30 August 2011

Certain Ignorance versus Uncertain Uncertainty

Writing in the quarterly newsletter of Risk Frontiers at Macquarie University in Sydney, Rob van den Honert has an excellent discussion (here in PDF) of the interim report of the Queensland Flood Inquiry. Some background on the topic can be found in this post from last January.

In his summary van den Honert writes of the decision of the dam operators to ignore weather forecasts of pending rainfall:
[P]redicted reservoir levels depend on expected water inflows into the dam, and that expectation would be based almost entirely upon the rainfall forecast for the catchment area. The Bureau of Meteorology supplies regular 24-hour forecasts of rainfall, and the operators also had access to the Bureau’s weather radar, even though the Bureau cautions that in some circumstances the radar can produce poor estimates, either over- or underestimating actual rainfall. Furthermore, there are far fewer rain gauges in the catchment immediately above the Wivenhoe Dam than in other areas, which means that rainfall in that area was not well recorded.

Thus [the dam operator] Seqwater claim that there were gaps in the information available on which operational decisions had to be made. This is despite Seqwater having the best rain/runoff gauge of all - the dam itself!

A 2001 Seqwater report (Feasibility of Making Pre-releases from SEQWC Reservoirs) concluded that the precipitation forecasts were not sufficiently reliable to form the basis of operational decision making for the dam. Thus this less than perfect available information was given zero weight, and not used at all to help predict reservoir levels. Effectively a “forecast” of zero rainfall was used to inform decisions about water release strategies. In other words, under the circumstances, it seems that the operators chose a scenario guaranteed to be wrong over a forecast that was likely to be uncertain.
The flood disaster arguably was exacerbated by poor decision making under flawed decision processes -- decision makers chose the certainty of ignorance over the uncertain nature of uncertainty judgments.  Indeed, as van den Honert describes the rainfall forecasts were inaccurate, but this did not mean that they would have been without value.

Ultimately, the only way that Queensland gets out of this situation will be to build sufficient water retention capacity to simultaneously meet the conflicting objectives of flood mitigation and water storage as a drought buffer. In other words, there is a technological fix here that can dramatically reduce uncertainties -- but such a strategy will cost money.