5)/C(10) for the Texas watersheds (Table 4.2) as well as from French
et al. (1974) and Young et al. (2009) are not equal to 1 as assumed in ASCE and WPCF
(1960) and published by Gupta (1989) and Viessman and Lewis (2003) and the design
manual (TxDOT 2002) (Table 4.3 and Figure 4.4) but less than 1.
The frequency factors determined for the 36 Texas watersheds and the 72 Kansas
watersheds (Young et al. 2009), larger than those mostly found in literature, are for
undeveloped watersheds with relatively small percent impervious areas. The frequency
factors mostly found in the literature, smaller than those determined for the 36 Texas
watersheds, are appropriate for urban watersheds with relatively large percentages of
impervious area, as supported and presented in literature (e.g., DRCG 1969; Stubchaer
1975; Jens 1979; Gupta 1989; Viessman and Lewis 2003; TxDOT 2002). Such frequency
factors are consistent with those proposed by Jens (1979). When the frequency factor is
applied, if resulted C(T) is greater than unity, Jens (1979), Gupta (1989) and TxDOT
(2002) suggested setting C(T) equal to 1.
4.7 Acknowledgments
101
The authors thank TxDOT project director Mr. Chuck Stead, P.E., and project
monitoring advisor members for their guidance and assistance. This study was partially
supported by TxDOT Research Projects 0?6070, 0?4696, 0?4193, and 0?4194.
4.8 Notation
The following symbols are used in this paper:
A = watershed area in hectares or acres;
C
f
(T) = C(T)/C(10), frequency factor or frequency multiplier;
C
max
= maximum runoff coefficient for the return period 100 years;
C(T) = ratebased runoff coefficient for return period T;
C
v
= volumetric runoff coefficient as the ratio of total runoff depth and total rainfall
depth;
I = average rainfall intensity (mm/hr or in. /hr) with the duration equal to time of
concentration;
m
o
= the dimensional correction factor (1.008 in English units, 1/360 = 0.00278 in SI
units);
Q
p
= peak runoff rate in m
3
/s or ft
3
/s;
Q(T) = peak discharge for return period T;
Q
T
= regional regression equation for natural basins developed for TxDOT;
T = recurrence interval or return period in years;
T
c
= time of concentration;
102
4.9 References:
ASCE, and WPCF. (1960). Design and construction of sanitary and storm sewers,
American Society of Civil Engineers (ASCE) and Water Pollution Control Federation
(WPCF).
Asquith, W. H., and Slade, R. M. (1997). "Regional equations for estimation of peak
streamflow frequency for natural basins in Texas." Water Resources Investigations
Report 964307, U.S. Geological Survey, Austin, Texas, 68 p.
Asquith, W. H., Thompson, D. B., Cleveland, T. G., and Fang, X. (2004). ?Synthesis of
rainfall and runoff data used for Texas department of transportation research projects
0?4193 and 0?4194.? U.S. Geological Survey, OpenFile Report 2004?1035, Austin,
Texas.
Asquith, W. H., Thompson, D. B., Cleveland, T. G., and Fang, X. (2006). ?Unit
hydrograph estimation for applicable Texas watersheds.? Texas Department of
Transportation Research Report 041934, Austin, Texas.
Asquith, W. H., and Roussel, M. C. (2007). ?An initialabstraction, constantloss model
for unit hydrograph modeling for applicable watersheds in Texas.? U.S. Geological
Survey Scientific Investigations Report 2007?5243, Austin, Texas, 82 p.
Bernard, M. (1938). ?Modified rational method of estimating flood flows.? Natural
Resources Commission, Washington D.C.
Chow, V. T., Maidment, D. R., and Mays, L. W. (1988). Applied hydrology, McGraw
Hill New York.
Dhakal, N., Fang, X., Cleveland, T.G., Thompson, D.B., Asquith, W.H., and Marzen, L.J.
(2012). ?Estimation of volumetric runoff coefficients for Texas watersheds using
landuse and rainfallrunoff data.? Journal of Irrigation and Drainage Engineering,
138(1):4354.
DRCG. (1969). ?Urban Storm Drainage Criteria Manual, Vols. 1 and 2.? Prepared by
WrightMcLaughlin Engineers for Denver Regional Council of Governments
(DRCG), Denver, CO.
Fang, X., Thompson, D. B., Cleveland, T. G., and Pradhan, P. (2007). ?Variations of time
of concentration estimates using NRCS velocity method.? Journal of Irrigation and
Drainage Engineering, 133(4), pp. 314?322.
Fang, X., Thompson, D. B., Cleveland, T. G., Pradhan, P., and Malla, R. (2008). ?Time
of concentration estimated using watershed parameters determined by automated and
manual methods.? Journal of Irrigation and Drainage Engineering, 134(2), pp. 202?
211.
103
French, R., Pilgrim, D. H., and Laurenson, B. E. (1974). ?Experimental examination of
the rational method for small rural catchments.? Civil Engineering Transactions, The
Institution Of Engineers, Australia CE16(2), pp. 95?102.
Gupta, R. S. (1989). Hydrology and hydraulic systems, Prentice Hall, Englewood Cliffs,
New Jersey.
Hotchkiss, R. H., and Provaznik, M. K. (1995). ?Observations on the rational method C
value.? Watershed management: planning for the 21st century; proceedings of the
symposium sponsored by the Watershed Management Committee of the Water
Resources, New York.
Jens, S. W. (1979). Design of Urban Highway Drainage, Federal Highway
Administration (FHWA), Washington DC.
Kerby, W. S. (1959). ?Time of concentration for overland flow.? Civil Engineering,
29(3), 174 p.
Kirpich, Z. P. (1940). "Time of concentration of small agricultural watersheds." Civil
Engineering, 10(6), 362.
Kuichling, E. (1889). ?The relation between the rainfall and the discharge of sewers in
populous areas.? Transactions, American Society of Civil Engineers 20, pp. 1?56.
Pilgrim, D. H., and Cordery, I. (1993). ?Flood runoff.? In Handbook of Hydrology, D. R.
Maidment, ed., McGrawHill, New York, 9.1?9.42.
Roussel, M. C., Thompson, D. B., Fang, D. X., Cleveland, T. G., and Garcia, A. C.
(2005). ?Timing parameter estimation for applicable Texas watersheds.? 0?4696?2,
Texas Department of Transportation, Austin, Texas.
Schaake, J. C., Geyer, J. C., and Knapp, J. W. (1967). ?Experimental examination of the
rational method.? Journal of the Hydraulics Division, ASCE, 93(6), pp. 353?370.
Stubchaer, J. M. ?The Santa Barbara urban hydrograph method.? National Symposium of
Urban Hydrology and Sediment Control, University of Kentucky, Lexington, KY.
Thompson, D. B. (2006). "The rational method, regional regression equations, and site
specific flood frequency relations." FHWA/TX06/044051, Center for
Multidisciplinary Research in Transportation, Texas Tech University, Lubbock,
Texas.
Titmarsh, G. W., Cordery, I., and Pilgrim, D. H. (1995). ?Calibration Procedures for
Rational and USSCS Design Flood Methods.? Journal of Hydraulic Engineering,
121(1), pp. 61?70.
104
TxDOT. (2002). Hydraulic design manual, The bridge division of the Texas Department
of Transportation (TxDOT), 125 East 11th street, Austin, Texas 78701.
Viessman, W., and Lewis, G. L. (2003). Introduction to hydrology, 5th Ed., Pearson
Education, Upper Saddle, N.J., 612 p.
Young, C. B., McEnroe, B. M., and Rome, A. C. (2009). ?Empirical determination of
rational method runoff coefficients.? Journal of Hydrologic Engineering, 14, 1283 p.
105
Table 4.1 Statistical Summary of C(T) for Select Return Periods (T) for Texas
Watersheds.
2 years 5 years 10 years 25 years 50 years 100 years
Minimum 0.08 0.14 0.18 0.24 0.30 0.34
Maximum 0.39 0.64 0.77 0.97 1.14 1.44
25
th
percentile 0.12 0.29 0.39 0.48 0.50 0.55
Median 0.15 0.32 0.43 0.62 0.77 0.94
75
th
percentile 0.22 0.37 0.50 0.70 0.91 1.12
Average 0.18 0.33 0.44 0.59 0.73 0.86
Standard deviation 0.075 0.091 0.110 0.163 0.238 0.319
Table 4.2 Statistical Summary of the Frequency Factors C(T)/C(10) for Texas
Watersheds.
C(2)/C(10) C(5)/C(10) C(25)/C(10) C(50)/C(10) C(100)/C(10)
Minimum 0.25 0.63 1.10
1.14
1.16
Maximum 0.69 0.89 1.55
2.45
3.14
25
th
percentile 0.30 0.68 1.23
1.34
1.47
Median 0.35 0.72 1.42
1.66
1.93
75
th
percentile 0.51 0.85 1.49
2.06
2.54
Average 0.41 0.75 1.36
1.68
1.98
Standard deviation 0.14 0.09 0.14
0.39
0.578
Table 4.3 Frequency Factor or Multiplier for Literaturebased Rational Runoff
Coefficient C from Different Sources.
Return period,
T, years
Frequency factor, C
f
(T), C(T)/C(10)
Gupta (1989)
1
0 % IMP
2
65 % IMP
2
Young et al. (2009) Texas watersheds
2 1.0 0.48 0.69 0.45 0.41
5 1.0 0.77 0.87 0.77 0.75
10 1 1 1 1 1
25 1.1 1.22 1.15 1.30 1.36
50 1.2 1.40 1.22 1.54 1.68
100 1.25 1.60 1.30 1.77 1.98
1
C
f
(T) from the Denver material (DRCG 1969; Jens 1979) and later published in other
textbooks (e.g., Gupta 1989, Viessman and Lewis 2003) and design manuals (e.g.,
TxDOT 2002)
2
From Jens (1979)
106
Fig. 4.1 Map showing the locations of U.S. Geological Survey streamflowgaging
stations in Texas associated with the 36 undeveloped watersheds considered for this study
(two stations are very close and overlapped each other).
107
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
R
u
n
o
f
f
C
o
e
f
f
i
c
i
e
n
t
,
C
(
T
)
0 10 20 30 40 50 60 70 80 90 100
Return Period, T(years)
08139000 Deep Creek Sub. 3 near Placid, Texas (Area = 8.86 km )
08178465 East Elm Creek at San Antonio, Texas (Area = 6.03 km )
08187000 Escondido Creek Sub. 1 near Kenedy, Texas (Area = 8.52 km )
U.S. Geological Survey streamflowgaging stations
2
2
2
Fig. 4.2 Estimation of C(T) versus T for three undeveloped Texas watersheds.
108
Fig. 4.3 Box plot for the distribution of runoff coefficients for different return periods
from Texas watersheds, from Hotchkiss and Provaznik (1995) and from Young et al.
(2009).
109
Fig. 4.4 Frequency factors for different return periods from Texas watersheds, Young et
al. (2009), French et al. (1974) and FHWA (Jens 1979) and Gupta (1989).
110
Chapter 5. Modified Rational Unit Hydrograph Method and Applications in Texas
Watersheds
5.1 Abstract
The modified rational method (MRM) is an extension of the rational method to
develop simple runoff hydrographs. The hydrographs developed using the MRM can be
considered application of a special unit hydrograph (UH) that is termed the modified
rational unit hydrograph (MRUH) in this paper. Being a UH, the MRUH can be applied
to nonuniform rainfall distributions and for watersheds with drainage areas greater than
typically used for the rational method (a few hundred acres). The MRUH was applied to
90 watersheds in Texas using 1,600 rainfallrunoff events. Application of the MRUH
involved three steps: (1) determination of rainfall excess using the runoff coefficient, (2)
determination of the MRUH using drainage area and time of concentration, and (3)
applying the unit hydrograph convolution. Times of concentration for the study
watersheds were estimated using four empirical equations. Runoff coefficients used were
estimated using two methods (literaturebased from landuse information and back
computed from observed rainfall and runoff data). The Gamma UH, ClarkHEC1 UH,
and NRCS dimensionless UH were also used to predict peak discharges of all events in
the database. The MRUH performed about as well as these UH methods when the same
rainfall loss model was used.
5.2 Introduction
111
The rational method was originally developed for estimating peak discharge, Q
p
,
for sizing drainage structures, such as storm drains and culverts. The peak discharge, Q
p
,
(in m
3
/s in SI units or ft
3
/s in English units) is computed using:
,CIAmQ
op
?
(5.1)
where C is the runoff coefficient (dimensionless), I is the rainfall intensity (mm/hr or
in./hr) over a critical period of storm time (the time of concentration, T
c
), A is the
drainage area (hectares or acres), and m
o
is the dimensional correction factor (1/360 =
0.00278 in SI units, 1.008 in English units). Kuichling (1889) and Llyod?Davies (1906)
are credited with independent development of the rational method (Singh and Cruise
1992). Texas Department of Transportation (TxDOT) guidelines for drainage design
recommend use of the rational method for watersheds with drainage areas less than 0.8
km
2
or 200 acres (TxDOT 2002).
Incorporation of detention basins to mitigate effects of urbanization on peak flows
requires design methods to include the volume of runoff as well as the peak discharge
(Rossmiller 1980). To use the rational method for hydraulic structures involving storage,
the modified rational method (MRM) was developed (Poertner 1974). The term
?modified rational method analysis" refers to ?a procedure for manipulating the basic
rational method techniques to reflect the fact that storms with durations greater than the
normal time of concentration for a basin will result in a larger volume of runoff even
though the peak discharge is reduced" [p. 54 Poertner (1974)]. Emil Kuichling (1889)
stated: ?in drainage areas of moderate size, the heaviest discharge always occurs when
the rain lasts long enough at its maximum intensity to enable all portions of the area to
contribute to the flow.? The MRM is based on the same assumptions as the conventional
112
rational method and is a conceptual extension of the rational method (Viessman and
Lewis 2003).
The MRM was revisited and reevaluated in this study. The hydrographs
developed using the MRM represent application of a special unit hydrograph (UH) that
will be termed the modified rational unit hydrograph (MRUH) in this study. The MRUH
and unit hydrograph convolution were used to compute the direct runoff hydrographs for
1,600 rainfallrunoff events for 90 Texas watersheds. The objectives of the MRUH
application were (1) to evaluate the applicability of the method if blindly applied to
watersheds of size greater than typically used with either the rational method or the
modified rational method (that is, a few hundred acres), and (2) to study the effects of the
runoff coefficient and the time of concentration on prediction of runoff hydrographs
using MRUH. In addition, three other unit hydrograph models?Clark unit hydrograph
developed for HEC?1?s generalized basin (Clark 1945; USACE 1981), Gamma unit
hydrograph for Texas watersheds (Pradhan 2007) and Natural Resources Conservation
Service Dimensionless (NRCS) unit hydrograph (NRCS 1972)?were used to compute
the direct runoff hydrograph for each of the 1,600 rainfallrunoff events in 90 Texas
watersheds for comparison of the results derived from application of the MRUH with
these unit hydrograph methods.
Revisit of MRM
For the MRM, an urban stormwater runoff hydrograph resulting from the design
storm is approximated as being either triangular or trapezoidal in shape (Smith and Lee
1984; Walesh 1989; Viessman and Lewis 2003), depending on the relation between the
113
storm duration (D) and the time of concentration (T
c
). The rising and the falling limbs are
linear because the timearea relation for a watershed is assumed to be linear. If the storm
duration (D) is equal to time of concentration (T
c
) of a watershed, the resulting
hydrograph is triangular with a peak discharge of CIAQ
p
? at time t = T
c
; that is Case
(A) in Fig. 5.1. If D is greater than T
c
, the resulting hydrograph is trapezoidal with
uniform maximum discharge, CIAQ
p
? from time D to T
c
; that is Case (B) in Fig. 5.1.
The linear rising and falling limbs each has duration of T
c
, as shown in Fig. 5.1 (e.g.,
from Walesh 1989; Viessman and Lewis 2003). If the storm duration D is less than T
c
,
then the resulting hydrograph is trapezoidal with a maximum uniform discharge of Q
p
?
(Eq. 5.2) from the end of the storm (D) to the time of concentration T
c
. The linear rising
and falling portions of the hydrograph each has duration of D < T
c
as shown in Case (C)
in Fig. 5.1. Smith and Lee (1984) and Walesh (1989) reported the modified rational
hydrograph for the case when D is less than T
c
and stated that Q
p
? can be calculated by:
?
?
?
?
?
?
?
c
p
T
D
CIAQ
'
(5.2)
Chien and Saigal (1974) used a linearized subhydrograph approach to derive three
runoff hydrographs depending on rainfall duration and time of concentration, although
there was an error for the case of D < T
c
as reported by Walesh (1975). Wanielista (1990)
discussed the rational hydrograph in the context of the contributing area and assumed that
the contributing area varies linearly with time. He derived a triangular hydrograph for D
= T
c
and a trapezoidal hydrograph for D > T
c
from the rational method, similar to results
by Chien and Saigal (1974) and Walesh (1975).
114
Smith and Lee (1984) examined the rational method as a unit hydrograph. They
noted that if the rate of change of the contributing area is constant so that the accumulated
tributary area increases and decreases linearly and symmetrically with the time, then the
instantaneous unit hydrograph (IUH) response function, u(t), is of rectangular shape
given by:
c
T
A
dt
dA
tu ??)( (5.3)
Using the rectangular response function (Eq. 5.3) in conjunction with a uniform rainfall
intensity, Smith and Lee (1984) derived the resulting direct runoff hydrographs, q(t) (in
watershed depth per time), by convolution as:
?
?? dtuitq
t
e
)()()(
0
??
?
(5.4)
where ? is the time with respect to which the integration is carried out and Cii
e
?)(? is
the effective rainfall intensity. Two types of outflow hydrographs, triangular and
trapezoidal shape (Fig. 5.1), were obtained from Eq. (5.4), depending on the duration of
rainfall. Similar to Smith and Lee (1984), Singh and Cruise (1992) assumed the
watershed is represented as a linear, timeinvariant system whose instantaneous unit
hydrograph (IUH) is a uniform rectangular distribution of base time equal to the time of
concentration of the watershed. They used convolution to derive the Shydrograph and D
hour unit hydrograph from application of the rational method. The unit hydrograph
developed thereby is called the modified rational unit hydrograph (MRUH) in this study.
The MRUH is trapezoidal in shape and three examples of the MRUH used in this study
are shown in Fig. 5.2. Cases (A), (B), and (C) of the MRM in Fig. 5.1 are runoff
115
hydrographs and none is a unit hydrograph, although Cases (B) and (C) have the same
shape as MRUH in Fig. 5.2.
Guo (2000, 2001) developed a rational hydrograph method (RHM) for
continuous, nonuniform rainfall events. The RHM was used to extract the runoff
coefficient and the time of concentration from observed rainfall and runoff data through
optimization. The RHM developed by Guo (2000, 2001) is not unit hydrograph method,
but is a practical procedure to compute the moving average rainfall intensity for
application of the MRM. Guo (2000, 2001) used a linear approximation from the
discharge Q(T
d
) at the end of the rainfall event to zero at the time T
d
+ T
c
for RHM. We
checked that, for nonuniform rainfall events, this approximation is incorrect because it
violates the conservation of mass between the rainfall excess and the runoff hydrograph.
Treating the MRM as a unit hydrograph method, such as the MRUH, will always
conserve mass.
Bennis and Crobeddu (2007) developed an improved rational hydrograph (IRH)
method for small urban catchments using a rectangular impulse response function (Singh
and Cruise 1992; Smith and Lee 1984). They considered impervious and pervious areas
separately. The IRH method was calibrated and validated using ten rainfall events from
two urban catchments. However, the Bennis and Crobeddu (2007) IRH is not a unit
hydrograph method.
The unit hydrograph for a watershed can be used to predict the direct runoff
hydrograph for any given rainfall excess hyetograph (uniform or nonuniform
distribution) using unit hydrograph convolution (Chow et al. 1988; Viessman and Lewis
2003). If the MRM is application of a unit hydrograph method, then the approach
116
establishes a continuity of hydrographdevelopment methods from very small watersheds
to relatively large watersheds. For the MRUH, the assumption and restriction of the
MRM to uniform rainfall distributions, as stated by Rossmiller (1980) and others, is not
necessary. The MRUH can be applied to nonuniform rainfall hyetographs to obtain direct
runoff hydrographs using convolution similar to application of other unit hydrograph
methods.
The Dhr MRUH results from a uniform excess rainfall intensity of 1/D in./hr
over D hrs and has a peak discharge of A/T
c
in ft
3
/s when drainage area A is in acres and
T
c
is in hours (taking into account that oneacre inch per hour is nearly equal to one cubic
foot per second). If SI units are used (drainage area A in hectare and rainfall intensity in
mm/hr), the peak discharge from the MRUH should be equal to A/(360T
c
) in m
3
/s. The
MRUH has only one control parameter?time of concentration of the watershed. The
runoff coefficient, C, for the rational method is not a control parameter of the MRUH.
This is because the MRUH results from one unit of rainfall excess depth and the runoff
coefficient is actually used to determine rainfall excess, not for transformation of
effective rainfall to direct runoff hydrograph (DRH) through application of the MRUH.
Application of the MRUH is straightforward and similar to application of other
unit hydrograph methods. Convolution of the unit hydrographs with the rainfall excess is
applied to obtain the direct runoff hydrograph for each storm event. The excess rainfall
or the net rainfall is obtained from the product of the incremental rainfall and C, similar
to Smith and Lee (1984). The MRUH was first tested using data obtained for concrete
surfaces from Yu and McNown (1964). The first dataset was based on a test bed with an
area of 152.4 m by 0.3 m (500 ft by 1 ft), surface slope of 0.02, and a uniform rainfall
117
intensity of 189 mm/hr. The second dataset was based on a test bed with an area of 76.8
m by 0.3 m (250 ft by 1 ft), surface slope of 0.005, and a variable rainfall intensity with
an initial rate of 43.2 mm/hr, increasing to 96 mm/hr at t = 6 minutes, decreasing to 45
mm/hr at t = 18 minutes, and ending at t = 32 minutes. The T
c
of about 5 minutes was
computed using the Kirpich method (Kirpich 1940) for both experiments. A trapezoidal
1minute MRUH was developed for each experiment (Fig. 5.2A). The runoff coefficient
was taken to be unity. The time interval used for unit hydrograph convolution was 1
minute. Predicted and observed hydrograph ordinates used the same time interval. For
both cases, the modeled results match the observed results well (Fig. 5.3).
The NashSutcliffe efficiency, EF, is a parameter to measure goodnessoffit
between modeled and observed data (Legates and McCabe 1999) and is defined by Eq.
(A.3) in Appendix A. For hydrograph simulation, a good agreement between the
simulated and the measured data is reached when EF is higher than 0.7 (Bennis and
Crobeddu 2007). For the experiment using the uniform rainfall intensity (Fig. 5.3A), EF
was 0.93 and for the experiment using the nonuniform rainfall intensity (Fig. 5.3B), EF
was 0.80, which are greater than 0.7 for both cases indicating a good fit.
5.3 Applications of MRUH in Texas Watersheds
Watersheds Studied and RainfallRunoff Database
Watershed data taken from a larger dataset (Asquith et al. 2004) accumulated by
researchers from the U.S. Geological Survey (USGS) Texas Water Science Center, Texas
Tech University, University of Houston, and Lamar University were used for this study.
The dataset comprises 90 USGS streamflowgaging stations in Texas, each representing a
118
different watershed (Fang et al. 2007, 2008). Location and geographic distribution of the
stations are shown in Fig. 5.4. There are 29, 21, 7, 13 watersheds in Austin, Dallas, Fort
Worth, and San Antonio areas, respectively. The remaining 20 watersheds are small
watersheds located in rural areas of Texas. The drainage areas of study watersheds ranged
from approximately 0.8 to 440.3 km
2
(0.3 to 170 mi
2
), with median and mean values of
17.0 km
2
(6.6 mi
2
) and 41.1 km
2
(15.9 mi
2
), respectively. There are 33, 57, and 80 study
watersheds with drainage areas less than 13 km
2
(5 mi
2
), 26 km
2
(10 mi
2
), and 65 km
2
(25
mi
2
), respectively. The stream slope of study watersheds ranged from 0.0022 to 0.0196,
with median and mean values of 0.0075 and 0.081, respectively. The percentage of
impervious area (IMP) of study watersheds ranged from approximately 0.0 to 74.0, with
median and mean values of 18.0 and 28.4, respectively.
The rainfallrunoff dataset comprised 1,600 rainfallrunoff events recorded during
1959?1986. The number of events available for each watershed varied?for some
watersheds less than 4 events were available, whereas for others as many as 50 events
were available (Cleveland et al. 2006). Rainfall depths ranged from 3.56 mm (0.14 in.) to
489.20 mm (19.26 in.), with median and mean values of 57.66 mm (2.27 in.) and 66.8
mm (2.63 in.), respectively. Maximum rainfall intensities calculated using the time of
concentration ranged from 0.01 mm/min (0.03 in./hr) to 2.54 mm/min (6.01 in./hr), with
median and mean values of 0.25 mm/min (0.58 in./hr) and 0.30 mm/min (0.72 in./hr),
respectively.
Roussel et al. (2005) developed a geospatial database containing watershed
drainage area, longitude and latitude of the USGS streamflowgaging station, (which was
treated as the outlet of the watershed), and 42 watershed characteristics of each individual
119
watershed. These watershed parameters were used to estimate time of concentration for
study watersheds using four empirical methods (Roussel et al. 2005; Fang et al. 2008).
Time of Concentration and Runoff Coefficients
Time of concentration, T
c
, and the runoff coefficient, C, are the required
parameters for application of the MRUH. The T
c
values for the study watersheds were
estimated by Fang et al. (2008) using four empirical equations: (1) Williams equation
(1922) developed from data for watersheds with drainage areas less than 129.5 km
2
(50
mi
2
), (2) Kirpich equation (1940) developed from NRCS data for rural watersheds with
drainage areas less than 0.45 km
2
, (3) Johnstone?Cross equation (1949) developed from
data for watersheds with drainage areas between 65 and 4206 km
2
(25?1620 mi
2
), and (4)
Haktanir?Sezen equation (1990) developed from data for watersheds with drainage areas
from 11 to 9867 km
2
(4 to 3811 mi
2
). For large watersheds, Fang et al. (2007, 2008)
demonstrated that the Kirpich equation provides as reliable an estimate of T
c
as other
empirical equations and the NRCS velocity method (Viessman and Lewis 2003). T
c
estimated using the Kirpich equation reasonably approximates the average T
c
estimated
from observed rainfall and runoff data (Fang et al. 2007). Application of the four
empirical equations requires watershed parameters watershed drainage area, channel
length, channel slope, and watershed shape (Fang et al. 2008). For the present study, T
c
was estimated using watershed parameters developed by USGS researchers through
automated watershed delineation using digital elevation models and geographic
information system (GIS) software (Fang et al. 2008). The range of T
c
values for the
120
study watersheds, along with median and mean values, estimated using the four empirical
equations are presented in Table 5.1.
Rainfall excess was computed using the volumetric interpretation of the rational
runoff coefficient. Wanielista et al. (1997) showed that rainfall loss for a uniform rainfall
input (intensity of i) is equal to (1

C)
iDA and rainfall excess is equal to CiDA. The
rational rainfall loss method is the constant fractional loss model?in which it is assumed
that the watershed immediately converts a constant fraction (proportion) of each rainfall
input into an excess rainfall fraction (McCuen 1998).
Two estimates of the runoff coefficient were examined for the application of
MRUH. The first is a watershed composite, literaturebased coefficient (C
lit
) derived
from landuse information for the watershed and published C
lit
values for appropriate
landuses (Dhakal et al. 2011). The composite C assigned to a watershed is the area
weighted mean C derived from the landuse classes in the watershed. Values of C
lit
for
the study watersheds ranged from 0.29 to 0.63, with median and mean values of 0.50 and
0.47, respectively.
The second runoff coefficient is a backcomputed, volumetric runoff coefficient,
C
vbc
, determined by preserving the runoff volume using observed rainfall and runoff data.
C
vbc
was estimated by the ratio of total runoff depth to total rainfall depth for individual
observed storm event. Computed C
vbc
ranged from 0.001 to 0.99, with median and mean
values of 0.29 and 0.33, respectively, for 1,600 rainfall events in the study watersheds.
The determination and comparison of C
lit
and C
vbc
for the study watersheds was
documented by Dhakal et al. (2011).
121
Estimated Runoff Hydrographs Using the MRUH
For the 90 Texas watersheds, observed rainfall hyetograph and runoff hydrograph
data were tabulated using a time interval of five minutes. Therefore, the fiveminute
MRUH was developed for each of the 90 study watersheds. The fiveminute MRUH
duration is less than the time of concentration for all study watersheds. The basic time
interval used for unit hydrograph convolution and hydrograph ordinates was five minutes.
Comparison between observed and simulated peak discharges and time to peak are
presented in Figs. 5.6 and 5.7.
The results for the event on 07/08/1973 at the USGS streamflowgaging station
08157000 Waller Creek, Austin, Texas are presented in Fig. 5.5 as an illustrative
example. The watershed drainage area is 5.72 km
2
(2.21 square miles). The back
computed volumetric runoff coefficient, C
vbc
, is 0.29. The T
c
values estimated using
Kirpich, HaktanirSezen, JohnstoneCross, and Williams equations are 1.7, 2.2, 1.4, and
3.4 hours, respectively. Peak discharge of the 5minute trapezoidal unit hydrograph (Fig.
5.2B) is 24 m
3
/s (cms). Duration of the rainfall event was 19 hours. Three distinct rainfall
episodes resulted in three distinct discharge peaks. These were reasonably represented by
results from the MRUH using T
c
estimated by Kirpich, HaktanirSezen, and Johnson
Cross equations. Results developed from the Williams equation appear to overestimate
time of concentration for the watershed and peak discharges were then underestimated
(Fig. 5.5). The NashSutcliffe efficiencies from MRUH model results using T
c
values
estimated from Kirpich, HaktanirSezen, JohnstoneCross, and Williams equations are
0.83, 0.86, 0.70, and 0.63, respectively. Simulated time to peak agrees reasonably well
with observed values (Fig. 5.5) when using T
c
estimated by Kirpich, HaktanirSezen, and
122
JohnsonCross equations. However, using T
c
estimated by Williams equation resulted in
the computed time to peak exceeding the observed time to peak. Although the drainage
area of Waller Creek watershed exceeds that usually accepted for MRM application,
results from application of the MRUH reasonably approximate watershed behavior.
Different combinations of T
c
and C were used for applications of MRUH to predict
the direct runoff hydrographs for 1,600 rainfallrunoff events in 90 Texas watersheds to
determine the sensitivity of the peak discharges and time to peak to different T
c
and C
values. Five combinations of T
c
and C were used:
(A) T
c
estimated using HaktanirSezen equation and C
vbc
,
(B) T
c
estimated using JohnstoneCross equation and C
vbc
,
(C) T
c
estimated using Williams equation and C
vbc
,
(D) T
c
estimated using Kirpich equation and C
vbc
, and
(E) T
c
estimated using Kirpich equation and C
lit
.
Figure 5.6 is a plot of the observed and computed peak discharges using C
vbc
and T
c
values calculated using the four different empirical equations (Fang et al. 2008). In
comparison to observed peak discharges, modeled peak discharges using T
c
estimated
from the HaktanirSezen, JohnstoneCross and Kirpich equations not only graphically
look alike (Fig. 5.6) but also are similar with respect to four statistical parameters (Table
1): relative root mean square error, RRMSE (Eq. A.1 in Appendix A); coefficient of
determination, R
2
(Eq.. A.2); NashSutcliffe efficiency, EF (Eq. A.3); and peak relative
error, QB (Eq. A.4). The results for EF and RRMSE using the Williams equation seem to
be inferior to others. The fraction of modeled peak discharges that are within 1/3 of a log
123
cycle from the 1:1 line are summarized in Table 1 and ranged from 67.7% (Williams
equation) to 89.1% (Kirpich equation). Fraction of storms with peak relative error (QB)
less than ?25% and ?50% is listed in Table 5.1 for applications of MRUH with different
combinations of T
c
and C. Applications of MRUH with T
c
estimated from Kirpich
equation and backcomputed C
vbc
resulted in 73% of storms with QB less than ?50%.
Use of C
vbc
results in preservation of event runoff volume. Ideally, computed and
observed peaks should plot precisely along the equal value line (black line in Fig. 5.6).
However, the unit hydrograph is a mathematical model that is an incomplete description
of the complexity of the combination of the rainfallrunoff process and runoff dynamics.
Therefore, the relatively simple approach cannot fully capture the nuances of watershed
dynamics and deviations from this ideal (the equalvalue line) are expected. For example,
Asquith and Roussel (2009) computed mean residual standard error about 1/3 of a log
cycle for annual peak discharges at 638 streamflow gauging stations in Texas.
Figure 5.7 is a plot of the observed time to peak (T
p
) and computed T
p
values
predicted using C
vbc
and T
c
values calculated using the four different empirical equations
(Fang et al. 2008). For T
p
, application of MRUH using T
c
estimated from the Haktanir
Sezen, JohnstoneCross and Kirpich equations produces the similar values of the
quantitative measures: median value of T
p
relative error (TB) and fraction of storms with
TB less than ?25% and ? 50% (Table 5.2), and TB equal to ? 50% is graphically
presented in Fig. 5.7 also. The T
p
results using the Williams equation seem to be slightly
inferior to others with respect to TB (Table 5.2). In summary, for predicting peak
discharge and time to peak, use of T
c
estimated from Williams equation with the MRUH
124
produces less accurate results than those computed using the Kirpich, HaktanirSezen and
JohnstoneCross equations.
Simulated peak discharges derived using MRUH with the forwardcomputed
(literaturebased) runoff coefficient (C
lit
) are compared against the results obtained from
the backcomputed (C
vbc
) runoff coefficient (Fig. 5.8) when T
c
values were estimated
using the Kirpich equation. For the peak discharges predicted using C
lit
, most of the
values are above the equal value line (1:1 line). Based on visual inspection, about one
third of the peak discharges from C
lit
(triangles) are quite far from the peak discharges
using C
vbc
(black circles). Peak discharges computed using C
vbc
are superior to those
using C
lit
with respect to all statistical measures used to assess goodness of fit (Table 5.3).
Therefore, use of literature based values (C
lit
) will tend to generate estimates of peak
discharge that exceed expected values (observations) when the C
lit
values are interpreted
as volumetric coefficients. Furthermore, literaturebased estimates (C
lit
) of the runoff
coefficient yield results that do not preserve runoff volume when applied to measured
rainfallrunoff events. In contrast, there is no difference in quantitative measures between
the observed and predicted time to peak values, regardless of which runoff coefficient is
used (Table 5.3). This is because T
p
is controlled by time of concentration and rainfall
hyetograph and the same T
c
values were used with different runoff coefficients (Fig. 5.8)
for each of 90 Texas watersheds.
Hence, the simulation results of peak discharge are more sensitive to the choice of
the runoff coefficients (C) or rainfall loss model. Furthermore, the time to peak results are
not related to C when MRUH was used.
125
5.4 Estimated Runoff Hydrographs from Different Unit Hydrograph Methods
In addition to application of the MRUH for 90 Texas watersheds, three other unit
hydrograph models?the unit hydrograph developed using the Clark method (Clark 1945)
with HEC?1?s generalized basin shape (USACE 1981), the NRCS unit hydrograph
(NRCS 1972), and the Gamma unit hydrograph (GUH) for Texas watersheds (Pradhan
2007)?were used to develop the direct runoff hydrograph for each rainfallrunoff event
in the database. The GUH used in this study was that developed by researchers at Lamar
University for TxDOT project 0?4193 ?Regional Characteristics of Unit Hydrographs.?
Linear programming was used to develop unit hydrographs from observed rainfall
hyetographs and runoff hydrograph, and the GUH was fitted to each derived unit
hydrograph. Regression equations were developed for fiveminute GUH parameters:
peak discharge Q
p
(in cfs) and time to peak T
p
(in hours) (Pradhan 2007),
,55075.0
06032.042612.026998.0 ?
? SLAT
p
(5.5)
,22352.93
5.0326.083576.0
SLAQ
p
?
?
(5.6)
where A is drainage area in square miles, L is main channel length in miles, and S is main
channel slope (ft/mile, elevation difference in feet divided by main channel length in
miles). The ordinates of the GUH can be obtained from (Viessman and Lewis 2003):
,
)](1[
?
?
p
T
t
p
p
e
T
t
QQ
?
?
?
?
?
?
?
?
(5.7)
where Q is the discharge ordinate at time t and ? is the shape parameter of GUH. Two of
the three GUH parameters (Q
p
, T
p
, and ?) are independent and the shape factor is
determined from Q
p
and T
p
(Aron and White 1982).
126
Clark?s (1945) instantaneous unit hydrograph (IUH) method is based on the time
area curve method (Bedient and Huber 2002). The Clark IUH method is one of the unit
hydrographs available in the flood hydrograph package HEC1 (USACE 1981) and the
hydrologic modeling system HECHMS (USACE 2000). A synthetic timearea curve
derived from a generalized basin shape is used to implement Clark?s IUH in HEC1 and
HECHMS. The equations for the timearea curve are
AI = 1.414 TI
1.5
, 0 ? TI ? 0.5 (5.8)
1 AI = 1.414 (1TI)
1.5
, 0.5 < TI <1 (5.9)
where AI is the cumulative area as a fraction of watershed area and TI is fraction of time
of concentration. These equations are applicable to most basins (Bedient and Huber
2002). In the Clark method and HEC1/HECHMS programs, the resulting hydrograph is
routed through a linear reservoir at the outlet of a watershed. The linear reservoir routing
was not implemented because this study is to only compare MRUH resulted from an
equal timearea curve (such as a rectangular watershed) and ClarkHEC1 UH resulted
from a generalized watershed shape (ellipselike shape having a nonuniform timearea
curve). Both MRUH and ClarkHEC1 UH implemented without routing assume that the
outflow hydrograph results from pure translation of direct runoff to the outlet.
The NRCS dimensionless unit hydrograph was developed in the late 1940s (NRCS
1972). NRCS personnel analyzed a large number of unit hydrographs for watersheds of
different sizes and in different geographic locations to develop a generalized
dimensionless unit hydrograph in terms of t/T
p
and q/Q
p
. The peak flow, Q
p
, for the unit
hydrograph is computed by approximating the unit hydrograph with a triangular shape
having base time of 8/3T
p
and unit area (Viessman and Lewis 2003):
127
p
p
T
A
Q
484
? , (5.10)
where Q
p
is the peak discharge (cfs) and A is the drainage area (mi
2
).
Unit hydrographs developed using all four UH models, including the MRUH, for the
watershed associated with the USGS streamflowgaging station 08048520 Sycamore
Creek in Fort Worth are shown in Fig. 5.9 (A). The purpose of this example is to
illustrate the differences and similarities of the UH models. The shape of the MRUH is
trapezoidal. Unit hydrographs from the ClarkHEC1, the Gamma, and the NRCS
methods are curvilinear. The unit hydrograph peak discharge from each model is
different (Fig. 5.9A). However, the area under the UH curves is the same. This is
because each UH corresponds to 1 inch of a uniform excess rainfall over 5minute
duration (one impulse).
Gamma, ClarkHEC1, and NRCS unit hydrographs developed for each watershed
were applied to the 1,600 rainfallrunoff events in the database to generate direct runoff
hydrographs. The constant fraction rainfall loss method (rational method) was used to
estimate rainfall excess for each rainfall event. The runoff coefficient C
vbc
determined for
each event was used. T
c
determined using the Kirpich method (1940) was used for those
methods that require T
c
. As an illustrative example, the results for the observed and
simulated direct runoff hydrographs for the rainfall event on 07/28/1973 at the USGS
streamflowgaging station 08048520 (Sycamore Creek in Fort Worth, Texas) by the four
models (base flow was assumed to be zero) is presented in Fig. 5.9 (B). The watershed
area is of 45.66 km
2
(17.63 square miles), the time of concentration is 3.96 hours from
the Kirpich method, and the backcomputed volumetric runoff coefficient, C
vbc
, is 0.20.
128
Simulated peak discharges from the four UH methods are different, but comparable. For
the particular example shown in Fig. 5.9 (B), the MRUH and the ClarkHEC1 model
appear to perform better than the other models with regard to prediction of peak
discharge. For the time to peak, simulated values using the four methods agree
reasonably well with the observed value (Fig. 5.9). Additionally, the area under the four
simulated hydrographs matches the observed curve because event C
vbc
was used.
Although the drainage area of Sycamore Creek watershed exceeds that usually accepted
for rational method application, results from the MRUH reasonably approximate
watershed behavior.
The observed and modeled peak discharges from all four UH models developed
using backcomputed runoff coefficient, C
vbc
and time of concentration from the Kirpich
method are presented in Fig. 5.10 for 1,600 rainfall and runoff events. Modeled peak
discharges from all the four UH models are similar (Fig. 5.10). Four statistical parameters
RRMSE, R
2
, EF, and QB (Appendix A) between the observed and modeled peak
discharges were computed for evaluation of model performance (Loague and Green
1991; Cleveland et al. 2006) and are listed in Table 5.4. Based on the statistical measures,
all the four UH models perform similarly. However, the GUH developed for Texas
watersheds perform somewhat worse than the other three UH models (Table 5.4).
Fractions of storms for each model meeting the acceptance tolerance of QB and TB are
also listed in Table 5.4 (Eqs. A.4 and A.6). Using this acceptance approach, again all the
models perform similarly.
5.5 Sensitivity of the MRUH to unit hydrograph duration
129
A sensitivity analysis was performed for peak discharges derived from application of
MRUH using different unit hydrograph durations. The simulated runoff hydrographs
were obtained for unit hydrograph durations of 5, 10, 20, 30, 40 and 50minutes. To
minimize error in developing discrete MRUH and DRH, a time interval of 10 minutes
was used for hydrograph convolution when unit hydrograph durations exceeded 10
minutes.
Predicted runoff hydrographs for the rainfall event on 07/28/1973 for the watershed
associated with the USGS streamflowgaging station 08178600 Salado Creek San
Antonio are shown in Fig. 5.11 as an illustrative example. No noticeable differences were
visible both in terms of the peak and the shape of the hydrographs regardless of the unit
hydrograph duration. The NashSutcliffe efficiency and relative change in the peak
discharge for the simulated results are presented in Table 5.5. Based on review of
statistical measures, results are not sensitive to changes in unit hydrograph duration
(Table 5.5, EF is derived from runoff hydrograph ordinates).
For the 1,600 rainfall events in the database, the median relative change in the peak
discharge (Q
RE
) for the changes in the UH durations are all 0% (Table 5.6). Fractions of
storms with Q
RE
less than ?5% and ?10% are listed in Table 5.6, and almost all of rainfall
events (95 to 99%) has Q
RE
less than ?10%. The application of MRUH is not sensitive to
the selection of the unit hydrograph duration so long as the same time interval is used for
hydrograph convolution.
5.6 Discussion and Summary
130
The modified rational method, MRM, is an extension of the rational method to
produce simple runoff hydrographs for applications that do not warrant a more complex
modeling approach. In this study, the MRM was revisited. The hydrographs developed
using MRM can be considered an application of a special unit hydrograph termed the
modified rational unit hydrograph, MRUH. The MRUH method was applied to develop
unit hydrographs for 90 watersheds in Texas. Unit hydrograph convolution was used to
determine the direct runoff hydrograph for 1,600 rainfallrunoff events associated with
the Texas database. The purposes were (1) to evaluate the applicability of the method if
blindly applied to watersheds of any size, and (2) to study the effects of runoff coefficient
and the time of concentration on prediction of runoff hydrograph using the MRUH.
Runoff coefficients estimated using two approaches by Dhakal et al. (2011) were
examined for application with the MRUH. The first was a watershed composite
literaturebased coefficient (C
lit
) derived using the landuse information for the watershed
and published C
lit
values for various landuses. The second was a backcomputed
volumetric runoff coefficient (C
vbc
) determined by preserving the runoff volume and
using observed rainfall and runoff data. Times of concentration for study watersheds
were estimated by Fang et al. (2008) from four empirical equations, which were based on
several watershed characteristics. Predicted and observed discharge hydrographs were
reported and compared. Simulated peak discharges and times to peak from MRUH agree
reasonably well with observed values. The drainage area of the study watersheds
(average 440 km
2
or 15.6 mi
2
) is greater than that usually accepted for rational method
application (0.8 km
2
or 0.3 mi
2
), yet results from the MRUH reasonably approximate
watershed behavior regardless of watershed size. Simulated peak discharges are more
131
sensitive to the choice of the runoff coefficient than the time of concentration. Simulated
times to peak are moderately sensitive to the time of concentration but independent of the
runoff coefficient. A sensitivity analysis of the MRUH to the unit hydrograph duration
was performed. The MRUH is not sensitive to the selection of the unit hydrograph
duration so long as the same time interval is used for hydrograph convolution.
Three other unit hydrograph models, the Clark (using HEC?1?s generalized basin
equations), the Gamma, and the NRCS unit hydrographs were also used to compute the
direct runoff hydrograph for each rainfallrunoff event in the database. Runoff
hydrographs simulated using all four methods were similar. Simulated peak discharges
for all events in the database were similar regardless of statistical or quantitative
measures used for comparison. For time to peak, simulated values using all four models
agree reasonably well with observed values. The four UH models produce similar values
of statistical and quantitative measures for both peak discharges and time to peak.
Three general conclusions for MRUH are: (1) Being a unit hydrograph, it can be
applied to nonuniform rainfall distributions and for watersheds with drainage areas
greater than typically used with either the rational method or the modified rational
method (that is, a few hundred acres). (2) The MRUH performs about as well as other
unit hydrograph methods used in this study for predicting the peak discharge and time to
peak of the direct runoff hydrograph, so long as the same rainfall loss model is used. (3)
Modeled peak discharges from application of the MRUH are more sensitive to the
selection of runoff coefficient, less sensitive to T
c
, and not sensitive to the selection of the
unit hydrograph duration. In predicting peak discharges and runoff hydrographs for
132
engineering design, rainfall loss estimation results in greater uncertainty and contributes
more model errors than variations of UH methods and model parameters for UH.
5.7 Notation
The following symbols are used in this paper:
? = shape parameter of gamma unit hydrograph;
A = drainage area in hectares or acres;
AI = cumulative area as a fraction of watershed area;
C = runoff coefficient;
C
lit
= composite literaturebased runoff coefficient;
C
vbc
= backcomputed volumetric runoff coefficient;
D = storm duration;
EF = NashSutcliffe efficiency;
I = average rainfall intensity (mm/hr or in. /hr) with the duration equal to time of
concentration;
L = main channel length in mile;
m
o
= the dimensional correction factor (1.008 in English units, 1/360 = 0.00278 in SI
units);
Q
p
= peak runoff rate in m
3
/s or ft
3
/s;
Q
p
? = peak runoff rate of the modified rational hydrograph for the case when the storm
duration is less than the time of concentration of the drainage area;
QB = peak relative error between the observed and simulated peak discharges;
q(t) = direct runoff hydrographs (in watershed depth per time) by convolution;
133
R
2
= coefficient of determination;
RRMSE = relative root mean square error;
S = main channel slope (ft/mile);
t = shape parameter of gamma unit hydrograph;
TB = relative bias of direct runoff hydrograph time to peak;
TI = fraction of time of concentration;
T
c
= time of concentration;
T
p
= time to peak;
u(t) = rectangular instantaneous unit hydrograph response function;
5.8 Appendix A: Statistical Measures to Evaluate Model Performance
Four statistical measures were used to analyze model results. They are the relative
root mean square error (RRMSE), the coefficient of determination (R
2
), the NashSutcliffe
efficiency (EF), and the peak relative error (QB) between the observed and simulated
peak discharges and times to peak (Loague and Green 1991; Feyen et al. 2000; Cleveland
et al. 2006). The equations used to compute these measures are:
,
)(
5.0
1
2
O
n
OP
RRMSE
n
i
ii
?
?
?
?
?
?
?
?
?
?
(A.1)
,
)()(
))((
2
1
22
1
12
?
?
?
?
?
?
?
?
?
?
?
?
?
?
??
??
?
??
?
??
?
n
i
i
n
i
i
n
i
ii
PPOO
PPOO
R
(A.2)
134
and,
)(
)()(
1
2
11
22
?
??
?
??
?
?
?
?
?
?
?
???
?
n
i
N
I
n
i
iii
OOi
OPOO
EF (A.3)
,
i
ii
O
OP
QB
?
? (A.4)
where, P
i
are the simulated peak discharge values, O
i
are the observed peak discharge
values, n is the number of observations and O is the mean of the observed peak
discharge values. RRMSE is a measure of overall spread of the residuals with respect to
the mean observed value. The target RRMSE value is 0 for acceptance of a model (Feyen
et al. 2000). The coefficient of determination R
2
is a measure of the proportion of the total
variability in observed data that can be explained by the model and ranges from 0 to 1.
According to Moriasi et al. (2007) values of R
2
greater than 0.5 are considered acceptable
for a model. Although R
2
and EF values are used often for the evaluation of a model,
Legates and McCabe (1999) suggested that EF is a more appropriate measure for
goodnessoffit. For hydrograph simulation, a good agreement between the simulated and
the measured data is reached when EF is higher than 0.7 (Bennis and Crobeddu 2007).
Peak relative error, QB, is the difference in magnitude between the modeled and observed
peak divided by observed peak discharge. Similar to RRMSE, values of QB near to 0
indicate correspondence between modeled and observed values. Cleveland et al. (2006)
suggested the following range of the QB for the acceptance of model performance:
25.025.0 ??? QB (A.5)
For evaluation of the time to peak results, relative bias of direct runoff hydrograph
time to peak (TB) was estimated (Zhao and Tung 1994) for each storm event using:
135
,
po
popm
t
tt
TB
?
? (A.6)
where, t
pm
is the modeled time to peak and t
po
is the observed time to peak. A positive TB
indicates that the observed peak occurs sooner (smaller) than the modeled peak (i.e. the
model predicts a late peak). Similarly, a negative TB indicates that the observed peak
occurs later (larger) than the modeled peak (i.e. the model predicts an early peak)
[Cleveland et al. 2006].
5.9 References:
Aron, G., and White, E. L. (1982). "Fitting a gamma distribution over a synthetic unit
hydrograph." Water resources Bulletin, 18(1), 9598.
Asquith, W. H., Thompson, D. B., Cleveland, T. G., and Fang, X. (2004). "Synthesis of
rainfall and runoff data used for Texas department of transportation research
projects 04193 and 04194." Austin, Texas 787544733.
Asquith, W. H., and Roussel, M. C. (2009). "Regression equations for estimation of
annual peakstreamflow frequency for undeveloped watersheds in Texas using an
Lmomentbased, PRESSMinimized, residualadjusted approach." U.S.
Geological Survey Scientific Investigations Report 20095087, 48 p.
Bedient, P. B., and Huber., W. C. (2002). Hydrology and Floodplain Analysis, Prentice
Hall, Upper Saddle River, NJ 07458.
Bennis, S., and Crobeddu, E. (2007). "New runoff simulation model for small urban
catchments." Journal of Hydrologic Engineering, 12(5), 540544.
Chien, J.S., and Saigal, K. K. (1974). "Urban runoff by linearized subhydrograph
method." ASCE Journal of the Hydraulics Division, 100(HY8), 11411157.
Chow, V. T., Maidment, D. R., and Mays, L. W. (1988). Applied hydrology, McGraw
Hill, New York.
Clark, C. O. (1945). "Storage and the unit hydrograph." Transactions of American
Society of Civil Engineering, 110, 14191446.
136
Cleveland, T. G., He, X., Asquith, W. H., Fang, X., and Thompson, D. B. (2006).
"Instantaneous unit hydrograph evaluation for rainfallrunoff modeling of small
watersheds in north and south central Texas." Journal of Irrigation and Drainage
Engineering, 132(5), 479485.
Dhakal, N., Fang, X., Cleveland, T.G., Thompson, D.B., Asquith, W.H., and Marzen, L.J.
(2012). ?Estimation of volumetric runoff coefficients for Texas watersheds using
landuse and rainfallrunoff data.? Journal of Irrigation and Drainage Engineering,
138(1):4354.
Fang, X., Thompson, D. B., Cleveland, T. G., and Pradhan, P. (2007). "Variations of time
of concentration estimates using NRCS velocity method." Journal of Irrigation
and Drainage Engineering, 133(4), 314322.
Fang, X., Thompson, D. B., Cleveland, T. G., Pradhan, P., and Malla, R. (2008). "Time
of concentration estimated using watershed parameters determined by automated
and manual methods." Journal of Irrigation and Drainage Engineering, 134(2),
202211.
Feyen, L., V?zquez, R., Christiaens, K., Sels, O., and Feyen, J. (2000). "Application of a
distributed physicallybased hydrological model to a medium size catchment."
Hydrology and Earth System Sciences, 4(1), 4763.
Guo, C. Y. J. (2000). "Strom hydrograph from small urban catchments." Water
International, 25(3), 481487.
Guo, J. C. Y. (2001). "Rational hydrograph method for small urban watersheds." Journal
of Hydrologic Engineering, 6(4), 352357.
Haktanir, T., and Sezen, N. (1990). "Suitability of two  parameter gamma and three 
parameter beta distributions as synthetic unit hydrographs in Anatolia."
Hydrological Sciences, 35(2), 167184.
Johnstone, D., and Cross, W. P. (1949). Elements of applied hydrology, Ronald Press,
New York.
Kirpich, Z. P. (1940). "Time of concentration of small agricultural watersheds." Civil
Engineering, 10(6), 362.
Kuichling, E. (1889). "The relation between the rainfall and the discharge of sewers in
populous areas." Transactions, American Society of Civil Engineers 20, 1?56.
Legates, D. R., and McCabe, G. J. (1999). "Evaluating the use of ?goodnessoffit?
measures in hydrologic and hydroclimatic model validation." Water Resources
Research, 35(1), 233241.
137
LloydDavies, D. E. (1906). "The elimination of storm water from sewerage systems."
Minutes of Proceedings, Institution of Civil Engineers, Great Britain, 164, 41.
Loague, K., and Green, R. E. (1991). "Statistical and graphical methods for evaluating
solute transport models: overview and application." Journal of Contaminant
Hydrology, 7(12), 5173.
McCuen, R. H. (1998). Hydrologic analysis and design, PrenticeHall, Inc., Upper
Saddle River, N.J.
Moriasi, D. N., Arnold, J. G., Van Liew, M. W., Bingner, R. L., Harmel, R. D., and
Veith, T. L. (2007). "Model Evaluation Guidelines for Systematic Quantification
of Accuracy in Watershed Simulations." Transactions of the ASABE, 50(3), 885
900.
NRCS. (1972). "Hydrology", National Resource Conservation Service (NRCS), U.S.
Dept. of Agriculture, Washington D.C.
Poertner, H. G. (1974). "Practices in detention of urban stormwater runoff: An
investigation of concepts, techniques, applications, costs, problems, legislation,
legal aspects and opinions." American Public Works Association, Chicago, IL.
Pradhan, P. (2007). "Rainfall loss and unit hydrograph estimation by nonlinear
programming for the Texas watersheds," Lamar University, Beaumont, Texas
77710.
Rossmiller, R. L. (1980). "Rational formula revisited." International Symposium on
Urban Storm Runoff, University of Kentucky, Lexington, KY., 112.
Roussel, M. C., Thompson, D. B., Fang, D. X., Cleveland, T. G., and Garcia, A. C.
(2005). "Timing parameter estimation for applicable Texas watersheds." 04696
2, Texas Department of Transportation, Austin, Texas.
Singh, V. P., and Cruise, J. F. (1992). "Analysis of the rational formula using a system
approach." Catchment runoff and rational formula, B.C. Yen, ed., Water
Resources Publication, Littleton, Colo. , 3951.
Smith, A. A., and Lee, K. (1984). "The rational method revisited." Canadian Journal of
Civil Engineering(11), 854862.
TxDOT. (2002). Hydraulic design manual, The bridge division of the Texas Department
of Transportation (TxDOT), 125 East 11th street, Austin, Texas 787012483.
USACE. (1981). "HEC1 flood hydrograph package, user?s manual (Revision in 1987)."
U.S. Army Corps of Engineers (USACE), Hydrologic Engineering Center (HEC),
Davis, CA.
138
USACE. (2000). ?Hydrologic modeling system HECHMS.? U.S. Army Corps of
Engineers (USACE), Hydrologic Engineering Center (HEC), Davis, California.
Viessman, W., and Lewis, G. L. (2003). Introduction to hydrology, Pearson Education,
Upper Saddle River, N.J.
Walesh, S. G. (1975). "Discussion of Chien and Saigal (1974)." ASCE Journal of the
Hydraulics Division, 101(HY11), 14471449.
Walesh, S. G. (1989). Urban Water Management, Wiley, New York.
Wanielista, M. P. (1990). Hydrology: Water quantity and quality control, John Wiley &
Sons, Inc.
Wanielista, M. P., Kersten, R., and Eaglin, R. (1997). Hydrology: Water quantity and
quality control, John Wiley & Sons, Inc.
Williams, G. B. (1922). "Flood discharges and the dimensions of spillways in india "
Engineering (London), 134, 321.
Yu, Y. S., and McNown, J. S. (1964). "Runoff from impervious surfaces " Journal of
Hydraulic Research, 2(1), 324.
Zhao, B., and Tung, Y. K. (1994). "Determination of optimal unit hydrographs by linear
programming." Water Resources Management, 8(2), 101119.
139
Table 5.1 Quantitative Measures of the Success of the?Peak Discharges Modeled Using MRUH
with BackComputed Volumetric Runoff Coefficient (C
vbc
) and Time of Concentration (T
c
)
Estimated Using Four Equations.
Statistical Parameters
Haktanir
Sezen
equation
1
Johnstone
Cross equation
2
Williams
equation
3
Kirpich
equation
4
RRMSE (Eq. A.1)
5
0.87 0.78 1.03 0.75
R
2
(Eq. A.2) 0.66 0.72 0.70 0.75
EF (Eq. A.3) 0.64 0.71 0.49 0.73
Median value of QB (Eq. A.4) 0.18 0.04 0.40 0.09
Fraction of storms with 0.25 ? QB ?
0.25
0.35 0.39 0.25 0.40
Fraction of storms with 0.5 ? QB ? 0.5 0.68 0.70 0.60 0.73
% of events within ? 1/3 of a log cycle 81.5 86.4 67.7 89.1
1
T
c
computed using HaktanirSezen equation ranged from 0.8 to 17.6 hours in the study
watersheds, with median and mean values of 2.7 hours and 3.7 hours, respectively.
2
T
c
computed using JohnstoneCross equation ranged from 0.7 to 7.7 hours in the study
watersheds, with median and mean values of 1.9 hours and 2.3 hours, respectively
3
T
c
computed using Williams equation ranged from 1.2 to 31.4 hours in the study watersheds,
with median and mean values of 4.0 hours and 5.9 hours, respectively
4
T
c
computed using Kirpich equation ranged from 0.6 to 16.2 hours in the study watersheds, with
median and mean values of 2.3 hours and 3.2 hours, respectively
5
Statistical parameters are defined in Appendix A.
Table 5.2 Quantitative Measures of the Success of the?Time to Peak Modeled Using MRUH with
BackComputed Volumetric Runoff Coefficient (C
vbc
) and Time of Concentration Estimated
Using Four Equations.
Statistical Parameters
HaktanirSezen
equation
Johnstone
Cross
equation
Williams
equation
Kirpich
equation
Median value of TB (Eq. A.6) 0.00 0.07 0.09 0.03
Fraction of storms with 0.25 ? TB ?
0.25
0.52 0.52 0.46 0.53
Fraction of storms with 0.5 ? TB ?
0.5
0.70 0.71 0.64 0.71
140
Table 5.3 Quantitative Measures of the Success of the?Peak Discharges and Time to Peak
Modeled Using MRUH with Time of Concentration Estimated Using Kirpich Equation and
Runoff Coefficients Estimated Using Two Different Methods (C
vbc
and C
lit
).
Statistical Parameters C
vbc
C
lit
RRMSE (Eq. A.1) 0.75 1.07
R
2
(Eq. A.2) 0.75 0.48
EF (Eq. A.3) 0.73 0.45
Median value of QB (Eq. A.4) 0.09 0.44
Fraction of storms with 0.25 ? QB ? 0.25 0.40 0.22
Fraction of storms with 0.5 ? QB ? 0.5 0.73 0.46
% of events within ? 1/3 of a log cycle 89.1 63.3
Median value of TB (Eq. A.6) 0.03 0.03
Fraction of storms with 0.25 ? TB ? 0.25 0.53 0.53
Fraction of storms with 0.5 ? TB ? 0.5 0.71 0.71
Table 5.4 Quantitative Measures of the Success of the Peak Discharges Modeled Using Four Unit
Hydrograph Models for 1,600 RainfallRunoff Events in 90 Texas Watersheds.
Statistical Parameters MRUH Gamma UH
ClarkHEC1
UH
NRCS
UH
RRMSE (Eq. A.1) 0.75 0.87 0.74 0.72
R
2
(Eq. A.2) 0.75 0.80 0.74 0.78
EF (Eq. A.3) 0.73 0.65 0.74 0.76
Median value of QB (Eq. A.4) 0.09 0.31 0.04 0.10
Fraction of storms with 0.25 ? QB ?
0.25
0.40 0.32 0.39 0.39
Fraction of storms with 0.5 ? QB ?
0.5
0.73 0.71 0.70 0.76
% of events within ? 1/3 of a log
cycle
89.1 81.2 87.0 90.6
Median value of TB (Eq. A.6) 0.03 0.03 0.04 0.00
Fraction of storms with 0.25 ? TB ?
0.25
0.53 0.53 0.53 0.53
Fraction of storms with 0.5 ? TB ?
0.5
0.71 0.72 0.74 0.73
141
Table 5.5 Sensitivity of Peak Discharges Modeled Using MRUH on Unit Hydrograph Duration
for the Rainfall Event on 07/28/1973 for the Watershed Associated with the USGS Streamflow
gaging Station 08178600 Salado Creek, San Antonio, Texas.
UH Duration EF
1
Change in UH duration
(minutes) Relative (%) change in Q
p
2
5min 0.89
10min 0.90 5 to 10 0.00
20min 0.91 10 to 20 0.30
30min 0.92 20 to 30 2.42
40min 0.91 30 to 40 1.27
50min 0.92 40 to 50 2.07
1
Nash Sutcliffe efficiency (Eq. A.3) derived from runoff hydrograph ordinates
2
Relative change in Q
p
= Q
RE
(%) = (Q
p10
? Q
p5
)/Q
p5
?
100%, where Q
p10
and Q
p5
are peak
discharges calculated using unit hydrograph durations of 10 and 5 minutes, respectively. Q
RE
for
other UH durations is calculated in a similar way.
Table 5.6 Sensitivity of Peak Discharges Modeled Using MRUH on Unit Hydrograph Duration
for 1,600 Rainfall Events in 90 Texas Watersheds.
Change in UH
duration
(minutes)
Median value of
Q
RE
1
Fraction of storms with
5% ? Q
RE
? 5%
Fraction of storms with
10% ? Q
RE
? 10%
5 to 10 0.00 0.93 0.99
10 to 20 0.00 0.98 0.99
20 to 30 0.00 0.96 0.99
30 to 40 0.00 0.90 0.96
40 to 50 0.00 0.87 0.95
1
Q
RE
is defined in Table 5.
142
Fig 5.1 The modified rational hydrographs for three different cases: (A) Duration of
rainfall (D) is equal to time of concentration (T
c
), (B) Duration of rainfall is greater than
T
c
, and (C) Duration of rainfall is less than T
c
.
143
0
5
10
15
20
25
30
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0.000
0.001
0.002
0.003
0.004
0.005
0.006
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Ti m e ( h r )
0 1 2 3 4 5 6
Ti m e ( m i n u t e s )
for Waller Creek, Austin, Texas
for 152.4 m x 0.3 m (Yu and McNown 1964)
for 76.8 m x 0.3 m (Yu and McNown 1964)
(A)
(B)
Fig. 5.2 The modified rational unit hydrographs (MRUH) developed for: (A) two lab
settings from Yu and McNown (1964) and (B) for the watershed associated with USGS
streamflowgaging station 08157000 Waller Creek, Austin, Texas.
144
0.0000
0.0002
0.0004
0.0006
0.0008
0.0010
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0.000
0.001
0.002
0.003
0.004
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0.0
2.0
4.0
0.0
4.0
8.0
0 4 8 12 16 20 24 28 32 36 40
Ti m e ( m i n u t e s )
0 4 8 12 16 20 24 28 32 36 40
Modeled discharge
Observed discharge
Incremental rainfall
I
n
c
r
e
m
e
n
t
a
l
r
ain
f
all
(
m
m
)
(A)
(B)
Fig. 5.3 Incremental rainfall hyetograph and observed and modeled runoff hydrographs
using the MRUH for the two lab tests on concrete surfaces: (A) 152.4 m
?
0.3 m and (B)
76.8 m
?
0.3 m reported by Yu and McNown (1964).
145
Fig. 5.4 Map showing the U.S. Geological Survey streamflowgaging stations (dots)
associated with the watershed locations in Texas.
146
0
1
2
3
4
5
6
7
8
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0
2
4
6
8
5 10 15 20 25 30
Ti m e ( h r )
Observed discharge
Incremental rainfall
T
c
(HaktanirSezen Equation)
T
c
(JohnstoneCross Equation)
T
c
(Williams Equation)
T
c
(Kirpich Equation)
I
n
cr
e
m
en
t
a
l
ra
i
n
f
a
l
l
(
m
m)
Fig. 5.5 Sensitivity of Peak Discharges Modeled using MRUH on Unit Hydrograph
Duration for the Rainfall Event on 07/28/1973 for the Watershed Associated with the
USGS Streamflowgaging Station 08178600 Salado Creek, San Antonio, Texas.
147
1
10
100
M
o
d
e
l
e
d
Q
p
(
c
m
s
)
1
10
100
M
o
d
e
l
e
d
Q
p
(
c
m
s
)
1 10 100
Observed Q
p
(cms)
1 10 100
Observed Q
p
(cms)
T
c
(Williams Equation)
T
c
(HaktanirSezen Equation)
1:1 line
? 1/3 of a log cycle
T
c
(Kirpich Equat ion)
T
c
(JohnstoneCross Equation)
(A)
(C)
(B)
(D)
Fig. 5.6 Modeled versus observed peak discharges for 1,600 rainfallrunoff events in 90
Texas watersheds. Modeled results were developed from MRUH using event C
vbc
and T
c
estimated using four different methods: (A) HaktanirSezen equation, (B) Johnstone
Cross equation, (C) Williams equation and (D) Kirpich equation.
148
0.01
0.1
1
10
100
M
o
d
e
l
e
d
T
p
(
h
r
)
0.01
0.1
1
10
100
M
o
d
e
l
e
d
T
p
(
h
r
)
0.01 0.1 1 10 100
Observed T
p
(hr)
0.01 0.1 1 10 100
Observed T
p
(hr)
T
c
(Williams Equation)
T
c
(HaktanirSezen Equation)
1:1 line
? 0.5 TB
T
c
(Kirpich Equat ion)
T
c
(JohnstoneCross Equation)
(A)
(C)
(B)
(D)
Fig. 5.7 Modeled versus observed time to peak for 1,600 rainfallrunoff events in 90
Texas watersheds. Modeled results were developed from MRUH using event C
vbc
and T
c
estimated using four different methods: (A) HaktanirSezen equation, (B) Johnstone
Cross equation, (C) Williams equation and (D) Kirpich equation.
149
1
10
100
M
o
d
e
l
e
d
Q
p
(
c
m
s
)
1 10 100
Observed Q
p
(cms)
C
vbc
C
lit
1:1 line ? 1/3 of a log cycle
Fig. 5.8 Modeled versus observed peak discharges developed from MRUH using C
vbc
(triangles) and C
lit
(circles) with time of concentration estimated using Kirpich equation
for 90 Texas watersheds.
150
0
10
20
30
40
50
60
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0
6
12
18
24
I
n
c
r
e
m
e
n
t
a
l
R
a
i
n
f
a
l
l
(
m
m
)
0
20
40
60
80
100
120
140
D
i
s
c
h
a
r
g
e
(
c
m
s
)
18 20 22 24 26 28 30 32 34 36 38 40 42
Time ( hr)
0 1 2 3 4 5 6 7 8 9 10 11
Observed
Modeled (MRUH)
Modeled (ClarkHEC1)
Modeled (Gamma)
Modeled (NRCS)
Incremental Rainfall
MRUH
ClarkHEC1 UH
Gamma UH
NRCS UH
(A)
(B)
Fig. 5.9 (A) Modified rational, Gamma, ClarkHEC1, and NRCS unit hydrographs
developed for the watershed associated with USGS streamflowgaging Station 08048520
Sycamore, Fort Worth, Texas; and (B) Rainfall hyetograph, observed and modeled runoff
hydrographs using the four different unit hydrographs for the rainfall event on
07/28/1973 for the same watershed.
151
1
10
100
M
o
d
e
l
e
d
Q
p
(
c
m
s
)
1
10
100
M
o
d
e
l
e
d
Q
p
(
c
m
s
)
1 10 100
Observed Q
p
(cms)
1 10 100
Observed Q
p
(cms)
ClarkHEC1 UH
MRUH
1:1 line
? 1/3 of a log cycle
NRCS UH
Ga mma UH
(A)
(C)
(B)
(D)
Fig. 5.10 Modeled versus observed peak discharges using: (A) MRUH, (B) Gamma UH,
(C) ClarkHEC1 UH, and (4) NRCS UH for 1,600 rainfallrunoff events in 90 Texas
watersheds.
152
0
10
20
30
40
D
i
s
c
h
a
r
g
e
(
c
m
s
)
0
4
8
12
0 2 4 6 8 10 12 14 16 18 20
Ti m e ( h r )
Observed
5min
10min
20min
30min
40min
50min
Incremental rainfall
I
n
c
r
e
m
e
n
t
a
l
r
a
in
f
a
ll
(
m
m
)
Fig. 5.11 Observed and modeled runoff hydrographs using MRUH with six unit
hydrograph durations (5, 10, 20, 30, 40 and 50 minutes) for the rainfall event on
07/28/1973 for the watershed associated with the USGS streamflowgaging station
08178600 Salado Creek, San Antonio, Texas.
153
Chapter 6 Conclusions and Recommendations
6.1 Conclusions
This research work is a part of TxDOT Project 06070 ?Use of the Rational and
the Modified Rational Methods for TxDOT Hydraulic Design?. The objective of the
project is to evaluate appropriate conditions for the use of the rational method and
modified rational methods for designs on small watersheds; evaluate and refine, if
necessary, current tabulated values of the runoff coefficient and construct guidelines for
TxDOT analysts for the selection of appropriate parameter values for Texas conditions.
The objective is achieved through four different phases of work.
For the first phase of our study, volumetric runoff coefficients were estimated for
90 Texas watersheds using three different methods?(1) a watershed composite, literature
based coefficient (C
lit
) was derived from landuse information for the watershed and
published C
lit
values for appropriate landuses, (2) backcomputed volumetric runoff
coefficient (C
vbc
) was estimated by the ratio of total runoff depth to total rainfall depth for
individual storm events and, (3) rankordered volumetric runoff coefficient (C
vr
) was
determined from the rankordered data; similar to the procedure used by Schaake et al.
(1967). The median value of C
lit
for developed watersheds exceeds that for undeveloped
watersheds. Watershedmedian values of C
vbc
for 45 developed watersheds in Texas with
percent imperviousness greater than 15% are consistent with median values of runoff
coefficient R
v
reported for 60 NURP watersheds by the USEPA.
154
The key conclusions of this phase of study are:
? Published runoff coefficients, even though they were not developed from
observed rainfallrunoff measurements and instead resulted from a survey on
engineering practices in 1950s, reflect the physical meanings of the original
runoff coefficients introduced by Kuichling in 1889 ? the runoff coefficient
is related to the percent impervious area within the watershed. Therefore,
published runoff coefficients remain useful for engineering design of drainage
systems.
? The distribution of C
vr
is different from that of C
lit
with about 80 percent of C
lit
value greater than C
vr
value. This result might indicate that literaturebased runoff
coefficients overestimate peak discharge for drainage design when used with the
rational method.
? Volumetric runoff coefficients are useful in transforming rainfall depth to runoff
depth such as is done in the curve number method (SCS 1963) and for watershed
rainfallrunoff modeling, e.g., the fractional loss model (McCuen 1998, p. 493).
For the second phase of our study, ratebased runoff coefficients, C
rate
, were
estimated using two different methods. First, C
rate
was estimated using rational equation;
the rainfall intensity was computed as the maximum intensity for a moving time window
of duration T
c
before and up to the time to peak, T
p
. Second, the frequencymatching
approach (Schaake et al., 1967) was used to extract a representative runoff coefficient
(C
r
) for each watershed. The key conclusions of this phase of study are:
155
? The ratebased C is dependent on rainfall intensity averaging time t
av
used for the
study, because estimates of the runoff coefficient based on observed data cannot
be decoupled from the selection of the timeresponse characteristic.
? The distributions of the watershedaverage and watershedmedian C
rate
are
similar. The C
r
values for the developed watersheds are consistently greater than
those for the undeveloped watersheds.
? The values of C
r
and C
rate
were compared with the literature based runoff
coefficients (C
lit
) developed from landuse data for these study watersheds. About
75 percent of C
lit
values are greater than C
r
.
? For typical applications of the rational method in urban watersheds, watershed C
lit
is less than C
r
; using smaller C
lit
would underestimate Q
p
for design.
For the third phase of our study, the runoff coefficients C(T) for different return
periods (T) were developed for the 36 undeveloped Texas watersheds using previously
published regional regression equations of peak discharge and countybased tabulated
empirical coefficients for a model of rainfall intensities at different T. The frequency
factors C
f
(T) = C(T)/C(10) determined in this study exceed those values in textbooks such
as Gupta (1989) and Viessman and Lewis (2003) and those from TxDOT (2002) when T
> 10 years. The key conclusions of this phase are:
? C(T) values increase with T and these increases are more than previously thought.
? The frequency factors determined for the 36 Texas watersheds and the 72 Kansas
watersheds (Young et al. 2009), larger than those mostly found in literature, are
for undeveloped watersheds with relatively small percent impervious areas.
156
? The frequency factors mostly found in the literature, smaller than those
determined for the 36 Texas watersheds, are appropriate for urban watersheds
with relatively large percentages of impervious area, as supported and presented
in literature (e.g., DRCG 1969; Stubchaer 1975; Jens 1979; Gupta 1989;
Viessman and Lewis 2003; TxDOT 2002).
? When the frequency factor is applied, if resulted C(T) is greater than unity, Jens
(1979), Gupta (1989) and TxDOT (2002) suggested setting C(T) equal to 1.
The modified rational method, MRM, is an extension of the rational method to
produce simple runoff hydrographs for applications that do not warrant a more complex
modeling approach. For the fourth phase of our study, the MRM was revisited. The
hydrographs developed using MRM can be considered an application of a special unit
hydrograph termed the modified rational unit hydrograph, MRUH. The MRUH method
was applied to develop unit hydrographs for 90 watersheds in Texas. Runoff coefficients
estimated using two approaches were examined for application with the MRUH. The first
was a watershed composite literaturebased coefficient (C
lit
) derived using the landuse
information for the watershed and published C
lit
values for various landuses. The second
was a backcomputed volumetric runoff coefficient (C
vbc
) determined by preserving the
runoff volume and using observed rainfall and runoff data. Times of concentration for
study watersheds were estimated by Fang et al. (2008) from four empirical equations,
which were based on several watershed characteristics. Simulated peak discharges and
times to peak from MRUH agree reasonably well with observed values. The drainage
area of the study watersheds (average 440 km
2
or 15.6 mi
2
) is greater than that usually
157
accepted for rational method application (0.8 km
2
or 0.3 mi
2
), yet results from the MRUH
reasonably approximate watershed behavior regardless of watershed size. Simulated peak
discharges are more sensitive to the choice of the runoff coefficient than the time of
concentration. Simulated times to peak are moderately sensitive to the time of
concentration but independent of the runoff coefficient. The MRUH is not sensitive to the
selection of the unit hydrograph duration so long as the same time interval is used for
hydrograph convolution. Three other unit hydrograph models, the Clark (using HEC?1?s
generalized basin equations), the Gamma, and the NRCS unit hydrographs were also used
to compute the direct runoff hydrograph for each rainfallrunoff event in the database.
Runoff hydrographs simulated using all four methods were similar. Simulated peak
discharges for all events in the database were similar regardless of statistical or
quantitative measures used for comparison. For time to peak, simulated values using all
four models agree reasonably well with observed values. The four UH models produce
similar values of statistical and quantitative measures for both peak discharges and time
to peak. Three key conclusions for MRUH are:
? Being a unit hydrograph, it can be applied to nonuniform rainfall distributions and
for watersheds with drainage areas greater than typically used with either the
rational method or the modified rational method (that is, a few hundred acres).
? The MRUH performs about as well as other unit hydrograph methods used in this
study for predicting the peak discharge and time to peak of the direct runoff
hydrograph, so long as the same rainfall loss model is used.
? Modeled peak discharges from application of the MRUH are more sensitive to the
selection of runoff coefficient, less sensitive to T
c
, and not sensitive to the
158
selection of the unit hydrograph duration. In predicting peak discharges and
runoff hydrographs for engineering design, rainfall loss estimation results in
greater uncertainty and contributes more model errors than variations of UH
methods and model parameters for UH.
6.2 Recommendations
Based on the analysis of the volumetric runoff coefficients for 45 developed
watersheds in Texas and 60 NURP watersheds, a polynomial regression equation was
recommended which can be used to estimate volumetric runoff coefficients for developed
urban watersheds that are similar to the 45 developed watersheds in Texas:
04.030.111.266.1
23
???? IMPIMPIMPC
v
(6.1)
Above equation is useful mainly for the urban (developed) watersheds. Further study is
recommended in future to correlate runoff coefficients for undeveloped watersheds to soil
types and other watersheds characteristics like slope, initial soil moisture condition and
landuse.
A single equation was recommended to estimate the ratebased runoff coefficient
C* (?Cstar?) for the unified rational method (URAT) developed for TxDOT:
15.085.0 ?? IMPC (6.2)
The above equation is consistent with the Kuichling?s original idea of the runoff
coefficient as the amount of imperviousness of the drainage area. Kuichling (1889)
concluded that the percentage of the rainfall discharged for any given watershed studied
is nearly equal to the percentage of impervious surface within the watershed. Several
researchers [Longobardi et al. (2003); Merz and Bl?schl (2009)] have shown that runoff
159
coefficients are strongly correlated with the initial soil moisture condition. With the
increase of the rainfall duration, the degree of land saturated also increases and the runoff
coefficient increases (or the imperviousness as proposed by Kuichling increases). So for
future study, it is recommended to study the variation of the runoff coefficient with the
degree of saturation of the land (temporal variation of C).
Many would argue that the application of the rational method is not appropriate
for the range of watershed areas presented in this study. The TxDOT guidelines
recommend the use of the rational method for watersheds with drainage areas less than
0.8 km
2
(200 acres) (TxDOT 2002). However our study showed that there is no
demonstrable trend in runoff coefficient with drainage area. We applied ratebased C to
estimate the peak discharge for the study watersheds and found out that the differences
between the observed and modeled Q
p
are generally within the expected errors from
typical hydrologic analysis. We do not advocate any specific limits that should be
imposed on drainage area for application of the rational method. However further study is
recommended in other watersheds and with more extensive database, to determine what
is the reasonable size that can be used with the rational method for the hydrologic design.
It is observed that application of the MRUH is simple and straightforward. Like
other UH methods, MRUH can be applied to large watersheds with nonuniform rainfall
distribution. However, using the runoff coefficient for the rainfall loss estimation doesn?t
account for the initial moisture condition of the watershed. We concluded that in
predicting peak discharges and runoff hydrographs for engineering design, rainfall loss
estimation results in greater uncertainty and contributes more model errors than
variations of UH methods and model parameters for UH. So for future study it is
160
recommended to incorporate the runoff coefficient with another loss parameter which
accounts for the initial moisture condition of the watershed for rainfall loss estimation in
application of the MRUH.
6.3 References
DRCG. (1969). ?Urban Storm Drainage Criteria Manual, Vols. 1 and 2.? Prepared by
WrightMcLaughlin Engineers for Denver Regional Council of Governments
(DRCG), Denver, CO.
Fang, X., Thompson, D. B., Cleveland, T. G., Pradhan, P., and Malla, R. (2008). ?Time
of concentration estimated using watershed parameters determined by automated and
manual methods.? Journal of Irrigation and Drainage Engineering, 134(2), pp. 202?
211.
Gupta, R. S. (1989). Hydrology and hydraulic systems, Prentice Hall, Englewood Cliffs,
New Jersey.
Jens, S. W. (1979). Design of Urban Highway Drainage, Federal Highway
Administration (FHWA), Washington DC.
Kuichling, E. (1889). ?The relation between the rainfall and the discharge of sewers in
populous areas.? Transactions, American Society of Civil Engineers 20, pp. 1?
56McCuen, R. H. (1998). Hydrologic analysis and design, PrenticeHall, Inc.,
Upper Saddle River, N.J.
Longobardi, A., Villania, P., Graysonb, R., and Westernb, A. (2003). "On the relationship
between runoff coefficient and catchment initial conditions." Proceedings of
MODSIM 2003, 867?872.
Merz, R., and Bl?schl, G. (2009). "A regional analysis of event runoff coefficients with
respect to climate and catchment characteristics in Austria." Water Resources
Research, 45(1), W01405.
Schaake, J.C., Geyer, J.C., and Knapp, J.W. (1967). ?Experimental examination of the
rational method.? Journal of the Hydraulics Division, ASCE, 93(6), pp. 353?370.
Stubchaer, J. M. ?The Santa Barbara urban hydrograph method.? National Symposium of
Urban Hydrology and Sediment Control, University of Kentucky, Lexington, KY.
TxDOT. (2002). Hydraulic design manual, The bridge division of the Texas Department
of Transportation (TxDOT), 125 East 11th street, Austin, Texas 78701.
161
Viessman, W., and Lewis, G.L. (2003). Introduction to hydrology, 5th Ed., Pearson
Education, Upper Saddle, N.J., 612 p.
Young, C. B., McEnroe, B. M., and Rome, A. C. (2009). ?Empirical determination of
rational method runoff coefficients.? Journal of Hydrologic Engineering, 14, 1283 p.