As floods continue to cover parts of Australia, scientists are calling for an urgent change in thinking on how we predict and cope with flood events.
They say poor records, changing background environmental conditions and human impact on a local scale all make accurate flood modelling something of a moving target.
Dr Jerry Maroulis, a fluvial geomorphologist with Landloch Pty Ltd, witnessed the Toowoomba flash flooding on January 10 first hand, and says it was caused by a freak convergence of several factors.
"We had a saturated catchment, a catchment area which is covered with impervious surfaces and then we have the 'perfect storm' - a storm which dumped this deluge at just the right time," Maroulis says.
"'Inland tsunami' was the perfect description for what happened."
Maroulis believes the urban environment was the main contributing factor in the flash flood that tore through Toowoomba.
He says records going back to the 19th century indicate that the amount of rainfall that fell in the Toowoomba catchment before the flash flood was not unusual and had been exceeded on several past occasions, but without the dramatic results.
As the city has grown, the catchment's natural vegetation has been increasingly replaced by impervious surfaces, which send runoff straight into the concreted watercourses, says Maroulis.
"Before the Toowoomba area was settled it was largely a chain of ponds, so there were a lot of reed beds and all the rest of it," he says.
"The nature of development is that we like to control things and use channels to control the creeks, and that's understandable. But the consequence is you change the flow hydrology and the flow dynamics. And when you've got smooth flowing channels, water flows very quickly."
Wake-up call to town planners
Maroulis says the events in Toowoomba should be a wake-up call to other towns in similar geographical situations. He says they should undertake flood modelling to predict what would happen under similar conditions.
Stephen Gale, a hydrologist at the University of Sydney, agrees that badly planned urban development is a good recipe for flood damage.
"The most significant effect is the fact that, particularly around cities like Brisbane, we gobble up agricultural land and we develop it," he says.
"We change a fairly permeable surface that lets water go into the soil moisture system or the groundwater system and move slowly from there into the river system. We've replaced that with concreted surfaces, [which is] very efficient at getting water into the river channel.
"Hydraulically it's superb, but of course it's a disaster if there's a big rainfall event."
Flood modelling inadequate
According to Gale, another problem is the inadequacy of flood modelling systems.
"There are two major things that the conventional approach to dealing with this doesn't take into account, and the first is that our instrumental records are short," he says.
"If you have 50 years of records, then you're doing well. The trouble is, the rule of thumb that's used in most places in the world is that it's the 'hundred year flood' which we need to deal with.
"Forty or fifty years of records don't allow us to predict that with any kind of reliability", he says.
Indeed some scientists say the flood that caused so much damage in Brisbane was only a one-in-25 year flood, and only the sixth highest on record.
Flood could arrive at any time
"The hundred year flood is one that has a probability of occurring once in a hundred years," Gale says.
"Now that's not to say that it can't occur tomorrow, or that you can't have hundred year floods in successive years, or even in successive months. The big difficulty from our point of view is that because our records are so short, we haven't actually experienced many of those big floods. So it's very hard to model their probability."
Gale says modelling is complicated by the fact that climate goes through natural oscillations - phenomena like El Niño and La Niña.
"There are lots of periodicities and quasi-periodicities that occur," he says.
"Now what that means is that if your record is derived under a set of low-flow and low rainfall conditions, if you then shift to a period of high rainfall conditions, then the game has changed.
"Your predictive capacity just doesn't work anymore. There's a lot of evidence to suggest that in southeast Australia there's an alternation in rivers between drought dominated regimes where you get decades where flows are pretty low, and then flood dominated regimes, where you get decades of big floods."
Natural archive contains clues
He says if we want a better understanding of what may be going on in long term cycles, we need to look at the natural archives. Sediment cores are one valuable source of information about flood events. Coral reefs provide an even better record, Gale says.
"Corals are a great indicator. They grow quickly, they have annual growth rings, so you can date them like trees," he says.
"You've got rivers discharging lots of silt into that area between the mainland and the Great Barrier Reef. That silt gets incorporated in the corals and you can actually look at the frequency with which big floods occur and date them very well.
"Indeed there are records that go back in the order of three hundred years that have been determined in that way."
"When people do that, what they find is that big floods occur much more frequently than the instrumental records suggest."