HIGH FLYERS THINK TANK

Extreme Natural Hazards

University of Melbourne, Tuesday 30 October 2007

Mr Trevor Dhu
Geoscience Australia

Trevor Dhu has a degree in geophysics and applied mathematics from Adelaide University. He joined Geoscience Australia in 2001 where he led the development of a probabilistic earthquake risk model for Australia. Trevor is interested in developing earthquake ground motion models and risk modelling techniques for tsunami. Since 2007 Trevor has led Geoscience Australia’s Natural Hazard Impacts Project which aims to define the national risk from a range of natural hazards including earthquake, tsunami, severe-wind, landslide, and flood.

Tsunami risk in Australia

I want to give you a flavour of some of our work and some of our understanding of tsunami risk in Australia. But, just as importantly, I want to pull out some of those things that I think are generating a lot of the uncertainties and a lot of the complexities that make it quite difficult to anticipate and to better mitigate and better manage tsunami as a risk.

However, in order to do that I think it is important to have a bit of context. Who are we trying to help with our scientific work?

(Click on image for a larger version)

At Geoscience Australia: we are very focused on informing emergency management. So we are talking about working with state emergency management agencies in Australia and fundamentally they are really asking two questions. They are pretty basic ones, conceptually at least: how big, how often? That's essentially what the emergency services would like to know. What will a tsunami do to an Australian community, and how often will that effect occur? How often will we be confronted with some kind of potential disaster to our communities?

This presentation is predicated on the assumption that you can communicate science effectively to emergency managers – and I am conscious that that is an art and a science in and of itself, effectively translating scientific knowledge to tangible, useful information. I am not really going to go into how we approach that at Geoscience Australia, but I think it is an important thing to acknowledge up front, that just doing this kind of science work and answering science questions isn't enough unless it is translated into tangible outcomes, and tangible things that can be dealt with and considered and inform decision-making.

In terms of modelling tsunami, I think it is fair to say that the modelling technologies – the tools to propagate a tsunami numerically across deep water and up onshore – exist today. As with any sort of model, I am sure they could do with more work, more validation, but generally they are there.

However, there are a lot of fundamental science questions that still need to be answered in order to feed these models, to try and reduce the uncertainty that underlies them and to reduce the uncertainty in the answers and the decisions that we are making. That is really what I want to try and draw out today.

I will break the talk into three sections:

• I will do a quick whip-through of some tsunami fundamentals.

• We generally break up tsunami modelling into two concepts: that of hazard, and then of impact. The reason we deal with hazard is that it is a bit simpler. You can deal with a tsunami model in deep water a little bit more easily than you can when you drive it onshore into shallow water, where it affects our communities. So we will think about it in those two terms, and look at the issues around both of those.

• I will then look at some conclusions.

The very first question is: where do tsunami come from? This is a pretty obvious question.

(Click on image for a larger version)

This slide shows one source, but it is a pretty unlikely one and we are not going to worry about it too much today.

There are actually quite a few sources that can cause tsunami, and you see here a break-up in percentages of different sources. In the Pacific, historically, some are caused by landslides, not a massive percentage, just a bit less than 5. In Australia they are a bit of a concern, because some of our continental shelves, which is where you get build-up of sediment that can slide down and cause a localised tsunami, look as though they have evidence of historic tsunami-generating slides. And they are quite close; they are very difficult to warn for.

Whilst I mention landslides now, in reality this talk is going to focus on earthquake-generated tsunami, which are responsible in the Pacific – and I think in general – for over 80 per cent of catastrophic tsunami that affect communities.

The question becomes: how does an earthquake generate a tsunami? They often occur on a subduction zone. That is where most things happen – where two big tectonic plates actually bash into each other and one slides underneath the other.

Tsunami waves can be huge and this is essentially how the Indian Ocean tsunami of 2004 was generated, and it was generated along a subduction zone that was about 1000 kilometres long – a huge length of zone.

Another consideration is that the wave is from the ocean floor occurs all the way up. So it is not a surface wave, it's not just sitting at the top. It actually is the entire water column, essentially, travelling along. There is massive momentum, massive energy behind these waves.

(Click on image for a larger version)

This map is to give you a sense of where a lot of the earthquakes in our region have occurred in the past. They quite nicely delineate where these quite threatening subduction zones are, and you can see that Australia is essentially surrounded by them. Not all of those areas cause tsunamigenic earthquakes, but quite a lot of them do, and Australia sits right in the middle.

The other question about tsunami is a historical one. The 2004 Indian Ocean tsunami pretty clearly demonstrated how catastrophic these events can be, but there is a very legitimate question: what have they done to Australia in the past?

I think it is fair to say that quite a lot of people probably don't realise that the Indian Ocean tsunami did affect Australia – not as dramatically or tragically as it affected the Indian Ocean countries, but it actually swept about 30 people off the sandbars in Rockingham, near Perth. It was Boxing Day, a lot of boats about, so those people were rescued pretty quickly, but there are actually significant threats. Even that event that wasn't really directed at Australia had some impacts along our coast.

(Click on image for a larger version)

Historically there have been impacts on the Australian coast. Each of the little bars on this map of Australia represents a spot where we have got recorded run-up, where a tsunami has come up onshore. A little bar represents 1 to 2 metres, so we have on the west coast a green bar representing about 4 metres, and further north another green one representing maybe about 6 metres. The pink bar represents about 8 metres, and I will draw your attention to that one. That was a fairly recent impact.

(Click on image for a larger version)

It was caused by the 2006 West Java earthquake, located to the north-west of Australia, and it actually occurred at Steep Point. What is interesting about that is that it draws out a couple of important things about tsunami.

(Click on image for a larger version)

Firstly, we can look at some tide gauge data for that event: this is quite near the event itself, and we see that there is a wave height of about 85 centimetres. So out in deep water, tsunami actually aren't very high. They are quite small waves at the surface, in deep water. Even near Perth there was only about a 30-centimetre wave.

At Steep Point, the wave went to a run-up height of 9 metres. That doesn't mean it was a 9-metre high wave; it means the wave was driven up to a highest point of 9 metres above sea level.

So the wave is less than a metre out in deep water; at this particular locality the topography and the bathymetry conspired to drive that water up.

(Click on image for a larger version)

It actually dragged a four-wheel drive inland about 10 metres, and destroyed a camp site. The story goes that there was a little baby about to be put to bed in that camp site, and it is quite fortuitous that the campers were able to escape just before the wave hit.

So tsunami impacts do occur in Australia. They have typically been far more localised, from what we understand at this stage. They haven't been the catastrophic, whole-of-coast impacts we have seen in other parts of the world.

Let's move on to hazard and impact. We will start with an idea of how we model the deep-water hazard, or what we have done to date.

(Click on image for a larger version)

For this map the approach we have taken is to simulate a magnitude 8.5 event, quite a decent-sized event, on each one of these little orange and yellow bars – hundreds of events. We simulate the tsunami from every single one of those, for an 8.5 event.

The biggest limitation of some maps is that it gives no sense of probability of an event. You don't know whether the events affecting north-west Western Australia are more or less likely than the events affecting Tasmania.

Just as importantly, you have no sense as to what the occurrence rate is on any of them. Are they 1:500 year events? Are they 1:10,000 year events? The answer to that really affects how you mitigate them, how you deal with them practically, and how you prioritise them against other risks.

In order to give that sense of probability, there are tools that allow you to integrate the information, but you really have to understand how often the earthquakes occur that cause tsunami, and you have to understand how big they can be. There are a lot of technologies, a lot of geological techniques that can be applied, but what I really want to show you is not how to solve those – I think one of the points of this workshop is to draw together some ideas – but to show you how influential those questions can be.

To consider the hazard in Australia, let's look at the Java Trench.

(Click on image for a larger version)

Fundamentally, we don't know how big the biggest earthquake on the Java Trench can be. It might be that that trench can only hold, say, an 8.5. It might be that it can hold something up to 9.5. Probably it is somewhere in the middle, but we really don't know the answer to that. So what does that tangibly mean? What sort of uncertainties does that produce?

(Click on image for a larger version)

If we consider Fremantle and look at the hazard, we can create a curve such as is shown on the next slide.

(Click on image for a larger version)

What this shows you is the return period at different wave heights. I will go into what each of the different curves is, but basically you can ask: what is the probability of seeing a tsunami of a certain height offshore from Fremantle, at 100 metres water depth?

If you say to yourself, 'Well, look, in reality it is only an 8.5; that's as big as the Java Trench can hold,' then what you get is an estimate of hazard for 1000 years at about 0.2 metres, about 20 centimetres. If, on the other hand, you think that the trench can hold about a magnitude 9.5 earthquake, at the other extreme, the hazard is actually a bit over 0.5 metres. So it is 2½ times different. And that's out in deep water still. You don't know what it is going to do onshore, but in deep water that question as to uncertainty in the one earthquake source can basically double or halve your understanding of the hazard that you face.

So they are important questions. They fundamentally change what the impacts are that tsunami will produce in Australia.

It is one thing to look at tsunami hazard in deep water, but the real question is: what does it do to our communities? It's one thing out at 100 metres water depth; we really want to know what it does onshore.

There are a lot of modelling tools available to do that. I will give Geoscience Australia's model a quick plug. Geoscience Australia has created, along with the Australian National University, a very inventively named model, ANUGA. It is a free, open source model. Like many others, it models the flow in a detailed shallow-water wave equation, basically. It is designed to bring a tsunami onshore.

The problem is that, whilst the model exists, to actually use it you have to have high-resolution elevation and bathymetry data – that is a map of the Earth's surface, both underwater and onshore, what the elevation is – because that is what fundamentally controls where a tsunami will go. That is a data set that, as I will show you a little bit later, is lacking in Australia in many places, and it influences many of the things we try to look at in coastline research. Storm-surge modelling, anything where you try and bring a wave onshore, needs to have this bathymetry and elevation data.

What I will show you now is meant to give a sense of some of the outputs we produce, using the best available data. Using what does exist in Australia, what can we see?

(Click on image for a larger version)

What we have done is to model this big event – again a deep-water tsunami, so this map only shows the water heights in deep water. We have taken that event and we have driven it onshore, using our modelling, for about half a dozen Western Australian communities.

There are a number of things we can model. The first is onshore impact: what does the wave do onshore?

The first thing you see is the draw-down. The wave sucks out the water and then propagates along the coast. It doesn't do very much to the community straight away; it sort of scoots along. But it banks up over on the far side of the bay and is reflected off there and drives back to the community again.

Then there can be a second wave that banks up (maybe about 15 or 20 minutes later) and this second wave comes back onshore and comes through to inundate parts of communities.

So it is quite often not the first wave with the tsunami that does all the damage. Thus there is a real emergency management issue of how to keep people away until you are sure all the waves have gone and all the risk is past.

You can take the outputs from models and use them, by turning them into maps for communities. For instance, it is possible to take the same tsunami and model it at high tide.

What about near shore impacts?

(Click on image for a larger version)

About a week ago I was stuck on an aircraft for 40 minutes before we took off, allowing me to watch Bondi Rescue, which I would normally not see. It drove home the point of the kinds of marine threats we face. This is on a normal day, this is without a tsunami. These words should be read in a dramatic radio voice, but to cut a long story short, rips and currents can be quite dangerous under normal conditions in our day-to-day environment.

If people think about tsunami, mostly they think about the massive onshore waves. One of the bigger risks in Australia is that they can cause dramatic currents, really quite dangerous and quite unexpected currents. You could imagine that in this particular case, if instead of just a dozen swimmers swept off Bondi you had a few hundred or a few thousand, it would be a catastrophe – or a disaster, in the true sense of the word.

(Click on image for a larger version)

So what we were able to do with these same models was to put out an estimate of what the velocities are, created by these events. So we are starting to move towards informing: what are the currents, what are the sorts of things we might see?

Port Hedland, which didn't see much impact onshore, obviously has a lot of marine infrastructure – a lot of big issues there. And it's getting currents greater than 6 metres a second in some places. In some of the communities we have modelled we have seen velocities up around 10 metres a second. To give you a feel for what that means: the fastest person in the world can run at about 10 metres a second for, say, 10 seconds. In the water, a current like that is a very, very dangerous beast indeed, and it is something we need to be aware of and we need to mitigate.

These models are now being used to assist emergency managers, emergency responders and local government in these areas, in preparation for an impact and response to a tsunami.

(Click on image for a larger version)

However, there are significant data gaps. Any model is only as good as the things you put into it. Whilst I think any of these models are quite well developed, there are serious holes in the data sets we feed in to them.

(Click on image for a larger version)

In the elevation stakes, this slide shows an example of two data sets that we saw in Onslow and the difference it makes. The line on each of these (the 'bad' data set, the 'good' data set) shows you where, based on an elevation model, high tide would be – no tsunami, just where high tide would be. What one of the data sets suggests is that Onslow couldn't handle a high tide; it is under about 2 metres of water every time it gets its highest astronomical tide. That is blatantly wrong.

(Click on image for a larger version)

The same problem is even worse for bathymetry. At Karratha I showed you that there was basically no inundation for the community at all. But this map shows where there are bathymetry soundings in the region. So each coloured dot is an estimate of what the water depth is. And you can see that there is quite a lot of what is called a 'white strip'.

(Click on image for a larger version)

In particular, in front of Karratha there is basically no data at all; we have got a dead straight line coming through there. So the question is: if you've got no data going into the model, how good is your answer? And if you plan around the results of that model as it stands at the moment, how good are the decisions you make? I think that's a real concern.

Detailed inundation modelling is time consuming and requires detailed input data, and, just as importantly, that data is expensive to get if you don't have it already. It can be quite hard to find.

So one question to think about is: in the absence of that data, are there alternative methods you can use to try and prioritise where you do the detailed modelling? We can do the deep-water hazard, but can you pick coastal regions that are perhaps more vulnerable, naturally, based on their shape? I honestly don't know. There is work going on looking into that, but I think it is a very interesting question.

There are a couple of things I haven't touched on in this talk. There is a lot of work that needs to be done in understanding what a tsunami will do to Australian structures and to Australian infrastructure, as well as the vulnerability of people. And there is an entire other issue around how Australia will deal with warnings of tsunami, given that most warnings will either be nil events or potentially just 'Marine: Get off the beach.' It is an area outside my expertise, but I think it is a critical question. Tsunami will rarely be catastrophic events in Australia, so how do you maintain effective responses in your communities to an event for which there are either false alarms or regular nil warnings?

To summarise, I think modelling of tsunami can give critical information for emergency management, but there are serious limitations or questions that need to be answered, particularly: how often do tsunami occur, how big can they be, and how do you model them or how do you get detailed elevation and bathymetry data to model them accurately?

Discussion

Question – Amongst the earthquakes in the last four years along that plate boundary there seems to be a migration from north to south, towards Australia. Do you think we are in more of a danger than ever?

Trevor Dhu – There was modelling work done after the 2004 tsunami suggesting that the next zone to the south-east of that would be more likely to have an event, and lo and behold, within a few months it went. Mind you, they were not suggesting it would happen within a few months; they were talking decades, potentially – but higher likelihood. I think the same modelling suggested then that the next segment down to the south-east is loaded up again.

Certainly as you move towards the south-east and you move to those subduction zones that are better directed, if you like – or, from an Australian community's point of view, worse directed, because they are pointed straight at us – yes, that certainly is a bigger risk.

We don't really know how likely earthquakes are on those subduction zones, and until we know that it is very hard to put a quantitative finger on what the risk is.