Vous êtes sur la page 1sur 11

933

Expanded analytical probabilistic stormwater


models for use in watershed and master drainage
planning
Yiping Guo and Jianping Dai
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

Abstract: In watershed and master drainage planning studies, information on flood frequency distributions must be avail-
able for locations throughout the study area and for conditions with and without planned stormwater control facilities. De-
sign storm modeling and continuous simulation approaches are used to obtain the required information. Analytical
probabilistic stormwater models are computationally more efficient alternatives but they were developed originally for ur-
ban stormwater management planning and design purposes. In this paper, methods and strategies are developed to expand
the capabilities of existing analytical probabilistic stormwater models so that they can also be used in watershed and mas-
ter drainage planning studies. The case study tests and validates the proposed methods and strategies and demonstrates the
versatilities of the new approach.
Key words: design storms, continuous simulation, flood frequency distribution, watershed planning, master drainage plan,
probabilistic models.
Résumé : Dans les études des bassins versants et de planification du drainage principal, de l’information sur les distribu-
tions de la fréquence des inondations doit être disponible pour les emplacements dans toute la zone à l’étude et pour les
conditions avec et sans installations planifiées de contrôle des eaux pluviales. La modélisation basée sur les averses typi-
ques et les approches de simulation continue sont utilisées pour obtenir l’information requise. Les modèles analytiques pro-
For personal use only.

babilistes des eaux pluviales sont des choix plus efficaces au point de vue calculs, mais ils ont été initialement développés
pour la planification et la conception de la gestion des eaux pluviales en milieu urbain. Cet article développe des méthodes
et des stratégies visant à étendre les capacités des modèles analytiques probabilistes des eaux pluviales existants afin qu’ils
puissent également être utilisés dans les études des bassins versants et de planification du drainage principal. L’étude de
cas vérifie et valide les méthodes et stratégies proposées et démontre la polyvalence de la nouvelle approche.
Mots-clés : averses typiques, simulation continue, distribution de la fréquence des inondations, planification des bassins
versants, plan de drainage principal, modèles probabilistes.
[Traduit par la Rédaction]

1. Introduction tershed plans, master drainage plans are developed to ad-


dress the current and future drainage needs of a specific
Watershed or subwatershed planning is a process that uses
drainage area. Master drainage planning is the preferred
the watershed or subwatershed drainage area as the natural
mechanism for the planning and design of urban drainage
boundary to establish plans and regulations for managing
systems to minimize the impacts of urban stormwater runoff
human uses of the river and connected wetlands, woodlands,
on receiving watercourses (MOE and MNR 1993b). One of
valley lands, and floodplains. An ecosystem-based water-
the main tasks of watershed and master drainage planning is
shed planning process has been advocated in Ontario by the
to sketch out and optimize the flood, erosion, and water
Ministry of the Environment (MOE) and the Ministry of
quality control systems throughout a study area. Flood fre-
Natural Resources (MNR) for more than 15 years (MOE
quency distributions at locations throughout a watershed are
and MNR 1993a). Often guided by watershed or subwa-
important information required for both watershed and mas-
ter drainage planning. Design storm-based single event rain-
Received 13 March 2008. Revision accepted 23 February 2009. fall–runoff modeling and continuous simulation modeling
Published on the NRC Research Press Web site at cjce.nrc.ca on are the two traditional methods of flood frequency estima-
9 June 2009. tion for the purpose of watershed and master drainage plan-
Y. Guo.1 Department of Civil Engineering, McMaster ning.
University, 1280 Main St. West, Hamilton, ON L8S 4L8, The assumption invoked when using the design storm ap-
Canada. proach is that a design storm of a given return period will
J. Dai. Ganaraska Region Conservation Authority, 2216 County produce a runoff peak of the same return period. Despite
Road 28, P.O. Box 328, Port Hope, ON L1A 3W4, Canada. the fact that this assumption is not proven to be always ac-
Written discussion of this article is welcomed and will be ceptable (Adams and Howard 1986), the consensus of the
received by the Editor until 31 October 2009. engineering community seems to be that the design storm
approach can produce peak discharges of desired return peri-
1Corresponding author (e-mail: guoy@mcmaster.ca). ods with acceptable levels of accuracy, if it is used properly

Can. J. Civ. Eng. 36: 933–943 (2009) doi:10.1139/L09-037 Published by NRC Research Press
934 Can. J. Civ. Eng. Vol. 36, 2009

(Packman and Kidd 1980; Marsalek and Watt 1984; Levy fall event characteristics follow exponential PDFs. The exact
and McCuen 1999). Continuous simulation models use performance of APSWM is location-dependent, detailed
long-term precipitation records as inputs to generate contin- rainfall analysis and experimental applications are necessary
uous series of flows. Frequency analyses on the generated to determine the suitability of APSWM for a specific region.
flow series are conducted to estimate flood peaks of differ- The APSWM was originally developed for use in urban
ent return periods. Although continuous simulation can pro- stormwater management analyses, it has not been tested spe-
vide more accurate estimates of flood frequency cifically for rural areas and applied to large watersheds.
distributions, the data and computation requirements limit However, in watershed and master drainage planning stud-
its application in practice. ies, the subject watershed or drainage area may be medium
A new approach — the analytical probabilistic approach or large in size (i.e., much larger than the catchment area
for flood frequency estimation — was developed to over- draining to a typical urban stormwater detention pond) and
come the problems of both the design storm and the contin- may include portions that are entirely rural. The main objec-
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

uous simulation approaches (Guo and Adams 1998a, tive of this study is to expand the existing capabilities of
1998b). Rather than employing design storms, the analytical APSWM so that it can also be used in watershed and master
probabilistic approach estimates flood frequency distribu- drainage planning.
tions directly using closed-form analytical equations devel-
oped based on the derived probability distribution theory. 2. Expansion of the capability of the
The probability density functions (PDFs) of rainfall event
characteristics (i.e., event volume, duration, and inter-event analytical probabilistic approach
time) were identified first. The functional relationships of 2.1. Development of a probabilistic rainfall areal
rainfall event and runoff event characteristics (i.e., peak reduction method
flow, runoff volume, runoff duration) were then established The analytical probabilistic approach uses probability
to derive the PDFs of runoff event characteristics. The re- models of storm event characteristics to represent the rain-
sults are closed-form analytical equations that can be used fall conditions of a location. To generate the PDFs of storm
for flood frequency estimation. event characteristics, continuous rainfall series at a specific
To expand the utility of the analytical probabilistic ap- location is separated into individual storms. A minimum
For personal use only.

proach, Guo and Zhuge (2008) derived additional equations time period without rainfall, referred to as the inter-event
for the probabilistic routing of floods through channel time definition (IETD), is used as the separation criterion.
reaches and detention ponds. For ease of reference, the ana- To ensure the statistical independence between consecutive
lytical equations derived in Guo and Adams (1998a, 1998b, storms, a 12 h IETD was found appropriate for most loca-
1999a, 1999b) and Guo and Zhuge (2008) are referred to as tions (Restrepo-Posada and Eagleson 1982). When the sepa-
the analytical probabilistic stormwater models (APSWM). rated storm (rainfall) event series are obtained, the
The accuracy of APSWM has been verified by performing characteristics including total event volume v, duration t,
comparisons with the continuous simulation approach for a and inter-event time b for individual rainfall events can be
number of test catchments in Toronto, Ontario (Guo and determined and statistically analyzed. For many locations, it
Adams 1998a, 1998b, 1999a, 1999b); comparisons between was found that exponential PDFs provide good fits to the
the analytical probabilistic approach, the design storm ap- histograms obtained from the frequency analyses conducted
proach, and the continuous simulation approach for an urban for each of the event characteristics (Adams and Papa 2000).
test catchment in Chicago, Illinois, USA (Guo 2001); and The exponential distributions for rainfall event character-
comparison between the analytical probabilistic approach istics are expressed as
and the design storm approach for a design case in King-
ston, Ontario (Quader and Guo 2006). Rivera et al. (2005) ½1 fV ðvÞ ¼ xexv ; v0
applied APSWM for two locations: Fort Collins, Colorado,
USA; and Santiago, Chile. The APSWM results were com-
pared with continuous simulation results. Their comparison ½2 fT ðtÞ ¼ lelt ; t0
studies showed that the APSWM results are fairly close to
continuous simulation results for Fort Collins. However, for ½3 fB ðbÞ ¼ jejb ; b0
Santiago, the comparisons are not as good as those for Fort
Collins. The authors pointed out that the dependence be- The annual average number of events, denoted as q, can
tween rainfall event volume and duration caused the poor also be determined from the previously described process of
performance of APSWM in Santiago and suggested methods analysis. For a specific geographic location, the four param-
to overcome this problem. eters x, l, j, and q need to be known to use the probability
The computation involved in APSWM is much less than models to describe local rainfall characteristics. The values
that involved in the design storm approach. Since accurate of the three distribution parameters (i.e., x, l, and j) may
tracing of the probability distribution transformation from be estimated as the inverse of, respectively, the average
rainfall to runoff is achieved through the use of the derived rainfall event volume, the average rainfall event duration,
probability distribution theory, results of APSWM are com- and the average of inter-event times. For locations through-
parable with those from continuous simulation followed by out Canada, these rainfall statistics may be obtained from
frequency analyses. Thus, APSWM has a great potential in Adams and Papa (2000).
becoming an efficient alternative to both design storm and The previously outlined probabilistic description of local
continuous simulation models for many regions where rain- rainfall characteristics has so far been used for point rainfall

Published by NRC Research Press


Guo and Dai 935

Fig. 1. Rainfall event volume frequency distributions affected by the same probability distribution as that followed by point
different areal reduction factors. rainfall events. Areally averaged inter-event times are simi-
lar to rainfall durations, therefore, eqs. [2] and [3] will still
be the same after areal reduction. The annual average num-
ber of rainfall events is also the same before and after areal
reduction.
In the study by Yoo et al. (2007), the same type of mixed
Gamma distributions was used for both point rainfall and
areal average rainfall. Sivapalan and Bloschl (1998) show
that the upper tails of the Gamma distributions (which are
followed by areal average rainfall) may be approximated by
exponential functions. Supported by these findings, it is as-
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

sumed here that the areally averaged rainfall event volumes


follow the same type of exponential distributions, as ex-
pressed in eq. [1]. Thus, to achieve proper areal reduction
of point rainfall, the same type of probability models used
only. To extend the application of APSWM to large water- to describe point rainfall events can still be used to describe
sheds, some form of rainfall areal reduction must be in- areally averaged rainfall events and x is the only parameter
cluded in the rainfall input. In using the design storm that needs to be adjusted.
approach, an areal reduction factor (ARF) is used to reduce
the point rainfall depth to its areal average. For a given Let the PDF of the areal average rainfall event volume be
storm duration, ARF is calculated as the quotient of the ½4 fVa ðva Þ ¼ xa exa va ; va  0
average of annual maxima of areal rainfall and the average
of the annual maxima of point rainfall (U.S. Weather Bu- where va is the areal average rainfall event volume and xa is
reau 1957). It is generally assumed that the ARFs so deter- the distribution parameter. Since the areal average rainfall
mined are independent of the return periods of storms and event duration is the same as the corresponding point rain-
For personal use only.

geographic locations. fall event duration and follows the same probability distri-
Although recent studies (Allen and DeGaetano 2005; bution, t will still be used to denote areal average rainfall
Veneziano and Langousis 2005; Yoo et al. 2007) all suggest event duration, and l the distribution parameter. The same
that ARFs are dependent on the frequency of occurrence of is true for b and q. Assume that rainfall event volume and
rainfall and geographic location, for engineering applica- duration are statistically independent (Guo and Adams
tions, the most common source of ARF is still technical pa- 1998a), the probability Pr per rainfall event with T £ t and
per 29 (TP-29), Rainfall intensity-frequency regime – Part 1 Va > va, where t and va are the specific values of, respec-
(U.S. Weather Bureau 1957). For planning and design pur- tively, random variables T (areal rainfall event duration)
poses, a suitable design storm duration is selected first. and Va (areal rainfall event volume), can be calculated as
Based on this selected duration and catchment area, the re- ½5 PrfðT  tÞ \ ðVa > va Þg ¼ PrfT  tgPrfVa > va g
quired ARF is determined from the set of curves in TP-29.
The same ARF is applied to design storms of any return pe-
riod of interest. For APSWM applications, a method of areal ¼ ½1  expðltÞexpðxa va Þ
reduction that would result in similar degrees of reduction of
point rainfall is needed. Since consecutive rainfall events are statistically inde-
pendent as ensured by the use of an appropriate IETD, the
The currently accepted areal rainfall reduction procedure
annual probability of having a rainfall event with duration
implies that rainfall durations themselves are not affected in
less than or equal to t and volume greater than va is q times
the areal averaging process, or that the duration of an areally
the exceedance probability per rainfall event, as expressed in
averaged rainfall event is approximately the same as that of eq. [5]. Converting the annual exceedance probability to re-
the corresponding event observed at a point. This may be turn period Tr, we have
supported by the following reasons. First, if a storm is lo-
cally generated and subsides locally (e.g., convective 1
storms), the areally averaged event duration is the same as ½6 Tr ¼
PrfðT  tÞ \ ðVa > va Þgq
the point-observed event, averaging of event volume is re- 1
quired because of the uneven distribution of rainfall depths. ¼
Second, if a storm moves across the catchment (e.g., frontal q½1  expðltÞexpðxa va Þ
storms), the start and end times of the event as observed at To calculate the areal average rainfall depth va corre-
different points in the catchment may be different, the dura- sponding to a specific return period Tr and duration t,
tions of the same meteorological event observed at different eq. [6] is rearranged as
points should still be approximately the same. Averaging of
the rainfall event depths for the latter type of storms does 1
not affect the true duration of the same event observed at ½7 va ¼ fln½1  expðltÞ þ lnq þ lnTr g
xa
different points, the areally averaged event should have the
same duration as its corresponding point-observed events. Equation [7] shows that va for a given return period Tr
Hence, the areally averaged rainfall event durations follow and duration t is inversely proportional to the value of xa.

Published by NRC Research Press


936 Can. J. Civ. Eng. Vol. 36, 2009

Table 1. Conversion between curve-number and infiltration capacities.

Ultimate infiltration ca- Curve number (CN) of Maximum infiltration


Hydrologic soil group (HSG) pacity fc (mm/h)a meadow coverb capacity fm (mm/h)c
A (sandy, loamy sand or sandy loam) 25 30(3) 250
B (silt loam or loam) 13 58 200
C (sandy clay loam) 5 71 125
D (clay loam, silty clay loam, sandy 3 78 75
clay, silt clay or clay)
a
Smith 2004.
b
U.S. Soil Conservation Service (1972).
c
Curve number (CN) of 30 is considered the lowest possible. Users should be warned that the Natural Resource Conservation Services
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

(NRCS) CN method for runoff does not work well for CNs less than 30; another infiltration model may be more suitable. When CN =
100, i.e., for impervious areas; fc = fm = 0.

Therefore, to reduce va by a certain proportion, xa should be 2.2. Conversion of runoff curve-number and Nash
increased by the same proportion. instantaneous unit hydrograph parameters
It was found that for large rainfall events in the Toronto For large rural catchments, the SCS (now NRCS, i.e.,
region, the rainfall event volume and duration are indeed Natural Resource Conservation Services) curve-number
statistically independent (Palynchuk and Guo 2008). In the (CN) method is widely used for runoff estimation, whereas
same study, it was also pointed out that although conven- APSWM rainfall-loss estimation is based on the Horton in-
tional design storms are often quoted as fixed duration de- filtration model (Guo and Adams 1998a). The total potential
sign storms, the complete interpretation of a design storm losses throughout a rainfall event as estimated in APSWM
should be the storm with a specific return period for event are comprised of three parts: (1) losses due to the intercep-
depth, under the condition that its duration is less than or tion and depression storage on impervious areas (Sdi); (2) in-
equal to a given duration. The derivations in the previous itial losses on pervious areas (Sil), which is the sum of the
paragraph incorporated the complete interpretation of design depression storage (Sdp) and the long-term average initial
For personal use only.

storms. Since design storms are applied as if they are actual soil-wetting infiltration depth (Siw); and (3) infiltration loss
individual storms, equivalence of storms defined in the left on pervious areas at a constant rate of fc, which equals the
side of eq. [5] to design storms is adequate for the purpose ultimate infiltration capacity as defined in the Horton model.
of ensuring that the same degree of areal reduction is The three parts of losses in an APSWM model can be ob-
achieved when the analytical probabilistic approach is used. tained from the NRCS-CN loss method parameters as fol-
Based on all these empirical evidences and theoretical lows.
reasoning, a simple method that makes use of the existing First, in using the NRCS-CN method, pervious and imper-
empirically derived ARFs (e.g., TP-29) is proposed for use vious areas are lumped together and treated as if the catch-
with the analytical probabilistic approach. First, a reduction ment surface is homogeneous. In APSWM, this is equivalent
factor value h is determined based on TP-29 curves for a to specifying the imperviousness ratio of the catchment as
specific watershed area using the average rainfall event du- zero and treating the entire catchment as pervious. Second,
ration of the geographic location as the storm duration. Sec- both Siw and Sdp are considered as constants in APSWM
ond, the parameter xa is obtained as x=h, where x is the and subtracted from incoming rainfall volumes before runoff
distribution parameter value for point rainfall event volume occurs, this is similar to initial abstraction (Ia) in the NRCS-
of that location. CN method. Thus, Sil should be equal to Ia. Third, the ulti-
The effect of the proposed areal reduction method on the mate infiltration capacity fc and the maximum infiltration
entire spectrum of rainfall event volume distributions can be capacity fm for dry soils are both functions of soil type. A
found in Fig. 1, where examples of point and areal average loose relationship between fc and hydrologic soil group
rainfall event volume distributions are plotted. It can be seen (HSG) was established in Maidment (1993, page 5.25). For
from Fig. 1 that with higher degrees of areal reduction (i.e., stormwater modeling purposes, Smith (2004) suggests more
lower ARF values), the frequencies of rainfall events with exact relationships between HSG and fc as well as HSG and
large volumes decrease, while the frequencies of rainfall fm; these relationships are listed in Table 1. The runoff CN
events with smaller volumes increase. Areal averaging of is determined by not only soil type but also land use. Table 1
storms that partially cover a watershed or move across a lists the CN for combinations of HSG and meadow land
watershed makes large point rainfall events smaller. The cover. Using Table 1 as a guideline, the corresponding fc
larger the watershed area, the more the rainfall events that and fm values for all CN values can be obtained through in-
partially cover the watershed. That is why, as the watershed terpolation. For the purpose of conversion from runoff CN
area increases, the frequency of larger areal-average rainfall parameter values to Horton infiltration model parameter val-
events decreases and that of smaller areal-average events in- ues, there is no need to use fm to calculate the value of Siw,
creases. These changes in frequency of occurrence are well since Siw is included in the conversion-determined Sil value
represented in Fig. 1, indicating that the proposed method already.
captures the essence of rainfall areal averaging from a prob- In conventional hydrologic models, the Nash instantane-
ability of occurrence point of view. Simply using xa in place ous unit hydrograph (IUH) method may be selected to trans-
of x, the existing APSWM can be applied to large water- form effective rainfall over a subcatchment to runoff
sheds where rainfall areal reduction is necessary. hydrograph. The Nash IUH method conceptualizes a sub-

Published by NRC Research Press


Guo and Dai 937

catchment as a series of n identical linear reservoirs with the The area-weighting technique cannot be used for the
outflow from one reservoir inflowing into the other. Two in- lumping of tc. If the components (i.e., subcatchments,
put parameters, i.e., the number of reservoirs n and time to reaches, or detention ponds) are in series, it is obvious that
peak tp, are required when Nash IUH is used. In APSWM, the hydrologically equivalent tc of the lumped catchment
however, the effective rainfall and runoff hydrograph trans- should be the sum of the equivalent tc of the individual com-
formation is based on a triangular hydrograph approximation ponents. If the components are parallel to each other, it is
(Guo and Adams 1998b), the only input parameter is the proposed here that the equivalent tc of the lumped catchment
time of concentration (tc) of the subcatchment. It was ini- be calculated as a flow-weighted average of tc of the indi-
tially postulated in Guo and Adams (1999b) and later on va- vidual component. The weight for an individual component
lidated in Guo and Zhuge (2008) that the equivalent time of is
concentration of a single linear reservoir is approximately
2K, where K is the storage coefficient of the reservoir. EðVRi Þ
½11 wi ¼
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

When n reservoirs are connected in series, the total equiva- t þ tci


lent time of concentration is 2Kn. By differentiating the where EðVRi Þ is the expected value of runoff per rainfall
Nash IUH with respect to time and equating it to zero, the event from component i, which is calculated by APSWM as
time to peak tp of a Nash IUH in terms of n and K can be a by-product (Guo and Adams 1998a); tci is the equivalent
expressed as time of concentration of component i; t is the average dura-
½8 tp ¼ ðn  1ÞK tion of rainfall events at that location.
The weight wi in eq. [11] is approximately the average
Thus, K can be determined as tp/(n–1). The equivalent flow rate from component i. Compared with other compo-
time of concentration that should be used in APSWM can nents, a component may have a relatively long or short tc
therefore be obtained as but hardly produce any flow; tc of the lumped catchment
should not be too heavily dependent on the tc of those com-
n
½9 tc ¼ 2Kn ¼ 2 tp ponents that generate little flows. Use of wi instead of Ai in
n1 eq. [10] assists in compensating for the uneven contribution
of flows from parallel components. It is recognized that in-
For personal use only.

Establishment of the above parameter relationships be-


tween rainfall loss and surface runoff routing models en- corporation of the above-described catchment aggregation
sures the hydrologically equivalent conversion of parameter strategy in APSWM will result in some loss of accuracy.
values when different methods are used. Using these rela- For small drainage areas typical of urban stormwater man-
tionships, APSWM can accept input parameters as if agement studies, this loss of accuracy is usually insignifi-
NRCS-CN and Nash IUH methods are used. The robustness cant. For larger watersheds, this may no longer be the case.
of this expanded capability will be examined later in the The following case study was conducted to assist in quanti-
case study. fying loss of accuracy of APSWM resulted from large catch-
ment areas.
2.3. Incorporation of a catchment aggregation strategy
In conventional hydrologic modeling, the modeled water- 3. Case study and analysis of results
shed or catchment is usually divided into a number of sub- 3.1. Description of study area, available data, and model
catchments according to the drainage system and the runs
physical characteristics, such as soil type and land use. Sub- The Ganaraska River watershed with a drainage area of
catchments might be connected by pipes and channel 277 km2 in Ontario (Fig. 2) was used as an example. The
reaches or joined together at junction points. Starting the main stream of the Ganaraska River originates in the Oak
calculation from the most upstream end and gradually mov- Ridges Moraine area and flows southeasterly for 42 km to
ing downstream, the hydrograph and its peak at the point of Lake Ontario. The watershed consists of a mixture of wood-
interest are obtained. land and agricultural lands. The soils map provided by the
The APSWM can also allow catchment discretization ex- Ganaraska Region Conservation Authority (GRCA) indicates
actly the same way as in conventional hydrologic modeling. that most of the watershed soils are gravel, sand, or sandy
However, in actual calculations to determine the flood fre- loam (classified as HSG A, AB, and B), which is particu-
quency distribution at a point of interest using the derived larly suitable for our purpose of investigating highly perme-
analytical equations, the area upstream of that location of in- able rural catchments. One stream gauge station (Dale
terest is treated as one lumped catchment. The characteris- Gauge) located on the main stem of the Ganaraska River
tics of this lumped catchment except tc are obtained by provides about 30 years of flow data, and one nearby Envi-
using an area-weighting technique, i.e., ronment Canada weather station (Peterborough station) pro-
Xm vides more than 32 years of meteorological data.
Ai Pi
½10 P ¼ X1 m The discretization and characterization of subcatchments
A were performed by GRCA using the ArcHydro GIS soft-
1 i
ware. The watershed was discretized into 47 subcatchments
where P represents the value of a parameter for the lumped based on stream orders, with some modifications to combine
catchment, Pi is the value of the same parameter of subcatch- small subcatchments. The areas of subcatchments range
ment i, Ai is the area of subcatchment i, and m is the total from 121 to 2609 ha. Weighted CN value for each subcatch-
number of subcatchments comprising the lumped catchment. ment was determined by ArcHydro based on land use, drain-

Published by NRC Research Press


938 Can. J. Civ. Eng. Vol. 36, 2009

Fig. 2. Study area map showing subcatchment discretization and junction point locations.
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14
For personal use only.

age area, and HSG. An OTTHYMO model (Greenland Inter- data from the Peterborough station were used in the third
national Consulting Inc. 2002) for the watershed was devel- set of runs. Two types of design storms, 24 h Chicago de-
oped and calibrated by Schaeffer & Associates Ltd. (2005). sign storms (Water Environment Federation and American
In the model, both the NRCS-CN and Nash IUH methods Society of Civil Engineers 1992) and 12 h Canadian Atmos-
were used. The Dale gauge streamflow data were used to pheric Environment Service (AES) design storms (Hogg
calibrate the OTTHYMO model. 1980) developed from rainfall intensity–duration–frequency
An APSWM model for the Ganaraska River watershed (IDF) curves, were used in all the OTTHYMO model runs.
was developed based on exactly the same subcatchment and The main characteristics for these design storms are listed in
channel reach discretizations used in the OTTHYMO model. Table 2. For Toronto, the average rainfall event volume, du-
The parameter values for each subcatchment of the APSWM ration, inter-event time, and annual number of events as in-
model were determined based on the corresponding OT- put to APSWM are 8.43 mm, 9.26 h, 93.9 h, and 64.7,
THYMO parameter values and the relationships for conver- respectively; those for Peterborough are 8.14 mm, 9.24 h,
sion as described earlier. No calibration of the APSWM 86.7 h, and 61.3, respectively (Adams and Papa 2000).
parameter values using observed flow data was performed
since the purpose is to compare uncalibrated APSWM re- 3.2. Comparison for individual subcatchments
sults directly with OTTHYMO results. As the OTTHYMO Four representative subcatchments (subcatchment Nos. 3,
model parameters had been calibrated already (Schaeffer & 47, 40, and 22) were selected for comparison. The main
Associates Ltd. 2005), OTTHYMO results were taken as ap- physical characteristics of the four subcatchments are sum-
proximately the same from frequency analysis using ob- marized in Table 3. It was found that the use of two types
served flow series. The capability of the expanded APSWM of design storms does not result in significant differences in
was illustrated by comparing the uncalibrated APSWM re- flood peaks. This may be a result of the following two facts:
sults with calibrated OTTHYMO results. (1) for each return period, although the Chicago-storm peak
To generate the necessary results for comparison, the first intensity is more than two times that of the AES storm, the
set of model runs was designed to investigate the perform- average intensity of the AES storm is almost two times that
ance of APSWM for individual subcatchments and to test of the Chicago storm (Table 2); and (2) all the four sub-
the robustness of the conversion methods for rainfall loss catchments have long time of concentration (from 9.0 to
and subcatchment routing calculations. The second set of 15.0 h, converted from the tp and n values using eq. [9])
runs was designed to examine the catchment aggregation and long time of concentration reduces the impact of peak
strategy of APSWM and the proposed probabilistic rainfall rainfall intensity on resulting flood peaks. As examples, the
areal reduction method. These two sets of runs were con- comparisons for No. 40 (higher infiltration rate) and No. 22
ducted using rainfall data from the Toronto Pearson Interna- (lower infiltration rate) are plotted in Figs. 3 and 4, respec-
tional Airport station. To examine the capability of the tively. Figures for the other two subcatchments repeat simi-
analytical probabilistic approach in representing accurately lar patterns. To facilitate the inter-comparison between
the differences in rainfall characteristics between geographi- catchments of different drainage areas, both the horizontal
cal locations and their effects on flood frequency, rainfall and the left vertical axes were plotted using the same scale

Published by NRC Research Press


Guo and Dai 939

Max. intensity (mm/h)


in all figures. The vertical axes on the right represent flood
peaks in m3/s (cms).
Subcatchments No. 47 and No. 40 both have highly per-
meable soils, while subcatchment No. 3 is almost 100% im-
pervious (CN = 100) and subcatchment No. 22 has low
infiltration rates (CN = 90). An almost perfect match be-

12.1
14.0
16.5
18.3
20.0
tween APSWM and OTTHYMO results was obtained for

9.2
subcatchment No. 3 (figure not presented). For subcatch-
ment 22, APSWM estimation for lower return periods is
slightly higher than those of OTTHYMO (Fig. 4). For sub-
Peterborough

catchment No. 40, flood peaks from APSWM are slightly


Total rainfall
depth (mm)

lower than those from OTTHYMO for lower return periods,


Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

whereas more comparable results were obtained for longer

Note: Chicago design storms were developed with 0.48 as their ratio of time to peak to total storm duration; AES: Canadian Atmospheric Environment Service.
43.9
57.7
66.8
78.4
86.9
95.4
return periods. Results for subcatchment No. 47 are similar
to those from subcatchment No. 40. The agreement between
APSWM and OTTHYMO model results for a variety of in-
dividual subcatchments suggests that the relationships for
conversion of rainfall loss and runoff routing parameter val-
Max. intensity

ues are generally acceptable. Hence, the NRCS-CN and


Nash IUH methods can be used together with APSWM.
(mm/h)

12.4
14.8
17.8
20.0
22.2
12 h AES design storm

3.3. Comparison for aggregated large catchments


8.9

Modeling results from four downstream locations with


various drainage areas are compared. The four locations are
illustrated in Fig. 2 as J1, J2, J3, and J4. The drainage area
Total rainfall
depth (mm)

of 2128 ha at J1 is comprised of three subcatchments and


For personal use only.

Toronto

two reaches. The three subcatchments are similar in their


105.7

soil and land use characteristics and their CNs are around
42.3
59.2
70.5
84.7
95.2

80. The comparison for J1 is plotted in Fig. 5, which shows


that the results from APSWM are at the worst 20% to 25%
different from OTTHYMO results. Compared with Figs. 3
Max. intensity

and 4, Fig. 5 indicates that when a limited degree of aggre-


Table 2. Main characteristics of design storms for Toronto and Peterborough, Ontario.

gation is performed with a few relatively homogeneous sub-


(mm/h)

catchments, loss of accuracy of APSWM is unnoticeable.


22.9
31.4
37.1
44.4
49.5
54.5

Junction J2 is the point where the tributary of North Ganar-


aska confluences with the Ganaraska River. The drainage
area at J2 is 7048 ha and consists of nine subcatchments and
four reaches. The CNs of the nine subcatchments vary from
Peterborough
Total rainfall
depth (mm)

45 to 91. The level of aggregation performed in APSWM for


J2 represents cases involving intermediate to high degrees of
102.8

lumping of widely varying subcatchment characteristics.


48.1
62.7
72.4
84.7
93.7

Since the drainage area at J2 exceeds the threshold of 25 km2


(MNR 1986), rainfall areal reduction must be applied and re-
sults at J2 from modeling runs with areal reduction (Table 4
Max. intensity

lists the reduction factors applied) are plotted in Fig. 6. The


maximum difference between APSWM and OTTHYMO esti-
24 h Chicago design storm

(mm/h)

mated flood peaks is about 1 mm/h of water over the contri-


25.0
33.5
39.3
46.5
51.7
57.0

buting catchment area. On a millimeter per hour (mm/h)


basis, the difference between APSWM and OTTHYMO re-
sults at J2 does not seem to increase significantly as com-
pared with J1 and individual subcatchments. On a cubic
Total rainfall
depth (mm)

meter per second (m3/s) basis, however, the difference in-


Toronto

creases as the contributing catchment area increases. From


109.3
121.5

the figures, it can be seen that as the catchment areas in-


47.4
67.2
80.4
97.0

crease, the difference between 24 h and 12 h design storm re-


sults also increase on a cubic meter per second (m3/s) basis.
Return period

One mm/h difference translates to 0.28 m3/s per 100 ha. If


this level of uncertainty is acceptable, the loss of accuracy
due to the use of the proposed aggregation strategy in
(years)

APSWM may be considered insignificant even when inter-


100
10
25
50

mediate to high degrees of aggregation are involved.


2
5

Published by NRC Research Press


940 Can. J. Civ. Eng. Vol. 36, 2009

Table 3. Physical characteristics of the four representative subcatchments.

Location Subcatchment No. 3 Subcatchment No. 47 Subcatchment No. 40 Subcatchment No. 22


Area (ha) 198.89 605.5 993.7 771.7
Slope (%) 1.23 1.8 1.74 1.65
Flow length (m) 3511 6392 7331 8603
CN 100 48 64 90
Time to peak (h)a 1.28 1.74 2.14 2.12
(for Nash IUH)
Number of reservoirsa 1.4 1.4 1.4 1.4
(for Nash IUH)
a
Obtained from calibration.
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

Fig. 3. Analytical probabilistic stormwater model (APSWM) and Fig. 5. Analytical probabilistic stormwater model (APSWM) and
OTTHYMO modeling results for subcatchment No. 40 (catchment OTTHYMO modeling results for Junction 1 (catchment area =
area = 994 ha). 2128 ha).
7 7 40
OTTHYMO 24 h Chicago
6 OTTHYMO 24 h Chicago 6 35
15 OTTHYMO 12 h AES
Flood peak (mm/h)

5 OTTHYMO 12 h AES 30

Flood peak (m3/s)

Flood peak (m3/s)


5 APSWM

Flood peak (mm/h)


4 APSWM 25
10 4
20
3 3
15
2 5 2
10
1
1 5
0 0
For personal use only.

0 0
1 10 100
1 10 100
Return period (year)
Return period (year)

Fig. 4. Analytical probabilistic stormwater model (APSWM) and Table 4. Rainfall areal reduction factors for different points of in-
OTTHYMO modeling results for subcatchment No. 22 (catchment terest and storm durations.
area = 772 ha).
7 15 Point of Catchment Areal reduction
6
interest area (hectares) Storm duration factor
J2 7049 9.26 0.88
Flood peak (mm/h)

Flood peak (m3/s)

5
10 J2 7049 12 0.89
4 J2 7049 24 0.93
3 J3 23 872 9.26 0.85
5 J3 23 872 12 0.87
2 OTTHYMO 24 h Chicago
J3 23 872 24 0.91
OTTHYMO 12 h AES
1
APSWM Note: Areal reduction factors for J4 are the same as those for J3. From
0 0 U.S. Weather Bureau 1957, TP 29.
1 10 100
Return period (year)

3.4. Validation of the probabilistic rainfall areal OTTHYMO is close to that determined by APSWM. The
reduction method comparison figure for J4 (not shown) illustrated a very sim-
Reduction factors for different storm durations and points ilar pattern. These figures suggest that the proposed rainfall
of interest determined using the adjustment curves in TP-29 areal reduction method for use with the analytical probabil-
are listed in Table 4. Results for J3 with and without areal istic approach may provide similar effect as the conventional
reduction are plotted in Fig. 7. Since OTTHYMO results us- areal rainfall reduction method used with IDF curves and
ing the two types of design storms are quite close to each design storms. It is therefore worthwhile to conduct further
other on a millimeter per hour (mm/h) basis, to avoid over- research about the application of this new rainfall areal re-
crowding, results from only the 12 h AES storms are plotted duction approach.
in Fig. 7. The drainage area at J3 is 23 872 ha which is From Fig. 7, it can also be seen that OTTHYMO and
much larger than that of J2. It can be seen from Fig. 7 that, APSWM results are quite close to each other for return peri-
for the same return period, the decrease of flood peaks re- ods greater than 10 years. For return periods below 10 years,
sulting from rainfall areal reduction determined by flood peaks estimated by APSWM are lower than those esti-

Published by NRC Research Press


Guo and Dai 941

Fig. 6. Analytical probabilistic stormwater model (APSWM) and Fig. 8. Analytical probabilistic stormwater model (APSWM) and
OTTHYMO modeling results for Junction 2 (catchment area = OTTHYMO modeling results for Junction 4 (catchment area =
7049 ha) with rainfall areal reduction. 27 691 ha) with the watershed located in Toronto and Peterborough.
7 7
OTTHYMO 24 h Chicago Toronto ‐ OTTHYMO 500
6 120
Flood peak (mm/h)

Flood peak (m3/s)

Flood peak (mm/h)


Peterborough ‐ OTTHYMO

Flood peak (m3/s)


5 OTTHYMO 12 h AES 5 400
80 Peterborough ‐ APSWM
4 APSWM 4
Toronto ‐ APSWM 300
3
3
2 40 200
2
1 100
1
0 0
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

0 0
1 10 100
1 10 100
Return period (year)
Return period (year)

Fig. 7. Analytical probabilistic stormwater model (APSWM) and the analytical probabilistic approach, only four parameters
OTTHYMO modeling results for Junction 3 (catchment area = (i.e., x, l, j, and q) are employed to represent the rainfall
23 872 ha) with and without rainfall areal reduction. conditions of a geographic location. Although not yet specif-
7 450 ically demonstrated, it is implied that the differences in the
OTTHYMO without areal reduction values of the four parameters are sufficient to reflect the dif-
6 400
OTTHYMO with areal reduction
Flood peak (mm/h)

Flood peak (m3/s)


5 350 ferences in rainfall conditions between geographic locations.
APSWM without areal reduction 300 To evaluate the adequacy of the analytical probabilistic ap-
4 APSWM with areal reduction 250 proach in reflecting locational differences, both APSWM
3 200 and OTTHYMO models of the entire watershed were run
150
2
100
again using rainfall data from the Peterborough station
which is approximately 150 km away from the Toronto
For personal use only.

1 50
0 0 Pearson International Airport.
1 10 100 Junction J4 is the outlet of the entire Ganaraska River
watershed. The comparison for J4 is plotted in Fig. 8. Again
Return period (year)
OTTHYMO results for the two types of design storms are
close to each other on a millimeter per hour (mm/h) basis,
mated by OTTHYMO. This is the same as for single sub- and thus only results from the 24 h Chicago storms are pre-
catchments with highly permeable soils (No. 40, Fig. 3). Up- sented in Fig. 8. The OTTHYMO results show that flood
stream of J2 and J3, the majority of the catchment soils are peaks for the same return period are larger if the catchment
also highly permeable. The comparison figure for J2 showed is located in Toronto; the higher the return period, the larger
a similar pattern as that shown in Fig. 7 for J3. The under- the difference. The same pattern of locational difference in
estimation of lower return period flood peaks of APSWM flood peaks is reflected by APSWM results. This seems to
for highly permeable areas is likely caused by the use of dif- suggest that using only four parameters, the analytical prob-
ferent rainfall loss calculation methods. After satisfying the abilistic approach is capable of representing the unique rain-
initial losses, the runoff CN method estimates rainfall losses fall conditions of a location. To verify this, more locations
as a portion of the incoming rainfall, and there is always should be investigated in the future.
some runoff regardless of how intense the rainfall event is.
While APSWM calculates the potential infiltration loss as
the product of infiltration rate and rainfall event duration, if 4. Summary and conclusions
the incoming rainfall event volume is less than this potential Since 1998, APSWM has been developed, verified, and
loss, all the incoming rainfall is lost and no runoff occurs. continuously improved so that it can be used for urban
For highly permeable soils, APSWM estimated potential stormwater management planning and analysis. This study
loss is high and the resulting runoff volume is lower than was conducted to expand the capabilities of APSWM so
that estimated by the runoff CN method. This is especially that it can also be used in watershed and master drainage
true for lower return period storms. Therefore, APSWM pre- planning studies. The proposed rainfall areal reduction
dicted lower return-period flood peaks are slightly lower method involves adjustment of only the value of the rainfall
than those predicted by OTTHYMO for catchments with event volume probability distribution parameter x. The de-
highly permeable soils. Under longer return-period storms, gree of adjustment can be determined following the same
rainfall losses are insignificant compared with storm vol- procedure as in estimating the areal reduction factors for
umes, thus APSWM and OTTHYMO results are closer. use with design storms. Comparison in this study showed
that the proposed rainfall areal reduction method provides
3.5. Representation of geographical locational differences very similar effects as those produced by the conventional
in rainfall conditions method used with the design storm approach.
In using the design storm approach, a set of design storms The proposed area-weighted lumping for imperviousness
of various durations and return periods is employed to repre- and soil infiltration parameters and the flow-weighted lump-
sent the rainfall conditions of a geographical location. Using ing for time of concentration of parallel hydrologic elements

Published by NRC Research Press


942 Can. J. Civ. Eng. Vol. 36, 2009

provide the essential catchment aggregation strategy re- catchments with event-based probabilistic models, 1. Runoff vo-
quired for the application of APSWM-analytical equations. lume. Water Resources Research, 34(12): 3421–3431. doi:10.
Application of this strategy in the case study demonstrated 1029/98WR02449.
that catchment aggregation does not noticeably increase the Guo, Y., and Adams, B.J. 1998b. Hydrologic analysis of urban
difference between design storm modeling and APSWM re- catchments with event-based probabilistic models, 2. Peak dis-
sults when compared on the basis of millimeters of water charge rate. Water Resources Research, 34(12): 3433–3443.
per unit catchment area per hour (mm/h) of water per unit doi:10.1029/98WR02448.
catchment area. Comparisons for cases involving various de- Guo, Y., and Adams, B.J. 1999a. An analytical probabilistic ap-
grees of aggregation seemed to illustrate that the proposed proach to sizing flood control detention facilities. Water Re-
strategy is appropriate. sources Research, 35(8): 2457–2468. doi:10.1029/
1999WR900125.
It is shown in this paper that the input parameters of run- Guo, Y., and Adams, B.J. 1999b. Analysis of detention ponds for
off curve-number (CN) and Nash instantaneous unit hydro-
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

storm water quality control. Water Resources Research, 35(8):


graph methods may be converted to their equivalent 2447–2456. doi:10.1029/1999WR900124.
counterparts required for input to APSWM. Comparison in Guo, Y., and Zhuge, Z. 2008. Analytical probabilistic flood routing
the case study confirmed that these conversions do not result for urban stormwater management purposes. Canadian Journal
in increased difference between design storm and APSWM of Civil Engineering, 35: 487–499. doi:10.1139/L07-131.
results for catchments with medium to high CNs. As the Hogg, W.D. 1980. Time Distribution of short duration rainfall in
catchment area increases, APSWM is still capable of provid- Canada. In Proceedings of the Canadian Hydrology Symposium,
ing accurate flood peaks as compared with design storm Toronto, Ont., May 26–27, pp 53–63.
modeling results. Only for return periods lower than 10 years Levy, B., and McCuen, R. 1999. Assessment of storm duration for
and catchments with very low CNs (i.e., dominated by hydrologic design. Journal of Hydrologic Engineering, 4(3):
highly permeable soils), the difference between design storm 209–213. doi:10.1061/(ASCE)1084-0699(1999)4:3(209).
and APSWM results becomes consistently larger. Design Maidment, D.R. (Editor). 1993. Handbook of hydrology. McGraw-
storm modeling results themselves may be highly imprecise Hill, Inc. New York.
for cases with very low CNs because the results are ex- Marsalek, J., and Watt, W.E. 1984. Design storms for urban drai-
tremely sensitive to antecedent soil moisture conditions that nage design. Canadian Journal of Civil Engineering, 11: 574–
For personal use only.

are difficult to estimate. Cases with low CNs should there- 584.
fore be dealt with caution. MOE and MNR. 1993a. Watershed management on a watershed
basis: implementing an ecosystem approach. Ministry of Envir-
Comparison for the same catchments but with Peterbor-
onment and Energy and Ministry of Natural Resources, Queen’s
ough and Toronto rainfall data inputs quantified that, in rep-
Printer for Ontario, Toronto, Ont.
resenting local rainfall conditions and their impact on the MOE and MNR. 1993b. Subwatershed Planning. Ministry of Envir-
frequency distribution of surface runoff-generated floods, onment and Energy and Ministry of Natural Resources. Queen’s
the four parameters used in the analytical probabilistic ap- Printer for Ontario, Toronto, Ont.
proach (i.e., x, l, j, and q) provide comparable degrees of MNR. 1986. Ontario Floodplain Management Technical Guide-
effectiveness as design storms. The four rainfall parameters lines, Ministry of Natural Resources. Queen’s Printer for On-
are perhaps sufficient descriptors of local rainfall conditions tario, Toronto, Ont.
for stormwater management purposes. Packman, J.C., and Kidd, C.H.R. 1980. A logical approach to the
design storm concept. Water Resources Research, 16(6): 994–
Acknowledgments 1000. doi:10.1029/WR016i006p00994.
Financial support from the Centre for Earth and Environ- Palynchuk, B., and Guo, Y. 2008. Threshold analysis of rainstorm
mental Technologies of Ontario and the Natural Sciences and depth and duration statistics at Toronto, Canada. Journal of Hy-
drology (Amsterdam), 348: 535–545. doi:10.1016/j.jhydrol.2007.
Engineering Research Council of Canada (NSERC) is grate-
10.023.
fully acknowledged. Assistance provided by the Ganaraska
Quader, A., and Guo, Y. 2006. Peak discharge estimation for urban
Region Conservation Authority is also greatly appreciated.
catchments using analytical probabilistic and design storm ap-
proaches. Journal of Hydrologic Engineering, 11(1): 46–54.
References doi:10.1061/(ASCE)1084-0699(2006)11:1(46).
Adams, B.J., and Howard, C.D.D. 1986. Design storm pathology. Restrepo-Posada, P.J., and Eagleson, P.S. 1982. Identification of in-
Canadian Water Resources Journal, 11(3): 49–55. dependent rainstorms. Journal of Hydrology (Amsterdam), 55:
Adams, B.J., and Papa, F. 2000. Urban stormwater management 303–319. doi:10.1016/0022-1694(82)90136-6.
planning with analytical probabilistic models. John Wiley & Rivera, P., Gironas, J., Montt, J.P., and Fernandez, B. 2005. An
Sons, Inc., New York. analytical model for hydrologic analysis in urban watersheds. In
Allen, R.J., and DeGaetano, A.T. 2005. Areal reduction factors for proceedings of the 10th International Conference on Urban Drai-
two eastern United States regions with high rain-gauge density. nage (on CD), Copenhagen, Denmark, 21–26 August.
Journal of Hydrologic Engineering, 10(4): 327–335. doi:10. Schaeffer & Associates Ltd. 2005. Ganaraska river watershed hy-
1061/(ASCE)1084-0699(2005)10:4(327). drology update. Report to Ganaraska Region Conservation
Greenland International Consulting Inc. 2002. Visual OTTHY- Authority. Available from Ganaraska Region Conservation
MOTM v2.0 Reference Manual. Ontario, Canada. Authority, Port Hope, Ontario.
Guo, Y. 2001. Hydrologic design of urban flood control detention Sivapalan, M., and Bloschl, G. 1998. Transformation of point to
pond. Journal of Hydrologic Engineering, 6(6): 472–479. areal rainfall: Intensity-duration-frequency curves. Journal of
doi:10.1061/(ASCE)1084-0699(2001)6:6(472). Hydrology (Amsterdam), 204: 150–167. doi:10.1016/S0022-
Guo, Y., and Adams, B.J. 1998a. Hydrologic analysis of urban 1694(97)00117-0.

Published by NRC Research Press


Guo and Dai 943

Smith, A.A. 2004. MIDUSS1 Version 2, Reference manual. fVa ðva Þ probability density function of areal average rainfall
Rev2.00, Produced by Alan A. Smith Inc., Dundas, Ontario. event volume;
U.S. Soil Conservation Service. 1972. National engineering hand- K reservoir storage coefficient of a linear reservoir (h);
book. Hydrology, U.S. Department of Agriculture, U.S. Govern- m total number of subcatchments in a lumped catch-
ment;
ment Printing Office, Washington, D. C.
n number of reservoirs in Nash Instantaneous Unit
U.S. Weather Bureau. 1957. Rainfall intensity-frequency regime – Hydrograph;
Part 1. The Ohio Valley, TP-29, U. S. Department of Com- P the value of a parameter for a lumped catchment;
merce, Washington, D.C., 44 pp. Pi the value of a parameter for subcatchment i;
Veneziano, D., and Langousis, A. 2005. The areal reduction factor: Sdi depression storage of impervious areas (mm);
A multifractal analysis. Water Resources Research, 41: W07008. Sdp depression storage of pervious areas (mm);
doi:10.1029/2004WR003765. Sil initial loss of pervious areas (mm);
Water Environment Federation (WEF) and American Society of Siw initial soil wetting infiltration volume (mm);
Civil Engineers (ASCE). 1992. Design and construction of urban
Can. J. Civ. Eng. Downloaded from www.nrcresearchpress.com by University of P.E.I. on 11/14/14

t rainfall event duration (h);


stormwater management systems. ASCE Manual and Reports of t average rainfall event duration (h);
Engineering Practice No. 77, WEF Manual of Practice FD-20. tc time of concentration of a catchment (h);
Yoo, C., Kim, K., Kim, H.S., and Park, M.J. 2007. Estimation of tci time of concentration of subcatchment i (h);
areal reduction factors using a mixed gamma distribution. Jour- tp time to peak of a unit hydrograph (h);
nal of Hydrology (Amsterdam), 335: 271–284. doi:10.1016/j. v rainfall event volume (mm);
jhydrol.2006.11.026. va areally averaged rainfall event volume (mm);
wi weight for subcatchment i used in the lumping of
List of symbols time of concentration;
x distribution parameter of point rainfall event vo-
Ai area of subcatchment i (ha); lume (1/mm);
b rainfall inter-event time (h); xa distribution parameter of areal average rainfall
E(VRi) expected value of runoff volume per rainfall event event volume (1/mm);
of subcatchment i (mm); q average number of rainfall events per year;
fB(b) probability density function of inter-event time; l distribution parameter of rainfall event duration (1/h);
fc ultimate infiltration capacity of soil (mm/h); j distribution parameter of inter-event time (1/h; and
For personal use only.

fm maximum infiltration capacity of dry soils (mm/h); h a specific rainfall areal reduction factor value (frac-
fT(t) probability density function of rainfall event dura- tion).
tion;
fV(v) probability density function of rainfall event volume;

Published by NRC Research Press

Vous aimerez peut-être aussi