Journal of Hydrology 554 (2017) 406–419

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

The economic value of the flow regulation environmental service in a

Brazilian urban watershed
Guilherme F. Marques a,⇑, Verônica B.F.S. de Souza b, Natália V. Moraes a
a
Instituto de Pesquisas Hidraulicas (IPH), Universidade Federal do Rio Grande do Sul (UFRGS), Campus do Vale da UFRGS – IPH, Av. Bento Gonçalves, 9500 – Agronomia, Porto
Alegre, RS 91501-970, Brazil
b
UNI-BH, R. Diamantina, 567 - Lagoinha, Belo Horizonte, MG 31110-320, Brazil

a r t i c l e i n f o a b s t r a c t

Article history: Urban flood management have often focused either on the capacity expansion of drainage systems or on
Received 13 February 2017 artificial detention storage. While flood control should take part early on urban planning, not enough is
Received in revised form 29 August 2017 known to guide such plans and provide incentive to land use decisions that minimize the vulnerability to
Accepted 30 August 2017
localized floods. In this paper, we offer a broader perspective on flood protection, by treating the original
Available online 6 September 2017
This manuscript was handled by G. Syme,
hydrologic flow regulation as an environmental service, and exploring how the value of this environmen-
Editor-in-Chief, with the assistance of Helge tal service drives economic land use decisions that convert original (permeable) land into urbanized
Bormann, Associate Editor (impermeable). We investigate the relationship between land use decisions and their hydrologic conse-
quences explicitly, and use this relationship to simulate resulting land use scenarios depending on the
Keywords: value attached to the environmental service of flow regulation. Rainfall-runoff simulation model results
Flood control are combined to an optimization model based on two-stage stochastic programming approach to model
Urban planning economic land use decisions. The objective function maximizes the total expected land use benefit in an
Environmental services urban area, considering the opportunity cost of permeable areas in the first stage and the resulting loss of
Economic instruments for urban flood the environmental service of flow regulation on the second stage, under several probable hydrological
management events. A watershed in the city of Belo Horizonte, Brazil, is used to demonstrate the approach.
Different values attached to the environmental service were tested, from zero to higher than the oppor-
tunity cost of land, and artificial detention infrastructure was included to calculate the resulting land use
change and the loss in the environmental service value. Results indicate that by valuing the environmen-
tal service loss and discounting it from the economic benefits of land use, alternative solutions to land use
are found, with decreased peak flows and lower flood frequency. Combined solutions including structural
and non-structural techniques provide more cost effective results, avoiding both the depletion of the
environmental service and the high opportunity cost associated to valuable commercial urban areas.
Urban development under such premises will be more resilient and adapted to local flooding, instead
of relying on increasingly expensive infrastructure.
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction Under natural conditions, the water cycle in a watershed is one

The process of urban growth and development inevitably brings
a reduction in the permeable area (Hamel et al., 2013) through the
construction of buildings, pavement and urban utilities. This
affects local climate, landscape and ultimately people’s lives
(Brown et al., 2008) with heat island effects, reduction in rainfall
infiltration and increase in runoff volume and velocity, in addition
to water quality issues.
⇑ Corresponding author.
E-mail addresses: Guilherme.marques@ufrgs.br (G.F. Marques), veronica.
bernardes.bh@gmail.com (V.B.F.S. de Souza), vmoraes.nati@gmail.com
(N.V. Moraes).
of the principal factors that determine the fluvial morphology and
floodplain dynamics (e.g. frequency, duration and magnitude of
floods). Urban growth interferes in these hydrological processes
by reducing infiltration and watershed storage, thus affecting the
flow regulation.
We can perceive the flow regulation as an environmental ser-
vice, as its presence contributes to less frequent flooding when
compared to conditions with reduced watershed storage and flow
regulation (Arico et al., 2005). Given that less frequent flooding
often means higher economic benefits from land use (lower main-
tenance, lower risk and losses to urban systems and people, higher
potential for occupation), the flow regulation service has economic
value to society. Similar to other common goods (natural resources

https://doi.org/10.1016/j.jhydrol.2017.08.055
0022-1694/Ó 2017 Elsevier B.V. All rights reserved.
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
and environmental services), the flow regulation service value is
not clearly signaled to society and users, resulting in the overex-
ploitation of the service as indicated in Hardim (1968). The pres-
ence of excess runoff and more frequent flooding is thus an
externality of the decisions taken by urban inhabitants under a
given economic context, public policy, urban planning and con-
struction market. While such decisions are individual, it is up to
urban planners and water managers to design and implement
instruments to signal the value and scarcity of environmental ser-
vices to the users. The term ‘‘excess runoff” is here used to define
the additional runoff produced by land use change in urban areas
due to reduced infiltration and storage when compared to natural
conditions.
Traditional engineering solutions have long managed the exter-
nalities associated with the loss of this environmental service with
drainage systems designed primarily to convey the water down-
stream as efficiently as possible. However, such drainage solutions
are now showing limitations in many urban areas to provide con-
sistent, reliable and cost effective protection against flooding
(Brown et al., 2008; Burns et al., 2012) giving room to other tech-
niques based on LID (low impact development) and conservation of
some natural functions of the watershed (Tucci, 2008; Canholi,
2005; Cembrano et al., 2004). Urban drainage issues are not
uncommon in Brazil, which had its early urban drainage develop-
ment based on conveyance of excess runoff through structural
measures rather than its control at the site.
Each urban land user including policy makers, urban planners
and construction industries bears responsibility over the occur-
rence of excess runoff when taking decisions that alter the hydro-
logic cycle (e.g. reducing infiltration by building a parking lot). In
this context, it is important to design efficient LID solutions and
to understand the behavior of the users (i.e. their land use
decision-making), so that regulatory and economic instruments
can be put into place to manage urban waters. It is also necessary
to evaluate the environmental services present in the watershed,
how such value drives the inhabitants’ decisions and how these
decisions contribute to the loss of the environmental service. The
objective of this paper is to address these important aspects.
This paper presents an approach to flood protection and man-
agement through a novel representation of urbanization decisions
integrated to their negative externalities in the watershed using a
stochastic simulation-optimization approach. While most existing
work as in Paule-Mercado et al. (2017) and Loperfido et al. (2014)
have simulated the effects of a given stormwater management
solution in order to verify its benefits, we integrate simulation into
a land use model that weighs economic benefits against environ-
mental service losses, bringing in user preferences as a key
parameter.
The results produced by the model should be useful to (a)
understand how the value of the environmental service of flow
regulation drives land use decisions in urban watersheds; (b) sup-
port application of economic instruments for managing urban
waters (e.g. flood tariffs or payment for environmental services)
and (c) support environmental impact assessment studies and def-
inition of mitigation measures.
Common approaches to study economic instruments for urban
stormwater management have focused on tariffs applied to imper-
meable areas (Haarhof, 1996; BARRAQUÉ, 1995; ANTENER, 1999).
A limitation of such methods is the lack of other factors responsible
for peak flows, such as topography and soil type. Thus, areas of the
same size, but with distinctive features, under the same hydrologic
scenario may produce different results. An economic instrument
may offer a credit to a user who leaves a fraction of his land perme-
able, or a fee if the instrument is based on impermeable area,
depending on the actual result (runoff) produced by his land parcel
and not solely on the area. In this paper, we adopt this broader
407
concept to calculate the loss of the environmental service of flow
regulation, replacing the traditional view based only on the area
or on the incorporation of a given LID.
Ossa-Moreno et al. (2017) highlight the benefits of Sustainable
Urban Drainage Systems (SuDS) and point out that economic fac-
tors are among the main obstacles to their implementation, espe-
cially when asset value of the SuDS and the full scope of its
benefits are not quantified. The authors present a detailed evalua-
tion of retrofitting SuDS in an urban area and the associated bene-
fits through a cost-benefit analysis. While Ossa-Moreno et al.
(2017) conclude that implementation of SuDS is improved when
wider benefits are considered, it is still not clear how, and where,
such benefits would drive local decisions regarding urban land
use change.
McNeill and Dollery (2007) present a method to calculate devel-
oper charges for urban infrastructure (water supply, sewerage and
drainage) but best management practices for stormwater control
are not included. The cost of best management practices and land
for urban stormwater management is calculated in a general
method by Sample et al. (2003), which considers housing (con-
struction and land) and transportation costs. It is pointed out that
a defensible method is still needed to allocate the costs among dif-
ferent uses of developed land, highlighting the need to better
understand the land development costs owned to stormwater
management.
The present paper contributes to fulfill those gaps by modeling
users’ decisions regarding urban land use integrated to the evalu-
ation of the environmental service provided by the watershed. Fur-
thermore, the hydrologic and economic consequences of such
decisions under stochastic conditions are calculated. Given that
our approach is based on distributed hydrologic modeling, it is
possible to identify where in the watershed the land use decisions
should favor more or less intense occupation, as well as the conse-
quences of the decisions. This should allow public policies and eco-
nomic instruments to be tailored according to each location.
To test our model and approach, a study area in the Cercadinho
watershed, in the city of Belo Horizonte (Brazil) was chosen. The
city of Belo Horizonte and neighboring municipalities have been
subject to high incidence of floods (flash floods) and landslides,
which affected more than 500,000 inhabitants from 1991 to 2010
(OECD, 2016). Therefore, Cercadinho watershed provides a good
example of the city’s exposure to natural disasters, given it has
shown local flooding issues, and it still maintains significant por-
tions with natural vegetation with steep terrain. According to
OECD (2016), this might be aggravated as a result of climate
change, urbanization and deforestation of remaining natural
vegetation.
Under the current regulatory system in Brazil, urban drainage
management is responsibility of the municipality, along with san-
itation and land use planning. However, urban drainage is often
managed by public works departments, which actions are discon-
nected from the sanitation, solid waste, water, sewage and land use
management actions. This fragmented planning leads to a limited
view of the drainage problem, its causal relationship with land
use and a lack of perception of the environmental benefits and
costs to users that take place at the watershed scale. Brazilian fed-
eral laws for water and urban planning provide several economic
instruments (e.g. water charges, fiscal and financial incentives,
property taxes, payment for environmental services, and drainage
tariffs, among others) at the disposal of urban planners to (a) signal
to users the scarcity of the environmental services, as well as their
economic benefit and (b) motivate users to adopt best manage-
ment practices to reduce runoff at their property. However, these
instruments need to be deployed as part of a comprehensive urban
plan connecting land use, watershed hydrologic responses and
economic value and benefit to users. The modeling approach
408 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
presented in this paper contributes to the understanding of this
connection, necessary for the design of effective economic instru-
ments and its acceptance by users.
2. Methods
The economic value of the flow regulation environmental ser-
vice is considered nonexistent when land use decisions are made.
In order to attach a value to this service one could look into (a)
what does it cost to society by not having it (e.g. flood damage cost)
or (b) what does society expect in terms of environmental quality
in the watershed, which goes beyond flood damage cost. We have
chosen the later approach because it is broader and not likely to
underestimate the environmental value like the former.
However, that option lacks a clear reference value for the envi-
ronmental service. To avoid this limitation, we incorporate multiple
reference values for the flow regulation environmental service to
weigh it against the economic returns of land use. Three values
are considered as input data to the model: below, equals and above
the economic return of land use. Each value thus represents a differ-
ent preference that will drive model results. In the model, a low ref-
erence value puts a lower weight at the economic loss to society if
the environmental service is depleted. In this case, the land use
decisions will favor the economic use of the land, rather than pre-
serving it to maintain the existing environmental services.
This approach has three benefits: (a) it avoids the use of an arbi-
trary value for the environmental service; (b) it provides a sensitiv-
ity analysis and (c) it enables the model to investigate the impact of
different society preferences regarding environmental quality on
land use decisions (permeable area left), as it will be shown later on.
Each reference value for the flow regulation environmental ser-
vice (represented by parameter F) is run against two scenarios with
an optimization model (CASE II and CASE III) to determine land use
decisions driven by economics (benefit maximization). The opti-
mization model runs also determines the loss of the environmental
service of flow regulation based on the peak flows produced by a
parcel of urban land as a consequence of the users’ decisions. The
higher the peak flow, the higher the loss of the flow regulation ser-
vice value (under a same rain event). A CASE I represents a ‘‘base”
case with no calculation of the environmental service value, indi-
cating the maximum potential economic return of land use when
environmental services are disregarded. Finally, CASE 0 represents
the possible peak flows under current watershed conditions.
Each optimization model run scenario describes a different land
use management strategy. CASE I has no environmental service
value considered and CASE II has it. Given that the adoption of
structural measures (e.g. detention storage) is a very common
solution to manage runoff, CASE III includes environmental service
value and water detention structures to provide additional flow
regulation capacity. The overall methodology is presented in Fig. 1.
A hydrologic simulation model, built with the Storm Water Man-
agement Model (SWMM) (Gironás et al., 2010; Jang et al., 2007) is
used to define a mathematical relationship between land use deci-
sions and peak flows, which is embedded into the optimization
model as polynomials. The study area is divided into 11 sub-
catchments and each one has different characteristics (slope,
roughness, among others). The SWMM is run successive times
varying the % impervious area from 0 to 100% to construct a rela-
tionship between land use decisions and peak flows. Each sub-
watershed has specific drainage infrastructure and its capacity is
based on Belo Horizonte’s Municipality Development Superinten-
dence database (SUDECAP).
The optimization model is implemented in the General Algebraic
Modeling System (GAMS) (McCARL et al., 2016) platform using a
two-stage quadratic stochastic programming approach. The two-
stage approach is suitable for modeling a decision structure with
permanent and temporary elements. Permanent decisions are
modeled in the first stage, while the consequences (or other deci-
sions) are considered in the second stage, given the realization of a
stochastic event. Applications of this approach to different water
problems can be found in Wilchfort and Lund (1997) and
Marques et al. (2005).
Fig. 2 presents the structure of the optimization model. The first
stage models permanent decisions of occupation/non-occupation
of a land parcel. For simplification, we consider ‘‘occupation” as
impermeable land and ‘‘non-occupation” as permeable. Non-
occupation basically means that the full scale of economic returns
from commercial use is forgone in favor of a land use that con-
tributes to infiltration and maintenance of the service of flow reg-
ulation in the watershed. While there are several intermediate
levels of occupation and land use, with different economic returns
and environmental impacts, we have chosen not to represent them
in detail here, focusing on the model concept instead. Although no
decision is indefinitely ‘‘permanent”, these are not expected to
change frequently. First stage decisions of occupation/non-
occupation produce an economic benefit based on the revenues
generated by land use. In the case study, we focus on areas where
the owner of a land parcel decides to develop it by building a
multi-store building and selling it in the market.
The second stage models the resulting flow regulation service
value loss associated to the peak flow generated for each probable
hydrologic scenario j. A hydrologic scenario is composed by a
specific rain event and a probability of occurrence. This value loss
reflects the loss of the environmental service of flow regulation in a
given urban watershed. The watershed is divided into smaller sub-
watersheds i.
The objective function (1) maximizes the total expected value
(TEV) of urban land use, considering the total combined land use
values of each sub-watershed. This total value is given by the eco-
nomic value of land use in the first stage C1 subtracted by the
expected flow regulation service value loss in the second stage
C2. The larger the occupied land (impermeable) in the first stage,
the higher the economic use and the value C1.
However, more occupied land also means more runoff and
higher peak flows, which increase the value C2 and the loss in
the flow regulation service. On the other hand, leaving more per-
meable land (e.g. parks and vegetated strips) reduces the occupa-
tion potential and the economic value C1, but it is compensated
by a smaller loss value C2 in the second stage given the reduced
runoff and peak flows. The optimization model searches for a ‘‘bal-
ance” between C1 and C2 that results in the largest possible total
expected value TEV and calculates the pervious and imperious
areas for each sub-watershed individually.
Finally, given CASE III has water detention structures to aid flow
regulation, their cost is included as C3. C3 is a parameter pre-
calculated according to the size of the structure and such calcula-
tion is performed separated from the optimization routine (i.e. it
has a single value depending on the sub-watershed characteristics
and design storm). It is important to note that C2 is not calculated
in CASE 0 and CASE I. Thus, while CASE I presents the maximum
value for TEV, it does not take into account the flow regulation ser-
vice value loss associated with it. For that reason, a direct economic
comparison of CASE I with CASE II and CASE III (which do calculate
C2) does not allow one to draw conclusions on the best decision.
CASE I is useful to illustrate the maximum potential economic ben-
efit when the environmental service is disregarded. On the other
hand, a comparison between CASE II and CASE III should allow the
evaluation of the best decision (i.e., the one with the higher TEV).
For i sub-watersheds, we have:
X X X
TEV ¼ C1i  C2i  C3i ð1Þ
i i i
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419 409

Fig. 1. Diagram for the overall methodology, where PA, QP, C1 and C2 are, respectively, permeable area, peak flow, economic value of land use and expected flow regulation
service value loss.

Fig. 2. Decision diagram for the two-stage stochastic optimization model (square represents decision, circles represent stochastic events, triangles represent consequences).

where in each sub-watershed QPi was recorded. The group of simulations

was repeated for each one of the probable hydrologic scenario. A
C1i ¼ ðTAi  PAi Þ  BLU i ð2Þ
3rd degree polynomial (4) was fit to the PA vs QP relationship so
X it could be represented in the optimization model.
C2i ¼ P j  SVLi;j ð3Þ
j QPi;j ¼ ðaai;j  PA3i Þ þ ðbbi;j  PA2i Þ þ ðcci;j  PAi Þ þ ddi;j ð4Þ
2
where TAi is the total area in sub-watershed i (km ); PAi is the non- where aai,j, bbi,j, cci,j e ddi,j are fitting parameters for each sub-
occupied, permeable area in sub-watershed i (km2); BLUi is the eco- watershed i and hydrologic scenario j.
nomic return of land use (annualized, in R$/km2); j is the hydrologic Lesser degree polynomials did not result in a good fit, while
scenario and pj is the probability of the hydrologic scenario. higher order tends to incorporate noise in the relationship. Thus,
SVLij is the flow regulation service value loss ($) in sub- we have chosen the least degree possible still able to produce a
watershed i and hydrologic scenario j, determined as an increasing very good fit (R2 above 0.99). Fig. 3 exemplifies the curve’ s behav-
function of the peak flow produced at the exutory of that sub- ior obtained from PA vs QP relationship.
watershed. The peak flows calculation is done with SWMM, where Eq. (5) limits the non-occupied, permeable area in sub-
a model of the study area and data was built. watershed i to the total area TAi.
An external script automated multiple SWMM runs, where PAi
varied from 100% to 0%, in 5% intervals, and the resulting peak flow
PAi 6 TAi ð5Þ
410 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
Fig. 3. Example polynomial fit for sub-watershed A and hydrologic scenarios 1 and 8.
Given that the fitted relationship QP vs. PA is an approximation,
it is necessary to constrain the QP values to the minimum the
watershed can produce (6).
QP i;j P Qmini;j
The relationship between the flow regulation service value loss
(SVL) and the peak flow (QP) depends on local conditions, includ-
ing spatial distribution of the economic value of the land across
the watershed and the relationship between land use changes
and QP. The investigation of these aspects is beyond the scope of
the paper, thus we opted to use the simplest form, which is linear.
This function however has two important features:
(a) The starting minimum point on each curve refers to the
smallest peak flow produced for each hydrologic event when the
watershed is 100% non-occupied. We thus assume a 100% non-
occupied watershed has full flow regulation service and a zero
value of SVL (zero value loss); (b) The ending, highest point on each
curve refers to the highest peak flow produced for each hydrologic
event when the watershed is 100% occupied. In this case, the flow
regulation service value loss SVL is equal to the reference values for
the flow regulation environmental service multiplied by the total
area. Thus it can be: (1) below the economic return of land use,
(2) equals the economic return of land use and (3) above the eco-
nomic return of land use. Each reference represents a preference,
describing how urban environmental quality and the associated
environmental services are valued compared to economic benefits
from developing the land for commercial and residential purposes.
The economic return of land use (R$/km2) is represented by the
parameter BLUi. The service value loss at 100% land occupation
considered were 0.8 ⁄ BLU; 1 ⁄ BLU; and 1.2 ⁄ BLU. The 0.8, 1.0
and 1.2 are the values applied to the parameter F.
If the users decide to fully occupy the watershed, the imperme-
able surface would produce a loss of flow regulation service that is
valued at above, equal or below the economic benefits produced by
the full occupation. The SVLij values are given by (7):
SVLi;j ¼ ðai;j  QP i;j Þ þ bi;j
where ai,j and bi,j are curve parameters and QPij is the peak flow in
sub-watershed i and hydrologic event j.
2.1. Hydrologic scenarios
The Two-stage stochastic model is run for a horizon of discrete
hydrologic scenarios, each one associated with a probability of
occurrence within the given discrete intervals. Intensity-
duration-frequency (IDF) curves were used to calculate the rain
intensity for different probabilities from 0.005 to 0.95 and duration
equals to 2 h. These curves were obtained through a IDF equation
proposed for Pinheiro e Naghettini (1998) and it is specific for Belo
Horizonte (MG/Brazil) region. Based on the probabilities of a given
rain intensity being equaled or exceeded, we subtracted the previ-
ous probability value (i.e. the probability of the previous rain event
among the discrete events), to obtain a probability associated with
a rain intensity interval. The result is a discretization of the rain
intensity probability curve into intervals, each one with a lower
ð6Þ
and upper bound for rain intensity and a probability. The intervals
are represented by the average of the upper and lower bounds
(Table 1) with a total of 8 hydrologic scenarios. Finally, Huff’s
method was used to generate the full histogram for the hydrologic
scenarios. Each scenario provided input data for a SWMM simula-
tion, already described.
2.2. The economic returns of land use
The economic benefits associated to urban land use may include
several different elements, associated to commerce, services, resi-
dences and industrial use, which also depend on the availability
of local infrastructure and public services. In the current paper,
we focus on a commercial area, where the owner of a land parcel
decides to develop it by building a multi-store building and selling
it on the market. By ‘‘commercial” we refer to general land use that
provides direct economic returns (i.e. building and selling a house
or commercial building). The owner thus decides which percentage
of his parcel will be ‘‘occupation” which means built area (build-
ings, parking lots and such) that bring economic return at the
expense of reducing the environmental service of flow regulation,
given this build area is impermeable. The remaining area is then
considered ‘‘non-occupation” which means gardens and natural
vegetation that retain the environmental service of flow regulation,
but forego the full scale of traditional economic return.
Intermediate levels of occupation, including LID, are not yet
included to avoid the natural complexity of all decision possibili-
ties in an urban area, allowing us to focus on presenting the model
concept. SWMM allows representation of different LID technolo-
gies as specific elements within a watershed or as separated ele-
ments. LID technologies, such as permeable pavers, green roofs,
ð7Þ rain barrels and others, have the potential to provide additional
flow regulating services (and also improve water quality, not mod-
eled here), which would allow further urban development and eco-
nomic benefits by reducing the permeable area.
The BLUi (economic return of land use in R$/km2) is based on
corresponding neighborhood real state market value in the water-
shed (Trem Util, 2013) including the average construction costs in
R$/m2 and other complementary costs (taxes, tariffs, advertise-
ment and such). The revenues to the owner/incorporator are given
subtracting the costs (construction and complementary) from the
real state market value. It is used the average value considering
the options available to the owner: residential building, commer-
cial building or selling the unpaved empty lot (which is soon fol-
lowed by construction and loss of the environmental service of
flow regulation). The average value for the total BLU is then
1346 R$ Million/km2 (Table 2).
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419 411

Table 1
Hydrologic scenarios and probabilities.

Hydrologic scenario j Lower bound (mm/h) Upper bound (mm/h) Average (mm/h) Probability
R1 15.56 18.76 17.160 0.15
R2 18.76 23.11 20.935 0.30
R3 23.11 32.79 27.950 0.40
R4 32.79 36.51 34.650 0.05
R5 36.51 37.08 36.795 0.01
R6 37.08 41.32 38.303 0.02
R7 41.32 44.89 43.105 0.01
R8 44.89 48.50 46.695 0.005

Table 2 expense of higher flood peaks moved downstream to Arrudas

Economic returns of urban land use for the neighborhood located at Cercadinho
watershed. Data from Trem Util (2013). Source: Trem Util (2013).
Remuneration of incorporator/owner R$/m2
Residential building R$ 1617.00
Commercial building R$ 1644.00
Empty lot R$ 777.00
Total BLU (average) R$ 1346.00
The BLU value was annualized over a five years’ period, which is
a common time frame for the owner to have returned the invested
capital, through Eq. (8):
X  
n
CF t r  ð1 þ rÞn
UAV ¼  ð8Þ
t¼1
ð1 þ rÞn ð1 þ rÞn  1
where CFt is the project cash flow considering the total return (R
$1346/m2), UAV is the uniform annualized value considering n = 5
periods (years); r is the interest rate (8.4% per annum). According
to Eq. (8) the annualized value is R$340.71/m2. This is the value
used for the annualized BLU.
2.3. Study area and model run cases
The model was applied to a watershed in a large metropolitan
area in Brazil (Belo Horizonte city). The study watershed (part of
Cercadinho watershed) has 4.4 km2 drainage area and it is sub-
divided in 11 sub-watersheds (labeled from A to L in Fig. 5) to bet-
ter represent the heterogeneity of the region. Fig. 4 depicts the
study watershed (delimited in red) indicating a denser building
occupation in the Northern, while parts of the Southern watershed
remain still vegetated.
Table 3 presents some of the characteristics of the sub-
watersheds.
The watershed data (slope, areas, roughness, etc.) and all the
data in Table 3 were provided by the Municipality Development
Superintendence (SUDECAP), which is responsible for stormwater
drainage works in the study area. Slope is based on topographic
survey conducted during the construction of the drainage network.
The permeable and impermeable areas identified in Table 3 were
calculated through manual map classification, also using the SUDE-
CAP land use map database.
The study area was chosen given intense recent building devel-
opment and accelerated reduction of the remaining vegetated
areas. The whole Cercadinho watershed drains to one of the major
rivers in Belo Horizonte city (Arrudas River), which is subject to
frequent flooding. If the urban expansion and development in
watersheds like Cercadinho is not properly done, with evaluation
of the flow regulation provided by the remaining vegetated areas,
the amount of runoff directed to Arrudas River will be higher. This
situation is aggravated if the excessive runoff results in local flood-
ing in Cercadinho watershed. This might trigger the deployment of
additional structural solutions, alleviating the local flood at the
River. This is a common situation in several metropolitan areas
that highlights the limitations of traditional drainage engineering
solutions. Out of the simulated sub-watersheds, A has the largest
area, presents steep topography compared to the others, and also
has little development (impermeable area). Sub-watershed J is
close to the exutory and it presents local flooding. Considering
these aspects, sub-watersheds A and J were chosen to illustrate
some of the results in further detail, as an analysis of all sub-
watersheds is beyond the scope of the paper.
2.4. The optimization model scenarios - land use management
strategies
To evaluate the land use decisions and the resulting loss in the
value of the environmental service of flow regulation, three cases
were prepared. The cases vary by including (or not) the environ-
mental service value and also by including detention tanks to rep-
resent common flood management measures in metropolitan
areas. While detention structures do not provide the same level
of environmental benefits as a vegetated area, they are usually
effective in compensating for the loss of flow regulation by tem-
porarily storing the excessive runoff. The presence of this type of
structural solution prevents the value of the environmental service
of flow regulation to be overestimated, given that the peak flows
can be reduced without resorting exclusively to vegetated areas
to regulate the flow (which often have a very high opportunity
cost). The details and additional equations for each case are pre-
sented as follows.
2.4.1. CASE I – No environmental flow regulation value and no water
detention structures
No value attached to the environmental service of flow regula-
tion and no water detention structures. This allows maximum eco-
nomic benefits of land use for construction, at the expense of
environmental service of flow regulation. C2 and C3 are equal to
zero in (1).
2.4.2. CASE II – With environmental flow regulation value and no
water detention structures
The environmental service of flow regulation is valued. The flow
regulation value loss (C2) is calculated based on the peak flow pro-
duced, and deducted from the land use benefit (C1). The higher the
land benefit C1, the higher the peak flow and the higher the flow
regulation value loss C2, which increases according to the SVL
function (7). This provides an economic incentive to increase the
vegetated area and with it the service of flow regulation in the
basin. Otherwise, a high peak flow would be produced, resulting
in a high flow regulation value loss C2 and a small total expected
value (TEV) in the basin. C3 is equal to zero. Each one of the 11
sub-watersheds in the 4.4 km2 study area has 8 SVL functions (7)
one for each probable hydrologic scenario ‘‘j” in the region
(Table 1).
412 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419

Fig. 4. Satellite image of the study watershed (in red). The remaining area in the map is the Cercadinho watershed. (Barbosa and Fonseca, 2006). (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. Study area divided in sub-watersheds from A to L.

2.4.3. CASE III – With environmental flow regulation value and with produced, and deducted from the land use benefit (C1). However,
water detention structures this is combined with water detention structures to absorb some
The environmental service of flow regulation is valued. The flow of the excess runoff and allow more area to be used for construc-
regulation value loss (C2) is calculated based on the peak flow tion. The water detention tanks result in value C3 that is also
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419 413

Table 3
Sub-watersheds characteristics used in SWMM.

Sub-watershed A B C D E F
Impervious area slope (%) 10.64 11.84 36.07 32.93 29.94 21.12
Pervious area slope (%) 80.90 26.33 31.23 20.13 6.26 65.78
Area (km2) 1.17 0.02 0.53 0.03 0.12 0.72
Average width (m) 23.962.69 840.22 15.146.95 720.65 1.994.90 18.397.02
% Impervious areaa 14.51 56.89 80.69 85.69 97.37 56.87
Impervious roughness coefficient 0.013 0.013 0.013 0.013 0.013 0.013
Pervious roughness coefficient 0.80 0.15 0.15 0.15 0.15 0.15
Sub-watershed G H I J L
Impervious area slope (%) 22.71 53.06 23.60 19.24 22.06
Pervious area slope (%) 23.12 101.98 49.40 4.44 56.14
Area (km2) 0.88 0.28 0.18 0.31 0.14
Average width (m) 20.653.20 6.260.94 4.522.96 7.572.41 4.409.75
% Impervious areaa 63.59 82.50 73.51 92.40 80.69
Impervious roughness coefficient 0.013 0.013 0.013 0.013 0.013
Pervious roughness coefficient 0.15 0.15 0.15 0.15 0.15
a
These were not the values used in SWMM. SWMM was run with a series of values from 0 to 100%.

deducted from C1. This case basically transfers part of the environ-
mental service of flow regulation from the vegetated areas to the
water detention structures, allowing some of the vegetated areas
to be used for construction and reducing the cost to the system.
The water detention structures were designed for 100 years
return period floods and the original sub-watershed conditions
(Table 1). A SWMM simulation model of the watershed was run
previously to identify the likely flooding spots so the location of
the water detention reservoirs could be determined. Only two
out of the eleven sub-watersheds received the reservoirs due to
previous local flooding reported: A and J (the remaining nine did
not present local flooding).
The representation of the dampening effect provided by a water
detention reservoir in the flow required a change in the PA  QP
relationship (4). The SWMM simulation model was run for each
of the two watersheds with the presence of the water detention
reservoir. The simulations were repeated varying the permeable
land area PA from 0 to 100% in 5% increments, and the entire pro-
cess was repeated for each one of the eight hydrologic scenarios.
The results (Fig. 6) showed how the detention reservoir affected
the peak flow and allowed a configuration of different PA vs. QP
relationship.
Fig. 6 shows that for smaller PA values (larger impermeable
area) the rate of increase of the peak flow with PA is smaller than
for higher values of PA (towards the end of the curve). This could
be explained by the fact that under smaller PA the presence of
the reservoir dampens the flow, affecting the increase in QP. Thus,
as the QP increases when PA becomes smaller, it does so at a
slightly smaller rate. On the other hand, for higher values of PA,
the effect of the tank is less evident, as a larger part of the runoff
is stored in the soil given the larger vegetated area, and QP
decreases more rapidly as PA increases. This change in behavior
is more evident under more intense hydrologic scenarios (such
as R7 and R8 in Fig. 7) and almost unnoticeable under less intense
scenarios. Further investigation of the effects of the detention stor-
age in the PA vs. QP relationship is beyond the scope of this paper,
given time and space limitations. Rather, we focused on represent-
ing this behavior change in the optimization model, which
required replacing equation (4) by a different set of curves for
the two sub-watersheds where the reservoirs were present. For
these watersheds, the original polynomial PA vs. QP (4) was
replaced by a curve with two linear segments. The first covers
the domain with the smaller slope in the PA vs. QP relationship
(the well aligned sequence of points in Fig. 8). The second covers
the higher slope in the PA vs. QP relationship (the ‘‘stepped” por-
tion at the end of the curves in Fig. 7). An example of this interpre-
tation for R7 appears in Fig. 6, where c1 and c2 are the two linear
segments that describe the QP vs. PA relationship.

Fig. 6. QP vs. PA relationship without (a) and with (b) the presence of a water detention reservoir (showed here for hydrologic scenario 7 and 8).
414 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419

R7 correct QP. If we need to find out the corresponding QP to

a PA = 1 (now we are in the domain of c2) then LSCc1 must be equal
to 0.72 (intercept of c1 and c2) and LSCc2 must be equal to the
remaining value to reach the given PA: 0.28. By multiplying each
LSC by the slope of its respective curve, adding them and subtract-
ing from the qmax we obtain the correct QP value. This procedure
allows the model to sweep across the compound QP vs. PA rela-
tionship. Eqs. (9)-(12) calculate the LSC values and determine
how a given PA is split.
Eq. (10) limits the size of LSC to the domain of each segment c
(sizec,i,j). For the first segment c1, this domain is the distance from
the origin to the intercept of c1 and c2. For c2, the domain is the
distance from the intercept to the total permeable area PA, which
is a decision variable.
Fig. 7. Representation of QP vs. PA relationship through linear segments c1 and c2
(example for hydrologic scenario R7 shown here). LSC c;i;j 6 SIZEc;i;j ð10Þ
The purpose of Eq. (11) is to make APCc2 to be equal to zero
whenever we are in the domain of curve c1.
Mathematically, this compound relationship is represented by
Eqs. (9)-(12).  
LSC
X SIZEc2;i;j  SIZEc1;i;j
QP i;j ¼ qmax þ LSC c;i;j  aaac;i;j ð9Þ LSC c2;i;j 6  c1;i;j
ð11Þ
i;j LSC
c 1  SIZEc1;i;j  2:34  105 þ 1
c1;i;j

where LSCc,i,j is the length of the segment c in sub-watershed i and

Finally, Eq. (12) determines that the sum APCc1 + APCc2 is
hydrologic event j (measured in the x axis in the chart of Fig. 8) and
always equal to AP.
aaac,i,j is the slope of this segment. qmaxi,j is the intercept of c1 with
the y axis in Fig. 8. X
A given PA is split into a LSC value for curve c1 (LSCc1) and a LSC LSC ¼ PAi ð12Þ
c
value for curve c2 (LSCc2) in order to return the correct QP. For
example, in Fig. 5 both c1 and c2 intercept each other at The total expected value (TEV) is given by (1) where C3 repre-
PA = 0.72. If we need to find out the corresponding QP to sents the cost of the building and maintenance of detention reser-
a PA = 0.5 then LSCc1 must be equal to 0.5 and LSCc2 must be equal voirs (annualized costs). Table 4 presents the costs based on
to 0. This way we are in the domain of c1 and (11) will yield the MOURA (2004).

Fig. 8. Variation of the peak flow produced against the different hydrologic scenario.
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419 415

Table 4 (F = 0.8), the weight of the environmental service in the objective

Stormwater detention reservoir costs. function is small and less than one third of the land is left perme-
Water detention reservoirs Sub-watershed Sub-watershed Total able. On the other hand, a SVL value placed higher than the oppor-
A J tunity cost of land BLU (F = 1.2) resulted in most of the area
Capacity (m3) 5000 11,000 remaining permeable. As expected, the introduction of detention
Operation and maintenance costs 0.173 0.380 0.553 structures allowed some of this permeable area to used economi-
(R$ Million/year) cally, and such effect is more significant when the environmental
Construction cost (R$ Million/year) 0.509 1.121 1.631
Total (R$ Million/year) 0.682 1.501 2.183
service has a higher value (it provides a cost-effective alternative
to regulate flow other than through permeable areas). However,
one should consider that permeable areas covered with vegetation
provide other environmental benefits that the detention tanks
3. Results and discussion don’t.
Considering F = 1 (SVL = BLU) for CASE II, as the model weighs
Considering the three reference values for the flow regulation the economic benefit of land use C1 against the flow regulation
environmental service (below, equal and above the economic value loss C2, the resulting peak flows are smaller than CASE I,
return of land use) different results are available for CASE II and now within the current conveyance capacity of the drainage infras-
III (CASE I has no value for the environmental service). Due to space tructure for all study area for all hydrologic scenarios up to hid7.
limitation, results are organized with a discussion first considering This reduction in the peak flow was result of improved flow regu-
SVL = BLU, and a sensitivity analysis is introduced afterwards with lation through the presence of the permeable areas (see Table 7).
the other values. Fig. 8 presents the variation of the peak flow produced against
For CASE I, the whole study area would be occupied by build- the different hydrologic scenarios, under the three cases simulated,
ings (PA = 0) resulting in the maximum benefit from land use C1. for sub-watersheds A and J considering F = 1. The flat dashed line
However, the peak flow produced exceeded the local conveyance represents the conveyance capacity of the drainage infrastructure.
capacity of the drainage infrastructure at two sub-watersheds (A We notice that for sub-watershed A, CASE I produced peak flows
and J) and at the exutory of the total study area, indicating that exceeding the conveyance capacity more frequently (in 6 out of 8
under current conditions local flooding is likely. hydrologic scenarios). This is what would likely happen in a near
Tables 5 and 6 present further detail, including the permeable future if sub-watershed A (which currently has only 14.51% of its
area resulting from user decisions under the three reference values area impervious – CASE 0) were developed without consideration
for the flow regulation environmental service: below (F = 0.8), of the environmental service value. Under CASE 0 (i.e. maintaining
equals (F = 1) and above (F = 1.2) the economic return of land current impervious level) the peak flows are under the capacity for
use. In general, as expected, when SVL is valued below BLU all hydrologic scenarios but one.

Table 5
Results for CASE II.

Permeable area (%)

Sub watershed TA (km2) F = 0.8 F = 1.0 F = 1.2
A 1.178 0 65.66 76.20
B 0.018 0 97.48 97.48
C 0.532 99.76 99.76 99.76
D 0.036 100.00 100.00 100.00
E 0.121 99.47 99.47 99.47
F 0.726 0 99.86 99.86
G 0.217 0 99.72 99.72
H 0.277 0 100.00 100.00
I 0.176 0 99.74 99.74
J 0.314 100.00 100.00 100.00
L 0.145 0 100.00 100.00
Total 100 26.77 89.07 92.38

Table 6
Results for CASE III.

Permeable area (%)

Sub watershed TA (km2) F = 0.8 F = 1.0 F = 1.2
A 1.178 0 0 0.41
B 0.018 0 97.48 97.48
C 0.532 99.76 99.76 99.76
D 0.036 100.00 100.00 100.00
E 0.121 99.47 99.47 99.47
F 0.726 0 99.86 99.86
G 0.217 0 99.72 99.72
H 0.277 0 100.00 100.00
I 0.176 0 99.74 99.74
J 0.314 93.91 93.91 94.91
L 0.145 0 100.00 100.00
Total 100 26.25 67.87 67.87
416 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
Table 7
Summary results from A and J sub-watersheds when F = 1.
Sub-watershed C1
(R$ Million)
CASE 0 - Without changes
A (85.49% permeable area) 58.24
J (7.6% permeable area) 98.85
Total (44.51% permeable area) 706.98
CASE I – No environmental flow regulation value and no water detention structures
A (0% permeable area) 401.36
J (0% permeable area) 106.98
Total (0% permeable area) 1274.27
CASE II – With environmental flow regulation value and no water detention structures
A (65.66% permeable area) 137.81
J (100% permeable area) 0.00
Total (89.07% permeable area) 139.33
CASE III – With environmental flow regulation value and with water detention structures
A (0% permeable area) 401.36
J (93.9% permeable area) 6.51
Total (67.87% permeable area) 409.17
The effect of each case varies according to the sub-watershed
characteristics. For example, for sub-watershed A, (SVL = BLU)
and CASE II (With environmental flow regulation value and no
water detention structures) the reduction in the peak flow is
higher than for CASE III (With environmental flow regulation value
and with water detention structures) across all different hydrologic
scenarios, except for the most severe one (R8). For sub-watershed J,
CASE II resulted in higher peak flows for most of the scenarios (R3
and above). This can be explained by the fact that, for larger sub-
watersheds, the larger amount of runoff produced would prove
very expensive to be controlled through vegetated areas alone.
Once detention structures are present (CASE III), the vegetated area
can be reduced. While this may result in increased peak flows, it is
still possible to maintain it under the drainage capacity for most
hydrologic scenarios. This result corroborates findings in Paule-
Mercado et al. (2017) who verified the impact of impervious cover
expansion on stormwater runoff and also that best management
practices (BMP) contributed to reduce the runoff volume. However,
the work presented here also discusses the potential cost reduction
when structural interventions (i.e. detention tanks) are present,
while Paule-Mercado et al. (2017) focus on the simulation of the
resulting runoff.
The peak flow increase is the trade off from the economic gains
possible when vegetated areas are converted to buildings. For
small sub-watersheds, the presence of detention structures allows
some reduction in the permeable area with a reduced trade off in
the peak flows (in our example, the peak flows for a small sub-
watershed such as J are actually smaller, thanks to the detention
structure, regardless of its permeable area being reduced).
Such difference in the results indicates how the value of flow
regulation and the effect of structural solutions vary throughout
the watershed, indicating the necessity of investigating individual
solutions to cope with local flooding. Indeed, Lopefido et al. (2014),
when calculating the effects of distributed and centralized
stormwater best management practices and land cover, verified
that the spatial distribution is important to stormwater manage-
ment, and pointed out that distributed management practices
improved stream hydrology if compared to centralized practices.
The results in our paper show that stormwater management prac-
tices (here represented by water detention structures) allowed
more intense land use without the runoff penalty in smaller sub-
watersheds, which indicates that distributed individual solutions
might produce better combined results (although this is not veri-
fied here). The difference from our work to Lopefido et al. (2014)
C2 Detention structures cost (R$ Million) C1 – C2 – C3
(R$ Million) (R$ Million)
– – 58.24
– – 98.85
– – 706.98
– – 401.36
– – 106.98
– – 1274.27
95.32 0.00 42.49
0.00 0.00 0.00
96.84 0.00 42.49
224.62 0.68 176.74
0.057 1.50 4.95
226.04 2.18 180.95
is that we have calculated the economic value associated to the
trade offs at each solution.
While sub-watershed A is not facing local flooding under cur-
rent conditions (CASE 0) it would become severely vulnerable if
nothing is done and urban development follows its past course
(expanding impervious areas with no stormwater control at the
source). This would result in 100% impervious area for sub-
watershed A (CASE I) and peak flows exceeding the current
capacity very frequently (Fig. 8). If the environmental service lost
is valued however, the exceeding runoff can be internalized and
the peak flows reduced to most of the hydrologic scenarios (CASE
II and CASE III), with CASE III providing the best results and the
highest economic benefit, as it will be shown in the next section.
However, depending on the value placed on the environmental
flow service, the peak flow may not be reduced enough to prevent
local flooding. Fig. 9 presents results for sub-watershed A indicat-
ing the potential peak flow reduction under CASE II for higher F
values. The 20–30 m3/s range shaded area is entirely above the
local drainage capacity in sub-catchment A, indicating that local
flooding is likely to be frequent if the value placed on the environ-
mental flow service is low (F = 0.8 in this case).
It is important to note that the drainage capacity adopted here
is an exogenous variable, based on investments already made. If
planning is being made for a new urban area not yet developed,
the drainage capacity can be evaluated against the costs to main-
tain the flow regulation service, as well as the trade offs involved.
Fig. 10 summarizes the sensitivity analysis for CASE II and CASE
III under different reference values for the flow regulation environ-
mental service and the consequence to land use decisions in the
entire study area. If society values the flow regulation service less
than the economic return of land use (F = 0.8) the permeable area
left covers about one third of the total watershed and detention
structures would bring no additional benefit.
If society values the flow regulation service the same as the eco-
nomic return of land use (F = 1) the permeable area left would be
significantly higher, as the C2 cost has a stronger effect in the
objective function and losing the flow regulation service would
negate the economic returns in the first stage, C1. At this point,
detention structures become significantly more interesting eco-
nomically, providing valuable flow regulation and allowing more
land to be used commercially. Beyond this point (at F = 1.2),
increasing the value of the flow regulation service above the eco-
nomic return of land use did not result in significant changes in
the land use decisions, which remained the same for CASE III and
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
Fig. 9. Sub-catchment A under CASE II.
Fig. 10. Sensitivity analysis for CASE II and CASE III under different reference values
for the flow regulation environmental service: below (0.8), equal (1.0) and above
(1.2) the economic return of land use.
with a slight increase in the permeable area in CASE II. This indi-
cates that there’s likely an effective value to the flow regulation
service that motivate users’ decisions to maintain it, and it does
not necessarily need to be higher than the economic return of land
use.
Thus, by valuing the environmental service, even if it still less
than the economic return of land use, urban planners would moti-
vate significant land use decisions to conserve the flow regulation
service, if compared to CASE I (which had permeable area equals
0). The model results can also help in the design of urban water
management instruments that signal the value of environmental
services to the users. Urban planners and regulators could compare
different land use results (e.g. distribution of permeable and non-
permeable areas) and identify the corresponding value that needs
to be attached to the environmental service of flow regulation in
order to drive land use decisions to a desired policy of environmen-
tal conservation. By adjusting the parameter F, the value placed on
the environmental service changes, and so does the land use. The
value placed on the environmental service will also determine
the loss C2 in the model (i.e. the higher the environmental value,
the higher the C2). The C2 can be later distributed among the users
with economic land management instruments (e.g. runoff tariffs).
The parameter F could then be adjusted to produce the desired
distribution.
417
Adopting the distribution of permeable and non-permeable
areas to represent a land use policy leaves out many details and
complexities and it is a limitation in our analysis. However, this
distribution captures the main driver of urban runoff and localized
flooding. The model approach and results are thus useful to point
out a broader plan, which can be later refined. Besides, the same
modeling approach and concept can be applied extending the
number of intermediate levels between permeable and non-
permeable, as well as LID techniques.
3.1. Economic values
The reduction of peak flows to mitigate urban flooding requires
some of the impact caused by individual decisions in urban occu-
pation to be internalized. Here we present the economic value
resulting from a combination of the urban land use for buildings
and for flood mitigation. Between CASE II and CASE III, the higher
the TEV, the better the result, meaning that a better balance was
found in terms of economic land use and environmental services
loss. A comparison of the TEV of CASES II and III with the TEV of
CASE I is not useful given that the later does not calculate the flow
regulation service value loss C2 (the purpose of CASE I is to illus-
trate the maximum potential economic benefit when the environ-
mental service is disregarded).
For CASE I, the absence of the environmental flow regulation
service value results in the full occupation of the watershed
(100% impermeable area) and maximum TEV value (R$1274 Million,
which is only the C1 value given C2 is not calculated). However, this
scenario also results in externalities as flooding in sub-watersheds
A, J and at the exutory of the study area. This is a common condition
in many urban centers worldwide where the value of the environ-
mental service of flow regulation is not considered.
When the environmental service of flow regulation is valued in
CASE II, the permeable area increases from 0% to up to 89.07% to
partially recover the flow regulation service and reduce the peak
flows down to the capacity of the current infrastructure, eliminat-
ing all the flooding present in CASE I. Such increase in the perme-
able area reduces the C1 down to R$139.33 Million. The remaining
impermeable area still produces a loss in the environmental ser-
vice of flow regulation of R$96.84 Million (C2) resulting in a net
expected TEV equals to R$42.49 Million. The high C1 value reflects
the opportunity cost of the urban land.
When detention storage is added in CASE III it effectively pro-
vides additional flow regulation service to partially replace the
one present in the vegetated areas, allowing the permeable area
to be reduced from the 84.07% in CASE II to 67.87% in CASE III. This
result brings two important consequences, economical and
hydrological:
(a) It increases the economic value of land use C1 to R
$409.17 Million. For example, sub-watershed A has the
occupied area (impermeable land) increased from 34.34%
to 100%.
(b) Not all additional runoff produced is captured by the deten-
tion storage, resulting in higher peak flows and higher loss of
the environmental service of flow regulation C2, now up to R
$226.04 Million.
However, comparing CASE III to CASE II, the increase in the eco-
nomic value of land use C1 is higher than the increase in the loss of
environmental service of flow regulation C2. The result is a TEV
value significantly higher in CASE III (R$180.95 Million, up from
R$ 42.49 Million in CASE II). This indicates a better balance
between urban economic land use, vegetated areas and structural
solutions.
418 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
Finally, given the value of the environmental service of flow
regulation in urban areas may be very high, depending on the
opportunity cost of the land and society preference, a combination
with structural measures is important to avoid overly constrained
urban occupation. However, as pointed out in Kaltenborn et al.
(2017), above a certain level of providing essential public services,
simplicity in life and availability of resources may be preferred
over other (economic) opportunities and services. This indicates
that urban dwellers will be likely worse off by fully replacing the
environmental services with economic benefits from land use,
and that some intermediate compromise is necessary. Depending
on the vegetated area maintained in the watershed, which varies
according to the F factor used, overall environmental quality will
be preserved.
3.2. Limitations
Given that our approach addresses broad land use planning and
drainage, some limitations arise from necessary simplifications,
although it must be highlighted that the main drivers to users’
decisions are currently represented and can be later refined (with
more land use alternatives other than permeable and non-
permeable and presence of LID structures).
In addition to the simplification on the land use decisions mod-
eled here, there is some level of uncertainty in the permeable area
results produced, which are an upper bound. The inclusion of low
impact development (LID) solutions, for example, may enable the
permeable area to be reduced while still producing economic ben-
efits and mitigating the loss of the environmental service of flow
regulation.
Other limitations to this paper include (a) the relationship
between land use and QP and (b) the relationship between SVL
and QP. The adoption of a quadratic optimization approach
requires well-defined equations for these relationships, which
demands simplification to the real processes. A more diverse land
use decision, beyond permeable vs. impermeable land as modeled
here is likely to result in a more complex land use vs. QP relation-
ship. Finally, the true form of the SVL vs. QP relationship depends
on factors including the spatial distribution of the economic value
of the land across the watershed, which are not modeled here.
4. Conclusions
Land use decisions in urban areas affect the environmental ser-
vice of flow regulation and produce externalities in the form of
excess runoff and flooding. Based on the results presented, we con-
clude that the value preference placed on the environmental ser-
vice of flow regulation can drive land use decisions to maintain
such service and bring down peak flows to desired limits, and that
there is an upper bound for this value, beyond which land use deci-
sions are not significantly affected. The modeling framework pre-
sented in this paper can be useful to evaluate the reference value
placed on the environmental service to produce desired land use
policies regarding permeable and non-permeable areas, which
would serve as a basis to develop economic instruments for
managing urban stormwater (e.g. payment for environmental ser-
vices, drainage tariffs or a ‘‘flooder-payer” tariff).
The application of tools such as the one presented here may be
especially useful in new urban developments, to support discus-
sion on the level of occupation and its future consequences, both
in economic and hydrologic terms, including drainage system
costs, forgone economic land use and flood risks. Most important,
the model results provide a clear cause-effect relationship between
individual land use decisions and impacts on neighboring society,
which contributes to raise awareness about user’s decisions.
Another benefit of the solutions discussed here is the cumula-
tive effect of peak flow reduction on the main streams of the urban
region. While the local conveyance capacity could always be
increased to cope with higher flows, their combined effect further
downstream would eventually lead to higher magnitude flooding,
which is also significantly more expensive to manage with hydrau-
lic structures alone. Instead, if local, distributed solutions involving
land use decisions that recover the environmental service of flow
regulation are implemented throughout the watershed, the latter
can be avoided.
Finally, Belo Horizonte’s city need improved resilience to face
natural hazards (OECD, 2016). If nothing is done, economic returns
to land use will likely drive an occupation pattern that may make
some of its regions (such as the one modeled here) even more vul-
nerable to intense hydrologic scenarios. It is urgent that new urban
and drainage planning favor new approaches other than increasing
infrastructure capacity. The results shown here indicate this is
possible.
Acknowledgment
The authors thank the CAPES, IPH and CEFET-MG for supporting
the development of this work.
References
ANTENER, M., 1999. La revalorisation des milieux naturels en sites périurbains et la
gestion de risque d’inondation para des actions à l’échelle privative – l’exemple
de la ville de Zurich. Journée d’Information Aménagement et Eaux Pluviales, 24–
35.
Arico, S., Bridgewater, P., El-beltagy, A., Harms, E., Program, S., Hepworth, R.,
Watson, R.T., et al., 2005. Ecosystems and Human Well-Being. Millennium
Ecosystem Assessment.
Barbosa, L.D., Fonseca, D.F., 2006. Diagnóstico do risco ambiental da bacia do
Córrego do Cercadinho, Belo Horizonte/MG: co-relação de seus aspectos físicos
e demográficos. Artigo de projeto de iniciação apresentado em um simpósio na
UFV, Centro Universitário de Belo Horizonte. Disponível em: <http://www.geo.
ufv.br/simposio/simposio/trabalhos/> (Acesso em: fev. 2013).
BARRAQUÉ, B., 1995. Les politiques del’eau em Europe. la Decouverte, Paris: Ed. 303.
Brown, R.R., Keath, N., Wong, T.H.F., 2008. Transitioning to water sensitive cities. In:
11th International Conference on Urban Drainage, Edinburgh, Scotland, UK.
Burns, M.J., Fletcher, T.D., Walsh, C.J., Ladson, A.R., Hatt, B.E., 2012. Hydrologic
shortcomings of conventional urban stormwater management and
opportunities for reform. Landscape Urban Plann. 105(3), pp. 230–240. ISSN
0169-2046. Disponível em: <http://www.sciencedirect.com/> (Acesso em: fev.
2013).
Canholi, A.P., 2005. Drenagem Urbana e Controle de Enchentes. Oficina de Textos,
São Paulo, 302p.
Cembrano, G., Quevedo, J., Salamero, M., Puig, V., Figureueras, J., Martí, J., 2013.
Optimal control of urban drainage systems. A case study. Control Eng. Practice
12(1), pp. 1–9. ISSN 0967-0661. Disponível em: <http://www.scopus.com/>
(Acesso em: fev. 2013).
Gironás, J., Roesner, L.A., Rossman, L.A., Davis, J., 2010. A new applications manual
for the storm water management Model (SWMM). Environ. Modell. Software 25
(6), 813–814.
Hamel, P., Daly, E., Fletcher, T., 2013. Source-control stormwater management for
mitigating the impacts of urbanisation on baseflow: A review. J. Hydrol. 485, pp.
201–211. Disponível em: <http://www.sciencedirect.com/> (Acesso em: jul.
2014).
Hardim, G., 1968. The tragedy of the commons. Science 162, pp. 1243–1248.
Disponível em: <http://cecs.wright.edu/~swang/cs409/Hardin.pdf> (Acesso em:
nov. 2014).
Jang, S., Cho, M., Yoon, J., Yoon, Y., Kim, S., Kim, G., Kim, L., Ksoy, H., 2007. Using
SWMM as a tool for hydrologic impact assessment. Desalination 212, 344–356.
Kaltenborn, B.P., Linnell, J.D.C., Baggethun, E.G., Lindhjem, H., Thomassen, J., Chan, K.
M., 2017. Ecosystem services and cultural values as building blocks for ‘the good
life’. a case study in the community of røst, Lofoten Islands, Norway. Ecol. Econ.
140, 166–176.
Lopefido, J.V., Noe, G.B., Jarnagin, S.T., Hogan, D.M., 2014. Effects of distributed and
centralized stormwater best management practices and land cover on urban
stream hydrology at the catchment scale. J. Hydrol. 519, 2584–2595.
Loperfido, J.V., Noe, G.B., Jarnagin, S.T., Hogan, D.M., 2014. Effects of distributed and
centralized stormwater best management practices and land cover on urban
strem hydrology at the catchment scale, 519, pp. 2584–2595. Disponível em:
<http://www.sciencedirect.com/> (Acesso em: jun. 2017).
Marques, G.F., Lund, J.R., Howitt, R.E., 2005. Modeling irrigated agricultural
production and water use decisions under water supply uncertainty. Water
Resour. Res., W08423.
 G.F. Marques et al. / Journal of Hydrology 554 (2017) 406–419
McCARL, B.A., Meeraus, A., Eijk, P., Bussieck, M., Dirkse, S., 2016. General Algebraic
Modeling System (GAMS). GAMS Development Corporation. February 28, 2016
McNeill, J., Dollery, B., 2007. Calculating developer charges for urban infrastructure:
a feasible method for applying marginal cost pricing. Eng. Econ. 48 (3), 218–
240.
MOURA, P. M., Contribuição para a Avaliação Global de Sistemas de Drenagem
Urbana, Dissertação (Mestrado em Saneamento, Meio Ambiente e Recursos
Hídricos, Belo Horizonte, Minas Gerais, 2004.
OECD. Resilient Cities, 2016. Disponível em: <http://www.oecd.org/cfe/regional-
policy/resilient-cities-report-preliminary-version.pdf> (Acesso em: mai. 2017).
Ossa-Moreno, J., Smith, k.M., Mijic, A., 2017. Economic analysis of wider benefits to
facilitate SuDS uptake in London, UK. Sustainable Cities Soc. 28, 411–419.
Paule-Mercado, M.A., Lee, B.Y., Memon, S.A., Umer, S.R., Salim, I., Lee, C.H., 2017.
Influence of land development on stormwater runoff from a mixed use and land
419
cover catchment. Sci. Total Environ. 599–600, pp. 2142–2155. Disponível em
<http://www.sciencedirect.com/> (Acesso em: jun. 2017).
Pinheiro, M., Naghettini, M., 1998. Análise regional de frequência e distribuição
temporal das tempestades na região metropolitana de Belo Horizonte - RMBH.
Revista Brasileira de Recursos Hídricos 3 (4), 73–88.
Sample, D.J., Heaney, J.P., Wright, L.T., Fan, C.-Y., 2003. Costs of best management
practices and associated land for urban stormwater control. J. Water Resour.
Plann. Manage. 129 (1).
Trem Util, 2013. Custo do m2 em Belo Horizonte. 2013. Disponível em: <http://
www.tremutil.com.br/bhc-metroquadrado.php> (Acesso em: abril. 2013).
Tucci, C.E.M., 2008. Águas urbanas - Urban waters. Estudos Avançados, pp. 97–112.
Disponível em: <http://www.sciencedirect.com/> (Acesso em: fev. 2013).
Wilchfort, O., Lund, J.R., 1997. Shortage management modeling for urban water
supply systems. J. Water Resour. Plann. Manage. 123 (4), 250–258.