You are on page 1of 10

105

ARTICLE
Critical review of the evolution of the design storm event concept
Ed Watt and Jiri Marsalek

Abstract: A critical review of the literature and practice indicates that design storm events, which have been used in specic
elds of Canadian and US engineering practice for more than 100 years, can be ascribed to six basic attributes: (a) design return
period, (b) storm duration, (c) intensitydurationfrequency (idf) relations (representing a summary of historical rainfall data,
with some extrapolation for longer return periods), (d) temporal distribution (design hyetograph), (e) areal reduction factor, and
(f) antecedent moisture conditions. Concerns about climate change (or variability) and the need to adapt to the associated
climatic conditions prompted many agencies, and particularly municipalities, to revisit the design storm event issue, particu-
larly in connection with drainage design. It would appear that this analysis has mostly focused on a single property of design
storms idf relations and projected increases in rainfall intensities. The review concludes that the design practice would be well
served by adopting a comprehensive approach considering all design storm event characteristics and their sensitivity to climate
change and inherent uncertainties in the existing idf relations as well as hydraulic design of sewer networks.

Key words: stormwater infrastructure, design rainfall (idf curves), design return period, antecedent conditions, areal reduction
factor.

Rsum : Une analyse critique de la littrature et des pratiques indique que les pluies de projet, qui ont t utilises dans des
domaines spciques de l'ingnierie au Canada et aux tats-Unis depuis plus de 100 ans, sont entirement dnies par six
attributs de base: (a) priode de retour de conception, (b) la dure de l'pisode pluvieux, (c) relations idf (reprsentant un rsum
des donnes historiques de prcipitation, avec extrapolation pour les priodes de retour plus longues), (d) la distribution
temporelle (hytogrammes de projet), (e) facteur de rduction de supercie et (f) les conditions antcdentes d'humidit. Les
proccupations concernant le changement (ou la variabilit) climatique et la ncessit de s'adaptor aux conditions climatiques
associes ont incit de nombreux organismes, et en particulier les municipalits, a rexaminer la question des pluies de projet,
en particulier dans le cadre de la conception des systmes de drainage. Il semblerait que cette analyse ait principalement port
sur une seule des proprits des pluies de projet les relations idf et les augmentations prvues de l'intensit des pluies. L'tude
conclut que la conception des pluies de projet pourrait bncier de l'adoption d'une approche globale tenant compte de
l'ensemble de leurs caractristiques et de leur sensibilit au changement climatique, des incertitudes inhrentes aux relations
idf existantes ainsi que de la conception hydraulique des rseaux d'gouts. [Traduit par la Rdaction]

Mots-cls : infrastructure des eaux pluviales, pluviomtrique (relations idf), priode de retour de conception, conditions antc-
dentes, facteur de rduction de supercie.

Introduction are attributed certain catchment conditions, or (ii) actual histori-


The use of design storms, or more precisely, design storm cal rainfall events that were well documented with respect to
events (DSEs), has been common practice in the United States and their characteristics and resulting ood damage. Only the rst,
Canada for well over 100 years, if one includes the block rainfall in more common and general type of design storms is addressed
the rational method. Applications that include a design rainfall here, even though the value of historical design storms, particu-
hyetograph date back to the late 1930s for larger watersheds and larly in major drainage design and oodplain delineation, is rec-
the 1940s for smaller urban watersheds. Research and develop- ognized and acknowledged.
ment reached its zenith in the 1970s and 1980s with the wide- This paper has two objectives: (i) to provide a critical review of
spread access to computers. However, following that period, work to date and (ii) to identify emerging issues, challenges, and
research and development fell off for about 20 years and only possible directions for future research. The rst part of the paper
relatively recently has interest picked up, possibly spurred on by includes (i) a brief summary of the historical need for small wa-
concerns over potential climate change. tershed design storm events and (ii) a critical review of design
As research scientists, developers of design storms, engineers storm event attributes and components proposed and employed
familiar with current practice, and interested observers of the over the past 35 years. The second part examines the scope of
current focus in some jurisdictions on solving the problem by considerations to be made in design storm event applications in a
changing only one parameter in design storm events, we think changing climate.
that a useful rst step might be to provide a critical review of
Historical need for design storm events
urban and small basin DSEs. The review is limited to planning and
design applications on small and intermediate watersheds. Fur- Storm sewers and drainage design
thermore, design storm events can be either (i) events that have The earliest application of design storm events in North Amer-
never occurred but are derived by synthesis of rainfall data and ica was in the design of sewers in Rochester, New York following

Received 22 December 2011. Accepted 26 November 2012.


E. Watt. Civil Engineering, Queen's University, Kingston, ON K7L3N6, Canada.
J. Marsalek. National Water Research Institute, Environment Canada, Burlington, ON L7R4A6, Canada.
Corresponding author: Ed Watt (e-mail: edw@xcg.com).

Can. J. Civ. Eng. 40: 105113 (2013) dx.doi.org/10.1139/cjce-2011-0594 Published at www.nrcresearchpress.com/cjce on 5 February 2013.
106 Can. J. Civ. Eng. Vol. 40, 2013

the pioneering work of Kuichling (1889). Using the rational Knowledge base of design storm events
method, peak discharge (in cfs) was computed as the product of
Design rainfall temporal distributions derived from
the runoff coefcient, the average rainfall intensity (in inches/
measured point rainfall distributions
hour) over a duration equal to the time of concentration for a
Following the work on urban design storms in Los Angeles and
specied return period, and the drainage area (in acres). The de-
Chicago, several temporal distributions were developed using ob-
sign rainfall hyetograph was then a block of rainfall, the anteced-
served storm rainfall data for areas in a number of countries.
ent conditions (if considered at all) were represented by the runoff These design hyetographs were intended for applications on both
coefcient and the design return period was given by the rainfall urban and (generally larger) rural watersheds. A number of these
return period. hyetographs, which were well publicized in the international lit-
For storm sewer design, the rational method remains to this day erature, are briey discussed below.
the method of choice in many jurisdictions, sometimes with out-
of-date intensitydurationfrequency (idf) relations and values of United States
design return period assigned 50 years ago. One reason for the Huff (1967) developed temporal distributions for the US Mid-
rational method's widespread use (in addition to its simplicity) is, west; he sorted observed storms into four groups with peak inten-
as noted by Chow (1964), that the underlying assumptions might sity occurring in rst, second, third or fourth quartile, and within
nearly hold for paved areas with gutters and sewers of xed di- individual quartiles, he presented distributions as probability
mensions and hydraulic characteristics. The method is also ap- groups, with the 10th percentile storm approximating an expo-
plied in many rural applications. However, the underlying nential distribution and the 90th percentile almost uniform. The
assumptions are less likely to hold as the drainage area increases median curve (50%) of the rst quartile storms was recommended
because of higher likelihood of surface storage and variation in for most applications (Terstriep and Stall 1974). Yen and Chow
inltration potential, which is particularly signicant in the cur- (1980) used the method of moments to develop a triangular di-
rently promoted low impact development (LID) approach to drain- mensionless hyetograph, based on over 9000 storms at four loca-
age design. tions. In a subsequent paper (Yen and Chow 1983), they presented
In the mid-20th century, the development of so-called a map of the United States showing the ratio of time to peak
hydrograph methods in selected urban centres in the United States intensity to storm duration. The ratio ranged somewhat system-
(e.g., Los Angeles and Chicago) marked a major breakthrough in atically from 0.250.30 in the east to 0.45 in the west.
the inclusion of hydrologic science in drainage design. These Australia
methods, which predated the development of urban runoff meth- Pilgrim et al. (1969), Pilgrim and Cordery (1975), and Pilgrim
ods by several decades, were based on the rainfallrunoff process. (1987) used local data to develop design rainfall patterns for Syd-
The City of Los Angeles method dates back to 1939 (Hicks 1944) ney, Australia. This work was extended to six climatological zones
and the Chicago Hydrograph Method to the late 1940s (Keifer and of Australia (Institution of Engineers, Australia 1977); temporal
Chu 1957; Tholin and Keifer 1960). Both methods are quite com- patterns for durations from 12 min to 24 h were provided.
prehensive in terms of realistic design hyetographs based on local
data, type of land use, consideration of ow paths, depression United Kingdom
storage, inltration based on local data, overland ow, gutter Packman and Kidd (1980) presented, an alternative to continu-
ow, conduit storage and, in the Chicago case, areal variation of ous simulation, a design storm event that included the following
rainfall. hydrological inputs: (i) a design rainfall depth, (ii) rainfall dura-
tion, (iii) design hyetograph, and (iv) catchment antecedent wet-
Design ood estimates for watersheds of small and ness based on 15 d antecedent precipitation index and soil
intermediate size moisture decit (estimated nationally by using a running water
Where records of streamow data are either short or non- balance).
existent, design ood estimates are typically determined by
simulating a design event with a prescribed rainfall input. Ap- Canada
plications include (i) culverts, bridges, and small dams; (ii) mi- Hogg (1980) used a method similar to that of Huff to analyze
nor and major urban drainage systems; (iii) master drainage storms of 1 and 12 h duration for 35 stations across Canada. He
planning; (iv) oodplain delineation; and (v) protection of crit- developed regional time probability distributions for the various
regions of Canada. In a subsequent article, Hogg (1982) developed
ical infrastructure.
AES Distributions for each station by matching the three most
The widespread availability of digital computers in the 1960s
important parameters characterizing the time distribution of
and 1970s led to the development of event-based models for rural
storm rainfall in a municipality: the magnitude of the peak with
and urban watersheds. As with their predecessors, hydrograph
respect to the total accumulation, the timing of the peak, and the
methods, these models incorporated hydrologic science and were
rainfall prior to the peak. He presented 1 and 12 h distributions (in
based on the rainfallrunoff process. For urban applications, the cumulative hyetograph form) for eight regions of Canada. Watt
Storm Water Management Model, SWMM (Metcalfe et al. 1971) has et al. (1986), using the same storm data set as Hogg, developed the
become the model of choice in the United States and Canada. For HYDROTEK two-component, dimensionless 1 h design hyetograph
rural and pre-post applications, the Hydrologic Modeling System (linear rise and exponential decay) for 45 stations across Canada.
HEC-HMS (USACE 2000) is a common model. With the advent of This model preserves both the average peakedness and the time to
tested and supported public-domain event models, there is no peak intensity of the observed rainfall and was intended for ap-
longer any problem in representing the drainage basin and mod- plications on urban watersheds. In comparison to Yen and Chow's
eling the rainfallrunoff process. However, there can be problems work, the ratio of time to peak to storm duration ranged from 0.25
in generating realistic results that are appropriate for planning in the BC interior and Alberta to 0.45 in Northern Ontario, Nova
and design. These problems can arise for a number of reasons, Scotia, and Newfoundland. Westhoff et al. (1993) extended the
including selection of an inappropriate design storm. Accord- HYDROTEK distribution to longer duration storms in Calgary and
ingly, key contributions to our knowledge base of design storm found that a third parameter improved the t for atter, longer
events are reviewed in the following section, building on two duration storms. Most recently, Peyron et al. (2002) used 52 year
previous publications (Watt and Marsalek 1977; Marsalek and rainfall series (Dorval Airport, Quebec) to develop an optimal de-
Watt 1984). sign storm pattern for urban runoff estimation in Southern

Published by NRC Research Press


Watt and Marsalek 107

Quebec as a hyetograph with a total volume of 1.3 1 h volume tween such ARFs and those calculated from a dense network of rain
from the idf curves (for a selected return period), a symmetrical gauges. Sivapalan and Bloschl (1998) developed catchment ARF fac-
peak intensity burst consisting of three 5 min intervals with in- tors using the spatial correlation structure of rainfall rather than
tensities of 0.8 i15, 1.4 i15, and 0.8 i15, respectively, and occurring observations from a rain gauge network. In Ontario, information
between the 20th and 35th minute from the event start, and the from two historical storms, Hurricane Hazel in southern Ontario,
remaining storm volume distributed in four 5 min intervals be- and the Timmins rain in northern Ontario, is used to dene ARFs
fore and ve 5 min intervals after the peak burst. prescribed for oodplain delineation (MNR 2002). Svensson and
Jones (2010) reviewed the methods used to produce ARFs and noted
Design rainfall temporal distributions based on that no method was unambiguously correct, though the traditional
generalized or derived idf relations data-intensive, empirical, xed-area methods offered a number of
The U.S. Soil Conservation Service (USDA 1968) developed four advantages, including producing probabilistically correct ARFs ap-
dimensionless 24 h storm distributions (presented in cumulative
plicable over a comprehensive range of spatial and temporal scales.
form) for four storm types applicable to various regions of the
United States. Type II was intended to represent the bulk of Specication of antecedent conditions or catchment
the country, excluding the Pacic maritime climate region and wetness
the Gulf of Mexico and Atlantic coastal areas. The distributions Specication of antecedent conditions or catchment wetness
are based on generalized depthdurationfrequency relations. has received much less attention than the form of the design
The location of the peak depends on the storm type: at 8 h for hyetograph, in spite of the fact that it has much more inuence on
types I and IA and at 12 h for types II and III. These design storms the magnitude of the peak ow for all watersheds except those
were intended to be applied on watersheds normally studied by that are highly impervious. The form of the specication varies
USDA engineers. with the computational method and the abstractions model. Kidd
Three common methods, based only on the derived idf relation and Packman (1980) developed an index of antecedent catchment
and not on individual storm events, termed blocked IDF distri- wetness formed from a combination of soil moisture decit and a
butions by Viessman and Lewis (1996), are the alternating block
5 day antecedent precipitation index. A value of this index was
method, the Chicago method, and the balanced method. The de-
selected such that the exceedance probability of the peak ows
sign storms developed by these methods generally have a maxi-
generated by the urban runoff model matched synthetic fre-
mum intensity in the centre of the hyetograph, and a total rain
quency distributions. In the SCS curve number (CN) method
depth for any duration matching the depth given by the idf rela-
(USDA 1986), antecedent moisture conditions (AMC) can be speci-
tion. Such design hyetographs are easy to construct, but have
ed in three categories (AMC I-III), depending on soil conditions
been criticized in that they do not resemble observed storm
and the hydrologic soil group. For low moisture dry soils with
events; this is particularly true for the Chicago storm, in which
high inltration rates (Group A), CN is reduced; for average mois-
the assumption that the design storm should contain all rainfall
maxima contradicts ndings for actual storms (ASCE 1992), in- ture soils with moderate inltration rates (common for design),
cluding those presented in the original paper (Keifer and Chu CN is not modied; and, for high moisture resulting from heavy
1957). Hogg (1980) pointed out that the Chicago-type distribution rain in last few (antecedent) days, high moisture and soils with
is totally inappropriate for some Canadian climates and, for most slow or low inltration rates, the CN is increased. Nishat et al.
of Canada, it is not among the most probable distributions. In a (2010) used a deterministic continuous simulation model to sim-
recent simulation experiment with ve design hyetographs, in- ulate antecedent moisture conditions for the Pearson Airport
cluding the Chicago method, Aleri et al. (2008) noted that all (Malton, Ontario) rainfall record and a wide range of different soil
design hyetographs produced ood peak estimates that were con- types. Statistical analysis of simulation results produced empiri-
sistently biased in most of the conditions studied, and the Chicago cal equations for the estimation of average antecedent soil mois-
method tended to overestimate ows. In some cases, such an ture conditions on the basis of known soil characteristics (such as,
overestimation was corrected by a scaling factor resulting in ro- porosity and the saturated hydraulic conductivity) and the equa-
bust and nearly unbiased estimates of design ows. tions were veried by comparisons with eld data.
In most cases, there is no direct relation between the value
Areal reduction factor specied for the abstractions model parameter and soil water
In the early studies, a point, as in point rainfall, is assumed to be content or average conditions are assumed (USDA 1986; Nishat
10 square miles or about 25 square kilometres (for non- et al. 2010). Occasionally, some form of antecedent precipitation is
mountainous areas), but in urban studies, much smaller point taken as a surrogate for soil water content. In the Chicago Hydro-
areas have been considered (Bengtsson and Niemczynowicz graph method, no antecedent rain was assumed prior to the be-
1986). For larger catchments, the leading design manual (ASCE ginning of the 3 h storm because of the duration of the storm.
1992) recommends reducing the point value by an areal reduction
factor (ARF) to account for the decrease in average rainfall depth Taxonomy of design storm events for various
away from the storm centre, however, without any further guid- applications
ance with respect to the catchment size. The rst comprehensive
analysis on this subject was conducted by the U.S. Weather Bureau Overview
(USWB 1963). The results are presented as curves of ARF vs. Area Design storm events are a key element in the determination
for areas up to 400 square miles for durations of 30 min, 1, 3, 6, of design ows and water levels for a number of applications.
and 24 h. Similar ARF curves, adopted by the U.S. Soil Conserva- The rainfall characteristics of a design storm depend on the
tion Service, the U.S. Bureau of Reclamation and the Tennessee task at hand and the design variable, as summarized in Table 1
Valley Authority were presented by Viessman and Lewis (1996). and discussed in the following sections. In the interests of clar-
Areal reduction factors have also been developed for specic re- ity and simplicity, antecedent moisture conditions, areal reduc-
gions, for example, Los Angeles and Chicago in the studies cited tion factor, and the possibility of a different specication of the
above, Southern Arizona (Woolhiser and Schwalen 1960), and design event are omitted from the table, but are included in the
more recently, for New Jersey and North Carolina (Allen and discussion.
DeGaetano 2005) and Edmonton (Jolly et al. 2008). Bengtsson The issue of climate change: its impacts on the variables listed
and Niemczynowicz (1986) calculated ARFs for a number of Swedish in Table 1 and the likelihood of any sudden change in engineering
cities from rain system movement and found a good agreement be- practice, as well as current inadequacies in current practice that

Published by NRC Research Press


108 Can. J. Civ. Eng. Vol. 40, 2013

Table 1. Taxonomy of design storms for various tasks.


Design return Duration
Sizing task Design variable(s) period (years) Rainfall properties (hours)
Sewers Peak discharge 210 idf relationa 0.251.0
Culverts, bridges & small dams Peak discharge 10100 idf relationa 112
Minor system Peak discharge 510 idf relationb hyetograph 0.251.0
Major system Peak discharge and peak volume 50100 idf relationc hyetograph 624
Master drainage planning Pre- and post-peak discharge, 2100 idf relationc hyetograph 312
peak volume
Floodplain delineation Peak water level 100300 idf relationc hyetograph 312
Critical infrastructure protection Peak water level 10000 or a different idf relationd hyetograph 0.2512
design approache
aFor rational method applications, the short-duration portion of the idf relation typically 10 minutes to 1 h, is usually of interest; the t line to the idf relation is

programmed into the design software.


bFor urban runoff model applications, only the mid-duration portion of the idf relation (e.g., 1 h and 3 h) is used.

cFor major system, master drainage planning and oodplain delineation applications, only the long-duration portion of the idf relation (12 h or 24 h) is used; in

earlier drainage practice, this portion was also used to design stormwater storage.
dFor protection of critical infrastructure, only the long-duration portion of the idf relation (12 h or 24 h) is likely to be used.

eCritical infrastructure is sometimes designed for the probable maximum ood (PMF), which is calculated for the probable maximum precipitation.

Table 2. Climate change (CC) and other tasks.


Item Sensitivity to CC Likelihood of change Other issues
Design return perioda (DRP) None None, except for ooding Major system design; need DRP for critical infrastructure
problems
Storm durationb None-to-low Low Need clarication for storage structures
idf relations Signicantc Modest Reliability of idf relations
Design hyetograph Unknown Modest Need design hyetographs based on Canadian rainfall data
Areal reduction factor Unknown Modest Availability of rainfall radar-dened ARF relations
Antecedent conditions Unknown Low Need will arise from water balance concerns
aIn some parts of sewer systems, DRP may be reduced when delaying adaptation measures for economic reasons or scheduled sewer rehabilitation; however, this

may create liabilities for the sewer system operator.


bSome authors (Cobbina 2007) predict changes in event durations (from statistical analysis of the existing rainfall data), but it is not known whether this applies to

design-type events.
cClimate change projections of short duration extreme events and for high spatial resolution involve high uncertainties.

should be addressed are summarized in Table 2. Detailed discus- with insufcient outow capacity). As a result of poor design, or in
sion for each variable follows the table. some cases, too low a DRP, extra inows to the sanitary system
occur, and result in backups during major ood events. Typically,
Design return period below-grade investment has increased dramatically since current
The design return period (DRP) has evolved to reect the conse- values of DRP were established. These increases and the resulting
quences of failure of the hydraulic element being sized (sewer, increases in damage costs have not been reected in increases in
culvert, spillway, etc.). Although the DRP typically is not deter- DRP.
mined by a detailed economic analysis, the value selected has It has been suggested that DRP values be increased as a hedge
stood the test of time, in most cases. Differences between values against climate change, but this should not be done without ad-
prescribed for storm sewer design by adjacent municipalities are ditional considerations. Before any adjustment to current DRP is
largely a consequence of historical accident, and may appear il- made as an adjustment to account for increases in rainfall regimes
logical. However, current values of DRP are unlikely to be changed due to climate change, causes of urban ooding should be as-
unless there are serious ooding problems. This issue should be sessed and classied as due to (1) too low a value of DRP, or (2) poor
revisited in areas where redevelopment led to potentially much design (e.g., inadequate or non-existent major system), or (3) poor
higher ood damages, due to, for example, increased below-grade maintenance (e.g., blocked inlets or partially blocked sewers). If
investments. there is a systematic tendency toward category 1 in certain areas,
Because there is often no historical ow record (e.g., for a sewer the DRP should be increased in those areas.
system that has not been built) on which to conduct a frequency
analysis to determine the T-year ow, practice is to specify the Use of the probable maximum ood in critical
T-year rainfall as input and assume that the return period of the infrastructure design
resulting calculated peak ow is approximately T. This is only Engineering design calls for high levels of public safety in de-
approximately true if all other variables (antecedent moisture, signing such critical infrastructures as dams or nuclear plants. In
temporal distribution, contributing area, etc.) are set at average many cases, the design criteria require that these structures with-
conditions. For events used to size the minor system in highly stand the probable maximum ood (PMF), which is dened as the
impervious watersheds, the assumption holds true. For all other ood that may be expected from the most severe combination of
cases, it is approximately true only if all other variables are set at critical meteorological and hydrologic conditions that are reason-
average conditions. This subtlety was once common knowledge, ably possible in the drainage basin under study.(US FERC 2002).
but over time has been somewhat forgotten. The PMF is typically estimated by modeling runoff produced by
Typically, not enough attention has been paid to major system the probable maximum precipitation (PMP) at a particular site
design. In some cases, drainage systems built before recognition and a time of year. Thus, the theoretical probability of exceeding
of the major system (UDFCD 1969) have not been designed for PMF is zero; however, since PMF estimates are based on historical
major drainage and ood damages may occur (e.g., in bowls data, there is a need to revisit PMF estimates as more data become

Published by NRC Research Press


Watt and Marsalek 109

available, particularly in a changing climate. The PMF estimates a difference is signicant for the design of major drainage and
are generally produced by specialized consultants and fall beyond also for master drainage planning and the design of storage
the scope of this review. Experience with such estimates in the US facilities. Perhaps adoption of an approach similar to that
and Alberta suggests the following: (a) PMP research is advancing used in the frequency analysis of oods, assuming observed
rapidly and this new knowledge, including probabilistic alter- data representing two populations and tting a combined
natives needs to be incorporated into operational procedures distribution, would resolve this issue. This concern is further
(England et al. 2011), (b) the existing procedures involve great supported by Hailegeorgis and Burn (2009) who noted for a
uncertainties, with the extent of conservatism largely unknown London (Ontario) rain gauge that different best t distribu-
(Alberta Transportation 2004), and (c) potential impacts of a tions applied to 1560 min storms and 624 h storms.
changing climate need to be addressed (Environment Canada (3) Finally, as noted by Metcalfe et al. (1997) when discussing the
2004). importance of the catch factor (CF) in the Environment Can-
ada tipping bucket rain gauge (TBRG) network, Presently, no
Storm duration CF can be calculated for TBRGs on Canadian autostations,
For transmission elements such as sewers, the storm duration is therefore errors are no longer being corrected as before. This
usually selected to match the response time of the watershed to suggests problems in data from the autostation network
the location of the element being sized. At the low end, the sum while creating an inhomogeneous time series. Note also that
of inlet time and sewer travel time yields storm durations of the CF is not the only source of uncertainties in the measured
2030 min for storm sewer applications. At the high end, dura- rainfall data; for example, Hoppe (2008) listed a number of
tions are about 24 h, except for spillways on large dams where a others, including wind caused losses, wetting losses, splash
48 h design storm is applied. Even longer durations are used for losses, and evaporation from heated gauges. The authors be-
rain-on-snow events. For storage elements, the storm duration is lieve that good rain gauge locations (i.e., with good exposure,
selected so as to yield maximum storage, and is often taken as without shading by tall buildings or trees) are almost impos-
24 h, though rain-on-snow combined with snowmelt may call for sible to nd in urban areas. Poor locations, combined with
longer durations. gauge relocations, may lead to statistically signicant jumps
There is some uncertainty in the selection of storm duration for in data series.
storage structures as to whether the critical duration is (1) the one
that maximizes the storage volume for a given rainfall input, or In summary, the discussion in this section indicates high uncer-
(2) the one that yields the T-year storage from the T-year rainfall, tainties in the existing idf relations. Yet such uncertainties are
or the T-year melt plus rain event. Category 2 has been preferred rarely addressed, or even acknowledged. Some appreciation of idf
historically, in that all the risk is contained in the DRP and that it uncertainties can be gained from recent work by Hailegeorgis and
is more consistent. In some jurisdictions, confusion may arise Burns (2009) who reported for a rain gauge station in London,
because of unclear design guidelines. In other cases, where cate- Ontario uncertainties in idf intensities ascribed to data tting as
gory 1 is selected, overdesign is recognized, but accepted as a follows: (a) 95% lower condence limits for 1 h storm and return
hedge against potential climate change effects. This is not consis- periods from 2 to 100 years ranged from 0.81 (T = 100 years) to 0.86
tent with the common assumption of a single DRP, though sug- (T = 2 years), and (b) 95% upper condence limits ranged from 1.11
gestions have been made in the past (Yen 1975) to increase DRP in (T = 2 years) to 1.23 (T = 100 years). In practical terms, these limits
parts of the catchment with higher potential ood damages aris- may represent up to 2 nominal pipe sizes in actual design. Thus,
ing from property damages or disruptions of business. However, the current idf relations contain substantial uncertainties and
this opens the issue of liabilities, where the property owners in even larger uncertainties can be expected in climate change pro-
areas with lower DRP may consider themselves under-serviced jections, recognizing that the current estimates of climate in-
compared to those with higher DRP. duced changes in rainfall intensity, duration, frequency and
spatial extent in urban areas are highly uncertain (Ashley et al.
Intensity durationfrequency (idf) relations 2008). Furthermore, approaches to addressing this issue are not
In Canada, idf relations have been developed by Environment standardized and widely varying results were reported in the lit-
Canada. Storm rainfall depths for durations from 5 min to 24 h are erature addressing urban drainage in a changing climate in Can-
assumed to follow the Extreme Value Type 1 or Gumbel distribu- ada. Large increases in idf intensities were suggested, e.g., by He
tion and sample values are tted by the method of moments. The et al. (2006) (+50%, for i5 min, 5 years), but much smaller increases
resulting information (annual maximum series, moments, and were suggested by Mailhot and Duchesne (2010), the design return
depth and intensity quantiles) are published in the form of a period would increase from 100 to 133 years, for an infrastructure
report for each station (http://climate.weatherofce.gc.ca/prods_ with expected lifetime 100 years; and, Waters et al. (2003) inter-
servs/index_e.html). There are three issues concerning the infor- preted projections by global circulation models as suggesting in-
mation contained in these reports: creases in rainfall depth quantiles for short durations of the order
of 15%.
(1) The sample sizes used are often much too small for reliable
frequency analysis, a fact not realized by some users. In many Temporal distribution (design hyetograph)
cases, a more accurate estimate would result from a regional The sensitivity of design hyetographs to climate change is gen-
idf relation. erally unknown. However, because in present practice, there is so
(2) Implicit in the method of analysis is the assumption that each little consistency regarding design hyetographs, climate change is
annual maximum rainfall event is generated by the same not a primary concern. For rational method applications, a rain-
underlying physical process. While this assumption may hold fall pattern is not required; however, in the remaining application
for short durations where the generating process is convec- types listed in Table 1, a rainfall pattern must be assigned or
tive rainfall, it is less likely to hold for longer durations such selected. Differences in peak discharge or maximum storage re-
as 12 and 24 h, which may be generated by late season hurri- sulting from different design hyetographs may be much larger
cane remnants. The frequency curve for Kingston storm rain- than differences due to climate change.
fall (61 data points) shown in Fig. 1 provides some anecdotal Although 1 and 12 h temporal distributions have been devel-
support for this assertion. The highest depth observed of oped for all regions of Canada, there is a tendency for analysts not
about 126 mm (by 2004), is about one third greater than that to use these distributions, but instead to use some form of the
obtained from the tted frequency curve (about 94 mm). Such Chicago distribution and the SCS distribution, possibly because of

Published by NRC Research Press


110 Can. J. Civ. Eng. Vol. 40, 2013

Fig. 1. Frequency curve for Kingston 12 h storm rainfall.

Fig. 2. Twelve hour design rainfall distributions.

their simplicity, but in many cases as the result of historical acci- Areal reduction factor
dent. The fact that under the present climate, design storms have Areal reduction factors are required on larger catchments, but
not been developed or developed design storms are neither se- are rarely prescribed or used in urban drainage design (i.e., the
lected nor assigned is a serious situation that should be resolved. ARF is assumed to be 1.0). Canadian exceptions to this generality
One problem arising is what might be termed worst case syn- include oodplain delineation in Ontario and large sewer design
drome wherein a designer will select more than one design in Edmonton and York Region. This assumption of an ARF of 1.0 is
storm, for example, the AES 12 h 30% and the so-called SCS Type II not a problem provided that either of the following two condi-
12 h (see, for example Haan et al. 1994) as shown in Fig. 2 and select tions applies: (1) the USWB curves apply or (2) the basin is not
the one that gives the largest discharge. (It is interesting to note large. However, there is some indication that the USWB curves
that the SCS did not publish a 12 h distribution.) The problem might not apply. The derived ARFs for Hurricane Hazel and
with this approach is that the AES 12 h 30% storm was intended to Timmins, both large storms with rainfall total depths exceeding
be representative of storms that might generate the 100 year ow, 100-year values, are less than the USWB values. In addition, the
whereas the SCS 12 h storm represents an extremely intense denition of small with respect to drainage basins should be re-
storm. With the same depth of rainfall and typical basin topogra- considered. For an area of 25 km2, the derived ARF of about 0.7 for
phy, the SCS storm will always generate a higher peak ow, which Edmonton is less than the USWB value (see Fig. 3) and is fairly
would yield a higher DRP. Furthermore, note that the delayed close to the values reported from Sweden for short duration
occurrence of the peak intensity burst in the SCS storm, after storms (20 min) (Bengtsson and Niemczynowicz 1986).
typical hydrological abstractions have been satised, may contrib- The consequence of using an ARF of 1.0, rather than a value
ute to higher simulated runoff peaks. based on Canadian rainfall data, is overdesign, which is a higher

Published by NRC Research Press


Watt and Marsalek 111

Fig. 3. Areal reduction factors.

DRP. This may appear as a useful safety factor, but again it repre- in most cases impossible (Greer 2008). Other potential issues in-
sents an inconsistency in sewer design and this issue should be clude LID performance under extreme rainfall events (producing
revisited and available Canadian storm data used to develop ARFs extreme ows) and maintaining LID performance on private prop-
that would be applicable for major system design. erties (Guillon et al. 2008).

Antecedent conditions General aspects of drainage design


In general, current practice regarding antecedent conditions is The current urban drainage design methodology encompasses
to either include a specication embedded in the abstractions an unprecedented multitude and variety of concepts and options
model (e.g., Antecedent Moisture Condition II in the SCS Curve requiring revisiting the issue of design rainfall data. Furthermore,
number method) or ignore them. Such practice is not consistent the authors believe that the priorities of such considerations vary,
with the current state-of-knowledge of ood generating mecha- depending on the design issue and concept addressed, and a ho-
nisms and available abstraction algorithms (Nishat et al. 2010). listic approach addressing all these issues rather than just one
The importance of antecedent conditions varies from not very (e.g., climate change) is needed.
important for minor system design on largely impervious areas Of paramount importance is major drainage. Dysfunctional ma-
(respectively, the short return period events may be dominated by jor drainage results in repeated ood damages and criticism of
contributions from impervious areas) to extremely important for both designers and political decisions makers. With well function-
major system design and the protection of critical infrastructure. ing major drainage, the importance of minor drainage design for
One more storm characteristic generally neglected in design, climate change is of secondary importance, because overows
but affecting simulated runoff peaks, is storm movement. Storms from minor drainage are conveyed by major drainage. For major
moving with runoff/ow wave in the sewer system produce drainage, extreme rainfall of long duration and low frequency of
higher peaks than those moving in the opposite direction occurrence is important. Under such circumstances, runoff gen-
(Bengtsson and Niemczynowicz 1986). However, in US design prac- eration on pervious areas becomes important and that is of par-
tice, it was suggested that including storm dynamics in urban ticular interest for the currently promoted low impact
drainage design appeared to be an unnecessary renement in
development (LID), comprising many measures reducing runoff
view of other uncertainties (ASCE 1992).
generation by enhancing hydrologic abstractions in the area
Specication of design event and selection of drainage drained and delaying runoff. So far, the performance of LID in
design rainfall data such conditions has not been documented and it remains to be
New concerns may require changes in criteria for the design seen how such systems will perform when exposed to long-
event from a storm rainfall event. There might be two design duration extreme rainfall, particularly when recognizing that
events for drainage design, a storm rainfall event for pipe sizing some inltrating rainwater may reappear as subsurface runoff.
and a snowmelt plus rain event for storage sizing. Compared to Other considerations mentioned earlier with direct applicability
traditional snowmelt hydrology, the situation in urban areas is to major drainage are antecedent moisture conditions and the
radically different. In particular, heat uxes in urban areas are areal reduction factor. Recognizing high uncertainties in these
largely modied by heat losses or refection of solar radiation from considerations, they should be probably best accounted for by a
buildings, and nally, applications of deicers result in chemical safety factor that would also consider climate change.
melts (Oberts 2003). Even though the main topic of this paper is design storm events,
Another concern is the need to preserve the water balance. In hydraulics, and, in particular sewer hydraulics, must be addressed
this case, continuous simulation might be used as a screening as well, because it may adversely impact DRP, which so far has
tool, but a storm rainfall event and (or) a melt plus rain event been considered as dependent on the choice of rainfall data only
might be used for design. Here, the requirement to preserve the (actually not just here, but also in drainage practice; see for exam-
water balance would require much better abstractions model and ple the ASCE drainage manual, ASCE 1992). Contrary to many
resulting specication of antecedent conditions as further dis- other elds of civil engineering practice (e.g., structural design),
cussed in the next section. While low impact development (LID) the hydraulic design of sewers does not allow for any safety fac-
applications have been successfully tested at small scales (site, tors. Yet, the hydraulics of sewer systems indicates ample oppor-
neighbourhood), large scale applications (watershed) in Delaware tunities for deviations from ideal steady-state conditions, for
indicate that mimicking predevelopment hydrology using LID is which sewers are designed. Firstly, pipes are designed for open

Published by NRC Research Press


112 Can. J. Civ. Eng. Vol. 40, 2013

channel ow, with peak capacity occurring for the relative depth (3) Depth of rainfall: Intensitydurationfrequency relations are
of about 0.95. Under unsteady ow and with presence of surface likely to be affected by climate change. There are some prob-
waves, transition to pressure ow may occur unexpectedly, and lems with existing published idf relations, the impacts of
this leads to further increases of local minor losses at sewer junc- which could be of the same order as those predicted due to
tions in manholes. Experimental studies indicate that local losses climate change (i.e., looking at the next 2540 years). These
somewhat throttle the system capacity and for a low degree of include (a) unreliable relations due to small samples, (b) un-
pressurization (pressure heads from 1.0 to 1.3 times the pipe di- derestimation of storm rainfall depths of longer durations at
ameter, measured from the pipe invert), system capacity increases some locations where the data series may include events from
only little. This throttling effect may be further magnied by two populations, and (c) the fact that a correction factor is no
junction designs (without benching) contributing to formation of longer applied to the tipping bucket data.
a vortex ow pattern in the junction manhole, resembling the (4) Temporal distribution (design storm): Differences in peak dis-
vortex brake used to control ow (Marsalek 1984). Other reduc- charge or maximum storage resulting from different design
tions in ow capacity arise from blockage of inlets (by debris or hyetographs are likely to be larger than differences due to
snow and ice in winter), or blockage of pipes by sediment or large climate change (i.e., looking at the next 2540 years). There is
debris, or cross-sections of sewers for a variety of reasons. It is only a need for consistency in specifying design storms.
most recently that calls have been made to apply safety factors in (5) Antecedent conditions: Current practice regarding anteced-
hydraulics design of sewers to account not only for climate ent conditions is not consistent with the current state-of-
change, but also for addressing the water ponding and ooding knowledge of ood generating mechanisms and available
issues caused by blockage of sewer inlets or pipes. The latter prob- abstractions algorithms. Correct specication of antecedent
lems are of the operational nature and can be mitigated by im- conditions is extremely important for major system design
proved reporting of maintenance problems, quick response of and the protection of critical infrastructure.
municipal maintenance crews to blockages at critical points of (6) Areal reduction factor: Updated ARFs should be specied for
the sewer network, and implementation of more anticipatory major system design, particularly in large urban areas, by
maintenance programs in general (Schimek et al. 2008). Thus, the revisiting the existing ARFs and updating them by analyzing
authors believe that some recognition of uncertainty in hydraulic available Canadian storm data.
design should be made, thereby increasing the robustness of (7) New concerns: These may require changes in criteria for the
drainage design. design event from just a storm rainfall event to two design
events, a storm rainfall event and a snowmelt plus rain event.
Summary (8) Uncertainty in the hydraulic modeling of storm sewers: Un-
The use of design storm events has been common practice in certainty should be allowed for to increase the robustness of
Canada and the United States for over 100 years, beginning with drainage design.
the sizing of sewers and evolving over time to include sizing of
bridge and culvert openings, minor and major stormwater infra- Postscript major systems
structure, and regulatory oodplains as well as master drainage If the list of recommendations appears to be formidable, and it
planning. Over the last 4050 years, the knowledge base with is, the authors believe that the maximum benet for effort ex-
respect to design storm events has expanded signicantly, in par- pended, either in the current climate or the future climate, would
ticular, in the development of idf relations, design rainfall tem- result from a guideline for the design of major systems. Such a
poral distributions, and areal reduction factors. In other areas, for guideline would include (i) a locally-specic DRP, (ii) the design
example, specication of antecedent wetness, advances have been depth of rainfall, (iii) the rainfall temporal distribution, (iv) the
more modest. parameters of the abstractions model, and (v) the areal reduction
As a reaction to the potential impact of climate change on the factor.
performance of hydraulic structures and stormwater infrastruc-
ture, some investigators have focused on storm rainfall, idf rela- References
tions, to the exclusion of other parameters in the design process. Alberta Transportation. 2004. Guidelines on extreme ood analysis. Civil Proj-
To put the impacts of climate change in perspective and to iden- ects Branch, Edmonton, Alta. (available on line: http://www.
tify inadequacies of current practice for the existing climate, the infratrans.gov.ab.ca/INFTRA_Content/docType125/Production/
gdlextrmd.pdf; visited Sep. 20, 2012).
taxonomy of design storm events for various applications has Aleri, L., Laio, F., and Claps, P. 2008. A simulation experiment for optimal
been developed. For each task, an estimate of the sensitivity to design hyetograph selection. Hydrological Processes, 22: 813820. doi:
climate change is made, the likelihood of changes in practice is 10.1002/hyp.6646.
evaluated, and other needs are identied to improve the results Allen, R.J., and DeGaetano, A.T. 2005. Areal reduction factors for two eastern
United States regions with high rain-gauge density. Journal of Hydrologic
for the existing climate. As a result, the following specic conclu- Engineering, 10(4): 327335. doi:10.1061/(ASCE)1084-0699(2005)10:4(327).
sions can be made. ASCE. 1992. Design and construction of urban stormwater management sys-
tems. ASCE Manual and Reports of Engineering Practice No. 77. American
Conclusions and recommendations Society of Civil Engineers, New York, NY.
Ashley, R., Blanksby, J., and Cashman, A. 2008. Adaptable urban drainage
(1) Design return period (understood here as an administrative addressing change in intensity, occurrence and uncertainty of stormwater
(AUDACIOUS). London, UK. UK Climate Impacts Programme, Department of
risk-management measure) does not appear to be sensitive to Energy and Climate Change.
climate change and, with two exceptions, is unlikely to Bengtsson, L., and Niemczynowicz, J. 1986. Areal reduction factors from rain
change. The two exceptions are (a) explicit specication of the movement. Nordic Hydrology, 17: 6582.
DRP for the protection of critical infrastructure and (b) possi- Chow, V.T. 1964. Handbook of applied hydrology. McGraw-Hill, New York, NY.
Cobbina, A.S. 2007. The impact of climate change on Toronto precipitation
ble increase in the DRP for urban infrastructure where there patterns and stormwater infrastructure. M.Sc. thesis, Department of Civil
are cases of ooding not caused by deciencies in design (e.g., Engineering, University of Toronto, Toronto, Ont.
no major system). The DRP should not be changed as a hedge England, J.F., Sankovich, V.L., and Caldwell, R.J. 2011. Review of Probable Maxi-
against climate change. mum Precipitation procedures and databases used to develop hy-
drometeorological reports. U.S. Dept. of Interior, Bureau of Reclamation,
(2) Critical storm duration: It is somewhat sensitive to climate Denver, Colo. (ftp://ftp.usbr.gov/jengland/NRC/./Review_PMP_
change (certainly on a natural watershed), but research and Procedures_SE_US.pdf; visited Sep. 20, 2012).
guidance is required for storage structures. Environment Canada. 2004. Threats to water availability in Canada. Nat. Wat.

Published by NRC Research Press


Watt and Marsalek 113

Res. Institute (NWRI), Burlington, Ontario. NWRI Scientic Assessment Re- Packman, J.C., and Kidd, C.H.R. 1980. A logical approach to the design storm
port Series No. 3. concept. Water Resources Research, 16(6): 9941000. doi:10.1029/
Greer, R.K. 2008. Mimicking predevelopment hydrology using LID: time for a WR016i006p00994.
reality check? Edited by N. She and M. Clar. In CD-ROM Proceedings of the 2008 Peyron, N., Nguyen, V-T-V., and Rivard, G. 2002. An optimal design storm pattern
International Low Impact Development Conference. Seattle, Wash., Novem- for urban runoff estimations in Southern Quebec. In Proceedings, Annual
ber 1619, 2008. Available from ASCE, Reston, Va., 5p. Conference of the Canadian Society for Civil Engineering, Montreal, Quebec,
Guillon, A., Kovacs, Y., Lowera, L., and Senechal, C. 2008. Investigation on how June 58, pp. 110.
Control of source-control facilities is organized in France. In Proceedings of Pilgrim, D.H. (Editor). 1987. Australian rainfall and runoff, a guide to ood esti-
the 11th International Conference on Urban Drainage, Edinburgh, U.K., Aug. mation. The Institution of Engineers, ACT, Australia.
31 Sep. 5, 10 p. Pilgrim, D.H., and Cordery, I. 1975. Rainfall temporal patterns for design oods.
Haan, C.T., Bareld, B.J., and Hayes, J.C. 1994. Design hydrology and sedimentol- Journal of Hydraulics Engineering (ASCE), 101(HY1): 8195.
ogy for small catchments. Academic Press Inc., San Diego, Calif. Pilgrim, D.H., Cordery, I., and French, R. 1969. Temporal patterns of design
Hailegeorgis, T.T., and Burn, D.H. 2009. Quantifying the uncertainty in modelled rainfall for Sydney. Institution of Engineers, Australia, Civil Engineering
estimates of future extreme precipitation events. Department of Civil and Transactions, CE11: 914.
Environmental Engineering, University of Waterloo, Waterloo, Waterloo, Schimek, G., Fleming, P., and Hartley, D. 2008. Preparing for an uncertain fu-
Ont. ture: Seattle Public Utility's climate change and urban drainage adaptation
He, J., Valeo, C., and Bouchart, F.J.-C. 2006. Enhancing urban infrastructure strategy. In Proceedings of the 11th International Conference on Urban Drain-
investment planning practices for a changing climate. Water Science & Tech- age, Edinburgh, U.K., Aug. 31 Sep. 5, 10 p.
nology, 53(10): 1320. doi:10.2166/wst.2006.292. Sivapalan, S., and Bloschl, G. 1998. Transformation of point rainfall to areal
Hicks, W.I. 1944. A method of computing urban runoff. Transactions of the rainfall: intensity-duration-frequency curves. Journal of Hydrology, 204(14):
American Society of Civil Engineers (ASCE), 109: 12171253. 150167. doi:10.1016/S0022-1694(97)00117-0.
Hogg, W.D. 1980. Time distribution of short duration rainfall in Canada. In Svensson, C., and Jones, D.A. 2010. Review of methods for deriving areal reduc-
Proceedings Canadian Hydrology Symposium: 80, Ottawa, Ont. pp. 5363. tion factors. Journal of Flood Risk Management, 3(3): 232245. doi:10.1111/j.
Hogg, W.D. 1982. Distribution of design rainfall with time: design consider- 1753-318X.2010.01075.x.
ations. Paper presented at the American Geophysical Union Chapman on Terstriep, M.L., and Stall, J.B. 1974. The Illinois urban drainage area simulator,
Rainfall Rates, Urbana, Ill., April 2729. ILLUDAS. Bull. No. 58. Illinois State Water Survey, Urbana, Ill.
Hoppe, H. 2008. Impact of input data uncertainties on urban drainage models: Tholin, A.L., and Keifer, C.J. 1960. The hydrology of urban runoff. Transactions,
climate changea crucial issue? In Proceedings of the 11th International ASCE, 125: 13081379.
Conference on Urban Drainage, Edinburgh, U.K., Aug. 31 Sep. 5, 10 p. UDFCD. 1969. Urban storm drainage manual. Urban Drainage and Flood Control
Huff, F.A. 1967. Time distribution of rainfall in heavy storms. Water Resources District, Denver, Colo.
Research, 3(4): 10071019. doi:10.1029/WR003i004p01007. USACE. 2000. Hydrologic modeling system HEC-HMS. Technical reference man-
Institution of Civil Engineers, Australia. 1977. Australian rainfall and runoff: ual. United States Army Corps of Engineers, Davis, Calif.
ood design and analysis. The Institution of Civil Engineers, Australia. USDA. 1968. Hydrology. Suppl. A to Sec. 4, Engineering Handbook. Soil Conser-
Jolly, J.P., Jobin, D.I., and Lodewyk, S. 2008. Weather radar derived rainfall areal vation Service, U.S. Department of Agriculture.
reduction factors. Presented at Weather Radar and Hydrology 2008 Interna- USDA. 1986. Urban hydrology for small watersheds. Tech. Release No. 55. Soil
tional Symposium, Grenoble, France. Conservation Service, U.S. Department of Agriculture, Washington, D.C.
Keifer, C.J., and Chu, H.H. 1957. Synthetic storm pattern for drainage design. USWB. 1963. Rainfall frequency atlas for the United States for durations from 30
ASCE Journal of Hydraulics, Division, 83(HY4): 1332:11332:25. minutes to 24 hours and return periods from 1 to 100 years. Paper 40. U.S.
Kidd, C.H.R., and Packman, J.C. 1980. Selection of design storm and antecedent Weather Bureau Tech.
condition for urban drainage design. Institute of Hydrology, Wallingford, U.S. Federal Energy Regulatory Commission. 2002. Engineering guidelines for
UK. (IH Report No.61) (unpublished). 54 p. evaluation of hydropower projects. Washington, D.C.
Kuichling, E. 1889. The relation between rainfall and the discharge in sewers in Viessman, W., and Lewis, G.L. 1996. Introduction to hydrology. HarperCollins
populous districts. Transactions of the American Society of Civil Engineers College Publishers, New York, N.Y.
(ASCE), 20: 3740. Waters, D., Watt, W.E., Marsalek, J., and Anderson, B.C. 2003. Adaptation of a
Mailhot, A., and Duchesne, S. 2010. Design criteria of urban drainage infrastruc- storm drainage system to accommodate increased rainfall resulting from
tures under climate change. Journal of Water Resources Planning and Man- climate change. Journal of Environmental Planning and Management, 46(5):
agement (ASCE), 136(2): 201208. doi:10.1061/(ASCE)WR.1943-5452.0000023. 755770. doi:10.1080/0964056032000138472.
Marsalek, J. 1984. Head losses at sewer junction manholes. Journal of Hydraulic Watt, W.E., and Marsalek, J. 1977. What the practicing urban hydrologist needs
Engineering (ASCE), 110: 17731776. from the hydrometeorologist. In Proceedings of the 2nd Conference on Hy-
Marsalek, J., and Watt, W.E. 1984. Design storms for urban drainage design. drometeorology, Toronto, Ont., Oct. 2527, 1977, pp. 1523.
Canadian Journal of Civil Engineering, 11(3): 574584. doi:10.1139/l84-075. Watt, W.E., Chow, K.C.A., Hogg, W.D., and Lathem, K.W. 1986. A 1-h urban design
Metcalf and Eddy, Inc., University of Florida, and Water Resources Engineers, storm for Canada. Canadian Journal of Civil Engineering, 13(3): 293300.
Inc. 1971. Storm water management model, Vol. I: Final Report, Water Pollu- doi:10.1139/l86-041.
tion Control Research series 11024 DOC 07/71. Environmental Protection Westhoff, D.R., Watt, W.E., and Deong, Y. 1993. Development of urban design
Agency, Washington, D.C. storms from historic data: an application to Calgary, Alberta, Canada. In
Metcalfe, J.R., Routledge, B., and Devine, K. 1997. Rainfall measurement in Proceedings of the Sixth International Conference on Urban Storm Drainage,
Canada: changing observational methods and archive adjustment proce- Niagara Falls, Ont. pp. 294299.
dures. Journal of Climate, 10: 92101. doi:10.1175/1520-0442(1997)010<0092: Woolhiser, D.A., and Schwalen, H.C. 1960. Area-depth-frequency relations for
RMICCO>2.0.CO;2. thunderstorm rainfall in southern Arizona. Technical paper 527. Arizona
MNR. 2002. Technical guide, river & stream systems: ooding hazard limit. Agricultural Experiment Station, The University of Arizona, Tucson, Ariz.
Ontario Ministry of Natural Resources, Peterborough, Ont. Yen, B.C. 1975. Risk based design of storm sewers. Rept. IN141. Hydraulics Re-
Nishat, S., Guo, Y., and Baetz, B.W. 2010. Antecedent soil moistures conditions of search Station, Wallingford, U.K.
different soil types in South-western Ontario, Canada. Hydrological Pro- Yen, B.C., and Chow, V.T. 1980. Design hyetographs for small drainage struc-
cesses, 24: 24172424. tures. Journal of the Hydraulics Division (ASCE), 106(HY6): 10551076.
Oberts, G.L. 2003. Cold climate BMPS: Solving the management puzzle. Water Yen, B.C., and Chow, V.T. 1983. Local design storm. Vol. II. Report No. FHWA/RD-
Science and Technology, 48(9): 2132. 82-064. Federal Highway Administration, Washington, D.C.

Published by NRC Research Press


Copyright of Canadian Journal of Civil Engineering is the property of Canadian Science Publishing and its
content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for individual use.