Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Journal of Hydrology 596 (2021) 126130

Contents lists available at ScienceDirect

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

Review papers

Modelling of sewer exfiltration to groundwater in urban wastewater


systems: A critical review
Hong Hanh Nguyen a, *, Aaron Peche b, c, Markus Venohr a
a
Department of Ecohydrology, Leibniz-Institute of Freshwater Ecology and Inland Fisheries, 12489 Berlin, Germany
b
Federal Institute for Geosciences and Natural Resources, 30655 Hannover, Germany
bpi Hannover, Mengendamm 16D, 30177 Hannover, Germany

A R T I C L E I N F O A B S T R A C T

This manuscript was handled by Corrado Cor­ Increasing evidence from studies in urban wastewater systems proves that sewer exfiltration may present a
radini, Editor-in-Chief, with the assistance of profound threat for rising levels of toxic substances and microbial pollution in groundwater. New and innovative
Wei Hu, Associate Editor methods for monitoring sewage exfiltration are continuously developed, which provide a solicit ground for a
better understanding of key processes and influencing characteristics of sewer exfiltration to support the
Keywords:
development of mathematical equations and models worldwide. However, modelling of sewer exfiltration and its
Exfiltration rate
associated impacts on groundwater are still constrained and not fully understood in large urban systems. This
Leakage factor
Standard leakage area paper reviews the knowledge advancements and challenges of sewer exfiltration modelling. Based on the present
Colmation layer review, we define five key research domains for developing more generic framework of large-scale models. These
Private sewers domains inquire to advance the representation of hydraulic modelling (e.g. testing exfiltration models coupled
with biofilm growth models), modelling of solutes originating from sewer leakages (e.g. heavy metals), define the
practical values of key model parameters of main model components and sub-processes for calibration and
uncertainty evaluations, develop upscaling approaches for integrated multi-dimensional sewer exfiltration
models, and incorporation of sewer exfiltration from private sewers.

1. Introduction (CCTV) is so far the most common method to evaluate the structural
integrity and defects of sewers and has been used in major cities
Sewer exfiltration is the process of wastewater losses from a defective worldwide (Amick and Burgess, 2000; Ellis and Bertrand-Krajewski,
sewer system to the environment (Rutsch et al., 2008). Sewer exfiltra­ 2010). Techniques using sewer tracers and wastewater indicators (e.
tion is often associated with soil undermining and bedrock dissolution, g., sodium chloride, fluorescent dyes) are increasingly applied since
which may lead to sinkhole development, urban infrastructure damages, these methods can distinguish substances from sewer exfiltration with
and further expansions of exfiltration areas (Selvakumar et al., 2004; pollutants from other sources entering groundwater (e.g., Ellis and
Hermosilla, 2012; Harvey and McBean, 2014). An increasing number of Revitt, 2001; Fenz et al., 2005; Rieckermann et al., 2007, 2010; Lee
investigations provide solicited evidence that sewer exfiltration can be et al., 2015; Roehrdanz et al., 2017; Stegeman et al., 2019). Other
the source of significant pollutions to groundwater (e.g., Reynolds and techniques for in-situ measurements of sewer exfiltration include
Barrett, 2003; Lee et al., 2015; Grimmeisen et al., 2017). Thus, reliable Distributed Temperature Sensing (DTS) (e.g., Nienhuis et al., 2013),
quantification of sewer exfiltration plays an important role in supporting Electric Resistivity Tomography (ERT) (e.g., Thompson et al., 2012),
sewer network rehabilitation and management decision of urban areas Ground Penetrating Radar (GRP) (Hao et al., 2012), using robots for
vulnerable to sewer exfiltration contamination (Ellis and Bertrand- pipe inspections (e.g., Kirkham et al., 2000), impact-echo method (e.g.,
Krajewski, 2010; Karpf, 2012; Dvory et al., 2018). Kang et al., 2017), or ring laser scanning (e.g., Clemens et al., 2015).
Traditionally, sewer exfiltration is measured using various moni­ Although a variety of instruments are available to monitor and detect
toring approaches, which serve as foundation for the classification of leaks, they are often labor-intensive and expensive to apply on the whole
sewer defects and exfiltration (Chisala and Lerner, 2008; Harvey and sewer network, thus can hardly represent large spatial pattern or be
McBean, 2014; Tscheikner-Gratl et al., 2020). Closed-Circuit Television applied at higher frequencies (Blackwood et al., 2005, 2006; Wolf and

* Corresponding author.
E-mail address: hanh.nguyen@igb-berlin.de (H.H. Nguyen).

https://doi.org/10.1016/j.jhydrol.2021.126130
Received 28 October 2020; Received in revised form 21 February 2021; Accepted 23 February 2021
Available online 25 February 2021
0022-1694/© 2021 Elsevier B.V. All rights reserved.
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

Hötztl, 2007). Additional to field observation techniques, there is a need 2. An overview of sewer exfiltration processes
to develop modeling tools to quantify emissions from sewer exfiltration
source to groundwater and to predict the structural state of sewers for Sewer exfiltration releases pollutants to the environment, which may
management decisions of urban wastewater systems (Yang et al., 1999; exceed emissions at background conditions and from other sources
Karpf et al., 2009; Roehrdanz et al., 2017). (Wolf et al., 2012; Lee et al., 2015; Roehrdanz et al., 2017) (Fig. 2).
Sewer exfiltration models have been developed to help identify Untreated wastewater from sewer exfiltration has been observed to
target areas with the highest leakage rate potentials at various temporal contain high levels of suspended solids, organic pollutants, and toxic
and spatial scales (Yang et al., 1999; Karpf et al., 2009; Peche et al., substances (USEPA, 1989; Reynolds and Barrett, 2003; Lee et al., 2015).
2017). The general principles and approaches for sewer exfiltration Pathogenic microorganisms in sewer exfiltration may potentially impair
models emerged more than a decade ago (Rutsch et al., 2008; Karpf the recreational values of surrounding surface waters, cause risks to
et al., 2009; Peche et al., 2017). However, model development progress public health, groundwater pollution and serious economic damages
is still slow, which is indicated by a limited number of published studies (Burston et al., 1993; Blackwood et al., 2005; Sercu et al., 2011; Kar­
on sewer exfiltration models when compared to models for other envi­ amouz et al., 2011; Harvey and McBean, 2014). However, emissions to
ronmental applications (e.g., Zoppou, 2001; Fu et al., 2019; Tscheikner- the environment from urban sewer leakage often perceive less attention
Gratl et al., 2020). Integrated 1D-3D models were developed to simulate than those from agricultural areas and are even neglected in studies of
sewer exfiltration dynamics but are hard to validate due to the large urban system emissions, particularly at large-scale studies (e.g., Eiswirth
number of data required for model parametrization (Rutsch et al., 2008; and Hötzl, 1997; Burn et al., 2007; Peche et al., 2017).
Peche et al., 2017; Tscheikner-Gratl et al., 2020). Karpf (2012) and Sewer exfiltration can be seen as a conditional process (Davies et al.,
Peche (2019) proposed simplified approaches for modeling sewer 2001; Rutsch, 2006; Karpf, 2012). The first condition is the existence of
exfiltration, but the methods have been tested only to site-specific and a defect in the sewer, either due to aging and deterioration of materials,
pipe scales. Wolf et al. (2012) considered the temporal and spatial poor construction of sewer pipes, physical stress (e.g., due to surface
variation of sewer exfiltration to groundwater by conducting sampling loads or plant root intrusion), or the effect from the surrounding soil
for the period from 2002 to 2008 in the city of Rastatt. However, no bedding (e.g., DeSilva et al., 2005; Chisala and Lerner, 2008). Another
significant findings on lateral and horizontal fluxes of sewer exfiltration prerequisite condition for the occurrence of sewer exfiltration processes
were achieved from the limited number of wastewater indicators. Lee is the partly or completely location of defective pipes above the
et al. (2015) and Roehrdanz et al. (2017) developed a spatial model of groundwater level (Ellis et al., 2009; Boukhemacha et al., 2015). Finally,
sewer exfiltration, which allows predicting the occurrence of exfiltration the hydraulic potential in the sewer pipe has to be higher than the hy­
in shallow groundwater and incorporates various wastewater indicators, draulic potential in the surrounding soil, such that gravity-induced
but the outputs of the model are subjected to the groundwater short- leakage flow can occur (Peche et al., 2017, 2019). The difference in
term observations in a small region of a city. Besides, the uncertainty hydraulic heads in the sewer and groundwater governs the direction and
of upscalling approaches in sewer exfiltration modelling remains sig­ the rate of sewer leakage (Selvakumar et al., 2004; Roehrdanz et al.,
nificant issue due to a lack of understanding of core parameters and 2017).
processes of sewer exfiltration (Wolf and Hötztl, 2007; Karpf, 2012; The process of sewer exfiltration from damaged pipes into urban
Peche, 2019). So far, the studies by Rutsch et al. (2005) and Rutsch et al. groundwater does not remain constant over time (Ellis et al., 2009;
(2008) remain the only reviews that incorporate modelling aspects of Boukhemacha et al., 2015; Peche et al., 2017). At first, the exfiltration
sewer exfiltration (Table 1), though many changes have occurred since rate is low when the soil environment is dry, i.e., the hydraulic potential
then. of the soil is small. The exfiltration rate into the soil increases with
In this study, we review and evaluate the current state of sewer increasing soil moisture (corresponding to increasing relative perme­
exfiltration models, with a particular focus on hydrological aspects of ability). At large hydraulic gradients (between the pipe and the soil in
model developments and applications. The ultimate aim is to review the the pipe vicinity) and soil moisture, increased sewer exfiltration leads to
up-to-date knowledge and understanding of the whole mechanism of downwards propagation of a wetting front and sewer outflow may
wastewater exfiltration from defective sewers to groundwater and recharge groundwater. Sewer exfiltration is governed by the presence of
define critical research gaps to reduce uncertainty in decision-making a colmation layer forming in the pipe defect vicinity (Karpf, 2012;
relating to outputs from sewer exfiltration models. The term ‘exfiltra­ Peche, 2019). After the occurrence of a pipe defect, high sewer exfil­
tion’ and ‘leakage’ are used interchangeably in this field, though the tration rates continue until an equilibrium pressure height between in-
latter term often includes both the infiltration of groundwater into sewer sewer and soil conditions is reached. These high rates often remain
pipes and the exfiltration of sewage into the environment, as well as over some hours or days (largely depending on soil properties, such as
diverse types of outflows through defective canals such as drinking initial soil moisture and soil hydraulic properties) before gradually
water delivery (Karpf, 2012; Hatam et al., 2019). In the scope of this decreasing (Karpf and Krebs, 2004; Boukhemacha et al., 2015). The
review, most of our discussions are targeted explicitly on the exfiltration reduction in exfiltration rates is due to the stabilization of the colmation
from wastewater sewers due to its impacts on groundwater. The overall layer, which causes a change of hydraulic properties (in the form of a
scheme of the review is presented in Fig. 1. reduction of porosity and intrinsic permeability) of the soil in the pipe
vicinity. This reduction continues until a mature colmation layer is
developed (Ellis et al., 2009; Boukhemacha et al., 2015).
Table 1 A kinematic wave pulse, i.e. a breakthrough, can be initiated by
Papers reviewing topics relating to sewer exfiltration modelling. diurnal peak flows or during storm events (Klinger et al., 2007). This is
Study focus References due to a temporary local increase in flow velocity, turbulence, shear
Overview of sewer exfiltration Rutsch et al. (2005), Rutsch et al. (2008) stress, changing hydraulic gradients, and accompanied by instability of
Sewer exfiltration rates Amick and Burgess (2000), Rutsch et al. colmation layers, leading to a breaking-up of the colmation layer (Ellis
(2008), Ellis et al. (2009) et al., 2009). As a result, a steady-state exfiltration condition is disturbed
Colmation of sewer exfiltration Rauch and Stegner (1994); Vollertsen and by a short-term increase of exfiltration a few orders of magnitudes above
Hvitved-Jacobsen (2003); Ellis et al. (2009);
Karpf (2012)
the equilibrium of the dry-weather flow (Wolf et al., 2007). Such ‘wash-
Impacts of sewer exfiltration on Reynolds and Barrett (2003) out’ events followed by the establishment of steady-state conditions
groundwater have been observed persistently in various studies, i.e., both in labora­
Sewer exfiltration as part of urban Lerner (2002); Benedetti et al. (2013); tory and field tests, but the effects are local and only remain over a short
drainage and wastewater systems Tscheikner-Gratl et al. (2020)
time (Vollertsen and Hvitved-Jacobsen, 2003, Wolf et al., 2007). Thus, it

2
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

Fig. 1. The general structure of the review on sewer exfiltration modelling.

3. Key influencing variables of sewer exfiltration processes

Stemming from literature, influential primary factors that affect


sewer exfiltration modelling include pipe characteristics and defective
conditions, wastewater level and composition in the pipe, soil proper­
ties, hydraulic potential (as expressed by water pressure and soil mois­
ture) in the soil in the pipe vicinity, and colmation layers (Davies et al.,
2001; Ellis et al., 2009; Harvey and McBean, 2014; Roehrdanz et al.,
2017). These driving factors and their associated attributes and (or)
important derivatives are therefore reviewed in detail in the following
section (Fig. 3) to explore the progress in understanding the key pa­
rameters and their role in sewer exfiltration modelling.

3.1. Sewer pipe characteristics

Pipe characteristics and conditions are reported to significantly


Fig. 2. Example of contaminants from leaking sewers to groundwater in
affect the sewer exfiltration rates among various locations in the sewer
comparison to other pollution sources (adapted from Eiswirth et al., 2001; Burn
et al., 2007). *VOCs: Volatile Organic Compounds
network (Dohmann, 1995; Ellis and Bertrand-Krajewski, 2010; Lee et al.,
2015). Major pipe characteristics reported in sewer exfiltration studies
include age (e.g., Kettler and Goulter, 1985), diameter (DeSilva et al.,
is acknowledged that sewer defects are never completely blocked and
2005), length (Le Gat and Eisenbeis, 2000), types of pipe junctions
are subject to certain degrees of exfiltration when evaluating over
(Berardi et al., 2008), and material (e.g., Harvey and McBean, 2014).
extended time periods (Reynolds and Barrett, 2003; Ellis et al., 2009).
Aging sewer systems are the most primary cause of sewer exfiltration in

3
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

Fig. 3. Graphical illustration of pipe sewer exfiltration (adapted from Ellis et al., 2009, Boukhemacha et al., 2015, and Peche et al., 2017): hS = hydraulic head in the
sewer, Hg = hydraulic head in the surrounding aquifer, kcolmationlayer = hydraulic conductivity of colmation layer, ksoil = hydraulic conductivity of the soil.

soil and groundwater worldwide (Amick and Burgess, 2000; Reynolds defect area’ or ‘representative elementary area’, refers to the mean
and Barrett, 2003; Ellis and Bertrand-Krajewski, 2010). Based on a defect size per unit pipe length (given as the defect area per pipe length
literature review on the status of sewers and their potential exfiltration in m2/m) in a given pipe network and represents the unique charac­
rate, Chisala and Lerner (2008) concluded that exfiltration risks increase teristics and defective condition of the entire pipe network within the
with the sewer system age and decrease in more modern and younger study boundary. In a study by Rauch and Stegner (1994), the rectangular
sewers. The higher durability of newer pipe infrastructure is achieved by hole (0.0015 m2/m) was used as the standard leak size for all experi­
more inert and stable materials and connections (Reynolds and Barrett, ments. Wolf and Hötztl (2007) suggested an average size of a defective
2003; Gibson et al., 2019). Regarding the conditions of sewer pipes, sewer around 0.002 m2/m based on their literature review and consid­
typical reasons causing sewer exfiltration include defects at lateral ered the estimated average defect size in their particular case study
connections, broken joints, transversal and longitudinal cracks, frac­ (0.008 m2/m) to be high because the study was focused on critical de­
tures, and holes (Decker, 1994). The rate of exfiltration also depends fects with large size. Similar to Rauch and Stegner (1994), Karpf (2012)
considerably on the defect types. For instance, Vollertsen and Hvitved- proposes the application of a standard quadratic size of a defective pipe
Jacobsen (2003) reported three times higher exfiltration rates for in sewer exfiltration calculations, and these values have been adopted
cracks than for open joints and holes, i.e., 0.02 L/d.cm2 and 0.06 L/d. successfully in sewer pipe models proposed by Peche et al. (2017) and
cm2, respectively. Peche et al. (2019).

3.2. Standard defect area 3.3. Colmation layer

Many studies suggest that the exfiltration rate is positively correlated Sewer exfiltration is known as a ‘self-sealing’ (or ‘self-repairing’)
with the pipe defect area, i.e., the exfiltration-relevant damaged area (e. process due to the effect of a colmation layer forming (Blackwood et al.,
g., Davies et al., 2001; Howard, 2007). Meanwhile, other studies 2006; Karpf et al., 2011). A colmation layer is defined as a layer
conclude that this effect might be overestimated since large defects developed of solid inputs from wastewater, biomass growth and chem­
allow more sealing material in the sewage to reach the gravel and thus ical precipitation and is characterized by fast declining hydraulic con­
reduce the rate of sewer exfiltration (e.g., Blackwood et al., 2006). Thus, ductivity compared to that in the surrounding underground (Rauch and
the term ‘standard defect area’ is often used to represent the status of Stegner, 1994; Ellis et al., 2009). Meanwhile, any overlying biofilm
defective sewers in sewer exfiltration models (Rutsch et al., 2008). It is should be referred to as a bridging layer because the formation of this
based on the assumption that similar pipe characteristics and defective external layer is assumed to not affect the exfiltration rate through a
conditions induce similar exfiltration rates (Franz and Krebs, 2005; defect (Ellis et al., 2009; Karpf et al., 2011). For modelling purposes,
Karpf, 2012). The term ‘standard defect area’, also named as ‘average colmation layers are often represented by the two constant properties:

4
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

colmation layer thickness and intrinsic permeability (expressed as hy­ complexities (Rutsch et al., 2008; Karpf, 2012). Only a few studies exist
draulic conductivity for simplification). Accordingly, the colmation where the non-linearity of the exfiltration process is considered (e.g.,
layer is modelled as a thin layer with a thickness varying between 10 and Kidmose et al., 2015; Peche et al., 2017).
20 mm possessing a relatively low permeability (Vollertsen and Hvitved-
Jacobsen, 2003; Karpf et al., 2009; Peche et al., 2017). Hydraulic con­ 3.6. Exfiltration rate
ductivity decreases with time until a constant value is reached. This
parameter is assumed to represent a mature and stabilized colmation Sewer exfiltration rate is the primary indicator addressed in sewer
layer. Up to the thickness of 50 mm, the colmation layer is stabilized and exfiltration studies to quantitatively estimate the losses from defective
hinders the process of exfiltration from defective pipes (Ellis and sewers. Numerous studies focusing on sewer exfiltration rates are re­
Bertrand-Krajewski, 2010). ported in the literature, including reviews by Gruenfeld (2000), Rutsch
et al. (2008), and Ellis et al. (2009). According to these studies, a broad
3.4. Soil matrix variability of values and units of reported exfiltration rates are recorded
from field and laboratory investigations. This variability depends on the
One of the main difficulties in estimating sewer exfiltration is asso­ measurement techniques, the scale and focus of studies, and the
ciated with the complex physico-chemical interactions between waste­ particular conditions at the study sites. As such, it is recommended to
water in the exfiltration and soil medium, which involves the chemical develop a standardization of the found quantification approaches for
composition in the soil (e.g., pH, sulfates, chlorides), soil moisture, hy­ modelling applications (Rutsch et al., 2008; Peche et al., 2017). Corre­
draulic conductivity, porosity and aeration levels (Ellis et al., 2003; sponding to results from laboratory and field measurements, modelling
Nickpay et al., 2015). Studies at the laboratory to pipe scales indicate a studies provided relatively consistent estimates of exfiltration rates,
wide range of removal of Dissolved Organic Carbon (DOC) and ammo­ though being tested at different spatial and temporal scales and using
nium, from 30 to 73% and from 0 to 99% respectively, depending on soil methods of different complexity. Table 2 collates reported values to
particle size, dissolved oxygen supply in the soil, and the residence time derive generalized patterns and trends. Accordingly, exfiltration rates
of sewage flow in the unsaturated zone (Fuchs et al., 2004). When are often expressed as the amount of exfiltration per standard area and
exposed to acidic or sulfate-bearing soil medium, concrete pipes are per unit length of pipes, i.e., l/day/cm2 or l/s/km, respectively (e.g.,
more susceptible to corrosion than clay pipes (Davies et al., 2001; Chisala and Lerner, 2008; Karpf, 2012). Ellis et al. (2009) pointed out
Jaganathan et al., 2010). Thus, it is often suggested to interpret the ef­ that an over-estimation of some reported data at the larger scale might
fects of soil on sewer exfiltration in modelling studies in the context of be due to upscaling the results from physical experiments and moni­
the interaction and exchange of the soil matrix, i.e., the matrix potential toring. Concluding from different studies, exfiltration rates between 1.0
of the soil below the colmation layer and the colmation layer (Ellis et al., × 10− 2 and 0.2 L/s.km or even less (in case of sewers carrying large
2009; Karpf et al., 2011). The soil matrix potential increases, and hy­ sediment loads) provide reasonable mean values for large-scale
draulic conductivity decreases with reduced water content in the soil, approaches.
together affecting the estimation of the hydraulic potential in the pipe
(Karpf et al., 2009). More details on this issue are described in the 4. Development and applications of sewer exfiltration models
following section on hydraulic potential.
Together with a reliable estimation of sewer exfiltration at defective
3.5. Hydraulic potentials in the pipe and in the soil (in the pipe vicinity) pipe sites, complex biogeochemical processes need to be considered to
model flow and matter transport into soil and groundwater (Karpf and
The process of sewer exfiltration depends on the hydraulic potential Krebs, 2011). Initial equations of sewer exfiltration were developed to
gradient between pipe water (as expressed by wastewater level) and quantify losses from sewer systems and to gain a better understanding of
water pressure in the soil in the pipe defect vicinity (Fig. 3). The pipe the effects of sewer exfiltration on groundwater (e.g., Yang et al., 1999;
water level may be represented in static mode by a constant water level Ly and Chui, 2012). Later studies integrated equations of sewer exfil­
or pressure head (representing, e.g., dry-weather flow) or calculated tration into empirical and process-based models to simulate the overall
dynamically using pipe flow equations or models (e.g., Rauch and sewer leakage processes (Reynolds and Barrett, 2003; Peche et al., 2017;
Stegner, 1994; Karpf et al., 2009). For dynamic pipe flow calculations, Dvory et al., 2018). This section reviews in detail the development and
the Saint-Venant equations (de Saint-Venant, 1871) are used in many advancements of sewer exfiltration equations and models.
studies (e.g., Karpf and Krebs, 2005; Kidmose et al., 2015; Peche et al.,
2017). Later studies include the effect of a colmation layer on the hy­
draulic potential in the sewer pipes and in the soil matrix surrounding
the pipes.
Table 2
The water pressure in the soil (in the pipe defect vicinity) is used for
Examples of sewer exfiltration rate per pipe length unit in modelling studies.
determining the hydraulic gradient, i.e., the actual three-dimensional
water pressure distribution in the soil is reduced to one spatial dimen­ Scale of study Exfiltration rate (L/s. References
km)
sion by assuming a water flow through the colmation layer to be one-
dimensional (e.g., Karpf, 2012; Peche et al., 2017). This is achieved by Pipe scale 0.1 – 2.0 Blackwood et al. (2005)
Pipe scale 1.7 × 10− 2
– 0.45 Trauth et al. (1995)
defining the hydraulic potential in the soil as the potential directly
below the colmation layer and enables to use a one-dimensional form of Pipe network of a city 0.012 Hoffman and Lerner
(1992)
Darcy’s law (or Richard’s) equation. According to this, the change of the
Catchment area of a city 1.4 × 10− 2 Yang et al. (1999)
hydraulic potential not only depends on the water level in the pipe and
Catchment area of a city 1.4 × 10− 9 – 0.179 Chisala and Lerner
the thickness of the colmation layer, but also on the soil matrix potential (2008)
below the colmation layer, including the local soil moisture (as District area of a city 3.0 × 10− 2
– 7.5 × 10− 2 Morris et al. (2006)
expressed by the relative permeability) and soil properties (as expressed Multiple scales of the 1.0 – 2.0 Amick and Burgess
by intrinsic permeability and retention curve parameters) (Beal et al., country (2000)
2004; Ellis et al., 2009). Most of the time, the sewer exfiltration process Single pipe to catchment 6.2 × 10− 2
– 17.9 × Reynolds and Barrett
scales 10− 2 (2003)
rarely occurs under partially saturated soil conditions and is highly non-
linear. However, sewer exfiltration is often linearized by using a modi­ Review of various scale 1.7 × 10− 9
– 0.179 Ellis et al. (2009)
studies
fied version of Darcy’s law in order to reduce computational

5
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

4.1. Estimation approaches of sewer exfiltration 4.2. Sewer exfiltration models

Three main approaches for calculations of sewer exfiltration are Sewer exfiltration models can broadly be grouped into data-driven or
defined in the literature: sewer exfiltration rates, leakage factor, and process-based categories based on model applications (Table 4). In
leakage functions (Table 3). general, none of the available model approaches provides a single so­
lution to the major issues in sewer exfiltration from site-specific to large
4.1.1. Sewer exfiltration rates scales, and all models are study-dependent because of the inherent
Equations of sewer exfiltration are based on a combination of theo­ complexity of pipe sewer exfiltration processes, the difference in pipe
retical assumptions and laboratory or field investigations. Two general network conditions, and data availability within the study context (Xu
approaches are referred to in the review by Rutsch et al. (2008). One et al., 2014a).
approach is based on the Torricelli equation (Eq. 1), calculating exfil­
tration rates as a function of the coefficient of outflow, gravitational 4.2.1. Data-driven models
acceleration, and pipe water level. Another approach is based on Darcy’s Data-driven models, or statistical models, employ data-driven tech­
law (Eq. 2). Later studies stressed the importance of a colmation layer, in niques to reveal the relationships between ‘explanatory’ variables and
particular the effect of hydraulic properties of the colmation layer on the leakage patterns of defective pipes at specific locations. The method
simulating the sewer exfiltration (DeSilva et al., 2005; Boukhemacha generalizes a predictive model with a certain degree of confidence (Liu
et al., 2015) (Eq. 3, Eq. 4). et al., 2010; Harvey and McBean, 2014). Such modelling approaches are
All above-stated calculations are based on the assumption of a sta­ most applicable to asset management and risk planning and modelling
tionary exfiltration flow, but in practice, this condition applies only the in-situ issues of sewer pipe damages in small catchments (e.g.,
under constant pipe and soil conditions (Han, 1997; Karpf et al., 2011; Mashford et al., 2011; Younis and Knight, 2010).
Boukhemacha et al., 2015). In practice, sewer exfiltration into a variably Defective sewer inspections, most commonly with CCTV techniques,
saturated soil is a non-linear process. For example, exfiltration rates resulted in an extensive database that is often abstracted at different
using the linear Darcy approach tend to be overestimated under wet levels to enable various analyses of the exfiltration pattern and failure
weather conditions when significant proportions of sewer pipes are rates of the whole sewer network (e.g., Le Gat, 2006; Mashford et al.,
pressurized, and soil resistance increases with increasing volume of 2011). For example, Davies et al. (2001) extracted 12.000 records of
wastewater flow (Rushton and Tomlinson, 1979; Rutsch et al., 2008). CCTV data and other explanatory variables for a 634 km long sewer
Similar to the pipe-aquifer interaction of sewer exfiltration, developing system to develop a statistical model for predicting sewer failure prob­
an equation of non-linear exfiltration from sewers into an aquifer re­ ability. Younis and Knight (2010) used extensive inspection data to
quires information of both water level and streambed properties (Eq. 5). develop a Generalized Linear Regression Model to predict the proba­
Corresponding to this approach, Karpf and Krebs (2004) propose a sewer bility of aged pipe deterioration in Ontario. Their results suggested that
exfiltration model which bases on a model for surface water leaking into the deterioration of vitrified clay (VC) pipes is age-independent, while
groundwater (Han, 1997) (Eq. 6, Eq. 7). the service life for reinforced concrete (RC) pipes is approximately 75
years.
4.1.2. Leakage factor Outputs of sewer exfiltration models are often considered to directly
Rauch and Stegner (1994) introduced the generic leakage factor KL correspond with sewer damages and this principle is applied for the
(Eq. 8). This factor incorporates the two parameters, hydraulic con­ development of various sewer deterioration models (e.g., Baur and Herz,
ductivity and thickness of the colmation layer, which are difficult to 2002; Harvey and McBean, 2014). One common technique is the cohort
determine and often unknown in practice (Peche, 2019). Introducing survival model, which was selected as the primary tool to predict the
such a lumped leakage factor as a single colmation layer parameter helps structural condition of sewer pipes at the network-level in various re­
to simplify the calibration of sewer exfiltration models. Eq. (10) is a gions, for instance, in Norway (Rostum et al., 1999), Germany (Baur and
more detailed version of the sewer leakage equation proposed by Ellis Herz, 2002), or Canada (Duchesne et al., 2012). Le Gat (2006) and Le
et al. (2009) in order to incorporate interactions and combined effects of Gat (2008) applied the non-homogenous Markov chains (NHMC) model
the colmation layer, the soil layer, and the transition zones on exfiltra­ to develop the time-dependent probability function to predict sewer
tion rates. Direct calibration of the exfiltration factor KL and its com­ deterioration conditions to support short-term repair and long-term
ponents are often not possible and are alternatively proposed to achieve maintenance planning of sewer networks. Another example is the
by experimental analysis (Karpf and Krebs, 2011). This method applies Gompitz model, which can predict pipe defective states even when
the Eq. (11) and requires data on a dynamic (i.e., colmation layer sewers are subjected to very sparse inspections (Caradot et al., 2017).
clogging and break-up processes) and a structural non-dynamic The method was successfully applied in many other countries, such as in
component (i.e., exfiltration area and soil properties). Canada (Harvey and McBean, 2014), Norway (Rokstad and Ugarelli,
2015), Germany (Caradot et al., 2017), and Austria (Fuchs-Hanusch
4.1.3. Leakage function et al., 2015).
Peche et al. (2017) introduced a nonlinear leakage function that Artificial intelligence-based and machine learning techniques pre­
relates the nonlinear dependency of sewer exfiltration to pipe water sent other alternative methods, particularly in cases where the inherent
level and water pressure below the colmation layer. That relationship uncertainties of inspection datasets constrain the use of statistical
can be represented by a 4th order polynomial equation (Table 3). Co­ models (e.g., Khan et al., 2010; Mashford et al., 2011). Artificial neural
efficients in Peche et al. (2017) were determined using a coupled nu­ networks (ANN) were successfully used by Najafi et al. (2005) and Khan
merical pipe flow-unsaturated flow model. For this leakage function, soil et al. (2010) to investigate the impact of explanatory parameters on the
water pressure and pipe water level act as input variables and specific prediction of the sewer pipe condition based on limited input data.
sewer exfiltration flux via the colmation layer as the output variable. Applications of machine learning models, such as the Support Vector
This method enables coupling pipe flow and unsaturated zone flow Machine (SVM), helped to overcome the constraints of the sparseness of
(both flow regimes may be calculated with numerical models). The sewer exfiltration data when predicting the condition of individual
leakage function represents an upscaling method because it enables to sewer pipes in Australia, Colombia, and Germany (Mashford et al., 2011;
omit otherwise very fine spatial discretization of the colmation layer. Hernández et al., 2018). The Neural Network machine learning model in
the study by Yin et al. (2020) also predicted well the deterioration state
of individual sewer pipes. The gained knowledge from this model can
ultimately be generalized with a certain degree of confidence to non-

6
H.H. Nguyen et al.
Table 3
Approaches of sewer exfiltration estimation using equations of exfiltration rate, leakage factor, and leakage functions.
No. Equation functions Parameters References
√̅̅̅̅̅̅̅̅ 3 − 1
Eq.(1) Qex = Aleak *μ* 2gh Qex = exfiltration rate (L T ) Rutsch et al. (2008)
Eq. kf Aleak = leakage area (L2)
Qex = Aleak *h* µ = coefficient of outflow (–)
(2) Δs
g = gravitational acceleration (LT− 1)
h = wastewater level in the pipe (L)
kf = coefficient of permeability (L T− 1)
Δs = depth of the soil column (L)
Eq. kc kc = hydraulic conductivity of colmation layer (L T− 1) DeSilva et al. (2005);
Qex = h* *Aleak
(3) B B = thickness of the colmation layer (L) Boukhemacha et al. (2015)
Eq. kc L = length of the sewer segment of interest (L),
Qex = h* *(Wp *L*%leak )
(4) B Wp = wetted perimeter generated with the water level in
the sewer (L)
%leak = percentage of leaking surfaces from the total
area under the wetted perimeter.
Eq. q = k1 *(ek2*Δh − 1) whenΔh > 0 q = the flow between sewer conduit and aquifer Rushton and Tomlinson (1979)
(5) (L3L− 1T− 1)
k1 = the streambed leakage coefficient (L T− 1)
k2 = a constant (L)
Δh = the difference in water level between sewer
conduit and aquifer (L)
7

Qex,T = exfiltration rate at time T (L3 T− 1)


∑i
Eq. Qex,T = KL i=1 Ai,T *(hG,i,T − hS,i,T ) whenhS,i,T > hG,i,T > hP,i Karpf and Krebs (2004)
(6) Ai,T = exfiltration effective surface area at time T (L2)
∑i
Eq. Qex,T = KL i=1 Ai,T *(hS,i,T − hP,i ) whenhP,i > hG,i,T hS,i,T = height of water level in pipe i at time T (L)
(7) hG,i,T = height of groundwater level pipe i at time T (L)
hP,i = pipe bottom level of pipe i (L)
Eq. kf KL = leakage factor (T− 1) Rauch and Stegner (1994)
KL =
(8) dl kf = coefficient of permeability (L T− 1)
Eq. Qex = Aleak *h*KL dl = thickness of the layer (L)
(9)
Eq. kcol Qex ΔH = hydraulic or matrix potential (m)hcol = thickness Ellis et al. (2009)
KL = −
(10) hcol Aleak *ΔH of the colmation layer (m)
kcol = permeability coefficient of the transition zone
beneath the leak (m/s)
Eq. KL = KL,dy *KL,n− dy KL,dy = leakage factor with dynamic input data (T− 1) Karpf and Krebs (2011)
(11) KL,n− dy = leakage factor with non-dynamic input data
(T− 1)

Journal of Hydrology 596 (2021) 126130


Eq. 1 ( ) qex = specific discharge (m/s) Peche et al. (2017)
qex * a00 + a10 p + a01 h + a20 p2 + a11 ph + a02 p3 + a21 p3 + a21 p2 h + a12 ph2 + a03 h3 + a40 p4 + a31 p3 h + a22 p2 h2 + a13 ph3
(12) 2.25*10− 4 p = soil water pressure below the comation layer (Pa)
h = height of water level in pipe (m)
axx = polynomial coefficients (units are omitted for
legibility)
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

Table 4
Modelling approaches and applications incorporating sewer exfiltration processes and parameters.
Model approaches Model description Model inputs* Model applications

(1) (2) (3) (4) (5) (6) (7) (8)

Data-driven models
Correlation analysis The correlation pattern is defined among pipe x x x x Dohmann et al. (1999); Davies et al. (2001);
damages attributes, other environmental DeSilva et al. (2005)
attributes, and resulting exfiltration in order to
identify important variables which affect sewer
exfiltration
Logistic and multiple Model variables are treated as random data and x x x Ariaratnam et al. (2001); Chughtai and Zayed
regressions have a probability distribution. Include (2008); Younis and Knight (2010); Salman
regression models, multiple linear regression and Salem (2012); Fuchs-Hanusch et al.
models, or binary logistic models. (2015); Lee et al. (2015); Roehrdanz et al.
(2017)
Cohort survival models Cohorts (sewers) are grouped based on similar x x Herz and Hochstrate (1987); Baur and Herz
characteristics, i.e., the period of construction or (2002); Baik et al. (2006); Duchesne et al.
failure probabilities. Inspection data are used to (2012)
develop ‘survival’ curves to forecast the residual
lifetime of the pipe and the specific time when
sewers enter a critical condition.
Markov-chain models A common type of deterioration model. The x x x Baur and Herz (2002); Le Gat (2006, 2008);
model incorporates the time-dependent Egger et al. (2013); Caradot et al. (2017);
transition probabilities, i.e., probabilities of Tscheikner-Gratl et al. (2016); Rokstad and
condition changes, which allows defining the Ugarelli (2016)
specific time when the condition changes in a
sewer system.
Artificial intelligence Allow identifying the non-linear relationships x x x x Tran et al. (2006); Khan et al. (2010);
and machine between sewer condition states and explanatory Mashford et al. (2011); Syachrani et al.
learning models variables by ‘learning’ the behavior of (2013); Sousa et al. (2014); Jiang et al.
deteriorated pipes from inspection data. It does (2016); Harvey and McBean (2014); Rokstad
not require a predefined structure of the model and Ugarelli (2015); Laakso et al. (2018);
to link model inputs and outputs. Commonly Hernández et al. (2018)
used models are Neural Networks, Random
Forests, and Support Vector Machines.

Physically-based models
Groundwater models A groundwater flow model. Sewer leakage is x x x x x x Yang et al. (1999); Lerner (2002); Reynolds
considered as a pathway contributing to aquifer and Barrett (2003); Fenz et al. (2005);
recharge. The simulated flow is calibrated with Wakida and Lerner (2005); Wolf et al. (2006,
data available from groundwater and other 2012)
monitoring sources.
Models based on water Sewer exfiltration is estimated by determining x x x x x x Karpf and Krebs (2005); Trauth et al. (1995);
balance principles water balance inputs, such as precipitation and Nakayama et al. (2007); Li et al. (2008); Xu
water consumption and discharge. et al. (2014a, 2014b))
Sewer pipe models Sewer exfiltration is represented as a function of x x x x x x x x DeSilva et al. (2005, 2007); Karpf and Krebs
pipeline system deterioration and incorporates (2011); Peche et al. (2017, 2019)
sewer exfiltration equations, its key parameters
and processes, and interactions among key
variables which occur in the sewer pipe system.
Coupled pipe- The transport of flow and solutes are simulated x x x x x x x x Wolf et al. (2007); Mohrlok et al. (2008);
subsurface- by coupling the sewer pipe models and (or) Maxwell et al. (2009); Vizintin et al. (2009);
groundwater and sewer exfiltration equation with various Ly and Chui (2012); Bhaskar et al. (2015);
integrated models hydraulic, hydrologic, and process-based models Kidmose et al. (2015); Thorndahl et al.
in tandem. (2016); Gogu et al. (2017); Peche et al. (2017)

* Model inputs: (1) Climate data; (2) Sewer pipe characteristics and conditions; (3) Wastewater content; (4) Soil; (5) Colmation layer; (6) Hydraulic potential; (7)
Groundwater data; (8) Administrative information

inspected pipes to better define rehabilitation planning (Caradot et al., from defect sources into aquifers (Yang et al., 1999; Karpf et al., 2011;
2017; Tscheikner-Gratl et al., 2020). In another study by Laakso et al. Boukhemacha et al., 2015). In general, physically-based models serve as
(2018), the machine learning approach identified pipes in critical con­ a useful tool for defining areas vulnerable to groundwater contamina­
ditions and revealed key predictor factors which affected the poor pipe tion and predicting sewer exfiltration scenarios with a certain degree of
conditions. In general, machine learning models show an advantage spatial and temporal representation (Mohrlok et al., 2008; Ellis and
over other statistical approaches in revealing the pipe-scale sewers in the Bertrand-Krajewski, 2010).
critical condition (e.g., Caradot et al., 2018; Tscheikner-Gratl et al., Groundwater and water balance models are commonly used in sewer
2020). exfiltration studies to estimate the exfiltration rates, particularly in areas
lacking site-specific information on defect conditions and inspections
4.2.2. Physically-based models (Karpf and Krebs, 2005; Xu et al., 2014b). A reliable representation of
Unlike data-driven models, physically-based models are based on areas susceptible to sewer exfiltration pollution typically requires an
physical process principles and aim to describe the relationships among accurate determination of water balance elements in urban groundwater
explanatory parameters relevant to the simulated process (Zoppou, models (Boukhemacha et al., 2015; Lee et al., 2015; Gogu et al., 2017).
2001). A dependency among explanatory variables that drive sewer The core water balance-specific inputs include precipitation, evapora­
exfiltration processes are often assumed or formulated, and dependent tion, interception, infiltration and percolation, transpiration and root-
variables are incorporated into equations of transport of sewage fluids water uptake, surface water flow, miscellaneous sink/source terms (e.

8
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

g., leaking water mains, groundwater infiltration into a leaky pipe expanded toward integrating multi-dimensional, coupled pipe-soil-
network), water consumption at households, or collected/treated groundwater models that represent better the spatial and temporal
wastewater. For example, Nakayama et al. (2007) combined data from pattern of sewer exfiltration (e.g., Schulz et al., 2005; Mohrlok et al.,
730 urban catchments with sewer pipes and data on precipitation to 2008; Peche et al., 2019). The calibration of integrated sewer exfiltra­
estimate the average leakage of sewer fluids into aquifers in Tokyo, tion models is often based on water sampling and groundwater moni­
Japan. The calculation of sewer exfiltration based on the water balance toring (e.g., Wolf et al., 2012; Boukhemacha et al., 2015; Roehrdanz
was used in a case study by Xu et al. (2014b) to represent the Shanghai’s et al., 2017). In the study by Mohrlok et al. (2008), the leakage from
drainage system, China (but with the main focus on dry-weather flow pipes is considered the point source which can generate the three-
conditions). The simulated groundwater flow is calibrated with data dimensional flow, and results from this study defined that the long-
from groundwater monitoring. To reduce model complexity of ground­ term residence times of wastewater leakage in the soil medium were
water models, some studies focusing on entire pipe networks often due to lateral flow spreading. Dvory et al. (2018) used coupled 1D-3D
consider certain water balance elements (including precipitation, unsaturated and saturated flow and transport models to evaluate the
evaporation, interception, transpiration, root-water uptake, and surface effects of a leakage event after a maintenance incidence. They found out
water flow) as lumped into groundwater recharge, which is assigned to a that sewer leakage was retained in the porous media of a karst/
Neumann-type boundary condition at the groundwater model top fractured-porous unsaturated zone and groundwater system and
boundary (e.g., Peche et al., 2019; Rodriguez et al., 2020). Such a contributed a continuous supply of the micro-contaminant carbamaze­
reduction in model complexity enables to omit the calculation of surface pine to the groundwater and nearby creeks long after the incidence has
water flow and its coupling to the groundwater flow regime (Rodriguez stopped. Bhaskar et al. (2015) used the coupled groundwater-surface
et al., 2020). It is crucial to include elements of the urban water balance water-land surface model ParFlow (Maxwell et al., 2009) to quantify
into such a lumped groundwater recharge because a mere consideration the long-term impact of exfiltration on groundwater storage in Balti­
of water balance elements would truncate the urban water balance. more, USA, and defined that sewer exfiltration is the second main driver
When modelling leakage from a specific pipe network, all other pipe causing the change in subsurface storage volume when compared to
networks providing exchange fluxes to and from groundwater (such as other factors. Peche et al. (2017) calculated sewer exfiltration based on a
sewer network, stormwater network, water mains, and drain tiles) and hydro-dynamic pipe water level and variably soils saturation subsurface
their respective temporal variations (from daily to seasonal flow pat­ flow approach and suggested that for groundwater levels far below leaky
terns) have to be considered as inputs of the urban water balance. Other pipes, sewer exfiltration rates are independent of groundwater and soil
inputs are irrigation and dewatering of construction sites. The incor­ water pressure and merely depend on the water level in the sewer pipe
poration of surface coverage elements in the model is particularly (Peche, 2019). This study also revealed that the characteristics of the
important in areas of low-pervious or impervious surface cover, in which aquifer strata and colmation layer, which in this case were sand and
the local groundwater recharge may not be precipitation-driven at all sandy loam, had significant effects on the breakthrough time of pipe
(Mitchell et al., 2003; Kenway et al., 2011). leakage to the groundwater table. In another case study by Bhaskar et al.
Sewer pipe models are developed to supplement the rough estima­ (2015), the hydraulic connection between the sewer exfiltration to
tion of water balance and groundwater models, to further advance the groundwater varied when passing through various unconfined and
model knowledge and to increase the accuracy in predicting sewer confined aquifers.
exfiltration effects on the environment when incorporating specified Incorporating geological and hydrogeological models in the inte­
exfiltration characteristics and patterns and wastewater flow equations grated model framework revealed that the exfiltration process from
in the model. The first sewer exfiltration model at the pipe scale was sewers depends strongly on the interaction between the conduit and the
proposed by DeSilva et al. (2005), in which the flow equation of sewer hydrogeological strata (Ly and Chui, 2012; Boukhemacha et al., 2015;
exfiltration was adapted from the studies of Rauch and Stegner (1994) Roehrdanz et al., 2017). The study coupling the urban 3D geological and
and Vollertsen and Hvitved Jacobsen (2003) to quantify the daily sewer system models in Bucharest city, Romania, showed that sewer
pattern of sewer exfiltration. At the city scale, the sewer network exfil­ conduits located completely in the aquitard strata were particularly
tration and infiltration model (NEIMO) was developed within the Eu­ susceptible to the exfiltration process. In contrast, conduits located
ropean funded project ‘Assessing and Improving the Sustainability of partially or entirely in the aquifer might exhibit either infiltration or
Urban Water Resources and Systems’ (AISWURS) and was applied in exfiltration (Gogu et al., 2017). Sewer conduits with a known hydraulic
various countries in Europe and Australia to reveal the magnitude of property of backfill material when crossing through different zones of
sewer exfiltration in regions with different pipe and soil conditions the heterogeneous aquifer and aquitard strata showed very distinct
(DeSilva et al., 2007; Ellis and Bertrand-Krajewski, 2010). Another patterns of hydraulic interaction with the hydrogeological matrix, and
example of a sewer exfiltration model was proposed by Karpf and Krebs consequently very different rates and loads of sewer leakage to
(2011), in which the dynamic and structural components of the exfil­ groundwater (Lee et al., 2015; Roehrdanz et al., 2017; Gogu et al.,
tration process, including clogging/break-up of the colmation layer and 2017). The volume of artificial recharge from sewer leakage sources to
groundwater infiltration in the vicinity of sewer leaks, were introduced groundwater varied in both space and time in the case study of Barce­
in the hydrodynamic model setting. Karpf and Krebs (2011) study lona city, Spain, due to the complex nature of the aquifer strata (Tubau
showed the complexity of exfiltration hydraulics and identified impor­ et al., 2017). Thus, studies focusing on the effects of sewer exfiltration on
tant parameters and components of exfiltration models which can be groundwater need to properly consider the intersection between the
managed on the basis of different data availabilities. The results of Karpf sewer conduits and the aquifer strata by correctly reproducing the hy­
and Krebs (2011) were adapted in the study by Peche et al. (2017), draulic processes and water balance of core domains of surface water,
which further reveals the mechanism and parametrisation of sewer pipe flow, hydrogeology, and groundwater strata.
exfiltration for upscaled studies on the basis of coupled pipe flow and
groundwater flow simulators. Approximation of sewage exfiltration by 4.2.3. Others
means of leakage flow functions, which were derived from numerical These include other models or management tools that do not strictly
simulations of pipe water levels and hydraulic pressures below the belong to the two above mentioned categories, such as visual mapping
colmation layer, proved to reasonably represent the spatial discretiza­ and GIS-operations. Such models are particularly used in areas with a
tion of pipe defects at broader network scales with reduced computa­ deficiency of available data for the setup of sewer exfiltration models.
tional times (Peche et al., 2017). Reconstructing spatial-temporal patterns of sewer defect distribu­
Stimulated by increasing availability of observed sewer exfiltration tions in urban areas by using GIS tools is considered a useful supple­
data and the improvement of computational power, model designs have mentary tool to analyze the risk and potential locations of sewer

9
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

exfiltration to groundwater (Mohrlok et al., 2008; Roehrdanz et al., sewer pipe models with urban infrastructure, hydrogeological, and
2017; Ducci, 2018). Selvakumar et al. (2004) developed a map on the groundwater models is essential to realistically simulate the effects of
depth-to-groundwater basing on sewer network and groundwater level sewer leakage to groundwater. Examples from the literature showed
data in the USA. Hopkins and Bain (2018) and Vystavna et al. (2018) diverse applications of data-driven and physically-based models that
applied a similar approach to map the spatial patterns of sewer networks primarily targeted only one or another above-mentioned aspect and
and defined potential sewer leakage hotspots by conducting the prox­ simplified the representation of other aspects of the sewer exfiltration
imity analysis between aging sewers and surface waterways. Ducci process (e.g., Rutsch et al., 2008; Tscheikner-Gratl et al., 2020). An
(2018) produced the Potential Nitrate Contamination (PNC) map, which accurate prediction of sewer exfiltration impacts on groundwater at
incorporates the source of sewer leakage in urban sewage networks, by large scales requires a thorough consideration of both aspects by care­
overlaying various thematic maps and statistical data available at the fully integrating the model domains from the local conduit surface to the
national scale. Lee et al. (2015) and Roehrdanz et al. (2017) used a lo­ entire sewer network and city scales.
gistic regression model and GIS tool to predict the high likelihood areas At the pipe scale, an accurate hydraulic simulation of sewer exfil­
of sewer exfiltration. tration needs to consider the condition of the following dependencies:
Using leakage indicators is an alternative approach in regions where initial soil type, vertical distance to the groundwater table and pipe
the sewer leakage data constrain the application of models. For instance, water level (that may be used to estimate a hydraulic gradient through
Barrett et al. (1997) proposed combining stable nitrogen isotopes and the colmation layer), and the colmation layer condition (Karpf, 2012;
microbiological indicators as the most effective parameters for the Guo et al., 2017; Peche, 2019). The initial soil type is often well-known if
quantification of groundwater recharge by sewer exfiltration. Their in­ pipes are embedded in well-defined backfill material (Karpf, 2012). The
vestigations provide evidence for the presence of sewage in shallow vertical distance to the groundwater table may be calculated using pipe
groundwater. An approach using various wastewater indicators to network information (in the form of pipe invert depth) and groundwater
analyse the process of sewer exfiltration to groundwater is also proposed maps (Selvakumar et al., 2004; Roehrdanz et al., 2017). The pipe water
in monitoring combined with modelling studies by Wolf et al. (2012), level is highly variable and may be simplified by using the dry-weather
Lee et al. (2015), and Roehrdanz et al. (2017). flow water level (Peche et al., 2017). However, data of the colmation
layer parameters are not readily available and difficult to predict and
5. Challenges and research needs for sewer exfiltration have to be calibrated or determined using in-situ physical experiments
modelling (e.g., Karpf et al., 2009). Sewer pipe models are often based on the
simplification that hydraulic properties and the depth of the colmation
Based on current literature, it can be concluded that the primary layer are constant. Results from physical experiments justified the latter
focuses of sewer exfiltration studies, including the identification of (e.g., Okubo and Matsumoto, 1983; Rauch and Stegner, 1994), which
important variables affecting sewer exfiltration, the behavior and in­ shows that these attributes of the colmation layer converge to a constant
teractions among sewer exfiltration parameters, and the presence and value within a short time of pipe defect occurrence. However, this
types of effects on the environment, are well addressed at the scales from simplification leads to the significant uncertainty in the model outputs of
single-site to certain defined sewer lengths (Section 2, Section 3). Based long-term predictions due to the negligence of colmation layer dy­
on those grounds, various approaches and models have been proposed in namics, such as colmation layer breaking-up effects (driven by high pipe
the literature to assist the prediction of deterioration status of sewer water pressure), leading to short-time large sewer exfiltration rates
pipes and to define the critical areas affected by sewer exfiltration to the (known as leakage pulses) (Ellis et al., 2009; Karpf, 2012). Developing
environment (Section 4). colmation layer forming models which allow predicting both the hy­
However, effects of sewer exfiltration on the surrounding environ­ draulic properties and depth of the colmation layer, and (or) filtration
ment are often not limited to single sites where the leakages from pipe models coupled with biofilm growth models are potential important
defects are inspected and modeled (Section 1). Long-term management research areas which can help to better represent the processes of hy­
decisions and rehabilitation planning often require a determination of draulic flow transport to groundwater in integrated pipe-subsurface-
the system deterioration and effects at the broader scales of whole sewer groundwater models. As the subject of future studies, it is also impor­
pipe networks or catchments (e.g., Wolf and Hötztl, 2007; Heinrich, tant to incorporate leakage pulses into sewer exfiltration models in order
2007; Roehrdanz et al., 2017; Peche, 2019). Applications of sewer to better address and to compare the effects of sewer exfiltration under
exfiltration models at higher spatial and temporal resolutions are still the dry-weather flow and stormwater flow conditions for both long-term
constrained due to the lack of information and understanding on the role studies and studies focusing on event-based impacts of sewer
and interactions of key driving factors, which are not directly derivable exfiltration.
from small-scale studies (e.g., Rutsch et al., 2008; Vystavna et al., 2018). At large spatial scales, the challenge in modelling the hydraulic
The following section focuses on the more recent literature addressing processes of sewer exfiltration is associated with the local geology
these knowledge gaps and defining research needs for future studies on characteristics (Geyer et al., 2007; Dvory et al., 2018). Caverns may
sewer exfiltration modelling. form from sewer exfiltration-bound percolation when sewer leakage
occurs in dissoluble rock (e.g., limestone) (Hermosilla, 2012). These
5.1. Model hydraulic processes of sewer exfiltration caverns represent a substantial variability in local geology, and this
variability needs to be accounted for when modelling sewer exfiltration.
To estimate and adequately predict the effects to groundwater from In other cases, the exchange between conduits and the aquifer strata
the sewer exfiltration source, studies on the hydraulic modelling aspects when sewer infiltrated via carbonate aquifers with preferential flow
of sewer exfiltration need to cover two main areas, which include (1) an paths may significantly impact on the local groundwater (e.g., Dvory
improved understanding of core processes and interactions among key et al., 2016, 2018). In contrast, the presence of areas with very small
parameters of sewer exfiltration at the pipe scale, and (2) a better un­ hydraulic conductivities (e.g., clay lenses) may lead to a soil-bound
derstanding of the transport and transformation processes of sewer sealing of pipe defects, and modelling sewer exfiltration in such areas
leakage into aquifer strata. For the first aim, advanced data-driven may not be necessary. Thus, the local geology and geological strata
techniques such as machine learning and artificial intelligence models characteristics should be well known when modelling the process of
proved to be efficient and reliable tools that can help an accurate pre­ sewer exfiltration to groundwater.
diction of pipe deterioration status and its key explanatory variables for A further challenge in modelling the sewer exfiltration hydraulic
a better exploration of sewer exfiltration processes at the pipe scale. processes is caused by the complex nature of underground in­
Meanwhile, coupling the physically-based, integrated surface water and frastructures and their interactions with the local hydrogeology, which

10
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

subsequently affect the potential distribution of sewer exfiltration to the wastewater largely depend on land-use within the urban area and
groundwater. Urban systems are often characterized by extensive net­ thus often requires intensive sampling and analysis with a focus on total
works of sewage pipes which intersect with other underground in­ filterable solids, components leading to chemical precipitation, and
frastructures and constructions (e.g., stormwater canals and drains) and bacteria leading to biofilm growth (Wolf et al., 2012; Lee et al., 2015;
various pervious and impervious hydrogeological strata. These in­ Roehrdanz et al., 2017). Thus, the knowledge gained and behavior
tersections contribute a significant source of disturbance to the local patterns of these contaminants are not readily available to enable a
urban groundwater (Bhaskar et al., 2015; Attard et al., 2016). Studies generalization for larger-scale modelling studies than the understanding
also revealed that the problem of sewer exfiltration in urbanized regions of hydraulic sewer flow models (e.g., Peche et al., 2017; Tscheikner-
is not limited to dry-weather conditions as often referred in previous Gratl et al., 2020). In other studies, wastewater pollutants from sewer
studies. Indeed, a significant amount of rainwater and surface flow via exfiltration are roughly represented as fixed concentration amounts or
pervious urban areas may infiltrate into conduits during storm events indirectly estimated from population dynamics and other indicators (e.
and contribute to an increased peak of sewer discharge to groundwater g., Vystavna et al., 2018). However, no uncertainty quantifications in
over a longer duration period depending on the local characteristics and such approaches have been reported so far in the literature.
conditions of pervious and impervious aquifers (e.g., Ly and Chui, 2012; For pipe and catchment-scale models, it is also critically important to
Boukhemacha et al., 2015; Tubau et al., 2017). Therefore, it is crucial to correctly characterize the local condition of geological strata and to
judiciously analyse local hydrogeology and groundwater maps to incorporate the physico-chemical adsorption and interactions of
accurately estimate the full sewage leakage process and to understand its wastewater constituents with the soil medium to better understand the
migration mechanism into groundwater. In regions where detailed transport and transformation of key wastewater leakage compositions
urban groundwater maps are not available, there is a need to apply into groundwater. Sewer exfiltration adsorption at pipe-soil interfaces is
alternative and more generic approaches (e.g., using leakage factors and considered as a key factor in the formation of a sedimentary media at the
functions) in order to gain the reasonably accurate estimation of sewer pipe-soil surface, which is known as a colmation layer and plays a
exfiltration to groundwater. critical role in defining the ultimate quantity and quality of exfiltration
Besides, current sewer exfiltration models do not take into account to the environment (Vollertsen and Hvitved-Jacobsen, 2003; Ellis et al.,
the potential ‘hydraulic disconnection’ of leaky pipes from groundwater, 2009; Karpf, 2012). The porous matrix of an unsaturated–saturated zone
i.e., the distance of groundwater and leaky sewer at which the discon­ is extraordinarily heterogeneous (Vollertsen and Hvitved-Jacobsen,
nection occurs and no significant effect on groundwater is observed 2003; Ellis et al., 2009). This often leads to the assumption of homo­
(Peche, 2019). Some studies with a focus on groundwater pollution geneity to characterize the medium for simplification in large-scale
suggested that the impact from sewer exfiltration sources was mostly studies. However, the hydraulic interaction between the sewer leakage
remarkable and relevant near the vicinity of defective pipe areas (e.g., flow and the hydrogeological matrix in the soil for the known hydraulic
Reynolds and Barrett, 2003; Rutsch et al., 2008; Bhaskar et al., 2015). properties of the colmation layer might be completely different. This
Such a hydraulic disconnection is well known for river-groundwater clogging phenomenon between the sewer exfiltration flow and the
interactions (e.g., Sophocleous, 2002; Brunner et al., 2009). Mean­ aquifer system environment has to be properly quantified and repre­
while, the concept was first described for sewer-groundwater in­ sented in sewer exfiltration models (including stabilization of gas bub­
teractions in Peche et al. (2017). However, to which spatial boundary (i. bles, transport of particles in soil pores, attachment of toxic metal and
e., depth to groundwater) this effect should be considered in sewer algae, or development of preferential flows) in order to improve the
exfiltration modelling is still largely unknown and should be subject to understanding and reduce the uncertainty in evaluating the impacts of
future research (Lee et al., 2015; Roehrdanz et al., 2017). A better- sewer exfiltration on the groundwater (Nickey et al., 2015). Besides, an
obtained knowledge of sewer pipe hydraulic disconnection has the po­ improved understanding of sewage solute transport and impacts on
tential for model simplification since solving the nonlinear and vulnerable groundwater sources can help to formulate a legal threshold
computationally costly Richards (1931) equation is not necessary for a on ‘allowable’ limits of sewer leakage in urban wastewater systems.
disconnected leaky pipe.
5.3. Model calibration and uncertainty analyses
5.2. Model wastewater compositions from sewer exfiltration
In general, sewer exfiltration models at large scales are not well
Sewer exfiltration is characterized by a diverse composition of sub­ established (e.g., Hopkins and Bain, 2018; Ducci, 2018). One main
stances from industrial, commercial, and residential sources such as constrain is due to the fact that the development of large-scale models
heavy metals, surfactants, or organic compounds (Reynolds and Barrett, based on site-specific experiments or small-scale models often requires
2003; Karamouz et al., 2011; Lee et al., 2015). These substances differ in careful realisation of model parameters, i.e., the knowledge of several
their characteristics and behavior (including, e.g., degradation, solubi­ parameters with high spatial and temporal resolutions, which in most
lity, adsorption-affinity) in the pipe and in groundwater, and as such cases are not available (Karpf, 2012; Peche, 2019). For instance, the
might not be compatible to model in a similar way as pollutant in­ calibration based on groundwater monitoring merely enables to esti­
dicators in surface water or groundwater models. However, very few mate a hydraulic conductivity for fully saturated conditions. Thus, it is
studies have been focused on modelling of wastewater contaminants in important to use the results of laboratory experiments and field samples
sewer exfiltration (e.g., Roehrdanz et al., 2017; Dvory et al., 2018). to determine the relative hydraulic conductivity function and retention
Thus, the behavior and effects of most contaminants from sewer exfil­ curve parameters to support the correct simulation of flow under un­
tration on the urban environment in the literature remain largely un­ saturated conditions at the pipe and sewer network scales. Adapting the
known, particular on large scales, though it is acknowledged that the leakage factor approach has proven a reliable method which is
significant pressure on groundwater from sewer exfiltration is caused by increasingly used in various pipe to city-scale studies to simulate the
pollutant loads rather than volume of sewer exfiltration discharge (e.g., interaction between groundwater and sewer systems with a less data
Wolf et al., 2012; Roehrdanz et al., 2017). demand on complex sewer network and local strata conditions (e.g.,
Some studies incorporated sewage solute transport module with Karpf and Krebs, 2011; Boukhemacha et al., 2015; Peche et al., 2017).
coupled sewer flow, variably saturated flow, and groundwater flow For model calibration, the following factors should be considered. A
models, but the model results are interpretable only within the context reasonable discretization of the groundwater model into zones of
of the local small-scale regions and limited observation data available different soil properties and aquifer characteristics, and distinguishing
for model calibration (Mohrlok et al., 2008; Ly and Chui, 2012; sewers into certain segments which can act either as a drain for the
Thorndahl et al., 2016). In fact, the physico-chemical compositions of aquifer or as an artificial recharge source to groundwater. Further,

11
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

consideration of an appropriate number of groundwater observations, colmation layer, enables to omit the spatial discretization of colmation
since the results of model calibration is often dependent on both the layers in numerical models and significantly reduces computation times.
model structure and the number of available groundwater observations Along with the generalization of general patterns of key parameters and
in the study area. Tubau et al. (2017) suggested that urban recharge processes of sewer exfiltration, integrating various modelling techniques
models incorporating sewer exfiltration can only be successfully vali­ with advanced GIS and planning tools to extract available large-scale
dated when an ‘accurate’ groundwater flow and transport model is data (such as detailed maps of sewer networks, urban infrastructure,
available in the study area. A successful calibration is also attributed to urban groundwater levels, and data on structural and operational states
the model uncertainty, which is an integral problem of all modelling of sewer conduits) is also a common but effective approach. These ap­
practices and is caused by various sources such as input data, model proaches are increasingly being used to supplement the data shortage
structure and parameterization, as well as model simplifications and from sewer observations required for large-scale studies of sewer exfil­
assumptions (Rutsch et al., 2008; Karpf et al., 2009). Thus, the uncer­ tration impacts on the environment (e.g., Lee et al., 2015; Ducci, 2018;
tainty of sewer exfiltration models should be addressed in tandem with Hopkins and Bain, 2018).
model calibration to improve the accuracy of modelling outputs. The ideal approach to accurately allocate urban areas susceptible to
The first calibration model incorporating sewer exfiltration was re­ sewer exfiltration contamination and groundwater pollution is to care­
ported in the study by Yang et al. (1999), though the main focus was on fully upscale the integrated surface water-pipe-hydrogeology-
fitting the simulated flow with available observed groundwater data. groundwater model from site-specific location to the pipe network and
Meanwhile, applying a Monte Carlo approach by calculating 2000 entire city scales in order to accurately simulate the key water balance
randomly selected combinations of seven independent input parameters components. However, such intensive data requirement of both sewer
for the sewer network of the city of Rastatt, Germany, Wolf et al. (2006) network and groundwater observations is often unavailable and cannot
and Wolf and Hötztl (2007) successfully defined the uncertainty ranges be satisfied in most urban systems worldwide. A potential solution to
for each parameter and the validated exfiltration rate was reported with this common issue might be to perform the ‘multi-component and multi-
the confidence interval of 95%. Karpf (2012) also incorporated Monte scale analyses. Such analysis requires a full calibrated model at the large
Carlo on 5000 simulations of a sewer pipe network and defined that city-scale that is re-calibrated under various simplification settings (e.g.,
water level in the pipe and duration of the clogging process have a reduced parameters and processes incorporated in the model) and,
significant impact on the accuracy of the exfiltration predictions. Peche separately, for smaller regions within the city, where distinctions in the
et al. (2017) performed 60 repetitive simulations to develop a functional sewer infrastructure and hydrogeological conditions are known.
relationship between sewer exfiltration fluxes, pipe water level and soil Applying a calibrated large-scale model on different ‘scenarios’, i.e.,
water pressure below the colmation layer, such that uncertainties including and excluding key model parameters and processes, is another
regarding the qualitative distribution of that relationship can be reduced research direction which may help to define the driving factors for sewer
significantly. Studies of Lee et al. (2015), Roehrdanz et al. (2017), and exfiltration modelling (e.g., Bhaskar et al., 2015; Peche, 2019).
Dvory et al. (2018) stand among few modelling studies incorporating Analyzing and comparing the results of different model framework
important contaminants from sewer leakage. However, the calibration is simplifications and scenarios with the results of the original full city-
constrained by the availability of short-term groundwater observations scale model may help to better evaluate the “accuracy” of various
in small-scale regions and the lack of information on the background development settings and calibration schemes of sewer exfiltration
concentration of contaminants in the groundwater. models, as well as to define parameters and processes that may be
Overall, more studies combining monitoring and modelling of sewer generalized or have negligible effects on groundwater. Addressing these
exfiltration at pipe and city scales are needed to further verify the issues can help to develop more case studies of large-scale urban models
important parameters in sewer exfiltration models and their practical incorporating sewer exfiltration processes, particularly in regions with a
default ranges, as well as to further test the uncertainty approaches in significant data-scarce issue.
predicting sewer exfiltration impacts of large-scale models.
5.5. Modelling of sewer exfiltration from private household sources
5.4. Upscaling and generalization of sewer exfiltration
While inspection and rehabilitation of public sewer systems are
Some efforts have been conducted to develop approaches to allow managed and regulated, information on house connection sewers is
upscaling the results of models from single defect points or pipe lengths often limited as it locates in private properties (Ellis and Bertrand-
to catchment scales in order to provide information for decision support Krajewski, 2010). As a result, data of sewer exfiltration from private
and rehabilitation implementation of urban wastewater systems. First sewer networks as well as privately conducted monitoring and reha­
efforts towards upscaling observed exfiltration rates aim at representing bilitation data are, in most cases, not available to third parties and for
sewer exfiltration on the whole sewer network scale. Dohmann et al. exfiltration modelling studies. Exfiltration from defective sewers of
(1999) quantified exfiltration rates at field scale and upscaled their re­ private sewer systems is often indirectly estimated from population
sults based on the statistical information on the structural condition of statistics (e.g., Ducci, 2018; Vystana et al., 2019) or using a population-
the sewer system. CCTV-based inspection data have been extrapolated to based risk/probability matrix (e.g., Ellis and Bertrand-Krajewski, 2010;
the city district scale in a study by Wolf and Hötztl (2007). However, Kaddoura and Zayed, 2018). Results of a few available studies indicate
results upscaled from field studies tend to overestimate the exfiltration high variability in damage rates of private sewers and hardly monitored
at the sewer network scale since the results are derived from observed exfiltration rates (e.g., Kaddoura and Zayed, 2018; Vystana et al., 2019;
pipes with significant damages and defects (Rutsch et al., 2008). Nguyen and Venohr, 2021). This lack of knowledge leads to the question
Possible solutions for large-scale modelling of sewer exfiltration on whether sewer exfiltration remains underestimated in urban catch­
target to closer implement the characteristics and behavior of defective ments if restricted to public sewers (Nguyen and Venohr, 2021). It is
sewer networks. Few studies have been conducted in this direction, but recognized that modelling exfiltration from private sewers may contain
the cases are still limited. For instance, Karpf et al. (2009) defined a fixed high uncertainties due to the considerable lack of appropriate data. The
value of colmation properties under stable conditions and proposed the topic of sewer exfiltration from private house connections represents a
application of a standardized sewer leakage area to ease the use of a pipe research gap that needs to be addressed thoroughly as an integral part of
to catchment scale modelling. This approach was further developed in the overall sewer leakage problems in urban areas.
the study by Peche et al. (2017), in which an upscaling of leakage, using
transfer functions, was derived from numerical simulations. This func­
tional relationship represents sewer exfiltration flux through the

12
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

6. Conclusions depth analyses of source, degradation and transport pathways is


required for different types of sewers. In most studies, private sewer
Sewer exfiltration models proved to be effective and important tools networks and household connections are missing and should be
for decision making and rehabilitation planning of urban wastewater considered. Upscaling and generalization of calibrated model results
environment. The following main findings can be drawn from the from the conduit to city scales could help to define the certain model
literature review on sewer exfiltration modelling: simplification schemes that are applicable for urban systems with a
significant data shortage.
• Sewer exfiltration involves various malfunction reasons caused by
pipe defects. Ageing-bound failure of structural integrity remains the
primary factor causing sewer leakage incidences in sewer systems Declaration of Competing Interest
worldwide.
• With the pressure from rapid urbanisation and climate change, the The authors declare that they have no known competing financial
topic of sewer exfiltration modelling has undergone a significant interests or personal relationships that could have appeared to influence
expansion in various directions, which can be broadly grouped into the work reported in this paper.
statistical, physically-based, and other approaches. Initial models
focus on identifying important variables affecting sewer exfiltration, Acknowledgements
the probability of sewer failures and potential impacts of sewer
exfiltration on groundwater in small case studies. More modern tools This work is part of the interdisciplinary project “Agro-Environ­
aim to represente the real-time spatial pattern of sewer exfiltration mental measures in Germany” (AGRUM-DE). The project is funded by
and to improve predictions of the failure of large sewer systems for the German Working Group on Water Issues of the Federal States and the
better rehabilitation and inspection plans. Federal Government, Germany (LAWA, Grant number O.4.19), which
• The general on-going trend of sewer exfiltration models are in favor aims at quantifying the water and nutrient emissions from multiple
of more integrated approaches, i.e., combining the strength of sources for all German river systems. The authors thank the State part­
diverse modelling techniques, upscaling tools, together with more ners for proposing the necessity of this research and providing com­
data from real sewer experiments developed at various spatial, ments to improve the content of this paper.
temporal, and geographical boundaries for a better definition of
model parameters and representation of key processes, as well as for References
confirming the validity of current approaches and advancing
methods for model calibration and uncertainty analyses. Amick, R.S., Burgess, E.H., 2000. Exfiltration in Sewer Systems. Report EPA/600/R-01/
034, US Environmental Protection Agency, National Risk Management Research
Laboratory, Ohio, US.
Modelling of sewer exfiltration processes and linking this emission to Ariaratnam, S., El-Assaly, A., Yang, Y., 2001. Assessment of infrastructure inspection
groundwater impacts at city or larger scales, yet, show considerable needs using logistic models. J. Infrastruct. Syst. 7 (4), 160–165. https://doi.org/
10.1061/(ASCE)1076-0342(2001)7:4(160).
short-comings. Here, lacking details on sewer conditions (including age/ Attard, G., Winiarski, T., Rossier, Y., Eisenlohr, L., 2016. Review: impact of underground
damage but also depth in the ground) together with a coupled sewer structures on the flow of urban groundwater. Hydrogeol. J. 24 (1), 5–19. https://doi.
pipe-groundwater modelling are the two major limiting factors. For org/10.1007/s10040-015-1317-3.
Baik, H.-S., Jeong, H.S., Abraham, D.M., 2006. Estimating transition probabilities in
future research, a more systematic approach with more model compo­ Markov chain-based deterioration models for management of wastewater systems.
nents and tools is recommended when modelling the water losses from J. Water Res. PL-ASCE 132 (1), 15–24. https://doi.org/10.1061/(ASCE)0733-9496
defective sewers into groundwater environment. At the pipe scale, (2006)132:1(15).
Barrett, M.H., Lerner, D.N., Hiscock, K.M., Pedley, S., Tellam, J.H., 1997. The use of
advanced machine learning and data-driven techniques could be used to
marker species to establish the impact of the City of Nottigham, UK on the quantity
find more accurately the deterioration state of the entire sewer network. and quality of its underlying groundwater. In: Chilton, J. (Ed.), Groundwater in the
At this scale, models incorporating hydraulic interactions between pipe Urban Environment. Balkema Publications, Rotterdam, The Netherlands.
Baur, R., Herz, R., 2002. Selective inspection planning with ageing forecast for sewer
and groundwater often assume constant damage rates or use simplified
types. Water Sci. Technol. 46 (6–7), 389–396. https://doi.org/10.2166/
leakage equations based on single pipe properties (e.g., pipe length, pipe wst.2002.0704.
diameter, or pipe age). Machine learning models could also allow an Beal, C., Gardner, T., Menzies, N.W., Rassam, D., Vieritz, A., 2004. Prediction of steady-
improved understanding of key driving factors affecting the sewer state flux through variably saturated zones within a septic absorption trench. 3rd
Australian New Zealand Soils Conference. University of Sydney, Australia.
leakage to groundwater, as the importance of these factors in sewer Benedetti, L., Langeveld, J., Comeau, A., Corominas, L., Daigger, G., Martin, C.,
leakage models at the scale of the whole sewer network might vary from Mikkelsen, P.S., Vezzaro, L., Weijers, S., Vanrolleghem, P.A., 2013. Modelling and
those defined in small-scale studies at major leakages. Among major monitoring of integrated urban wastewater systems: review on status and
perspectives. Water Sci. Technol. 68 (6), 1203–1215. https://doi.org/10.2166/
parameters affecting the modelling of hydraulic and solute contami­ wst.2013.397.
nants from sewer exfiltration, colmation layer properties and hydraulic Berardi, L., Kapelan, Z., Giustolisi, O., Savic, D.A., 2008. Development of pipe
potential gradient between pipe and aquifer strata are of critical deterioration models for water distribution systems using EPR. J. Hydroinform. 10
(3), 113–126. https://doi.org/10.2166/hydro.2008.012.
importance inquiring further research. While the hydraulic potential Bhaskar, A.S., Welty, C., Maxwell, R.M., Miller, A.J., 2015. Untangling the effects of
distribution in time and space may be calculated with numerical models urban development on subsurface storage in Baltimore. Water Resour. Res. 51,
on the basis of a well-known initial condition, a major challenge is the 1158–1181. https://doi.org/10.1002/2014WR016039.
Blackwood, D.J., Ellis, J.B., Revitt, D.M., Gilmour, D.J., Stainer, A., 2005. Factors
accurate prediction of the extent and hydraulic properties of the col­
influencing exfiltration processes in sewers. Water Sci. Technol. 51 (2), 147–154.
mation layer. In present times, no reliable model for the prediction of Blackwood, D.J., Ellis, J. B., Revitt, D.M., Gilmour, D.J., Stainer, A., 2006. Exfiltration
these properties exist and colmation layer properties have to be deter­ from sewers: is it a serious problem? in: Mikkelsen, P. S. (ed.) Proceedings of the IWA
10th International Conference on Urban Drainage, CD-ROM. Copenhagen, Denmark.
mined with probing or physical experiments. Complete physically-
IWA Publishing, London. ISBN 9781843395744.
based, integrated surface water, pipe and urban infrastructure, hydro­ Boukhemacha, M.A., Gogu, C.R., Serpescu, I., Gaitanaru, D., Bica, I., 2015.
geological, and groundwater model framework based on an accurate A hydrogeological conceptual approach to study urban groundwater flow in
determination of core water balance components is essential to analyse Bucharest city, Romania. Hydrogeol. J. 23, 437–450. https://doi.org/10.1007/
s10040-014-1220-3.
the effects of sewer exfiltration on groundwater quantitatively rather Brunner, P., Cook, P.G., Simmons, C.T., 2009. Hydrogeologic controls on disconnection
than only qualitatively. A more thorough model calibration incorpo­ between surface water and groundwater. Water Resour. Res. 45 (1), W01422.
rating the uncertainty analyses is another integral part of future sewer https://doi.org/10.1029/2008WR006953.
Burn, S., Eiswirth, M., Correll, R., Cronin, A., DeSilva, D., Diaper, C. , Dillon, P., Mohrlok,
exfiltration models at large-scales. As groundwater monitoring is still U., Morris, B., Rueedi, J., Wolf, L., Vizintin, G., Vött, U., 2007. Urban infrastructure
often patchy, the consideration of additional contaminates and more in- and its impact on groundwater contamination. in: Howard, K.W.F. 2007. Urban

13
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

groundwater – meeting the challenge. Selected papers from the 32nd International Fu, B., Merritt, W.S., Croke, B.F.W., Weber, T.R., Jakeman, A.J., 2019. A review of
Geological Congress (IGC), Florence, Italy, August 2004. ISBN 0-203-94705-3. catchment-scale water quality and erosion models and a synthesis of future
Burston, M.W., Mehrdad, I., Nazari, M., Bishop, P.K., Lerner, D.N., 1993. Pollution of prospects. Environ Model Softw. 114, 75–97. https://doi.org/10.1016/j.
groundwater in the Coventry region (UK) by chlorinated hydrocarbon solvents. envsoft.2018.12.008.
J. Hydrol. 149, 137–161. https://doi.org/10.1016/0022-1694(93)90104-H. Fuchs, L., Scheffer, C., Verworn, H.R., 2004. Modellbeschreibung: Hystem-Extran 6.
Caradot, N., Sonnenberg, H., Kropp, I., Ringe, A., Denhez, S., Hartmann, A., Rouault, P., Institut für technisch wissenschaftliche Hydrologie GmbH (ITWH), Hannover.
2017. The relevance of sewer deterioration modelling to support asset management Fuchs-Hanusch, D., Günther, M., Möderl, M., Muschalla, D., 2015. Cause and effect
strategies. Urban Water J. 14, 1007–1015. https://doi.org/10.1080/ oriented sewer degradation evaluation to support scheduled inspection planning.
1573062X.2017.1325497. Water Sci. Technol. 72 (7), 1176–1183. https://doi.org/10.2166/wst.2015.320.
Caradot, N., Riechel, M., Fesneau, M., Hernández, N., Torres, A., Sonnenberg, H., Geyer, T., Birk, S., Licha, T., Liedl, R., Sauter, M., 2007. Multitracer test approach to
Eckert, E., Lengemann, N., Waschnewski, J., Rouault, P., 2018. Practical characterize reactive transport in karst aquifers. Ground Water 45, 36–45. https://
benchmarking of statistical and machine learning models for predicting the doi.org/10.1111/j.1745-6584.2006.00261.x.
condition of sewer pipes in Berlin, Germany. J. Hydroinf. 20, 1131–1147. https:// Gibson, J., Karney, B., Guo, Y., 2019. Predicting health risks from intrusion into drinking
doi.org/10.2166/hydro.2018.217. water pipes over time. J. Water Resour. Plann. Manage. 145 (3), 04019001. https://
Chisala, B.N., Lerner, D.N., 2008. Distribution of sewer exfiltration to urban doi.org/10.1061/(ASCE)WR.1943-5452.0001039.
groundwater. Proc. Inst. Civ. Eng.-Water Manage. 161 (6), 333–341. https://doi. Gogu, C.R., Gaitanaru, D., Boukhemacha, M.A., Serpescu, I., Litescu, L., Zaharia, V.,
org/10.1680/wama.2008.161.6.333. Moldovan, A., Mihailovici, M.J., 2017. Urban hydrogeology studies in Bucharest
Chughtai, F., Zayed, T., 2008. Infrastructure condition prediction models for sustainable city, Romania. Procedia Eng. 209, 135–142. https://doi.org/10.1016/j.
sewer pipelines. J. Perform. Constr. Facil. 22 (5), 333–341. https://doi.org/ proeng.2017.11.139.
10.1061/(ASCE)0887-3828(2008)22:5(333). Grimmeisen, F., Lehmann, M.F., Liesch, T., Goeppert, N., Klinger, J., Zopfi, J.,
Clemens, F.H.L.R., Stanić, N., Van der Schoot, W., Langeveld, J.G., Lepot, M., 2015. Goldscheider, N., 2017. Isotopic constraints on water source mixing, network
Uncertainties associated with laser profiling of concrete sewer pipes for the leakage and contamination in an urban groundwater system. Sci. Total Environ. 583,
quantification of the interior Geometry. Struct. Infrastruct. E. 11, 1218–1239. 202–213. https://doi.org/10.1016/j.scitotenv.2017.01.054.
https://doi.org/10.1080/15732479.2014.945466. Gruenfeld, M., 2000. Exfiltration in Sewer Systems: A Draft Report to the U.S.
Davies, J.P., Clarke, B.A., Whiter, J.T., Cunningham, R.J., Leidi, A., 2001. The structural Environmental Protection Agency. National Risk Management Research Laboratory,
condition of rigid sewer pipes: a statistical investigation. Urban Water J. 3, 145–154. Edison, New Jersey 08837.
https://doi.org/10.1016/S1462-0758(01)00036-X. Guo, S., Yang, Y., Zang, Y., 2017. An approximate model on three-dimensional
de Saint-Venant, A.B., 1871. Théorie et équations générales du mouvement non- groundwater infiltration in sewer systems. Water Sci. Technol. 75 (2), 306–312.
permanent des eaux courantes. Comptes Rendus des séances de l’Académie des https://doi.org/10.2166/wst.2016.518.
Sciences 73, 147–154. Han, Z., 1997. Modellierung von Stofftransport und -umsetzung in kleinen
Decker, J., 1994. Pollution load of subsoil, groundwater and surface water by leaky Einzugsgebieten unter Berücksichtigung der Wechselbeziehung zwischen
sewers. Proceedings from Hydrotop. 94, Marseille, France. Grundwasser und Oberflächenwasser. in: Mitteilungen, (84), Institut für
DeSilva, D., Burn, S., Tjandraatmadja, G., Moglia, M., Davis, P., Wolf, L., Held, I., Wasserwirtschaft, Hydrologie und landwirtschaftlichen Wasserbau, University of
Vollertsen, J., Williams, W., Hafskjold, L., 2005. Sustainable management of leakage Hanover, Germany, pp. 145–330.
from wastewater pipelines. Water Sci. Technol. 52 (12), 189–198. https://doi.org/ Hao, T., Rogers, C.D.F., Metje, N., Chapman, D.N., Muggleton, J.M., Foo, K.Y., Wang, P.,
10.2166/wst.2005.0459. Pennock, S.R., Atkins, P.R., Swingler, S.G., Parker, J., Costello, S.B., Burrow, M.P.N.,
DeSilva, D., Burn, S., Moglia, M., Tjandraatmadja, G., Gould, S., Sadler, S., 2007. Chapter Ansprach, J.H., Armitage, R.J., Cohn, A.G., Goddard, A.G., Lewin, P.L., Orlando, G.,
2 the models: network exfiltration and infiltration model NEIMO. In: Wolf, L., Pedfern, M.A., Royal, A.C.D., Saul, A.J., 2012. Condition assessment of the Buried
Morris, B., Burn, S. (Eds.), Urban Water Resources Toolbox. IWA Publishing. Utility Service Infrastructure. Tunn. Undergr. Space Technol. 28, 331–344. https://
Dohmann, M., Decker, J., Menzenbach, B., 1999. Water Contamination Due to Sewer doi.org/10.1016/j.tust.2011.10.011.
Leakage. Springer, Berlin, Heidelberg (in German). Harvey, R.R., McBean, E.A., 2014. Predicting the structural condition of individual
Dohmann, M., 1995. Vergleich der Boden- und Grundwasserbelastung undichter Kanäle sanitary sewer pipes with random forests. Can. J. Civ. Eng. 41, 294–303. https://doi.
mit anderen Schmutzstoffeinträgen. Gesellschaft zur Förderung d. org/10.1139/cjce-2013-0431.
Siedlungswasserwirtschaft an d. RWTH Aachen, Gewässerschutz Wasser Abwasser Hatam, F., Blokker, M., Besner, M.C., Ebacher, G., Prévost, M., 2019. Using Nodal
(GWA) 152. Infection Risks to guide interventions following accidental intrusion due to sustained
Ducci, D., 2018. An Easy-to-Use Method for Assessing Nitrate Contamination low pressure events in a drinking water distribution system. Water 11, 1372. https://
Susceptibility in Groundwater. Wiley. Hindawi. DOI:10.1155/2018/1371825. doi.org/10.3390/w11071372.
Duchesne, S., Beardsell, G., Villeneuve, J.P., Trombou, B., Bouchard, K., 2012. A survival Heinrich, A., 2007. Transferability of exfiltration rates from sewer systems. J. Soils
analysis model for sewer pipe structural deterioration. Comput.-Aided Civ. Sediments 7, 69–74. https://doi.org/10.1065/jss2007.02.209.
Infrastruct. Eng. 28 (2), 146–160. https://doi.org/10.1111/j.1467- Hermosilla, R.G., 2012. The Guatemala City sinkhole collapses. Carbonates Evaporites 27
8667.2012.00773.x. (2), 103–107. https://doi.org/10.1007/s13146-011-0074-1.
Dvory, N.Z., Livshitz, Y., Kuznetsov, M., Adar, E., Yakirevich, A., 2016. The effect of Hernández, N., Caradot, N., Sonnenberg, H., Rouault, P., Torres, A., 2018. Support Tools
hydrogeological conditions on variability and dynamic of groundwater recharge in a to predict the critical structural condition of uninspected pipes for case studies of
carbonate aquifer at local scale. J. Hydrol. 535, 480–494. https://doi.org/10.1016/j. Germany and Colombia. Water Pract. Technol. 13, 794–802. https://doi.org/
jhydrol.2016.02.011. 10.2166/wpt.2018.085.
Dvory, N.Z., Kuznetsov, M., Livshitz, Y., Gasser, G., Pankratov, I., Lev, O., Adar, E., Herz, R., Hochstrate, K., 1987. Erneuerungsstrategien für städtische Infrastrukturnetze.
Yakirevich, A., 2018. Modeling sewage leakage and transport in carbonate aquifer Jahrbuch für Regionalwissenschaften Jg.8. 67–105.
using carbamazepine as an indicator. Water Res. 128, 157–170. https://doi.org/ Hoffman, J.M., Lerner, D.N., 1992. Leak free sewers – who needs them? Water Waste
10.1016/j.watres.2017.10.044. Treatment 35 (8), 18–19.
Egger, C., Scheidegger, A., Reichert, P., Maurer, M., 2013. Sewer deterioration modeling Hopkins, K.G., Bain, D.J., 2018. Research Note: mapping spatial patterns in sewer age,
with condition data lacking historical records. Water Res. 47, 6762–6779. https:// material, and proximity to surface waterways to infer sewer leakage hotspots. Landsc
doi.org/10.1016/j.watres.2013.09.010. Urban Plan. 170, 320–324. https://doi.org/10.1016/j.landurbplan.2017.04.011.
Eiswirth, M., Hötzl, H., 1997. The impact of leaking sewers on urban groundwater, in: Howard, K.W.F., 2007. Urban groundwater - meeting the challenge: IAH Selected Papers
Chilton, J. (Eds), Groundwater in the urban environment, ISBN-13: 978- on Hydrogeology 8. ISBN 9780415407458.
0415434935, pp 399-404. Jaganathan, A.P., Allouche, E., Simicevic, N., 2010. Numerical modeling and
Eiswirth, M., Hötzl, H., Burn, L.S., Gray, S. and Mitchell, V.G., 2001. Contaminant loads experimental evaluation of a time domain UWB technique for soil void detection.
within the urban water system – scenario analyses and new strategies. in: K.P. Seiler Tunn. Undergr. Space Technol. 25 (6), 652–659. https://doi.org/10.1016/j.
and S. Wohnlich (Eds.), New Approaches to Characterising Groundwater Flow: vol. tust.2009.08.006.
1, 493–498. Netherlands. Jiang, G., Keller, J., Bond, P.L., Yuan, Z., 2016. Predicting concrete corrosion of sewers
Ellis, J.B., Bertrand-Krajewski, J.L., 2010. Assessing Infiltration and Exfiltration on the using artificial neural network. Water Res. 92, 52–60. https://doi.org/10.1016/j.
Performance of Urban Sewer Systems (APUSS). IWA Publishing, London, UK. watres.2016.01.029.
Ellis, J.B., Revitt, D.M., 2001. Sewer losses and interactions with groundwater quality. Kaddoura, K., Zayed, T., 2018. An integrated assessment approach to prevent risk of
Wat. Sci. Tech. 45 (3), 195–202. https://doi.org/10.2166/wst.2002.0079. sewer exfiltration. Sustain. Cities Soc. 41, 576–586. https://doi.org/10.1016/j.
Ellis, J.B., Revitt, D.M., Lister, P., Willgres, S.C., Buckley, A., 2003. Experimental studies scs.2018.05.032.
of sewer exfiltration. Water Sci. Technol. 47 (4), 61–68. Kang, M.J., Song, S., Park, D., Choi, C., 2017. Detection of cavities around concrete
Ellis, J., Revitt, D., Vollertsen, J., Blackwood, D., 2009. Sewer exfiltration and the sewage pipelines using impact-echo method. Tunn. Undergr. Space Technol. 65,
colmation layer. Water Sci. Technol. 59 (11), 2273–2280. https://doi.org/10.2166/ 1–11. https://doi.org/10.1016/j.tust.2017.02.002.
wst.2009.271. Karamouz, M., Ahmadi, A., Akhbari, M., 2011. Groundwater Hydrology: Engineering,
Fenz, R., Blaschke, A.P., Clara, M., Kroiss, H., Mascher, D., Zessner, M., 2005. Monitoring Planning, and Management, second ed. CRC Press, p. 750.
of carbamazepine concentrations in wastewater and groundwater to quantify sewer Karpf, C., 2012. Modellierung der Interaktion zwischen Grundwasser und Kanalisation.
leakage. Water Sci. Technol. 52 (5), 205–213. https://doi.org/10.2166/ PhD thesis. TU Dresden.
wst.2005.0135. Karpf, C., Krebs, P., 2004. Sewers as drainage systems – quantification of groundwater
Franz, T., Krebs, P., 2005. Statistical methods towards more efficient infiltration infiltration. Proc. Vol. 2, pp. 969–975, NOVATECH conference, Lyon, France.
measurements, in: Proc. 10th International conference of Urban Drainage, Karpf, C., Krebs, P., 2005. Application of a leakage model to assess exfiltration from
Copenhagen, Denmark. sewers. Water Sci. Technol. 52 (5), 225–231. https://doi.org/10.2166/
wst.2005.0137.

14
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

Karpf, C., Krebs, P., 2011. A new sewage exfiltration model – parameters and calibration. Peche, A., Graf, T., Fuchs, L., Neuweiler, I., 2019. Physically based modeling of
Water Sci. Technol. 63, 2294–2299. https://doi.org/10.2166/wst.2011.167. stormwater pipe leakage in an urban catchment. J. Hydrol. 573, 778–793. https://
Karpf, C., Traenckner, J., Krebs, P., 2009. Hydraulic modelling of sewage exfiltration. doi.org/10.1016/j.jhydrol.2019.03.016.
Water Sci. Technol. 59 (8), 1559–1565. https://doi.org/10.2166/wst.2009.172. Peche, A. 2019. Numerical modeling of pipe leakage in variably saturated soil. Doctoral
Karpf, C., Hoeft, S., Scheffer, C., Fuchs, L., Krebs, P., 2011. Groundwater infiltration, dissertation, Hannover: Institut für Strömungsmechanik und Umweltphysik im
surface water inflow and sewerage exfiltration considering hydrodynamic conditions Bauwesen der Leibniz Universität Hannover.
in sewer systems. Water Sci. Technol. 63 (9), 1841–1848. https://doi.org/10.2166/ Rauch, W., Stegner, T., 1994. The colmation of leaks in sewer systems during dry
wst.2011. 388. weather flow. Water Sci. Technol. 30 (1), 205–210. https://doi.org/10.2166/
Kenway, S., Gregory, A., McMahon, J., 2011. Urban water mass balance analysis. J. Ind. wst.1994.0022.
Ecol. 15 (5), 693–706. https://doi.org/10.1111/j.1530-9290.2011.00357.x. Reynolds, J.H., Barrett, H.M., 2003. A review of the effects of sewer leakage on
Kettler, A.J., Goulter, I.C., 1985. An analysis of pipe breakage in urban water distribution groundwater quality. J. Chart. Inst. Water Environ. Manage. 17 (1), 34–39. https://
networks. Can. J. Civil Eng. 12 (2), 286–293. https://doi.org/10.1139/l85-030. doi.org/10.1111/j.1747-6593.2003.tb00428.x.
Khan, Z., Zayed, T., Moselhi, O., 2010. Structural condition assessment of sewer Richards, L.A., 1931. Capillary conduction of liquids through porous mediums. J Appl
pipelines. J. Perform. Constr. Facil. 24, 170–179. https://doi.org/10.1061/(ASCE) Phys. 1 (5), 318–333. https://doi.org/10.1063/1.1745010.
CF.1943-5509.0000081. Rieckermann, J., Bareš, V., Kracht, O., Braun, D., Gujer, W., 2007. Estimating sewer
Kidmose, J., Troldborg, L., Refsgaard, J.C., Bischoff, N., 2015. Coupling of a distributed leakage from continuous tracer experiments. Water Res. 41, 1960–1972. https://doi.
hydrological model with an urban storm water model for impact analysis of forced org/10.1016/j.watres.2007.01.024.
infiltration. J. Hydrol. 525, 506–520. https://doi.org/10.1016/j. Rieckermann, J., Borsuk, M.E., Sydler, D., Gujer, W., Reichert, P., 2010. Bayesian
jhydrol.2015.04.007. experimental design of tracer studies to monitor wastewater leakage from sewer
Kirkham, R., Kearney, P.D., Rogers, K.J., Mashford, J., 2000. PIRAT - a System for networks. Water Resour. Res. 46, W08513. https://doi.org/10.1029/
quantitative sewer pipe assessment. Int. J. Robotics Res. 19, 1033–1053. https://doi. 2009WR008630.
org/10.1177/02783640022067959. Rodriguez, F., Delliou, A.L.L., Andrieu, H., Gironás, J., 2020. Groundwater contribution
Klinger, J., Turkovic, R., Wolf, L., Hotzl, H., 2007. Long term investigations of the to sewer network baseflow in an urban catchment-case study of Pin Sec Catchment,
colmation processes at a real sewer defect. Geophys. Res. Abstr. 9, 1–3. Nantes, France. Water 12 (3), 689. https://doi.org/10.3390/w12030689.
Laakso, T., Kokkonen, T., Mellin, I., Vahala, R., 2018. Sewer condition prediction and Roehrdanz, P.R., Feraud, M., Lee, D.G., Means, J.C., Snyder, S.A., Holden, P.A., 2017.
analysis of explanatory factors. Water 10, 1239. https://doi.org/10.3390/ Spatial Models of sewer pipe leakage predict the occurrence of wastewater indicators
w10091239. in shallow urban groundwater. Environ. Sci. Technol. 51, 1213–1223. https://doi.
Le Gat, Y., 2006. Modeling the deterioration process of drainage pipelines. Urban Water org/10.1021/acs.est.6b05015.
J. 5, 97–106. https://doi.org/10.1080/15730620801939398. Rokstad, M.M., Ugarelli, R.M., 2015. Evaluating the role of deterioration models for
Le Gat, Y., 2008. Modelling the deterioration process of drainage pipelines. Urban Water condition assessment of sewers. J. Hydroinf. 17, 789–804. https://doi.org/10.2166/
J. 5, 97–106. https://doi.org/10.1080/15730620801939398. hydro.2015.122.
Le Gat, Y., Eisenbeis, P., 2000. Using maintenance records to forecast failures in water Rokstad, M.M., Ugarelli, R.M., 2016. Improving the benefits of sewer condition
networks. Urban Water J. 2 (3), 173–181. https://doi.org/10.1016/S1462-0758(00) deterioration modelling through information content analysis. Water Sci. Technol.
00057-1. 74, 2270–2279. https://doi.org/10.2166/wst.2016.419.
Lee, D.G., Roehrdanz, P.R., Feraud, M., Ervin, J., Anumol, T., Jia, A., Park, M., Tamez, C., Rostum, J., Baur, R., Saegrov, S., Horold, S., Schiling, W., 1999. Predictive service-life
Morelius, E.W., Gardea-Torresdey, J.L., Izbicki, J., Means, J.C., Snyder, S.A., models for urban water infrastructure management. In: Joliffe, I.B., Ball, J.E. (Eds.),
Holden, P.A., 2015. Wastewater compounds in urban shallow groundwater wells Proceedings of the 8th International Conference on Urban Storm Drainage,
correspond to exfiltration probabilities of nearby sewers. Water Res. 85, 467–475. pp. 594–601.
https://doi.org/10.1016/j.watres.2015.08.048. Rushton, K.R., Tomlinson, L.M., 1979. Possible mechanisms for leakage between aquifers
Lerner, D.N., 2002. Identifying and quantifying urban recharge: a review. Hydrogeol. J. and rivers. J. Hydrol. 40 (1–2), 49–65. https://doi.org/10.1016/0022-1694(79)
10, 143–152. https://doi.org/10.1007/s10040-001-0177-1. 90087-8.
Li, T., Zhou, Y.C., Li, H., 2008. Quantifying non-stormwater discharges to stormwater Rutsch, M., Rieckermann, J., Cullmann, J., Ellis, J.B., Vollertsen, J., Krebs, P., 2008.
systems with model analysis. J. Environ. Eng. 134, 928–932. https://doi.org/ Towards a better understanding of sewer exfiltration. Water Res. 42 (10–11),
10.1061/(ASCE)0733-9372(2008)134:11(928). 2385–2394. https://doi.org/10.1016/j.watres.2008.01.019.
Liu, Z., Sadiq, R., Rajani, B., Najjaran, H., 2010. Exploring the relationship between soil Rutsch, M., 2006. Assessment of sewer leakage by means of exfiltration measurements
properties and deterioration of metallic pipes using predictive data mining methods. and modelling tests. Doctoral dissertation, Dresden. pp. 204.
J. Comput. Civil Eng. 24 (3), 289–301. https://doi.org/10.1061/(ASCE)CP.1943- Salman, B., Salem, O., 2012. Risk assessment of wastewater collection lines using failure
5487.0000032. models and criticality ratings. J. Pipel. Syst. Eng. Pract. 3, 68–76. https://doi.org/
Ly, D.M., Chui, T.F.M., 2012. Modeling sewage leakage to surrounding groundwater and 10.1061/(ASCE)PS.1949-1204.0000100.
stormwater drains. Water Sci. Technol. 66 (12), 2659–2665. https://doi.org/ Schulz, N., Baur, R., Krebs, P., 2005. Integrated modelling for the evaluation of
10.2166/wst.2012.496. infiltration effects. Water Sci. Technol. 52 (5), 215–223. https://doi.org/10.2166/
Mashford, J., Marlow, D., Tran, D., May, R., 2011. Prediction of sewer condition grade wst.2005.0136.
using support vector machines. J. Comput. Civil Eng. 25, 283–290. https://doi.org/ Selvakumar, A., Field, R., Burgess, E., Amick, R., 2004. Exfiltration in sanitary sewer
10.1061/(ASCE)CP.1943-5487.0000089. systems in the US. Urban Water J. 1 (3), 227–234. https://doi.org/10.1080/
Maxwell, R.M., Kollet, S.J., Smith, S.G., Woodward, C.S., Falgout, R.D., Baldwin, C., Bosl, 15730620410001732017.
W.J., Hornung, R., Ashby, S., 2009. ParFlow User’s Manual. International Ground Sophocleous, M., 2002. Interactions between groundwater and surface water: the state of
Water Modeling Center Report GWMI 2009-01. pp. 129. the science. Hydrogeol. J. 10 (1), 52–67.
Mitchell, V.G., McMahon, T.A., Mein, R.G., 2003. Components of the total water balance Sousa, V., Matos, J.P., Matias, N., 2014. Evaluation of artificial intelligence tool
of an urban catchment. J. Environ. Manage. 32 (6), 735–746. https://doi.org/ performance and uncertainty for predicting sewer structural condition. Autom.
10.1007/s00267-003-2062-2. Constr. 44, 84–91. https://doi.org/10.1016/j.autcon.2014.04.004.
Mohrlok, U., Wolf, L., Klinger, J., 2008. Quantification of infiltration processes in urban Stegeman, B., Langeveld, J., Bogaard, T., Clemens, F., 2019. Detection of exfiltration in
areas by accounting for spatial parameter variability. J. Soils Sediments 8 (1), 34–42. sewer systems with tracers. In: Mannina, G. (Ed.), New Trends in Urban Drainage
https://doi.org/10.1065/jss2007.05.225. Modelling. Green Energy and Technology. https://doi.org/10.1007/978-3-319-
Morris, B.L., Darling, W.G., Cronin, A.A., Rueedi, J., Whitehead, E.J., Gooddy, D.C., 99867-1_141.
2006. Assessing the impact of modern recharge on a sandstone aquifer beneath a Syachrani, S., Jeong, H., Chung, C., 2013. Decision tree–based deterioration model for
suburb of Doncaster, UK. Hydrogeol. J. 14, 979–997. https://doi.org/10.1007/ buried wastewater pipelines. J. Perform. Constr. Facil. 27 (5), 633–645. https://doi.
s10040-006-0028-1. org/10.1061/(ASCE)CF.1943-5509.0000349.
Najafi, M., Kulandaivel, G., 2005. Pipeline condition prediction using neural network Thompson, S., Kulessa, B., Luckman, A., 2012. Integrated Electrical Resistivity
models. Proc., Pipelines 2005, ASCE, Reston, VA, pp. 767–775. Tomography (ERT) and Self-potential (SP) Techniques for Assessing Hydrological
Nakayama, T., Watanabe, M., Tanji, K., Morioka, T., 2007. Effect of underground urban Processes within Glacial Lake Moraine Dams. J. Glaciol. 58, 849–858. https://doi.
structures on eutrophic coastal environment. Sci. Total Environ. 373, 270–288. org/10.3189/2012JoG11J235.
https://doi.org/10.1016/j.scitotenv.2006.11.033. Thorndahl, S., Balling, J.D., Larsen, U.B.B., 2016. Analysis and integrated modelling of
Nguyen, H.H., Venohr, M., 2021. Harmonized assessment of nutrient pollution from groundwater infiltration to sewer networks. Hydrolog. Process. 30 (18), 3228–3238.
urban systems including losses from sewer exfiltration: a case study in Germany. https://doi.org/10.1002/hyp.10847.
Environ. Sci. Pollut. Res. DOI:10.1007/s11356-021-12440-9. Tran, D., Perera, A.W.M., Davis, P., 2006. Application of probabilistic neural networks in
Nienhuis, J., De Haan, C., Langeveld, J.G., Klootwijk, M., Clemens, F.H.L.R., 2013. modeling structural deterioration of stormwater pipes. Urban Water J. 3 (3),
Assessment of detection limits of fiber-optic distributed temperature sensing for 175–184. https://doi.org/10.1080/15730620600961684.
detection of illicit connections. Water Sci. Technol. 67, 2712–2718. https://doi.org/ Trauth, R., Hahn, H.H., Xanthopoulos, C., 1995. In-situ method to determine water
10.2166/wst.2013.176. exchange between sewerage and groundwater. Abwassertechnik 4, 55–57 (In
Okubo, T., Matsumoto, J., 1983. Biological clogging of sand and changes of organic German).
constituents during artifcial recharge. Water Res. 17 (7), 813–821. https://doi.org/ Tscheikner-Gratl, F., Sitzenfrei, R., Rauch, W., Kleidorfer, M., 2016. Enhancement of
10.1016/0043-1354(83)90077-5. limited water supply network data for deterioration modelling and determination of
Peche, A., Graf, T., Fuchs, L., Neuweiler, I., 2017. A coupled approach for the three- rehabilitation rate. Struct. Infrastruct. E. 12, 366–380. https://doi.org/10.1080/
dimensional simulation of pipe leakage in variably saturated soil. J. Hydrol. 555, 15732479.2015.1017730.
569–585. https://doi.org/10.1016/j.jhydrol.2017.10.050. Tscheikner-Gratl, F., Caradot, N., Cherqui, F., Leitão, G.P., Ahmadi, M., Langeveld, J.G.,
Gat, Y.L., Scholten, L., Roghani, B., Rodríguez, J.P., Lepot, M., Stegeman, B.,

15
H.H. Nguyen et al. Journal of Hydrology 596 (2021) 126130

Heinrichsen, A., Kropp, I., Kerres, K., Almeida, M.C., Bach, P.M., Vitry, M.M., Wolf, L., Klinger, J., Held, I., Hötzl, H., 2006. Integrating groundwater into urban water
Marques, A.S., Simões, N.E., Rouault, P., Hernandez, N., Torres, A., Werey, C., management. Water Sci. Technol. 54 (6–7), 395–403. https://doi.org/10.2166/
Rulleau, B., Clemens, F., 2020. Sewer asset management – state of the art and wst.2006.614.
research needs. Urban Water J. 16, 662–675. https://doi.org/10.1080/ Wolf, L., Klinger, J., Hoetzl, H., Mohrlok, U., 2007. Quantifying mass fluxes from urban
1573062X.2020.1713382. drainage systems to the urban soil-aquifer system. J. Soils Sediments 7 (2), 85–95.
Tubau, I., Vázquez-Suñé, E., Carrera, J., Valhondo, C., Criollo, R., 2017. Quantification of https://doi.org/10.1065/jss2007.02.207.
groundwater recharge in urban environments. Sci. Total Environ. 59, 391–402. Wolf, L., Zwiener, C., Zemann, M., 2012. Tracking artificial sweeteners and
https://doi.org/10.1016/j.scitotenv.2017.03.118. pharmaceuticals introduced into urban groundwater by leaking sewer networks. Sci.
USEPA, 1989. Results of the Evaluation of Groundwater Impacts of Sewer Exfiltration. Total Environ. 430, 8–19. https://doi.org/10.1016/j.scitotenv.2012.04.059.
Municipal Facilities Division, Office of Water, US Environmental Protection Agency, Xu, Q., Liu, R., Chen, Q., Li, R., 2014a. Review on water leakage control in distribution
Washington, DC. networks and the associated environmental benefits. J. Environ. Sci. 26, 955–961.
Vizintin, G., Souvent, P., Veselic, M., Curk, B.C., 2009. Determination of urban https://doi.org/10.1016/S1001-0742(13)60569-0.
groundwater pollution in alluvial aquifer using linked process models considering Xu, Z., Yin, H., Li, H., 2014b. Quantification of non-stormwater flow entries into storm
urban water cycle. J. Hydrol. 377(3-4), 261-273. DOI:10.1016/j. drains using a water balance approach. Sci. Total Environ. 487, 381–388. https://
jhydrol.2009.08.025. doi.org/10.1016/j.scitotenv.2014.04.035.
Vollertsen, J., Hvitved-Jacobsen, T., 2003. Exfiltration from gravity sewers: a pilot scale Yang, Y., Lerner, D.N., Barrett, M.H., Tellam, J.H., 1999. Quantification of groundwater
study. Water Sci. Technol. 47 (4), 69–76. https://doi.org/10.2166/wst.2003.0223. recharge in the city of Nottingham. Environ. Geol. 38 (3), 183–198. https://doi.org/
Vystavna, Y., Diadin, D., Rossi, P.M., Gusyev, M., Hejzlar, J., Mehdlzadeh, R., Huneau, F., 10.1007/s002540050414.
2018. Quantification of water and sewage leakages from urban infrastructure into a Yin, X., Bouferguene, A., Al-Hussein, M., 2020. Data-driven sewer pipe data random
shallow aquifer in East Ukraine. Environ. Earth Sci. 77, 748. https://doi.org/ generation and validation. J. Constr. Eng. Manag. 146, 103181 https://doi.org/
10.1007/s12665-018-7936-y. 10.1016/j.autcon.2020.103181.
Wakida, F.T., Lerner, D.N., 2005. Non-agricultural sources of groundwater nitrate: a Younis, R., Knight, M., 2010. A probability model for investigating the trend of structural
review and case study. Water Res. 39, 3–16. https://doi.org/10.1016/j. deterioration in wastewater pipelines. Tunn. Undergr. Space Technol. 25 (6),
watres.2004.07.026. 670–680. https://doi.org/10.1016/j.tust.2010.05.007.
Wolf, L., Hötztl, H., 2007. Upscaling of Laboratory Results on Sewer Leakage and the Zoppou, C., 2001. Review of urban storm water models. Environ. Model Softw. 16,
Associated Uncertainty. International Contributions to Hydrogeology, Balkema, 195–231. https://doi.org/10.1016/S1364-8152(00)00084-0.
London.

16

You might also like