Satellites spotted part of the crater of Whakaari / White Island collapsing in the days before the deadly December 2019 eruption, it has emerged.
Scientists believe such images could be integrated into a supertool to forecast eruptions and save lives.
Geothermal scientist Dr Mark Harvey is currently on a boat with his family, heading to the Bahamas to escape Covid in the US, where he runs his business out of Connecticut.
Previously, he was ploughing through radar photos of Whakaari taken as satellites passed overhead every three-five days, "like a three-dimensional cartoon over time", he said.
"You could see a particular area within the crater floor in the lead-up to the eruption, is rising ... like a little bit of a dome starting to form."
This occurred from 2018 on, then reversed in the fortnight before the fatal eruption that killed 22 people.
"The dome looks like it starts to collapse."
A simple explanation was that magma was coming up from depth, "a little bit like popping this high-pressure geothermal system with a pin".
A scientific journal has just published his research.
"This finding has immediate relevance for ongoing hazard monitoring at White Island and other volcanoes worldwide where sudden dome contraction may presage an eruption," his article stated.
Using satellite imaging to help forecast eruptions depended on having enough data crunchers to keep up with the big advances in technology, Harvey told RNZ.
"It's an incredible volume of information that just wasn't there five years ago.
"I think there's an amazing opportunity to utilise it in volcano forecasting" in the next five years, he said.
The Whakaari disaster is giving up key volcanic secrets to help in what Auckland University Professor of Volcanology Shane Cronin describes as the "dark art" of forecasting - secrets revealed in detecting changes that preceded the eruption, not just in the crater's ground formation, but in gas emissions, electromagnetic fields and seismic tremors.
Integrate all those, and you could have a supertool that saved lives, not just at volcanoes but at geothermal fields, Cronin said.
"It's not pie in the sky.
"I think it's an achievable outcome. It's something that we could look forward to, I would say, over the five-year timescale, I'm not saying 20 years.
"Takes a little bit of bravery though ... to run a tool like that.
"You've got to take it on the chin that there's going to be some false alarms."
The most progress has so far been made with seismic forecasting, with Cronin helping develop a system that would have given warning of a seismic surge 16 hours before the Whakaari eruption.
Mount Ruapehu
His collaborator, Dr David Dempsey of Canterbury University, hurriedly applied the tool at Mount Ruapehu when the crater lake began heating up just before last Christmas.
The surge of seismic energy the system spotted at Ruapehu looked similar to one preceding the 2006 eruption, unnerving observers.
"Now that surge is different in size and in duration to Whakaari, which means we can't simply map over that model to the Ruapehu data," Dempsey said.
"But it is a good sign that there is something there to be recognised, and something there to learn from and potentially issue warnings on."
It was trickier at Ruapehu as there were fewer past eruptions to get data from, plus the background seismic noise from storms - and even skilifts - was higher.
"We've certainly learned a lot about what I think is possible from Whakaari," Dempsey said.
"There is still quite a way to go before this can really be generalised to a lot of other volcanoes, but that is something we are pursuing as fast as we can."
Scientists overseas are helping out; some are trying out the seismic forecasting on volcanoes that erupt regularly in Europe and South America, others are providing Dempsey with data to crunch from closely-watched peaks in Alaska and Mexico.
The other forecasting tools were in a more experimental phase, and integrating them would be a big job, Dempsey said.
"It's entirely feasible. It's exciting that there are so many data streams.
"The satellite uplift data is really interesting, because ... you can rely on that satellite passing over regularly to give you information.
"The electromagnetic fields is something we're investigating as well."
Gas geochemistry, while sampled less often, was useful.
What was really scarce was the time to evaluate the models to gauge what data was genuinely useful, and then to integrate them into, say, a "machine-learning" approach, he said.
GNS Science heads up volcano monitoring and raises or lowers alert levels, and is among 13 parties charged with breaching health and safety laws over the Whakaari eruption.
It had been kept up to date about the seismic tool's evaluation since it began a year ago, as part of reviewing many types of monitoring data, the agency said in a statement.
"The forecasts generated by this technique are not accurate enough yet to give a firm prediction of an eruption happening at a particular time, nor do they indicate the size or potential impact," it said.
"So the tool it is not operational in the wider sense of feeding directly into decisions about access to the island by external parties."
Time to take courageous approach
Professor Cronin is urging action.
It was time for more "courage" in place of the caution that typically slowed down implementation of forecasting technologies.
"For New Zealand, we're in landmark position," Cronin said.
"The Whakaari situation really has sort of thrown the spotlight on us... Are we being too conservative in some of the tools we're applying to forecasting?
"So it's a national moment of reckoning."
The scientists have been in talks with GNS for months about deploying the seismic forecasting tool.