Forecast ignored before the floods

THE fuzzy outline of what looks like a mountain appears in the cockpit windscreen of the passenger jet.

Wivenhoe Dam

THE fuzzy outline of what looks like a mountain appears in the cockpit windscreen of the passenger jet.

The experienced and highly trained crew can make out the potential danger through the fog. At the aircraft's present altitude there will be a crash if the outline is indeed a mountain.

The captain, co-pilot and first officer are in a position to weigh the risk of a catastrophe. But something else is occurring on this fateful flight.

As the outline becomes larger, the pilots are transfixed by the instrumentation and on-board computers, not the terrain.

The captain had programmed the computer with a flight plan based on "no mountains". As the computer had been instructed to factor in "no mountains", it is not warning of an imminent collision.

When it is too late, the crew members take desperate action. But in the ensuing inquiry, the airline and its paid experts and witnesses try to persuade investigators that the crew reacted prudently based on the computer's advice. The airline insists that at all times the crew followed the aircraft's operating manual.

The airline's lawyers argue that if what had appeared to be (and turned out to be) a mountain had in actual fact been merely a mirage, any action taken to change course could have put people at needless risk.

It is an odd explanation. How can it be logically argued that, instead of mitigating the risk of a catastrophe by taking early avoidance measures that have limited downside, it is better to multiply the risk by taking little or no action? After all, the consequence of not taking early action is a disaster.

The precis is fictional; however, it is analogous with one of the worst disasters in Queensland's history: the January floods, which inflicted a multibillion-dollar hit on the economy, turned upside-down the lives and homes of thousands of residents, and contributed to Australia's present economic malaise.

Think of the aircraft in this tale as Brisbane's Wivenhoe Dam. The pilots are the dam's operators. The airline is the dam's owner, SEQWater. The mountain is the rainfall, which was heavy, and forecast to continue to fall heavily, as the lake level in the dam rose rapidly with run-off from a saturated catchment. And the onboard computer's programming can be compared to the "no rainfall" modelling relied on by the dam's operators to determine the volume of releases of water from the dam.

The fact the modelling did not take into account the Bureau of Meteorology's forecast rainfall after a monsoon front had been identified needs to be well understood. The modelling's "no rainfall" input led the operators to make relatively small releases of water early in the crisis as the run-off into the dam quickly escalated.

As the lake in the dam rose to worrying levels (because of the run-off that had not been factored into the modelling), the operators realised, too late, that drastic action had become crucial to protect the structural integrity of the dam. They reacted by opening the gates for huge and unprecedented releases, a massive flow of 7500 cubic metres a second. This flow was most of the Brisbane River flood, more than the water coming down the Lockyer Creek, and more than the water flowing in the Bremer River.

If the decision by the dam's operators not to factor in what laypeople see as an obvious input -- forecast rainfall -- means a breach of the manual, then under the legislation SEQWater and the state government (which owns Wivenhoe Dam) lose their indemnity from claims for damages.

On the evidence at the ongoing royal commission-style inquiry, run by Supreme Court Justice Cate Holmes, the failure to take into account forecast rainfall is a breach. Other important deficiencies may be traced to the inaction of senior Water Grid bureaucrats and politicians who are accountable for managing vital and potentially dangerous infrastructure, yet appear, from the evidence, to be better at handballing their responsibility to others.

The breaches of the manual look plain. One of the flood engineers was not even registered. But whether there will be serious financial repercussions depends on crucial evidence not yet in. The inquiry has not released independent hydrodynamic modelling showing the difference to flood levels across Brisbane and Ipswich if releases of water had occurred earlier as a result of forecast rainfall being taken into account.

SEQWater has failed to release such modelling and it is unclear if it has attempted it. It is the most important missing piece of the puzzle and a significant omission among the thousands of pages of evidence to date.

HOW the operators of Wivenhoe Dam could have overlooked or reinterpreted or decided not to follow the requirements of the manual might be in part due to its recent revision. Unlike earlier manuals, the current version requires the operators to use the predicted lake level, not the actual lake level, as the basis to select strategies for releases of water.

The other key requirement of the current manual in this context states: "The strategy chosen at any point in time will depend on the actual levels in the dam and the following predictions, which are to be made using the best forecast rainfall and stream flow information available at the time. Strategies are likely to change during a flood event as forecasts change and rain is received in the catchments." (emphasis added)

Independent engineer Michael O'Brien's concerns, reported extensively in The Australian, about the operation of Wivenhoe Dam dam have been vindicated by the direction and focus of the floods inquiry. He points out that SEQWater, seeing the rainfall and the forecasts but not responding to the predicted consequences in January, was always behind the curve and too slow to respond to releases until it was too late.

Phil Hassid, a former mathematician who has followed the evidence, tells Inquirer: "The group think, silo effect and overdependence on models is alarming. You have to counter-balance risks. If you are in a situation where you know there is a theoretical potential for the imminent occurrence of something that will significantly worsen things, then the less confident you are about predictions as to the occurrence of that thing, the more you should make contingent allowance for it, not less as the [dam] engineers argue.

"The [dam] engineers have succeeded in inverting the correct logic by saying, instead, that in the absence of solid reliable forecasts that there will be large following rain, you have to act as though there won't be."

When the inquiry's senior counsel assisting, Peter Callaghan SC, questioned the dam's most senior engineer, Robert Ayre, he asked: "You have just acknowledged, I think, in your evidence, and it certainly comes through in the statements made by yourself and the other engineers, that the means by which you ascertain the predicted lake level is the 'no further rainfall model', is that correct?"

Ayre: "That's correct, yes."

Callaghan: "For the purposes of choosing which of the strategies is to be deployed?"

Ayre: "Yes."

Subsequent questioning by Darryl Rangiah SC on behalf of flood victims produced this exchange: "Now, if you don't take into account the predicted levels of rainfall in decision-making, then do you proceed on the assumption that the predicted rainfall won't happen?"

Ayre: "The 'no rainfall' scenario is indeed that, yes."

Rangiah: "So, in other words, the 'no rainfall' models ignore the probability or even any possibility of there being the predicted rainfall?"

Ayre: "Yes, they're based on what we actually know, which is the rainfall that's already occurred on the ground."

Rangiah: "So, the 'no rainfall' models use what is the best case scenario; that is, there won't be any further rain?"

Ayre: "That's correct, yes, it's a lower level -- lower limiting model scenario, if you like."

Rangiah: "And the worst case scenario is rain significantly exceeding the Bureau of Meteorology predictions?"

Ayre: "Quite likely, yes."

Rangiah: "And what, in effect, happened was that the [January] rainfall significantly exceeded the predicted rainfall?"

Ayre: "On the Sunday and Monday and Tuesday, yes."

Rangiah: "Wouldn't predicted rainfall also have a relevance in the rate of release that you adopt within a particular strategy?"

Ayre: "The predicted rainfall is highly variable and so depending on how consistent those forecasts have been, it may be incorporated but generally speaking, no, we don't -- we do not take it into account."

Rangiah: "In other words, you base your rate of release solely upon the rainfall that has actually fallen and not on the rainfall that is predicted to fall?"

Ayre: "In most cases, yes."

The reality is that the forecasts of rainfall were serious. The rainfall did eventuate and it was significantly greater than predicted. The Bureau of Meteorology's unprecedented warnings of extremely wet La Nina-related weather -- including a briefing of Premier Anna Bligh's cabinet in October 2010 and another in early January -- were right, along with its warnings of a very large monsoon front shortly before the flood. The problem was that the forecasts did not influence the operations of the most important piece of public infrastructure for flood mitigation.

Part of SEQWater's reasoning for not operating the dam based on rainfall forecasts involves their accuracy; often the actual rain is well below what was forecast and often it is well above. But it is highly unlikely that Wivenhoe Dam will ever again be operated without regard to the forecasts.

The dam's operators accept that the rapid increase in releases on January 11 and 12 to 7500 cubic metres a second did "certainly produce flooding in Brisbane" (4000 cubic metres a second is the level at which flooding and damage is known to occur).

They have agreed that if there had been higher releases starting from Saturday, January 8, lower releases would have been required on January 11 and 12. But they don't believe, based on the information at hand at that time, "that necessarily higher releases were justified".

Giving no weight to the rainfall forecasts when deciding release rates made a difference to the inundation of thousands of homes. Some say the difference was profound, that the flood was almost completely avoidable. Others say the difference was not great. Quantifying it with analysis independent of SEQWater should be a priority for the inquiry, particularly so that those who were most affected, such as the uninsured who lost their houses, can weigh up legal action.