CHAPTER 3 METHOD OF ANALYSIS In this research the goal is to determine an IUH from observed rainfall-runoff data.

This research assumes that an IUH exists, and that it is the response function to a linear system, and the research task is to find the parameters (unkno n coefficients! of the transfer function. To accomplish this task a database must be assembled that contains appropriate rainfall and runoff values for analysis. "nce the data are assembled, the runoff signal is analy#ed for the presence of any base flo , and this component of the runoff signal is removed. "nce the base flo is removed, the remaining hydrograph is called the direct

runoff hydrograph ($%H!. The total volume of discharge is determined and the rainfall input signal is analy#ed for rainfall losses. The losses are removed so that the total rainfall input volume is e&ual to the total discharge volume. The rainfall signal after this process is called the effective precipitation. 'y definition, the cumulative effective

precipitation is e&ual to the cumulative direct runoff. If the rainfall-runoff transfer function and its coefficients are kno n a-priori, then the $%H signal should be obtainable by convolution of the rainfall input signal ith the

IUH response function. The difference bet een the observed $%H and the model $%H should be negligible if the data have no noise, the system is truly linear, and selected both the correct function and the correct coefficients. If the analyst postulates a functional form (the procedure of this thesis! then searches for correct values of coefficients, the process is called de-convolution. In the present ork by guessing at coefficient values, convolving the effective precipitation e have

((

signal, and comparing the model output

ith the actual output,

e accomplish de-

convolution. ) merit function is used to &uantify the error bet een the modeled and observed output. ) simple searching scheme is used to record the estimates that reduce the value of a merit function and hen this scheme is completed, the parameter set is

called a non-inferior (as opposed to optimal! set of coefficients of the transfer function. 3.1. Database Construction U*+* small atershed studies ere conducted largely during the period spanning the early (,-./s to the middle (,0./s. The storms documented in the U*+* studies can be used to evaluate unit hydrographs and these data are critical for unit hydrograph investigation in Texas. 1andidate stations for hydrograph analysis substantial database as assembled. Table 2.( is a list of the 33 stations eventually keypunched and used in this research. The first t o columns in each section of the table is the atershed and sub atershed name. The urban portion of the database does not use the sub atershed naming convention, but the rural portion does. The third column is the U*+* station I$ number. This number identifies the gauging station for the runoff data. The precipitation data is recorded in the same reports as the runoff data so this I$ number also identifies the precipitation data. The last numeric entry is the number of rainfall-runoff records available for the unit hydrograph analysis. The details of the database construction are reported in )s&uitn et. al (4..5!. ere selected and a

(4

Table 2.(.*tations and 6umber of *torms used in *tudy

Austin Watershed BartonCreek BartonCreek BearCreek BearCreek BearCreek Bo%%&Creek Bo%%&SouthCreek Bu''Creek *itt'eWa'nutCreek +nionCreek +nionCreek Shoa'Creek Shoa'Creek Shoa'Creek Shoa'Creek S'au%hterCreek S'au%hterCreek Wa''erCreek Wa''erCreek Wa'nutCreek Wa'nutCreek Wa'nutCreek Wa'nutCreek Wa'nutCreek WestBou'dinCreek Wi'bar%erCreek Wi''ia"sonCreek Wi''ia"sonCreek Wi''ia"sonCreek Sub-Shed Station ID 08155200 08155 00 08158810 08158820 08158825 08158050 08158880 0815#700 08158 80 08158700 08158800 0815$$50 0815$700 0815$750 0815$800 081588#0 081588$0 08157000 08157500 08158100 08158200 08158#00 08158500 08158$00 08155550 08152150 08158220 081582 0 08158270 #Events 5 8 8 2 2 10 1# 1 2 $ 2 1 1$ 1 2# 2 2 #0 8 15 17 10 1# 22 10 22 1# 18 1$ Watershed AshCreek Ba!h"anBran!h CedarCreek Coo"bsCreek CottonWoodCreek Du!kCreek E'a"Creek (ive)i'eCreek (ive)i'eCreek ('o&dBran!h ,oesCreek -e.tonCreek /rairieCreek 0ushBran!h South)es1uite South)es1uite S3ank&Creek 4urt'eCreek Wood&Bran!h Watershed Dr&Bran!h Dr&Bran!h *itt'e(ossi' *itt'e(ossi' S&!a"ore S&!a"ore S&!a"ore S&!a"ore Da''as Sub-Shed Station ID 08057 20 08055700 08057050 08057020 080571#0 080$1$20 08057#15 08057#18 08057#20 080571$0 08055$00 08057# 5 08057##5 080571 0 080$1220 080$1250 08057120 0805$500 08057#25 Station ID 080#8550 080#8$00 080#8820 080#8850 080#8520 080#85 0 080#85#0 SSSC #Events 5 #1 7 $ 8 8 7 10 8 1# 8 5 2 1 # #2 1 #Events 25 27 20 2# 2# 28 2# 21

(ort Worth Sub-Shed

Watershed A'a5anCreek *eonCreek *eonCreek *eonCreek +'"osCreek +'"osCreek +'"osCreek Sa'adoCreek Sa'adoCreek Sa'adoCreek Sa'adoCreek Sa'adoCreek

San Antonio Sub-Shed Station ID 08178 00 08181000 08181#00 08181#50 08177$00 08177700 08178555 08178$00 08178$#0 08178$#5 08178$20 081787 $

#Events 0 10 15 22 12 2 10 1 10 5 2 12

Watershed BrasosBasin BrasosBasin BrasosBasin BrasosBasin Co'oradoBasin Co'oradoBasin Co'oradoBasin Co'oradoBasin Co'oradoBasin SanAntonioBasin SanAntonioBasin SanAntonioBasin 4rinit&Basin 4rinit&Basin 4rinit&Basin 4rinit&Basin 4rinit&Basin 4rinit&Basin 4rinit&Basin 4rinit&Basin

S"a''0ura'Sheds Sub-Shed Station ID Co.Ba&ou 0802$800 6reen 0802#000 /ond-E'" 08028 00 /ond-E'" 08108200 Dee3 081 2000 Dee3 081#0000 )uke.ater 081 $200 )uke.ater 081 7000 )uke.ater 081 7500 Ca'averas 08182#00 Es!ondido 08187000 Es!ondido 08187200 E'"(ork 08050200 7one& 08057500 7one& 08058000 *itt'eE'" 08052$ 0 *itt'eE'" 08052700 -orth 080#2$50 -orth 080#2700 /in+ak 080$ 200

#Events #8 28 12 21 27 28 22 8 # 2# 1 21 # 1 22 22 58 1# 5$

3. . Data Pre!aration )n additional processing step used in this thesis is the interpolation of the observed data into uniformly spaced, one minute intervals.

(2

3. .1. "ase F#o$ Se!aration Hydrograph separation is the process of separating the time distribution of base flo from the total runoff hydrograph to produce the direct runoff hydrograph (7c1uen (,,3!. 'ase flo separation is a time-honored hydrologic exercise termed by hydrologists as 8one of the most desperate analysis techni&ues in use in hydrology9 (He lett and Hibbert (,-0! and 8that fascinating arena of fancy and speculation9 ()ppleby (,0.: 6athan and 7c7ahon (,,.!. Hydrograph separation is considered more of an art than a science ('lack (,,(!. *everal hydrograph separation techni&ues such as constant discharge, constant slope, concave method, and the master depletion curve method have been developed and used. ;igure 2.( is a sketch of a representative hydrograph that ill be used in this section to explain the different base flo separation methods.
$ischarge (<2=T!

Time (T!

;igure 2.( %epresentative Hydrograph

(5

1onstant-discharge method The base flo is assumed to be constant regardless of stream height (discharge!.

Typically, the minimum value immediately prior to beginning of the storm is pro>ected hori#ontally. )ll discharge prior to the identified minimum, as ell as all discharge

beneath this hori#ontal pro>ection is labeled as 8base flo 9 and removed from further analysis. ;igure 2.4 is a sketch of the constant discharge method applied to the

representative hydrograph. The shaded area in the sketch represents the discharge that ould be removed (subtracted! from the observed runoff hydrograph to produce a directrunoff hydrograph.
$ischarge (<2=T!

Time (T!

;igure 2.4. 1onstant-discharge base flo separation. The principal disadvantage is that the method is thought to yield an extremely long time base for the direct runoff hydrograph, and this time base varies from storm to storm, depending on the magnitude of the discharge at the beginning of the storm

(?

(<insley et, al, (,5,!. The method is easy to automate, especially for multiple peak hydrographs. 1onstant-slope method ) line is dra n from the inflection point on the receding limb of the storm hydrograph to the beginning of storm hydrograph, as depicted on ;igure 2.2. This method assumes that the base flo arbitrarily sets to the inflection point.
$ischarge (<2=T!

began prior to the start of the current storm, and

inflection point identified as location here second derivative of the hydrograph passes through #ero

Time (T!

;igure 2.2. 1onstant-slope base flo separation. The inflection point is located either as the location here the second derivative atershed area. This ill have

passes through #ero (curvature changes! or is empirically related to

method is also relatively easy to automate, except multiple peaked storms multiple inflection points.

(-

1oncave method The concave method assumes that base flo flo continues to decrease hile stream

increases to the peak of the storm hydrograph. Then at the peak of the hydrograph,

the base flo is then assumed to increase linearly until it meets the inflection point on the recession limb. ;igure 2.5 is a sketch illustrating the method applied to the representative hydrograph. This method is also relatively easy to automate except for multiple peak hydrographs hich, like the constant slope, method ill have multiple inflection points.
$ischarge (<2=T!

inflection point identified as location here second derivative of the hydrograph passes through #ero

Time (T!

;igure 2.5 1oncave-method base flo separations $epletion curve method This method models base flo as discharge from accumulated ground ater

storage. $ata from several recessions are analy#ed to determine the basin recession constant. The base flo is modeled as an exponential decay term qb (t ! = qb ,o exp(kt ! . The time constant, k, is the basin recession coefficient that is inferred from the recession portion of several storms.

(0

Individual storms are plotted

ith the logarithm of discharge versus time. The

storms are time shifted by trial-and-error until the recession portions all fall along a straight line. The slope of this line is proportional to the basin recession coefficient and the intercept ith the discharge axis at #ero time is the value for q b ,o . ;igure 2.? ith a test storm here the base flo separation is

illustrates five storms plotted along being determined. The storm

ith the largest flo

at the end of the recession is plotted ere

ithout any time shifting. The recession is extrapolated from this storm as if there

no further input to the ground ater store. The remaining storms are time shifted so that the straight line portion of their recession limbs come tangent to this curve. 'y trial-anderror the master depletion curve can be ad>usted and the storms time shifted until a reasonable agreement of all storms recessions ith the master curve is achieved.
100

Dis!har%e 8!:s9

10

1 0 100 200 00 4i"e 8hours9 )aster;De3'etion;Curve #;11;2$ 2;22;#0 4est;Event #;1;#$ #;11;#1 ;2 ; 7 #00 500 $00

;igure 2.? 7aster-$epletion 1urve 7ethod ($ata from 7c1uen, (,,3, Table ,-4, pp 53-!

(3

"nce the master curve is determined, then the test storm is plotted on the curve and shifted until its straight-line portion come tangent to the master curve, and the point of intersection is taken as the base flo value for that storm. In the example in ;igure

2.?, the base flo for the test event is approximately 9.1 cfs, the basin recession constant is 0.0045/hr, and the base flo at the beginning of the recession is 17 cfs. "nce the base flo value is determined for a particular test event, then base flo separation proceeds

use the constant discharge method. The depletion curve method is attractive as it determines the basin recession constant, but it is not at all easy to automate. ;urthermore, in basins here the stream

goes dry (such as much of Texas!, the recession method is difficult to apply as the first storm after the dry period starts a ne master recession curve. "bserve in ;igure 2.? the storms used for the recession analysis span a period of nearly 5. years, and implicit in the analysis is that the basin recession constant is time invariant and the storms are independent. The follo ing ;igure 2.- is a multiple peak storm event from $allas )sh1reek station.3.?024.. To automate the rest of data set using this method because of the change of master recession curve for different peaks. ill be a challenge

(,

File : #IUH_1_sta08057320_1977_0327.dat Dallas AshCre ek

A!!u"< De3th 8in!hes9 2<50E-02

2<00E-02

1<50E-02

1<00E-02

5<00E-0

0<00E =00 0 500 1000 4i" e 8"inutes9 1500 2000 2500

#0A 4E;/0E CI/

#0A 4E;0>-+((

;igure 2.- 7ultiple peak storms from $allas module Se#ection o% Met&o' to E(!#o) The principal criterion for method selection that as based on the need for a method

as simple to automate because hundreds of events needed processing. )ppleby separation techni&ue based on a %icatti-type e&uation for e&uation is a rational functional that is

((,0.! reports on a base flo

base flo . The general solution of the base flo

remarkably similar in structure to either a <a@lace transform or ;ourier transform. Unfortunately the paper omits the detail re&uired to actually infer an algorithm from the solution, but it is useful in that principles of signal processing are clearly indicated in the model. 6athan and 7c7ahon ((,,.! examined automated base flo techni&ues. The ob>ective of their ork separation

as to identify appropriate techni&ues for

determination of base flo

and recession constants for use in regional prediction ere a smoothed minima techni&ue and ork!. 'oth

e&uations. T o techni&ues they studied in detail

a recursive digital filter (a signal processing techni&ue similar to )pplebyAs techni&ues

ere compared to a graphical techni&ue that extends pre-event runoff (>ust ith the point of greatest curvature on the

before the rising portion of the hydrograph!

4.

recession limb (a constant-slope method, but not aimed at the inflection point!. They concluded that the digital filter as a fast ob>ective method of separation but their paper

suggests that the smoothed minima techni&ue is for all practical purposes indistinguishable from either the digital filter or the graphical method. ;urthermore the authors ere vague on the constraint techni&ues employed to make the recursive filter values and to produce peak values that did not exceed the

produce non-negative flo

original stream flo . @ress et.al. ((,3-! provide convincing arguments against timedomain signal filtering and especially recursive filters. 6evertheless the result for the smoothed minima is still meaningful, and this techni&ue appears fairly straightfor ard to automate, but it is intended for relatively continuous discharge time series and not the episodic data in the present application. The constant slope and concave methods are not used in this ork because the

observed runoff hydrographs have multiple peaks. It is impractical to locate the recession limb inflection point ith any confidence. The master depletion curve method is not

used because even though there is a large amount of data, there is insufficient data at each station to construct reliable depletion curves. %ecursive filtering and smoothed minima ere dismissed because of the type of events in the present continuous!. Therefore in the present discharge method. The constant discharge method as chosen because it is simple to automate and ork (episodic and not

ork the discharge data are treated by the constant

apply to multiple peaked hydrographs. @rior researchers (e.g. <aurenson and "A$onell, (,-,: 'ates and $avies, (,33! have reported that unit hydrograph derivation is insensitive to base flo separation method hen the base flo is not a large fraction of

4(

the flood hydrograph B a situation satisfied in this in this research determined discharge before that time average rate

ork. The particular implementation

hen the rainfall event began on a particular day: all as accumulated and converted into an average rate. This as

as then removed from the observed discharge data, and the result

considered to be the direct runoff hydrograph. The candidate models ill be run in t o cases ith or ithout base flo

separation, so one can compare ho prediction. 3. . . E%%ecti*e Preci!itation

much the separation

ould effect the runoff

The effective precipitation is the fraction of actual precipitation that appears as direct runoff (after base flo separation!. Typically the precipitation signal (the

hyetograph! is separated into three parts, the initial abstraction, the losses, and the effective precipitation. Initial abstraction is the fraction of rainfall that occurs before direct runoff. "perationally several methods are used to estimate the initial abstraction. "ne method is to simply censor precipitation that occurs before direct runoff is observed. ) second method is to assume that the initial abstraction is some constant volume (Ciessman, (,-3!. The 6%1* method assumes that the initial abstraction is some fraction of the maximum retention that varies ith soil and land use (essentially a 16 based method!. <osses after initial abstraction are the fraction of precipitation that is stored in the atershed (depression, interception, soil storage! that does not appear in the direct runoff hydrograph. Typically depression and interception storage are considered part of the initial abstraction, so the loss term essentially represents infiltration into the soil in the

44

atershed. *everal methods to estimate the losses includeD @hi-index method, 1onstant fraction method, and infiltration capacity approaches (HortonAs curve, +reen-)mpt model!. @hi-index model The -index is a simple infiltration model used in hydrology. The method assumes that the infiltration capacity is a constant E(in=hr!. Fith corresponding observations of a rainfall hyetograph and a runoff hydrograph, the value of can in many cases be easily guessed. ;ield studies have sho n that the infiltration capacity is greatest at the start of a storm and that it decreases rapidly to a relatively constant rate. The recession time of the infiltration capacity may be as short as (. to (? minutes. Therefore, it is not unreasonable to assume that the infiltration capacity is constant over the entire storm duration. Fhen the rainfall rate exceeds the capacity, the loss rate is assumed to e&ual the constant capacity, hich is called the phi (! index. Fhen the rainfall is less

than the value of , the infiltration rate is assumed to e&ual to the rainfall intensity. 7athematically, the phi-index method for modeling losses is described by ;(t!G I(t!, for I(t! H ;(t!G ,for I(t!I , (2.(! (3.2)

here ;(t! is the loss rate, I(t! is storm rainfall intensity, t is time, and is a constant. If measured rainfall-runoff data are available, the value of can be estimated by separating base flo from the total runoff volume, computing the volume of direct

runoff, and then finding the value of that results in the volume of effective rainfall being e&ual to the volume of direct runoff. ) statistical mean phi-index can then be 42

computed as the average of storm event phi values. Fhere measured rainfall-runoff data are not available, the ultimate capacity of HortonAs e&uation, fc, might be considered. HortonAs model Infiltration capacity (fp! may be expressed as fp G fc J (fo B fc! e-Kt, (2.2!

here fo G maximum infiltration rate at the beginning of a storm event and reduces to a lo and approximately constant rate of fc as infiltration process continues and the soil is saturated K G parameter describing rate of decrease in fp. ;actors assumed to be influencing infiltration capacity, soil moisture storage, surface-connected porosity and effect of root #one paths follo the e&uation f G a*a(.5J fc, here f G infiltration capacity (in=hr!, a G infiltration capacity of available storage ((in=hr!=(in!(.5! (Index of surface connected porosity!, *a G available storage in the surface layer in inches of ater e&uivalent ()-hori#on in agricultural soils - top six inches!. ;actor fc G constant after long etting (in=hr!. The modified Holton e&uation used by U* )gricultural %esearch *ervice is f G +Ia *a(.5 Jfc, (2.?! (2.5!

here +I G +ro th index - takes into consideration density of plant roots hich assist infiltration (... - (..!. +reen-)mpt 7odel

45

+reen L )mpt ((,((! proposed the simplified picture of infiltration sho n in ;igure 2.0.

;igure 2.0.Cariables in the +reen-)mpt infiltration model. The vertical axis is the distance from the soil surface: the hori#ontal axis is the moisture content of the soil. (*ourceD Applied Hydrology by 1ho =7aidement=7ays (,33! The wetting front is a sharp boundary dividing soil belo ith moisture content i

from saturated soil done ith moisture content i above. The etting front has penetrated to a depth L in time t since infiltration began. Fater is ponded to a small depth h. on the soil surface. The method computes total infiltration rate at the end of time t, ith the follo ing e&uation, ;(t! G Mt J here M G Hydraulic conductivity, NO lnP ( J ;(t!=( NO!Q, (2.-!

4?

t G time in hrs, ;(t! G Total infiltration at the end of time t, ! " Fetting front soil suction head, and NO G increase in moisture content in time t. Unlike the *1* curve method, this method gives the total amount of infiltration in the soil at the end of a particular storm event. $epending on this value and the total amount of precipitation, e can easily calculate the amount of runoff. 1onstant ;raction 7odel The constant fraction model simply assumes that some constant ratio of precipitation becomes runoff: the fraction is called a runoff coefficient. )t first glance it appears that it is a rational method disguise, but the rational method does not consider storage and travel times. Thus in the rational method, if one doubles the precipitation intensity, and halved the duration, one ould expect the peak discharge to remain unchanged, hile in a unit hydrograph such changes should have a profound effect on the hydrograph. )s a model, the method is simple to apply, essentially
p e (t ! = %rp R p raw (t !

Ap
(2.0! here %rp G the runoff coefficient,

(t !dt = #$H (t ! dt

pe G the effective precipitation,

praw G the ra

precipitation,

) G drainage area.

4-

The first e&uation states that the effective precipitation is a fraction of the ra precipitation, hile the second states that the total effective precipitation volume should e&ual the total direct runoff volume. 3.3. Su((ar) o% Data Pre!aration 'ase flo separation as accomplished using the constant discharge method Fe analy#ed the data ith and ithout a Sffective

because it is amenable to automation. separation to test precipitation hether separation

as necessary in our data set.

as al ays modeled using the constant fraction model, because of the need as

to automate and also because of the sheer magnitude of the dataset, but the fraction

left as a fitting constant. Ideally, the fitted result should preserve the re&uired mass balance (precipitation volume G runoff volume!. )n important detail in this research 8pseudo data9 for IUH analysis. minute. This time length as the conversion of the original data into as one-

The time-step length used in the research

as chosen because it is the smallest increment that can be

represented in the current $)TSTTI7S format in the database. It should be noted that there are very fe actual one-minute intervals in the original data, so linear interpolation

as used to convert the cumulative precipitation into one-minute intervals, then numerical differentiation is performed to obtain the rainfall rates. The resulting units are inches per minute. ;igure 2.3 is a sketch sho ing the incremental rate and the cumulative depth relationship. The cumulative depth scale is the left vertical scale and the incremental rate scale is the right vertical scale. 7athematically the cumulative rainfall distribution is the integral of the incremental rainfall distribution (or rainfall density! over the entire rainfall

40

event. S&uation 2.3 expresses this relationship: integration over the entire number line is intended to indicate the entire lifetime of the individual rainfall event.
& (t ! = p (t ! dt

(2.3!

0<07 0<07 2 2 0<0$ 0<0$ 0<05 0<05 0<0# 0<0# 0<0 0<0 0<02 0<02 0<5 0<5 0<01 0<01 0 0 0 0 0 0 100 100

1<5 1<5 Cu"u'ative /re!i3<8in<9

1 1

2020

#0#0

$0$0 4i"e 4i"e 8"in<9 8"in<9

8080

A!!/re!i3

Inst/re!i3

;igure 2.3. 1umulative @recipitation and Incremental @recipitation %elationship In ;igure 2.3 the cumulative precipitation, &'t(, is indicated by the open circles, hile the rate, p't!, is indicated by the open s&uares. In practice only the cumulative depth is recorded as a function of time: so to determine the rate the cumulative precipitation.
p (t ! = d& (t ! d = P p (t ! dtQ . dt dt

e simply differentiate

/re!i3< 0ate 8in<?"in<9

2<5 2<5

0<08 0<08

(2.,!

The present

ork used a simple centered differencing scheme, except at the first

and last time interval, hen for ard and back ard differencing ere used, respectively.
p (t ! & (t + t ! & (t t ! . 4t

(2.(.!

43

$etails of the 8pseudo data9 conversion ere reported by 1leveland et. al, (4..2!. The (-minute data for roughly (-54 storms are located on a University of Houston server and can be publicly accessed at the U%< associated ith this citation. 3.+. NRCS ,nit H)'ro-ra!& The 6atural %esources 1onservation *ervice (6%1*!, formerly the *oil 1onservation *ervice, developed a unit hydrograph (UH! in the (,?.s. This UH as used to develop storm hydrographs and peak discharges for design of conservation measures on small agricultural atersheds. 7ockus ((,?-! discussed development of the standard 6%1* unit hydrograph and the peak rate e&uation, &pGM)U=Tp, (2.((!

here the peak discharge rate &p is a function of drainage area ), direct runoff volume U, factor M, and time to peak of the unit hydrograph Tp. He indicated that the peak rate factor (@%;! of M is e&ual to MG(4,..-=((JH!, (2.(4!

here H is the ratio of the time of recession to the time peak (Tr= Tp!. He also indicated that M as a function of the UH shape and that 2=3 of the storm runoff volume in the rising limb and ?=3 in the recession limb ere typical of small agricultural atersheds. M also includes a conversion factor to make the e&uation dimensionally correct. 7ockus used the triangular UH shape in development of above t o e&uations. It appears that 7ockus analy#ed many flood hydrographs to >ustify the selection of the peak rate factor M of 535. ) UH ith @%; of M of 535 as felt to be representative of small agricultural atersheds in the U.*.

4,

NRCS.D,H /0a((a a!!ro1i(ation2 The 6%1* $imensionless Unit Hydrograph (U*$), (,3?! used by the 6%1* (formerly the *1*! as developed by Cictor 7ockus in the late (,5.As. The *1* atersheds of different si#es and in

analy#ed a large number of unit hydrographs for

different locations and developed a generali#ed dimensionless unit hydrograph in terms of t/tp and q/qp here, tp is the time to peak. The point of inflection on the unit graph is as observed to be ..4 the base

approximately (.0 the time to peak and the time to peak time (hydrograph duration! ()b!.

The functional representation is presented as tabulated time and discharge ratios, and as a graphical representation. Table 2.4 is the tabulation of the 6%1* $UH from

the 6ational Sngineering Handbook. Table2.4. %atios for dimensionless unit hydrograph and mass curve
Time ratios (t=Tp! ... ..( ..4 ..2 ..5 ..? ....0 ..3 .., (.. (.( (.4 (.2 (.5 (.? (.(.0 (.3 (., 4.. 4.4 $ischarge ratios (&=&p! ... ...2 ..(. ..(, ..2( ..50 ..-..34 ..,2 ..,, (... ..,, ..,2 ..3..03 ..-3 ..?..5..2, ..22 ..43 ..4.0 7ass 1urve %atios (U=Up! ..... ....( .......(4 ...2? ...-? ..(.0 ..(-2 ..443 ..2.. ..20? ..5?. ..?44 ..?3, ..-?. ..0.. ..0?( ..0,. ..344 ..35, ..30( ..,.3

2.

4.5 4.4.3 2.. 2.4 2.5 2.2.3 5.. 5.? ?..

..(50 ..(.0 ...00 ...?? ...5. ...4, ...4( ...(? ...(( ....? .....

..,25 ..,?2 ..,-0 ..,00 ..,35 ..,3, ..,,2 ..,,? ..,,0 ..,,, (....

-0CS Di"ension'ess >nit 7&dro%ra3h and )ass Curve

1 0<2 0<8 0<7 0<$ 1?13 0<5 0<# 0< 0<2 0<1 0 0 1 2 t?t3 Dis!har%e 0atios 81?139 )ass Curve 0atios 8@?@39 # 5 $

1 0<2 0<8 0<7 0<$ 0<5 0<# 0< 0<2 0<1 0 @?@a

;igure2.,. @lot of $UH and 7ass 1urve ;igure 2., is a plot of these ratios. This figure is identical to ;igure (-.( in the 6ational Sngineering handbook (except this figure is computer generated!.

2(

The IUH analysis assumed that the hydrograph functions are continuous and the database as analy#ed using discrete values calculated from continuous functions. %ather than use the 6%1* tabulation in this ork, the fit as tested of a function of the same family (the gamma distribution! as the IUH function and this function as used in place of the 6%1* tabulation. ) similar approach as used by *ingh (4...! to express common unit hydrographs (*nyderAs, *1*, and +rayAs! by a gamma distribution. The gamma function used to fit the tabulation is
&* ( + ! = k * (e * = (! .

(2.(2! The variables , and k are unkno n, and ere determined by minimi#ation of the sum of s&uared errors bet een the tabulation and the model (the function! by selection of numerical values for the unkno n parameters. 8Sxcel solver9 minimi#ation. The values for parameters , and k as used to perform the

ere 2.33, 5.3( and (.4,

respectively. *o the 6%1* $UH approximation is

&* ( + ! =(.4, R 2.33 5.3( * 2.3(e 2.33 *

(2.(5!

;igure 2.(. is a plot of the model and tabulation, the variable * in the e&uations is the dimensionless time. Uualitatively the fit is good. The maximum residual(s! occur early in dimensionless time and at -.V of the runoff duration, but the magnitudes are &uite small, and thus this model of the 6%1* $UH is deemed acceptable for use.

24

-0CS Curve-(ittin% >sin% 6a""a :un!tion 1 0<2 0<8 0<7 0<$ 1?13 0<5 0<# 0< 0<2 0<1 0 0 1 2 t?43 +bserved )ode'ed # 5 $

;igure2.(.. @lot of Tabulated and +amma-7odel $UH ) 1hi-s&uare fitness test as performed to further support the decision to use the

model in lieu of the tabulation. The test statistic for the chi-s&uare test as calculated as

%4 = (-i ,i ! 4 = ,i .
i =(

(2.(?!

The test statistic is ..?242. ;or t o degrees of freedom and ,. V confidence limits the value as (..hich is greater than the test statistic (..?242! therefore the hypothesis

(model! represent the observed values. The 6%1* $UH as presented in the 6SH integrates to a little over (.5 and thus it is not a true unit hydrograph as presented. It is likely that it originally as a UH: then it

as ad>usted procedurally so that the peak value of the dimensionless distribution is (.. (thus the factor that scales the integral correctly is imbedded in the & p value!. The research assumes that all unit graphs and the accompanying functional representations of

22

IUHs integrate to one: so in this

ork the 6%1* $UH approximation is ad>usted by

dividing by the integral of the original $UH, in this case the value is (.4,.2. Therefore the final approximation to the 6%1* $UH as a functional representation useful in IUH analysis is
&* ( + ! = 2.33 5.3( * 2.3(e 2.33 * .

(2.(-! "r ith all the constants evaluated and simplified and expressed in the 6%1* terminology the 6%1* $UH (as an IUH function! is
2.33 q (t ! t tp = 23.?230( ! 2.3( e . qp tp t

(2.(0!

3.3. Co((ons H)'ro-ra!& 1ommons ((,54! developed a dimensionless unit hydrograph for use in Texas, but details of ho the hydrograph ere developed are not reported. The labeling of axes in the original document suggests that the hydrograph is dimensionless. ;or the sake of completeness in this ork, an approximation as produced for treatment as another ere

transfer function by fitting a three-gamma summation model. Sssentially there three integrated gamma models ith different peaks and

eights to reproduce the shape

of 1ommonsA hydrograph. The 1ommons hydrograph is &uite different in shape after the peak than other dimensionless unit hydrographs in current use (i.e. 6%1* $imensionless Unit Hydrograph! B it has a very long time base on the recession portion of the hydrograph.

25

;igure 2.((. Hydrograph developed by trial to cover a typical flood 1ircles are tabulation from digiti#ation of the original figure. 1urve is a *mooth 1urve )pproximation. ;igure 2.(( is 1ommonsA hydrograph reproduced from a manual digiti#ation. The smooth curve is given by the follo ing e&uation that as fit by trial and error.
00...( t 5.0.0 t p q (t ! = ( ! ..(0- e (..((3! 5.0.0t p t

0.?3 t 4.-,5 t p ( ! ..,-? e (..,4?! 4.-,5t p 2.33 t ?.-5(t p ( ! ..(24 e (..433! ?.-5(t p t

. (2.(3!

The numerical values are simply the result of the fitting procedure. The time axis as reconstructed (in the fitting algorithm! so that the tp parameter could be left variable for consistency ith the other hydrograph functions. The tabulated function integrates to

2?

approximately ((-.: thus the function above is divided by this value to produce a unit hydrograph distribution. 3.4. 0a((a S)nt&etic H)'ro-ra!&s The gamma distribution is given in the e&uation
f ( * ! = .e * = b * a .

(2.(,!

In the e&uation 1 e&uals

a+ (

( to make the area enclosed by the curve e&ual to ( a +(!

unity. is called the gamma function. Calues can be found tabulated in mathematical handbooks. The gamma distribution is similar in shape to the @oisson distribution that is

given the form as f ( * ! =

/ * e / .The curve starts at #ero *W

hen the variable x is #ero,

rises to a maximum, and descends to a tail that extends indefinitely to the right. The values that the variable x can take on are thus limited by . on the left. Calues can extend to infinity on the right. The gamma distribution differs from the @oisson distribution is that it has t o parameters instead of the single parameter of the @oisson. This allo s the curve to take on a greater variety of shapes than the @oisson distribution. The parameter a is a shape parameter hile b is a scale parameter. The shape of the +amma distribution is similar to the shape of a unit hydrograph, so many researchers started looking for the application of the +amma distribution into hydrograph prediction. This first started ith Sdson ((,?(!, ho presented a theoretical expression for the unit hydrograph assuming U to be proportional to t * e yt

2-

0=

%Ay ( yt ! * e yt , ( * + (!

(2.4.!

here UG discharge in cfs at time t: )G drainage area in s&uare miles: x and y G

( * +(! is the parameters that can be represented in terms of peak discharge: and

gamma function of (xJ(!. 6ash ((,?,! and $ooge ((,?,!, based on the concept of n linear reservoirs ith e&ual storage coefficient M, expressed the instantaneous UH (IUH! in the form of a +amma distribution as
q= ( t 1( n! 1
n (

e t = k ,

(2.4(!

in hich n and MG parameter defining the shape of the IUH: and &Gdepth of runoff per unit time per unit effective rainfall. These parameters have been referred to as the 6ash model parameters in the subse&uent literature. @revious attempts to fit a +amma distribution to a hydrograph ere by 1roley((,3.!, )ron and Fhite ((,34!, Hann et al. ((,,5!, and *ingh ((,,3!. The procedure given by 1roley ((,3.! to calculate n for kno n values of &p and tp re&uires programming to iteratively solve for n. 1roley also proposed procedures to obtain a UH from other observable characteristics. The method by )ron and Fhite ((,,4! involves reading the values from a graph, in hich errors are introduced. 'ased on their methods, 7c1uen ((,3,! listed a step-by-step procedure to obtain the UH, hich maybe briefly described by the follo ing e&uations, nG(..5?J..?fJ?.-f4J..2f2, (2.44!

20

in hich

f =

0 pt p A

, here Up is in cubic feet per second, tp is in hours, and ) is in

acres. These t o e&uations re&uire careful attention for the units, and these cannot be used as such hen Uptp is re&uired to be computed for a value of n kno n from other sources. Hann et al. ((,,5! gave the follo ing expression to calculate n,
n = ( + -.?( 0 pt p 2 !(.,4 ,

(2.42!

here CGtotal volume of effective rainfall. )n e&uation provided by *ingh ((,,3! to obtain the value of n may be ritten,
n =(.., + ..(-5 + -.(, 4 ,

(2.45!

here = q p t p (dimensionless!, in hich q p is the peak runoff depth per unit time per effective rainfall. *ingh observed that the error in n obtained from the e&uationD
n =(.., + ..(-5 + -.(, 4

(2.4?!

is ..?2V hen = ..4? and ...?V hen =(.. . The error in n calculated decreases ith increasing values of . 3.5. 6eibu## Distribution Historically a t o-parameter Feibull distribution is employed to define the configuration of a natural hydrograph of direct runoff and is given in the follo ing forms as (1anavos, (,35!
0 = 3t n (e ( t = k ! ,
n

(2.4-!

here U is the discharge ordinate of the natural hydrograph corresponding to the time t after the commencement of direct runoff, n is the dimensionless shape factor, and k is the

23

storage time constant. 'oth n and k reflect the basin characteristics and are related to the time to peak tp in the follo ing manner.
(t p = k ! n = (n (! = n .

(2.40!

The constant of proportionality ' in S&uation (2.4-! is evaluated as

3= 0p (t p ! n ( e
( t p = k ! n

(2.43!

here Up is the peak discharge and e is the base of the natural logarithms. 1ombining S&uation (2.4-!, (2.40! and (2.43! yields
0 = 0 p = (t = t p ! n ( e
( n (!((( t = t p ! n ! = n

(2.4,!

S&uation (2.4,! is the desired form of the dimensionless Feibull distribution as used in this study. )nalytical formulation of the parameter n can be developed as follo s. $esignating qR = 0 = 0 p , and t R = t = t p , S&uation (2.4,! may be ritten as

qR = (t R ! n ( e ( n (!((tR

!=n

(2.2.!

Taking natural logarithms of both sides of the above e&uation and solving for n, one obtains
n = ln(n ln t R +( (n (! ln qR = n! = ln t R .

(2.2(!

The value of n can be obtained from S&uation (2.2(! through graphical means. "nce the value of n has been ascertained properly, the value of k can then be determined from S&uation (2.2(! conveniently. 3.7. Reser*oir E#e(ents )n alternative ay to construct the hydrograph functions is to model the The

atershed response to precipitation as the response from a cascade of reservoirs.

response function is developed as the response to an impulse of input, and the response to 2,

a time series of inputs is obtained from the convolution integral. The end result is the same, a function that is a distribution function, but the parameters have a physical interpretation. The kernel (response function! to an impulse in this ork is an

instantaneous unit hydrograph (IUH!. reservoirs is ell studied and orks

The conceptual approach for a cascade of

ell for many unit hydrograph analyses (e.g. 6ash, ork the cases are examined

(,?3: $ooge, (,?,: $ooge, (,02: 1roley, (,3.!. In this

here a +amma, %ayleigh, and Feibull distribution govern the individual reservoir element responses, respectively. In addition, e have also converted the 6%1*-$UH

into its o n response model (a special case of gamma!.

e*%e44 pre%ipitation 'depth( A

51,t q1,t

56,t q6,t water4hed 4y4te/ re4pon4e /odel

57,t q7,t 58,t q8,t ob4er9ed dire%t r:noff 'depth(

;igure 2.(4 1ascade of %eservoir Slements 1onceptuali#ation ;igure 2.(4 is a schematic of a identical reservoirs atershed response conceptuali#ed as a series of of each reservoir is related to the

ithout feedback. The outflo

5.

accumulated storage in the reservoir. The behavior of the individual reservoir elements determines hether the model becomes a +amma, %ayleigh, or Feibull distribution. ;igure 2.(2 is a schematic of a reservoir response element. In the sketch, the element

area is A, the accumulated excess storage is 5, and the outlet flo area is a. pe

A a q"a9

;igure 2.(2 %eservoir Slement 7odel The outlet discharge is the product of the outlet area, a, and the flo input is pe. 3.7.1 Linear /0a((a2 Reser*oir E#e(ent The first response model is a linear reservoir model, here the reservoir discharge is proportional to the accumulated depth of input. The constant of proportionality is %. The discharge e&uation is velocity 9. The

q = a9 = a%5 .
) mass balance of the reservoir is
A d5 = a%5 . dt

(2.24!

(2.22!

The input pe is applied over a very short time interval: so the resulting depth, before outflo begins is 5o. The solution to this "$S ("rdinary $ifferential S&uation! is
5 (t ! = 5 . exp( a% t! . A

(2.25!

The ratio A/a% is called the residence time of the linear reservoir.

5(

t =

A . a%

(2.2?!

Thus in terms of residence time the accumulated depth in a linear reservoir is

t 5 (t ! = 5 . exp( ! . t

(2.2-! The discharge rate is the product of this function and the constant of proportionality
t A t q (t ! = a%5 . exp( ! = 5 . exp( ! . t t t

(2.20!

;igure 2.(5 <inear %eservoir 7odel t = 4 , AG(, 50G1 This particular atershed model has the follo ing propertiesD 1umulative discharge is related to accumulated time. Instantaneous discharge is inversely related to accumulated time. The peak discharge is proportional to the precipitation input depth, and occurs at time #ero. The peak discharge is proportional to the atershed area.

54

3.7. Ra)#ei-& Reser*oir E#e(ent The next response model assumes that the discharge is proportional to both the accumulated excess precipitation (linear reservoir! and the elapsed time since the impulse of precipitation as added to the atershed (translation reservoir!. The constant of

proportionality in this case is 6%.

q = a9 = a 4%5t .

(2.23!

) mass balance for this model is

A d5 = a 4%5t . dt

(2.2,!

The solution (using the same characteristic time re-parameteri#ation as in the linear reservoir model! is
t 5 (t ! = 5 . exp(( ! 4 ! . t

(2.5.!

The discharge function is

t 4t t q (t ! = a 4%t5 . exp(( ! 4 ! = A5 . 4 exp( ( ! 4 ! . (2.5(! t t t

This result is a %ayleigh distribution

eighted by the product of

atershed area

and the initial charge of precipitation (hence the name %ayleigh reservoir!. The discharge function for unit area and depth integrates to one: thus it is a unit hydrograph, and it satisfies the linearity re&uirement, thus it is a candidate IUH function.

52

;igure 2.(?. %ayleigh %eservoir Fatershed 7odel. t = 4 , AG(, 50G1 "f particular interest, the %ayleigh model &ualitatively looks like a hydrograph should, ith a peak occurring some time after the precipitation is applied (unlike the

linear reservoir! and a falling limb after the peak ith an inflection point. Sxamination of the discharge function includes the follo ing relationshipsD 1umulative discharge is proportional to accumulated time. Instantaneous discharge is proportional to accumulated time until the peak, then inversely proportional after ards. The peak discharge is proportional to the precipitation input depth, and occurs at some non-#ero characteristic time. The peak discharge is proportional to the atershed area.

3.7.3 6eibu## Reser*oir The Feibull response model assumes that the discharge is proportional to both the accumulated excess precipitation (linear reservoir! and the elapsed time raised to

55

some non-#ero po er since the impulse of precipitation

as added to the

atershed

(translation reservoir!. The constant of proportionality in this case is p%.

q = a9 = ap%5t p ( .

(2.54!

) mass balance for this model is

A d5 = ap%5t p ( . dt

(2.52!

The solution (using the same characteristic time re-parameteri#ation as in the linear reservoir model! is
t 5 (t ! = 5 . exp(( ! p ! . t

(2.55!

The discharge function is

t pt p ( t q (t ! = ap%t p ( 5 . exp(( ! p ! = A5 . exp(( ! p ! . p t t t

(2.5?!

This result is a Feibull distribution

eighted by the product of

atershed area

and the initial charge of precipitation (hence the name Feibull reservoir!. The discharge function for unit area and depth integrates to one, thus it is a unit hydrograph, and it satisfies the linearity re&uirement, thus it is a candidate IUH function. These three models constitute the reservoir element models used in this research. 3.8. Casca'e Ana#)sis ;igure 2.(4 is the schematic of a cascade model of research atershed response. In our atershed could

e assumed that the number of reservoirs 8internal9 to the

range from . to J . "ur initial theoretical development assumed integral values, but others have suggested fractional reservoirs can be incorporated into the theory. To develop the cascade model(s!, start ith the mass balance for a single reservoir element, e

and the discharge from this reservoir becomes the input for subse&uent reservoirs and 5?

determine the discharge for the last reservoir as representative of the entire response. 3.8.1. 0a((a Reser*oir Casca'e S&uation 2.5-,

atershed

here 5i represents the accumulated storage depth, a% is the for a particular reservoir, and A is the

reservoir discharge coefficient, qi is the outflo

atershed area, represents the discharge functions for a cascade of linear reservoirs that comprise a response model. The subscript, i , is the identifier of a particular reservoir in the cascade.
Aqi ,t = a%5 i ,t .

(2.5-!

S&uation 2.50 is the mass balance e&uation for a reservoir in the cascade. In S&uation 2.5-, the first reservoir receives the initial charge of ater, 5o over an

infinitesimally small time interval, essentially an impulse, and this impulse is propagated through the system by the drainage functions.
i ,t = Aq i (,t a%5 i ,t A5

(2.50! The entire atershed response is expressed as the system of linear ordinary

differential e&uations, S&uation 2.53, and the analytical solution for discharge for this system for the 8-th reservoir is expressed in S&uation 2.5,.

5-

( = 5 4 = 5 2 = 5 8 = 5

5o

( 5( t ( 5( t

( 54 t ( 54 t

( 52 t ( 5 8 ( t ( 58 t

(2.53!

The result in e&uation 2.53 is identical to the 6ash model (6ash (,?3! and is incorporated into many standard hydrology programs such as the 1"**)%% model (%ock ood et. al. (,04!. The factorial can be replaced by the +amma function (6auman and 'uffham, (,32! and the result can be extended to non-integer number of reservoirs.
t 8 ( t ( exp( ! . q 8 ,t = A5 . 8 ( t t ( 8 (!Wt

(2.5,! To model the response to a time-series of precipitation inputs, the individual responses (S&. 2.5,! are convolved and the result of the convolution is the output from the atershed. If each input is represented by the product of a rate and time interval

(5o't( " qo't( dt! then the individual response is (note the +amma function is substituted for the factorial!
8 ( t ( (t ! exp( dqi , = Aq. ( ! ! d . 8 ( t t ( 8 !t

(2.?.!

The accumulated responses are given by

t 8 ( t ( (t ! exp( q 8 (t ! = Aq. ( ! ! d . 8 ( t t ( 8 ! t .

(2.?(!

50

S&uation 2.?( represents the

atershed response to an input time series. The of that ork, is

convolution integral in 1hapter 0 in 1ho , et al ((,33!, an overvie repeated as S&uation 2.?4,

0 (t ! = ; ( !: (t ! d .
. t

(2.?4!

The analogs to our present ork are as follo s (1ho As variable list is sho n on the left of the e&ualities!D
0 (t ! = q 8 (t ! ; ( ! = q . ( !
8 ( t ( (t ! exp( : (t ! = A ! 8 ( t t ( 8 !t

(2.?2! Fe call the kernel ( :'t <( ! for the linear reservoir a gamma response because the kernel is essentially a gamma probability distribution. The reason for representing the function as being derived from a cascade is that this derivation provides a 8physical9 meaning to the distribution parameters. The analysis is repeated for the %ayleigh and Feibull distributions. 3.8. . Ra)#ei-& Reser*oir Casca'e ) %ayleigh response is developed in the same fashion as the gamma, except the %ayleigh reservoir element is used instead of the linear (gamma! response. The discharge and mass balances for the %ayleigh case are given as S&uations 2.?5 and 2.??, respectively,
Aq i ,t = 4a%t5 i ,t ,

(2.?5! (2.??!

i ,t = Aq i (,t 4a%t5 i ,t . A5

53

The entire

atershed response is expressed as the system of linear ordinary

differential e&uations in S&uation 2.?-.

( = 5 4 = 5 2 = 5 8 = 5 5o t 4 5( t t 4 5( t

t 4 54 t t 4 54 t

t 4 52 t t 4 5 8 ( t t 4 58 t

(2.?-!

The analytical solution for any reservoir is expressed in S&uation 2.?0.

t q 8 ,t = 4 A5 . 4 t
4 8 ( t 4 (t ! ( 8 !(t 4 ! 8 ( exp(( t ! ! ,

(2.?0!

S&uation 2.?3 gives the convolution integral using this kernel.

4 8 ( (t ! 4 t ((t ! ! qi (t ! = 4 Aq . ( ! 4 exp( !d , 4 8 ( t4 t ( 8 !(t ! . t

(2.?3!

This distribution is identical to <einhardAs 8hydrograph distribution9 (<einhard, (,04! that he developed from statistical-mechanical analysis. 3.8.3. 6eibu## Reser*oir Casca'e ) Feibull response is developed in the same fashion as the gamma by substitution of the Feibull reservoir element in the analysis. The discharge and mass balances are given as S&uations 2.?, and 2.-., respectively,
Aq i ,t = 4a%t5 i ,t ,

(2.?,! (2.-.!

i ,t = Aq i (,t 4a%t5 i ,t . A5

The entire

atershed response is again expressed as a system of linear ordinary

differential e&uations: S&uation 2.-(.

5,

( = 5 4 = 5 2 = 5 8 = 5

5o

p p

t p ( 5( t

t p ( 5( t p p

t p ( 54 t

t p ( 54 t p

t p ( 52 t t p ( 58 t

(2.-(!

p (

5 8 (

The analytical solution to this system for any reservoir is expressed in S&uation 2.-4,
t p ( (t p ! 8 ( t p q 8 ,t = pA5 . tp ( 8 !(t p ! 8 ( exp(( t ! ! .

(2.-4!

The accumulated responses to a time series of precipitation input are given by S&uation 2.-2.
t (t ! p ( ((t ! p ! 8 ( (t ! p qi (t ! = pAq . ( ! exp( ! d . p ( 8 !t 8 ( t t .

(2.-2! The utility of the Feibull model is that both the linear cascade (exponential! and the %ayleigh cascade are special cases of the generali#ed Feibull model, thus if a Feibull-type model as the IUH, e program

e can investigate other models by restricting

parameter values. The parameters have the follo ing impacts on the discharge functionD (. The po er term controls the decay rate of the hydrograph (shape of the falling limb!. If p is greater than one, then decay is fast (steep falling limb!: if p is less than one then the decay is slo (long falling limb!. 4. The t term controls the scale of the hydrograph. It simultaneously establishes the location of the peak and the magnitude of the peak.

?.

2. The reservoir number, 8, controls the lag bet een the input and the response, as ell as the shape of the hydrograph. The next chapter describes ho observations. the distribution parameters are determined from

?(