Search This Blog

Friday, February 4, 2011

The value of hindsight, or how to say "I told you so" without ever having done so

On the 9th of March, 2010, the opposition (Liberal National Party) called for a reduction of the flood mitigation capacity of Wivenhoe dam to enhance Brisbane's water supply.  Not content with a review of options, Mr Seeney (LNP Shadow Minister for Natural Resources) demanded that extra water be retained until the completion of any review.  "It would be absurd", he said, "to release water from Wivenhoe until all options are thoroughly investigated."   The LNP renewed its calls after Wivenhoe dissipated a flood event in early October of 2010.   Then Lawrence Springborg renewed the call on December 20th, in the middle of yet another flood event, from which Wivenhoe again protected Brisbane.

These calls represented the ideal time for anyone to argue for enhanced flood mitigation at Wivenhoe.  Those such as Andrew Dragun, who now believes that water should have been released to prevent the risk of flooding, said nothing.  He thought it was far more important to attack the Federal Governments Murray Darling Basin Scheme.  Hedley Thomas, who has run so hard on this story for the Australian, ignored the story.  And Michael O'Brien, the engineer who has so publicly condemned the dam operators was not to be heard. Their wisdom, it appears, is exclusively 20-20 hindsight.

Let's put this in context.




Late on Sunday the 9th, the engineers at Wivenhoe dam had no reason to suspect what was coming.  They had already passed through three flood events (13th October, 21st December, and 29th December) without difficulty.  Having lowered the dam levels to near 100% of the Full Supply Level (67 meters) after the last flood event, waters were on the rise again.  They had increased flows over the spillway to a level that flooded all bar two of the cross river crossings above the Centenary Bridge to compensate, but had no reason to suspect what was coming.  The weather forecast was for heavy rain, but they were experiencing heavy rain, and had not reason to suspect heavier.

(Wivenhoe Dam levels from 22nd Sept 2010 to 3 Feb 2011)

The charge against them is that during this period they should have released more water.  They should have flooded all the lower middle river crossings (between Wivenhoe and Jindalee) and low lying areas of Brisbane in anticipation of what was to come.  How realistic is that?

According to GHD's report on the auxiliary spillway, a dam level of 71 meters represents 1,647 gigaliters of water (table 3.15.2).  That is 141.4% of normal capacity.  According to SEQ Water's report, on 9:00 am of Monday, the dam was at 148.4% of capacity, which is less than 71.25 meters of water.  I consider 71 meters to be a reasonable approximation of the water levels of the dam at midnight of Sunday the 9th.


(Wivenhoe Dam storage by level, purple line is the Full Supply Level, Green line is 71 meters, while the Red line represents dam levels  as at 9:00 am Monday 10th of January, 2011.)

That is interesting because SEQ Water's report on different methods to increase Wivenhoe's storage capacity considers the effects of increasing the supply level to 71 meters with no other adjustments.

In particular, the consider the impact of a 1 in 200 rainfall event on the probability of flooding.  A 1 in 200 48 hour rainfall event, according to this study, would result in peak inflows of 8,433 cumecs.  It would necessitate peak releases of 6,037 cumecs.  And it would result in dam levels rising to 74.82 meters.


In fact, over the 48 hours following Midnight of Sunday 9th, peak outflows reached 7,500 cumecs, average inflows over the 48 hours exceded 6,000 cumecs, and peak inflows certainly topped 8,000 cumecs (and may have topped 10,000 cumecs), and dam levels rose to 74.85 meters.

For comparison, a 1 in 500 48 hour rainfall event is expected to result in peak inflows of 10,543 cumecs, peak outflows of 7,649 cumecs, and a peak level of 75.664 meters (ie, 6 mm below the level of the first fuse plug).

Very clearly the events at Wivehoe lie somewhere between these two.  They were significantly worse than a 1 in 200 year event, but not quite as bad as a 1 in 500 year event.

So when push comes to shove, what the engineers at Wivenhoe "did wrong" is that on the weekend preceding the flood, they did not anticipate an unpredicted 1 in 200 year (at minimum) event.

I am certainly open to suggestions that the operating manual needs to be reviewed in light of experience.  With hindsight, it may have been better to allow the fist fuse plug to be overtopped and keep peak dishcharge below 6,000 cumecs.  Or not.  That is something only a full review can tell.

But suggestions that the engineers did the wrong thing, given what they knew are simply false.  They are needless vilification of people who under trying circumstances saved Brisbane from a much larger disaster.

And what of the suggestions that Wivenhoe's dam level should be reduced before hand in anticipation of the flood?  Had that advice been carried out during the last significant La Nina, Brisbane's water supply would have dropped to  5% of capacity in the recent drought.  Had it been followed the last time the dam was full, it would have run dry a year before the end of the recent drought.  I doubt Andrew Dagun or The Australian will indemnify the people of Brisbane against the risk of a similarly long drought following this La Nina.

Update  (9:30 am, 8/2/11)

I have just come across this BOM graph of river levels at the Wivenhoe dam wall from 12 PM, Sunday 9/1/11 to approximately 9:00 AM, Friday 14/1/11:

 (Graph recovered from Ryan Independant,  Ryan being one of the federal electorates in Brisbane.)

17 comments:

  1. TC - is there anyway to calculate intake and outflow levels at which they would have lost the ability to defend the dam wall? Or was a runaway situation impossible?

    If they allowed the fuse plugs to go, there would be no regulation of outflows possible until the level fell below that spillway, right? Seems like that could be a "tsunami" of water.

    Scroll to the bottom of the first link and there are pictures of unmanaged dams that were overtopped. One gave way. USA flood records categorize it as extreme rainfall, which it was, but at the time the news articles placed a lot of blame on the Canyon Lake Dam's failure.

    Rapid City flood, 1972:

    http://sd.water.usgs.gov/projects/1972flood/photos.html

    http://www.rapidcitylibrary.org/lib_info/1972flood/inmemory.asp

    ReplyDelete
  2. JCH, there are five spillway gates, each 12 meters wide, with the crest of the spillway having an elevation of EL 57 meters

    The fuse plugs are divided into three bays. The first bay triggers at EL 75.7 meters, and is 34 meters wide. The second triggers at EL 76.2 meters, and is 64 meters wide. The third triggers at 76.7 meters, and is 65 meters wide. The Ogee crest on which the fuee plugs sit has an elevation of EL 67 meters (the Full Supply Level).

    If the dam was sufficiently full to trigger the first fuse plug, but not quite high enough to trigger the second ie, 76.1 meters, then a cross sectional area of 309.4 square meters for water to flow through. Each spillway gate opened to maximum capacity has a cross sectional area of 229.2 square meters. Therefore, by closing the spillway gates by 27%, outflow from the spillway could be reduced to compensate for outflow from the auxilliary spillway.

    If the second fuse plug is triggered, but not quite the third (water level = EL 76.6 meters) the cross sectional area for the two bays becomes 940.8 square meters. This can be compensated for by closing the spillway gates by about 80%.

    If the third fuse plug is over topped, but water levels are still below that of a proposed fourth fuse plug in a second auxiliary spillway (water level = EL 78.2 meters, the combined cross sectional area becomes 1825.6 square meters. In this case the total flow is at minimum 140% of that with the spillway gates full open, and if the spillway gates are opened, can be 240% of that flow.

    If the first fuse plug went, it would certainly be possible to regulate, but not eliminate the discharge, and it could probably could be kept under the peak discharge during the 2011 Brisbane flood. Assuming the water was flowing at around 40 km/hour, the gates were only about three quarters open for that event. On that basis, with only one fuse plug initiated, flow rates could be kept as low as 3,500 cumecs. With two fuse plugs initiated, flows would have exceded 10,000 cumecs excluding the main spillway. With three, the flow becomes 20,000 cumecs excluding the main gates.

    The maximum possible flow is 37,400 cumecs (from which I calculated the approximate water velocity).

    ReplyDelete
  3. I'm just trying to get my head around the logic for the big release. The newspapers indicate 225% total capacity, but they pushed the big button at ~190%. All I can imagine is either the rate of change upward in the water level somehow forced that decision or the conditions downstream, or a combination of both. If it had been a slower rise, my bet is they would have gone higher than ~190%, and possibly used the fuse plugs tactically as you have laid out.


    You did not comment on the Rapid City flood. You can imagine my shock when I turned on the car radio the next morning and heard a deejay tell the numbers of dead, missing and injured, and stories of new cars stacked up like toys. I thought it was a radio-show hoax. Probably drove a kilometer, this was before kilometers existed in the United States, before it sank in he was serious.

    ReplyDelete
  4. 1:200 sounds plausible ,Tom.

    Pure speculation warning:I'd suggest that the Tuesday evening forecast for heavy rain to continue past dawn helped prompt the bigger release late on Tuesday,given how rapidly the dam had been filling. Any 6 hour or 12 hour continuation of rain rates similar to what had just been experienced would have easily put the dam past 200%,and any decision to try and equalise inflow and outflow may have meant bigger peak rates.

    ReplyDelete
  5. With regard to Rapid City, you may be interested in this account of the flood. According to it, the peak flow at Rapid City was 50,000 cubic feet per second, or approximately 1,500 cubic meters per second. The very high death toll was due to the fact that the flood happened with no warning at midnight, so nobody got out of the way. From what I can tell, the events at Rapid City resemble events in the Lockyer valley far more than those in Brisbane, the difference being that Rapid City was a reasonable sized town, while the towns in the Lockyer valley are very small.

    If Wivenhoe were overtopped, the death toll would probably be quite small. That is because it would take around 24 hours for the flood peak, which probably would resemble a tsunami, to reach Brisbane. As a result, almost everyone in danger would be evacuated should the dam break. On the other hand, the heart of the city would be flattened.

    As to the strategy, as the dam level approaches the first fuse plug, maximum releases through the spillway are approximately 12,000 cumecs. A net inflow of 12,000 cumecs will take the dam from the level of the first fuse plug to overflowing in less than six hours. The estimated peak flow for a 1 in 100,000 year event (the greatest flood the dam is designed to survive) is 42,600 cumecs, at which rate the dam will go from first fuse plug to overflow in around 2 hours if only the spillway drains, and in around 12 hours with full drainage from all spillways including all three bays of the auxiliary spillway.

    With that background, the current operating manual requires by law that if the dam level goes over 74 meters, the dam operator is to do everything they can to prevent the first fuse plug from being over topped, or if that cannot be prevented, to minimize down stream flooding by allowing for the extra release through bay one. Evidently the feeling is that if inflows excede 12,000 cumecs for long enough to over top bay one, then there is way to little safety margin for preventing bay two from being triggered, with subsequent almost uncontrollable flooding.

    It is also felt that any flood that would trigger bay three leaves too little safety margin against dam failure if only the primary spillway was used for releases.

    I suspect the first assessment is correct, but a case could be argued that allowing bay 1 (only) to trigger might be a sensible option if events similar to those in January repeated. I am certain the second assessment is correct.

    ReplyDelete
  6. Nick, I'm not sure that the modeled estimates truly capture the probability of this event. I suspect they are based on the assumption of steadier rainfall than was in fact experienced. However, on the existing modeling I would say the event was closer to a 1 in 300 or 1 in 400 year event.

    As to your speculation, I am almost certain of it.

    ReplyDelete
  7. Pure speculation warning part 2:

    Tom,did you see the three aerial shots of the Wivenhoe spillway,flip bucket and channel,post buffer drawdown, in the Courier-Mail a week or so back?

    They show a lot of damage to the substrate,with very large fresh sandstone blocks scoured out of the pool and banks,and piled across the discharge channel just downstream of the lower obs deck. This indicates what those 6,000 + cumecs releases were capable of. Contemplating releases at 8 or 10,000 cumecs is scary. The manual mentions the imperative of methodical sequential gate operation with the aim of minmising turbulence,and possible erosion. If the channel becomes choked with scoured rubble and forces the pool levels to rise,who knows what this may mean to the structure?

    The 7465 cumec transient was the highest ever release through the spillway,so comparisons with the modelling that underlies manual assumptions will be very informative.

    ReplyDelete
  8. Nick, I did see the picture.

    I'm not sure there would be a risk to the wall from greater outflows resulting in debris blocking the channel. That is because the wall at the point of the spillway is entirely clad in concrete, protecting the embankment; and also because the channel would overflow before it rose above the wall. Further, the channel is quite deep, and as the water backed up, it would provide a buffer against further erosion.

    On the other hand, it is probably a good thing investigate in a post flood review.

    I absolutely agree about the modelling. This is the first time Wivenhoe has really been tested, so it is an ideal opportunity to check out the performance of meteorological and hydrological models when put to the test. I am very much looking forward to the results of such comparisons when they become public.

    I am a bit apprehensive about the forthcoming judicial inquiry. Some judicial inquiries have performed brilliantly (Fitzgerald and the inquiry into Victoria's bush fires come to mind) but many others have had their outcomes determined by political considerations.

    ReplyDelete
  9. The fuse plugs make the news:

    http://www.theaustralian.com.au/national-affairs/dam-storage-capacity-was-compromised/story-fn59niix-1226001786766

    ReplyDelete
  10. Hedley at The Oz is not sure what this information means,JCH. The actual flood storage volume to 75.7m,the level of the lowest fuse plug,is 1182GL,but this does not mean this is the limit of flood storage. In extreme floods there are extreme inflows[!] and when one fuse plug is triggered,the dam can keep rising if inflow outweighs outflow,until the next fuse plug [76.2m] and then the next [76.7m] are triggered. These flood outflows through the fuse plugs are 'uncontrolled' in a sense,but they are through designed,measured 'apertures' so their flow rates at any level are calculable. The dam can keep rising to 77m and get to that theoretical 1450GL capacity that Hedley thinks has been compromised,or more if inflows are sufficiently high.

    I think what is not generally appreciated is that the 1450GL point is not reached without big releases going on simultaneously. I think it is likely that people think the dam does not have to pass a drop until 1450GL is reached.

    The dam was upgraded to pass higher maximum floods in extreme scenarios. An auxilliary spillway was built in the right abutment [looking downstream]. This can be seen on Google Earth to the left of the dam wall with orientation north. An obliging poster has placed an informative photo captioned "The levy that was made to fail" just to the left of the aux spillway on the roadway west of the overbridge. This excellent photo shows the three fuse plugs sitting across the spillway,above a concrete wall,an ogee crest at 67m. The three plugs are separated by concrete dividers,and the middle,and narrowest plug is the lowest. If the dam tops this at 75.7m it crumbles,leading to a possible maximum erosion of the plug down to the ogee crest. So the dam will empty down to 100% without control,but this is not likely to happen because the gated spillway has the capacity to pass,under control, higher flows than it did recently.

    ReplyDelete
  11. JCH, I don't have a lot to add to Nick's comments. The additional information I have is already included in the second and fifth comments above. The essential points:

    a) The auxilliary spillway was built because the BOM predicted higher maximum possible rainfalls as a result of global warming, rendering the existing design potentially unsafe;

    b) At the maximum flood level the dam is designed to cope with under Australian Standards, it has expected inflows of 42,600 cumecs (153 gigaliters per hour);

    c) At that rate of inflow, with the auxiliary spillway in operation, it takes 15 hours to go from the level of the first fuse plug to over flowing the dam wall, with catastrophic failure; and

    d) If there were no auxiliary spillway, it would take 2.5 hours.

    To my mind, the occasional risk of a 1974 style flood that exists with the current design is much preferable to the risk of 5 sydney harbours worth of water being released in a couple of hours with the dam failing. Not living downstream of Wivenhoe, Hedley Thomas may take a different view.

    The only other pertinent facts to note are that contrary to Hedly Thomas's introduction, these facts are not hidden away, but are well known and easily accessible. Probably the easiest place to get full details are the two reports linked in the main post above (although the workd is discussed in GHD's in house magazine, in the Queensland Engineering trade magazine, and multiple other places.) That Hedley Thomas could only find out about it in the manual simply demonstrates that he is not bothering to do any background research, ie, he is speaking in ignorance.

    Also, the auxiliary spillway is only the first part of the upgrade. A second auxiliary spill way was going to be installed after completion of the first. Work on that has been held of for reasons not stated by SEQ Water, but probably because consideration was being given to major design work on Wivenhoe which may have included raising the Dam Wall, the height of the main spillway, and the height of the ogee crest on the auxiliary spillways. It makes little sense to install the second auxiliary spillway, then immediately raise is height.

    ReplyDelete
  12. I realize the flood-mitigation level of 1.45 million ML is a dynamic concept, but what height corresponds to 1.45 million ML? 76.7m?

    I read the manual when TC posted the link. Somehow is sticks in my head that one of the rules is they are never to allow water to go over the tops of the tainter gates (which I believe is 73m.) Is that right? I'm certain that I read that, but I cannot find it tonight. I think that is a relevant factor in their decisions. It would mean they cannot fully close the tainter gates until the level falls below 73m, so there could be a period of less than optimal offset of the auxiliary spillway release.

    Do you guys agree they were in a W4B scenario, page 30 of the manual?

    ReplyDelete
  13. I believe EL 77 meters corresponds to a full flood compartment.

    You are completely correct about the Tainter gates. A flood that initiated the first fuse plug would require the gates to be kept open by approximately 3 meters, which represents a flow of approximately 2,200 cumecs. Therefore the minimum flow with one fuse plug triggered would be closer to 5,700 cumecs, about 2,000 cumecs less than was actually released in order to avoid triggering the fuse plug. Of course, by not triggering the fuse plug the operators were able to rapidly scale back that release, something they would not have been able to do once the fuse plug triggered.

    Given the inflows from the Bremer and Lockyer, I doubt allowing the fuse plug to trigger would have dropped more than half a meter of the flood height. It would have resulted in a longer lasting flood, however, and depending on the timing of the peaks, may have resulted in a higher flood at Brisbane. This is where you really need a hydrological model to back up judgement calls.

    Good catch on the gate heights, by the way, and thanks.

    ReplyDelete
  14. The Queensland Floods Commission of Enquiry had held its first hearing. Presumably in a little while it will start releasing some of the data we want to see.

    ReplyDelete
  15. TC - the bullet points are good as far as they go, but this wording:

    implementation of systems operation plans for dams, and

    makes me nervous.

    ReplyDelete
  16. JCH, I assume you are referring to the terms of reference of the inquiry which states (amongst other things):

    "UNDER the provisions of the Commissions of Inquiry Act 1950, Her Excellency the Governor, acting by and with the advice of the Executive Council, hereby appoints the Honourable Justice Catherine Holmes to make full and careful inquiry in an open and independent manner with respect to the following matters:-

    ...

    f) implementation of the systems operation plans for dams across the state and in particular the Wivenhoe and Somerset release strategy and an assessment of compliance with, and the suitability of the operational procedures relating to flood mitigation and dam safety,

    ..."

    IMO that is a subject that had to be inquired into. I have doubts that a royal commission is the best place for that inquiry in that it sets up an adverserial framework, and gets lawyers involved (almost always a bad move). Whether the Commission can overcome those disadvantages depends too much on the individual wisdom and integrity of the commissioner and the council assisting the inquiry. History shows that this cannot be relied upon, although in individual inquiries both have performed admirably with far better outcomes as a result than can be achieved by a other forms of inquiry. We can only hope.

    However, whatever the form of inquiry, the operation of the dams needs to be looked at. In fact, the dam operation manual itself requires both a review after every flood (even those prevented by the dam), and every five years.

    In any fair review, I expect it will be found that:

    a) The dam operators operated according to the manual, and also according to most reasonable proceedure given the information they had available at the time of their decisions;

    b) The manual itself represents best practise, or very close to best practise given the constraints on peak river flows imposed by low bridges, and low lying properties downstream; but that

    c) The low lying bridges downstream should be raised to allow larger, earlier releases in large rainfall events (costed at $10 million dollars in 2007); and

    d) The lowest lying properties in Brisbane should be resumed (cost unknown); so that

    e) In future flood events, releases should be allowed to rise to 4,000 cumecs at any time the dam level is above the Full Supply Level to preserve maximum amount of storage in the flood compartment to deal with rare flooding events.

    Adjusting the manual to allow (e) is only reasonable once the capital works and resumptions are completed.

    A further possible recommendation is that the dam wall (including spillway and auxiliary spillway) be raised, the former by 8 meters; and that of the improved capacity, at least half be assigned to the flood compartment. This is not a necessary step except that even greater floods can be expected later this century as a result of global warming. The expected cost of such a modification was $300 million dollars in 2007.

    It may also be found that meteorological warmings need to be improved for the dam operator, possibly by including a doppler radar at the dam wall to better predict rainfalls and inflows. To this end, hydrological models may need to be improved to better model inflows when the soil is completely saturated; and it may be wise to have a staff meteorologist to immediately advise the dam operator (ie, have on board a meteorologist whose only concern is events in the catchment, and who is always available for immediate consultation).

    ReplyDelete
  17. TC - my mistake. I did not see that link. That inspires a great deal more confidence.

    ReplyDelete