You are on page 1of 15

Hydrological Sciences Journal

ISSN: 0262-6667 (Print) 2150-3435 (Online) Journal homepage: https://www.tandfonline.com/loi/thsj20

Estimating effective impervious area in urban


watersheds using land cover, soil character and
asymptotic curve number

Ali Ebrahimian, John S. Gulliver & Bruce N. Wilson

To cite this article: Ali Ebrahimian, John S. Gulliver & Bruce N. Wilson (2018) Estimating
effective impervious area in urban watersheds using land cover, soil character
and asymptotic curve number, Hydrological Sciences Journal, 63:4, 513-526, DOI:
10.1080/02626667.2018.1440562

To link to this article: https://doi.org/10.1080/02626667.2018.1440562

Published online: 09 Mar 2018.

Submit your article to this journal

Article views: 1409

View related articles

View Crossmark data

Citing articles: 8 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=thsj20
HYDROLOGICAL SCIENCES JOURNAL, 2018
VOL. 63, NO. 4, 513–526
https://doi.org/10.1080/02626667.2018.1440562

Estimating effective impervious area in urban watersheds using land cover, soil
character and asymptotic curve number
a,b
Ali Ebrahimian , John S. Gulliverb and Bruce N. Wilsonc
a
Department of Civil and Environmental Engineering, Villanova University, Villanova, Pennsylvania, USA; bSaint Anthony Falls Laboratory,
Department of Civil, Environmental, and Geo-Engineering, University of Minnesota, Minneapolis, Minnesota, USA; cDepartment of
Bioproducts and Biosystems Engineering, University of Minnesota, Saint Paul, Minnesota, USA

ABSTRACT ARTICLE HISTORY


Knowledge of the effective impervious area (EIA) or the degree to which impervious surfaces are Received 20 March 2017
hydraulically connected to the drainage system is useful for improving hydrological and environ- Accepted 8 January 2018
mental models and assessing the effectiveness of green stormwater infrastructure in urban EDITOR
watersheds. The goal of this research is to develop a method to estimate EIA fraction in urban R. Woods
watersheds using readily available data. Since EIA is dependent on rainfall–runoff response and
cannot be solely determined based on the physical characteristics of a watershed, the EIA is ASSOCIATE EDITOR
linked with the asymptotic curve number (CN), a watershed index that represents runoff char- R. van Nooijen
acteristics. In order for the method to be applicable to ungauged watersheds, the asymptotic CN KEYWORDS
is predicted using land cover and soil data from 35 urban catchments in Minnesota and Texas, effective impervious area;
USA. Similar data from 11 other urban catchments in Wisconsin and Texas, USA, are used to impervious area;
validate the results. A set of runoff depth versus EIA fraction curves is also developed to assess the stormwater; runoff; rainfall;
impact of EIA reduction on discharge from an urban watershed in land-use planning studies. urban catchments;
ungauged watersheds;
urban watersheds

1 Introduction defined as the portion of TIA that is directly connected


to the storm sewer system. DCIA and EIA are different
The impervious character of an urban area can be
measures of connectedness in urban watersheds. While
defined by its total extent and the degree of connectiv-
DCIA is independent of rainfall and runoff and can be
ity to the drainage system. Total impervious area (TIA)
measured solely based on field surveys and map mea-
is a characteristic defining the extent of imperviousness
surements, the determination of EIA inherently con-
in a watershed. TIA is often used as an integrative
siders the losses along the flow paths of runoff over
indicator of urban development and its consequences
impervious surfaces. For example, an urban street that
for the urban ecosystem. However, the inability of TIA
is directly connected to the sewer system (i.e. DCIA)
to represent the alterations of hydrological processes
may not be fully EIA because of surface depressions
and the prediction of pollutant loads in environmental
and cracks in the pavement, vegetation interception, or
models suggests the need for another parameter that
clogged storm drains. Hence, EIA is a more precise
better represents the impact of impervious surfaces on
measure of the impact of impervious area on urban
the hydrology, physiography, water quality, and biol-
runoff. More details on the difference between EIA and
ogy of watersheds (Booth and Jackson 1997, Hatt et al.
DCIA can be found in Ebrahimian et al. (2016a,
2004, Shuster et al. 2005, Ebrahimian et al. 2016a). In
2016b).
the context of urban hydrology, effective impervious
Conventional stormwater networks in urban areas
area (EIA), or the portion of TIA that is hydraulically
often need to be expanded to meet the larger runoff
connected to the storm sewer system, is a better para-
volumes and rates due to urban development. This
meter in determining urban runoff and water quality
expansion is usually difficult because of several techni-
response (Alley and Veenhuis 1983, Lee and Heaney
cal, traffic, and environmental limitations and issues
2003, Ebrahimian 2015a). A brief review of the impli-
(Ebrahimian et al. 2015). To address this issue, green
cations and applications of EIA can be found in
infrastructure stormwater control measures are being
Ebrahimian et al. (2016b).
adopted to reduce the runoff volume entering storm-
Directly connected impervious area (DCIA) is
water networks. One of the main functions of green
another parameter related to impervious surfaces. It is

CONTACT Ali Ebrahimian ali.ebrahimian@villanova.edu


© 2018 IAHS
514 A. EBRAHIMIAN ET AL.

infrastructure is disconnecting impervious areas (i.e. cross storm sewers may seem reasonable. However,
reducing EIA) and enhancing infiltration of rainfall in determining the connectivity of rooftops to the drai-
catchments (Mueller and Thompson 2009). EIA reduc- nage system requires field investigation. In addition,
tion by green infrastructure practices needs to be deter- the results from this process would likely represent
mined for planning, crediting, and assessing the DCIA, not EIA.
effectiveness of those practices. Quantification of EIA To address the need for a GIS-based method to
is also essential for accurate hydrological modeling and estimate EIA in urban watersheds, Han and Burian
investigating environmental degradation and water (2009) presented a method that has the advantage of
quality in urban watersheds (Jones et al. 2003). being an automated GIS tool for developing a spatial
map of DCIA related to impervious surfaces such as
streets and parking lots in ungauged urban watersheds.
1.1 Background
However, there are some challenges that hinder the
Current methods for determining the amount of EIA wide use of Han and Burian’s method: First, one
mostly rely on empirical equations (e.g. Alley and needs to have a dataset containing the locations of
Veenhuis 1983, Sutherland 1995) that have been devel- conveyance elements and inlet points of the storm-
oped for specific areas and, in general, are ineffective in water collection system, which may not be available
estimating EIA in terms of TIA (Roy and Shuster in an urban area. Second, the method is not able to
2009). Field investigations along with map measure- recognize the connectivity of rooftops to the drainage
ments can be used to estimate EIA. However, field system, requiring determination by the user, a process
investigations are often time consuming, costly, and that can add significant time and expense to the
even impractical in large watersheds, and ultimately analysis.
result in DCIA (not EIA) values. The best estimate of
EIA is typically obtained from the analysis of rainfall–
1.2 Objective
runoff data in a watershed (Boyd et al. 1993, 1994,
Ebrahimian et al. 2016a, 2016b). Ebrahimian et al. While knowledge of EIA is needed for a wide range of
(2016a, 2016b) presented a method that improves and studies including but not limited to hydrology,
generalizes the method of Boyd et al. (1993, 1994) for watershed management, and land-use planning, ade-
estimating EIA fraction (i.e. fEIA or EIA=At where At is quate methods for EIA determination based on readily
the total drainage area of watershed) in gauged urban available data do not exist (Yang et al. 2011, Fletcher
catchments where rainfall–runoff data are available. et al. 2013, Ebrahimian et al. 2016a). The overall goal of
Based on the analysis of observed rainfall–runoff data this research is to develop a method to estimate EIA
in 50 urban catchments using the successive weighted fraction for urban watersheds using only readily avail-
least square (WLS) method (Ebrahimian et al. 2016a) able spatial data.
for the EIA estimation, Ebrahimian et al. (2016b) In the following sections, the sources for the input
showed that determining EIA based on a representative GIS data as well as the study sites utilized for develop-
average value for the EIA/TIA ratio can be misleading. ing the method and validation of the results are pre-
Collecting runoff data from storm sewers requires sub- sented and described. The different steps towards
stantial time and instrumentation investment. Hence, development and validation of the method are then
most of the urban watersheds are ungauged (i.e. miss- elaborated and discussed. A set of curves is also devel-
ing runoff measurements), and a method is needed to oped to investigate the sensitivity of runoff depth to
estimate EIA in ungauged urban watersheds using data EIA in an urban watershed of interest. We conclude
that are readily available. Thus, the use of geographic with a discussion on advantages and limitations of the
information systems (GIS) and remote-sensing-based proposed method.
tools to estimate EIA becomes particularly attractive
due to its applicability to ungauged watersheds, and
2 Data sources
to the increasing quality and availability of spatial data.
Several studies (Taylor et al. 2004, Carle et al. 2005, The rainfall–runoff data from 35 urban catchments in
Yang et al. 2011, Vrebos et al. 2014) have estimated the USA were analyzed in this study to determine CN1
EIA by intersecting impervious area and drainage sys- in each catchment and develop a general framework for
tem GIS layers combined with field surveys, and using the fEIA estimation. Similar data from 11 other urban
additional assumptions on connectivity of impervious catchments in the USA were used to validate the
surfaces to drainage systems. Making assumptions on results. These sets of 35 and 11 catchments are referred
the connectivity of the streets that are located over or to herein as “calibration catchments” and “validation
HYDROLOGICAL SCIENCES JOURNAL 515

catchments,” respectively. All of the catchments are refers to combined residential and commercial land
presented in Table 1. The location of catchments in uses. All of the study catchments operate local mon-
the USA is also shown in Figure 1. As seen in Table 1 itoring programs to measure rainfall and runoff from
and Figure 1, among the 35 calibration catchments, 16 urban areas. The monitoring period in each catchment
are located in the Twin Cities metro area of Minnesota from which the required rainfall–runoff data in this
and 19 in the City of Austin, Texas. Also, of the 11 study were extracted and analyzed is presented in
validation catchments, three are located in Madison, Table 1. The rainfall data that were used for determin-
Wisconsin, and eight in Austin, Texas. The main land ing CN1 in calibration catchments were the same data
use of the study catchments is presented in Table 1. that were used in Ebrahimian et al. (2016b) for esti-
Most of the catchments have a residential land use but mating fEIA . A comprehensive discussion on the quality
with different development ages. In Table 1, transpor- of utilized rainfall data including outlier identification
tation land use denotes the drainage areas that include and addressing spatial variation of rainfall can be found
highways and airports. Also, mixed urban land use in Ebrahimian et al. (2016b).

Table 1. Description of the study catchments used for developing the method (calibration catchments) and validating the results
(validation catchments).
Row Catchment name Location Drainage Percentage of HSGs Land use Monitoring years
area (ha) A B C D
Calibration:
1 AHUG Saint Paul, MN 15.9 100 Residential 2007–2012
2 GCP Saint Paul, MN 51.8 100 Residential 2008–2012
3 Como 3 Saint Paul, MN 185.8 100 Residential 2009–2012
4 Sarita Saint Paul, MN 376.0 100 Institutional 2006,2008–2009
5 TBEB Saint Paul, MN 377.2 8.9 91.1 Residential 2006–2012
6 EK Saint Paul, MN 451.6 100 Residential 2005–2012
7 PC Saint Paul, MN 579.9 43.5 56.5 Residential 2005–2012
8 SAP Saint Paul, MN 1007.3 100 Residential 2005–2012
9 TBO Saint Paul, MN 2034.8 16.4 83.6 Residential 2007–2012
10 MG1 Maple Grove, MN 5.5 100 Residential 2001–2003, 2005–2006
11 MG2 Maple Grove, MN 3.5 100 Residential 2001–2003, 2005–2006
12 P1 Plymouth, MN 5.1 100 Residential 2001–2003, 2005–2006
13 P2 Plymouth, MN 6.8 100 Residential 2001–2003, 2005–2006
14 P3 Plymouth, MN 5.6 100 Residential 2001–2003, 2005–2006
15 Tapestry Minnetonka, MN 11.1 100 Residential 2010
16 Smith Bloomington, MN 55 NA Transportation 2004–2005
17 BW1 Austin, TX 146.3 100 Residential 2012–2014
18 EBA Austin, TX 14.3 71.8 28.2 Residential 2000–2003
19 EHA Austin, TX 20.8 100 Residential 1994–2002
20 ERA Austin, TX 40.4 90.9 9.1 Transportation 1994–1999
21 HI Austin, TX 1.2 38.6 61.4 Residential 1985–1987
22 HPA Austin, TX 17.4 100 Residential 2000–2003
23 LCA Austin, TX 84.9 0.5 86.6 12.9 Residential 1992–1999
24 LOA Austin, TX 5.4 100 Residential 2008–2011
25 LUA Austin, TX 5.5 100 Mixed Urban 1989–1996
26 MBA Austin, TX 82.1 80.4 19.6 Residential 1993–1995
27 OFA Austin, TX 0.6 47.5 52.5 Commercial 1993–1997
28 PP1 Austin, TX 2.0 100 Residential 2009–2012
29 PP2 Austin, TX 1.8 100 Residential 2009–2012
30 PP3 Austin, TX 0.9 100 Residential 2009–2012
31 RRI Austin, TX 6.4 100 Residential 2003–2007
32 SCA Austin, TX 2.3 100 Residential 2006–2010
33 TBA Austin, TX 20.0 91.9 8.1 Residential 1996–2000
34 TCA Austin, TX 16.5 100 Residential 1993–1997
35 TPA Austin, TX 16.8 100 Residential 1993–1997
Validation:
1 ASB Madison, WI 21.2 77.1 22.9 Residential 2001–2006
2 HFBB Madison, WI 23.8 76.4 23.6 Residential 2002–2006
3 CB Madison, WI 36.3 77 23 Residential 2001–2006
4 CBL Austin, TX 339.8 13.1 4.8 82.1 Mixed Urban 2001–2016
5 CMI Austin, TX 40.5 96.3 3.7 Mixed Urban 1996–2002
6 CTJ Austin, TX 11.7 92.1 7.9 Residential 2005–2007
7 CTK Austin, TX 9.6 55.5 44.5 Residential 2005–2007
8 HLA Austin, TX 133.2 3.4 96.6 Residential 1985–1996
9 SBI Austin, TX 4.4 100 Commercial 2009–2013
10 SWI Austin, TX 6.6 100 Industrial 1995–1997
11 W3A Austin, TX 1061.7 8.5 44.9 46.6 Mixed Urban 1993–1995, 1997–2016
516 A. EBRAHIMIAN ET AL.

Figure 1. Location of the study catchments in the USA.

The required GIS data for this study include land delineation of drainage area for that catchment. The
cover and hydrologic soil group (HSG). High-resolu- SSURGO database is a general view of the soils in a
tion spatial land-cover data for the City of Saint Paul, region, and urban soils are known for being disturbed
Minnesota, developed by Kilberg et al. (2011) were and redistributed. Even so, transportation costs will
used in this study. Land-cover data for other study typically dictate that disturbed and redistributed soils
sites were provided by the Three Rivers Park District, do not come from areas that are remote to the
City of Minnetonka, and City of Bloomington in watershed, and SSURGO is the best source available
Minnesota, USGS-Wisconsin Water Science Center, to all watersheds.
and the City of Austin Watershed Protection
Department for their corresponding catchments.
Land-cover data for urban areas in the USA are usually
3 Method
available through Cities, Counties, Watershed Districts A new method is developed herein to obtain EIA
or other responsible entities. However, in the absence estimates for urban watersheds based on the spatial
of land-cover data from those entities, they can be analysis of watershed characteristics. Our goal is to
extracted from the National Land Cover Database link EIA using a simple hydrological framework rather
(NLCD) (http://www.mrlc.gov). Hydrologic soil group than relying only on a statistical framework obtained
spatial data for the study areas were extracted from the by a regression analysis. The asymptotic curve number
NRCS-Soil Survey Geographic (SSURGO) database, (CN), which is a rainfall-dependent CN and defined in
which is publicly available (http://websoilsurvey.sc. this study by the asymptote of an exponential relation-
egov.usda.gov/App/HomePage.htm). The percentages ship between CN and rainfall depth, provides the fra-
of different HSGs in each catchment are summarized mework for capturing the runoff characteristics of
in Table 1. The HSG distribution was not determined watersheds. The EIA fraction is defined as a function
for Smith catchment, as we did not access to the of this index. Then, rainfall–runoff, land-cover, and
HYDROLOGICAL SCIENCES JOURNAL 517

soil data from 35 urban catchments with different sizes then uses this CN with a given rainfall depth (P) to
and physical characteristics are analyzed to predict the determine runoff depth (Q). The CN is not conven-
asymptotic CN from watershed characteristics. Land- tionally considered variable with P. However, research-
cover and soil datasets (which are both readily available ers have back-calculated CN corresponding to
through the national and sometimes local datasets) will measured Q and P and have found variability in CN.
be the only input data for estimating EIA fraction in Based on this fact, Hawkins (1993) developed the
the proposed method. asymptotic fitting method using observed rainfall–run-
off data. In this method, the rainfall and runoff depth
data (P and Q) are sorted separately and then realigned
3.1 Asymptotic curve number based on rainfall– on a rank-order basis. This is called frequency match-
runoff data ing and is done to equate return period of rainfall and
The CN method, which was first introduced by the US runoff events (Hjelmfelt 1980, Hawkins 1993). By using
Department of Agriculture (USDA), Soil Conservation Equations (1) and (2) with λ = 0.2, the storage index, S,
Service (SCS) (currently referred to as Natural for each pair of rainfall and runoff depth is deter-
Resources Conservation Service or NRCS) in 1954 mined as:
h  1 i
(Ponce and Hawkins 1996, NRCS 2004), is an empirical
S ¼ 5 P þ 2Q  4Q2 þ 5PQ 2 (3)
event-based hydrological model which has been pre-
dominantly used in hydrological analyses of ungauged
A unique CN is then determined for each storm event
watersheds. The CN values for urban areas are pre-
in the ordered dataset using the relationship between
sented based on land-cover type and conditions as well
CN and S. When the obtained CN values are plotted
as HSGs by the technical release 55 (TR-55) of USDA
against rainfall depths (P), three different patterns of
(USDA 1986). The NRCS classifies soils in the USA
CN vs P have been observed by researchers working on
into four HSGs, A, B, C, and D, based on estimates of
different watersheds around the world (Hawkins 1993,
soil runoff potential. Soils in Group A have the smallest
Hawkins et al. 2009, D’Asaro et al. 2014). The most
runoff potential and those in Group D the greatest. The
common CN–P pattern in small watersheds is labeled
details of this classification can be found in USDA
the standard behavior, where CN declines with increas-
(1986). Since the CN values for urban districts in TR-
ing P, but asymptotically approaches a constant value
55 are based on assumptions for both impervious and
for larger storms. This asymptotic value theoretically
pervious areas that tend to be approximate, we evalu-
corresponds with a rainfall depth of infinity (i.e. CN1 )
ated the asymptotic CN at the basin scale from rain-
(Hawkins 1993, D’Asaro et al. 2014). In the CN–P
fall–runoff events using the asymptotic fitting method
graph, if CN declines with increasing P but does not
of Hawkins (Hawkins 1993, Hawkins et al. 2009).
approach a constant value in larger storms, the CN–P
The CN equations for estimating direct runoff depth
pattern is labeled complacent behavior. The asymptotic
in terms of rainfall event depth were introduced by
CN cannot be determined for the watersheds with this
USDA-SCS (SCS 1964, 1972, 1985, NRCS 2004) as
CN–P behavior. The third CN–P pattern is the violent
follows (Q ¼ 0 for P < Ia ):
behavior in which CN declines with increasing P for
ðP  Ia Þ2 smaller storms but suddenly increases and approaches
Q¼ for P  Ia (1) a constant value in larger storms (i.e. CN1 ). To deter-
P  Ia þ S
mine CN1 in catchments with a standard CN–P pat-
Ia ¼ λ S (2) tern, Hawkins (1993) and other researchers have used
where Q is direct runoff depth of the storm (inches or the asymptotic Equation (4) to be fitted to P–CN data.
mm), P is rainfall depth (in. or mm), S is potential CN ¼ CN1 þ ð100  CN1 ÞekP (4)
maximum retention (in. or mm), Ia is the initial
abstraction of rainfall, and λ is the initial abstraction where k is a fitted constant.
ratio (Ia =S). The standard value of the initial abstrac-
tion ratio (λ) has been established by SCS as 0.2. The
3.2 Asymptotic curve number and EIA fraction
potential maximum retention (S) in Equation (1) is
transformed to a dimensionless index, CN, as The area weighted CN approach is used to determine a
CN = 25400/(S + 254), with S in mm. single CN for a watershed with mixed land uses and
Conventionally, one determines a CN for a soil types. This approach is widely used in the analysis
watershed based on its physical characteristics (i.e. of runoff from urban watersheds. By dividing the entire
land cover/land use and HSG) through TR-55 and drainage area of a watershed into two parts, impervious
518 A. EBRAHIMIAN ET AL.

and pervious areas, the weighted average CN for a CN data in the catchments of study and determine
watershed can be found from Equation (5) (Pandit CN1 and the fitting constant k for each catchment. The
and Regan 1998): obtained values of CN1 and k are presented in Table 2.
Table 2 also presents the values of fTIA , fEIA , and α for the
CNw At ¼ CNi Ai þ CNp Ap (5)
catchments of study. The TIA fraction ( fTIA ) is defined as
where CNw is the weighted average CN of the TIA=At , where At is the total drainage area of the
watershed, At is the total drainage area, CNi is the watershed. TIA in urban watersheds can be determined
CN for the impervious area, Ai is the area of imper- from land-cover data by summing up all the impervious
vious surfaces in the watershed, CNp is the CN for the surfaces. The fTIA values in Table 2 for calibration catch-
pervious area, and Ap is the area of pervious surfaces in ments were taken from Ebrahimian et al. (2016b). This
the watershed (Ap ¼ At  Ai ). value was not determined in the Smith catchment due to
An important assumption in the TR-55 CN for not accessing the delineation of this catchment. The fTIA
urban areas is that it considers only directly connected values for the validation catchments in Wisconsin and
impervious areas (or effective impervious area) as Texas were provided by USGS-Wisconsin Water Science
impervious. Equation (5) can then be modified as:
CNw At ¼ CNEIA AEIA þ CNr Ar (6) Table 2. Asymptotic CN, k, TIA fraction, EIA fraction and α
values in both calibration and validation catchments of study.
where CNEIA is the CN corresponding to effective Row Catchment name CN∞ k f TIA f EIA α
impervious area, AEIA is the area of effective imper- Calibration:
vious surfaces in the watershed, CNr is the CN corre- 1 AHUG 55.8 0.0168 0.507 0.137 0.5006
2 GCP 62.3 0.0117 0.438 0.250 0.5144
sponding to remaining area in the watershed, and Ar is 3 Como 3 44.7 0.0153 0.405 0.102 0.3945
the remaining area of the watershed (Ar ¼ At  AEIA ). 4 Sarita 48.6 0.0238 0.367 0.030 0.4804
5 TBEB 70.6 0.0219 0.447 0.193 0.6534
For our application, the CNs in Equation (6) are taken 6 EK 79.3 0.0153 0.562 0.376 0.6946
as the asymptotic CNs. 7 PC 76.1 0.0202 0.587 0.310 0.6755
8 SAP 73.2 0.0223 0.613 0.224 0.6745
CNw1 At ¼ CNEIA1 AEIA þ CNr1 Ar (7) 9 TBO 77.6 0.0204 0.473 0.284 0.7089
10 MG1 82.5 0.0600 0.405 0.117 0.8215
We hypothesize that the asymptotic CN for the total 11 MG2 71.9 0.0264 0.388 0.151 0.6867
12 P1 62.5 0.0219 0.380 0.204 0.5452
drainage area can be expressed as a weighted average 13 P2 63.9 0.0261 0.351 0.089 0.6176
CN (i.e. CNw ). For a fully impervious, connected sur- 14 P3 79.7 0.0569 0.273 0.096 0.7932
15 Tapestry 54.9 0.0175 0.237 0.169 0.4706
face the asymptotic CN is also approximated by the 16 Smith 75.1 0.0553 NA 0.076 0.7476
USDA TR-55 value of 98. By dividing both sides of 17 BW1 78.3 0.0387 0.460 0.152 0.7630
18 EBA 48.9 0.0179 0.404 0.093 0.4474
Equation (7) by At , the CN1 is defined as: 19 EHA 71.1 0.0132 0.434 0.336 0.5865
20 ERA 85.3 0.0467 0.460 0.179 0.8420
CN1 ¼ 98 fEIA þ CNr1 ð1  fEIA Þ (8) 21 HI 92.3 0.0902 0.500 0.281 0.9192
22 HPA 64.5 0.0096 0.450 0.300 0.5116
where fEIA is the EIA fraction and CNr1 is the asymp- 23 LCA 60.2 0.0258 0.225 0.064 0.5876
24 LOA 87.1 0.0474 0.422 0.196 0.8620
totic CNr . Given the availability of rainfall–runoff data, 25 LUA 95.1 0.0616 0.974 0.484 0.9421
CN1 in Equation (8) is known, but both fEIA and 26 MBA 90.9 0.0898 0.609 0.322 0.8939
27 OFA 95.6 0.1554 0.862 0.452 0.9549
CNr1 are unknowns. We define α as the ratio of 28 PP1 84.9 0.0403 0.497 0.229 0.8271
CNr1 =CNEIA1 or CNr1 =98: 29 PP2 82.2 0.0300 0.511 0.270 0.7794
30 PP3 62.3 0.0178 0.494 0.172 0.5605
CNr1 ¼ 98α (9) 31 RRI 84.2 0.0477 0.305 0.107 0.8428
32 SCA 77.7 0.0318 0.409 0.157 0.7536
33 TBA 65.5 0.0221 0.452 0.106 0.6296
By substituting Equation (9) into Equation (8), fEIA can 34 TCA 82.1 0.0721 0.374 0.097 0.8203
be evaluated as a function of CN1 and α: 35 TPA 74.5 0.0337 0.415 0.142 0.7204

CN1  98α Validation:


fEIA ¼ (10) 1 ASB 64.9 0.0256 0.343 0.109 0.6206
98ð1  αÞ 2 HFBB 52.4 0.0206 0.314 0.091 0.4882
3 CB 44.4 0.0167 0.298 0.086 0.4016
To investigate α and the possibility of presenting fEIA 4 CBL 78.2 0.0264 0.374 0.115 0.7719
5 CMI 80.3 0.0322 0.547 0.195 0.7753
solely as a function of CN1 , we analyzed the rainfall– 6 CTJ 85.8 0.0438 0.290 0.234 0.8369
runoff data from the 35 calibration catchments intro- 7 CTK 91.1 0.0494 0.392 0.348 0.8925
8 HLA 79.6 0.0351 0.391 0.138 0.7819
duced in Table 1 to determine CN1 in each catchment. 9 SBI 86.3 0.0284 0.680 0.381 0.8072
The aforementioned asymptotic fitting method 10 SWI 94.4 0.0804 0.604 0.463 0.9322
11 W3A 84.7 0.0388 0.490 0.227 0.8246
(Hawkins 1993) was applied to fit Equation (4) to P–
HYDROLOGICAL SCIENCES JOURNAL 519

Center and the City of Austin Watershed Protection validation. However, the effect of those uncertainties
Department, respectively. In another paper (Ebrahimian in CN1 values is noted later.
et al. 2016b), the authors determined the fEIA using the Figure 2 provides a plot of α against CN1 for the 35
analysis of observed rainfall–runoff data from urban study sites using values from Table 2. Weighted least
catchments in the USA including the 35 calibration square (WLS) regression with a similar weighting
catchments in the current study by the successive method as Ebrahimian et al. (2016a) was used in
weighted least square (WLS) method (Ebrahimian et al. Figure 2 to estimate α in terms of CN1 . To assign
2016a). The fEIA values in Table 2 were taken from larger weights to more reliable data and vice versa,
Ebrahimian et al. (2016b). Knowing both CN1 and fEIA the weight of observations in the utilized WLS method
for the study catchments, the α values in Equation (9) can was considered proportional to the inverse of variance
be obtained by rewriting Equation (10) as: of residuals. More details about this method can be
found in Ebrahimian et al. (2016a). A pseudo coeffi-
CN1  98 fEIA cient of determination (pseudo R2WLS ), according to
α¼ (11)
98ð1  fEIA Þ Willett and Singer (1988), was used to measure the
goodness of fit in the WLS regression. The value of
The α values in Table 2 are calculated based on pseudo R2WLS was found to be about 0.95, which indi-
Equation (11). cates a strong correlation between α and CN1 in
The CN–P pattern in all the study catchments was Figure 2, and hence, α can be estimated as a function
found to be standard. It should be noted that nine of of CN1 :
the calibration sites (Sarita, EK, P3, Smith, EHA,
α ¼ 0:0116 CN1  0:1601 (12)
HPA, LCA, TBA, and TCA) and two of the validation
sites (CBL and SBI) appeared to have less reliable Substituting Equation (12) into Equation (11) makes it
CN1 values because of issues such as (a) lack of possible to estimate the fraction of EIA as a function
large rainfall depths in the dataset (EK, SBI), (b) of CN1 :
presence of a gap between small and large rainfall 16  0:14 CN1
depths in the dataset and lack of enough large rainfall fEIA ¼ (13)
114  1:14 CN1
depths, which result in long extrapolations (Sarita,
Smith, EHA, HPA, TBA, CBL), (c) poor curve fit To investigate the applicability and limitations of the
(LCA), and (d) long extrapolation and poor curve fit presented relation between fEIA and CN1 , we note that
(P3 and TCA). Because of the limited number of Equation (9) should work in extreme conditions of fully
study sites, the analysis was performed using all of impervious ( fTIA ¼ 1) and pervious ( fTIA ¼ 0) water-
the 35 study sites for calibration and 11 sites for sheds. While fTIA ¼ 1 corresponds to CNr1 ¼ 98,

Figure 2. Plot of α versus CN1 in the calibration catchments.


520 A. EBRAHIMIAN ET AL.

fTIA ¼ 0 would correspond to a CNr greater than or in ungauged watersheds, are investigated herein to explore
equal to the minimum value of CN in urban areas (i.e. if they have any significant correlation with CN1 .
40, as reported by Hawkins et al. 2009, as the minimum The area-weighted average CN of a watershed,
CN in practice). This results in a range of [0.4,1] for α which is also called composite CN, can be determined
values: using the available CN values in TR-55 by the spatial
analysis of land cover/land use and HSG data (Zhan
0:4  α  1 (14) and Huang 2004, Hernández-Guzmán and Ruiz-Luna,
2013). We determined the composite CN values for
Limiting α in the range of [0.4,1] causes CN1 to be
nine catchments in Saint Paul, Minnesota (calibration
limited to the range of 49 and 100. This indicates that
catchments 1–9 in Table 1). However, no significant
Equation (13) is valid only for CN1 values greater than
correlation was found between the asymptotic and
49. For 40  CN1  49, α was assigned to be constant
composite CN values for those nine study sites. This
and equal to 0.4. Replacing α = 0.4 in Equation (10)
finding agrees with D’Asaro et al. (2014), who reported
leads to a two-criteria function for fEIA :
no significant relationship between asymptotic and
( composite CN values in 36 watersheds in Sicily, Italy.
0:0175 CN1  0:6987 if 40  CN1  49
fEIA ¼ 160:14 CN1
With the purpose of estimating asymptotic CN in
1141:14 CN1 if 49 < CN1  98 ungauged watersheds, D’Asaro et al. (2014) also inves-
(15) tigated the links between CN1 and some hydrological
and geomorphic–climatic characteristics of 46 water-
Figure 3 shows the plot of fEIA versus CN1 based on sheds in Sicily, Italy, but they found no significant
Equation (15). It is seen that fEIA increases with an relationships between CN1 and those characteristics.
increase of CN1 , where the rate of the change is low D’Asaro et al. (2014) recommended the estimation of
while 49 < CN1 < 80 and higher when CN1 is asymptotic CN values in terms of watershed character-
greater than 80. istics for small ungauged watersheds as a future
research topic.
We investigated the correlation between CN1 and
TIA fraction as another watershed parameter by using
3.3 Correlations between asymptotic curve the fTIA and CN1 values for study catchments from
number and watershed characteristics
Table 2. The regression slope was statistically signifi-
According to Equation (15), the determination of fEIA for a cant but the correlation was not strong enough to be
watershed requires an estimate of CN1 for that watershed. proposed for estimating CN1 . In order to improve
However, the direct determination of CN1 is based on accuracy of the regression analysis of CN1 , HSG was
rainfall–runoff data, which are not available in ungauged also considered as a potential factor explaining the
watersheds. Area-weighted-average CN, TIA, and HSG, all variability in CN1 . Our study sites have soils that
of which can be determined based on readily available data cover all of the HSGs (see Table 1). To avoid using

Figure 3. Plot of fEIA versus CN1 based on Equation (15) for the estimation of fEIA in urban watersheds in terms of CN1 .
HYDROLOGICAL SCIENCES JOURNAL 521

Table 3. Range of saturated hydraulic conductivity (Ksat) for that at least one of the slope terms in the regression
different hydrologic soil groups (HSGs) according to NRCS model is significantly different than zero. The signifi-
(2007). cance of each parameter in a model was determined
HSG Ksat (mm/h)
using a t-test. The best predictive model for the estima-
A > 36.07
B > 14.48 to ≤ 36.07 tion of CN1 in terms of fTIA and Ksat in urban water-
C > 1.52 to ≤ 14.48 sheds was obtained as:
D ≤ 1.52
CN1 ¼ 67:8 þ 30:0 fTIA
 
0:5 Ksat
 15:1 ð1  fTIA Þ log (17)
categorical variables in the regression analysis, the Ksat Ref
HSGs were represented by a saturated hydraulic con-
ductivity ðKsat Þ. Table 3 provides a range of Ksat for All of the parameters in this model (Equation (17))
each HSG. are significant; R2 , RMSE, and F  are equal to
A median log Ksat value between the upper and lower 0.39,10.9, and 10.24, respectively, and the model rea-
limits in each range was assigned to HSGs B and C. For sonably estimates CN1 in the extreme conditions of
the A and D type soils the value that Ksat was greater fTIA ¼ 1 and 0: The term “ð1  fTIA Þ” is included in the
than or less than, respectively, was used in place of the model to ensure that the characteristics of the per-
median. To use a dimensionless parameter in the regres- vious area (here log (Ksat/Ksat_ref)) have no effect on
sion analysis, we picked HSG D as a reference HSG and CN1 when fTIA ¼ 1. Details of the 16 investigated
considered the assigned value to each HSG as log(Ksat/ models and their corresponding statistics can be
Ksat_ref) where Ksat ref denotes the Ksat of HSG D. Then, found in Ebrahimian (2015b). The graphical represen-
the assigned values for HSGs A, B, C, and D are found as tation of Equation (17) is shown in Figure 4 for
1.37, 1.18, 0.49, and 0, respectively. For example, the different HSGs. By the use of Figure 4, CN1 can be
calculation for HSG C is as follows: estimated for a given fTIA and HSG.
 
Ksat
log ¼ logðKsat Þ  log ðKsat ref Þ 4 Results
Ksat ref
¼ 0:67  0:18 ¼ 0:49 (16) The general framework of the proposed GIS–CN
method for estimating fEIA in urban watersheds can
It should be noted that any other HSG could be used as be summarized as the following three steps.
reference. Obviously, by choosing HSGs other than D
as the reference, the assigned values of log (Ksat/Ksat_ref) (1) Obtain the total impervious area (TIA) and dis-
for HSGs would be different than the above-mentioned tribution of hydrologic soil groups (HSGs) for
values. Using the corresponding values of log (Ksat/ the watershed of study by the geospatial analysis
Ksat_ref) with the HSG D as reference, a weighted aver- of readily available land-cover and soil data
age value was calculated for each study site based on from national, regional, or local datasets.
the percentage of different HSGs in each site. Sixteen Calculate the TIA fraction ( fTIA ) and determine
regression models using different combinations of weighted average saturated hydraulic conductiv-
CNp , fTIA and log (Ksat/Ksat_ref) as independent vari- ity (Ksat ) of the soil for the watershed as
ables for the estimation of CN1 in calibration catch- described in Section 3.
ments were developed and investigated. Three (2) Estimate the asymptotic curve number of the
statistics, including R2 (coefficient of determination), watershed (CN1 ) as a function of fTIA and Ksat
RMSE (root mean square error or square root of the using Equation (17) or Figure 4.
average sum of the squared deviation between each (3) Determine the fraction of effective impervious
observed and predicted values of dependent variable) area ( fEIA ) as a function of CN1 using Equation
and F  (i.e. MSR/MSE, where MSR is the mean square (15) or Figure 3.
regression or the average sum of the squared deviation
between predicted value and the mean of observed In the following two sections, first the validity of
values of dependent variable, and MSE is the mean proposed GIS–CN method is evaluated and then the
square error), were used to investigate the goodness usefulness of the proposed method is further explored
of fit in each predictive model. Variables were evalu- by developing curves of runoff depth versus EIA frac-
ated for significance at the 5% level. A large F  means tion for different rainfall depths.
522 A. EBRAHIMIAN ET AL.

Figure 4. Graphical representation of Equation (17) for estimating CN1 based on fTIA and HSG.

4.1 Evaluating the proposed framework on the performance of the proposed method because the
rate of change of fEIA with CN1 is very low for CN1
In order to see how compatible the presented fEIA vs CN1
values between 50 and 80. Hence, the proposed curve in
plot (Fig. 3) is with the data from gauged watersheds (i.e.
Figure 3 (or Equation (15)) can be used for the planning
validation), pairs of (CN1 ; fEIA ) for both the calibration
level of hydrological, environmental, and urban land use
and validation catchments were added to Figure 3. The
studies.
result is presented in Figure 5. The data points in Figure 5
represent the actual CN1 and fEIA values for the study
catchments from Table 2. The 11 catchments with less
reliable CN1 values (i.e. Sarita, EK, P3, Smith, EHA, 4.2 Runoff depth as a function of EIA fraction
HPA, LCA, TBA, TCA, CBL, and SBI, as explained earlier The usefulness of the proposed method is further
in Section 3) are denoted with open symbols in Figure 5. explored by developing curves of runoff depth as a
While a good agreement is seen between the proposed function of EIA fraction and rainfall depth. These
curve and actual data in larger CN1 values, scatter of the curves provide useful information for land-use plan-
actual data around the curve is seen for smaller CN1 ners and allow the sensitivity of runoff depth to EIA
values. This scatter may be explained by the uncertainty fraction to be easily assessed.
in CN1 values, as the data points associated with the less The runoff depth versus EIA fraction curves are
reliable CN1 values (open symbols) have larger devia- derived for a typical range of rainfall depths for the
tions from the curve in Figure 5. In general, the scatter in design of green stormwater infrastructure using the
smaller CN1 values is not expected to have a severe effect NRCS CN method. To determine the CN from

Figure 5. Comparison of the actual fEIA and CN1 values in gauged watersheds with the proposed curve for ungauged watersheds.
HYDROLOGICAL SCIENCES JOURNAL 523

Figure 6. Histogram of the fitting parameter k for the calibration catchments with a standard CN–P pattern.

Equation (4) for specified rainfall depths, both CN1 fraction for three rainfall depths of 12.5, 25, and
and the exponential parameter (fitting constant) k are 50 mm, which lie in a typical range of rainfall depths
needed. CN1 for different fEIA values can be estimated for designing green stormwater infrastructure prac-
from Equation (15) or Figure 3. To find a representa- tices. Since the actual fEIA values in our study sites
tive k value for ungauged watersheds, a histogram of k were all less than 0.5, Figure 8 is presented for fEIA
was first plotted using the k values for the 35 calibra- between zero and 0.5.
tion catchments from Table 2. Figure 6 shows this To compare the presented curves of Q vs fEIA with
histogram. the observed values (i.e. validation), runoff depth was
Because of the variability of k in Figure 6, using a calculated in a similar manner to Figure 8, except
constant representative k value (e.g. mean or median) using the observed (actual) fEIA , CN1 and k values
is not reasonable. However, a power function was fit to of both the calibration and validation catchments
the k–CN1 plot for the 35 calibration sites (Fig. 7) as from Table 2 for the same rainfall depths (i.e. 12.5,
given in Equation (18). The k and CN1 values in 25, and 50 mm). The obtained pairs of ( fEIA , Q) from
Figure 7 were taken from Table 2. actual data were added to the plot of Q versus fEIA in
Figure 8.
k ¼ 6  1017 CN7:67
1 þ 0:0149 (18) As seen in Figure 8, the general trend of observed
By having both CN1 and k, CN can be estimated for data agrees with the proposed curves for all rainfall
different rainfall depths (P) from Equation (6). depths. Also, while agreement is seen between the
Consequently, runoff depth (Q) is obtained for differ- observed data and proposed curves for lower rainfall
ent rainfall depths using the NRCS CN method. in Figure 8 (i.e. P = 12.5 and 25 mm), more scatter is
Figure 8 illustrates the runoff depth in terms of EIA seen in higher rainfall (i.e. P = 50 mm). This may be

Figure 7. Variation of the fitting parameter k with asymptotic curve number (CN1 ) in the study catchments with a standard CN–P
pattern.
524 A. EBRAHIMIAN ET AL.

Figure 8. Comparison of the proposed curves of runoff depth against fEIA for urban watersheds with the observed (actual) data.

explained by the amount of contributing pervious area The proposed GIS–CN method is relatively simple
in a watershed. In higher rainfall depths, more pervious and can be applied to ungauged urban watersheds.
area is expected to contribute to the runoff generation While the asymptotic CN is determined as an inter-
process. Hence, other characteristics of contributing mediate step in the proposed GIS–CN method, it can
pervious area such as soil type, infiltration capacity be attractive for practitioners and modelers involved
and initial moisture content of the soil, as well as initial in computing and modeling runoff from urban
abstraction of the pervious surfaces, are added to the watersheds. Our investigations on the CN–P patterns
runoff generation process, which make it more com- in the 46 study catchments (35 calibration and 11
plex and can cause scatter in Figure 8. In general, it is validation catchments) with different sizes and phy-
concluded that the developed curves could be useful sical characteristics showed a standard exponential
tools in urban planning, especially for evaluating the pattern for CN–P plots in all of the catchments.
effect of implementing green stormwater infrastructure Using the GIS–CN method, a set of curves was also
systems throughout an urban watershed. developed to investigate the sensitivity of runoff
depth to EIA fraction in different rainfall depths.
The developed curves provide a tool to assess the
5 Conclusions
impact of EIA reduction (e.g. by implementing green
Available literature shows that EIA is a better para- stormwater infrastructure practices) on discharge
meter than TIA for representing the impact of imper- from an urban watershed of interest in a variety of
vious surfaces on the hydrology of urban watersheds. urban planning studies. One limitation of the
To estimate the EIA fraction ( fEIA ) in urban water- method is that it does not determine the location
sheds, a new method was developed using the analysis of EIA in the watershed. Also, the proposed method
of rainfall–runoff, land-cover and soil data from 35 has been developed using catchments with a drainage
urban catchments in the USA. The method was area of 1–2035 ha. Applicability of the proposed
further validated by the use of similar data from 11 method to larger catchments needs to be investigated
other urban catchments in the USA. The proposed in future studies. The presented general framework is
GIS–CN method is able to estimate EIA fraction as a capable of being applied to more urban watersheds
function of asymptotic CN of watersheds (CN1 ) and in order to further generalize the predictive models
predict CN1 in terms of TIA fraction ( fTIA ) and in the method.
representative saturated hydraulic conductivity of the
soil (Ksat ) as estimated from HSG in an urban
watershed. The required GIS data for the proposed Acknowledgements
method include land cover and HSG, both of which The authors would like to acknowledge the Capitol Region
are readily available from national, regional, or local Watershed District, Minnesota, Three Rivers Park District,
datasets. Minnesota, Cities of Minnetonka and Bloomington,
HYDROLOGICAL SCIENCES JOURNAL 525

Minnesota, the USGS-Wisconsin Water Science Center, and Ebrahimian, A., Gulliver, J.S., and Wilson, B.N., 2016b.
the Watershed Protection Department, City of Austin, Texas, Effective impervious area for runoff in urban watersheds.
for providing rainfall–runoff and GIS data. We also thank Hydrological Processes, 30, 3717–3729. doi:10.1002/hyp.
the Department of Forest Resources, University of Minnesota v30.20
for providing land-cover data in Saint Paul, Minnesota. Ebrahimian, A., Wilson, B.N., and Gulliver, J.S., 2016a.
Improved methods to estimate the effective impervious
area in urban catchments using rainfall-runoff data.
Disclosure statement Journal of Hydrology, 536, 109–118. doi:10.1016/j.
jhydrol.2016.02.023
No potential conflict of interest was reported by the authors. Fletcher, T.D., Andrieu, H., and Hamel, P., 2013.
Understanding, management and modelling of urban
hydrology and its consequences for receiving waters: a
Funding state of the art. Advances in Water Resources, 51, 261–
279. doi:10.1016/j.advwatres.2012.09.001
This research was funded by the Minnesota Local Road
Han, W.S. and Burian, S.J., 2009. Determining effective
Research Board.
impervious area for urban hydrologic modeling. Journal
of Hydrologic Engineering, 14 (2), 111–120. doi:10.1061/
(ASCE)1084-0699(2009)14:2(111)
ORCID
Hatt, B.E., et al., 2004. The influence of urban density and
Ali Ebrahimian http://orcid.org/0000-0002-3992-170X drainage infrastructure on the concentrations and loads of
pollutants in small streams. Environmental Management,
34 (1), 112–124. doi:10.1007/s00267-004-0221-8
References Hawkins, R.H., 1993. Asymptotic determination of runoff
curve numbers from data. Journal of Irrigation and
Alley, W.M. and Veenhuis, J.E., 1983. Effective impervious Drainage Engineering, 119 (2), 334–345. doi:10.1061/
area in urban runoff modeling. Journal of Hydraulic (ASCE)0733-9437(1993)119:2(334)
Engineering, 109 (2), 313–319. doi:10.1061/(ASCE)0733- Hawkins, R.H., et al., 2009. Curve number hydrology, state of
9429(1983)109:2(313) the practice. Reston, VA: ASCE.
Booth, D.B. and Jackson, C.R., 1997. Urbanization of aquatic Hernandez-Guzman, R. and Ruiz-Luna, A., 2013. SARA – an
systems: degradation thresholds, stormwater detection, and enhanced curve number-based tool for estimating direct
the limits of mitigation. Journal of the American Water runoff. Journal of Hydroinformatics, 15 (3), 881–887.
Resources Association, 33 (5), 1077–1090. doi:10.1111/ doi:10.2166/hydro.2013.145
jawr.1997.33.issue-5 Hjelmfelt, A.T., 1980. Empirical investigation of curve num-
Boyd, M.J., Bufill, M.C., and Knee, R.M., 1993. Pervious and ber technique. Journal of the Hydraulics Division, 106 (9),
impervious runoff in urban catchments. Hydrological 1471–1476.
Sciences Journal, 38 (6), 463–478. doi:10.1080/02626669 Jones, T., Johnston, C., and Kipkie, C., 2003. Using annual
309492699 hydrographs to determine effective impervious area.
Boyd, M.J., Bufill, M.C., and Knee, R.M., 1994. Predicting Journal of Water Management Modeling, R215–14,
pervious and impervious storm runoff from urban drai- 291–306.
nage basins. Hydrological Sciences Journal, 39 (4), 321–332. Kilberg, D., Martin, M., and Bauer, M., 2011. Digital classi-
doi:10.1080/02626669409492753 fication and mapping of urban tree cover: city of St. Paul.
Carle, M.V., Halpin, P.N., and Stow, C.A., 2005. Patterns of Minneapolis-St. Paul, MN: Department of Forest
watershed urbanization and impacts on water quality. Resources, University of Minnesota, Final Report.
Journal of the American Water Resources Association, 41 Lee, J.G. and Heaney, J.P., 2003. Estimation of urban
(3), 693–708. doi:10.1111/jawr.2005.41.issue-3 imperviousness and its impacts on storm water systems.
D’Asaro, F., Grillone, G., and Hawkins, R.H., 2014. Curve Journal of Water Resources Planning and Management,
number: empirical evaluation and comparison with curve 129 (5), 419–426. doi:10.1061/(ASCE)0733-9496(2003)
number handbook tables in sicily. Journal of Hydrologic 129:5(419)
Engineering, 19 (12), 04014035. doi:10.1061/(ASCE) Mueller, G.D. and Thompson, A.M., 2009. The ability of
HE.1943-5584.0000997 urban residential lawns to disconnect impervious area
Ebrahimian, A., 2015a. Towards sustainable urban stormwater from municipal sewer systems. Journal of the American
infrastructure: improving the estimation of effective imper- Water Resources Association, 45, 1116–1126. doi:10.1111/
vious area. World Environmental and Water Resources j.1752-1688.2009.00347.x
Congress, 2015, 2261–2268. doi:10.1061/9780784479162.223 Natural Resources Conservation Service (NRCS), 2004.
Ebrahimian, A., 2015b. Determination of effective impervious National engineering handbook, part 630 hydrology.
area in urban watersheds. Ph.D. Dissertation. Department Washington, DC: USDA.
of Civil, Environmental, and Geo-Engineering, University Natural Resources Conservation Service (NRCS), 2007.
of Minnesota, Twin Cities. National engineering handbook, part 630 hydrology,
Ebrahimian, A., et al., 2015. Urban stormwater construction “Hydrologic Soil Groups”. Washington, DC: USDA.
method selection using a hybrid multi-criteria approach. Pandit, A. and Regan, J., 1998. What is the impervious area
Automation in Construction, 58, 118–128. doi:10.1016/j. curve number. Advances in Modeling the Management of
autcon.2015.07.014 Stormwater Impacts, 6, 437–450.
526 A. EBRAHIMIAN ET AL.

Ponce, V.M. and Hawkins, R.H., 1996. Runoff curve number: mechanisms to management. Freshwater Biology, 49 (6),
has it reached maturity? Journal of Hydrologic Engineering, 835–851. doi:10.1111/fwb.2004.49.issue-6
1 (1), 11–19. doi:10.1061/(ASCE)1084-0699(1996)1:1(11) USDA (US Department of Agriculture), 1986. Urban
Roy, A.H. and Shuster, W.D., 2009. Assessing impervious hydrology for small watersheds. 2nd ed. Washington,
surface connectivity and applications for watershed man- DC: Natural Resource Conservation Service, Technical
agement. Journal of the American Water Resources Release 55.
Association, 45 (1), 198–209. doi:10.1111/j.1752-1688. Vrebos, D., et al., 2014. Water displacement by sewer infra-
2008.00271.x structure in the Grote Nete catchment, Belgium, and its
SCS (Soil Conservation Service), 1964. National engineering hydrological regime effects. Hydrology and Earth System
handbook, section 4 hydrology. Washington, DC: USDA. Sciences, 18 (3), 1119–1136. doi:10.5194/hess-18-1119-
SCS (Soil Conservation Service), 1972. National engineering 2014
handbook, section 4 hydrology. Washington, DC: USDA. Willett, J.B. and Singer, J.D., 1988. Another cautionary note
SCS (Soil Conservation Service), 1985. National engineering about R2: its use in weighted least-squares regression ana-
handbook, section 4 hydrology. Washington, DC: USDA. lysis. The American Statistician, 42 (3), 236–238.
Shuster, W.D., et al., 2005. Impacts of impervious surface on Yang, G., et al., 2011. The impact of urban development on
watershed hydrology: a review. Urban Water Journal, 2, hydrologic regime from catchment to basin scales.
263–275. doi:10.1080/15730620500386529 Landscape and Urban Planning, 103 (2), 237–247.
Sutherland, R.C., 1995. Methodology for estimating the effec- doi:10.1016/j.landurbplan.2011.08.003
tive impervious area of urban watersheds. Watershed Zhan, X. and Huang, M.-L., 2004. ArcCN-runoff: an ArcGIS
Protection Techniques, 2 (1), 282–284. tool for generating curve number and runoff maps.
Taylor, S.L., et al., 2004. Catchment urbanisation and Environmental Modelling & Software, 19, 875–879.
increased benthic algal biomass in streams: linking doi:10.1016/j.envsoft.2004.03.001

You might also like