Most people had the impression that the four flood engineers, Robert Ayre, John Ruffini, John Tibaldi and Terry Malone, had done pretty well during the Brisbane 2011 flood event, which began on 6 January when the Wivenhoe Dam first exceeded full storage capacity of 67m bringing the flood compartment into play. The gates were finally closed ending the event on 19 January. Later some argued that on the Saturday and Sunday 8-9 Jan there could have been more aggressive releases, flooding the remaining rural bridges below the dam at Fernvale and Mt Crosby Weir, thus lessening the later Brisbane flood peak. Against that, it was argued that the weather forecasts didn’t justify such action, that the effect would have been minimal and that some low level flooding may have resulted from that act alone. What’s more, the forecast rain may not have eventuated.
In February the Queensland Floods Commission of Inquiry reconvened as a result of reporting in The Australian by Hedley Thomas which suggested there had been a major breach of the flood manual for the operation of the Wivenhoe and Somerset Dams (7th revision) and that there had been a cover-up to disguise the fact. The Commission’s Interim Report of August 2011 had been thorough and sympathetic. The Commission noted such things as non-compliance in relation to weather forecasts (the forecasts were so erratic the engineers ignored them in determining gate settings, contra the manual). They also noted deficiencies in recording decisions and actions in real time. They made a host of recommendations about improvements (see Recommendations relating to Chapter 2) but the impression was of a job well done. Chapter 2 (Section 2.7) remains perhaps the best blow by blow account of how the flood was managed.
Chapter 16 of the Final Report does in fact find that the manual was breached from 8am on Saturday 8 Jan for about 36 hours until the evening of Sunday 9 January. It found that the engineers were operating in strategy W1 rather than W3 as claimed in the Flood Event Report of 2 March 2011 (downloads at bottom of screen). It also found that three of the engineers knowingly misled the Government and the Commission about the “W” strategies. It recommended that Ayre, Tibaldi and Malone be investigated by the Crime and Misconduct Commission in relation to the brief to the Minister of 17 January 2011, the Flood Event Report of 2 March and other documents, as well as in testimony to the Commission to see whether offences had been committed against the Criminal Code and the Crime and Misconduct Act, 2001.
Serious stuff. Hedley Thomas in these articles was well-satisfied with his work. He spoke of “mismanagement”, a lack of competence and assumed that compensation was clearly due to flood victims who had had their homes and properties needlessly damaged. Here Thomas, presumably with presumed hydrological expertise not available to the Commission, has gone beyond what they said. They had asked an independent hydrologist, Mr Mark Babister, to review the flood documentation and reporting and to undertake modelling of alternative strategies. Babister’s report had become available just as the Interim Report was to be printed. His finding could not be incorporated in the document, but a statement was inserted saying that nothing in his findings conflicted with the Commission’s interim report. He found that:
the flood engineers managed Wivenhoe Dam so that its flood mitigation effect was ‘very close’ to the maximum achievable within the constraints of the manual.
Furthermore, he said that an aggressive early release would have been “highly risky” and would have used the Wivenhoe as a flood amplification dam rather than a flood mitigation dam. That’s all from the Final Report. The Commission did not back off Babister’s findings as alleged by Thomas. What they did say was:
Ascertaining the practical result of acting more quickly also is subject to the uncertainties inherent in the modelling; but again, the possibility exists of at least some improvement in the flooding outcome for Brisbane and Ipswich.
The possibility of “at least some improvement”. And if the class action lawyers want to know exactly which homes and buildings would have been saved they will have to do their own hydrological modelling. The models Babister used were not sophisticated enough to show this, and in any case his technical modelling has not been made available.
The problem, you see, was that the engineers thought and acted in relation to the information they got from rain gauges, dam input and output flows, dam levels and conditions outside the catchment areas, modelled to forecast a stream flow at Moggill 16 hours after dam releases.
What they were meant to do was to pause there, check which “W” strategy they were in, and then make decisions. What they did was to go directly to making decisions. The “W” strategies were filled in later as part of the reporting. In doing so they followed the same methodology that had always been employed from back in the 1990s.
The Commission took the view that the engineers would make better decisions if they operated a different way, that the “W” strategies be consciously chosen and be front of mind at all times. What’s more, compliance with the manual demanded it, they said.
I’ll come back to these issues in a second post. In this one I’d like to give an account of the actual event, which was a good deal more threatening than most realise.
The Wivenhoe Dam in context
The Wivehoe Dam catchment drains some 7,000 square kilometres of the Upper Brisbane River. It includes the Stanley River, a tributary flowing in from the north which is dammed by the smaller Somerset Dam. The Locker Creek draining some 3,000 square kilometres joins from the west just below the Wivenhoe and just above Lowood. The Bremer River draining some 2,000 square kilometres flows through Ipswich and joins just above Moggill. It takes about 16 hours for water to flow from the Wivenhoe to Moggill.
A flow of 4,000 cubic metres per second (cumecs) at Moggill is taken in the manual as a marker for urban inundation downstream in Brisbane.
You can get an overview of the geography from the two maps in this post. The total average rain during the flood event is given in this map:
Three points to note. First, the Pine River is an entirely separate system, not a tributary of the Brisbane River. Second, there is a slab of country in the segment designated “Lower Brisbane” which would also drain into the river above Moggill. Finally, the cloudburst that flooded Toowoomba and devastated Grantham and the Lockyer Valley doesn’t show up in the overall. That might be because there was no rain gauge on the eastern escarpment of the Toowoomba range. The other catchment numbers might also be underdone as the engineers found that there was more water coming at them than was implied by the modelling based on the gauges.
A monster flood
In the 2011 event 2,650 gigalitres of water flowed into the Wivenhoe dam. The comparable figures for 1893 and 1974 would have been 2,740 and 1,410 respectively. In February 1999 there was an event with 1,220 GL that most people don’t remember, because the Wivenhoe had been built and coped with it. Sydney Harbour holds about 500 GL. Here is a graph showing historical flood levels:
That’s at the Port Office in the city reach. The 2011 flood was 4.6 metres, but the Flood Event Report tells us that it would have been two metres higher without the Wivenhoe. Later modelling by Sinclair Knight Mertz suggested about 1.2m higher.
In fact we had a flood in two parts, 30 hours apart. The dams would have coped pretty well with the first, but were overwhelmed by the second. The pattern is shown in this image from the report:
The Executive Summary of the Flood Event Report (well worth a read) tells us that the maximum flow rate of the first peak has been estimated to be 200% of the 1974 event and the second 230%.
The story, in brief
The flood operations centre was established, as is required, on 6 January when the Wivenhoe Dam reached the full storage level of 67 metres. Malone and Tibaldi were engineers from Seqwater, which owns and manages the dam. Ayre was a senior engineer from Sunwater, which was contracted by Seqwater to operate the centre. Ruffini was a senior engineer from DERM. They were supported by 13 flood officers at various times. My impression is that the flood operations centre was in the Brisbane CBD.
Strategy W1 was then automatically in operation on 6 January, where the focus is on the impacts on rural life below the dam, principally the river crossings. By Friday night the five lower crossings were inundated, leaving the Fernvale Bridge just below Lowood, which goes under with 2,000 cumecs, and the Mt Crosby Weir Bridge down near Moggill just above where the Bremer joins, which goes under with 1,900 cumecs. The dam level was predicted to rise above 68.5m, which triggers either W2 or W3. At 5am on Saturday morning Ruffini issued a gate opening schedule to cope with this situation, based on modelling done by Malone the previous day. At 8am on Saturday when the 68.5m trigger point was breached there was about 171 GL in the flood compartment, with inflows at 1,500 to 1,800 cumecs. Outflows were already above 900 cumecs and were increased to about 1,200 cumecs during the day.
Under both W2 and W3 priority must be given to preventing urban inundation, with 4,000 cumecs at Moggill the marker. Consideration is also given to lower level objectives, principally keeping the two higher bridges open. It should be noted that at 8am natural flows at Moggill, that is flows other than those from the dam release, were at 770 cumecs. At least two engineers (Ayre and Tibaldi) knew that water troubles in Brisbane start way lower than 4,000 cumecs at Moggill, in fact around 2,000 cumecs.
The increased releases saw the dam rise to 68.65m at 11pm then begin to fall again. Five of the six situation reports issued on Saturday and Sunday mention the predicted impact in millimetres at the Port Office in the Brisbane CBD. The exception was Sunday morning, which mentioned the dam level falling.
Overnight rain on Saturday night caused concern. It was decided to have a meeting of all four engineers on Sunday afternoon, with three present and Tibaldi on the phone. Malone compiled a scenario in preparation. Both this document and the meeting notes mention the rain influence tending to miss the upper Brisbane segment and move south to the Lockyer and Bremer catchments according to the forecast. They decided to keep releases to about 1,400 cumecs with the possibility of holding back more water to allow the Bremer and Lockyer water to pass.
Although there was considerable rain to the south the weather system did in fact settle over the Upper Brisbane and especially the Stanley catchment and Mt Glorious. An outflow level of 1400 cumecs decided at 3.30pm was passed at 6pm and increased steadily from there.
The first peak inflow of 10,095 cumecs was reached at 8am on Monday 10.
The cloudburst near Toowoomba which devastated Grantham occurred from about 2pm on Monday. The engineers didn’t know about it for about three hours as it fell between the gauges on the eastern escarpment of the range and didn’t show on radar. It takes about 24 hours for the water to get to the Brisbane River so they had plenty of notice. At 8pm the engineers were advised that the rain in the Upper Lockyer was estimated at 600mm.
Monday was a wild day. It began with a phone call at 12.45am from the Brisbane City Council advising that according to their maps inundation would occur with 3,500 cumecs at Moggill, not 4,000. There was a second phone call from the BCC at 9.40am. I think it was during this one that the engineers were told that a flow of 3,500 cumecs at Moggill would fully submerge 322 properties and impact 7,000. The engineers agreed to try to hold it to 3,500, but had to abandon this by 3pm. By that time the situation had deteriorated to the point were they had to warn of possible Brisbane flooding and the dam was forecast to rise to 75.2m with rainfall forecasts included. Then they heard about the Toowoomba/Lockyer downpour.
And then as the clock went past midnight the fun really began.
The inflows fell until 2am on Tuesday when they reached 3,594, but the rain had increased. In the early hours of Tuesday morning it became apparent to the engineers that the water coming at them in the Wivenhoe catchment was more than the models indicated, based on the gauge readings. Estimates indicated falls of up to 700mm.
Inflows rose again to reach a new peak of 11,561 at 1pm. Meanwhile the dam level had reached 74m at 11am.
When the dam level is predicted to go above 74m the main focus must be on preserving the dam wall, because 244,000 people are said to be at risk. The first of three fuse plugs in the auxiliary spillway automatically opens at 75.7 metres, the third at 76.7. The wall is 80m.
When the inflows peaked at 1pm the dam was at 74.39m and forecast to go to 76.2. The Moggill flow was approaching 5000 cumecs and rising fast. The sluice gates at Somerset were closed to hold back a bit more there. Permission was gained to let the level at Wivenhoe float above 74m for a time. But even though the inflow was tapering the outflow had to be increased to a level above the inflows. This occurred at 7pm when inflows were 6,876 and outflows were 7,464. The dam peaked at 74.97 and tapered from there.
At 74m the dam was approaching 180% of normal storage capacity. At its peak it was 191% of capacity. The nominal peak flood is 225%, but the intent is to bring it back to under 74m as soon as possible without blowing the fuse plugs with releases as necessary, albeit with consideration of downstream damage.
The dam was above 74m from 11am on Tuesday until 6am on Friday.
The requirement is to drain down the lake to full storage capacity within seven days, which is why the outflows were held steady at about 3,500 cumecs during that process.
Moggill peaked at 12,095 cumecs at 11am on Wednesday 12 January. At the Port Office in the Brisbane CBD the flood peaked at 3.00am on Thursday 13. At that time the flood height was 4.6m and the flow 9,500 cumecs according to the Flood Event Report. During the week tides were up to half a metre lower than the previous and following weeks.
From this article:
An estimated 14,100 Brisbane properties were affected, while in Ipswich about 3000 homes were flood damaged. In Brisbane, the number of residential properties inundated totalled 1203.
The engineers decided to double team from Sunday night, with each doing a 12 hour shift every 24 hours. Later as they dealt with the dam level peak I understand all four were there at times, conducting reviews at 30 minute intervals.
As the floods rose in Brisbane three were cut off from home and slept in a meeting room. I’ve not heard how the flood officers who supported the engineers got on.
Precision in weather forecasting at these times is not really adequate to dealing with floods. For the first few days from 6 January actual rain was considerably less than the forecasts. On Sunday morning a 24-hour forecast of 50mm was issued for the catchment, whereas 149mm actually fell. On Tuesday at 4pm when decisions were being made to increase the outflows to above the inflows the forecast was for 75mm in the next 24 hours. In fact the system weakened and only 12mm arrived. Thankfully. And thankfully the engineers held their nerve and didn’t up the release in anticipation.
What difference would a 75% dam level make?
Turns out Babister’s modelling came up with 30cm (see Final Report Section 16.14.3, or p525, Scenario 3). But if the manual strategy trigger points were ratcheted down by 25%, the gain would be 60cm at the Port Office and 1.3m off the 17.6m level reached at Moggill. That’s perhaps worth doing, but I still worry, as did John D in the post, that in a drought you would leave yourself short of time to construct additional desalination plants.
Also it seems to me insane to wind down the dam in September if an El Nino is predicted. It is also insane to be letting water out at the end of March, as happened last week, to bring the dam below 75% at the end of the wet season when the La Nina has broken up.