Search This Blog

Saturday, March 19, 2011

Mick O'Brien and the Brisbane Flood of 2011

Chemical engineer Mick O'Brien has released his submission to the Queensland Floods Commission of Inquiry. Mick O'Brien has been the chief informant of the Australian's Hedley Thomas, so unsurprisingly, the submission is just a rehash of many of the inaccurate and unwarranted claims that Hedley Thomas has being making over the last few months.  However, there are a number of points that Mick O'Brien raises that I wish to address, and will do so briefly here.  In later posts, I will give a more detailed survey of information now becoming available as part of the preparation of my own submission to the inquiry.

Mick O'Brien defines a "simple operating philosophy" which he believes should govern Dam operations:

"From this would flow the following simple operating philosophies: - 
• The flood storage should generally be kept empty by releasing all the water flowing into the flood storage system until the defined downstream flood event is reached. 
• Releases from the storage system should then be selected to remain below the defined downstream flood event until the flood event has passed or the storage system becomes full. 
• If the flood storage system becomes full, the storage system must then release all the inflows but releases should never exceed inflows. "

Clearly this operating philosophy is to simplistic for real life.  On the Brisbane River between Wivenhoe and Brisbane there are several locations effected by flooding, and each is effected by flooding at different levels.  Mick O'Brien's operating philosophy would require us to pick just one of these, presumably flood levels in Brisbane, and ignore all the others.  Such a strategy would result in unacceptable disruption of rural life downstream of Wivenhoe from even minor flooding events.  Floods that cause no significant risk to Brisbane could not be mitigated for the benefit of users of rural bridges because the operators would be required to release "all water flowing into the flood storage".

Further, by modifying the strategy based on current flood levels downstream, rather than projected levels, the philosophy would result in modification to release strategies being reactive and delayed. If flood levels in Brisbane represent the "defined downstream flood event", then releases in Wivenhoe would only be changed from a strategy or "releasing all the water flowing into the flood compartment" 24 to 48 hours after at least minor flooding occurred in Brisbane.  It may fairly be argued that I am being unduly literal here.  In practice an attempt to implement O'Brien's philosophy would surely incorporate modelling of Wivenhoe releases and other downstream flows so that changes in release strategies could be undertaken in a timely manner.  However, it is not clear how O'Brien's "simple philosophy" would then differ from the actual strategies used by Wivenhoe operators as laid down in the Dam manuals.  In fact, much of O'Brien's criticism throughout his report is based on assuming that his simple philosophy could in fact be followed without considerations of  the practical matters about timing of releases, and intermediate flooding levels that shape the actual operating strategies at Wivenhoe.  Had he developed a practical operating procedure rather than a simple philosophy, his criticism would largely come down to the choice his choice that Wivenhoe operators should ignore the effects of their releases on rural life, caring only for O'Brien's much loved Brisbane.

Finally, O'Brien ignores in his third point the fact that peak inflows into Wivenhoe can exceed any possible release level.  Indeed, for a Maxium Flood Event, the peak inflows excede maximum outflows even with all fuse plugs triggered and the gates opened to the maximum possible extent.  Because of this, any strategy that does not always leave some reserve flood storage capacity in the dam as a buffer for those peak inflows risks the dam being over topped, and destroyed.  Ignoring this risk is the sole basis of O'Brien's criticisms relating to unused flood storage capacity and the fuse plugs.

In essence, O'Brien's simple philosophy comes down to this:

1)  There is no level of risk, no matter how small that can justify mitigating the effects of minor floods on rural areas; and

2) In mitigating urban floods, no level of risk is to large in retaining water till the last possible minute.

Rather than a simple philosophy, O'Brien has produced a simplistic philosophy.

Mick O'Brien also criticizes the delay in increasing releases from Wivenhoe:

We know from the Flood Event Log Appendix M (3), that SEQWater were indeed aware as early as 19:10 Sunday 9th  January that substantially increased discharge rates of 3000 cubic metres per sec were required. 
Despite this knowledge, the rapid escalation through the Strategies for Wivenhoe, the forecasts and real time data available, SEQWater did not increase the release rate to 3000 cubic metres per sec until after 09:00 Tuesday 11th  January."

At this point it is worthwhile looking again at the inflows and releases from Wivenhoe:

If you look at that graph, you will see the vary sharp rise in inflows on Sunday the 9th.  Look closely and you will see that the rate of increase in the inflow dropped sharply at about 6:00 pm of Sunday the 9th, and peaked within a few hours.  You might also recognize that that peak coincided with the flash flooding in the Lockyer Valley, which by itself would have been sufficient to cause minor flooding in Brisbane.  In other words, you would recognize that very shortly after the notation in the log, the situation changed, both as to the releases that were apparently needed, and as to the risks to Brisbane of even very low release levels.  Indeed, had it not been for the second peak in inflows, dam levels would have peaked below 74 meters, and the operators decision to delay releases until well after the Lockyer Creek peak had passed would have met with nothing but praise.

O'Brien does not consider the probability of rainfall events in his report, but when judging the reasonableness or otherwise of the operators actions, the probability of the events that unfolded is an essential issue.  Had the second peak been a near certainty, condemnation of the operators would be universal.  But it was not.  It is associated with a rainfall event with an Annual Excedance Probability greater than 1 in 2000.  O'Brien is defacto arguing that the operators operated unreasonably because they did not predict a 1 in 2000 year rainfall event within two days of the preceding 1 in 2000 year event.

O'Brien also disputes the quantity of rainfall involved in the second peak:

"The period from 9:00 to 16:00 Tuesday 11 th  contains the three highest calculated hourly inflow rates and 8 of the 12 highest calculated inflow rates for the whole period reported in the Flood Event Report. The average inflow over this period is 9132.5 cubic metres per sec which is just slightly lower than the highest hourly rate of 9174 calculated for all other hourly periods and higher than 8820 cubic metres per sec which is the second highest calculated hourly rate for all other periods. 
Refer to Section 8.9 of the Flood Event Report (3) for further discussion on rainfall modelling necessary to generate such high inflow rates to Wivenhoe. However to replicate the rate of rise in the dam level measured using the manual gauge board it was necessary to impute an unmeasured rainfall event that was twice the recorded rainfall of the Mt Glorious measurement station which already had the highest recorded 10 hour rainfalls. SEQWater state: - 
'To model the rapid rise of the recorded Wivenhoe Dam levels between 03:00 to 15:00 on Tuesday 11 January 2011, the Mt Glorious rainfall data was repositioned to the ungauged area immediately upstream of the Dam, where the BoM radar indicated was the centre of the heavy rainfall during that period. It was then necessary to scale this rainfall up by a factor of two to match the rapid lake level rises. This factored Mt Glorious rainfall data had an average intensity of 68mm/hr, which exceeds an annual recurrence interval of 1 in 2,000 years and may be well into the extreme category. Rainfall of this intensity and duration over the Wivenhoe Dam lake area at such a critical stage of a Flood Event was unprecedented. The resulting runoff could not be contained without transition to Strategy W4, as discussed in Section 2 and Section 10.' 
There are two possible explanations: - 
• That this period of 8 hours was the period of highest rainfall in the Wivenhoe catchment, or 
• The manual gauge board was reading high and the dam level was not increasing at the rate 
shown by the gauge board. In this case the calculated inflow rates would then be lower than the currently estimated values."

There are in fact two possible explanations.  Because the total outflows of the dam are known over the whole period of the floods up until Full Supply Level was restored, the total inflows are also known.  There is no discrepancy on SEQWater figures.  This means that if the unusually large inflows did not occur at the time that SEQWater calculates them to have occurred, then they occurred a few hours later when the automatic gauge was rising sharply, while the manual gauge was fairly steady.  In other words, the extraordinary rainfalls occured regardless of which gauge we consider accurate.  The only difference is a question of timing.  On that basis, it is far more parsimonious to accept SEQWater's explanation and assume the timing is best indicated by the manual gauge.

What is more, even if that were not true, if the operators instruments indicated the dam was at risk, as they did, then the most reasonable thing for them to do is to act on that basis.  Doing otherwise would be to gamble the safety of the dam on whether the instruments were faulty or not.  So, even if the manual gauge was faulty, the operators acted reasonably.  (A major problem I have with O'Brien's report is his failure to think through even simple matters such as this on the basis of what was reasonable for the operators given the information they had at their disposal.  He at all times judges everything with the full knowledge he has before him, therefore imputing an assumption of omniscience into his definition of reasonable action.  That alone renders his report of little merit.)

These are not the only problems I have with O'Brien's report.  He makes a number of less important errors as well.  Of the other errors, the largest is failing to consider changes in river conditions in calculating the relationship between flow rates and river levels at the City Gauge.  However, these are, I believe, the crucial errors that render his analysis void as an attempt to understand the Brisbane flood of 2011  


  1. This comment has been removed by the author.

  2. Tom thanks for your careful consideration of my report, I was not aware that it would generate so much interest. I provide the following additional information that may assist in your understanding of my thought process and would be happy to discuss it further with you.
    You will already have my email address from my report and I would be happy to discuss any issue with you via email.
    I have also added some comments to your blog on the Insurance Council Report.
    For my “simple” operating philosophy there is nothing to stop a dam operator utilising forecast dam levels to improve the operation. I also agree with your conclusion that a multitude of different downstream “flood events” probably detracts attention from the main game. I think that I have said before, it is like driving headlong for a cliff and worrying about the potholes in the road. There has been a lot of other discussion about the lack of attention to the issues in the Brisbane Valley and I don’t need to add to these.
    I am very interested in your point that the dam/s may not be capable of handling the Maximum Flood Event without significant spare capacity and would be very interested in any additional data that you could give me regarding this issue.
    The devastating flooding in the Lockyer occurred on Monday not Sunday as you have suggested.

    Michael O'Brien

  3. Michael O'Brien, I am sorry for the delayed posting of your comment. For some reason it was diverted to the spam folder, and I have only just caught it.

    With regard to the "simple operating philosophy", the analogy of a cliff is inappropriate in my opinion, for it suggests a known peril. In contrast, in real life the operators did not know of the cliff until the second peak of inflows into Wivenhoe. Had that extraordinary event not occured, ie, had the rainfall followed BoM predictions, the dam was "approaching" not a cliff, but a slight incline that would have brought it to a "gentle stop".

    To put this in context, given the dam's capacity as of Monday the 10th, modelling has shown it would take a greater than 1 in 100 AEP flood event to force the operators to release water at a level that would threaten Brisbane. It would take a greater than 1 in 500 AEP flood event to place the dam at risk. Having just handled an at least AEP flood event on that day, it is unsurprising that the operators did not expect an unpredicted AEP 200 + event on the next day.

    In other words, the operators followed a strategy that would only have required them to increase flood waters in Brisbane above major flood levels just once in every 200 years on average. Given the limited loss of life, and even with the large property damage, that would seem a reasonable risk. Unfortunately it is a risk, and in this case, with a greater than 1 in 2000 AEP rainfall event onto the surface of the lake, it was a risk that caught them short; but that was because nature was unkind, not because they were unwise.

    I have discussed these risks in my blog post on "The Value of Hindsight".

    With regard to the Lockyer Valley flooding, although it occurred on the Monday, it is associated with a system of storms whose impact on Wivenhoe Dam levels were first felt late on the Sunday, as shown in the graph of inflows from the SEQWater report.