Six papers from IWA Glasgow share one lesson on resilience: the utilities that coped best had tested their tools before the crisis, not during it.
Resilience Is Rehearsed: Why Utilities Should Use the Unfamiliar Before the Crisis Arrives
Six papers, four types of extreme event, one shared lesson from Glasgow: the utilities that coped best had already tested what they relied on.
By Dr Jo Burgess
What a session on extreme events taught me about ordinary decisions
In October I had the privilege of chairing a technical session at the IWA World Water Congress & Exhibition in Glasgow, “Resilient Utilities: Collaboration and Innovation for Managing Extreme Events”. The session speakers took us from a nationwide blackout in Iberia to record rainfall in Valencia, from post-fire water quality in Los Angeles to AI flood forecasting in Northumberland, from a continuity management system in Porto to sensor placement in Seoul. Different hazards, different continents, different budgets.
Yet they all said the same thing underneath. None of the successful responses was improvised. Each rested on something that had been mapped, tried and agreed with the people who would have to use it, before the crisis occurred. That is the same conclusion I reached in my October blog: the water sector does not have an innovation problem, it has an implementation problem. Resilience is where that problem bites hardest, because you cannot try out a new tool in the middle of a crisis. By then you will use whatever you already trust, and nothing else.
Power, water and communications: one system, not three
The Iberian blackout of 28 April 2025 was, in the words of EPAL, Portugal’s primary water supply utility, a real-time stress test for water utilities (Cristóvão, 2026). Power and telecommunications failed together, and EPAL lost real-time visibility of reservoir levels, flows, pressures and treatment parameters. Teams had to move to local manual control of assets that normally run under automated supervision, while the same loss of communications undermined the coordination of field crews and crisis decision-making. Porto’s utility, Águas e Energia do Porto, faced the same blackout, along with severe downtown flooding and a cyberattack in recent years, and their paper describes how a digital activation tool guided decisions while its 24/7 Operations Control Centre kept monitoring reservoirs, pumping stations and drainage systems (Araújo et al., 2026).
Then look at Valencia. When more than 700 mm of rain fell in 24 hours in October 2024, many sensors were put out of action. Global Omnium’s hydraulic digital twin, fed by the sensors that survived, estimated flow, pressure and water quality at the points where instruments had failed, and the network performed as the twin had predicted (Ponz Carcelen and Macián Cervera, 2026).
Put these together and the lesson for the nexus is clear. Water is not an independent system with some dependencies attached. It is just one layer of a stack in which electricity and communications sit underneath, and EPAL’s paper is explicit that utilities need to map these interdependencies, find the weak points and design redundancy before the cascade begins (Cristóvão, 2026). I would define self-reliance in that context not as isolation, but as the ability to keep delivering a safe, if limited, service when the neighbouring systems you depend on have gone dark.
That self-reliance has to be designed, and it has to be tested.
What the session showed actually worked
They knew their risks in advance. EPAL argues for structured risk mapping translated into crisis scenarios, each with a roadmap of triggers, thresholds, actions and recovery milestones (Cristóvão, 2026). Porto’s continuity system rests on business impact analysis that identifies critical processes and acceptable downtimes (Araújo et al., 2026). In Seoul, researchers built twelve design rainfall scenarios, combining four return periods and three storm durations, to decide where a limited number of water-level sensors would do the most good (Kim et al., 2026). Each starts with the question “what could go wrong, and what would we need then?” rather than “what product is available?”.
They tested the tool before they needed it. Global Omnium’s digital twin had already been built and validated, which is why it could be trusted in emergency conditions (Ponz Carcelen and Macián Cervera, 2026). In Los Angeles, the authors attribute the low levels of volatile organic compounds found in 53 sampled homes after the 2025 fires, with a single benzene detection above California’s 1 ppb limit, to lessons learned from earlier fires that led local water agencies to adopt proactive testing and flushing policies (Olivares et al., 2026). That is a lesson that made it all the way from experience into standard practice, which is exactly the journey most innovations fail to make.
They built trust with the people who would act. The FloodAI project, part of Defra’s Flood and Coastal Resilience Innovation Programme in the UK, is working with six pilot communities in Northumberland and testing its prototype live with flood wardens, local authorities and Environment Agency representatives to build trust and gather feedback (Corrigan et al., 2026). Porto backs its continuity plans with training, exercises and simulations so that teams know their roles before they need them (Araújo et al., 2026). A tool that operators have never seen before is not a resilience measure, no matter how superb its technical merits.
They made information usable by others. In Valencia, laboratory results were uploaded to a central platform that triggered a traffic-light system identifying “Do Not Drink” areas, allowing authorities to issue public notices within hours, and more than 1,500 extraordinary analyses were processed in the first week (Ponz Carcelen and Macián Cervera, 2026). In Los Angeles, the team produced data briefs, ran consultations with residents and shared surface-water data with agencies for a monitoring dashboard (Olivares et al., 2026). Data that cannot be read, trusted and acted on by the next organisation in the chain is not resilience.
Where trialling fits: the pathway before the trial
In October I described five mechanisms that reduce adoption uncertainty: demand, validation, procurement, financing and evidence. Resilience supplies the first one in abundance. Few problems are more acute than a service that has just failed. What it does not supply is the other four, and that is where I see utilities and technology developers repeatedly miss the window.
After a major event there is a rush to buy whatever is available, and then the urgency fades and the budget goes with it. Before a crisis, the new and unfamiliar looks like risk, so it waits for a pilot or trial that may never be connected to a purchase. Both leave utilities less resilient than they could be.
In their FloodAI paper Corrigan et al. show us what is at stake in practice. Their approach uses forecast rainfall as an input as well as observed rainfall, which adds uncertainty but, according to the authors, gives the lead time that matters, since a warning around eight hours ahead is when people start to act on their flood plans (Corrigan et al., 2026). It also uses explicable AI, so that users can see how results are derived, which builds trust. Whether wardens and responders will act on a forecast that carries uncertainty is not something a model can settle. It can only be settled in a trial, with the people who will make the call. The prototype is designed for scalability and the team has ambitions for national adoption (Corrigan et al., 2026). Whether it gets there will depend on the same questions I raised in October: who buys it, on what terms and from which budget?
The Seoul paper offers a different, and equally useful, lesson. With 454 water-level sensors already installed across the city’s sewer networks, the team’s scenario-based method narrowed candidate sites to just four optimal locations, and the warning thresholds derived from them reproduced historical flood occurrences. Kim et al. (2026) position this as a cost-effective way to get more from limited budgets. To me, that is a template for designing a trial: use scenarios to decide where evidence will be most valuable, rather than trialling wherever is convenient.
This is the logic behind the Trial Reservoirs Initiative. Proof-of-value demonstrations begin only when a trial is designed to lead to full-scale implementation if it succeeds, with agreed technical, economic and service performance targets and a described route to implementation. Over 70% of the trials run through the Initiative have led to adoption. That figure is not directly comparable to the 12.5% company-level success rate cited in the World Economic Forum’s recent report (Uplink—World Economic Forum, 2026), but the gap suggests that design matters.
So what next?
First, the digital resilience paradox. Global Omnium shows how much digitalisation can do in a crisis (Ponz Carcelen and Macián Cervera, 2026) and at the same time EPAL shows what happens when the power and communications that digital tools need disappear (Cristóvão, 2026). Both are true. Every solution that is trialled for resilience should therefore be tested in degraded mode, with the power off and communications down, and the trial agreement should say what “good enough” looks like in an emergency. Digital tools that improve everyday efficiency can quietly increase fragility if no one has asked this question.
Second, the evidence is mostly self-reported. These are single-utility, single-crisis, or single-project accounts, presented by the people closest to them, and several are early in their journey. That does not diminish their value, but it does mean we should be cautious about generalising, and we should be asking about long term outcomes.
Third, resilience benefits are avoided losses. They are hard to put in a business case and easy to defer. As with the trial-to-purchase gap I described in October, nobody tends to own the budget that sits between a resilience trial and a resilience investment. Until someone does, the unfamiliar will keep stalling.
What this means for each of us
If you lead or advise a utility:
- Map your interdependencies on power and communications, and decide which services must keep running in skeleton mode (Cristóvão, 2026).
- Treat exercises and simulations as part of procurement, not as an afterthought (Araújo et al., 2026; Cristóvão, 2026).
- Before approving any trial, ask what it is a trial for: who will buy this if it works, on what terms, and from which budget?
- Share what you learn. The Los Angeles data briefs and dashboards show how quickly one organisation’s findings can strengthen others’ (Olivares et al., 2026).
If you develop technology for utilities:
- Understand your buyer’s worst day. These papers describe operators working with no SCADA, no radio, no lights, no help, and no certainty about what is happening in the network.
- Design for bare minimum operation. Ask what your product does when power and connectivity fail, and be honest in your answer.
- Make your outputs explainable, auditable and compatible with existing command and dashboard structures. Operators need to trust a result, and be able to show others how it was reached.
- Plan your pathway with the utility before the trial begins, including who needs training, which budget pays for adoption and what happens next if the targets are met.
Conclusion: self-reliance is built, not declared
The session left me optimistic. The utilities in these papers were not the largest or the best funded. What they shared was preparation: scenarios, exercises, validated tools and relationships with the people they would need to rely on. The next step for the sector is to extend that discipline to new technologies, so that trials stop being isolated demonstrations and become the first step of a deployment that makes utilities more able to stand firm when the systems around them fail.
If you take one thing from this article, make it a question to ask this week: when the lights go out, which of the tools you rely on have you tested without power and/or communications, and who has agreed to buy the ones that work?
Thank you to the six speakers whose work informed this piece, and to the IWA for the opportunity to chair the session.
Jo Burgess is CEO of Tech Ascend and head of Trial Reservoir Global Services at Isle. If you want to know more about the Trial Reservoirs Initiative or discuss a potential trial, visit Iniziativa serbatoio di prova or email her at [email protected].
References
Araújo, J.I., Fonseca, J.B., Oliveira, M.J. and Barros, C.F. (2026). Strategic resilience: Integrated operations and business continuity for Porto’s water and energy services. IWA World Water Congress & Exhibition, Glasgow, UK, 4 – 8 October 2026.
Corrigan, J., Wallace, C., Watson, N. and Doyle, T. (2026). FloodAI for more resilient communities. IWA World Water Congress & Exhibition, Glasgow, UK, 4 – 8 October 2026.
Cristóvão, A. (2026). When the Lights Go Out: A Stress Test for Water Utilities’ Preparedness and Operational Resilience. IWA World Water Congress & Exhibition, Glasgow, UK, 4 – 8 October 2026.
Kim, H., Kim, E., Oh, Y., Park, H., Kim, S., Baek, Y. and Koo, J. (2026). A study on the optimal selection of water-level monitoring locations and estimating thresholds under climate-change-driven extreme rainfall. IWA World Water Congress & Exhibition, Glasgow, UK, 4 – 8 October 2026.
Olivares, C.I., Tilzey, A., Morgan, L., Greenwald Healy, H., Palta, M., Azimi, P. and Allen, J. (2026). Data sharing of water quality impacts after the 2025 fires in Los Angeles. IWA World Water Congress & Exhibition, Glasgow, UK, 4 – 8 October 2026.
Ponz Carcelen, R. and Macián Cervera, J. (2026). Global Omnium’s 2024 High-Altitude Isolated Depression (DANA) weather event management: Digitalisation as key driver for water recovery and operational resilience. IWA World Water Congress & Exhibition, Glasgow, UK, 4 – 8 October 2026.
UpLink—World Economic Forum (2026). Making Water Innovation Work: Enabling Large-Scale Deployment. https://reports.weforum.org/docs/WEF_Making_Water_Innovation_Work_2026.pdf

