The family of tornadoes that destroyed more than 20 Kiama homes and let rip a 16-kilometre-long trail of destruction to the outskirts of Nowra were part of an East Coast Low that started its life off the coast of south-east Queensland.
The system developed on February 18, almost a week before it hit Kiama.
It crossed the northern NSW coast on Friday, bringing strong winds and heavy rain.
On Friday it seemed to tire, after moving inland to the northern ranges, up around Glen Innes, close to Tamworth and past Dubbo.
But a band of strong winds continued to buffet the coast, generating gusts of 80-90km/h in the Hunter region on Saturday, and 135km/h in Sydney towards midnight.
In the early hours of Sunday morning it moved down the Illawarra coast.
Stephen Jarrett, an Oak Flats meteorologist-in-training and a 12-year veteran of the storm-chasing movement, was waiting for it.
He parked his car at Barrack Point and sat inside with his camera and a laptop showing weather radar and the results of a subscription lightning tracker service.
Mr Jarrett planned to sit out the storm, photograph the lightning and report his observations in real time to the 4000-odd fans of his Illawarra Stormchasers Facebook page.
At about 12.30am he felt his car rocking from the force of the winds. Soon afterwards, he drove home.
"The lightning rates were indicating to me that there was a lot going on in the atmosphere," Mr Jarrett told the Mercury.
"The winds were starting to pick up and that's when I made the decision to head back home and hunker down. That was based not only on my instincts and observations, I was also watching the radar very closely. I was starting to see things.
"It looked like there was a bit of rotation in the storms as they were coming through Sydney. It was telling me it was too much of a safety risk to continue chasing this storm. I was thinking there may be tornadoes."
The storm passed Wollongong at 2.30am, still about 50 kilometres out at sea.
Experts from the Bureau of Meteorology believe they have since spotted a "little system" form within the broader storm system, registering as a subtle shift in radar imagery at about 2.30am.
The storm - tens of kilometres across - made landfall in Kiama at 3am.
The bureau is still investigating the extreme weather events that followed, but believes the little low-pressure system was the trigger for four, possibly five, tornadoes that hit the region.
Tornadoes are not uncommon in Australia but the number formed on Sunday morning, their closeness in distance and timing, is a rarity not witnessed by Michael Logan in his 10 years with the bureau's severe weather team.
"I'd say it [the smaller system] was the trigger for all of them," he said.
"What we'll have to go back and have a good look at is why nothing like that formed further north in the Hunter or in Sydney."
Mr Logan was one of two bureau experts sent to the region on Tuesday to study areas of destruction resulting from the storm and determine if tornadoes were to blame.
The bureau confirmed and rated three tornadoes according to the six-tiered Enhanced Fujita Scale (EFO) of tornado intensity.
They say the weakest - an EF0 - swept inland eight kilometres from the coast to Jamberoo at speeds of about 125km/h.
There was an EF1 at Kiama, lasting four kilometres at about 160km/h.
The strongest hit Seven Mile Beach and drove a 16-kilometre scar towards Nowra, reaching speeds of about 200km/h.
The bureau also accepts as credible reports of a tornado at Albion Park Rail, and say it is possible a fifth hit at Gerringong.
"From some of the stories we heard it's incredibly lucky no-one was seriously injured," Mr Logan said.
"It was lucky it came through when it did."
Mr Logan said there was no warning of the tornadoes because the trigger system developed so soon before they hit, and because the distance of the tornadoes from the nearest Doppler radar, located 18 kilometres north-west of Wollongong, made radar readings inconclusive, even after the event. Some tornadoes don't reach high enough into the atmosphere to be recorded on Doppler radar.
"Even as an organisation we didn't know there'd been tornadoes until we started to get damage reports from the SES," Mr Logan said.
Mr Jarrett believes he saw tell-tale "signs of rotation" in the storm cells as they were coming through Sydney.
"It's generally accepted that you have a possible tornadic environment when that occurs," said Mr Jarrett, a first-year bachelor of science physics student.
"But it is very hard to call [a tornado] initially because you need to have the damage path surveys completed to assess if the path was confined, or if it was a broad area of general destruction."
In the aftermath of the storm, the damage was attributed to multiple weather phenomena, including a waterspout and a downburst - a strong downward current of air capable of causing wind speeds in excess of 100km/h.
The bureau says a downburst, or multiple downbursts, may have occurred as part of the same system and may be responsible for wider, more generalised areas of damage.
Mr Jarrett said a waterspout could not have caused such serious damage.
"A waterspout will take off some tiles and break some windows whereas a tornado can, as we've seen, take an entire storey off a house, because there are stronger wind processes involved."