This is the first of a series of posts that will look at aspects of flood risk assessment that are key to a good analysis. In this post, I set the scene of a risk-based approach to flood management.
As engineers, our mindsets often lead us to think of hazard, not risk
Remember the riddle that questions whether a falling tree makes noise when it falls over in a remote forest, with nobody to observe it? Sometimes thinking about flood risk is like trying to answer that riddle.
Flooding can occur where there is no-one to observe it. When this happens, it has very little impact on people or our built environment. In these cases, the flooding goes by unnoticed: It does not impact us.
And this is the crux: We are not concerned when flooding has a negligible impact. In other words, we are not so concerned about flooding per se. The potential impact of flooding and how extensive a flood may be is what is important to us.
So flooding does happen if there is no-one to observe it, but the risk to us of that flood does not.
Many flood studies do not focus much on the impacts of flooding. Yes, characterising flood extent, flood depths and flood velocities is very important! Get this wrong, and the entire analysis loses its value. But, the purpose of flood assessment is to minimise flood impacts and the cost of flood management. If this is the end goal of our flood management, then our analytical studies should focus on this too.
We don’t have a good way to model losses from flood events
We have many complex models for simulating the way floods move across our landscapes. These are very detailed models. In contrast, our modelling of loss is often simple, static, inaccurate, and uncertain.
Modelling flood movement
These models usually divide the land surface into small areas (sometimes no larger than one square metre). Then the models solve complex equations that describe how water flows in each of these small areas. The equations take into account momentum, friction, water pressure, gravity and turbulence.
Because of the complexity, modellers often use supercomputers for this task. The complexity and detail pays off – when modellers spend a large effort on calibrating these models, they are highly accurate.
The way we model losses from flood events is very different.
Models for flood loss are simple, in that the equations used to calculate loss are usually only based on flood depth. These are equations that a year 8 student would understand; you do not need a PhD or physics degree to grasp them.
Yet we know that other factors apart from depth also affect loss. Factors such as:
- the speed with which flood waters are moving
- whether people had warning before the flood came, so they can prepare themselves and their belongings
- how early the warning came before the flood reached them
- the time of day in which the flood arrived (such as whether people were sleeping, at work or home).
We also know that different types of flood events can result in very different impacts.
For example, flash floods, coastal surge, overbank flow and groundwater rise are different types of flood events. These are rarely taken into account using our simple quantification of loss.
Finally, the form of the landscape and built environment can be significant. For example, your flood risk at home depends on
- how close your home is to flow conveyance pathways or flood storage
- how accessible it is to emergency services
- how quickly you can escape from the area.
We need better characterisation of vulnerability
The equations that estimate loss are called fragility curves. They are also called vulnerability or damage curves.
Unfortunately, there are few people developing fragility curves. So, we often take curves developed for a particular region and flood type and apply them to very different contexts. This can result in large errors in impact assessment.
And, before we move onto other topics, I would like to emphasise that simplicity is not necessarily a bad thing. Indeed, it is generally a good thing: The simpler the better! The KISS principle is one of the most important in modelling. Yet, more effort and time needs to be spent understanding what factors are significant in estimating loss, and how these factors change loss.
We can then incorporate these factors into the fragility curves we use, and better characterise the uncertainties/errors in their application. This is important, for the greatest uncertainty we have when calculating risk often comes from the fragility curves used. This will help us contextualise fragility curves for our case studies.
Design based on risk, not hazard
Thinking risk, not hazard has some major advantages for mitigation analysis. It moves us away from ‘blindly’ designing mitigation for a particular ‘design specification’.
While risk-based approaches are now routine, past methods of flood management used public infrastructure, such as dams and levees, to reduce risk. These structures were often designed to protect against a certain hazard level, for instance, a 1 in 100-year flood event.
However, in areas where very large floods result in very severe loss, protecting people at a 1 in 200-year level, say, may be better. Even higher return periods are used in the Netherlands: Some are at the 1 in 1000-year level.
Likewise, in catchments where not much damage results from flood events, protecting at the 1 in 50-year event may be preferred.
The cost of protecting at a higher level is much smaller than the long-term losses from large flood events, which makes protecting at a higher level pay-off. In the later case, the cost of protecting at a higher level is more than the value of what is protected, which does not result in a pay-off for more protection.
An economic approach to flood risk mitigation
Rather than basing protection on some level of hazard, an economic approach is better. In this approach, flood protection is based on the expected benefits and costs of different mitigation portfolios. I will cover more of this in later posts.
It is important to understand the risk preferences of stakeholders in choosing the level of safety for our mitigation. For example, the risk appetite of residents in a flood plain may be very different to other flood plain users such as primary producers, industry or mining bodies.
Where to from here?
A risk mindset is becoming more encultured within flood risk analysis. For example, it is a core concept at the heart of the latest rainfall and runoff guidelines in Australia. Yet, there is much scope for improving risk-based flood analysis’ in practice.
To improve our practice, later posts will consider some key challenges to good characterisation of risk.