Академический Документы
Профессиональный Документы
Культура Документы
Paper: A07
1 Introduction
One of the most significant parameters characterizing runoff from urban catchments is
a hydraulic width of its subcatchments. Two approaches to the width evaluation
described in literature were extended by an original authors approach. The presented
paper contains a comparative study of specified approaches performed for a real
420
urban catchment in the city of Poznan. The study is based on theoretical analysis and
comparison of hydrographs presenting measured and simulated variation from the
considered catchment. Hydrographs were developed for three historical rainfalls. A
set of special indicators was used for comparison of both hydrographs and assessing
correctness of fit for the considered approaches.
(1)
5
Q = B
s 3
h
n
(2)
dh
B s 3
= q
h
123
dt
1A
42
4 n 43
4
123
retention
inf low
(3)
outflow
421
2.2 Evaluation of Hydraulic Width Based on Dimensions of the Particular
Subcatchments Option 1
For a subcatchment which is symmetric in respect of its main collector (Fig. 1a) it is
assumed that a hydraulic width B is two times larger than the length of the main
collector Lcol of subcatchment [3]:
B = 2 L col
(4)
(5)
Hence an average length of runoff path Lav can be computed as a ratio of A and B
and expressed as a function of b:
L av =
A b L col b
=
=
B 2 L col 2
(6)
For an asymmetric subcatchment or by its irregular shape (Fig. 1b) the hydraulic
width is computed as a function of skewness coefficient Sk. The latter is defined as
the ratio of subcatchment areas difference on both sides of collector to total area of
subcatchment:
SK =
A 2 A1
A 2 + A1
(7)
The average length Lav of runoff path is obtained from the expression:
(8)
422
L av =
b L col
A
b
=
=
B (2 S K ) L col (2 S K )
(9)
L
b
L av = col + c
2
16
(10)
B=
A imperv
L av
(11)
423
3 Case Study
3.1 Characteristics of the Piasnica Catchment
Analyzed in the case study catchment of Piasnica collector is situated in the city of
Poznan in the area of Rataje (Fig. 2a). Its surface measures almost 700 ha. It is mainly
covered by dwelling houses built in the second half of the last century (blocks of flats
and some villas), two large industrial quarters (a brewery and a paper mill), a
shopping center and a hospital. Runoff from catchment is collected by storm sewer
network and transported to Cybina River at cross-section downstream from the Malta
Lake.
3.2 Computer Model of the Catchment
Simulation of runoff from the Piasnica Catchment was performed using the
stormwater model SWMM5. For computation of the surface runoff a model of
kinematic wave was applied, while flow in drainage network was computed using a
model of dynamic wave (the simplified version of the full Saint Venant equation).
The data for construction of the Piasnica Catchment model was taken from the plans
of drainage network in scale 1:500, longitudinal profiles of storm sewers, and
additionally from topographic maps in scale 1:10.000 and 1:25.000.
By modeling drainage network only larger sewers (with diameters of 500 mm and
larger) were considered. The model of the flow through smaller storm sewers was
replaced by a surface runoff model. This simplification resulted from the lack of
424
complete drainage network documentation and was justified by its small influence on
characteristics of outflow (depth and discharge) at the outlet from the exemplary
catchment. Altogether 82 stretches of storm sewers and 55 subcatchments were taken
into consideration by modeling of the catchment (Fig. 2b).
Its total area was evaluated as 700 ha; an average percentage of impervious
surfaces were assessed as 29% but it was evaluated separately for each individual
subcatchment. A slope of specific subcatchment was identified with an average slope
of the terrain determined along its main collector. The roughness of storm sewers was
computed on the basis of depth and discharge measurements more specifically using
measured hydrographs of these variables [12]. The values of remaining parameters
used in modeling of an exemplary catchment were taken from literature [3, 4] and are
presented in Table 1.
Table 1. The values of parameters used in modeling Piasnica catchment.
Parameter
Slope of catchment (-)
Manning roughness coefficient of surface (s/m1/3)
Surface retention (mm)
Maximum infiltration (mm/h)
Minimum infiltration (mm/h)
Infiltration constant (h-1)
Manning roughness coefficient of sewers (s/m1/3)
Type of surface
impervious
pervious
0.005
0.015
0.100
1.5
5.0
50
1.5
4
0.018
425
4 Scope of Analysis
4.1 Selection of a Hydraulic Width Evaluation Option
Two options of hydraulic width evaluation were compared in the study:
- Option 1 (based on specific dimensions of subcatchment) with the assumption of
subcatchment symmetry with respect to its main collector,
- Option 3 (based on equal route of runoff for all subcatchments) for different
values of this characteristic.
An analysis of subcatchment delimitation procedure reveals that they are usually
traced along the largest neglected storm sewers (of diameters smaller than 500 mm).
The width of these sewers subcatchments (denoted by b) doesnt exceed their length
Lcol i.e. b< Lcol. Taking into account this relation it can be shown (by comparing Eq.
(6) and (10) that the average length Lav of runoff path computed according to Option 2
is always longer than the one computed according to Option 1. The relations between
hydraulic width of subcatchments are reciprocal (Eq. 6). Hence the hydraulic width of
subcatchment computed according to Option 1 can be treated as the special case of
the width computed according to Option 2. It justifies neglecting Option 2 in the
further analysis.
In Option 3 of catchment width evaluation it was assumed that an average length of
runoff path can be approximated by a distance between runoff inlets. According to
literature [1, 2] this distance should not exceed 25 m. Taking into consideration
remarks enclosed in section 2.4 one can expect that an average length of runoff path
assumed in the model should be longer. Therefore in this study the runoff path of 25
m and its multiplications: 50, 75 and 100 m were considered.
4.2 Measurement Data
Three rainfall events selected from historical series were used in analysis. The
maximum depth of water measured in the main collector of Piasnica catchment due to
the above specified rainfalls was located in the range 50-80% of a collector height.
Table 2. The depth of rainfall used in the analysis
Rainfall
Date
A
B
C
25.06.2007
07.11.2007
12.04.2008
Station A
[mm]
21.4
18.2
15.0
Station B
[mm]
29.2
22.0
13.0
426
Measured volume of runoff was determined in an indirect way using results of
continuous measurements of outflow from the analyzed catchment in the main
collector.
Table 3. Comparison of measured and computed runoff volumes.
Rainfall
A
B
C
Measured
volume of
runoff
[m3]
41900
34300
25700
Computed
volume of
runoff
[m3]
49200
39100
27200
Relative
error of
volume
[%]
17
14
6
The computed runoff volumes presented in Table 3 are larger than respective
measured values. It can be explained by the assumption that a runoff from pervious
surface does not participate in the total outflow from a catchment. The difference
between computed and measured values of runoff, which can be measured by a
relative error of volume is directly proportional to the duration time of rainfall, the
number of peaks in hydrograph and the distance between them. The influence of these
factors can be explained by neglecting hydraulic losses (wetting losses, surface
retention) and evaporation in modeling. This situation results in an increase of a
computed rainfall volume and consequently in a positive value of its relative error.
For instance, for a rainfall A characterized by a relatively long duration time and
several peaks separated by interruptions of rainfall, participation of hydraulic losses
and evaporation in total volume of runoff is significant and a generates high value of a
relative error.
A relative error can be additionally influenced by neglecting spatial variability of
the rainfall in computation. The depth of the rainfall measured at station A and B is
substituted by an average value which is used for the whole catchment. In the
presented analysis the largest differences in rainfall depth measurements were
observed for rainfall A, which probably was the reason of the high value of a relative
error.
4.3 Criteria of Results Assessment
For comparing the results of simulations with measurements of outflow the following
indicators advised in literature [6, 7] were used in this study:
- ISE
integrated square error
- AVR
average values ratio
-R
linear correlation coefficient
- RS
special correlation coefficient;
- R2
model efficiency by Nash & Sutcliffe [5]
- Qmax
relative error of maximum flow
Indexes m and s denote respectively measurements and simulations. Symbol n
refers to the number of points used for description of hydrographs. Ordinates of them
are denoted by Qmax
427
Depending on the values of the considered indicators the model can be classified as
belonging to one of the five classes specified in Table 4.
Table 4. Classification of the model quality based on the values of selected indicators [7]
Class of model
excellent
very good
good
fair
poor
For indicators not specified in the head of the table i.e. AVR, R2 and Qmax
described above classification cannot be applied. Instead of that the extreme values
responding to ideal agreement of hydrographs are defined. They are equal to 1.0 for
AVR, R2 and for Qmax.
[%]
[-]
[-]
[-]
[-]
[%]
Rainfall
A
0.853
0.979
0.885
0.969
0.759
-12.8 (0.1)
B
0.913
0.992
0.903
0.973
0.777
-0.1
C
1.045
0.953
0.988
0.987
0.949
-5.4
428
Fig.3. Changes of model quality indicators as a function of average runoff path length
For this rainfall the highest differences between rainfall measurements on station A
and B were observed (among all considered in analysis rainfall events). Due to the
presence of two comparable maxima of the measured value of outflow, the indicator
Qmax was computed two times) for each peak separately) and both values were
shown in Table 5. The highest values of indicators were obtained for the rainfall C
indicating the smallest spatial variation.
For all three rainfalls the computed maximum runoff values were smaller than the
respective measured values. These differences were enclosed in an acceptable interval
of 15% [4]. An evaluation of hydraulic width according to Option 2 resulting in
smaller values of subcatchments hydraulic width in comparison with Option 1, would
increase the difference between computed and measured values of discharge and
depth and make the values of indicators worse.
5.2 Simulations for Option 3 of Hydraulic Width Evaluation
The largest number of the best indicators was obtained for the runoff path length Lav=
50 m (for rainfalls A and C) and Lav=75 m for the rainfall B. An analysis of the results
presented in Fig. 3 leads to the conclusion that the majority of the indicators
considered (ISE, AVR, R and RS) are dependent to a small degree on the runoff path
length. Only R2 and the relative error of maximum runoff Qmax are more sensitive to
this parameter. Generally one can conclude that a change of the runoff path length has
a relatively small influence on the results of runoff simulations.
429
6 Conclusions
The results of runoff simulations from the exemplary Piasnica catchment obtained for
Option 1 of subcatchment hydraulic width evaluation are comparable with the results
for option 3. If referred mainly to the shape of hydrographs (mainly the sequence of
peak flows), the computed peak flows are generally a bit lower than the respective
measurement results. Taking into consideration the fact that the outflow from
catchment is computed from kinematic wave equations, peak flows are directly
proportional to the square root of the surface slope and inversely proportional to the
Manning surface roughness coefficient. A better adjustment of peaks would require
the slope increasing of the subcatchment or a reduction of the Manning coefficient.
Changing of the third basic model parameter subcatchment hydraulic width in
option 1cannot be taken into consideration, because the hydraulic width is determined
by the length of the main collector Lcol (Eq. 4). In case of Option 3 the highest values
of indicators are obtained for runoff path length from the interval 50 -75 m. For a
catchment with comparable to Piasnica catchment parameters and drainage network
simplifications (neglecting storm sewers of D<500 mm) computations of hydraulic
width of subcatchments can base on equal value of average runoff path selected from
this interval.
Assuming that the sequence of peak flows depends on the values of three basic
parameters specified above i.e. slope s, roughness coefficient n and subcatchment
width B or runoff path Lav a computed length of runoff satisfies the condition.
0.005
s
0.005
0.015 75 n L av 0.015 50
(12)
Based on Eq. (12) a range of computed length of runoff path can be computed for
other catchments characterized by two remaining parameters (slope s and surface
roughness coefficient n). A sensitivity of SWMM model to changes of subcatchment
hydraulic width depends on parameters of rainfall [9]. One value of hydraulic width
assuring compatibility of computed and measured hydrographs does not exist.
Acknowledgments
This work has been financially supported by the Ministry of Science and Higher
Education of Polish Government (Grant No. 1253/H03/2006/30).
References
1. Edel R.: Drainage of highways (in Polish), WKL, Warsaw (2000)
2. Gudelis-Taraszkiewicz K., Suligowski Z.: An alternative management of rainfall waters.
Vademecum for businessmen (in Polish), part C (2007)
3. Huber W. C., Dickinson R.E.: Storm Water Management Model, version 4: Users Manual,
US EPA, Athens, Georgia (1992)
4. James W.: Rules for responsible modeling, CHI, Guelph, Ontario (2003)
430
5. Nash J. E., Sutcliffe J. V.: River flow forecasting through conceptual models. Part I a
discussion of principles, Journal of Hydrology, 10 (1970) 282-290
6. Ozga-Zieliska M., Brzeziski J.: Applied hydrology (in Polish), PWN, Warszawa (1997)
7. Sarma P. B. S., Delleur J. W., Rao A. R.: Comparison of rainfall-runoff models for urban
areas, Journal of Hydrology, 18 (1973) 329-347
8. Sieker F.: Arbeitsbericht der ATV/DVWK-Arbeisgruppe 1.2.6 Hydrologie der
Stadtenwsserung: Die Berechnung des Oberflchenabflusses in Kanalnetz Modellen, Teil
II-Abfluskonzentration, Korrespondenz Abwasser, 3 (1987) 263-269
9. Skotnicki M.: A comparative sensitivity analysis of selected runoff models for urban
catchment (in Polish), II Seminar Hydraulic of transient systems in environmental
engineering, Gdansk University of Technology, Gdansk, 21.10.2005, Journal no 2 (2005)
43-51
10.Skotnicki M., Sowinski M.: Measurements of rainfall and runoff for urban catchment
modeling (in Polish) Hydraulic measurements in environmental engineering, Gdansk
University of Technology, Gdansk, 7.04.2007, Journal no 4 (2007a) 63-71
11.Skotnicki M., Sowinski M.: Spatial rainfall variability in runoff modelling from urban
catchment (in Polish), In: Modelling of hydrological processes, Namyslowska-Wilczynska
B. (edit.), Wrocaw (2008)
12.Skotnicki M., Sowinski M.,: The investigation of storm channel roughness coefficient
variability based on flow measurement (in Polish), I National Conference INFRAEKO
2008, Rzeszow-Paczoltowice (2008) 193-204
13.Zaghloul N. A.: SWMM Model and Level of Discretization, Journal of the Hydraulics
Division, ASCE, vol. 107, 11 (1981) 1535-1545