Comment: After July’s heavy rain, Kaikōura was left with damaged roads, bridges, water systems and critical infrastructure.
It was a local emergency, but it tells a national story. When one part of an infrastructure network fails, communities can be cut off, freight delayed, utilities disrupted, and repair resources stretched thin.
After events like this, the familiar call is to spend more. New Zealand needs infrastructure investment. But with the Government running operating deficits and net core Crown debt forecast to rise further, money alone cannot solve every problem. Every dollar has to do more.
The question then becomes: how smartly are we spending it?
That means spending in the right place, at the right time, with the right action: identifying where failure would cause the greatest harm, acting before deterioration becomes an emergency, and choosing between maintenance, an upgrade, new construction or backup capacity.
Repairing failure cannot be the cure-all
Emergency repairs will always be necessary. But waiting for something to break is an expensive way to manage infrastructure.
Preventive maintenance repairs or replaces assets before failure. Predictive maintenance goes further, using information about condition, failures, operating loads and the environment to identify where intervention is most urgent.
Failure is uncertain. We cannot know which pipe will burst, which bridge will become unsafe, or which slope will fail in the next storm. Nor do we have complete information about every asset.
Uncertainty does not make prediction useless. It changes the question. Instead of demanding certainty, decision-makers should ask which failures are more plausible, how confident we are, and what the consequences would be if we were wrong.
A small failure at a critical location may isolate a community or disrupt several services, while an older asset elsewhere may continue operating safely for years. Maintenance should therefore reflect both likelihood and consequences.
Build new, or strengthen what we have?
Te Waihanga estimates that as much as 60 cents in every future infrastructure dollar may need to go towards maintaining and renewing what New Zealand already has. This does not mean we should stop building. Population growth, new technology and exposure to natural hazards will still require new infrastructure.
But new construction, maintenance and upgrades compete for the same limited resources. Sometimes a new bridge, road or water system is essential. Elsewhere, strengthening an existing asset, replacing one vulnerable component or bringing maintenance forward may offer greater protection at lower cost.
Future demand may differ from forecasts. Assets may deteriorate faster than expected. Environmental risks may change. The best option is not simply the one that looks strongest under a single forecast, but the one that still performs reasonably well when conditions differ.
How much backup is enough?
Resilient systems need redundancy: alternative roads, backup power, spare equipment, extra storage or more than one supplier. Too little leaves a system brittle. Too much means paying for capacity that may sit unused for years.
The 2026 Middle East conflict showed how quickly a distant shock can reach New Zealand through fuel markets and supply chains. Larger reserves and more diversified supply routes could offer greater protection, but both carry costs.
The answer cannot be unlimited backup for every possible disruption. We need targeted redundancy where failure would have the widest consequences. Which services must continue during an emergency? How long can disruption be tolerated? Where would backup capacity prevent the greatest harm?
A prediction is not a fact
My Marsden-funded research examines how failures occur and spread across large infrastructure networks, how uncertainty affects their prediction, and how that uncertainty should be carried into maintenance and resource-allocation decisions.
Infrastructure data is incomplete and noisy. Statistical models contain estimation error. Future operating and environmental conditions are unknown. All this uncertainty flows into the final decision.
Yet a predicted risk score is often treated as fact. An asset labelled ‘highest risk’ may be only marginally riskier than several others. A maintenance plan that looks optimal under one forecast may change once prediction error is considered.
The aim is not a model that claims to know the future. It is a decision that remains sensible when forecasts are imperfect.
As the election approaches, political parties will announce how many billions they intend to invest. Those figures matter, but they do not tell us whether the investment will work.
We should also ask: Which locations will be prioritised? Why now? Why is the proposed intervention better than the alternatives? What happens if the forecast is wrong?
New Zealand cannot afford perfect protection from every future shock, nor wait for perfect information before acting. But it can make uncertainty part of the decision and put limited resources where they will matter most: in the right place, at the right time, with the right action.
Dr Xun Xiao is a senior lecturer in statistics at the University of Otago. His Marsden-funded research develops statistical methods for managing uncertainty and supporting maintenance and resource-allocation decisions in complex infrastructure networks.
