George Heard

We could forecast floods better. Why don’t we?

Floods are New Zealand’s most frequent disaster, and one of the most costly. But regions have varying abilities to predict floods depending on local councils’ ability to buy weather data. And though that data is publicly funded, scientists who have created a national flood-forecasting system cannot put it into practice without free access to the same information. Should we prioritise profitability of our research institutes, or public safety?

Written by       Photographed by George Heard

Jo Paterson grew up around disaster. She spent her childhood in Hokitika, on the West Coast: a smattering of weatherboard homes, a regal main street, and tall milk tanks on the corner where the braids of the Hokitika River meet the Tasman. Each year, between two and three metres of rainfall drenches the coast. In the mountains, visible in the distance, annual rainfall can reach 15 metres.

On particularly wet days, or when the Alps have been inundated, the Hokitika River starts to rise. The water turns oyster grey, speckled by white where the accelerating flow churns bubbles to the surface. As it rushes down the riverbed, it crashes into once-gentle curves, eating away at the riverbank and surging over the islands that normally obstruct its flow.

In July 2021, more than 2000 people were evacuated from 826 properties in Westport after the town flooded. Brianna Fox shows New Zealand Herald photojournalist George Heard through her family home, which was badly damaged.

On particularly bad days, new rivers form as the sky unloads, flowing through the township and along its roads. Residents find water lapping at their doorstep, or pushing through their homes. The entire town can flood in a matter of hours. By this time, hopefully, the helicopters and trucks have swooped in, evacuating hundreds of residents to safety, as has happened several times in recent years. But the success of those evacuation efforts are no guarantee: they rely on a handful of experts managing several unwieldy streams of data to predict what will flood, where and when. Until recently, Paterson was one of them.

After training as a geologist, Paterson passed up more lucrative opportunities in order to work in civil defence on the West Coast. “My why is to reduce the impact on the community, inch by inch,” she says. “In a nutshell, it’s all about the community members being evacuated and out of ‘life risk’, full stop.” In the process, however, she has found that predicting and managing floods in New Zealand isn’t just a technical task, but a political one.

There’s no single source of data Paterson can use, no clear predictive tool. The best-quality weather data is expensive, and not all regional councils can afford it.

The result, according to one expert, is a failed experiment that threatens to slow our responses to accelerating storms, floods and catastrophes in an era of deepening climate change.

“Delayed planning means delayed decision-making,” says Paterson. “Then, worst case, they don’t evacuate.”

Floods are particularly difficult disasters to manage. Earthquakes, for example, “are really good warnings in themselves: you feel it, things are damaged, you don’t need to communicate to the public that there’s risk.” By contrast, floods are much more likely to surprise people as they develop. In addition, “the science of flooding across the country is really complicated” and “there’s a lot of moving parts, with MetService, NIWA, river engineers, natural hazard analysts”. As a result, floods “are underestimated for their complexity”.

Given that complexity, clear and accurate flood modelling is critical. “That’s what we’re really lacking: real-time spatial data or data that makes sense to everyone,” says Paterson. “We have really good hydrology networks across the country, but we don’t have real-time or near-time forecasts of flow.”

[chapter-break]

New Zealand used to have one weather forecasting and climate research institution: the Meteorological Service. Run by the government, as is the case in practically every developed country, it had an operational arm that forecast the weather for New Zealanders, and a research arm that studied a climate that was beginning to markedly change. To do both, the Service developed a vast network of weather stations, gathering unmatched quantities of data about the wind, rain and heat that swept across the country.

But by 1992, the Meteorological Service had been split in two. Its operational arm, now MetService, became a state-owned enterprise, while its research arm became the National Institute for Water and Atmospheric Research, or NIWA.

The new MetService had a requirement to make money for the government, so it starting charging for access to its weather data—even where other public agencies were concerned, such as NIWA or the country’s regional councils. Instead, MetService would only provide its data for free to NIWA after a 48-hour delay—an eternity in weather forecasting.

Locked off from much of MetService’s data, NIWA and the regional councils began to collect information of their own and erected their own weather stations across the country. For some councils, like those in Auckland, the mass of ratepayers meant they had a large enough budget to do so. For others, like that which managed the sparsely populated West Coast, it was a far harder task. Suddenly, access to weather data depended in large part on where one lived.

[chapter-break]

As a child, growing up in France, Céline Cattoën-Gilbert dreamed of being an astronaut. She had always found herself attracted to the big questions. What’s in the universe? Where are we from? Space exploration seemed the perfect way to find out. In pursuit of that goal, she went further than most and applied to the European Space Agency. They didn’t accept her, but by that point she was hooked: she trained as an engineer and obtained a masters in applied mathematics in Toulouse.

In search of a different world, she came to New Zealand as an exchange student. Later, at Victoria University of Wellington, she began a PhD in theoretical physics focused on cosmology and black holes. If she couldn’t explore the universe herself, perhaps she could study it from afar.

The New Zealand Water Model is based on inputs such as rainfall and snowfall, allows for evaporation, how water is stored in soil, and how the balance runs off into tributaries either over land or through groundwater.

But where once Stephen Hawking could dream of spacetime, now the field largely involved the management of supercomputers that could chew through the enormous sums of data involved and turn them into something close to an understandable model of black holes’ behaviour. After graduating, Cattoën-Gilbert found that potential employers were more interested in her mastery of those tools than in the distant objects she had used them to study.

Which is how she found herself at NIWA, where the processing of vast quantities of data is practically everything. Her first job was to study whether it was possible to model the behaviour of a small river catchment: could she figure out when it would idle and when it would flood? It took two years, but she developed a model sophisticated enough to do it. The implication broke over her: “Oh, this could run in real time, four times a day, across the whole country.”

These maps show, from left to right: the steepness of New Zealand’s terrain (white being mountaintops and black being sea-level); the amount of annual rainfall, and the major waterways.

The idea was revolutionary. No single institution in New Zealand forecasts floods, just as no single institution coordinates emergency responses to them. Instead, MetService, NIWA and local councils all feed weather, river and rain data to the local civil defence coordinators, who make a call—often on gut intuition—on when to issue flood warnings and when to evacuate.

If Cattoën-Gilbert streamlined the flow of information through a single flood forecast, it could simplify matters immensely for decision-makers on the ground.

In New Zealand’s largest flooding disasters, speed and clarity of information have consistently been the two most important factors determining whether an emergency response succeeds or fails.

It took Cattoën-Gilbert and her team another two years to develop a framework for a national model, a task made difficult by the complexity of New Zealand’s landscape. With towering mountains, deluges of water, and more than 50,000 small rivers to track, she needed all the data she could get to feed into this nascent flood forecasting system.

At that point, she confronted another challenge. “How do we update the system with the latest climate data that we have available,” she asks, “when it’s only available from two days ago, because there’s a 48-hour data embargo?” The MetService-NIWA divide had emerged once again. The breakdown fascinated and infuriated her in equal measure. “It’s not like the data’s not there and we can’t obtain it. It’s a frustrating situation where there is data, but because of constraints, we can’t access it.”

[sidebar-1]

When she compared the situation in New Zealand with other developed economies, including her home in France, the result was distressing.

“New Zealand has the most commercially oriented and restrictive weather observation data model compared to the USA, Norway, Australia, the UK and France,” her study reported, before going on to propose a solution. “A change in policy and funding to lift barriers to near-real-time data-sharing may be required to advance national flood-forecasting developments.”

[chapter-break]

Those barriers, however, have hardened recently due to an ongoing standoff between MetService and NIWA. The former isn’t the only institution facing commercial pressures; NIWA is also expected to bear some of the costs of its research. Partly as a result, it has in recent years begun bidding for the lucrative weather forecasting contracts that were once MetService’s domain.

When NIWA landed one of those contracts, with the Department of Conservation, MetService’s senior leadership sought a meeting with government to complain that NIWA’s “automated forecasts with no intervention from professional meteorologists” could jeopardise “public safety outcomes”, according to material obtained by the New Zealand Herald.

Westport is built on a floodplain, with the Buller River on one side and the Orowaiti Estuary on the other. In July 2021, both waterways flooded, and the town was submerged.

MetService went on to complain that “competition with NIWA for media presence during severe weather events may increase risks to public safety through conflicting narratives”. The relationship between the two has been “tense” ever since.

According to other material obtained by the New Zealand Herald, the relationship grew so challenging that MetService refused to move into a proposed science hub at NIWA’s home at Greta Point in Wellington. This final snub prompted the government to begin an inquiry into New Zealand’s meteorological system, with the goal of assessing what has gone wrong. Not only did that breakdown mean it was highly unlikely Cattoën-Gilbert would get the data she needed, it meant that an entire spectrum of potential relationships and collaborations were cut off.

[chapter-break]

The answer, according to Norm Henry, MetService’s chief of science and innovation, is simple: our current system isn’t fit for purpose.

Henry points to a 2018 report by the New Zealand Institute of Economic Research which found that the benefit of the services MetService provided around weather forecasting and warnings exceeded the cost by almost ten to one.

“So really, if you think of the economic benefit that MetService delivers to New Zealand, it massively dwarfs any return as a state-owned enterprise,” says Henry. In other words, as New Zealand deals with intensifying climate change, MetService’s weather forecasting and warning work is more important than the financial returns it makes to the government.

Yet as MetService and NIWA provide that work, says Henry, the commercial structures of both agencies mean they have to compete rather than collaborate. “At the science level, we’re like brothers and it works really well,” he says. “But we’re expected to deliver and are under some pressure to deliver reasonable returns to our shareholders. And so we’re driven with that commercial imperative. And that creates tension. And we lock horns.

“It probably made good sense back in the late 1980s, but the world is a very different place,” he says.

“In some ways it was a really useful and interesting experiment, in terms of a national meteorological service having that kind of structure, but it doesn’t feel like it’s giving us what we need right now.”

How to change that system is slightly more complex. “The ideal outcome for us,” says Henry—before pausing and rephrasing to, “The ideal outcome for New Zealand”—would be to take all the aspects of government weather forecasting, from rain assessment to flood prediction, and place them within a new agency. As it stands, he says, civil defence responders receive different predictions from different sources, leaving them confused during critical weather events.

After 30 years, Henry feels the cultural differences between MetService and NIWA run too deep to simply unify the two institutions. But a new agency would be able to “protect New Zealand and ensure New Zealand is resilient in the face of these kinds of events.”

[chapter-break]

Jo Paterson now works in civil defence for Southland’s regional council. She doesn’t much care who is in charge of what or why. “To be blunt, it’s not just MetService to NIWA. It’s MetService and NIWA to every other layer of civil defence and response,” says Paterson. “Getting data from both of them, from a local government perspective, can be difficult.”

Paterson remembers one “oh shit” moment where she was hovering in a helicopter above a town, watching the engorged arms of a local river beginning to wrap around the houses. It was a sunny day, so residents weren’t particularly concerned.

Luckily, that time she had good data in hand. Rain was pelting down in the river’s headwaters. That night, the water would reach them, and the risk of flooding skyrocketed.

Without that solid forecast, “it would have been very hard” to convince people to leave, she says. But her access to all those puzzle pieces on that day was as much a matter of luck as anything else.

“I just care about the end result for my community,” she says with a snort. “There shouldn’t be any handbrake on information sharing in my view.”

More by

More by George Heard