🔍 What’s in today’s flow
🤖 New research shows explainable AI techniques like XAI (eg.g. SHAP) are helping utilities build trust in water quality predictions while meeting regulatory requirements.
🌊 The U.S. Bureau of Reclamation awarded a $680,000 contract to deploy HydroForecast AI across five Western river basins for drought prediction.
📊 GPT-5.2 just proved AI can make original scientific discoveries—a capability that could accelerate breakthroughs in water treatment and climate modeling.
🔧 This week's featured tool, FIDO AI, uses machine learning to analyze acoustic data and prioritize leak repairs by severity.
⚡ A Senate briefing revealed 70% of U.S. water utilities failed basic cybersecurity checks, highlighting urgent risks as AI adoption accelerates.
🔬AI research spotlight: AI boosts confidence in water quality predictions
The details
A study in Environmental Modelling & Software examines how AI can be effectively applied to water quality monitoring and resource management across utilities globally.
Key points
Explainable AI (XAI) techniques like SHAP build trust in predictions, help move AI from a “black box” to a transparent decision-support tool
Hybrid models combining AI with physics-based approaches improve accuracy
Data quality remains the primary barrier to effective AI deployment
Why it matters
Utilities need AI they can trust and explain to regulators. This research provides a roadmap for choosing tools that deliver accuracy and transparency.
🤖Latest in AI: GPT-5.2 proves AI can contribute to scientific discovery
On February 13, OpenAI announced GPT-5.2 derived a new result in theoretical physics. The AI identified a loophole in decades-old assumptions about gluon particles, spending 12 hours reasoning before producing a formal mathematical proof verified by physicists from Harvard, Cambridge, and the Institute for Advanced Study.
Why it matters
This demonstrates AI's growing ability to solve complex problems. For water, similar reasoning could accelerate research in contaminant modelling, treatment optimisation, and infrastructure planning.
🔧 Case study: US Bureau of Reclamation deploys AI river forecasting
What happened
The U.S. Bureau of Reclamation awarded a $680,000 contract to Upstream Tech to deploy its HydroForecast AI platform across major Western river basins. The system uses machine learning to generate 10-day probabilistic streamflow forecasts for regions including the California Great Basin, Columbia-Pacific Northwest, Lower Colorado, Missouri Basin, and Upper Colorado.
Why it matters
As climate change disrupts seasonal water flows, federal agencies need better forecasting tools. HydroForecast will directly support reservoir storage decisions, water releases, and hydropower generation. This marks a significant step in government adoption of AI for critical water infrastructure management.
🔧AI tool of the week: FIDO AI
FIDO AI uses machine learning to analyse acoustic data from water networks. It adapts to local conditions and prioritises repairs by severity. United Utilities reports halving leak response time.
Key features
93% leak detection accuracy
Automatic severity ranking
Integrates with SCADA systems
⚖️ Category scores
Ease of use: designed for operations teams, not just data scientists. Minimal workflow disruption, 4/5
Cost: Moderate investment, but ROI improves significantly in high-NRW networks. 3/5
Security: 4/5
Integration: Strong interoperability with existing SCADA and telemetry systems.4/5
Overall
If your utility is focused on reducing non-revenue water through proactive leak management, this category of tool is worth serious evaluation.
🕵️AI’s shadows: Water utilities face rising cyber threats
Issue
A U.S. Senate briefing highlighted urgent cyber threats to water systems. Nation-state actors are increasingly targeting operational technology, and 70% of utilities inspected in 2024 failed basic security standards.
Why it matters
Utilities adopting AI must also defend against AI-powered attacks. New CIRCIA regulations in 2026 require reporting cyber incidents within 72 hours.
Takeaway
Balance AI adoption with robust cybersecurity and incident response planning.
Thanks for reading! I hope you’ve enjoyed this week’s edition and look forward to seeing you next week!
Dr. Andrea G.T