Water Leak Detection: A Comprehensive Review of Methods, Challenges, and Future Directions

Abstract

This paper provides a comprehensive review of the methods and techniques developed for detecting leaks in water distribution systems, with a focus on highlighting their strengths, weaknesses, and areas for future research. Given the substantial economic, social, and environmental impacts of undetected leaks, timely detection and precise location of leaks are critical concerns for water authorities. This review categorizes existing methods into traditional approaches, such as manual sounding, and modern techniques involving smart water management and sensor technologies. A multidimensional bibliometric analysis was employed to systematically identify, select, and evaluate 600 scholarly articles on water leak detection, sourced from the Scopus database over a 23-year period (2000–2023). The paper evaluates each method based on leak sensitivity, burst detection, continuous monitoring, alarm accuracy, and implementation costs. Novel insights include an analysis of emerging smart water technologies and their integration into real-world water distribution networks, offering improved efficiency in leak detection. The paper also identifies key gaps in current research and suggests future directions for advancing the accuracy and cost-effectiveness of these technologies.

1. Introduction

Water leakage is a significant issue in water distribution systems, with the greatest proportion of unaccounted water being lost through supply line leaks. Leaks can also occur at joints, valves, fire hydrants, and service connections up to the customer meter. Several factors can contribute to leaks, including the material, age, composition, and joining methods of pipes. In addition, environmental and external conditions, such as the type of surrounding soil, stress from traffic vibrations, and frost loads, can lead to leaks [1]. Water conditions, including velocity, pressure, and temperature, are important factors when dealing with water losses. For instance, higher water pressure can increase leakage rates by widening existing cracks in pipes, while extreme temperature variations can cause pipe materials to expand or contract, leading to more frequent breaks. The flow range and velocity of the leaking water also influence the detection difficulty, as slow leaks may go unnoticed for longer periods, leading to more significant water loss over time.

Typically, between 20 to 30% of water in distribution systems is lost before reaching consumers, and in some regions with aging infrastructure, this percentage can rise to as high as 50% [2]. These losses fall under the category of non-revenue water (NRW), which refers to water that is produced but never reaches the consumer due to leaks, theft, or metering inaccuracies. NRW represents a significant inefficiency that intensifies pressure on freshwater sources, such as rivers, lakes, and aquifers, which are already stressed due to over-extraction, population growth, and climate change. Excessive water extraction disrupts ecosystems, lowers water tables, and depletes vital reserves, such as aquifers, which are slow to replenish, threatening biodiversity and ecosystem functions [3]. Water losses can result in severe economic impacts, including repair costs and non-revenue water. Social effects may also occur due to consumer inconvenience caused by low pressure, service interruptions, and health risks. In addition to the immediate economic and social effects, water losses can have significant environmental consequences, including the depletion of valuable water resources, particularly in regions facing water scarcity. Water treatment and distribution are energy-intensive processes. When leaks occur, additional energy is required to treat, pump, and distribute extra water to compensate for the losses [4]. This inefficiency not only wastes energy but also raises the overall carbon footprint of water utilities, as more power is consumed to deliver the same amount of water [5]. Moreover, soil erosion and infrastructure degradation are commonly observed as secondary effects of continuous water leakage. Excessive leakage can increase the moisture content in soils, destabilizing foundations and contributing to subsidence and the deterioration of roads and buildings [6]. Additionally, the loss of potable water from pressurized systems increases the risk of contaminants entering the water supply due to negative pressure zones around leaks [7]. This issue is particularly problematic in older or poorly maintained water networks, posing a public health risk by contaminating drinking water supplies with pathogens or chemical pollutants.

To minimize the financial, social, and environmental impacts of water losses, various methods have been developed over the years to detect and manage water leaks [8,9,10,11]. The first methods include acoustic monitoring [12,13] and non-acoustic techniques [14,15,16,17]. The advancement of technology has led to the development of inline inspection methods for pipeline surveys [18,19,20,21]. Progress in software and measurement tools has enabled water utilities to easily calculate water volume balance [22] and apply statistical analysis to predict leakage [23]. However, these techniques are time-consuming and find it challenging to pinpoint leaks, and they cannot anticipate water-related issues. Consequently, they may not prevent potential catastrophic damage.

Recent developments in smart technology, encompassing monitoring, data transmission, and data analysis, have given rise to innovative water leakage detection methods [24,25,26]. This article comprehensively analyzes cutting-edge techniques and technologies developed for detecting water leaks. It investigates the primary benefits and limitations of each approach. Incorporating bibliometric analysis into the study effectively assesses the broader research trends and highlights connections among relevant research institutions. In this review, a comprehensive report on water leak detection methods spanning a 23-year period (2000–2023) across various countries is presented. The study utilizes bibliometric analysis to analyze the characteristics of scientific articles in Scopus, minimizing subjective biases, and then proceeds to offer an analysis of the overall publication trends in water leak detection techniques. This review not only assesses existing methods but also identifies research gaps and future directions for improving the accuracy, efficiency, and cost-effectiveness of water leak detection systems. Additionally, the study examines the potential for integrating smart technologies into leak detection, offering insights into how these technologies can address the limitations of conventional methods.

2. Research Methodology

This research employs a multidimensional bibliometric analysis to comprehensively investigate water leak detection techniques. The analysis explores the diverse range of assessed publications, encompassing academic papers, conference proceedings, reports, and more, to gain a solid understanding of the field. This comprehensive analysis spans two decades and extends beyond content exploration to include a geographical dimension, mapping the global distribution of research contributions. Recognizing the institutions actively engaged in water leak detection research and identifying the top contributors to scholarly literature are integral aspects of the study. Additionally, the analysis presents a spotlight on the key journals that have been central to knowledge dissemination in this field. This comprehensive approach aims to provide valuable insights into the evolution of water leak detection research, enhancing the collective understanding of this critical domain.

Data Collection

In the context of this bibliometric analysis, data were gathered from the Scopus database from 2000 until the end of 2023. The search focused on “water” and “distribution systems” and “leak detection” or “leakage detection”. In total, 626 documents were obtained. They were then classified into various document types. This categorization aimed to comprehensively understand the scholarly contributions within this specific research domain.

Table 1. Document type distribution (%) in water leak detection techniques.

Type of Documents	Number	Percentage (%)
Article	298	48
Conference paper	284	45
Review	18	3
Book chapter	12	2
Conference review	7	1
Note	5	1
Short survey	1	0.2
Data paper	1	0.2

presents the breakdown of these document types and their respective percentages. The categories with percentages less than 1% were considered not statistically representative and, therefore, were excluded from the study. The duplicates and the non-English studies were removed. Then, 600 papers were filtered for the bibliometric analysis and the systematic review. The data collection process is shown in Figure 1.

3. Bibliometric Analysis Results

Understanding the global landscape of research on water leak detection is crucial for identifying how the field has evolved, where contributions are most significant, and which regions or institutions are leading innovation. This analysis serves as a strategic tool for policymakers, funding agencies, and industry leaders. By recognizing trends and gaps in research, decision-makers can make informed choices about resource allocation, international collaboration, and priority areas for future research and technological investment.

3.1. Publication Trends

The global research output on water leak detection has shown marked growth since 2000, particularly over the past decade. Figure 2 shows fluctuations in annual publications, with a steady rise beginning in 2009, indicating a growing awareness of water distribution challenges and innovations in detection methods. The surge in scientific contributions during this period correlated with the rise of global water sustainability initiatives, advancements in sensor technologies, and the increasing application of machine learning and artificial intelligence to detect and predict leaks.

It is important to examine the factors behind the noticeable spikes in publication numbers during certain years, such as 2014 and 2017. These peaks corresponded to a period of increased global attention to infrastructure resilience, influenced by major initiatives, such as the United Nations’ Sustainable Development Goals (SDGs), particularly SDG 6, which addresses clean water and sanitation. The progressive nature of the publication trend reflects the growing complexity of leak detection technologies, moving from traditional hardware-based methods to the integration of smart water networks. The steady increase in research output highlights the critical need to address water leakage in aging infrastructures.

3.2. Mapping Research Distribution

A significant aspect of this bibliometric analysis is understanding which regions and countries are leading the research effort in water leak detection. As shown in Figure 3, the United States (98 publications), the United Kingdom (70 publications), and China (64 publications) were the top contributors to the field. This trend underscores the economic and technological resources available in these regions, which allow for sustained investment in research and development of water management technologies.

The geographical distribution of research showed significant differences in contributions. For instance, despite the severe water scarcity, many countries in Africa and the Middle East produce less research on water leak detection. This highlights the need for more international collaboration and funding in regions where water leakage threatens water security. A likely reason for this gap is the difference in research funding and access to advanced technology. In developing countries, the priority is often immediate water access rather than advanced leak detection technologies. Building partnerships between leading institutions in developed countries and universities or research centers in water-stressed regions could help close this gap. For policymakers, this suggests the importance of directing global research funding to areas with high water stress but low research activity, ensuring more equitable distribution of knowledge and innovation.

3.3. Institutions in Water Leak Detection Research Publications

At the institutional level, the leading contributors included the Universitat Politécnica de Catalunya (28 publications), The University of Adelaide (24 publications), and the CSIC-UPC Instituto de Robótica e Informática Industrial (20 publications), as shown in

Table 2. Contributions of institutions to research publications on water leak detection.

Institution	Number of Publications
Universitat Politécnica de Catalunya	28
The University of Adelaide	24
CSIC-UPC—Instituto de Robotica e Informatica Industrial IRII	20
University of Exeter	18
Bentley Systems Incorporated	14
Clemson University	12
Zhejiang University	11
NC State University	10
University of Waterloo	10
Deakin University	10

. The leadership of European and Australian institutions indicates regional priorities in addressing water leakage, where aging infrastructure in Europe and climate-change-induced water scarcity in Australia have made this issue a high priority. Companies, such as Bentley Systems Incorporated, have made significant contributions to the field, demonstrating the critical role that industry partnerships play in fostering innovation.

This analysis of institutional contributions also highlighted potential opportunities for collaboration. For instance, leading research institutions could establish partnerships with underrepresented regions. Such collaborations can enhance the global exchange of knowledge, leading to the development of more effective, locally adapted solutions to water leakage challenges.

3.4. Top Contributors to Scholarly Literature

Table 3. Top-ranked scientists in the field of water leak detection.

Top-Ranked Scientists	Number of Publications
Puig, V.	24
Gong, J.	17
Lambert, M.F.	15
Simpson, A.R.	14
Wu, Z.Y.	14
Kapelan, Z.	11
Piratla, K.R.	11
Zecchin, A.C.	11

provides a list of the nine top-ranked scientists in the field of water leak detection. It was observed that during the period (2000 to 2023), they made significant contributions to this field, with the number of publications ranging between 10 and 24. Their research covers various methodologies and innovations that have significantly influenced current knowledge and practices in leak detection.

3.5. Key Journals in Water Leak Detection Methods

Figure 4 illustrates the top-ranked journals in publishing papers in the field of water leak detection. The “Journal of Water Resources Planning and Management” is ranked first with 36 papers, followed by “Procedia Engineering” with 26 papers, “Water Switzerland” with 25 papers, and “Water Resources Management” with 18 papers. The “Journal of Pipeline Systems Engineering and Practice”, “Urban Water Journal”, and “Water Science and Technology Water Supply” also made significant contributions, with 13, 12, and 12 publications each, respectively. These journals are notable for publishing research related to methodologies, technologies, and case studies relevant to water leak detection.

3.6. Mapping of Keywords’ Co-Occurence

Keywords serve as efficient means to convey the primary research focus within an article. They define the various research subfields within a given domain [27]. Using VOSviewer, a map of authors’ keywords was created based on the fractional counting method. A thesaurus file was used to merge keywords with the same semantic meanings. The minimum number of occurrences was set to 5. Out of the 1128 keywords, 32 met the thresholds, as shown in Figure 5.

The visualization map allowed us to identify four clusters of methods to detect water leakage (Figure 6):

• Cluster A covers hardware-based techniques.
• Cluster B consists of leak localization methods.
• Cluster C assembles software-based methods.
• Cluster D regroups the methods based on smart monitoring.

4. Leak Detection Technologies

Based on the identified clusters, this research was completed by a systematic review on leak detection methods: hardware-based methods, software-based methods, and smart water networks. Hardware-based methods are able to detect and locate the leaks; therefore, clusters A and B were grouped together for the comprehensive review.

4.1. Hardware-Based Methods

Within the domain of water leak detection, hardware-based techniques have evolved significantly, offering a diverse array of methods to address this pervasive issue. A comprehensive examination of these methods revealed distinct advantages and limitations, presenting critical analysis and opportunities for improvement.

4.1.1. Non-Acoustic Methods

Non-acoustic methods are integral within the hardware-based approaches for water leak detection. They encompass techniques such as visual surveys, gas injection, thermography, ground-penetrating radar (GPR), and negative-pressure waves.

The visual survey method, a traditional mainstay, leans heavily on the observational judgment of the surveyor [28]. While this approach is accessible, its subjectivity and limited coverage present persistent challenges. The adjunct use of trained dogs, with their ability to detect leak-related substances through their acute sense of smell [15], offers promise. However, standardized training protocols and a comprehensive assessment of canine reliability across diverse field conditions are imperative to realize this potential.

The gas injection technique has been used to detect water leaks. This method involves injecting a non-toxic gas that is both water-insoluble and lighter than air into a specific section of a water pipe that has been isolated [29]. This technique, especially the employment of sulfur hexafluoride SF6, stands under the ecological spotlight due to its high global warming potential. The gas injection method can detect water leakage in all pipe materials, with a diameter range of 75 mm to 1000 mm. However, the success of this method is heavily dependent on the specific site conditions, including soil type, wind speed, and temperature. Notably, its limitations in detecting smaller or pinhole leaks necessitate innovation to broaden its applicability [30].

Thermography is a method rooted in detecting thermal anomalies in the soil adjacent to a buried pipe due to water leakage [31,32]. Thermal anomalies above pipes can be detected using infrared cameras mounted on a vehicle or aircraft or held by hand [33]. This technique holds theoretical promise, and it is capable of covering large areas without requiring any excavation [34] but remains vulnerable to the influence of external factors, such as solar radiation, wind speed, and ground moisture. These environmental variables necessitate a meticulous consideration when employing this technique.

Ground-penetrating radar (GPR) has been used as a non-destructive geophysical method for leak detection [16,35,36]. This technique operates by emitting electromagnetic waves into the ground, relying on anomalies in dielectric properties to reflect these waves back to the surface. The depth of these anomalies is determined by analyzing the time lag between the transmission and reception of waves, ultimately revealing the size and shape of buried objects. When it comes to detecting water leaks, GPR offers two methods: identifying soil voids formed due to leaks or pinpointing sections of pipes that appear deeper than they are, owing to changes in dielectric properties caused by saturated soil. This method applies to both metallic and plastic pipes, but it demands a substantial investment of both time and resources. Moreover, the challenge of selecting the appropriate radio frequency for diverse soil responses complicates its implementation. Additionally, the potential for false alarms stemming from reflected waves due to soil anomalies, such as metal objects, imparts uncertainty regarding its dependability [34].

The propagation of an incoming transient signal through a system generates reflections from any anomalies, encompassing leaks, blockages, or irregularities in surface conditions [37,38]. Notably, the presence of a leak induces an increase in the damping rate or gives rise to reflections within the pressure trace of the transient signal. Inducing hydraulic transient waves, frequently through a water hammer, has become a conventional method [39]. This hydraulic process triggers a negative-pressure wave upon abrupt pipe rupture, traversing in both directions and reflecting at the pipeline boundaries. The leak’s location is ascertained by the temporal arrival of the known-speed negative wave at each pressure transducer, while the magnitude of the transient wave permits estimation of the breach size. The transient-based approach exhibits several merits: non-invasiveness, cost-effectiveness, and a broad operational range [40]. Presently, leak detection strategies grounded in transient signals are categorized into two primary domains: time-domain techniques [41] and frequency-domain methods, which can be derived from time-domain data through fast Fourier transforms or wavelet transforms [42,43]. Despite its advantages, the transient-based approach for leak detection is not without limitations. One notable constraint lies in the necessity for specific boundary conditions and signal characteristics, which may not be applicable to all pipeline systems. Furthermore, the accurate estimation of the leak size and location can be compromised by complexities arising from multi-component pipe networks. Additionally, the reliance on computational methods, such as fast Fourier transforms and wavelet transforms, for signal analysis introduces the need for rigorous calibration and validation. Future research endeavors should prioritize the development of more robust algorithms and models capable of accommodating diverse pipeline configurations. Furthermore, the integration of advanced sensor technologies and data fusion techniques could enhance the precision and reliability of transient-based leak detection.

4.1.2. Acoustic Methods

Acoustic methods represent a significant subset of hardware-based techniques for water leak detection, relying on sound and pressure variations to identify leaks. Since the 1850s, the utilization of listening sticks, or stethoscopes, has served as a conventional technique for leak detection [44]. These mechanical instruments, fabricated from metal, wood, or plastic, are deployed to audibly discern leaks on fittings, hydrants, or service connections and precisely determine their location. When water forcefully escapes from a high-pressure leak, it generates vibrations that propagate through the pipeline as structure-borne noise. This auditory signal is subsequently conveyed through the steel shaft of the listening stick, audible through an earpiece [28]. Primarily implemented on metallic pipelines featuring diameters ranging from 75 mm to 250 mm and operational under pressures exceeding 10 m [30], listening sticks underwent development in the mid-1960s. This evolution facilitated the integration of battery-powered sound amplifiers into these devices, consequently enhancing their leak detection capabilities [44]. These amplified devices, known as ground microphones or geophones, function analogously to their mechanical counterparts. Geophones identify subterranean noise from surface-level positions, which is particularly valuable in areas characterized by minimal fittings. Distinguishingly, geophones exhibit the capacity to detect leaks in plastic pipelines, low-pressure environments, and regions beset by elevated noise interference originating from factors such as traffic, water utilization, ground perturbations, and wind [45]. Nonetheless, this method is time-intensive, and its effectiveness is contingent on the operator’s experience [46].

The leak noise correlation method relies on comparing the noises detected at two different measurement points within buried pipelines. Assuming consistent pipe material and diameter, noise travels at a constant velocity from both directions of the leak. The sensors detect the leak signals, which are wirelessly transmitted to a correlator that uses the cross-correlation method to pinpoint the leak location. Ideally, if the leak is equidistant from the two sensors, the noise would be detected at the same time. However, in most cases, the leak is located asymmetrically between the measurement points, resulting in a time shift measured by the correlation process [46]. Figure 7 demonstrates the fundamental principle of this method.

The time delay between the two signals is determined by the location of the sensors at easily accessible points, such as valves or fire hydrants, and can be calculated as follows:

$$\Delta \mathrm{T}={\mathrm{T}}_2-{\mathrm{T}}_1={\displaystyle \frac{{\mathrm{L}}_2-{\mathrm{L}}_1}{\mathrm{V}}}$$

(1)

Here, T₁ and T₂ denote the arrival times of signals 1 and 2, respectively, L₁ and L₂ represent the distances from the leak to sensors 1 and 2, and V is the sound propagation velocity in the pipe. The distance from the closer sensor to the leak site can be calculated by substituting L₂ = D − L₁ in Equation (1), where D is the distance between the two sensors:

$${\mathrm{L}}_1={\displaystyle \frac{\mathrm{D}-\mathrm{V}\Delta \mathrm{T}}{2}}$$

(2)

Leak noise correlation requires a noise signal that can be generated using accelerometers and hydrophones. Accelerometers sense the acceleration of vibration induced by leak signals and can be installed directly on the pipe or attached to fire hydrants or underground valves. These sensors are more responsive at high frequencies and work well for metallic pipes [46]. On the other hand, hydrophones sense the sound induced by leak noise in the water core of the pipe and are placed at fire hydrants or air-release valves using special fittings. These devices are more effective than accelerometers for low-frequency leak signals encountered in the case of plastic pipes and larger-diameter mains [30]. The limitations of the leak noise correlation technique in pinpointing leaks in plastic pipelines are due to the viscoelastic properties of the plastic pipes, which absorb sound energy and weaken the sound waves as they travel along the pipe [47]. Although the cross-correlation method was accurate within a 30 to 60 cm distance between the sensors, it was deemed too expensive, time-consuming, and had a limited range [48]. This method is enhanced by the use of noise loggers, which are compact units consisting of an acoustic sensor coupled with a programmable data logger and a communication module [46]. The loggers are arranged in a group of six or more at adjacent pipes, spaced between 200 to 500 m apart in metallic networks and as low as 80 m in plastic pipes. They are designed to automatically activate during nighttime hours to monitor the acoustic vibrations generated by leakages in the water system and provide actionable information about the location of the leaks. Using permanent acoustic noise loggers with wireless data transmission has significantly improved leak detection efficiency in water distribution networks. The data collected from these loggers can be analyzed using the limitations of leak noise loggers, including their reliance on acoustic detection susceptibility to external noise interference. Future research should focus on enhancing their capabilities through advanced sensor technologies and noise-filtering algorithms.

4.1.3. Inline Technologies

Inline technologies for detecting water leaks in pressurized pipelines have witnessed significant advancements, offering promising alternatives to traditional methods for more efficient and accurate leak detection. Tethered systems, exemplified by the Sahara System, and free-swimming systems, typified by the SmartBall, constitute two examples of inline technologies to detect water leakage in pressurized pipelines. The Sahara System incorporates a hydrophone acoustic sensor for inline leak detection [18]. Operating within a pipeline pressurized between 0.3 and 13.8 bars, the hydrophone is propelled through the pipeline by water flow, detecting leakage noise and marking the leak’s location [49]. While capable of detecting leaks as small as 1 L/h and unaffected by pipe material or soil type, the technology is relatively expensive and may face challenges with tether cable path blockages by various factors, such as drag, friction at bends, and the presence of inline valves [50]. In contrast, free-swimming systems, exemplified by the SmartBall, offer an innovative approach. The SmartBall, equipped with an acoustic sensor, accelerometer, magnetometer, GPS-synchronized ultrasonic transmitter, and temperature sensor, silently traverses the pipeline using water flow [19]. Its spherical design eliminates noise interference, allowing the acoustic sensor to operate effectively [51]. The SmartBall can survey extended pipeline lengths, up to 43 km, with specific flow rates and battery life [30]. It detects leaks as small as 0.15 gallons per hour, notwithstanding the pipeline material, and can identify pockets of trapped gas in pressurized pipes. However, the potential for false alarms due to acoustic characteristics resembling real leaks remains a concern. This technology has been tested in several case studies, including a successful survey by the Birmingham Water Works Board (BWWB), which used the SmartBall for inspecting 12 km of 1000 mm reinforced concrete pipe, detecting 26 leaks of varying sizes with close location accuracy and preventing long-term water loss [52].

The application of fiber optic sensor systems represents another inline approach to pipeline leak detection. By installing fiber optic cables along the pipeline’s length, changes in the thermal characteristics of the surrounding soil can be monitored to detect and locate leaks [53]. These systems measure temperature and strain along the cable, enabling real-time monitoring to prevent failures, timely detect issues, and facilitate repairs [54]. Additionally, existing fiber optic telecommunication lines may be utilized for temperature monitoring and leak detection [55]. Nonetheless, these advanced methods warrant further research and development to address potential limitations and optimize their effectiveness for various pipeline conditions and materials.

Future research on hardware-based methods should focus on refining non-acoustic approaches, acoustic tools, and inline technologies. The continuous refinement of gas injection techniques, including exploring alternative tracers, holds promise for improving leak detectability in varied environments. Research into machine learning algorithms can automate detection processes, particularly in thermography. Ground-penetrating radar technology advancements can enhance resolution and depth capabilities for effective subsurface leak detection. Optimizing negative-pressure-wave systems and exploring smart valves or controls can improve their efficacy. Regarding acoustic methods, signal-processing and machine learning advancements can automate analysis, and compact, cost-effective leak noise loggers are essential for widespread deployment. Inline technologies, such as Sahara, SmartBall, and fiber optic systems, require inquiry into heightened sensitivities, expanded applicability, and innovative methodologies for increased detection precision. These research efforts aim to enhance the water leak detection accuracy, efficiency, and adaptability across diverse water distribution infrastructures.

4.2. Software-Based Methods

The software-based approaches collect relevant data from water distribution systems, such as water flow, temperature, and pressure. They can be classified into six groups: flow variation, pressure point examination, water balance, and numerical methods, as well as artificial intelligence and pattern recognition.

4.2.1. Flow Variation Method

The flow variation method is based on the premise that a significant increase in the flow rate at the inlet or outlet of the water distribution system indicates the presence of a leak. A leak alarm is triggered if the flow rate surpasses a pre-set threshold within a specific timeframe. Van Thienen [56] introduced a technique called the comparison of flow pattern distributions (CFPD), which can detect and measure modifications in the volume of water supplied during two distinct timeframes. The technique involves plotting one dataset against the other and identifying anomalies through the best linear fit with slope and intercept. Consistent changes, which may be caused by variations in weather or population size, are indicated by the slope, while inconsistent changes, attributed to increased leakage, are indicated by the intercept. This approach has proven effective in detecting leaks in water distribution systems (WDS). Additionally, it can be used to examine archived flow data for irregularities [57]. Van Thienen and Vertommen [58] created Cuboid, an automated feature recognition tool for CFPD analysis of flow data in supply areas, to enable real-time detection of new leaks or identification of unregistered changes in valve status. The flow variation method faces limitations due to its sensitivity to minor fluctuations and potential false alarms from everyday water usage. Future research should refine the method by developing advanced algorithms to distinguish genuine leaks from other variations. Integrating complementary sensor technologies, such as pressure and acoustic sensors, could enhance the reliability, and scalability should be a key focus, especially for large water distribution networks.

4.2.2. Pressure Point Analysis

The pressure point analysis (PPA) method operates on the premise that a pipeline’s pressure decreases when a leak develops. To detect a leak, pressure measurements can be employed to monitor the rate of pressure change, identify abnormally low-pressure levels, or compare current pressure readings to a running statistical analysis created from prior data [22]. Pressure measurements are conducted using sensors that are installed in the pipeline and can range from a single measurement point to multiple pressure transducers distributed along the pipeline [59]. A leak alarm is triggered if the pressure drop exceeds a pre-defined level. The PPA technique can be utilized for pressurized pipelines ranging from 75 to 1000 mm in diameter and is unaffected by the material of the pipes or properties of the liquid. Employing this method for pressure monitoring in a network offers advantages in terms of simplicity and cost. However, it cannot detect minor leaks because they do not significantly impact the pressure. Additionally, a pressure drop in a network can be caused by events other than leaks, leading to the possibility of false alarms with the PPA technique.

4.2.3. Water Balance Method

Water balance, mass balance, or water audit is the most employed technique for quantifying total water losses and leakage in a network. This approach is based on the principle of mass conservation, which asserts that the amount of fluid that enters a pipe section must either remain in the section or exit it [22]. To identify a leak, the water balance method compares the amount of water introduced into the distribution network to the sum of the water consumed or used components, with any significant difference indicating the presence of a leak beyond an established tolerance level [60]. The accuracy of the mass balance technique is dependent on precise measurement of the mass flowing into and out of a pipeline [61]. The primary limitation of the water audit technique is its reliance on the assumption of steady-state conditions. This means that the volumes being balanced must be calculated over a longer time interval to prevent false alarms [62]. Additionally, while the water balance method is effective in quantifying total water losses, it is limited in pinpointing the exact locations of leaks in the network. As such, it is often necessary to complement this approach with other active leakage detection methods.

4.2.4. Numerical Methods

Numerical modeling techniques for detecting water leakage primarily involve the conventional hydraulic modeling approach, used through specialized software tools, and the methods based on transient events. Hydraulic modeling is another software-based tool for monitoring and managing water supply networks in the present and the future. This technique is used to anticipate problems in planned networks, design effective interventions, and operate and manage water supply systems. The hydraulic modeling involves organizing data into a hierarchical structure and calibrating the model until it closely reflects reality. However, model calibration remains a challenging issue [63]. The calibration process aims to reduce the discrepancies between the model results and measured values by adjusting model parameters, such as roughness, water consumption, and water losses [64]. The genetic algorithm is a commonly used method for calibrating hydraulic models [65,66,67]. The accuracy of the hydraulic model can be influenced by various factors, such as the quality of the database used (e.g., topography, pipe diameter, and length), uncertainty in valve status, and the accuracy of measurement systems (such as sensors and data loggers). The hydraulic model can estimate the amount of leakage within specific areas by reducing the differences between the measured and numerically estimated pressure and flow rates [68,69]. Additionally, it can serve various purposes for managing leakage, such as modeling leakage as pressure-dependent demand [70,71,72] and creating pressure management plans to control leakage [73]. As hydraulic modeling continues to play a significant role in water network management, future research should focus on enhancing model calibration techniques, improving data accuracy, and developing innovative strategies for more proactive and precise water supply network management.

Inverse transient analysis (ITA) is a sophisticated method used for detecting leaks in water distribution systems by analyzing pressure transients that propagate through the pipeline due to disturbances, such as valve closures or pump failures. This method operates by comparing the measured transient pressure data from the pipeline with simulated data obtained from a mathematical model of the system. The goal is to minimize the difference between the actual and simulated data, treating the leak detection problem as an inverse problem solved using optimization algorithms. ITA has been particularly successful in identifying leaks by exploiting pressure waves’ sensitivity to anomalies, such as leaks, within the pipeline [74].

Kapelan et al. [75] applied ITA in controlled laboratory conditions and quasi-field environments, demonstrating its efficacy in identifying leak locations and sizes when the system’s physical and hydraulic characteristics are well known. Their findings showed that the accuracy of ITA depends significantly on accurate system modeling, making it highly suitable for controlled environments. Covas and Ramos [76] further investigated ITA by applying it to polyethylene pipe systems at Imperial College London and Thames Water Utilities. The results of their study confirmed that ITA could successfully detect and locate leaks, provided that the pipeline’s boundary conditions and physical characteristics were well understood and controlled. However, they also highlighted the challenges of extending ITA to real-world applications, where noise and system uncertainties can interfere with the accuracy of leak detection.

To address the computational and noise-related challenges of ITA, recent studies have incorporated advanced optimization techniques. Brahami et al. [77] introduced a genetic-algorithm-based ITA framework, which significantly reduced the computation time and enhanced the leak localization accuracy, even in systems affected by noise. Despite its advantages, ITA faces limitations when applied to large-scale systems with complex topologies, noise interference, and uncertain boundary conditions. Future research aims to improve ITA’s robustness by integrating more advanced signal-processing techniques, machine learning, and hybrid approaches that combine ITA with other leak detection methods, such as acoustic sensors or fiber optic technologies [78].

4.2.5. Support Vector Machines

The support vector machines (SVMs) have recently become a valuable tool in water leak detection, providing robust performance in handling complex, high-dimensional data from pipeline systems. The primary mechanism of SVMs involves finding the optimal hyperplane that separates different classes, such as “leak” and “no leak”, based on input features, such as pressure, flow, and acoustic signals. The SVM’s effectiveness in leak detection is derived from its capacity to handle both linear and non-linear relationships through the kernel trick, which projects data into a higher-dimensional space where they become linearly separable. This flexibility enables SVMs to adapt to complex, real-world water distribution systems, where leak detection can be challenging due to environmental noise, variability in operating conditions, and the presence of non-leak signals [79,80]. The SVM was utilized to detect leaks in a simulated water pipe network, analyzing pressure data from various junctions. The system used data from sensors in a water distribution network and the SVM algorithm was trained using the radial basis function (RBF) kernel to classify and detect leakages in the network. The system achieved a root mean square error (RMSE) of 0.06785 for leakage size prediction and 0.1382 for location prediction, indicating a high level of accuracy [81]. Another study focused on the application of a one-class SVM (OCSVM) for anomaly detection in water distribution systems. The OCSVM was used to detect leaks in a cold climate using a hydrant-mounted sensor. The system was designed for minimally invasive monitoring, and the results demonstrated 97% accuracy in classifying leak events from acoustic data [82]. The main advantage of using the SVM in water leak detection lies in its strong classification ability, even when data are noisy or when the relationships between variables are non-linear. However, the limitations of the SVM include its reliance on high-quality training data and its sensitivity to parameter tuning, particularly with kernel selection, which can affect model performance. Additionally, SVMs can be computationally intensive when working with large datasets, which may restrict real-time application for large-scale water networks [83]. Future improvements in SVM-based systems for leak detection include hybridizing SVMs with other algorithms, such as random forests or neural networks, to improve robustness and accuracy. Furthermore, the integration of IoT technologies and advanced sensors could enable more comprehensive data collection, enhancing the precision of SVM models in identifying and locating leaks.

4.2.6. Deep Learning

Deep learning techniques, particularly artificial neural networks (ANNs) and convolutional neural networks (CNNs), have become useful tools for enhancing the detection and localization of water leaks in distribution networks [84,85]. Artificial neural networks (ANNs) have recently been employed to detect and locate leaks in water supply networks [86]. The ANN establishes a relationship between input and output data. In the case of water networks, input data characterize the normal and abnormal states of the system, with and without leaks. The ANN acts as a classifier to estimate the current state of the system and identify the leaks based on available data [45]. The ANN method has been used to detect and localize leaks based on flow and pressure data. It effectively detected bursts in a real case study of 144 zones in the United Kingdom [87]. Furthermore, the ANN model can be integrated with Geographic Information Systems (GIS) to assess water leakage and prioritize pipeline replacement [23]. Jafar et al. [88] utilized data collected over 14 years to predict pipe failures in the water supply system of Wattrelos City in France using an ANN method. The model also determined the optimal time for pipe renovation in the system [45,83]. While ANNs have showcased significant promise in leak detection, certain limitations must be highlighted. These include the necessity for extensive, high-quality training datasets, challenges in model interpretability, and computational resource requirements. To drive future research in this domain, it is imperative to focus on refining training strategies, enhancing model robustness across diverse operational conditions, and exploring techniques for improving model interpretability and human–AI collaboration in leakage management.

Convolutional neural networks (CNNs) are particularly useful in analyzing complex datasets, such as acoustic emissions and pressure residual maps. Unlike ANNs, which rely on structured numerical input, CNNs process more complex data representations, making them ideal for applications such as acoustic-based leak detection. Several studies have employed CNNs to convert pressure maps into images, enabling models to identify leaks with high precision by detecting subtle changes in pressure or acoustic patterns. For instance, one study used CNNs to analyze acoustic wave files collected from a smart water network in Adelaide, Australia, achieving an accuracy of over 90% in distinguishing leaks from background noise [86]. Javadiha et al. [89] explored deep learning for leak localization using CNNs. The study introduced a novel approach that converted pressure residual maps into images, applying CNNs to identify leak locations with high accuracy. This method used pressure data from a District Metered Area (DMA) in Hanoi, Vietnam, and demonstrated strong generalization by applying Bayesian reasoning over time to handle uncertainty and noise. The study reported a significant classification accuracy, showcasing how CNNs can process pressure data for leak detection in real-world scenarios. Nam et al. [90] focused on using CNNs for leak detection based on actual leak sounds collected from real-world pipelines. The study utilized recurrence plots to transform time-series acoustic data into 2D images that could be processed by the CNN model. The model achieved a detection accuracy of over 80% across 15 of the 20 leak cases tested, demonstrating that acoustic-based CNN models could effectively differentiate between leak sounds and background noise. This method was validated with field data, making it particularly valuable for early leak detection in practical settings.

The combination of ANNs and CNNs has enabled real-time, automated leak detection, significantly reducing the need for manual inspections and minimizing water losses. However, challenges remain, including the need for large, high-quality datasets to train these models and the computational resources required for deploying them in real-world systems. Future research should focus on improving the scalability and efficiency of these models, making them more robust in diverse operational environments, and enhancing their integration into smart water management systems for optimized leak detection and network efficiency [8].

5. Assessment of Leak Detection Methods

The effectiveness of leak detection methods can be evaluated using five criteria:

• Leak sensitivity—the ability to detect small leaks.
• Leak location—the capability to pinpoint the leak.
• Real-time monitoring—the possibility to continuously monitor the WDS.
• False alarm—the frequency of generating false alarms when leaks do not exist.
• Cost.

Table 4. Comparative analysis of various techniques for detecting water leaks.

Category	Methods	Leak Sensitivity	Accuracy	Leak Location	Real-Time Monitoring	False Alarms	Cost
Non-Acoustic	Visual survey	Low	Moderate	Yes	No	Low	Low
	Gas injection	High	High	Yes	No	Low	High
	Thermography	Medium	Medium	Yes	No	Medium	High
	Ground-penetrating radar	Medium	Medium	Yes	No	Medium	High
	Negative-pressure waves	High	High	Yes	No	High	Medium
Acoustic	Manual listening sticks	Medium	Medium	Yes	No	Medium	Low
	Leak noise correlation	Medium	Medium	Yes	Yes	Medium	High
	Leak noise loggers	Medium	Medium	Yes	Yes	Medium	High
Inline	Sahara	High	High	Yes	No	Low	High
	SmartBall	High	High	Yes	No	Low	High
	Fiber optic	High	High	Yes	Yes	Low	High
Software-based	Flow variation method	Low	Low	No	Yes	Low	Low
	Pressure point analysis	Medium	Medium	No	Yes	Medium	Medium
	Water balance method	Low	Low	No	Yes	Low	Low
	Numerical methods	Medium	Medium	No	Yes	Medium	Medium
	Support vector machines	High	High	Yes	Yes	Low	Medium
	Deep learning	High	High	Yes	Yes	Low	High

outlines the performance characteristics of the water leak detection methods detailed in this study. Non-acoustic methods, such as visual surveys and gas injection, provide high sensitivity and an accurate leak location but lack real-time monitoring, with associated low costs and occasional false alarms. Acoustic methods, such as leak noise correlation and loggers, offer real-time monitoring with medium sensitivity and infrequent false alarms. Inline technologies, including Sahara, SmartBall, and fiber optic, showcase high sensitivity and accurate leak location but do not provide real-time monitoring and have high costs. Software-based methods, such as water balance, hydraulic modeling, and artificial intelligence, vary in sensitivity, leak location capability, and real-time monitoring, with associated costs ranging from low to high.

Table 5. Limitations and required improvements of various techniques for detecting water leaks.

Category	Methods	Limitations	Improvements for Future Research
Non-Acoustic	Visual survey	Detection of visible surface water leaks only [91].	Develop complementary technologies for a more comprehensive leak assessment beyond detecting surface water leaks only.
	Gas injection	Accuracy affected by environmental factors, such as wind, temperature, and gas dispersion characteristics [92].	Refining the gas injection process, including the selection of suitable tracer gases.
	Thermography	Insufficient temperature differential [93].	Improve its sensitivity to detect minor temperature fluctuations.
	Ground-penetrating radar	Difficulty in differentiation between water pipes and other buried objects [94].	Adapt to diverse soil types and differentiate various subterranean characteristics.
	Negative-pressure waves	Accuracy highly affected by pipe material, diameter, and network complexity [95].	Integrate data analysis techniques to address these complexities.
Acoustic	Manual listening sticks	Less efficient in large systems or noisy urban environments [96].	Develop noise-filtering technologies and automated detection mechanisms to reduce human error.
	Leak noise correlation	Sensitive to pipeline materials, diameters, and network complexity [97].	Integrate calibration methods to consider diverse pipeline configurations.
	Leak noise loggers	Interference between the leak signals and environmental noise [2].	Improve sensitivity by focusing on advanced signal-processing algorithms.
Inline	Sahara	Need for suitable access points in the pipeline [98].	Enhance the system to cover pipelines with restricted access.
	SmartBall	Difficulty navigating complex pipeline geometries [99].	Improve adaptability for undiggable pipelines.
	Fiber optic	Accuracy affected by pipe materials and complex geometries [100].	Optimize fiber optic technology for a wider range of pipeline materials and configurations.
Software-based	Flow variation method	Changes related to consumer behavior or water demand rather than leaks [101].	Develop advanced algorithms to differentiate between various causes of flow changes.
	Pressure point analysis	Pressure fluctuations caused by water hammer rather than leaks [102].	Integrate pressure analysis algorithms to account for transient pressure effects.
	Water balance method	Inaccuracies in meter data, apparent losses [103].	Combine this method with other techniques to improve data accuracy and identify hidden sources of water loss.
	Numerical methods	Calibration [104].	Complement the model with adaptive techniques that can integrate real-time data for effective calibration.
	Support vector machines (SVMs)	Requires labeled training data, computationally intensive for large networks [105].	Enhance SVM algorithms with unsupervised learning or online training methods.
	Deep learning	Requires large datasets, prone to overfitting without adequate data [106].	Improve model generalization and diversify the data and parameters.

highlights the limitations and the potential improvements for various water leak detection methods. In non-acoustic methods, visual surveys are constrained to detecting visible surface water leaks only, suggesting the need for complementary technologies to assess leaks comprehensively. Gas injection accuracy is affected by environmental factors, urging the refinement of the process and the selection of suitable tracer gases. Thermography faces challenges in detecting minor temperature fluctuations, indicating the necessity for sensitivity improvements. Ground-penetrating radar encounters difficulty in differentiating between water pipes and other buried objects, prompting adaptation to diverse soil types. Negative-pressure waves’ accuracy is impacted by pipe characteristics, suggesting the integration of data analysis techniques. Acoustic methods, such as manual listening sticks, are less efficient in large systems, warranting the development of noise-filtering technologies. Leak noise correlation is sensitive to pipeline parameters, emphasizing the need for integration with calibration methods. Inline technologies, such as SmartBall, face challenges in finding the exact leak location, necessitating the integration of data-processing algorithms. Fiber optic accuracy is affected by pipe materials, indicating the need for optimization. Software-based methods, such as water balance, may show inaccuracies, suggesting a combination with other techniques for improved data accuracy. Hydraulic modeling requires effective calibration techniques, and artificial intelligence performance can be enhanced by diversifying data and parameters. These identified limitations provide clear directions for future research to refine and advance water leak detection technologies.

6. Smart Water Methods (SWMs)

Implementing information and communication technologies (ICT) in urban networks has given rise to the concept of a Smart City, in which infrastructure components are made more intelligent, interconnected, and effective [107]. After the successful implementation of the smart grid in the electrical sector [108], water utilities were motivated to adopt a similar approach in water distribution systems (WDSs). This led to the emergence of the concept of smart water networks. This concept is based on equipping pipes with sensors that monitor pipe performance and condition [109,110]. The primary aim of a smart water network is to establish a comprehensive monitoring system that integrates sensor technology, data acquisition, data interpretation, and decision-making processes for the real-time management of pipeline assets [111]. The concept and architecture of a smart water network are demonstrated in Figure 8.

A smart water network can be broken down into five distinct layers, each with its own set of responsibilities [113]:

• Physical layer—includes all physical components of the WDS, such as pipes, valves, reservoirs, and pumps.
• Sensing and control—involves measuring water parameters, such as flow, pressure, and quality, among other important characteristics.
• Collection and communication—entails collecting and transmitting measured data automatically and continuously.
• Data management and display—requires creating a database platform to organize collected data and presenting it through visualization tools, such as GIS, and setting up a customer information system.
• Data fusion and analysis—involves using modeling software to study network responses, applying predictive analytics for event detection, leak detection and localization, decision support, etc. The entire network can be managed remotely and automatically through communication channels.

Günther et al. [114] developed a smart water network using an experimental water distribution system called EWDS-TUG, including artificial customers, sensors, and control magnetic valves. By applying the smart grid concept, water utilities can minimize water loss through consistent and reliable monitoring of water distribution networks [115]. Real-time monitoring of water distribution networks offers several advantages, including improved network visibility, water and energy savings, early detection of network inefficiencies, reduced need for on-site inspection, and better quality of services for customers.

A popular way to determine the level of leakage in smart water networks is by analyzing the minimum night flow (MNF) between midnight and 4 a.m. This analysis is typically performed between 2 a.m. and 4 a.m., when customer demand and network pressures are high, making leakage more apparent in a District Metered Area (DMA) [28]. To create a District Metered Area (DMA), the water distribution network is divided into discrete zones by closing valves [116]. The volume of water entering and leaving the zone is metered to determine the amount of leakage. A leak alarm is triggered when the minimum night flow (MNF) in the DMA exceeds a threshold set by water utility companies. The MNF depends on several factors, including the age and length of pipes, the number of connections, and the pressure in the network.

A wireless sensor network (WSN) is an application of the smart grid in water networks. It consists of interconnected wireless nodes that can sense, compute, and communicate wirelessly, but also have power supply requirements. In a pipeline-monitoring WSN, sensor nodes work together to collect and transmit data to a base station [117]. However, certain factors need to be considered when implementing a smart pipe, such as the radio propagation channel, power and memory limitations, and efficient routing protocols [118]. PipeNet is a wireless sensor network system that is designed to detect, locate, and quantify leaks and bursts in water transmission pipelines [119]. The development of PipeNet is segmented into three stages. The first stage concerns the validation of a small-scale prototype deployment to assess the sensors’ durability. In the second stage, advanced data acquisition and analysis techniques are tested and validated in a laboratory setting, including time-synchronized data collection and acoustic leak detection. The third and final stage involves merging the first two stages to create a comprehensive real-time monitoring solution.

Another prominent example of WSN application is the Wireless Water Sentinel project, WaterWise@SG, operating in Singapore, known for its comprehensive management of hydraulic, acoustic, and water quality parameters. This platform enables applications such as predicting water demand and hydraulic state, detecting events, such as pipe bursts, and identifying longer-term trends through data mining [120]. Sensor nodes are used to cover an area of 60 km², packaged in clear plastic acrylic tubes with PVC caps at each end, supporting the attachment of pressure, hydrophone, and flow meter sensors. The system detects and locates pipe bursts by analyzing pressure, flow, and acoustic signals, providing a comprehensive solution for smart water management.

The ICT Solutions for Efficient Water Resources Management (ICeWater) project also aims to enhance freshwater supply stability by leveraging smart grid technology [121]. The system includes three layers: sensor installation, data collection and processing, and user interaction. The project is being tested in Milan (Italy) and Timisoara (Romania), where statistical and machine learning methods are used to predict water loss. The Milan demo site classified data to identify typical urban water demand and provide hourly short-term forecasts [122].

The iWIDGET project is another notable smart water project, providing an online platform for efficient urban water management through information and communication technology (ICT), rendering near-real-time water consumption data and decision-support tools to households and utilities [123]. The project primarily focuses on the processing and analysis of water use data and providing feedback. Operating across Greece, Portugal, and the United Kingdom, this project empowers water utilities to devise water pricing models and forecast demand, offering a pivotal utility in the efficient management of water resources [124].

Within the Scientific City in northern France lies the SunRise Smart Water project, a demonstration of smart water systems within the “SmartWater4Europe” European initiative. Encompassing a wide range of urban networks, including water distribution systems, the Scientific City faces water loss challenges due to various factors, including aging infrastructure, soil movement, and unaccounted water volumes for specific purposes, such as network cleaning or construction works. To overcome these challenges, hydraulic sensors, automatic metered readers (AMRs), and other techniques are employed to measure flow, pressure, and noise levels in pipelines and inside the buildings. The combination of various methods, including water balance, minimum night flow, leak noise loggers, hydraulic modeling, and artificial neural network approaches, has helped to detect multiple leaks quickly and to reduce the non-revenue water by 36% [125].

Smart water management systems play a crucial role in improving water leak detection. However, they face challenges, including high implementation costs, cybersecurity concerns, and data volume management. The heterogeneity of existing water infrastructure, often aged and incompatible with the sophisticated sensor and communication technologies necessary for smart monitoring, is also another challenge. Moreover, the lack of standardized protocols and interoperability intensifies the complexity of integrating smart systems. These challenges must be addressed, and future research should focus on several areas that require improvement. First, enhancing the integration of advanced data analytics, machine learning, and artificial intelligence can refine the leak detection accuracy and facilitate automated decision-making processes. Research into more cost-effective and energy-efficient sensors and communication technologies is essential to lower barriers to adopting smart water management solutions. Furthermore, developing standardized cybersecurity protocols and data encryption measures is crucial for safeguarding these systems from cyber threats. Initiatives promoting universal communication protocols may enhance interoperability, facilitating seamless integration of diverse technologies across water networks.

7. Discussion

The findings of this review highlighted both the advancements and ongoing challenges in water leak detection technologies, emphasizing the need for continuous innovation. Traditional methods, such as acoustic monitoring and manual surveys, remain useful for small-scale operations due to their relatively low cost and straightforward implementation. However, these methods show significant limitations in terms of real-time detection, accuracy in complex networks, and their dependence on environmental conditions. For instance, studies conducted by Negm et al. [9] and Yu et al. [126] showed that acoustic methods are primarily effective in metallic pipelines but struggle to detect leaks in plastic pipes due to the sound absorption properties of such materials. Furthermore, traditional methods are labor-intensive and prone to operator error, while environmental noise can lead to false positives or missed leaks.

Environmental factors, such as soil conditions, temperature fluctuations, and external vibrations from urban infrastructure, further complicate the use of non-invasive methods, such as thermography and ground-penetrating radar. These environmental variables can produce misleading signals, leading to false alarms or missed leaks. Similar findings from Latif et al. [127] support the need for integrated approaches, as combining multiple technologies—such as thermographic data with GPR—could mitigate some of these issues, improving detection accuracy under diverse conditions.

In contrast, modern technologies, including inline systems such as SmartBall and fiber optic sensors, offer high precision and continuous monitoring, making them ideal for large-scale or high-pressure networks. Similar to the findings of Mergelas and Henrich [49], who also emphasized the strengths of inline technologies in specific contexts, this review recognized that high costs and complexity limit widespread adoption, particularly in developing regions, where financial constraints and lack of infrastructure pose significant challenges. Additionally, the need for skilled operators to manage these technologies creates barriers for regions with limited technical expertise.

The application of artificial intelligence (AI) and machine learning (ML) in water leak detection introduces new possibilities for enhanced accuracy and automation. Techniques such as support vector machines (SVMs) and deep learning models enable the analysis of large datasets, facilitating more precise leak detection and localization. Similar investigations by Hu et al. [11] affirmed the growing importance of AI-driven systems, yet these technologies require access to high-quality, labeled datasets, which may not be readily available in many regions. Additionally, the computational demands of these methods can limit their use in real-time applications, particularly for utilities with limited processing capabilities. Future research should focus on optimizing AI models to function effectively with smaller datasets and developing real-time algorithms that reduce the computational load.

Smart water networks represent a promising approach to water management, integrating sensors, data analytics, and real-time monitoring to provide actionable insights into leak detection and system performance. However, the high implementation costs, cybersecurity risks, and lack of standardization across different sensor technologies remain significant barriers to large-scale deployment. Without established standards for interoperability, integrating different sensor technologies within a single system becomes complex, increasing both costs and operational challenges. Furthermore, the rising dependence on interconnected systems raises concerns about cybersecurity vulnerabilities, as attacks on critical infrastructure could disrupt water supplies or compromise sensitive data.

An important issue to address is the inequality in access to these advanced technologies. While developed regions can adopt modern technologies more easily, many developing countries struggle with the financial burden and lack of technical expertise required to implement smart systems. In regions where water scarcity is a pressing issue, the focus is often on basic water access rather than advanced leak detection technologies. Future research should prioritize the development of cost-effective, scalable solutions that can be implemented globally. Hybrid systems that combine affordable traditional methods with AI-driven data analysis might offer a feasible solution for regions with limited resources.

8. Conclusions

Water loss constitutes a significant challenge for many cities, which requires developing and implementing effective methods for water leak detection. This review offered a comprehensive evaluation of existing water leak detection technologies, presenting several key insights, as follow:

• The bibliometric analysis conducted over a 23-year period (2000–2023) provided crucial insights into research trends, key contributors, and publication patterns in the field of water leak detection. Analyzing 600 scholarly articles revealed a growing interest in innovative detection methods, particularly in the past decade, corresponding to increased global awareness of water sustainability challenges. The analysis also identified leading institutions and authors, highlighting the collaborative nature of research in this domain. This information serves as a valuable resource for policymakers and researchers seeking to understand the landscape of water leak detection and to identify opportunities for future collaboration.
• Traditional methods, including acoustic techniques, tracer gas methods, thermography, and ground-penetrating radar (GPR), have demonstrated efficacy in detecting leaks in specific contexts, such as metallic pipelines and underground infrastructures. Acoustic methods are particularly adept at identifying noise or vibrations generated by leaks, but their effectiveness is limited in non-metallic pipes and environments with significant background noise. Tracer gas methods, while reliable, often involve logistical challenges in their application. GPR can reveal underground voids but is influenced by soil conditions and requires careful site selection. Despite their strengths, these methods face limitations, such as high operational costs, dependence on skilled personnel, and the potential for false positives. Future research should focus on enhancing the cost-effectiveness of these traditional methods and minimizing inaccuracies, thereby increasing their applicability across various contexts.
• Software-based methods, including flow variation analysis, pressure point monitoring, and water balance techniques, leverage data analytics and computational models to enhance real-time leak detection capabilities. These approaches provide valuable insights into the operational efficiency of water distribution systems. However, challenges remain, particularly regarding calibration accuracy and the reliability of input data. The successful implementation of these methods hinges on developing advanced calibration techniques and robust data collection systems, ensuring that they can effectively pinpoint leaks and facilitate proactive maintenance strategies.
• The advent of smart water networks represents a major development in leak detection and management. By integrating hardware and software, these systems enable continuous monitoring, real-time data analysis, and predictive maintenance, thereby enhancing the operational efficiency of water utilities. Smart water technologies offer significant benefits, such as early detection of bursts, management of water quality, and optimization of resource allocation. However, the high initial investment costs and the complexity of deploying these systems pose barriers to widespread adoption. Future efforts should prioritize optimizing the cost-effectiveness of smart technologies while simplifying deployment processes to facilitate their integration into existing infrastructures.

Effective leak detection is not only critical for operational efficiency but also for minimizing environmental consequences. Water leaks lead to unnecessary water extraction, depleting vital resources, such as rivers and aquifers, which exacerbates water scarcity, especially in vulnerable regions. Additionally, leaks increase energy consumption for water treatment and distribution, raising the carbon footprint of water utilities. Addressing water leaks contributes to environmental sustainability by conserving water resources, reducing energy waste, lowering greenhouse gas emissions, and preventing soil erosion and infrastructure degradation. Future research should focus on solutions that are not only technologically advanced but also environmentally sustainable, providing long-term benefits for both water utilities and ecosystems.

This review not only assessed the current state of leak detection methods but also highlighted critical research gaps and outlined directions for future studies. The multi-dimensional bibliometric analysis highlighted current trends in the field and emphasized the need for interdisciplinary approaches that combine the strengths of traditional and emerging technologies. By identifying the potential of smart water networks and the integration of advanced data analytics, this study provided actionable insights for researchers and water utilities seeking to enhance their leak detection capabilities.