As the sun lightens the horizon and the weak dawn light spreads across the city, nine million inhabitants of New York City’s five boroughs begin their weekday morning routine. For some, it starts as normal with a cup of coffee, a shower, and a water bottle filled in advance of a heart-pumping spin class. For several neighborhoods, the day is off to a rocky start. Bathroom and kitchen faucets sputter and hiss, toilets go unflushed, and morning showers are skipped as bewildered faces stare at nonresponsive water knobs. At first, water outages are isolated to several blocks scattered throughout the city. As city workers struggle to identify and resolve water outages, the problems cascade and grow. Soon, entire sections of the boroughs are blinking out like faulty Christmas bulbs. Panic sets in as people realize it is not just their apartment, their building, their street. Cellular systems light up with frustrated calls asking, “When water will return?”, “When will someone come?”, and “What is being done!” In this way, water, the lifeblood of the city, slowly seeps away and leaves behind a convulsing, terror-filled wreck as roughly nine million inhabitants cry out in rage and fear.
While this scenario sounds like something from a popular sci-fi series, the reality is much more terrifying. As the water infrastructure of many large cities around the United States (US) age and deteriorate, outages will become more common as population growth and urbanization places greater demands on our already strained water infrastructure. For many cities, the problem of crumbling water infrastructure is all too well known. In 2015, approximately 87% of the US population relied on publically provided fresh water supplies.i As of 2020, US water infrastructure relied on 2.2 million miles of piping. As of a 2024 report, with an average age of 45 years old, this water infrastructure experienced an estimated 250,000 to 300,000 breaks each year in the US.ii Though water infrastructure statistics appear gloomy, advances in artificial intelligence (AI) combined with technology improvements may offer timely solutions for city planners to avert this seeming apocalyptic inevitability. AI solutions integrated with advances in sensor technology may hold the promise of predictive and preventative maintenance, load optimization, and prioritized replacement strategies. Such measures and solutions can improve city planners’ current and future water infrastructure understanding in near real-time to make decisions well in advance of crisis and collapse. While AI is not a panacea for America’s aging and crumbling water infrastructure, it could be the exact technology we need at time when AI could make all the difference for urban centers facing potential water crises with blinders on.
AI & Refined Predictive Modeling
As population growth and aging water infrastructure threaten urban disaster, water planners require a better way to anticipate water distribution network (WDN) pipe failures. AI combined with data analytics may hold the answer and support water planners in their effort to prioritize infrastructure upgrades. Historically, water infrastructure planners used rudimentary and basic statistical pipe modeling to anticipate maintenance or replacement of pipe sections. Planners calculated pipe failure rates based on single-variable, time-exponential models which combined age and pipe type to predict which sections required service.iii These simplistic models failed to consider factors besides pipe type and age which resulted in unanticipated breakage and inefficient water distribution. Additionally, the previous models grouped all pipes of the same type and age together into one homogenous cluster which failed to account for environmental factors which significantly alter pipe failure calculations.
By using AI, and incorporating large datasets which previously did not exist, water infrastructure planners can gain better fidelity in failure modeling. Many cities and municipalities now have datasets for both pipe and soil conditions surrounding pipes. Different soils apply pressure or distribute pressure across lengths of pipe which can either lengthen pipe lifespan or reduce it based on pipe type. Additionally, soil composition combined with weather data can determine impacts of frost or freezing, soil moisture, shrink-swell potential, soil corrosivity, and other environmental factors.iv By including these data sets into pipe modeling, city planners can more accurately anticipate failures in small pipe sections vice addressing predicted failure at the network and pipe type level.v
City planners and water infrastructure modelers are considering AI and machine learning (ML) algorithms to quickly determine networks at greatest risk of failure with more granular certainty based on newly available datasets. AI/ML modeling supports second and third-order risk analysis derived from modeled or simulated failures. Such complex determinations allow city and water planners to prioritize, with a better level of certainty, at risk sections of water distribution networks and resulting impacts if failures do occur. While AI and ML will not prevent collapse of the system, it is helping city planners efficiently get ahead of major infrastructure failures based on better understanding of pipe failure probabilities.
AI-Augmented Water Management
Water management is another area where AI can assist water infrastructure specialists in their battle against aging infrastructure and inefficiency. While aging pipes fail and leak, leading to loss and inefficiency, city planners are looking for solutions to help lessen consumer demand on aging water infrastructure and gain efficiencies in current networks. AI and the internet-of-things (IoT) approach may hold innovative answers for desperate water specialists. With the rise of miniaturization, meshed communications networks, and highly specialized sensors, water infrastructure networks are the beneficiaries of technology innovation. By incorporating networked sensors into water networks, AI solutions can highlight and anticipate problems for water management personnel. As an example, flow sensors installed throughout the network can closely track flow rates, usage data, and pressure levels within the network.vi Using AI to analyze this data, leaks or breakages become readily apparent as an aberration in this data. Abnormally high consumption rates detected by smart sensors which fall outside of historical time-usage parameters, could signify a leak or break in a specific line. City planners could use sensor location data to more accurately identify and locate the leaking or broken section of pipe. AI and IoT sensor networks will directly assist water maintenance personnel to fix outages quickly and thereby reduce loss throughout the whole of the network.
Adopting IoT and AI use in water infrastructure could be more than just a temporary fix for today’s aging infrastructure ailments. As automation, sensor placement, and AI become integrated into water networks, city planners could use digital twin technology to improve water availability and optimize performance as discussed above. Digital twins are the virtual representation of a physical system like a city’s water infrastructure system, which gives engineers, maintenance workers, and planners greater insights, information, and analysis possibilities to maintain the system and improve efficiencies. Water planners in Germany are testing initial prototypes of AI-augmented water infrastructure to improve efficiency and prevent pipe failure. KaSyTwin, a research project funded by the German government, is working to create digital twins for water networks throughout Germany. Like the US, Germany faces aging water infrastructure with an estimated 15% of their infrastructure over one hundred years old. KaSyTwin, using multi-sensor robotic platforms equipped with laser scanners and cameras, is surveying Germany’s physical sewer systems.vii This data will help build digital twins of German water networks with the goal of semi-automating structural health monitoring of networks in real-time. When combined with AI algorithms, KaSyTwin anticipates that the system will help planners develop proactive maintenance strategies and move maintenance personnel away from overly manual, time-consuming network monitoring.viii Historically, such monitoring involved operators watching a video stream while a camera moved throughout the sewer system. KaSyTwin hopes the project will enable water planners to understand current networks while developing plans for network repair, replacement, and expansion.
Just as AI and IoT will assist with maintenance and repair, this same pairing of technology can help water specialists reduce wear and tear on aging infrastructure. Pressure within water networks can contribute to pipe damage over time or accelerate failures in a network. Much like metering for electricity in some urban centers to gain efficiency in electrical grids, this same concept could be applied to water infrastructure through AI and IoT technologies.ix By monitoring usage trends and regional consumption, AI solutions could optimize pipe pressures to reduce impacts on water distribution networks while ensuring sufficient pressure to supply demands.x Consumption rates within cities follow certain patterns and by modulating pressures within networks based on these patterns, AI innovation could potentially lengthen pipe infrastructure lifespan by reducing constant, needless pressure within portions of the distribution network. Such an approach will limit over pressurization on leaks and thereby reduce losses as maintenance repairs are made. This approach will also gain efficiency at a network level. In the instance of NYC, business districts experience higher demand during weekday business hours with significantly lower demand during weekends. Boroughs known for large commuting populations display trends opposite of business districts. In this case, AI control systems could regulate and optimize network pressures to support these differing patterns while reducing strains on the water infrastructure. Any approaches extending current water infrastructure lifespans will provide additional time for city planners to develop alternative solutions to meet increasing demand.
Efficient Reuse & AI Analytics
Not all wastewater is equal. While AI augmentation in front of user demand could make water management more efficient, the same can be said on the backend. Each gallon of wastewater repurposed, is one less gallon placed into front end demand and thereby reducing load on water distribution networks. As currently designed, all user demand is the same and pulls from the same water source networks. But all demanded water is not used the same. Currently, water distribution networks send potable, highly treated water to supply all demands such as toilets. This system of water management is called “single-quality drinking water” and regardless of whether users are drinking from the tap, bathing, washing a vehicle, or watering the lawn, water distribution networks supply the same highly processed water even if not required.xi This makes the feeder network highly inefficient as all demand water must meet health standards regardless of the use.
However, repurposing many wastewater sources could supply lower-level demand. Most household wastewater falls into three distinct types which current waste networks dispose of in the same way. Shower and bathroom sinks generate light greywater, washing machines and kitchen taps produce heavy greywater, and toilets produce blackwater.xii To support reuse initiatives, wastewater from bathroom sinks or bathing could be minimally treated and applied back to the network. This wastewater could supply demands such as watering lawns or industrial uses since these water demands do not require potable standards. This concept, known as potable reuse or resource recovery, can be accomplished through lower-level on-site treatment mechanisms which are more efficient than blackwater treatment processes.xiii Such an approach will require a massive shift in water dissemination and recovery networks, but AI sensor and switching built into future networks will increase efficiency and prevent cross-source contamination between networks. Integrated AI sensors will monitor wastewater contaminants and ensure optimal reuse and reapplication in water distribution networks. Through similar technology-augmented approaches, distribution networks could identify and process light greywater, heavy greywater, and blackwater separately and more efficiently.
Similarly, this same AI and sensor technology applies to sewer or rain runoff systems and can lessen the burden on feeder and distribution networks. Current urban designs emphasize rain capture to prevent urban inundation from normal rainfall. With significant pavement and roofing surfaces, most rain flows directly into storm drain and sewer systems without contributing or feeding into distribution networks. While rain runoff does capture contaminants in this process, for many urban areas, rain runoff surprisingly also carries significant quantities of nutrients, organic matter, and heat.xiv By integrating AI and sensor technologies into rain runoff and wastewater systems, distribution networks can monitor rainfall rates across diverse geographic areas and recapture high volumes of water to augment supply-side water distribution.xv Such techniques will make water provisioning more efficient by leveraging natural rainfall to augment current networks and reduce demands on single-quality drinking water supply networks.
AI, Intelligence, and Water Infrastructure Protection
Implementing AI and a network of integrated sensors will not only improve efficiency, but also help security personnel respond to water infrastructure threats. Networked sensors will help water professionals identify anomalies in pipe pressure, backflow, or rapid changes in water purity. Some changes could indicate natural phenomenon such as high rainfall or an accidental breakage of a pipe. However, by implementing AI and a federated monitoring system, infrastructure managers could also identify nefarious actions against water networks. Comprehensive AI monitoring will track system pressures across large geographic areas. An explosive attack against a pipeline portion will create pressure changes that look very different from either an accidental collapse or unanticipated leak. Likewise, if multiple attacks against water infrastructure take place simultaneously, AI monitoring could quickly detect the attacks, provide planners context to the events, and automatically redistribute or isolate portions of water infrastructure to best protect the entire network.
Just as AI monitoring could help identify and address an explosive attack, such monitoring could also identify purposeful contamination. If a malicious actor attempted to poison or contaminate water supplies with a toxin or biological weapon, integrated sensors could rapidly identify changes in water quality. Using AI, the contamination could be tracked, isolated, and enable critical responders to quickly locate and neutralize the contamination. Automated valve or distributions systems could close or reroute portions of the network to protect consumers and mitigate exposure. Similar processes could help authorities trace the contamination back to the source and potentially even identify the contamination mechanism or time.
Security and intelligence are at the heart of AI and IoT-augmented water infrastructure. Such an approach would protect critical water networks from attack and provide vital information to understand the type, complexity, and potential impacts of any threats to the infrastructure. By quickly understanding a threat, authorities can share information in near real-time to provide warnings to other water network managers or provide recommendations for preventative actions. The intelligence community (IC), acting as a force multiplier, could provide threat information to help AI systems identify and mitigate biological agents. IC technical data indicators from previous terrorist attacks may inform AI algorithms to anticipate vulnerable or susceptible nodes in water infrastructure. If nodes cannot be improved, infrastructure security specialists could prioritize AI monitoring of critical nodes to improve response time should an attack take place. Advances in AI, supported by IC collaboration, will help security personnel anticipate or address threats to water infrastructure in ways current networks fall short.
Conclusion
While aging water infrastructure across the US poses a near-term risk to constantly growing urban centers, AI and advancing technologies could help water infrastructure specialist improve efficiency and security. AI integration into current water distribution networks promises to help city planners and water specialist address pipe network issues before they arise and work ahead of calamity. Such preemptive initiatives will prevent water distribution outages before they occur and avoid water loss from breakage or leaks. Addressing current operations, AI integration offers city planners ways to optimize water distribution systems to reduce stress on aging infrastructure. Pressure modulation and network flow rate adjustments enabled by AI solutions will increase the longevity of current water infrastructure and allow more time for much needed upgrades and repairs. To make current and future water distribution networks more efficient and address increasing consumer demand, AI and sensor technologies may hold the keys necessary to maximize reuse and recovery techniques. Rainfall runoff recapture, enabled by AI monitoring, could augment water supplies and limit network strains associated with single-quality drinking water paradigms. AI, while unlikely to address all water infrastructure issues, promises to more efficiently use, maintain, and sustain critical water infrastructure in the US and allow time for water specialists to address systemic issues.
The author is responsible for the content of this article. The views expressed do not reflect the official policy or position of the National Intelligence University, the Office of the Director of National Intelligence, the U.S. Intelligence Community, the Department of Defense, or the U.S. Government.
Endnotes
i Center for Sustainable Systems. “U.S. Water Supply and Distribution Factsheet.” University of Michigan. 2024. Pub. No. CSS05-17 https://css.umich.edu/publications/factsheets/water/us-water-supply-and-distribution-factsheet
ii Center for Sustainable Systems. “U.S. Water Supply and Distribution Factsheet.” University of Michigan. 2024. Pub. No. CSS05-17 https://css.umich.edu/publications/factsheets/water/us-water-supply-and-distribution-factsheet
iii Neal Andrew Barton, Stephen Henry Hallett, Simon Richard Jude, Trung Hieu Tran, “Predicting the risk of pipe failure using gradient boosted decision trees and weighted risk analysis,” NPJ Clean Water (2022), 26 July 2021, https://doi.org/10.1038/s41545-022-00165-2, 1
iv Neal Andrew Barton, Stephen Henry Hallett, Simon Richard Jude, Trung Hieu Tran, “Predicting the risk of pipe failure using gradient boosted decision trees and weighted risk analysis,” NPJ Clean Water (2022), 26 July 2021, https://doi.org/10.1038/s41545-022-00165-2, 4
v Neal Andrew Barton, Stephen Henry Hallett, Simon Richard Jude, Trung Hieu Tran, “Predicting the risk of pipe failure using gradient boosted decision trees and weighted risk analysis,” NPJ Clean Water (2022), 26 July 2021, https://doi.org/10.1038/s41545-022-00165-2, 2
vi Matthew Moy de Vitry, “Smart Urban Water Systems: What Could Possibly Go Wrong?” Environmental Research Letters #14, IOP Publishing (2019), https://doi.org/10.1088/1748-9326/ab3761, 2
vii Hartmann, Sabine, Raquel Valles, Annette Schmitt, Thamer Al-Zuriqat, Kosmas Dragos, Peter Gölzhäuser, Jan Thomas Jung, Georg Villinger, Varela Rojas Diana, Matthias Bergmann, Torben Pullmann, Dirk Heimer, Christoph Stahl, Axel Stollewerk, Michael Hilgers, Eva Jansen, Brigitte Schoenebeck, Oliver Buchholz, Ioannis Papadakis, Dominik Robert Merkle, Jan-Iwo Jäkel, Sven Mackenbach, Katharina Klemt-Albert, Alexander Reiterer, and Kay Smarsly. “Digital-Twin-Based Management of Sewer Systems: Research Strategy for the KaSyTwin Project.” Water 17, no. 3 (2025): 299, https://niu.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/digital-twin-based-management-sewer-systems/docview/3165914935/se-2, 3. viii Hartmann, Sabine, Raquel Valles, Annette Schmitt, Thamer Al-Zuriqat, Kosmas Dragos, Peter Gölzhäuser, Jan Thomas Jung, Georg Villinger, Varela Rojas Diana, Matthias Bergmann, Torben Pullmann, Dirk Heimer, Christoph Stahl, Axel Stollewerk, Michael Hilgers, Eva Jansen, Brigitte Schoenebeck, Oliver Buchholz, Ioannis Papadakis, Dominik Robert Merkle, Jan-Iwo Jäkel, Sven Mackenbach, Katharina Klemt-Albert, Alexander Reiterer, and Kay Smarsly. “Digital-Twin-Based Management of Sewer Systems: Research Strategy for the KaSyTwin Project.” Water 17, no. 3 (2025): 299, https://niu.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/digital-twin-based-management-sewer-systems/docview/3165914935/se-2, 7.
ix Matthew Moy de Vitry, “Smart Urban Water Systems: What Could Possibly Go Wrong?” Environmental Research Letters #14, IOP Publishing (2019), https://doi.org/10.1088/1748-9326/ab3761, 3
x Matthew Moy de Vitry, “Smart Urban Water Systems: What Could Possibly Go Wrong?” Environmental Research Letters #14, IOP Publishing (2019), https://doi.org/10.1088/1748-9326/ab3761, 2
xi Larsen, Tove A., Sabine Hoffmann, Christoph Lüthi, Bernhard Truffer, and Max Maurer. “Emerging Solutions to the Water Challenges of an Urbanizing World.” Science 352, no. 6288 (May 20, 2016): 928-33, https://niu.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/emerging-solutions-water-challenges-urbanizing/docview/1790069786/se-2, 930
xii Larsen, Tove A., Sabine Hoffmann, Christoph Lüthi, Bernhard Truffer, and Max Maurer. “Emerging Solutions to the Water Challenges of an Urbanizing World.” Science 352, no. 6288 (May 20, 2016): 928-33, https://niu.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/emerging-solutions-water-challenges-urbanizing/docview/1790069786/se-2, 931
xiii Matthew Moy de Vitry, “Smart Urban Water Systems: What Could Possibly Go Wrong?” Environmental Research Letters #14, IOP Publishing (2019), https://doi.org/10.1088/1748-9326/ab3761, 3
xiv Larsen, Tove A., Sabine Hoffmann, Christoph Lüthi, Bernhard Truffer, and Max Maurer. “Emerging Solutions to the Water Challenges of an Urbanizing World.” Science 352, no. 6288 (May 20, 2016): 928-33, https://niu.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/emerging-solutions-water-challenges-urbanizing/docview/1790069786/se-2, 929
xv Prajna Mahardhika Sakti and Putriani Okkie. “Deployment and use of Artificial Intelligence (AI) in Water Resources and Water Management.” IOP Conference Series. Earth and Environmental Science 1195, no. 1 (06, 2023): 012056, https://niu.idm.oclc.org/login?url=https://www.proquest.com/scholarly-journals/deployment-use-artificial-intelligence-ai-water/docview/2826813905/se-2.
