As I mentioned in the other post, which is necessary reading for background, the Queensland Floods Commission of Inquiry in its Final Report did suggest that the engineers could have done better – a little, perhaps – but were sufficiently upset with Robert Ayre, Terry Malone and John Tibaldi to refer them to the Crime and Misconduct Commission.
In this post I’ll address the question of whether the engineers could have done better, or to use the Commission’s terms, examine whether any non-compliance with the manual was consequential.
As I reported in the earlier post, the Commission found that the dam had been operated in strategy W1 in non-compliance with the manual for about 36 hours from 8am on Saturday, 8 January. The dam crossed the 68.5m threshold at 8am on the Saturday. Three hours earlier, at 4.50am, John Ruffini had authorised increased outflows of about 1,200 cumecs, as against 351 at 5am. Inflows at 3am were 1,300 cumecs.
1,200 was achieved by 2pm with inflows about 1,800. By 10pm outflows exceeded inflows, 1,242 to 1,091. This situation pertained until 10am on Sunday, when inflows became 1536 as against outflows of 1,332. One of the difficulties for the engineers is that releases of less than 1,900 are compatible with strategy W1. This threshold was not crossed until 8pm on Sunday. It should be noted that there is no minimum release rate specified for W3, or indeed for W1 and W2.
The engineers claim that Ruffini’s decision at 4.50am constituted a move out of W1 as it was designed to deal with the dam level rising above 68.5m threshold. Which it did, and the dam level fell for 12 hours from 11pm Saturday. The Commission was uncharitable about this because the Ruffini’s releases were based on modelling done on Friday by Malone, when the lake level was not projected to go above 68.5. However, Malone’s release rates were not implemented. Ruffini chose to do so in the light of the new situation, modifying them slightly. The Commission obsessed with the fact that Ayre, who came on duty at 7am, changed nothing at 8pm, the time the level past 68.5 and the time nominated by the Flood Event Report as transitioning to W3.
Much of the problem here is that the engineers did not have the global “W” strategies front of mind, but operated at the next level of detail. In particular they didn’t discriminate at all between W2 and W3, where the broad parameters are almost identical, and forgot about the fine print, where the differences are quite stark. However, anything you can do under W2 you can also do under W3, so in a sense it is redundant. Their practice was to work the dam and fill in the fine detail at the reporting stage.
The Commission’s view was that engineers are likely to make better decisions if the have the global situation in mind as represented specifically by the “W” strategies of the manual. This is questionable, as they had some external expert advice to the effect that it didn’t matter. In addition, their independent hydrologist, Mark Babister was comfortable with manual compliance, given the ambiguities of the manual. What is relevant here is that if the engineers had consciously chosen W2 on Saturday they would have been obliged to reduce the outflows from Wivenhoe to less than the 550 cumecs, which was the assessed natural flow at Lowood, sourced mainly from Lockyer Creek, excluding Wivenhoe releases.
Now take a look at the graph depicting the inflow and release summary for the dam from the Flood Event Report of March 2011:
Hedley Thomas’s contention that the dam was mismanaged is based on the notion that a more aggressive release of water from 8am on Saturday could have ramped up the light blue line more steeply. This would have lowered the dam level enabling a smaller peak in the light blue line later, which determined the subsequent height of the flood.
Babister modelled a scenario where enough was released from 8am on Saturday to produce a total flow of 4,000 cubic metres per second (cumecs) at Moggill, that being the indicator for urban inundation (see Scenario 7 in Section 16.14.3, or p525 of the Final Report). He found that as a result the Port Office peak would have been reduced by 0.6m. So the flood would have been 4m, still above the major flood level, which I think is 3.6m:
Graph from modelling by Sinclair Knight Mertz (see Section 16.14.2 of Final Report).
A clearer view of the pattern of the actual flood is shown in this graph from the Flood Event Report (Figure 6.5.13):
It is relevant here to show the tide levels for the period. This graph is Figure 6.5.14 in the Flood Event Report:
Whyte Island is near mouth of the Brisbane River. It seems from the above graphs that the normal tides commonly take the river to the point of minor flooding in the tidal reach.
It’s also worth bearing in mind that the river spreads and flattens out as it goes. The flood levels at Moggill, Jindalee and Oxley were 17.6m, 13.1m and 8.3m respectively. Babister’s modelling showed that there would be a reduction of 1.3m at Moggill and Jindalee and 0.9m at Oxley. This is to me a surprisingly small amount. In many cases the amount of damage to a house with a metre or so less water would be minimal.
Babister also found that such an aggressive release would have been “highly risky”. The 4,000 cumecs at Moggill guard rail in the manual is problematic. At least two of the engineers, Ayre and Tibaldi, knew that significant water problems started with 2,000 cumecs at Moggill. The Flood Event Report recommended that 2,000 be considered for inclusion in a revision of the manual. This was taken up by the Commission’s Interim Report and incorporated into the Ninth Revision of the manual published last November, not as a marker between strategies, as I think would have been appropriate, but as a point to keep in mind.
2,000 cumecs also happens to be the point at which the Fernvale Bridge goes under. That bridge is not primarily a connector between rural areas. It is on the Brisbane Valley Highway, which is one of three ways of getting from SEQ to Central Queensland. From memory all three were cut by floods when the Fernvale went under. Between Fernvale and Moggill the Bremer joins the Brisbane, so 2,000 at Fernvale is a higher bar than 2,000 at Moggill.
Once the significance of 2,000 cumecs at Moggill is understood the matter of whether the engineers could have done better is easily resolved. At 8am on Saturday as the dam hit 68.52m the flow at Moggill, with water from the Wivenhoe excluded, was assessed as 770 cumecs, deemed the “natural flow”. The engineers had decided to increase the release rate to 1,200 cumecs which was achieved at 2pm. The BOM had forecast increased rain in the Wivenhoe catchment but for the weather system to move south over the Lockyer and Bremer catchments. Estimating the flow at Moggill in 16 hours time is not an exact science so there was little room to move before you could expect complaints downstream and people perhaps ringing their lawyers.
The difficulty with 4,000 cumecs at Moggill is further highlighted by a phone call made by the Brisbane City Council to the engineers on Monday morning. The flood engineers were told that according to the BCC damage tables 3,500 cumecs at Moggill would see 322 properties fully inundated and an impact on 7,000 properties. 3,000 cumecs would see 2,600 properties impacted in some way.
Babister’s modelling was done without consideration given to rainfall forecasts. I’m not sure what difference that makes.
It is relevant here that on Friday at 4pm the 24-hour forecast was for an average of 25mm across the catchment. Only 6mm fell. The forecast at Saturday 4pm was for 40mm. A total of 80mm fell.
Of course major damage was eventually done in Ipswich, where water in the Brisbane River backed up the Bremer for 15 kilometres. Ipswich people have been among the most vocal in criticising the engineers. No modelling was done by Babister to show what difference different release strategies would have made to Ipswich. The class action lawyers may have to work that one out for themselves.
As far as I know no modelling has been done which shows the conditions that produce damaging backup in Ipswich. This should have received greater attention from the Commission.
There have been no significant complaints about how the engineers operated the dam from Sunday evening, but there were suggestions that they could have moved faster on Sunday in response to overnight rain. Inflows were as low as 773 cumecs at 8am on Sunday, but by 3pm when they had a meeting of all four engineers the rate had increased to over 4,000.
Given the rain forecasts (the 24-hour forecast of 10am was exceeded by 300%), the action taken was reasonable. The notes of the meeting state, in part:
The rainfall system is currently in the N-E part of the catchment and expected to travel south over the next 24-36 hours according to the BoM forecasts. This has the potential to significantly increase flows in the Lockyer Ck & the Bremer River which potentially could close Fernvale Bridge and Mt Crosby Bridge and increase risk of flooding in Lower Brisbane. Releases from Wivenhoe Dam will be maintained at the current level of ~ 1,400 cumecs. If required, release from Wivenhoe will be reduced to contain the flow in the Mid-Brisbane to 1,600 cumecs and 3,000 cumecs in the Lower Brisbane. At this stage it is anticipated that levels below 102.5 in Somerset and 72.5 in Wivehoe can be maintained.
Remember even with the unexpected overnight rain on Monday morning Tibaldi was able to promise the BCC to try to keep the flow at Moggill below 3,500. At that stage there was every prospect of reasonably containing the first peak shown in Figure 1 (inflows peaked at 8pm). The second peak was kicked off by heavy rain in the Wivenhoe catchment early Tuesday morning. The water from the Toowoomba/Lockyer cloudburst would have started arriving at the Brisbane River junction from Tuesday afternoon.
Moggill peaked with 12,095 cumecs at 11am on Wednesday, just 16 hours after the peak release of 7,464 from the Wivenhoe at 7pm on Tuesday. The black swan event, if you like, was a combination of the Toowoomba/Lockyer downpour and intense heavy rain in the Wivenhoe catchment arriving in the Brisbane River at much the same time.
Perhaps the Engineers, on Sunday, should have taken account of the risk of higher rainfall than was forecast on Sunday. But the Flood Event Report rates the whole event at between 1 in 100-year and 1 in 2,000-years. So that too is an inexact science and I’m not sure that reminding the engineers of those odds would have inspired them to act differently. As the events unfolded they did seem to have access to records of what happened in previous floods and be mindful of precedents.
There was also a risk that the rain system might weaken or go somewhere else. Quite recently we had a forecast of 60-100mm falls in Brisbane City. We got about 6mm from memory, but they got 300mm about 100km north of here. That kind of thing is not uncommon in this part of the world, like several times a season rather than 1 in 100 years or more.
To sum up, I think the engineers did pretty well. Indeed the Commission seemed to accept Babister’s opinion that the “flood mitigation effect was ‘very close’ to the maximum achievable within the constraints of the manual.” Thomas’s comment about “mismanagement” is I think risible.
I’ve addressed the issue of whether the engineers could have done better first so that you could see that any non-compliance was of little consequence to flood mitigation outcomes. I think the underlying problem of how the engineers operated in relation to the manual turns on the nature of the W2 strategy as formulated in the 7th revision of the manual, and the confusion it caused. I’ll take a look at that in the final post, which addresses how the engineers fell foul of the Commission.
It should be noted that the Commission’s standard of proof was on the balance of probabilities, not beyond reasonable doubt.