11 April 2011

Blind Spots in Australian Flood Policies

John McAneney of Risk Frontiers at Macquarie University in Sydney identifies some opportunities for better flood policies in Australia. First, he explains that better management of flood risks in Australia will depend up better data on flood risk.  However, collecting such data has proven problematic:
As many Queenslanders affected by January’s floods are realising, riverine flood damage is commonly excluded from household insurance policies.

And this is unlikely to change until councils – especially in Queensland – stop dragging their feet and actively assist in developing comprehensive data insurance companies can use.

Why? Because there is often little available information that would allow an insurer to adequately price this flood risk.

Without this, there is little economic incentive for insurers to accept this risk. It would be irresponsible for insurers to cover riverine flood without quantifying and pricing the risk accordingly.

The first step in establishing risk-adjusted premiums is to know the likelihood of the depth of flooding at each address. This information has to be address-specific because the severity of flooding can vary widely over small distances, for example, from one side of a road to the other.

Risk Frontiers is involved in jointly developing the National Flood Information Database (NFID) for the Insurance Council of Australia with Willis Re, a reinsurance broking intermediary. NFID is a five year project aiming to integrate flood information from all city councils in a consistent insurance-relevant form.

The aim of NFID is to help insurers understand and quantify their risk. Unfortunately, obtaining the base data for NFID from some local councils is difficult and sometimes impossible despite the support of all state governments for the development of NFID.

Councils have an obligation to assess their flood risk and to establish rules for safe land development. However, many are antipathetic to the idea of insurance.

Some states and councils have been very supportive – in New South Wales and Victoria, particularly. Some states have a central repository – a library of all flood studies and digital terrain models (digital elevation data).
Council reluctance to release data is most prevalent in Queensland, where, unfortunately, no central repository exists.

A litany of reasons is given for withholding data. At times it seems that refusal stems from a view that insurance is innately evil. This is ironic in view of the gratuitous advice sometimes offered by politicians and commentators in the aftermath of extreme events, exhorting insurers to pay claims even when no legal liability exists and riverine flood is explicitly excluded from policies.
Second, models of flood risk are sometimes misused:
Another issue is that many councils only undertake flood modelling in order to create a single design flood level, usually the so-called one-in-100 year flood. (For reasons given later, a better term is the flood with an 1% annual likelihood of being exceeded.)

Inundation maps showing the extent of the flood with a 1% annual likelihood of exceedance are increasingly common on council websites, even in Queensland. Unfortunately these maps say little about the depth of water at an address or, importantly, how depth varies for less probable floods. Insurance claims usually begin when the ground is flooded and increase rapidly as water rises above the floor level.

At Windsor in NSW, for example, the difference in the water depth between the flood with a 1% annual chance of exceedance and the maximum possible flood is nine metres.

In other catchments this difference may be as small as ten centimetres. The risk of damage is quite different in both cases and an insurer needs this information if they are to provide coverage in these areas.

The ‘one-in-100 year flood’ term is misleading. To many it is something that happens regularly once every 100 years — with the reliability of a bus timetable. It is still possible, though unlikely, that a flood of similar magnitude or even greater flood could happen twice in one year or three times in successive years.

The calculations underpinning this are not straightforward but the probability that an address exposed to a 1-in-100 year flood will experience such an event or greater over the lifetime of the house – 50 years say – is around 40%. Over the lifetime of a typical home mortgage – 25 years – the probability of occurrence is 22%. These are not good odds.
More on Risk Frontiers at Macquarie University here.


  1. It is still possible, though unlikely, that a flood of similar magnitude or even greater flood could happen twice in one year or three times in successive years..

    My understanding of the statistics of most climate data, and particularly that of precipitation and river flow, is that the probability of another flood is increased for several years following a flood. Hurst in his study of the several century long record of Nile river flows detected this long term persistence.

  2. Isn't this just another example of the shockingly naive linear model that you are usually (rightly) ridiculing? Like, they need actual scientific understanding before formulating their pricing policy? Some mistake surely.

  3. -2-James Annan

    Smarmy to the end, eh? ;-) I do not believe that you've actually mastered the material.

    There is a big difference between science informing decisions (not the linear model) and science determining decisions (the linear model). McAneney's commentary is about the importance of science to inform decisions (ie, data needed to calculate risk), and as well a push back against the idea that science that determines decisions (ie, the 1% threshold). Good stuff.

    Thanks for visiting.

  4. Blind spots?
    More like delusions based on obsessions over CO2.

  5. Oh, Roger, you are too kind in your praise. But certainly, I don't claim to understand this at all, hence my comment.

    In particular, the separation between "informing" vs "determining" seems like a semantic game, a distinction without a difference. In reality the influence of science is always going to be somewhat fuzzy, no? It can only ever be one component in the mix. So in that case, is there any a priori method for discriminating between the "right" approach (informing) and "wrong" (determining), or is this just a post-hoc rationalisation adjusted to suit the rhetorical purposes?

  6. -5-James Annan

    In The Honest Broker I discuss this at length. I suggest two criteria that can be useful in helping to think about the role of science (and expertise generally) in decision making. In short, the role of science is a function of context.

    In situations where uncertainties are known (or knowable) and values are generally agreed upon, science is more likely to be useful and effective in dictating decisions. The canonical example I use is that of decision making in the context of an approaching tornado.

    In situations where uncertainties are unknown, (or unknowable or contested) or values are in dispute, science is much less likely to be useful and effective in dictating decisions. Here I use decision making about abortion as an example.

    I argue that "abortion politics" and "tornado politics" imply very different approaches to science in decision making.

    Additional complexities include how you'd like to view the role of the expert in a democracy, and in particular, if you seek to reduce or expand the scope of available choice.

    There is much more to say of course (and much more than what I cover in my book), but perhaps this gives you a sense.


  7. Roger,

    This seems to be a bit of a change of topic. You started out with "inform" vs "determine" which is what I asked about, but you are now talking about whether the science and values are widely agreed or disputed. I don't see these dichotomies as in any way isomorphic.

  8. -7-James Annan

    If you and I agree that we will evacuate if a tornado gets within 1 km of our building, then the information of (science) tornado location DETERMINES our decision.

    If you and I agree that we need to develop a policy governing abortions, scientific information INFORMS (at best) that decision, the information does not determine what we do.

    Why does science "determine" in the first case and "inform" in the second? The answer I suggest has to do with the context of decision making, and specifically the degree of values agreement and the nature of the relevant scientific information.

    Ask again if unclear, thanks.

  9. Roger, if you and I agree that the threshold for abortions should be independent viability, then science determines that decision too - although perhaps one of us would try to manufacture fake scientific uncertainty as an easier alternative to arguing a real (hidden) values difference. This is hardly the fault of the well-intentioned and honest scientist who attempts to answer the question as stated, even if you lambast them for their naivety in assuming a "linear" model.

  10. -9-James Annan

    Sure, if we agree that scientific information will determine a decision then we are in the realm of tornado politics. In THB I discuss the frequent desire to turn cases of abortion politics into tornado politics. In those cases where abortion politics does not yield to such efforts, then that is one way that science becomes pathologically politicized. One need not introduce fake scientific certainty for this to happen.

    Sorry if you are feeling lambasted here, but I am not sure why. Thanks.