Drought resilience starts underground, AI is learning to look

By Gokul Krishna Govindu, Co-founder, SmartTerra

When Indian cities discuss drought resilience, the conversation gravitates toward dams, desalination, and inter-basin transfers. But there is another source of water that rarely features in policy debates, largely because it is neither glamorous nor new. It is the water that cities already produce, treat, and then silently lose underground.

This is not uniquely an Indian problem. Water utilities in England and Wales lose roughly three billion litres a day from their distribution networks, according to Ofwat. Globally, the International Water Association puts the cost of non-revenue water at $39 billion a year, or approximately ₹3.3 lakh crore. These are systemic losses, and they are everywhere. But India’s water arithmetic makes the problem particularly unforgiving.

A country that cannot afford to waste what it has
India supports roughly 18% of the world’s population with about 4% of its freshwater. According to WRI’s Aqueduct Water Risk Atlas, the country ranks among 25 nations facing “extremely high” water stress, using over 80% of its available supply. NITI Aayog’s Composite Water Management Index projects that by 2030, demand will be twice the available supply, with an eventual 6% loss in GDP if the trajectory holds. And yet, according to the CPHEEO, non-revenue water (NRW) in Indian cities averages 40–50%. World Bank estimates, cited in a 2024 CSIS analysis, put the annual cost to Indian utilities at roughly $5 billion, or about ₹42,000 crore.

Or to put it another way, Indian cities collectively spend billions treating water that disappears before anyone drinks it. If this were a line item in any company’s budget, someone would have been asked to explain it by now.

The causes are familiar: ageing pipelines, undocumented networks, faulty meters, illegal connections, and intermittent supply that makes measurement, never mind accurate measurement, difficult at best.

From reactive repair to something more useful
Traditionally, leak detection has been a maintenance activity: send a crew, listen for anomalies, fix what is visibly broken. This worked tolerably well when distribution networks were smaller. It is less convincing when you are responsible for thousands of kilometres of pipeline, much of it decades old, some of it unrecorded on any map.

What has changed is the emergence of digital tools that can overlay analytical intelligence on existing infrastructure. Network digital twins can model water flows across an entire distribution system using data that utilities already collect but rarely integrate: SCADA readings, meter data, GIS maps, maintenance logs, even consumer complaints. Machine learning can then flag where losses are most likely occurring, whether from physical leaks, faulty meters, or unauthorised connections.

Crucially, some of these platforms are designed for conditions that are distinctly Indian: intermittent supply, incomplete data, networks without formal district metering. Many global water-tech solutions assume 24/7 pressurised systems and continuous sensor coverage, conditions that describe relatively few Indian cities today. The tools gaining traction here are the ones built to handle the mess rather than wish it away.

For utilities and urban water operators, the shift is significant. Instead of waiting for a pipe to burst spectacularly enough to warrant attention, it becomes possible to prioritise segments most likely to fail. Capital expenditure moves from emergency response to strategic renewal. And in some cases, even a few weeks of structured data can surface anomalies that years of manual monitoring missed.

The 24/7 transition will test everything
India’s push toward continuous pressurised supply makes this doubly urgent. AMRUT 2.0 has set a target of reducing NRW to below 20%. The Jal Jeevan Mission (Urban), with an outlay of ₹2.87 lakh crore, is working toward universal tap water coverage across 4,378 statutory towns. These are laudable programmes. But pressurising a network that already leaks at 40–50% without first understanding where and why it leaks is a bit like inflating a tyre without checking for punctures. The cities that treat NRW as a data and analytics challenge, not merely a plumbing one, will be the ones that manage this transition without simply accelerating their losses.

Climate resilience, measured in litres
There is something appealing about the arithmetic of NRW reduction. Every litre recovered is a litre that has already been sourced, treated, and pumped. It does not require new environmental clearances, new land acquisition, or new raw water sources. It is, in the driest sense of the phrase, low-hanging fruit.

For drought-prone regions, this makes NRW reduction one of the most immediate and measurable forms of climate adaptation available. It lowers energy consumption, reduces operating costs, and strengthens supply reliability. And unlike large infrastructure projects that take years to commission, digital approaches layered on existing networks can begin delivering results in months, not decades.
India’s urban water pipes may not feature in speeches about climate resilience. They probably should.

The infrastructure is already in the ground. The water is already being produced. What is missing, in most cities, is the intelligence to know where it goes after that. Fixing that gap may not be glamorous, but it is increasingly difficult to ignore.