Wellington floods and tropical cyclones: How weather forecasting is dealing with an uncertain future

Source: Radio New Zealand

Explainer – Has climate change made weather forecasting harder to do – and how is technology changing the game?

There was plenty of warning about the dangers of Tropical Cyclone Vaianu earlier this month, but the fast-moving thunderstorms that hit Wellington this week came with less notice.

After the fact, there was a certain amount of complaints on social media about weather forecasting – one thread was titled, “Heaps of warnings for the small cyclone, no warnings for worse rain drop in decades, go figure.”

But armchair comments online ignore the complexity of predicting weather, experts say.

“Weather is not static,” Earth Sciences New Zealand (ESNZ) principal scientist Chris Brandolino said.

“Some warnings you can give days in advance and some you can’t,” MetService senior meteorologist Chris Noble said.

There have been five red weather warnings – the highest there is – so far in 2026, and 46 severe thunderstorm warnings.

Weather forecasting is, much like the weather itself, constantly changing, and new tools are being brought in all the time to increase accuracy. And the precise kind of weather that’s coming in matters a lot when it comes to forecasts.

The recent merger between MetService and Earth Sciences New Zealand (formerly NIWA) is allowing the formerly competitive organisations to work together to improve forecasting abilities – and new technologies including artificial intelligence is being brought to the table.

Why was there such a difference with the warnings for Wellington’s flooding and those for Cyclone Vaianu?

Size matters. The tropical cyclone was an immense system flagged well in advance when it formed earlier this month, and red and orange warnings were issued across most of the country before Vaianu struck. In the end, the cyclone moved slightly east and didn’t cause damage on the scale of Cyclone Gabrielle in 2023.

On the other hand, the Wellington flooding and other thunderstorms that hit most of the country over the past week came from small, fast moving storms passing through, some in the middle of the night – exacerbated by the effects of climate change.

New Zealand is also full of microclimates, particularly in hilly areas with widely varying topography, and it’s a small island nation in the middle of a very large ocean full of varying currents and conditions that wash over us.

“There’s always been challenges in forecasting for small scale events like we saw in Wellington,” Noble said.

“There’s a big difference in how you forecast something big that you can see and plays out over days versus something like Wellington, where it’s driven by a thunderstorm that develops and blows up really, really quickly, you know, in the space of under an hour.”

Earlier this week MetService meteorologist Devlin Linden told RNZ the Wellington event was highly unusual.

”The thing that makes it so impressive … is how quick the rainfall occurred,” he said.

One monitoring station recorded an extreme burst, with 77 millimetres of rain falling in just one hour.

“We consider a downpour to be anywhere between 15 to 25 millimetres,” he said.

“Severe weather events (like thunderstorms) and their associated hazards (like landslides or flash flooding) can move very quickly and unpredictably,” said John Price, director of Civil Defence Emergency Management.

Which means that while forecasts matter, your eyes matter too, he said.

“This is why people in New Zealand need to trust their ‘danger sense’ – if they encounter rising floodwaters, indications of a potential landslide or a long or strong earthquake in a coastal area, they should not wait for official advice or warnings, and should act immediately.”

Noble used what he calls a “popcorn analogy” to break down how tricky it can be to predict sudden bursts. Thunderstorms need ingredients like moisture, unstable environments and a trigger to kick off.

“Now, if we think about popcorn, making popcorn is kind of the same. You need a heat source. You’ve got to have your ingredients. You need your corn kernels. Put it in a pan, turn the heat on. Can you tell me which kernel’s going to pop first?

“And thunderstorm forecasting is a little bit like that. We can identify where the risk is, where’s the pan. I can say, oh, well, tomorrow, the Wellington region, there’s an elevated risk of thunderstorms somewhere in the Wellington region.

“But until those storms start forming, until that kernel starts to wiggle a little bit, it starts to pop, I can’t point to exactly where the storm will form. But I have been able to tell you something about the risk leading into the event. And so that’s the real challenge.”

How does weather forecasting work today, anyway?

Brandolino described the four pillars of weather predictions as observation, data management, modelling and forecasting – and then boiling that all down and passing it on the public in the clearest way possible.

“It’s connecting all that up to the communications, to the warnings, to the impact.”

MetService has a series of weather watches and warnings ranging from yellow watches to red warnings. A team of senior meteorologists makes the call on when to issue alerts, and the agency frequently consults with councils, Civil Defence and other agencies.

ESNZ uses multiple systems that range from short-term to seasonal forecasts. Data points can run as small as 200 metres apart or scale up to a global level.

“We are getting increasingly large amounts of information coming in,” Noble said. “The number of observations I’ve watched over my career have gone up and up and up.”

Radar networks in New Zealand scan the skies, and “every 7 1/2 minutes, we get a new complete three-dimensional picture of what’s happening”.

Computer modelling is a key tool, helped by immense supercomputers that can sift through all that data far faster.

How is technology like artificial intelligence playing a part?

Because there is so much data, AI can process results far faster than humans alone.

ESNZ launched its new multimillion dollar supercomputer Cascade last year, which is capable of computing speeds of 2.4 petaflops – a petaflop is one quadrillion calculations per second.

Brandolino uses terms like “neural weather modelling” and “nowcasting“, all of which have the ultimate goal of using new technology to get accurate weather information out faster. Nowcasting can even generate short-term predictions in real time.

With climate change making speedy forecasts more important than ever, the models have to keep up.

“So the benefit with these AI-based models is that they can update quite quickly,” he said.

“So physics-based models, the kind of gold standard of what has been happening over the past … is that every six hours, you get a new model update, you get a new sort of version of what the future is predicted to be from the eyes of the model. But with AI modelling, that can update much more quickly.”

Another method is what’s known as the New Zealand Ensemble System, which uses probability forecasting on a group of 18 different model members run five days into the future.

If you’re a fan of Marvel superhero movies, think of it as “the multiverse of weather,” showing all the ways a system could evolve.

“Just because you have a low chance of something happening doesn’t mean it’s a zero chance,” Brandolino said. “But it’s how do you know when to pick that outcome? How do you know when to rely on that outcome? And these are things we’re hoping to advance.”

“In those scenarios where we’ve got really good agreement across a range of different models or a range of different scenarios, the forecaster can have a higher confidence in that outcome,” Noble said.

This kind of forecasting came in key when predicting possible paths for Cyclone Vaianu.

“When you’re communicating to say, stakeholders or NEMA or civil defence or the public or media or whatever, say, (we say) look, some uncertainty, but we think here’s two or three outcomes that we think are most likely, A, B, and C. And that can help sort of storytell and sort of drive home a narrative,” Brandolino said.

What does the future of forecasting hold?

The merger between MetService and ESNZ will ultimately improve forecasting, Brandolino said, although he noted “it’s a multi-year journey”.

“The forecast value chain is a term that’s often used in our world.”

That means ensuring weather predictions communicate information in a way that’s “actionable”.

One of big talking points in weather forecasting is pivoting to more of an “impact based warning system”, which looks closer at the ground-level impacts wet weather can have.

“One hundred millimetres of rain that falls in central Wellington is much different, irrespective of the time, than if it falls over an open field,” Brandolino said.

He recalled a saying by former NIWA chief scientist Murray Poulter – “his famous line was, ‘Rain never did anything till it hit the ground’”.

An impact-based approach to weather warnings would crunch together more data on the particular threats locations face.

“We want to link up that really good flood advice and flood information and knowledge with the warning system so that when a rainfall warning goes out, it’s pitched at the right level because of what that rain is going to do, the impact of that rain on the ground,” Noble said.

“So in the future, an impact-based warning system will actually start to convey more about what the weather is going to do, not just what the weather is going to be. But what’s it going to do? How dangerous is that weather? Will it delay my commute because of surface flooding on the highway? Will it create river flooding or surface flooding elsewhere? Will it create landslides or is that landslide risk higher or lower?”

What does the public need to understand about weather forecasts today?

Pay attention and be prepared are key messages. Don’t just passively wait for a weather forecast on the 6pm news like the old days.

“You have to stay on top of it,” Brandolino said. “I don’t think you can check the weather and be like, oh, then check again in three days. I think the public would do themselves a favour by staying on top of the weather from reputable weather sources.”

While warnings and pre-emptive states of emergency may sometimes be seen as excessive – or apparently even “woke” according to Wairoa District Mayor Craig Little – they are still critical in making sure people are prepared, even if it turns out things aren’t as bad as forecast.

“Alerts are not issued lightly and the reality is, severe weather events are increasing in frequency and severity,” NEMA’s Price said.

“When emergencies happen, we all need to do our part so that emergency services can focus on helping the people who need it most. That includes being ready for emergencies even if they turn out not to have a large impact on you or your community – next time, it could be different.”

Noble said it’s also critical to know your own risks – that means know your home, its surrounding environment, and what could happen.

“Me personally, I live up on one of these hills in Wellington, so I know I’m above the tsunami safe zone. I know that I’m not particularly prone to flooding, but being on a hill, I do have that landslide awareness in my head.”

Weather warnings are important but can’t always take in a system as chaotic as the Wellington floods.

“Sometimes those warnings can give good lead time, like Vaianu,” Noble said.

“Sometimes they can’t.”

Sign up for Ngā Pitopito Kōrero, a daily newsletter curated by our editors and delivered straight to your inbox every weekday.

– Published by EveningReport.nz and AsiaPacificReport.nz, see: MIL OSI in partnership with Radio New Zealand