Can AI Stop Cities from Running Out of Water?

Cape Town’s “Day Zero” crisis reveals the perils of water scarcity for cities.

Before 2018, it was widely assumed that modern, technologically capable cities would not run out of water. That assumption changed when Cape Town came within weeks of “Day Zero”—the moment the city expected to shut off its municipal water system, limiting millions of residents to just 25 liters per person per day collected from distribution points.

Since then, city after city has faced similar pressures: Chennai, Nelson Mandela Bay, Montevideo, large parts of Mexico City. Even Dublin, a famously rainy city, is confronting water constraints in part due to rising demand from data centers.

Practical solutions exist—and many can be enabled by AI.

The encouraging news is that AI and machine learning can help manage water resources better. With better management, there is less water scarcity.

Below are three practical examples.

Mapping Boreholes in Lagos

Lagos, a city of about 20 million people, has piped water coverage for less than 20% of its population. Many residents rely on groundwater pumped from boreholes. Yet little is known about where those boreholes are located.

Without measurement, management is impossible.

Using Google Street View and machine learning, an algorithm can identify community boreholes automatically. This approach maps boreholes quickly, accurately, and at extremely low cost, orders of magnitude faster than manual surveys. With better visibility into groundwater extraction, aquifers can be protected from contamination and overuse.

Optimizing Irrigation Systems

Irrigation systems are vital for growing food, yet they are often highly inefficient. Many over-irrigate, wasting water; others under-irrigate, damaging crops. In many cases, operators lack reliable performance data.

Satellite remote sensing now provides temperature data, evapotranspiration measures, and soil moisture indicators. When processed through machine learning models, this information can determine whether irrigation systems are operating efficiently.

With that insight, incentives can be created to reward effective management. More efficient irrigation means less water required to grow food and more water available overall.

Detecting Leaks in Urban Networks

In many cities, up to 50% of treated water is lost before it reaches customers, largely due to leakage. Most leaks are invisible.

Where utilities deploy pressure sensors throughout their networks, machine learning can analyze pressure patterns to detect exactly when and where a leak occurs—sometimes almost before it becomes visible. Rapid detection allows rapid repair, reducing water loss dramatically.

These are only a few examples. There are many more opportunities to use AI to better manage water resources and reduce scarcity.

What will it take to scale these solutions?

Two things: talent and data.

While machine learning expertise is expanding rapidly, relatively few professionals working in water utilities have deep AI or data science training. At the same time, many AI specialists are drawn to sectors such as finance or technology.

The more difficult challenge is data. The water sector is notoriously weak in collecting reliable, consistent data. In two of the three examples above, the data did not even originate from water utilities; it came from satellite systems and commercial mapping platforms. There is a clear call for governments and water utilities to invest in collecting reliable data about what is happening in water networks.

With better data and the right skills, AI can help optimize water systems.