You are on page 1of 13

Urban Water Journal

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/nurw20

Evaluation of the tabulated, NEH4, least squares


and asymptotic fitting methods for the CN
estimation of urban watersheds

Marcus Vinícius Galbetti, Antonio Carlos Zuffo, Tais Arriero Shinma, Vassiliki
Terezinha Galvão Boulomytis & Monzur Imteaz

To cite this article: Marcus Vinícius Galbetti, Antonio Carlos Zuffo, Tais Arriero Shinma, Vassiliki
Terezinha Galvão Boulomytis & Monzur Imteaz (2022) Evaluation of the tabulated, NEH4, least
squares and asymptotic fitting methods for the CN estimation of urban watersheds, Urban
Water Journal, 19:3, 244-255, DOI: 10.1080/1573062X.2021.1992639

To link to this article: https://doi.org/10.1080/1573062X.2021.1992639

Published online: 25 Oct 2021.

Submit your article to this journal

Article views: 122

View related articles

View Crossmark data

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


https://www.tandfonline.com/action/journalInformation?journalCode=nurw20
URBAN WATER JOURNAL
2022, VOL. 19, NO. 3, 244–255
https://doi.org/10.1080/1573062X.2021.1992639

RESEARCH ARTICLE

Evaluation of the tabulated, NEH4, least squares and asymptotic fitting methods for
the CN estimation of urban watersheds
a a b a,c
Marcus Vinícius Galbetti , Antonio Carlos Zuffo , Tais Arriero Shinma , Vassiliki Terezinha Galvão Boulomytis
and Monzur Imteaz d
a
School of Civil Engineering, Architecture and Urban Design, State University of Campinas, Campinas, Brazil; bSchool of Environmental Engineering,
State University of Mato Grosso Do Sul, Dourados, Brazil; cSchool of Civil Engineering, Federal Institute of Science, Education & Technology of São
Paulo, Caraguatatuba, Brazil; dDepartment of Civil and Construction Engineering, Swinburne University of Technology, Hawthorn, Australia

ABSTRACT ARTICLE HISTORY


The Soil Conservation Service Curve Number (SCS-CN) model is widely used to estimate the watershed Received 16 February 2021
runoff. However, the model may result in inaccurate estimations when assuming a tabulated CN value Accepted 7 October 2021
and an initial abstraction (λ) of 0.20. Therefore, in this paper, we evaluated the suitability of four CN KEYWORDS
estimation methods for an urban watershed: Tabulated CN, National Engineering Handbook 4, Least Curve number; initial
Squares and Asymptotic Fitting. Also, the effect of ordering the data and different λ values (0.20 or 0.05) abstraction ratio; watershed
were analyzed. Our findings suggest that the transient approach of Asymptotic Fitting method, a variance runoff
from the original Asymptotic Fitting method, provides the best runoff estimations using rank ordered
data and λ equal to 0.05 (R2 = 0.79, PBIAS = 0.01% and NSE = 0.78). It is to be noted that similar results
were also obtained through a simple linear model.

Introduction
In developing countries, most of the urbanization process has estimating the runoff from rainfall events, which assists infra­
occurred without an appropriate public urban planning, parti­ structure projects through proper design of urban drainage
cularly from the second half of the 20th century onwards (Brasil systems.
2007). It has caused a significant increase in floodplain occupa­ One of the main models used to estimate runoff is the Soil
tion and impermeable areas. As a result, intense rainfalls tend Conservation Service Curve Number (SCS-CN) model, which
to cause frequent floods in the urban areas of developing was developed in the mid-1950s by the former Soil
countries. Conservation Service (SCS) of USA, currently known as Natural
According to the Brazilian Institute of Geography and Resources Conservation Service (NRCS). It has been widely used
Statistics (IBGE 2010), floods are among the main types of by water professionals due to its convenience, simplicity, and
natural disasters that adversely affect Brazilian urban areas. institutional reliability (Ponce and Hawkins 1996). One of the
Among 5,570 Brazilian municipalities, 2,274 (41%) reported main advantages of this method is the requirement of few
to have urban drainage problems, 1,381 (25%) reported the parameters such as Curve Number (CN) and initial abstraction
irregular occupation of floodplains, and 1,094 (20%) ratio (λ) for the runoff estimation.
reported the irregular occupation in protected areas (IBGE According to Hawkins et al. (2009), the CN parameter is
2010). Between 2010 and 2020, these problems affected based on variables such as LULC, Hydrologic Soil Group (HSG),
about 3.14 million people causing economic impacts such and Antecedent Runoff Condition (ARC). The LULC data pro­
as damage to urban infrastructure and private properties, vides information regarding the current landscape situation.
and were responsible for 1,791 deaths (EM-DAT 2020). The The HSG is related to the soil classification scheme from the
occurrences of frequent floods also caused social and envir­ lowest (type A) to the highest (type D) runoff potential (USDA
onmental impacts, including the increase of water-borne 2009). The combination of LULC and HSG are named
diseases and degradation of environmental quality Hydrologic Soil Cover Complex (USDA 2004b).
(Haddad and Teixeira 2015). The ARC attempts to represent the causes for the CN varia­
Therefore, the Land Use and Land Cover (LULC) planning bility, such as the rainfall intensity and duration, total rainfall,
integrated with water resources management are essential to soil moisture conditions, vegetation cover, vegetation growth
ascertain a sustainable development and avoid environmental, stage, and temperature (USDA 2004b). Thus, the ARC spatial
economic, and social vulnerability of the region (Boulomytis variation is according to the local features (Hawkins et al. 2009).
et al. 2016). For this purpose, hydrological models are efficient The ARC is divided into three classes: I for dry condition, II for
tools that play important roles in providing information for average condition and III for wetter condition. The ARC I and III
decision-making (Abbaspour et al. 2015). Among the hydrolo­ represent the runoff distribution limits for a given CN, and the
gical models, the rainfall-runoff models are widely used for ARC II is the central trend (Sartori, Hawkins, and Genovez 2011).

CONTACT Marcus Vinícius Galbetti marcusgalbetti@gmail.com


© 2021 Informa UK Limited, trading as Taylor & Francis Group
URBAN WATER JOURNAL 245

Therefore, the reliable performance of the SCS-CN model Brief historical background
depends on the appropriate estimation of the CN values.
The SCS-CN model was the final product of a major field
Several works show that it can be achieved through field
investigation in small agricultural watersheds in the first half
experiments (Elhakeem and Papanicolaou 2009; Edwards
of the 20th century in the USA (USDA 2004a), which resulted in
2017; Liu, Chen, and Feng 2018) or using proper empirical
a conceptual model of hydrologic abstraction of storm rainfall,
equation (El-Hames 2012; Kadam et al. 2012). The most com­
supported by empirical data. This characteristic enabled the
mon method for the CN estimation uses the suggested values
generalization of the SCS-CN model use for areas without
from the NRCS tables, which are based on the data achieved
stream gauges. According to Ponce and Hawkins (1996), the
from rural watersheds of the American Midwest. For the hydro­
SCS-CN model has the following features:
logic simulations, the CN is usually adopted for the ARC II
condition, and the ratio (λ) between the initial abstraction (Ia)
(1) The runoff depth Q is bounded in the range of 0 ≤ Q ≤ P,
and the potential retention (S) of a watershed is normally
assuring its stability;
assumed as 0.20 (USDA 2004a).
(2) As rainfall depth grows unbounded (P → ∞), the actual
Because of the original location where these parameters were
retention (P-Q) asymptotically was a constant value S,
developed, the application of tabulated CN values for Brazilian
the potential maximum retention;
conditions tends to overestimate the size of hydraulic infrastruc­
(3) The maximum potential retention is related to the CN
tures, which are neither consistent with the watershed scale nor
value; and,
the local flood observations (Ponce and Hawkins 1996; Cunha
(4) The CN estimates are based on the LULC, the HSG and
et al. 2015; Lal et al. 2017). Villanueva and Piccilli (2007) esti­
AMC.
mated that for every R$ 1.00 (Brazilian currency, one real)
invested in the urban macro-drainage network, the CN estima­
Many authors have addressed the historical developments of
tion error of 10% caused extra costs that varied between R$ 0.30
to R$ 1.10, depending on the local drainage conditions, making the SCS-CN model (Singh et al. 2010; Mishra, Singh, and Singh
the method as unreliable. In the literature, several authors stated 2018). However, some particularities have still been
that when λ was used as 0.20, significant uncertainties were approached in the literature, despite the widespread use of
observed in their model outcomes (D’Asaro, Grillone, and the model. According to Van Mullem et al. (2002), there is no
Hawkins 2014; Yuan et al. 2014; Ajmal and Kim 2015; Lal et al. consensus regarding how the infiltration rate interferes in the
2017; Valle Junior, Rodrigues, and Oliveira 2019). Hence, the CN loss. As a matter of fact, the SCS-CN model has been mis­
current recommended value for λ is 0.05 (Hawkins et al. 2009). takenly considered as an infiltration model same as the Horton
Due to the watershed physical characteristics, the appli­ and Green-Ampt methods (Van Mullem et al. 2002).
cation of the CN estimation methods and evaluation of λ The SCS-CN model was reviewed in 1990 by the Natural
value from observed rainfall-runoff (P-Q) data makes the Resources Conservation Service (NRCS) and the Agricultural
model outcomes more reliable. To perform this task, it is Research Service (ARS), from the U.S. Department of
necessary to: (i) achieve P-Q data by hydrological assessment Agriculture (Van Mullem et al. 2002). Some of the significant
(Angelini Sobrinha et al. 2014); (ii) treat this data by setting updates consisted on the Antecedent Moisture Condition
a minimum precipitation value to avoid bias (D’Asaro, (AMC) terminology, which changed to Antecedent Runoff
Grillone, and Hawkins 2014; Lal et al. 2017); and (iii) analyze Condition (ARC) as the variation of the CN is not entirely due
the effect of observed dataset (natural data) vs. dataset to prior rainfall events. The ARC varies from place to place, and
realigned by the return period (rank ordered data) is a function of local climate, soil, vegetation, and land use
(Hawkins 1993; Hawkins et al. 2009). Among the methods, (Hawkins et al. 2009).
the National Engineering Handbook 4 (NEH4), the Least However, as pointed out by Mishra, Singh, and Singh (2018),
Squares (LS) and the Asymptotic Fitting (AF) methods are any change in the AMC (or ARC, considering the new terminol­
noticeable and highly recommended by the ‘Curve Number: ogy) causes a sudden jump in the CN value, and, thus a sudden
State of Practice’ (Hawkins et al. 2009), which presented jump in the computed runoff, which does not match with
a critical analysis of the SCS-CN model for its dissemination a general actual gradual behavior of a watershed. In order to
and improvement. Despite efforts, these studies are still adjust this issue, some authors proposed models based on the
incipient in Brazil, (Andrade et al. 2017; Cunha et al. 2015; concept that the CN should vary continuously (Neitsch et al.
Oliveira et al. 2016; Sartori, Hawkins, and Genovez 2011; 2009; Verma et al. 2020,; Williams and LaSeur 1976), instead of
Valle Junior, Rodrigues, and Oliveira 2019), especially for adopting a discrete CN value for a particular ARC condition
urban and peri-urban areas. (Mishra, Singh, and Singh 2018).
With the aim of achieving a proper estimation of the CN value The use of the SCS-CN model has been extrapolated to
under the ARC II condition, and for the evaluation of the effects larger and more heterogeneous areas than those used for its
of λ on the urban runoff, this study investigated four different implementation. Some authors evaluated the effect of hetero­
methods. The aspects considered in the study were: (1) the geneous LULC in the runoff generation process. Grove,
effectiveness of the tabulated compared to estimated CN values, Harbor, and Engel (1998) studied the feasibility of the distrib­
based on observed P-Q events (NEH4, LS and AF methods); (2) uted CN approach (i.e. the weighted runoff from areas with
the application of natural or rank-ordered data; (3) the value of different CNs) compared to the single composite approach
0.20 versus 0.05 for the initial abstraction ratio. Finally, the (i.e. the area-weighted average of the CN values). They
proposed performance measurement indexes were discussed. obtained better results with the use of the former approach,
246 M. V. GALBETTI ET AL.

due to the non-linear form of the SCS-CN model (Soulis and


Valiantzas 2012). Soulis and Valiantzas (2012) proposed
a modified AF method (Hawkins 1993) to estimate CN by
incorporating the concept of an equivalent two-CN heteroge­
neous system, which depends on a threshold value of rainfall.
Soulis and Valiantzas (2013) expanded this concept for multi­
ple-CN heterogeneous system, which can be achieved using
GIS and non-linear optimization techniques. Thus, these meth­
ods take into consideration the soil-cover spatial variations in
the estimation of CN values from measured P-Q data, improv­
ing the SCS-CN runoff estimates. Santikari and Murdoch (2018)
analyzed the effects of Ia heterogeneity throughout the
watershed on the runoff estimates. They observed an
improvement in the performance of the SCS-CN model when
considering the spatial distribution of Ia.
Some authors evaluated the effect of the watershed slope in
the CN estimates (Ajmal et al. 2016; Huang et al. 2006,; Sharpley
and Williams 1990). They proposed adjustment equations, once
the CN values in NRCS tables assumed a slope of 5%. However,
these proposed equations need to be validated for other
regions that have similar climatic and slope conditions
(Mishra, Singh, and Singh 2018).

The SCS-CN model


For the application of the SCS-CN model it is important to
understand two fundamental concepts (Mishra, Jain, and
Singh 2004; USDA 2004b). First, the model considers the
water balance (Equation 1), through which it derives the
amount of direct runoff (Q). Then it correlates the ratio of the
Figure 1. Methodological approach of the study.
actual direct runoff (Q) to the total rainfall (P) subtracted by the
initial abstraction (Ia) and the ratio of the cumulative infiltration
(F) to the potential retention (S), as shown in Equation (2):
Materials and methods
Q¼P Ia F (1)
The methodological approach that we carried out in this study
is shown in Figure 1. This section details the methods used in
Q F each step, from the data collection up to the analysis of the
¼ (2) performance indexes.
P Ia S

Where, all parameters are measured in mm. Additionally, Ia is Study area


correlated to the maximum potential retention, as shown in
Equation (3): The study was conducted for Mineirinho watershed, in the
municipality of São Carlos, State of São Paulo, Brazil. It is
Ia ¼ λ � S (3) located between latitudes 21°59ʹ00” S and 22°01ʹ15” S and
longitudes 47°54ʹ30” W and 47°56ʹ30” W (Figure 2). The drai­
Where, λ is the dimensionless initial abstraction coefficient. nage area is 5.85 km2 and the elevation varies between 780 m
Combining Equations (1), (2) and (3), yield: and 872 m. The main river has a total length of 4,320 m and
equivalent slope of 2.10%. According to the Köppen-Geiger
2
ðP λ�SÞ classification (Alvares et al. 2013), the studied watershed is
f Q ¼ Pþð1 λÞ�S ; for ðP > λ � SÞ; (4) humid and subtropical (Cwa), with dry winter (April to
Q ¼ 0; otherwise
September) and rainy summer (October to March). The aver­
The S values are derived by Equation (5) for the different age annual precipitation is 1,521 mm, and the average tem­
P-Q dataset, assuming that they have the same return period: perature is 22°C (Tolentino 2007).
According to the taxonomy of the World Reference Base for
25; 400 Soil Resources (IUSS Working Group WRB 2015), the watershed
S¼ 254 (5) soil is Ferralsol (Rossi 2017), with no spatial variability. This soil
CN
is in the HSG classified as B (Genovez, Neto, and Sartori 2005),
Where, the CN values adopt the ARC II condition. where the runoff potential is moderately low when completely
URBAN WATER JOURNAL 247

Figure 2. Location of the study area.

wet and the infiltration is not constrained (USDA 2009). was not necessarily matched with the one caused by the original
According to Paulino (2014), the predominant LULC classes rainfall (P) (Hawkins 1993; Lal et al. 2017). This procedure aimed to
are: Urban areas (36.7%), Forest (22.7%), Bare Soil (18.2%), reduce the dispersion among the observed data.
Agriculture (14.1%) and Pasture (8.3%). The spatial distribution
of the LULC classes is displayed in Figure 3.
CN estimation methods
We evaluated the Tabulated CN, National Engineering
Collection and treatment of rainfall-runoff data Handbook 4 (NEH4), Least Squares (LS) and Asymptotic Fitting
The P-Q data events were measured between the years 2013 (AF) methods for the CN estimation. In Brazil, these methods
and 2015. Precipitation data were measured by 4 tipping are commonly used to estimate CN and subsequent watershed
bucket rain gauges (model TB4) and the discharge was mea­ runoff (Cunha et al. 2015; Oliveira et al. 2016; Sartori, Hawkins,
sured by an acoustic flow meter (model SonTek- IQ Plus) and Genovez 2011; Valle Junior, Rodrigues, and Oliveira 2019).
installed at the watershed outfall (Figure 2). The temporal The tests performed aimed to estimate the CN values under
discretization of precipitation and discharge corresponded the ARC II condition, considering both the natural and the rank-
to two minutes. Subsequently, the average rainfall in the ordered data, and assuming λ equal to 0.20 or 0.05. The proce­
basin was derived by the Thiessen polygon method. We dures are described in the following sections.
used rainfall events higher than 15 mm uniformly distributed
in the watershed. This criterion was chosen to avoid a biasing Tabulated CN method
effect, and to retain a sufficient number of P-Q data for ana­ This method consists of using tabulated values of CN according
lysis (Lal et al. 2017). to the Hydrologic Soil Cover Complex. For this study, we
In the present study, we used both the natural and rank- obtained the recommended CN values from the NRCS tables
ordered data for analysis. The natural P-Q data consist of the (USDA 2004b), which are presented in Table 1.
actual observed dataset. In rank-ordered data series, the observed Due to the lack of data for the Mineirinho watershed, we
P and Q values were first sorted separately and then realigned used the total rainfall for the period of 5 days (P5d) prior to the
using a common rank-order basis to form a new set of P-Q pairs. P-Q event as a criterion to estimate the ARC, which is a common
These new pairs had the same return period, and the runoff (Q) procedure in Brazil. When the ARC is classified as average, the
248 M. V. GALBETTI ET AL.

Table 2. CN value according to the ARC based on the P5d.


ARC Classification P5d CN Conversion
I dry P5d < 15mm CN
CNI ¼ 2;281 0;01384�CN
II average 15 � P5d � 40mm CN
III wet P5d > 40mm CN
CNIII ¼ 0;427þ0;00573�CN
Source: Adapted from Hawkins et al. (2009).

approaches. According to Grove, Harbor, and Engel (1998),


with the composite approach, an area-weighted average CN
is calculated for the entire area. The total runoff for each event
is obtained by Equations (4) and (5). With the distributed
approach, there is no CN averaging. Instead, separated CN
values are determined for each subarea, and separated runoff
values are calculated. Then, these runoff values are averaged to
find the total runoff depth for the watershed.

National engineering handbook 4 (NEH4) method


The National Engineering Handbook 4 (NEH4) method calcu­
lates the CN values from observed P-Q events that caused
annual floods or regarding which their magnitudes are sub­
stantial (USDA 2004a). Initially, the S value for each event was
determined using Equation (7) (Hawkins et al. 2009), which is
the valid quadratic solution of Equation (4):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 � λ � P þ Q � ð1 λÞ ðQ � ð1 λÞÞ2 þ 4 � λ � Q � P
S¼ (7)
2 � λ2

Where, S is the potential retention of the watershed (mm) for


a given P-Q event.
Figure 3. LULC of the Mineirinho watershed. Source: .Paulino (2014) Subsequently, the CN value for each event was derived
using Equation (5) and the final CN value in ARC II was deter­
Table 1. CN values for the hydrologic soil cover complex of the mineirinho mined by the median of all values. Finally, runoff was calculated
watershed. for each event using Equation 4. The procedure was performed
Class CN Description using λ equal to 0.20 and 0.05 in Equation 7, and for natural and
Urban 85 Residential districts presenting a lot size of 1/8 acre or less rank-ordered data.
(town houses)
Forest 55 Woods in good conditions
Bare Soil 86 Fallow – Bare Soil Least squares (LS) method
Agriculture 70 Row Crops Contoured & Terraced + Crop residue cover in
good conditions The Least Squares (LS) method comprises the calibration of the
Pasture 61 Pasture in good conditions CN value by the minimization of the sum of the quadratic errors
Source: Adapted from USDA (2004). of Q (SQLS) (Hawkins et al. 2009):

N
X �2
estimated CN value is used. In case the classification of the ARC SQLS ¼ min QobsðiÞ QsimðiÞ (8)
is dry (condition I) or wet (condition III), the CN value must be i¼1
converted according to the equations shown in Table 2.
To assess the sensitivity of CN regarding λ, we applied both Where, Qobs is the total observed runoff, Qsim is the total
the values using Equation (6) to convert the CN values (Hawkins simulated runoff (mm) and N is the number of events.
et al. 2009): Combining Equations (8), (4) and (5), the following Equation
(9) is obtained:
100
CN0:05 ¼ h �1:15 i (6) � 25;400
�� 2 ! 2
100 N
X
1:879 � CN 1 þ1 PðiÞ λ� CN 254
SQLS ¼ min QobsðiÞ � �� (9)
i¼1 PðiÞ þ ð1 λÞ � 25;400
CN 254
Where, the CN is estimated when λ is 0.20 and CN0.05 is the CN
value for λ equal to 0.05. Where, the CN value is set for all P-Q events (for λ equal to 0.20 and
After estimating the CN value as a function of ARC and λ, the 0.05) using Solver, which is an analysis tool in the Microsoft Excel
runoff was estimated by the composite (i.e. area-weighted spreadsheet that provides a simple means of fitting experimental
average CN) and the distributed (i.e. area-weighted runoff) data to solve nonlinear functions (Walsh and Diamond 1995).
URBAN WATER JOURNAL 249

Asymptotic fitting (AF) method Finally, the fitted CN∞ was used to calculate the
The Asymptotic Fitting (AF) method proposed by Hawkins runoff Q (Equations 4 and 5) for each event, considering λ
(1993), considers that the CN derived from rainfall runoff data equal to 0.20 and to 0.05, both for natural and rank-ordered
is related to the precipitation, presenting three distinct beha­ data.
viors: standard, complacent, and violent. The standard behavior Additionally, we also analyzed the use of the transient
is featured by the decay of the CN value alongside with the values, i.e. CN(P), to estimate CN for each event and calculate
increase in precipitation; but approaching a constant or near- the runoff Q, which we defined as transient approach of the AF
stable value called CN∞ at larger storms. This is assumed to be method. This is equivalent to considering a non-constant,
the characterizing watershed CN, which is applicable to larger P-dependent CN, which is a variation from the original AF
design storms (D’Asaro, Grillone, and Hawkins 2014), and can method (Hawkins et al. 2009).
be modeled by the Equation (10):

CNðPÞ ¼ CN1 þ ð100 CN1 Þ � e k1 �P


(10) Performance evaluation
For the performance evaluation of the CN estimation methods,
Where, CN(P) is the CN value according to the precipitation, we used three indices suggested by Moriasi et al. (2015): the
CN∞ is the CN value for the precipitation that tends to infinite coefficient of determination (R2) (Equation 13), the percentage
and k1 is the decay constant (mm−1). of bias (PBIAS) (Equation 14), and the Nash-Sutcliffe Efficiency
The complacent behavior is defined by the decay of CN (NSE) coefficient (Nash and Sutcliffe 1970) (Equation 15) pre­
concomitantly with the increase in precipitation without reach­ sented as follows:
ing a constant value. Despite the possibility of modeling the
0 12
complacent behavior by Equation (10), the true value of CN∞ is Pi¼1 � � ��
not achieved (Hawkins 1993). In this case, Hawkins et al. (2009) B N QobsðiÞ μ � QsimðiÞ μ C
R2 ¼ B ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiobs
@rffiP ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffisim
ffiffiffiffiq ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiC
ffi
suggest two alternatives: the application of the AF method
i¼1 �2 Pi¼1 �2 A
using the standard fitting, acknowledging the insecurity and N QobsðiÞ μobs � N QsimðiÞ μsim
inapplicability of extrapolation; or the application of a simple (13)
linear function Q = CP, where C is the runoff coefficient.
The violent behavior is related to the descent of CN due to
low precipitations, similarly to the complacent behavior, but PN �
i¼1 QobsðiÞ QsimðiÞ
with a sudden increase in both CN and runoff from a given PBIAS ¼ PN (14)
P value (Hawkins 1993; Hawkins et al. 2009). This behavior can i¼1 QobsðiÞ
be modeled by the Equation (11):
� � "PN �2 #
CNðPÞ ¼ CN1 � 1 e k2 �ðP Ps Þ (11) i¼1 QobsðiÞ QsimðiÞ
NSE ¼ 1 PN �2 (15)
−1 i¼1 QobsðiÞ μobs
Where, k2 is the increase constant (mm ) and Ps is the thresh­
old rainfall value (mm).
Where, µobs and µsim are the means of observed and simulated
According to Hawkins (1993), the correct determination of
runoff (mm), respectively.
the CN behavior is found by plotting the rainfall versus the CN
The R2 describes the degree of collinearity between the
value for each event, calculated by the procedure described in
observed and predicted data, while the PBIAS verifies
Section ‘National Engineering Handbook 4 (NEH4) Method’.
whether the model results are underestimated (positive
Additionally, the rainfall versus minimum CN (CN0) values for
values) or overestimated (negative values). The NSE deter­
runoff occurrences, is recommended to be plotted on the same
mines the relative magnitude of the residual variance com­
graph, based on Equation (12):
pared to the observed data variance (Nash and Sutcliffe
2:540 1970). Values for NSE between 0 and 1 are generally
CN0 ¼ (12) viewed as acceptable levels of performance, whereas
ð25; 4 þ P=2Þ
values less than 0 indicate that the mean observed value
Subsequently, a visual analysis of the scatterplots must be is a better predictor than the simulated value, which indi­
performed to identify the data basin behavior and to adjust cates unacceptable performance (Moriasi et al. 2015). The
the parameters of the corresponding behavior equation (i.e. performance indexes were calculated using the natural
Equations 10 or 11). The fitting was performed by minimizing data set even when the rank-ordered data were used for
the sum of the quadratic errors of CN using the Solver tool of model fitting. Classifications of the model performance are
the Microsoft Excel spreadsheet. presented in Table 3.

Table 3. Classification according to performance indexes.


Index Unsatisfactory Satisfactory Good Very Good
R2 R2 ≤ 0.50 0.50 < R2 < 0.70 0.70 ≤ R2 ≤ 0.80 R2 > 0.80
PBIAS PBIAS > ±25% ±10% ≤ PBIAS ≤ ±25% ±5% < PBIAS < ±10% PBIAS ≤ ± 5%
NSE NSE ≤ 0.50 0.50 < NSE < 0.60 0.60 ≤ NSE ≤ 0.80 NSE > 0.80
Source: Moriasi et al. (2015).
250 M. V. GALBETTI ET AL.

Results and discussion P-Q data for analysis, and the CN estimation was as coherent
as it would if we had a long annual series (Ponce 2014).
Observed P-Q events
Therefore, we could reinforce the importance of continuous
We analyzed 42 P-Q events, comprising total rainfall values hydrological assessments in Brazilian urban watersheds. This
between 15.3 mm and 72.0 mm, and total runoff between is extremely relevant for São Carlos, State of São Paulo, where
1.8 mm and 12.7 mm (Figure 4). We observed a small dispersion there were 53 floods between 1980 and 2009 (Lima and
between the P-Q data for rainfall depths lower than 25.0 mm, Amorim 2015).
which might be explained by the analysis of Mineirinho basin
LULC (Figure 3). For the rainfall depths that are lower than this
threshold, the part of the urban watershed (located on the right Analysis of CN estimation methods
side of the main river) is probably the most significant area to The CN estimated values by each method and their respective
generate runoff. For rainfall depths higher than 25.0 mm, the performance are presented in Table 4. There was a high dis­
entire watershed contributes to runoff generation, leading to crepancy between the CN values estimated by the proposed
a higher dispersion. methods. This result strengthens the need to use several meth­
It was not possible to achieve a long annual flood series for ods to estimate the CN values and then choose the best
the study watershed, which is a common issue for Brazilian method according to the study area, since there is no consen­
hydrologists (Andrade et al. 2017; Oliveira et al. 2016; Sartori, sus about which method is more appropriate for the CN esti­
Hawkins, and Genovez 2011; Valle Junior, Rodrigues, and mation (Tedela et al. 2012).
Oliveira 2019). To overcome this issue, we selected all events The transient approach of the AF method presented results
that suited to the proposed criteria (Section ‘Collection and classified as at least ‘good’ in all indexes evaluated for both λ
treatment of rainfall-runoff data’). Thus, we had more values and for the natural and rank-ordered data. The other

Figure 4. Observed P-Q events in Mineirinho basin.

Table 4. Estimated values of CN and the respective performance.


Natural Data Rank-Ordered Data
Method λ CN R2 PBIAS NSE CN R2 PBIAS NSE
Tabulated CN 0.20 74.3 0.54 28% −3.99 - - - -
Composite 0.05 64.3 0.62 17% −2.68 - - - -
Tabulated CN 0.20 - 0.63 1% −3.96 - - - -
Distributed 0.05 - 0.67 −10% −3.17 - - - -
NEH4 0.20 86.9 0.78 −67% −7.48 86.1 0.77 −56% −6.35
0.05 77.9 0.77 −45% −3.48 77.7 0.77 −44% −3.40
LS 0.20 74.5 0.69 49% −0.97 75.0 0.70 46% −0.97
0.05 63.0 0.74 33% −0.20 63.7 0.74 31% −0.20
AF 0.20 36.41 - 100% −3.63 46.11 0.23 100% −3.58
0.05 34.91 0.58 92% −2.70 42.01 0.66 84% −1.96
Transient 0.20 66.3–90.63 0.80 1.24% 0.80 67.6–90.43 0.80 −0.13% 0.79
AF2 0.05 52.2–84.03 0.80 1.74% 0.80 54.4–83.83 0.79 0.01% 0.78
1
CN∞; 2AF method using transient values of CN; 3range of CN(P) values.
URBAN WATER JOURNAL 251

Figure 5. Relationship between the observed and the simulated runoff values, considering: (a) natural data and λ equal to 0.20; (b) natural data and λ equal to 0.05; (c)
rank-ordered data and λ equal to 0.20; and (d) rank-ordered data and λ equal to 0.05. TCNC, TCND, NEH4, LS, AF and TAF refer to Tabulated CN Composite, Tabulated CN
Distributed, National Engineering Handbook 4, Least Squares, Asymptotic Fitting and Transient Asymptotic Fitting methods, respectively.

methods presented at least one of the indexes classified as The CN values estimated by the NEH4 method were slightly
‘unsatisfactory’. The relationship between the precipitation higher than the ones estimated by the composite approach of
and runoff values (Figure 5) showed that this method is fea­ the Tabulated CN method. Similar results were found at some
tured by presenting simulated runoffs close to the observed other Brazilian watersheds by Oliveira et al. (2016) and Valle
ones, whilst the others presented large discrepancy. Junior, Rodrigues, and Oliveira (2019). Since the NEH4 method
The estimated CN values for the composite approach of is based on the median value of the calculated CN values of
the Tabulated method were 74.3 and 64.3 for λ equal to each event, there is a tendency to prioritize the adjustment of
0.20 and 0.05, respectively, whereas in the distributed the CN for the most frequent events (Figure 5).
approach, the CN values presented in Table 1 were used. The objective function used for the CN calibration
Regardless of the approach applied, this method had the (Equation 9) in the LS method prioritizes the adjustment of
tendency to underestimate the runoffs generated by low great magnitude events (Shinma 2015). As a result, LS method
precipitation values and overestimate the runoffs from estimated lower CN values than the ones estimated by the
high precipitations. This behavior was also found by NEH4 method (Table 4), providing a closer to observed values
Cunha et al. (2015), Oliveira et al. (2016) and Valle Junior, runoff estimates for a few large-scale events. However, several
Rodrigues, and Oliveira (2019) in other regions of Brazil. As minor magnitude events were systematically underestimated,
a consequence, the Tabulated CN method tends to over­ leading to positive PBIAS values. Figure 5 presents similar
estimate the design flood and, therefore leads to oversizing curves for the LS and the composite approach of the
the hydraulic infrastructure (Cunha et al. 2015) and reduces Tabulated CN methods, once the calculated CN values in both
their economic viability (Villanueva and Piccilli 2007). cases were also similar.
Comparing the Tabulated CN method approaches, we Regarding the AF method, the visual inspection of the CN
found that the distributed approach systematically estimated values as a function of precipitation (Figure 6) indicates that the
lower runoff values than the composite approach. This beha­ study area showed complacent behavior in every evaluated
vior could be explained by the lower CN values adopted for combination, with CN∞ values varying between 34.9 and 46.1.
some areas in the distributed approach. These areas pro­ The complacent behavior was also observed in other Brazilian
duced no runoff due to the threshold precipitation value basins (Sartori, Hawkins, and Genovez 2011; Oliveira et al. 2016;
adopted in the SCS-CN model (Equations 4 and 5). Valle Junior, Rodrigues, and Oliveira 2019), which are featured
Nevertheless, none of the Tabulated CN approaches accu­ by the action of subsurface mechanisms in the generation of
rately represented the runoff generation process in the runoff due to their high infiltration and retention potentials of
study area. soils. This feature is also observed in the present study basin,
252 M. V. GALBETTI ET AL.

Figure 6. Curves of the Asymptotic Fitting Method for: (a) natural data and λ equal to 0.20; (b) natural data and λ equal to 0.05; (c) rank-ordered data and λ equal to
0.20; and (d) rank-ordered data and λ equal to 0.05.

since 63.3% of its area consists of permeable LULC on Ferrasol, recommend the use of the obtained parameters for broader
soil with a high potential for infiltration and retention (Bono watersheds, sub-watersheds or other similar nearby water­
et al. 2012; Souza et al. 2018). sheds, once they are not related to any specific character­
However, a higher number of P-Q events is necessary for the istic of the watershed. Therefore, the transient approach of
accurate definition of the Mineirinho basin behavior, prefer­ the AF method might be used to estimate CN for the
ably, including higher magnitude events. According to Mineirinho basin, based on the parameters obtained in
Hawkins et al. (2009), without any evidence that the declining any combination of λ and data set presented in Table 4,
CN with P will achieve a stable value, this condition might be when the rainfall depth is between 15.3 and 72.0 mm.
considered complacent, potentially violent or merely an incom­ For comparison purposes, we also evaluated the linear
pletely developed standard response. Therefore, the impor­ model (Q = CP). The C values obtained were equal to 0.175
tance of continuous hydrological monitoring in Brazilian (Figure 4) and 0.179, for natural and rank-ordered data,
urban watersheds is strengthened. respectively. This method produced runoff estimations clas­
Using the approach proposed by Hawkins (1993), the AF sified as ‘very good’ for all the indexes evaluated (R2 and
method did not present adequate results, regardless of the λ NSE were 0.95 and 0.80 for both data sets, and PBIAS was
value or the data set used. For instance, the method estimated 0.80% and −1.00% for natural and rank-ordered data sets,
no runoff for natural dataset and λ equal to 0.20. Consequently, respectively) and presented the same limitations of the
we were not able to calculate R2, and PBIAS was equal to 100% transient approach of the AF method. Thus, considering
in this combination. The extremely low and null runoff values the simplicity of this method, we suggest that runoff com­
are related to the adoption of low CN values (i.e. CN∞) and the puting by the linear model is more suitable for this study
observation of low rainfall events. area.
The transient approach of the AF method presented Among the evaluated CN estimation methods, only the
results classified as at least ‘good’ (Table 4) in every perfor­ transient approach of AF method was able to estimate CN
mance index, regardless of the λ value or the type of data and produce acceptable predicted runoffs in the SCS-CN
used. However, the use of CN(P) to estimate runoff is model for the Mineirinho watershed, however presenting sev­
a different approach than the one proposed by Hawkins eral limitations. This result does not invalidate the use of SCS-
(1993) and involves an additional parameter (k1). This CN model, rather indicates the need for further studies to
approach might be used acknowledging the impossibility evaluate more robust methods for CN estimation (Singh et al.
of extrapolation beyond the P values used in the construc­ 2013; Soulis and Valiantzas 2012, 2013,; Verma et al. 2020) than
tion of the curve, which is very limiting. Also, we do not the ones commonly applied in Brazil.
URBAN WATER JOURNAL 253

Evaluation of data ordering and λ value For the NSE coefficient, we observed that every method
presented ‘unsatisfactory’ performance, except for the transient
We evaluated the effect of ordering the data and observed
approach of the AF method. This approach almost achieved the
a minor improvement of the majority of the performance
‘very good’ performance threshold and presented consistent
indexes by the use of rank-ordered data (Table 4). Thus, this
numerical and graphical results (Table 4 and Figure 5).
procedure was more suitable for the CN estimation. As the
We verified that the use of an isolated performance index
Tabulated CN method did not use observed data, the natural
could be misleading. Thus, a combination of statistical and
and rank-ordered data performance indexes were identical.
graphical performance methods is recommended to evaluate
In general, the SCS-CN model presented better results when
the SCS-CN model performance, such as the findings of Gupta
the adopted λ value was 0.05, corroborating with recent tech­
et al. (2020) and Verma et al. (2020).
nical recommendations and other literature results performed
in urban areas (Singh et al. 2013). The only significant exception
was the worsening of the PBIAS index by 11% in the distributed Conclusion
approach of the Tabulated CN method. However, the isolated
In the present study, we evaluated four CN estimation methods
analysis of the PBIAS index value or of any other performance
for the Mineirinho basin using 42 observed rainfall-runoff events:
index may not express the efficiency of the CN estimations
Tabulated method (composite and distributed approaches),
properly.
NEH4, LS and AF (original and transient approaches). The calcu­
lated CN values presented considerable discrepancies among
the methods, and, in general, underestimated the runoffs gen­
Analysis of performance indexes
erated by lower precipitation values and overestimated the ones
We observed that most of the R2 values met at least the generated by higher precipitation values. The only exception
‘satisfactory’ class (Table 4), a result that contradicts the ones was the transient approach of the AF method, which estimated
presented by other performance indexes (PBIAS and NSE) and CN values that produced results classified as ‘Good’ or ‘Very
by the visual analysis of Figure 5. The R2 (Equation 13) expresses Good’ in all of the performance indexes.
the degree of linear agreement between the observed and Furthermore, the influence of the data ordering and the
predicted values, but not necessarily their closeness. Krause, adopted λ value were evaluated. It was found that results
Boyle, and Bäse (2005) recommend complementing the inter­ improved with the use of the rank-ordered data and a λ value
pretation of R2 results with the analysis of the intercept (a) and of 0.05, regardless of the adopted CN estimation method.
gradient (b) values from the linear model (y = a + bx) between Also, we pointed out that the isolated analysis of one per­
the observed and simulated data. The values of (a) and (b) are formance index can lead to erroneous conclusions about the
respectively distant from 0 and 1 for every method, except for model efficiency. Hence, we highlight the need of a joint
the transient approach of the AF method, especially regarding numerical-and-graphical assessments of the results and the
λ equal to 0.05 (Table 5). In other words, even though the R2 use of diverse performance indexes.
values for these methods were classified as ‘satisfactory’, the The Mineirinho basin presented a complacent behavior
simulated runoffs were not close to the observed ones. based on the analysis of the observed data. We found that
PBIAS may rate the model performance improperly if the the transient approach of the AF method might be used for
model overpredicts as much as it underpredicts (Moriasi et al. the CN estimation and runoff computation. Taking its limita­
2015). This behavior was observed in the distributed approach tions into consideration, the recommended equation
of the Tabulated CN method, especially using λ equal to 0.20. (Equation 10) used the adjusted parameters for the rank-
Despite the PBIAS value of 1%, Figure 5 shows that the perfor­ ordered data and a λ value of 0.05 (i.e. CN∞ equal to 42.0 and
mance of this approach was poor. Nevertheless, we found that k1 equal to 0.0214). The performance indexes obtained were
the transient approach of the AF method resulted in the best 0.79 for R2, 0.01% for PBIAS and 0.78 for NSE.
values of PBIAS. For comparison purposes, we also evaluated the linear
model, which achieved ‘very good’ performance considering
all the evaluated indexes. However, we emphasize that the
Table 5. Intercept (a) and gradient (b) values of the linear model between linear model exhibited the same limitations as the transient
simulated and observed data. approach of the AF method.
Natural Data Rank-ordered Data Finally, we reinforce the need for continuous hydrological
Method λ a b a b monitoring, particularly in urban watersheds, and the applica­
Tabulated CN 0,20 −6,62 2,04 - - tion of more robust methods to improve CN estimation. Since
Composite 0,05 −5,97 2,02 - - the use of SCS-CN model is simple and practical, the impor­
Tabulated CN 0,20 −6,77 2,34 - -
tance of having a proper estimation method is significant for
Distributed 0,05 −5,86 2,27 - -
NEH4 0,20 −6,85 3,04 −6,84 2,92 water resources professionals.
0,05 −4,84 2,42 −4,83 2,40
LSM 0,20 −4,59 1,43 −4,75 1,49
0,05 −3,54 1,38 −3,61 1,42 Acknowledgements
AF 0,20 - - −0,06 0,01
0,05 −0,88 0,25 −1,53 0,47 We gratefully acknowledge the Brazilian Innovation Agency (FINEP) for the
Transient 0,20 0,99 0,79 0,65 0,87 support under the grant “Transversal Action - Environmental Sanitation and
AF 0,05 0,90 0,80 0,58 0,88 Housing - 7/2009”, and the Ministry of Science & Technology for the support
254 M. V. GALBETTI ET AL.

under the grant CT-HYDRO 01/2010. We also express our gratitude to Prof. D’Asaro, F., G. Grillone, and R. H. Hawkins. 2014. “Curve Number: Empirical
Dr. Victor M. Ponce, from San Diego State University, for the technical Evaluation and Comparison with Curve Number Handbook Tables in
discussions regarding the content of this research, and for revising the paper. Sicily.” Journal of Hydrologic Engineering 19 (12): 04014035. doi:10.1061/
(ASCE)HE.1943-5584.0000997.
Edwards, D. R. 2017. “Long-Term Spatio-Temporal Variation in Runoff Curve
Number under Consistent Cover Conditions: A Southeastern US Case
Disclosure statement
Study.” Water Resources Management 31 (11): 3491–3505. doi:10.1007/
No potential conflict of interest was reported by the author(s). s11269-017-1680-z.
Elhakeem, M., and A. N. Papanicolaou. 2009. “Estimation of the Runoff Curve
Number via Direct Rainfall Simulator Measurements in the State of Iowa,
USA.” Water Resources Management 23 (12): 2455–2473. doi:10.1007/
Funding s11269-008-9390-1.
El-Hames, A. S. 2012. “An Empirical Method for Peak Discharge Prediction in
This work was supported by the Financiadora de Estudos e Projetos
Ungauged Arid and Semi-Arid Region Catchments Based on
[Transversal Action - 7/2009]; Ministério da Ciência [CT-HYDRO 01/2010].
Morphological Parameters and SCS Curve Number.” Journal of
Hydrology 456–457: 94–100. doi:10.1016/j.jhydrol.2012.06.016.
EM-DAT. 2020. The Emergency Events Database. Brussels, Belgium:
ORCID Université Catholique de Louvain (UCL). https://www.emdat.be
Genovez, A., F. Neto, and A. Sartori. 2005. “Classificação Hidrológica De
Marcus Vinícius Galbetti http://orcid.org/0000-0002-1606-0273 Solos Brasileiros Para a Estimativa Da Chuva Excedente Com O Método
Antonio Carlos Zuffo http://orcid.org/0000-0002-2186-9755 Do Serviço De Conservação Do Solo Dos Estados Unidos Parte 1:
Tais Arriero Shinma http://orcid.org/0000-0002-8536-9063 Classificação.” Revista Brasileira De Recursos Hídricos 10 (4): 5–18.
Vassiliki Terezinha Galvão Boulomytis http://orcid.org/0000-0001-7730- doi:10.21168/rbrh.v10n4.p5-18.
7033 Grove, M., J. Harbor, and B. Engel. 1998. “Composite Vs. Distributed Curve
Monzur Imteaz http://orcid.org/0000-0003-1800-3157 Numbers: Effects on Estimates of Storm Runoff Dephts.” Journal of the
American Water Resources Association 34 (5): 1015–1023. doi:10.1111/
j.1752-1688.1998.tb04150.x.
References Gupta, S. K., P. K. Singh, J. Tyagi, G. Sharma, and A. S. Jethoo. 2020.
“Rainstorm-generated Sediment Yield Model Based on Soil Moisture
Abbaspour, K.C., E. Rouholahnejad, S. Vaghefi, R. Srinivasan, H. Yang, and
Proxies (SMP).” Hydrological Processes 34 (16): 3448–3463. doi:10.1002/
B. Kløve. 2015. “A Continental-scale Hydrology and Water Quality Model
hyp.13789.
for Europe: Calibration and Uncertainty of A High-resolution Large-scale
Haddad, E. A., and E. Teixeira. 2015. “Economic Impacts of Natural
SWAT Model.” Journal of Hydrology 524: 733–752. doi:10.1016/j.
Disasters in Megacities: The Case of Floods in São Paulo, Brazil.”
jhydrol.2015.03.027.
Habitat International 45 (2015): 106–113. doi:10.1016/j.
Ajmal, M., and T-W. Kim. 2015. “Quantifying Excess Stormwater Using SCS-
habitatint.2014.06.023.
CN–Based Rainfall Runoff Models and Different Curve Number
Hawkins, R. H. 1993. “Asymptotic Determination of Runoff Curve Numbers
Determination Methods.” Journal of Irrigation and Drainage Engineering
from Data.” Journal of Irrigation and Drainage Engineering 119 (2):
141 (3): 04014058. doi:10.1061/(ASCE)IR.1943-4774.0000805.
334–345. doi:10.1061/(ASCE)0733-9437(1993)119:2(334).
Ajmal, M., M. Waseem, J. H. Ahn, and T-W. Kim. 2016. “Runoff Estimation
Hawkins, R. H., T. J. Ward, D. E. Woodward, and J. A. Van Mullem. 2009.
Using the NRCS Slope-Adjusted Curve Number in Mountainous
“Curve Number Hydrology.” In Curve Number Hydrology: State of the
Watersheds.” Journal of Irrigation and Drainage Engineering 142 (4):
Practice, edited by Richard H. Hawkins, Timothy J. Ward, Donald
04016002. doi:10.1061/(ASCE)IR.1943-4774.0000998.
E. Woodward, and Joseph A. Van Mullem, Reston, VA: American
Alvares, C. A., J. L. Stape, P. C. Sentelhas, J. L. M. Gonçalves, and G. Sparovek.
Society of Civil Engineers. doi:10.1061/9780784410042.
2013. “Köppen’s Climate Classification Map for Brazil.” Meteorologische
Huang, M., J. Gallichand, Z. Wang, and M. Goulet. 2006. “A Modification to
Zeitschrift 22 (6): 711–728. doi:10.1127/0941-2948/2013/0507.
the Soil Conservation Service Curve Number Method for Steep Slopes in
Andrade, E. M, J. R Araújo Neto, M. J. S. Guerreiro, J. C. N. Santos, and
the Loess Plateau of China.” Hydrological Processes 20 (3): 579–589.
H. A. Q. Palácio. 2017. “Land Use Effect on the CN Model Parameters in
doi:10.1002/hyp.5925.
a Tropical Dry Environment.” Water Resources Management 31 (13):
IBGE. 2010. Pesquisa Nacional De Saneamento Básico 2008. Rio de Janeiro,
4103–4116. doi:10.1007/s11269-017-1732-4.
Brasil: Instituto Brasileiro de Geografia e Estatística. Ministério do
Angelini Sobrinha, L., L. G. B. Martins, R. M. Genova, M. V. Galbetti,
Planejamento, Orçamento e Gestão.
M. M. Mine, and J. L. B. Brandão. 2014. “Challenges of the Hydrological
IUSS Working Group WRB. 2015. World Reference Base for Soil Resources
Monitoring of Small Hydrographic Basins: Case Study of the Mineirinho
2014, Update 2015 International Soil Classification System for Naming Soils
Stream Basin, São Carlos (SP, Brazil).” 6th International Conference on
and Creating Legends for Soil Maps. Rome: World Soil Resources Reports
Flood Management, 1–12.
No. 106. FAO.
Bono, J. A. M., M. C. M. Macedo, C. A. Tormena, M. R. Nanni, E. P. Gomes, and
M. M. L. Müller. 2012. “Infiltração De Água No Solo Em Um Latossolo Kadam, A. K., S. S. Kale, N. N. Pande, N. J. Pawar, and R. N. Sankhua. 2012.
Vermelho Da Região Sudoeste Dos Cerrados Com Diferentes Sistemas “Identifying Potential Rainwater Harvesting Sites of a Semi-Arid, Basaltic
De Uso E Manejo.” Revista Brasileira De Ciência Do Solo 36 (6): 1845–1853. Region of Western India, Using SCS-CN Method.” Water Resources
doi:10.1590/S0100-06832012000600019. Management 26 (9): 2537–2554. doi:10.1007/s11269-012-0031-3.
Boulomytis, V. T. G., M. A. Imteaz, A. C. Zuffo, and C. D. Alves. 2016. “Analysis Krause, P., D. P. Boyle, and F. Bäse. 2005. “Comparison of Different Efficiency
of the Urbanisation Effects on the Increase of Flood Susceptibility in Criteria for Hydrological Model Assessment.” Advances in Geosciences 5:
Coastal Area.” Theoretical and Empirical Researches in Urban 89–97. doi:10.5194/adgeo-5-89-2005.
Managemennt 11 (4): 30–45. Lal, M., S. K. Mishra, A. Pandey, R. P. Pandey, P. K. Meena, A. Chaudhary,
Brasil. 2007. “Mapeamento De Riscos Em Encostas E Margem De Rios”. R. K. Jha, A. K. Shreevastava, and Y. Kumar. 2017. “Evaluation of the Soil
Ministério das Cidades e Instituto de Pesquisas Tecnológicas (IPT), Conservation Service Curve Number Methodology Using Data from
Brasília, Brasil. Agricultural Plots.” Hydrogeology Journal 25 (1): 151–167. doi:10.1007/
Cunha, S., F. Silva, T. Mota, and M. Pinheiro. 2015. “Avaliação Da Acurácia s10040-016-1460-5.
Dos Métodos Do SCS Para Cálculo Da Precipitação Efetiva E Hidrogramas Lima, A. P., and M. C. C. T. Amorim. 2015. “Análise De Episódios De
De Cheia /Evaluation of the SCS Method for Effective Rainfall and Flood Alagamentos E Inundações Urbanas Na Cidade De São Carlos a Partir
Hydrograph Estimation.” Revista Brasileira De Recursos Hídricos 20 (4): De Notícias De Jornal.” Revista Brasileira De Climatologia 15. doi:10.5380/
837–848. doi:10.21168/rbrh.v20n4.p837-848. abclima.v15i0.33406.
URBAN WATER JOURNAL 255

Liu, W., W. Chen, and Q. Feng. 2018. “Field Simulation of Urban Surfaces Runoff Soulis, K. X., and J. D. Valiantzas. 2012. “SCS-CN Parameter Determination
and Estimation of Runoff with Experimental Curve Numbers.” Urban Water Using Rainfall-runoff Data in Heterogeneous Watersheds-the two-CN
Journal 15 (5): 418–426. doi:10.1080/1573062X.2018.1508597. System Approach.” Hydrology and Earth System Sciences 16 (3):
Mishra, S. K., M. K. Jain, and V. P. Singh. 2004. “Evaluation of the SCS-CN- 1001–1015. doi:10.5194/hess-16-1001-2012.
Based Model Incorporating Antecedent Moisture.” Water Resources Soulis, K. X., and J. D. Valiantzas. 2013. “Identification of the SCS-CN
Management 18 (6): 567–589. doi:10.1007/s11269-004-8765-1. Parameter Spatial Distribution Using Rainfall-Runoff Data in
Mishra, S.K., V.P. Singh, and P.K Singh. 2018. “Revisiting the Soil Heterogeneous Watersheds.” Water Resources Management 27 (6):
Conservation Service Curve Number Method.” In Hydrologic Modeling. 1737–1749. doi:10.1007/s11269-012-0082-5.
Water Science and Technology Library, edited by Vijay P Singh, Shalini Souza, J. L. M., K. F. Fezer, B. C. Gurski, D. Jerszurki, P. E. Pachechenik, and
Yadav, and Ram Narayan Yadava, Vol. 81, 667–693. Singapore: Springer. A. W. P. Evangelista. 2018. “Atributos Físicos E Balanço Hídrico Do Solo
doi:10.1007/978-981-10-5801-1_46. Com Floresta Ombrófila Mista Em Latossolo Vermelho-amarelo Em
Moriasi, D. N., M. W. Gitau, N. Pai, and P. Daggupati. 2015. “Hydrologic and Telêmaco Borba – PR.” Ciência Florestal 28 (1): 90. doi:10.5902/
Water Quality Models: Performance Measures and Evaluation Criteria.” 1980509831583.
Transactions of the ASABE 58 (6): 1763–1785. doi:10.13031/ Tedela, N. H., S. C. McCutcheon, T. C. Rasmussen, R. H. Hawkins,
trans.58.10715. W. T. Swank, J. L. Campbell, M. B. Adams, C. R. Jackson, and
Nash, J.E., and J.V. Sutcliffe. 1970. “River Flow Forecasting through E. W. Tollner. 2012. “Runoff Curve Numbers for 10 Small Forested
Conceptual Models Part I — A Discussion of Principles.” Journal of Watersheds in the Mountains of the Eastern United States.” Journal
Hydrology 10 (3): 282–290. doi:10.1016/0022-1694(70)90255-6. of Hydrologic Engineering 17 (11): 1188–1198. doi:10.1061/(ASCE)
Neitsch, S. L., J. G. Arnold, J.R. Kiniry, and J. R. Williams. 2009. “Equations: HE.1943-5584.0000436.
Surface Runoff.” In Soil and Water Assessment Tool Theoretical Tolentino, M. 2007. Estudo Crítico Sobre O Clima Da Região De São Carlos.
Documentation: Version 2009, 98–122. Temple, TX: Grassland, Soil and São Carlos: EdUFSCar. São Carlos, Brazil.
Water Research Laboratoty, Agricultural Research Service. http://swat. USDA (United States Department of Agriculture). 2004a. ““Estimation of
tamu.edu/documentation/ Direct Runoff from Storm Rainfall”. Chapter 10.” In National Engineering
Oliveira, P. T. S., M. A. Nearing, R. H. Hawkins, J. J. Stone, D. B. B. Rodrigues, Handbook: Part 630 Hydrology, edited by, 1–79. Washington: United
E. Panachuki, and E. Wendland. 2016. “Curve Number Estimation from States Department of Agriculture (USDA) and Natural Resources
Brazilian Cerrado Rainfall and Runoff Data.” Journal of Soil and Water Conservation Service (NRSC).
Conservation 71 (5): 420–429. doi:10.2489/jswc.71.5.420. USDA (United States Department of Agriculture). 2004b. ““Hydrologic Soil-
Paulino, P. F. 2014. Estudo Sobre a Sensibilidade Dos Parâmetros Do Método SCS Cover Complexes”. Chapter 9.” In National Engineering Handbook:
Na Determinação De Hidrogramas De Cheia Em Bacias Urbanas. São Carlos, Part 630 Hydrology, edited by, 1–20. Washington: United States
Brazil: Escola de Engenharia de São Carlos - Universidade de São Paulo. Department of Agriculture (USDA) and Natural Resources Conservation
Ponce, V. M. (2014). “Engineering Hydrology, Principles and Practices”. Service (NRSC).
Accessed January 18 2021. http://ponce.sdsu.edu/enghydro/ USDA (United States Department of Agriculture). 2009. ““Hydrologic Soil
Ponce, V. M., and R. H. Hawkins. 1996. “Runoff Curve Number: Has It Groups”. Chapter 7.” In National Engineering Handbook: Part 630
Reached Maturity?” Journal of Hydrologic Engineering 1 (1): 11–19. Hydrology, edited by, 1–79. Washington: United States Department of
doi:10.1061/(ASCE)1084-0699(1996)1:1(11). Agriculture (USDA) and Natural Resources Conservation Service (NRSC).
Rossi, M. 2017. Mapa Pedológico Do Estado De São Paulo: Revisado Valle Junior, L. C. G., D. B. B. Rodrigues, and P. T. S. Oliveira. 2019.
E Ampliado. Vol. 124. São Paulo: SMA/SP, Secretaria do Meio Ambiente “Initial Abstraction Ratio and Curve Number Estimation Using Rainfall
do Governo do Estado de São Paulo. https://smastr16.blob.core.win and Runoff Data from a Tropical Watershed.” Revista Brasileira De
dows.net/iflorestal/2017/11/Livro_Solos1.pdf Recursos Hídricos 24: 1–9. doi:10.1590/2318-0331.241920170199.
Santikari, V. P., and L. C. Murdoch. 2018. “Including Effects of Watershed Van Mullem, J. A., D. E. Woodward, R. H. Hawkins, A. T. Hjelmfelt, and
Heterogeneity in the Curve Number Method Using Variable Initial Q. D. Quan. 2002. “Runoff Curve Number Method: Beyond the
Abstraction.” Hydrology and Earth System Sciences 22 (9): 4725–4743. Handbook”. Proceedings. 2nd Federal Interagency Hydrologic
doi:10.5194/hess-22-4725-2018. Modeling Conference, Las Vegas, Nevada, July.
Sartori, A., R. H. Hawkins, and A. M. Genovez. 2011. “Reference Curve Verma, S., P. K. Singh, S. K. Mishra, V. P. Singh, V. Singh, and A Sinsh. 2020.
Numbers and Behavior for Sugarcane on Highly Weathered Tropical “Activation Soil Moisture Accounting (ASMA) for Runoff Estimation
Soils.” Journal of Irrigation and Drainage Engineering 137 (11): 705–711. Using Soil Conservation Service Curve Number (SCS-CN) Method.”
doi:10.1061/(ASCE)IR.1943-4774.0000354. Journal of Hydrology 589 (May): 125114. doi:10.1016/j.
Sharpley, A.N., and J.R. Williams (Eds.), 1990. “EPIC-Erosion Produc- Tivity jhydrol.2020.125114.
Impact Calculator, 1. Model Documentation.” US Department of Villanueva, A., and D. Piccilli. 2007. “Custo Da Incerteza Na Macrodrenagem
Agriculture, Agricultural Research Service, Tech. Bull. 1768 Urbana: Influência Do Erro Na Estimativa Do CN Sobre O Custo De
Shinma, T. A. 2015. Avaliação De Incertezas Na Calibração Automática Do Investimento Da Rede.” Revista Brasileira De Recursos Hídricos 12 (1):
Modelo SWMM. São Carlos, Brazil: Escola de Engenharia de São Carlos - 79–90. doi:10.21168/rbrh.v12n1.p79-90.
Universidade de São Paulo. Walsh, S., and D. Diamond. 1995. “Non-Linear Curve Fitting Using Microsoft
Singh, P. K., M. L Gaur, S. K. Mishra, and S. S. Rawat. 2010. “An Updated Excel Solver.” Talanta 42 (4): 561–572. doi:10.1016/0039-9140(95)01446-I.
Hydrological Review on Recent Advancements in Soil Conservation Williams, J. R., and W. V. LaSeur. 1976. “Water Yield Model Using SCS Curve
Service-curve Number Technique.” Journal of Water and Climate Numbers.” Journal of the Hydraulics Division 102 (9): 1241–1253.
Change 1 (2): 118–134. doi:10.2166/wcc.2010.022. doi:10.1061/JYCEAJ.0004609.
Singh, P. K., B. K. Yaduvanshi, S. Patel, and S. Ray. 2013. “SCS-CN Based Yuan, Y., W. Nie, S. C. McCutcheon, and E. V. Taguas. 2014. “Initial
Quantification of Potential of Rooftop Catchments and Computation of Abstraction and Curve Numbers for Semiarid Watersheds in
ASRC for Rainwater Harvesting.” Water Resources Management 27 (7): Southeastern Arizona.” Hydrological Processes 28 (3): 774–783.
2001–2012. doi:10.1007/s11269-013-0267-6. doi:10.1002/hyp.9592.

You might also like