Paradise Dam's reservoir glitters sapphire blue under a brilliant central Queensland sky, the monolithic spillway a monument to nature tamed. It's a placid scene — and an illusion.

The concrete wall — all that stands between the reservoir and 75,000 people, farms and livelihoods downstream — is eating itself from the inside.

Commissioned to last 100 years, it has been damned by experts, condemned by authorities and marked for demolition after just 19.

Years of tests indicate it was destined to fail from the first concrete pour — a disaster, literally, in the making.

It's one of Australia's worst infrastructure failures, a time bomb counting down to catastrophe for a city and farming community unaware of the danger until one man's alarming discovery.

Since it went up quickly and cheaply, thanks to modern big-build techniques, the troubles with Paradise Dam have been laid out in a stream of headlines that have embarrassed political leaders, infuriated the community and set off a major inquiry and ballooning remediation costs.

But it's also the story, untold until now, of an unassuming engineer who identified the threat to tens of thousands of people through his ingenuity, a forensic pursuit of answers and an ability to see a worrying future in a smear of concrete dust.

To do that, Brisbane-based global dam specialist Jon Williams and his small team invented a virtual time machine that not only saved an unsuspecting regional population from a potential inland tsunami but changed what engineers across the globe thought they understood about dam construction.

Built for about $200 million, Paradise Dam has already cost Queensland taxpayers more than $165 million in attempts at fixes even before the state government, amid community anger and resistance, decided to knock it down and build a second headwall directly in front of the original. By the time that happens the cost will be at least $1.2 billion more.

How did it come to this? Why was a critical infrastructure project meant to last 100 years defective from the start? Who's accountable? And is it safe?

After Jon Williams turned up to inspect problems exposed by a giant crater in the spillway, he had questions too. Like why could he pluck stones from the dam's face with his fingers?

Paradise found and lost

Water has long been a source of prosperity and heartache in the Wide Bay-Burnett region, one of the nation's most important food bowls, about four hours' drive north of Brisbane.

Renowned for sugar cane fields and grazing properties, it also produces about 75 per cent of Australia's sweet potatoes, 50 per cent of our macadamias, almost all Queensland's mandarins and is the nation's largest source of avocados, passionfruit and chillies.

Just over 100 kilometres upstream from Bundaberg, Paradise Dam stores 170,000 megalitres on the Burnett River (controversially it was 300,000ML — a megalitre is a million litres — when the dam was completed, but we'll get to that).

The river rises in the eucalypts, hoop pines and forest reserves of the Dawes Range north-west of Bundaberg.

For 400km it passes through the traditional lands of the Taribelang Bunda, Wakka Wakka, Wulli Wulli, and Gooreng Gooreng people, almost three million hectares of farmland and Bundaberg city before emptying into the Coral Sea at Burnett Heads.

Hundreds of the original inhabitants were killed in 1849 and 1850 in massacres described by historians as among the bloodiest in Queensland's colonisation.

Many survivors used the river to escape.

In 1889, prospectors discovered gold on the river's southern bank about 100kms south-west of Bundaberg. A town sprang up and the new arrivals called it Paradise.

By 1898 the gold was played out, miners had abandoned the town and, for a century, the land was left to livestock. Downstream, Bundaberg quickly developed the industry for which it would become famous: sugar.

But water was unreliable and the region prone to drought. After decades of simmering tension over water availability, producers turned up the volume on irrigation demands at the 1998 state election.

Two days before voters went to the ballot boxes, then-opposition leader Peter Beattie promised a 300,000ML dam on the Burnett River within five years.

Labor swept to victory and the Beattie government got to work. By 2005 the dam, 38 metres high and 600m long, was complete, expedited by a roller-compacted concrete (RCC) construction method and costing just $200 million.

Behind the wall the river swelled and submerged old mining and cattle townships. A push by the Wakka Wakka people to call the dam Degilbo — "big rock in the river" — was rejected. The government named it Paradise.

On completion of the largest volume dam of its type in the country, a proud Peter Beattie posed for photographers on the wall.

By January this year he was apologising for it.

There was a deep flaw in Paradise that only started to emerge after extreme weather devastated Bundaberg and the farming community and led, eventually, to the dam's terminal diagnosis.

The storm that ate concrete

In late January 2013, ex-tropical cyclone Oswald was moving slowly down the coast when it developed into an intense rain system and stalled west of Rockhampton.

For 48 hours the skies opened, dumping 1,000 millimetres in the Burnett River catchment.

Paradise Dam overflowed, a torrent pouring over the spillway and crashing into the river below before surging downstream, covering farmland and inundating much of Bundaberg.

Dam operator Sunwater estimates 1.5 million megalitres of water a day — about the equivalent of three Sydney Harbours at full tide — spilled over the wall at the flood's peak, churning at the base of the concrete spillway for days.

The 2013 flood was a disaster downstream from the dam.

All photos supplied: Flickr/Rod Savidge

It was almost two months before the overflow eased enough for inspectors to get close to the wall's base, where they found a 15-metre cavity carved from rock and concrete.

That was just the first sign of trouble.

Erosion beneath the dam's apron after the 2013 flood. A commission of inquiry later found the apron was not wide enough to prevent severe scouring and was at risk of collapsing in a major overflow.

Supplied: GHD

Enter global engineering firm GHD's dam design manager James Willey and principal engineer Jon Williams, whose masterful knowledge of concrete was about to be put to its toughest test.

For more than 30 years, Brisbane-based Williams has been GHD's go-to guy on its biggest concrete dam projects, helping build them in Malaysia and Oman and designing the concrete mix for Miel 1 Dam in Colombia which, when it was completed in 2002, was the world's tallest RCC dam at 188m.

He says a brief explanation of concrete helps to understand the issues with Paradise Dam.

"People often call concrete 'cement'," he says. "Cement is a component of concrete."

The rest is a mix of gravel, water, sand and other fine aggregates, and sometimes additives like fly ash to "improve the concrete performance". In this case, how well a dam holds back water.

Paradise Dam, Williams says, was built with "very fast and cheap" RCC, a construction method developed in the 1970s that has become a popular choice for big dams.

All photos supplied: Flickr/Rod Savidge

When a dam requires hundreds of thousands of cubic metres of concrete, RCC can mean massive financial savings, Williams says, because it uses less cement, the binding ingredient.

RCC refers not to the mix but to how it's applied.

"It looks like wet gravel, not like very liquid concrete," Williams says.

"You move it with a conveyor or a truck, you push it with a 'dozer and you compact it with a road roller, hence roller-compacted concrete."

Each RCC layer is about 300mm thick and, once placed, rolled and compacted, the next layer goes on top.

After the 2013 flood, as engineers repaired the void at the base of Paradise Dam, they noticed the layers of concrete were not holding together as they should.

Months of testing in 2019 concluded another severe flood could create a risk of the layers shearing and the wall effectively coming apart at the seams.

"The stability was less than required by the dam safety guidelines," James Willey says.

Unaware there were more problems with the RCC wall than it knew, the Queensland government ordered work to lower the top of the spillway by 5.8m and almost halve the water in the reservoir, reducing pressure on the wall.

That decision had a profound and immediate impact on the region, which in 2019 received just 320mm of rain, its lowest since 1942.

Farmers protested, outraged at a lack of consultation on dropping the dam wall.

Little did they know what they were demanding, that the spillway stay at its original height, could put their community in grave danger.

Farmers fire up for a fight

The region's producers could barely believe the increased water access they'd lobbied for and celebrated was about to be reduced by almost half.

At the time, Bree Watson, then managing director of Bundaberg Fruit and Vegetable Growers, now an LNP candidate for the seat of Bundaberg at this weekend's state election, argued the move put the region's agriculture value of "well over $1 billion" at risk and demanded RCC experts re-assess the dam.

"We find it really difficult when they talk about saving $100 million here or there," she told ABC's Landline.

"What would happen if we didn't have this agriculture industry in the region?"

Strawberry and macadamia nut grower Tina McPherson worried the lost water security would not just squeeze irrigation for farmers but also discourage investment.

"You don't put those macadamias in without knowing that you've got water to buy into the future," she told the ABC.

As debate raged, the Queensland government in 2019 ordered a commission of inquiry while farmers failed in a legal bid to stop the remediation work.

The inquiry report in May 2020 found the dam had structural and stability concerns due to its design and construction. Namely that:

• The RCC layers lacked sufficient "shear strength" and were at risk of sliding along their joints.
• The apron, or flat bed of concrete at the foot of the dam, was not wide enough "to resist the erosive force of water overtopping the main spillway".

By early 2021, work to lower the dam wall was complete, reducing storage capacity from 300,000ML to 170,000ML.

But by that time, Jon Williams had picked up worrying signs of an even bigger problem in concrete and rock samples stored for testing in a shed tucked away to the side of the dam.

Core samples of concrete meant to survive the elements for a century were breaking down.

The devil in the dust

"We found the signs of something unusual in the concrete," Williams says.

"There was dust on samples. You could pick stones out of the outer face of the dam itself.

"There were just anomalies in the performance of the concrete, which were causing some questions among the team.

Experts from the US, Canada, New Zealand and Australia were engaged as a technical review panel to assess the next steps.

"Those experts came back and said, 'Yeah, you actually need to investigate this further,'" Williams recalls.

But there was no way to do that easily on dams because it had not been done before.

"You don't normally test for degradation of concrete structures, as concrete is normally a very durable material," Williams says.

"The literature search we did … did not come up with a method of testing degradation for dams."

Degradation, or the weakening of concrete over time, had only been tested on RCC in US airport runways that endure freezing winters and subsequent thawing — not an issue in Queensland's subtropics.

Soak, dry, crush, repeat

The GHD team realised it had to replicate the rigours of wet and dry seasons on Paradise Dam by artificially aging the samples.

That meant creating conditions of "accelerated degradation" by, effectively, speeding up time.

"We were trying to extrapolate decades of performance in months of laboratory testing," Williams says.

It took just over a year to develop methods, gain approval from the international panel and set up a lab.

In mid-2022 the team began a series of tests including repeatedly soaking samples in water from the dam reservoir, in some cases for four months, drying them, soaking them again, then testing their strength by compressing them until they crumbled, asking the panel to challenge their work at each step.

"One of our review panel members commented that the process we'd come up with was worthy of a master's thesis," Williams says.

"So that gives you an idea of the rigorous nature of what we were doing."

By the end of 2023, the tests had revealed the concrete was degrading at an alarming rate.