Chris de Freitas: Flood prediction fraught with uncertainty

A view of the city high rises after the Brisbane river burst its banks. Photo / Getty Images

The recent floods in Australia, like the droughts that preceded them, are reminders of the power of nature.

Unlike many other natural hazards, their frequency puts them high in importance on the list of natural disasters for most countries.

Clearly, heavy rainfall is implicated in causing a flood disaster.

But vulnerability to flood loss is generated by human action. Many settlements are located near rivers, which also serve as transport routes and sites for protected harbours. Fertile alluvial soils adjacent to waterways attract farming communities.

Common methods for dealing with this vulnerability include one or more of six options - structural protection, compensation, regulation, relocation and education, along with improved long-term flood prediction and shorter-term warnings.

But all are problematic, for a variety of reasons. Structural protection measures such as stopbanks prevent the river in peak flow flooding naturally into depressions which detain water, releasing the flow in the river, thus preventing the river reaching danger levels further downstream.

Flood dams need to be carefully managed, which includes resisting the temptation of having them double as reservoirs for municipal water supply.

Another problem with large-scale investment in and reliance on structural protection measures to ameliorate the impact of flooding is that they can induce a false sense of security.

Both of these considerations are relevant in the case of Brisbane, which had put its faith in the Wivanhoe dam built to protect the city from a recurrence of the great flood of 1974.

The compensation option raises issues related to fairness and freedom of choice. Should a home or business owner who locates on flood-prone land reasonably expect to receive reparation from the Government when the inevitable deluge occurs?

What is an unacceptable risk in one place, where there is no chance of government relief when a flood occurs, may turn out to be quite an acceptable risk at the same place where there is an expectation of financial bailout or government compensation package. Relief aid generally works to ensure the same people and areas will be affected again and again, since it discourages taking action to avoid, reduce or mitigate future events.

Another option is to make flood insurance mandatory with the view that insurance premiums will work with market forces of cost-benefit.

Regulation, relocation and education are often the preferred flood management tools. Where the flood damage risk is clearly high, people can be guided towards adopting appropriate land use practices.

Land could be classified according to flood risk and land use, regulating land use accordingly, such as rezoning commercial and residential land as recreational and replacing buildings with parks and golf courses on flood-prone land and relocating flood-sensitive activities.

Flood predictions can be quite misleading and are often misunderstood. The return period probabilities of floods are statistics that are really human inventions based on conjecture, not factual data.

For example, it is false to assume a 50-year flood will occur or be exceeded once every 50 years. A case in point is flooding during 1998 in Greymouth where a 13-year as well as 36-year flood occurred in the same year.

Instrument records of past climate for most places seldom exceed 50 to 100 years.

In terms of identifying patterns' natural processes, this is a minute fragment of time. Using it as a reference is like composing an annual budget based on one week's expenses.

The biggest problem of all stems from the almost universal expectation that climate of the near future will resemble the recent past and that it is highly probable that past experience will repeat itself.

In other words, decision makers in flood and drought risk assessment operate in a framework where climate is treated a stationary process.

Stationarity is a statistical concept that means the fundamental statistics of climate do not vary and change over timescales of relevance to decision makers.

It assumes that the probability of occurrence of climatic events leading to flood peaks, droughts, annual river flow peaks does not change when shifted in time. Consequently, statistical parameters such as the mean and the variance also do not change over time.

Important questions arise. What if the stationarity assumption has been compromised by climate change? What if it's true that changing climate has affected the global water cycle, as has been claimed by some in the scientific community under the catchy banner "stationarity is dead"?

- Chris de Freitas is an associate professor in the School of Environment at the University of Auckland.