Towards a decentralized solution for sewer leakage detection – a review

The sewer system plays a critical role in the current urbanized world. As urbanization increases, the existing sewer system continuously expands due to new demands for residential, sanitary, and industrial sewer waters. Developed countries usually have a higher sewer infrastructure coverage rate in which complete network connections for multiple cities have already been achieved. In Europe, for example, according to the European Environment Agency (EEA 2021), 82% of the population on average is connected to sewer systems, which is increasing over time (EU-27). However, sewer pipes deteriorate due to physical, environmental, and operational factors, which ultimately create defects and ex-infiltration based on the groundwater elevation. These defects and exfiltration, sometimes, go unnoticed during the expansion of sewer infrastructures (Haurum & Moeslund 2020). This is very problematic because the unnoticed pollution from sewers can lead to potential risks such as public health and environmental impacts. Despite the hazards of urban ex-infiltration on public health, the environmental impact of urban sewer leakages receives less attention and is even ignored in urban system emissions, especially in large-scale studies (Blackwood et al. 2005). Sewer ex-infiltration is a dynamic process varying with precipitation and groundwater elevation as well as flow conditions within the pipe, which will deteriorate the colmation layer formed on defects (Nguyen et al. 2021). Although sewer exfiltration is known as a self-sealing process, high-velocity flow during a storm event will relatively wash the colmation layer, which will increase the leakage rate up to 20 times for the entire network (Held et al. 2007). Despite the uncertainty of estimating a leakage rate for the entire sewer system, the conclusion from different studies would suggest that leakage rates vary between 0.01 and 0.2 l/s.km, which is considerable at the network scale (Ellis et al. 2009; Karpf & Krebs 2011). Sewer exfiltration introduces pollutant to the pipe's surrounding environment and ultimately contaminate the groundwater. It has been widely examined that untreated wastewater contains high levels of organic pollutants, suspended solids, and toxic compounds (Li et al. 2008). These released micropollutants will potentially contaminate surface water and risk public health, which will cause serious socio-economic damage. On the other hand, groundwater infiltration to the sewer system is destructive and crucial mainly due to (i) hydraulic overload caused by infiltrated volume for the treatment plant (up to 100% in some cases) and (ii) pollution concentration dilution, which leads to removal efficiency reduction (Bertrand-Krajewski 2006). Therefore, structural health inspection of the sewer system is the key component to assure wastewater transport without ex-infiltration or groundwater infiltration.

Due to the sewer system's invisibility and long piping, sewer inspection is costly, time consuming, and challenging. Currently, various sewer inspection methods (SIMs) are present, and their accuracy of deterioration detection are increasing with the evolution of technology. Various in-pipe SIMs have been developed and documented ranging from single techniques such as acoustic methods, closed-circuit television (CCTV), and electromagnetic methods to multi-sensory inspections, which combine various techniques like sonar and ultrasound as well (Gokhale & Graham 2004; Harris & Dobson 2006; Ling et al. 2019; Yu et al. 2021). However, these SIMs usually have a low area of implementation (AoI) and inspect only a small area at a time, which makes them time consuming and hence costly to apply at the network scale. Inspection difficulties and invisibility of the potential leakages often lower the sewer inspection frequency, which results in defects and leakages remaining hidden unless a major failure happens. Moreover, applying only one SIM cannot offer sustainable monitoring since it needs to fill the gap between precision and AoI or in other terms, i.e., time and cost.

General structural health assessment of a sewer system provides an approximate understanding of the pipes, which could further lead to localized but more precise inspections by reducing the inspection costs of the entire system as a by-product. T-I methods tend to evaluate the spatiotemporal conditions of the pipes and their surrounding environment to assess a wide but reliable perspective of the sewer system deterioration process.

Sewer deterioration can be categorized into structural and hydraulic deterioration (Water Research Centre (Great Britain) & WAA Sewers and Water Mains Committee 1986). Structural deterioration is defined by structural faults (such as cracks and fractures) that can lead to structural failures, e.g., pipe collapse (Tran 2007). Hydraulic deterioration is manifested by hydraulic faults (such as tree root incursion and debris) that impairs transport capacity and may result in hydraulic failures, e.g., blockage and overflow. Ultimately, a defect could be derived from both structural and hydraulic deteriorations (Chughtai & Zayed 2008).

Structural health assessment of a sewer system can be determined by deterioration models. A wide variety of deterioration models has been developed and proposed to predict the deterioration process based on existing data and real-observed sewer conditions. Hansen et al. (2021) studied a database containing 318,457 pipes from 35 utilities which are implementing deterioration models and found that the age, groundwater, and pipe shape are the most common deterioration factors used by these models. Based on Kley & Caradot (2013), these models can be categorized into three main non-exclusive groups, i.e., (i) deterministic models, (ii) statistical models, and (iii) artificial intelligence (AI) models.

Deterministic models examine the quantitative relationship between deterioration factors and sewer conditions. These known relationships will be modeled either linear or exponentially to assess the current and future structural health of the sewer system. Deterministic models, regardless of their complexity, are often too simplistic and the lack of investigation of interdependent deterioration processes decreases their applicability (Rajani & Kleiner 2001). The outcome from a recent study by Yoon et al. (2021) reveals that the deterministic model predicts the life expectancy of a reinforced concrete (RC) almost 52% more than a statistical model which is due to simplification of the deterministic models.

To increase the efficiency of the deterministic approach, statistical modeling was developed in which parametric density for the deterioration process was assessed. The outcome (e.g., predicted pipe condition) of the statistical approach is derived from the probabilistic relationship of the input factors which makes this method more reliable than deterministic models, which are based on quantitative influences (Tscheikner-Gratl et al. 2020). Among various statistical models, cohort survival function and Markov chain are the most applied methods on a network scale (Baur & Herz 2002; Rokstad & Ugarelli 2015; Caradot et al. 2017). Logistic regression and discernment analysis were also implemented and documented successfully which further approves the higher precision level of the statistical approach over deterministic model (Ariaratnam et al. 2001; Fuchs-Hanusch et al. 2015).

Machine learning approaches on the other hand develop a non-linear relationship between input (e.g., deterioration factors) and output (e.g., predicted pipe condition) data. AI models, unlike the statistical approach, are needless for predefined deterioration correlations and ‘learn’ the deterioration process from pipe inspection data and aggregate it to non-inspected parts of the network (Scheidegger et al. 2011; Tscheikner-Gratl et al. 2020). The most frequent machine learning methods applied for sewer deterioration are neural networks (Jiang et al. 2016; Alsaqqar et al. 2017), random forest (Hansen et al. 2019; Tavakoli et al. 2020), and support vector machine (Mashford et al. 2010; Ye et al. 2019). Random forest as an example provides 66.7% efficiency in-sewer condition classification compared to CCTV data (Caradot et al. 2018).

The leakage hot spot map spatially represents the sewer leakages potential using the Geographic Information System (GIS). Previous publications indicate that several primary factors can influence sewer ex-infiltration including pipe characteristics, colmation layer, defective conditions, wastewater level, wastewater composition, soil properties, and hydraulic potential, e.g., water pressure and soil moisture (Davies et al. 2001). Moreover, factors contributing to the deterioration process are classified into physical factors, environmental factors, and operational factors (Malek Mohammadi et al. 2020). Influencing factors on sewer deterioration could be assessed and implemented in a spatial leakage likelihood map to predict the potential leakage hotspots based on available data, which complies with sampling and deterioration models.

Roehrdanz et al. (2017) developed a spatial sewer exfiltration probability map based on a physical factor (sewer infrastructure) and an environmental factor (groundwater elevation) without prior knowledge of pipes defects. The spatial model was further examined by sampling and tracer tracking, which proves the capability of the model to predict the sewer exfiltration to some degree. Furthermore, a detailed investigation and case-study implementation of the sewer leakage hotspot map is under preparation by the author of this paper considering deterioration factors such as age, material, sewer type, flow velocity, and surface vegetation at the city scale followed by CCTV data validation.

Overall, mentioned methods in T-I (see Table 1) tend to grasp a general but reliable understanding of the sewer structural health condition which has the highest AoI in exchange for the relatively low implementation cost. Furthermore, SIMs belonging to T-1 aim to localize and narrow the further required inspections areas to be suitable for Tier-II solutions.

After localizing the potential defect areas for leakage, intermediate solutions for investigating the temporal behavior of the pipes and their surroundings would be recommended as T-II methods. The advantage of these methods is to decrease the high operational cost of in-sewer inspections methods for relatively large areas by localizing the leakage source to some extent. Here, we are going to mention some suitable SIMs for T-II.

Following to ERT, MLM relies on receiving and source electrodes which report on electrical continuity. The source electrode should be placed within the medium (e.g., wastewater in the pipe) and assess the electrical continuity with surrounding receiving electrodes. Wood & Palmer (2000) investigated detecting pollution on four sites in Sydney using the MLM method followed by CCTV validation. The results show a strong correlation in clay and sandy soil, respectively. Ling et al. (2019) developed a methodology for detecting the leakages from water reservoirs by implementing 50 electrodes surrounding the reservoir and applying the MLM method. They further implemented the inversion algorithm and proved the capability of the MLM method for water leakage detection. Moreover, an ongoing investigation under project INCIDENT (Utilising sewer network characteristics for the identification of the optimized point-based monitoring systems) by the authors of this paper focuses on sensitivity analysis of this method in saturated and unsaturated zones equipped with 72 electrodes based on a 3-D matrix.

Sampling can be also used for sewer leakage detection when soil samples around the pipe are analyzed for their contaminants and moisture, which can be further calibrated by sewer grab sampling (Liu & Kleiner 2013). Long-term soil sampling and analyses for E. coli and wastewater micropollutants can detect the defects and quantify the sewer leakage for dry and wet weather conditions (Guérineau et al. 2014). By applying microbial source tracking (MST) and soil sampling, Calderon et al. (2022) were able to identify human fecal pollution in soil which is a potential source of microbial pollution to surface and subsurface water as they can last for several months after the contamination.

Overall, discussed methods in Tier-II (see Table 2) would provide a deeper and more detailed understanding of the leakage on an intermediate scale, which further needs to be examined with in-sewer inspection methods offered in Tier-III.

In this part of the solution, based on pipes' local conditions and surroundings, single or combination of methods can be selected and applied. Due to their inspection precisions, these methods are often costly and time consuming; therefore, a wise selection of the methods could be beneficial. Although most of these methods are well established and documented, mentioned here are only some of the significant ones.

The most straightforward methods for sewer leakage assessment are the physical approaches. The smoke test for instance can detect cracks above the waterline in the pipes. Highly visible smoke will be injected in an isolated pipe section, which is likely to escape the pipe from present cracks and further travel through the porous parts and cavities of the subsoil and eventually observe at the ground level. Smoke testing is unable to quantify the severity and leakage rate of the defect, while large pipe investigation is limited by the smoke blower capacity (Eiswirth et al. 2000; Noshahri et al. 2021).

The widely applied pressure test can detect the leakage presence and further determine the exfiltration rate for the inspected pipes' length. In this method, a section of the sewer pipe will be isolated by flammable devices and pressurized air, or water will be added. The reduction of pressure and volume within the pipe will identify and assess the leakage presence and rate over time. The pressure test overestimates the leakage rate due to the artificially introduced pressure to the pipe, therefore free surface method tends to calculate the leakage rate by introducing a known volume of water and performing the mass balance downstream (Bertrand-Krajewski 2006). Within a recent laboratory investigation by Ulutaş et al. (2022), 29 specialist contractors carried out different leak tests (air under/over pressure and water pressure) and it was observed that all of the applied methods are able to detect the leakage in case the experts did not make any serious test errors.

A compact laser-based scanner (also known as LiDAR) will be mounted on a pipeline inspection gauge (PIG) to provide a detailed and accurate assessment of the pipe by comparing the profile to the reference shape (Hao et al. 2012). Laser inspection is limited to the dry surface above the water waterline and its goal is to generate 2-D images along the pipe providing further information on pipe inner geometry, material reduction, debits, and defects (Lepot et al. 2017b). Laser scanning is generally coupled with closed-circuit television (CCTV) inspection to provide a wider perspective of the pipe. Currently, LiDAR is being used in multi-sensory inspections.

The most widespread SIMs are camera-based approaches that aim to provide qualitative images along the pipe. CCTV mounted on a crawling in-pipe robot allows inspectors to detect any structural defects by analyzing the returning video and documenting the defect by a snapshot of the inner surface (Duran et al. 2002). CCTV requires dewatering, debris removal, and a human operator for sewer inspection, which makes this approach costly and time consuming. Moreover, CCTV is unable to detect any voids in the backfill soil and is often carried out after a structural failure or blockage report was made (Selvakumar et al. 2014). Therefore, an inspection schedule would be recommended and beneficial for CCTV investigation. Zoom camera, which is a stationary investigator, is able to inspect the inner surface of the pipe of any materials same as conventional CCTV and provide a preliminary schedule for further inspections. The main difference among them is that the zoom camera is stationary and attached to a tripod above the manhole and captures fragmented pictures or videos along the pipe. This immobility of the zoom inspection makes it less costly and less time consuming; however, observed data resolution is lower than conventional CCTV (Moradi et al. 2019). Digital scanning is another visual SIM that implies a 360° fish-eye digital scanner to produce a unified side scan of the pipe instead of the forward imaging offered by conventional CCTV. Digital scanners propagate light beams to every millimeter of the pipe, and they are often mounted with an inclination meter and gyroscope (Jung & Sinha 2007). PANORAMO is another visual inspector equipped with a high-definition (HD) 185° fish-eye lens which has a higher crawling speed (0.3 m/s) than other robotic approaches (Rayhana et al. 2021). For more in-depth inspections, multi-sensory inspection often combines visual inspection with other SIMs such as sonar and acoustic for a broader understanding of the pipe deterioration process.

Acoustic leak analysis (ALD) is a passive non-invasive approach to detect leakage in pipelines. ALD systems are bounded with a listening device to detect and record the leak-induced sounds. Ultimately these methods require an onsite receiving station to detect the leakage using hydrophones or accelerators at the required location (Rizzo 2010). In addition, ground microphones can be installed to extend the assessment area and pinpoint leakage points by listening directly at the pavement or on the pipe's topsoil. More recently, noise correlators are being applied to increase the efficiency of leakage detection, which further expands the investigation area. In case of a leak presence between the sensors, there will be a distinct peak in the cross-correlation function, which can be located due to the travel time (Karney et al. 2009). The listening stick for instance is the simplest method to detect leakages within the pipe. The stick has an earpiece implemented at the end of a stainless-steel tube. Contact between the stick with the pipe will allow the sonic vibration caused by the leakage to be transmitted through the earpiece or potentially an acoustic detector. This method functions best in pressurized metallic pipes but its dependency on the human operator makes it costly (Hamilton & Charalambous 2013). Typically, these inspections are carried out based on an accident or report, meaning that they will only inspect a fraction of the sewer network (Yu et al. 2021).

Sonar propagates high-frequency sound waves and visualizes the surrounding by investigating the return waves from any reflector. In the case of wastewater assessment, the sonar inspection device generates pulse waves as it travels through the pipe and the receiving head analyzes the return waves and illustrates the inner surface allowing to detect any defects. Sonar inspection is limited to one medium, therefore, to investigate both dry and wet parts of the pipe, it is often combined with the laser technique to provide a 360° evaluation of the inner pipe (Selvakumar et al. 2014).

Smartball (Figure 5(c)) is another acoustic concept mounted by various receivers such as accelerometers, temperature, and pressure sensors. The ball is inserted upstream and travels within the pipe due to the flow. Furthermore, by collecting the transmitted acoustic data, it is possible to locate the potential leakage points as well as the ball's location. Ultimately, the smart ball will be collected downstream (Liu & Kleiner 2013). Currently, Smartball and SAHARA techniques developed by the Pure Technology group (Figure 5) can inspect low diameter pipes (from 65 cm) and precisely pinpoint leakages as small as 0.1 l/min according to their products' brochures.

Impact echo (IE) is a non-destructive method widely used to determine the thickness of concrete structures. IE applies stress pulses into the test surface by the mechanical impact which provides information on the concrete's thickness, quality, and delamination by sound velocity assessments within the concrete (Malhotra & Carino 2003). IE requires dewatering and human access to the pipe, but it can also be applied externally if the required exteriors are available. Sack & Olson (1998) have applied IE on a 2-m diameter pre-stressed concrete pipe. They have reported on the effective capability of IE in measuring the pipe wall thickness, detecting the delamination, and recognizing the reduced strength of the pipe. During an investigation done by Kang et al. (2017), a quantitative index based on resonance frequency and the spectrum was developed. They reported that the implemented index with IE can detect the cavities formed by sewer exfiltration around the concrete pipes.

Ultrasound sewer inspection is based on high-frequency sound waves to detect various structural features of the material such as thickness, attenuation, defect presence, size, and orientation. Ultrasound pulsed bulk beams generated from piezo-electric transducers will reflect from the material carrying amplitude and travel time information, which could further assess the defect criteria (Iyer et al. 2012).

The multi-sensor system combines various SIMs to provide an in-depth understanding of the deterioration process by benefiting from a system of sensors. PIRAT, a quantitative technique for sewer inspection, applies CCTV, sonar, and laser profiling data to the AI algorithm for calculating the inner geometry as well as classifying the existing defects (Kirkham et al. 2000). MAKRO is a flexible multi-sensor robot that is equipped with internal and external sensors mainly designed for 300–600 mm diameter pipes. Internal sensors such as odometry (0.25° precisions), inclinometer, ultrasound, and optical lens functioning along the external sensors like an obstacle, collision, and landmark detector aim for a more in-depth inspection of the deterioration process (Rome et al. 1999). MAKRO technology is not commercially available and originally it was designed to operate as a prototype (Rayhana et al. 2021). Responder is another multi-sensory technology that is equipped with 3-D LiDAR and is suitable for large pipe diameter inspection. Responder can investigate up to 2,400 meters in a single run but at a cost of its heavy weight of 300 kg (Tur & Garthwaite 2010). SAM is another sophisticated multi-sensor inspector mounted with a variety of sensors and technologies. SAM is equipped with optical triangulation for a 3-D geometry assessment, a microwave sensor for backfill soil inspection, a geoelectrical sensor for leak detection from its resistivity reduction, a hydro-chemical sensor for groundwater infiltration detection, a radioactive sensor for backfill soil moisture content, and an acoustic sensor for detecting the cracks and defects (Eiswirth et al. 2000). Surveyor is the 5th generation in-pipe crawler from SewerVUE group (Figure 5) which is mounted with CCTV, LiDAR, and in-pipe GPR (PPR) and it can detect inner and outer defects on non-ferrous pipes, structural integrity, and pipe thickness with high accuracy level as small as 10 mm (Figure 7).

Electromagnetic technologies are widely applied for pressurized and ferrous pipes. Magnetic flux leakage (MFL) is a non-destructive method that uses strong magnets to create a magnetic field inside the pipe and investigate the changes in magnetic permeability of the pipe resulting from cracks and corrosion (Bertrand-Krajewski et al. 2021). Pulsed eddy current (PEC) inspection is based on a time-varying alternative current (AC) to the pipe material generated from a magnetic coil. present cracks and material losses will influence the coils' impedance which will further intensify the defects (Liu & Kleiner 2013). In-pipe GPR mounted on a PIG is another electromagnetic method that penetrates microwave-band radiation to the inner pipe wall and analyses the reflected wave from backfill material. Despite the conventional GPR method mentioned in Tier-II solutions, in-pipe GPR uses either ‘look- through’ or ‘look-out’ approaches which receive the reflected wave on the ground surface or with a PIG inside the pipe respectively (Hao et al. 2012). Focused electro leak location (FELL) is an electrical method developed within the SAM project that relies on electrical resistivity reduction of the pipe in case of any defects or cracks presence. FELL employs the inlet electrode below the water table in the pipe as well as the receiving electrode mounted on the surface allowing FELL to detect any cracks regardless of the pipes' material and functionality (Gokhale & Graham 2004). Furthermore, Tuccillo et al. (2012) compared the results from commercially used FELL-41 with joint pressure test results and suggested its capability to detect defects.

Assuming that sewer ex-infiltration will cause temperature variation within the pipe and its surrounding environment, thermography approaches aim to detect this difference and locate the leakage area. Distributed temperature sensing (DTS) is an in-pipe inspection technology that requires a fiber optic cable installed along the pipe, which can detect the temperature, vibration, and acoustic pressure via stimulated changes in the optical indicator over time (Tanimola & Hill 2009).

Infrared reflectometry (IR) is an alternative to the DTS method since it can be applied from the surface. It has been reported that infrared cameras can detect illicit connections to the main sewer system, however, this temperature difference is lower than the detection limit for leakage rate quantification (Lepot et al. 2017a).

Overall, the abovementioned in-pipe SIMs tend to grasp an in-depth understanding of the pipe's structural health condition and precisely pinpoint the defects. However, these SIMs are often expensive and time consuming to apply in return for their precision. Table 3 summarizes all the T-III methods.

As illustrated in Figure 10, SIMs belonging to the T-I methods offer the highest AoI with relatively low technicality which makes them almost necessary for any sustainable monitoring campaign. Among T-I methods, AI, statistical, and deterministic deterioration models offer broad, dynamic, and a relatively precise (accordingly) understanding of the deterioration process. LHM, on the other hand, breaks down and implements a GIS-based matrix of available deterioration factors which provides a criticality situation snapshot of the sewer network with a less complex approach than the deterioration models. Decreasing the AoI, ATI provides an intermediate inspection while ERT and MLM offer more robust results, since underground temperature anomalies more easily detected by geoelectric than infrared means.

Furthermore, DTS and Smartball are long-term investments in this range of AoI since Smartball offers mobility, while DTS has the potential to monitor large areas with long-term temperature inspection. The main advantage of the Smartball method over DTS, ERT, and ATI is the mobility function which allows the ball to travel long distances in functioning pipes without disruption.

The lowest level of technicality belongs to pressure test, listening stick, and smoke test which offer an affordable and relatively fast inspection. The main disadvantage of these methods is the dewatering necessity (pressure and smoke tests) and being uncertain of the exact leakage rate and location. Soil sampling as a non-destructive method is able to reveals any existing wastewater compounds in soil, while it can potentially cover a large area of investigation in case of repetition. Soil sampling at the city scale will be too costly and time consuming, therefore FELL, DTS, Smartball, and MLM could be a promising alternatives due to their broad and long-term inspection. Following with non-destructive methods, GPR is relatively easy to implement but only can inspect the outer top surface of the pipe, while QUEST-C could be a supportive method since it can detect defects below the water surface.

In-pipe SIMs usually tend to have a higher level of technicality and precision with lower AoI which make them a favorable yet challenging inspection validator. Electromagnetism often offers high levels of precision, while MFL and PEC are being limited to ferrous pipes but not demanding dewatering. Acoustic methods on the other hand such as IE and AR are limited to concrete pipes and demand dewatering. However, sonar profiling can be performed during dry weather conditions, while mounting it on a PIG will potentially increase its AoI. Sonar profiling could be a superior SIM to FELL as it provides a cross-sectional inspection, while FELL is limited to the upper side of the pipe. Moreover, ultrasonic approaches such as bulk wave piezo-electric provide a deeper understanding of the deterioration process, while offering a relatively low AoI. A common limitation of ultrasonic methods, which is being limited to only locating the first defect, was overcome by the development of the guided ultrasonic wave, which propagates the wave through a waveguide allowing it to travel several meters to kilometers. Although ultrasonic inspection for ferrous and above-ground pipes is widely applied, the approach could be implemented on concrete pipes with some considerations. Concrete, due to its structural heterogeneity, highly attenuate the generated sonic wave; Therefore, the wave frequency should be chosen according to the required inspection range. Another challenge of this method is to have access to an excavated outer surface of the pipe or dewatered inner access for the device and the operator.

Ultimately, multi-sensory inspection, which is a visual-based PIG combined with other SIMs such as sonar, PPR, and laser scan, has the highest level of technicality and offers the most possible inspection precision. CCTV, as a standard method, provides raw images from inner pipe which should be further analyzed and related to the potential defects above the water line. Required image processing, while being limited to the dry surface of the pipe has the main disadvantages of CCTV. However, implemented sonar and LiDAR allow the crawling PIG to investigate the entire pipe cross-section, while other supplementary technologies such as hydro-chemical, microwave, radioactive, and geoelectrical sensors will further advance the inspection complexity which eventually increases the inspection cost. Although multi-sensory inspection requires dewatering of the pipe, some SIMs such as tracer test, SAHARA, Smartball, FELL, MLM, and thermal inspection can be applied during the dry-weather flow.

As a sewer system ages, it becomes more vulnerable to sewer deterioration factors (physical, environmental, and operational), which will eventually cause more leakages and environmental pollution. Therefore, continuous structural health monitoring of the sewer system is crucial and yet challenging at the network scale. Although there have been a variety of in-pipe SIMs developed and implemented for different conditions, their AoI is often low, which makes them costly and time consuming at the network scale. The necessity of sewer inspection as well as inspection cost and AoI demand a SIM recommendation for a sustainable monitoring campaign. This paper suggests a hierarchy for available SIMs based on their AoI and proposes a decision-making system for detecting potential leakages at the city scale through a decentralized approach.

Deterioration models and LHM as T-I solutions are preliminary SIMs that can provide a broad and reliable understanding of the sewer system's structural health. These methods are relatively easy to apply and offer the highest AoI which makes them necessary for a sustainable inspection plan.

An intermediate inspection on a cluster of neighboring pipes could be achieved from T-II solutions due to their simplicity and relatively high AoI, which makes them very cost effective and efficient. The main advantage of the T-II solutions is their capability to be applied during dry-weather conditions and their intermediate AoI makes them a desirable choice for validating the outcome from T-I solutions. Smartball, MLM, DTS, FELL, SAHARA, pressure/smoke test, and soil sampling are promising approaches among these intermediate and non-destructive methods, which offer acceptable AoI with relatively low implementation cost.

The early-warning and dynamic inspection plan derived from Tier-I and Tier-II solutions would offer substantial benefits such as:

• Prediction and detection of potential leakages at the early stages, which is crucial for the environment and groundwater quality.

• The independence of these methods to pipe's material provides a robust and scalable inspection.

• Early defect detection prior to severe pipe failure lowers the maintenance cost.

• A considerable amount of time and effort will be saved since dewatering is not required.

Obtained information should be further validated with in-pipe SIMs before excavation and rehabilitation. In-pipe and high technical SIMs offered in the T-III have the highest precision rate for detecting any defect within the pipe. Although these methods often require dewatering, it is logical to implement them before the excavation to avoid any extra costs. Multi-sensory inspections, which often include CCTV, LiDAR, and PPR as the base inspection, could be further modified and implement more techniques on the PIG, which will further increase its technicality level and ultimately its implementation cost. Combining more techniques together brings centralization which will increase the cost and vulnerability. Therefore, centralization (multi-sensory in this case) should be a supportive part of a sustainable inspection plan, not its main one.

A healthy sewer inspection plan should be decentralized, and it would be achieved by choosing at least two methods belonging to different tiers (T-I, T-II, and T-III). Solutions belonging to different tiers provide a dual investigation of precision and AoI, a balance between cost and time, as well as an equilibrium between advantages and disadvantages, which ultimately offers a decentralized and sustainable solution for sewer leakage detection and its structural health inspection at the city scale.

Although this study tends to cover SIMs with different techniques, future studies could further increase these SIMs and compare them from different perspectives. More detailed and comprehensive comparisons among SIMs will allow authorities to come up with a more sustainable inspection plan. A sustainable sewer inspection plan at the city scale should be decentralized, fractal, case specified, and cost effective. Future sewer inspection practices could be dynamic and reasonable by implementing low-cost sensors as a long-term investment in more vulnerable areas for early defect detection. Moreover, this early detection system could be interconnected through the wireless technology and creates a decentralized early inspection network which could potentially save the huge replacement cost after the pipe failure.

The authors would like to announce their gratitude to the Deutsche Forschungsgemeinschaft (DFG) regarding the funded project entitled INCIDENT (Utilising sewer network characteristics for the identification of the optimized point-based monitoring systems). The authors are grateful to the colleagues of the Institute of Urban and Industrial Water Management for their support in this review. The authors are also thankful to IWA Publishing for their open access (S2O) platform.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.