HIGH FLYERS THINK TANK

Extreme Natural Hazards

University of Melbourne, Tuesday 30 October 2007

Mr Alan Sharp
Bureau of Meteorology

Alan Sharp has a degree in physics from the University of Western Australia and a masters degree in applied physics from Curtin University. Alan worked at the Bureau of Meteorology from 1982 to 1995 as a weather forecaster in Perth and Darwin. Since 1995, he has worked at the Bureau’s head office evaluating new meteorological instruments and observation systems, and as the national manager of tropical cyclone warning services. Alan has also been part of two expert teams for the World Meteorological Organisation.

Assessing tropical cyclone risk in a global warming environment

My subject today is cyclone trends in a global warming environment. Global warming is a topical issue and there is a lot of research going on. But some of the research is a bit more solid, more scientifically based, than other research. Part of the reason behind that is that the amount of data can be limited – often when you are looking at the extreme end of the risk spectrum, as has been noted earlier by Trevor Dhu.

Firstly, I would like to acknowledge Blair Trewin and Jeff Kepert, from the Bureau of Meteorology – some of the work involved here has been thanks to them – and also other Bureau staff who have helped me in this process.

Also, I would like to mention the Bureau's position in what has been referred to as PPRR, a well-known phrase in the emergency management community. The Bureau is more involved in the Prevention and Preparedness area – that is, before the event and getting into the event – than in the Response and Recovery area. It is not that we don't have any role in response and recovery, because when events have occurred and chaos is reigning supreme it is as important as ever to know what is going to come behind the cyclone, the fire or whatever else has caused the disaster.

(Click on image for a larger version)

Prevention is more involved in building structures, town planning and so forth, making sure that you haven't made mistakes before the event. This is based more on climate data – you look at return periods, records, etc. You are not likely to get tropical cyclones in places like Melbourne so you don't have to build for them, but decision-making processes involve such things as where the cyclone is going to occur and where the risks are. It is not always going to cover every case. The Australian Standard didn't go far enough inland to have the structures robust enough where two people were killed inland from the west coast when Cyclone George went through last year.

Forecasts and outlooks look more at the preparedness stage. When you are getting close to the event there is a cyclone forecast warning about clearing up your yard, making sure you have got all your food and batteries ready in case it happens, evacuation, securing buildings and so on.

To set the picture:

(Click on image for a larger version)

Hurricane Katrina occurred about two years ago. It flooded New Orleans and caused a major disaster. It wasn't just Hurricane Katrina; there were 25 other cyclones and hurricanes which affected the North Atlantic that year, which beat the previous record by five. It was a remarkable year.

Hurricane Katrina caused the major damage, and coupled with a record year of 26 tropical cyclones, stoked the fire of the global warming debate.

Many people have jumped on the bandwagon to suggest that this is a sign of things to come as the Earth warms, and we will agree that the Earth is warming. But are these particular events a sign of global warming?

One thing to look at is the evidence. We are going to be examining the possible effect of global warming on cyclones whilst managing the pitfalls of the actual data, using a limited database. We haven't had that many cyclones in Australia and in the world in the last 100 years, when we have been looking at them closely, and so we are not looking at large statistical databases, particularly over the zone.

(Click on image for a larger version)

Firstly, let's have a look at the tropical cyclone zones. Our particular area is covered by Australia.

(Click on image for a larger version)

You see here the cyclone tracks which have been recorded over the last 50 or so years. The colours indicate the intensity, using the American Saffir-Simpson scale. (That is a bit different from ours but it gives you the idea.) The area to our north has had quite a few more intense storms than anywhere else in the world. We are less affected, but the area at the latitude of northern Australia, for instance, does get quite a lot of landfalling activity.

Also notice the relative distances from the equator. That will come into play later in the discussion.

The Australian database is managed by the Bureau of Meteorology. It runs back to 1906. (The Bureau was actually created in 1908.) Track data is contained in this database; other details are limited, so we don't have too much information on intensity, especially in the early days, and size and so forth. There is some information but it is fairly limited.

There are 961 cyclones on the record for 1909 to 2006, so let's have a look.

(Click on image for a larger version)

Shown here is the annual frequency of all cyclones recorded in the Australian region since 1909.

It looks dramatic – if you noticed, it did have an increase, but even the most naïve scientist should realise that in the early days we were not picking up as many cyclones as we are now because we didn't have so many means of detection.

In 1906 we only saw the ones which hit the populated areas or isolated ships which never came in to port again. There are a few cases like that, and the worst disaster since the turn of the last century was such a situation.

In the 1940s, with the Second World War, we realised that there was a need for a greater forecasting presence. It was after the war that general emergency management came into play.

In the 1960s, early satellite imagery allowed us to see much better over the ocean and pick up many of the cyclones which previously would have come and gone without anyone knowing, and then geostationary satellites managed to give us full coverage of the globe.

There are many ways around trying to manage a database which is not based on a constant method of population.

Firstly, the post-1970 data should be more complete so we can use that with more confidence than the early parts.

Data during the postwar period, particularly into the 1960s, will have some value.

And the older data can be used for studies over populated areas, like return periods for somewhere like Cairns. Cairns has been around for a while and there are records for Innisfail, for instance. For Larry it can be said that it was the biggest one since the 1917 or 1918 storm which occurred in that area.

(Click on image for a larger version)

Let's revisit this graph which I showed earlier, and have a look at some of its problems. What can we make of the data, how limited is it, and what other warts does the data have?

Too many people take this data on face value, without taking into account any of the problems or caveats that exist therein – which makes for easy science. I point that out because I remember that many years ago, when I first started getting involved in scientific research, the work I was doing involved complex nonlinear relationships. Yet too many of the researchers in those days had just discovered multivariate linear analysis on their computers, and they threw it at this problem and came up with results, but they weren't looking at linear events. It was very easy to throw it into the computer but they were not coming up with the right answers.

Let's have a look at some of the obvious database issues.

It can be seen that 1962 was a big year – or was it? Closer analysis has shown that many of these arose where a keen young staff member named lots of subtropical systems, including east coast lows, as cyclones – particularly the one which occurred between New Zealand and New South Wales in July, I believe.

It appears that 1956–58 is fairly busy too. Some of those are due to duplications. There are duplications throughout the database, but that was a particularly bad time. I think there were problems because these databases had been converted from paper to one type of computer to another, and so forth, and there have been problems in the combination of databases in different parts of the country.

There have also been a few other problems.

(Click on image for a larger version)

This slide shows the cyclones which have been recorded. Note especially that there are only a few cyclones just around the west and north-west coast but nothing further out recorded – maybe it has something to do with more shipping.

(Click on image for a larger version)

Put a few satellites in there and add 100 years, and there are lots of cyclones out there! It's just that in 1906 we didn't see them. They were roaming out there but there weren't very many ships.

(Click on image for a larger version)

Let's have a look at Cyclone Monica, which is one of the most powerful cyclones at landfall that we have had in the last 100 years. It actually hit the coast at Junction Bay – which is a fairly remote area, the nearest communities being Maningrida and inland at Jabiru.

So how would it have been recorded in 1906? Well, there wasn't very much to the north. I think the original Lockhart River community might have been affected by it when it came across the coast, but there would have been nobody in the area who might have recorded it. People would have noted the swells at any settlements that might have been up near Nhulunbuy at the time, and the larger swells down along the coast. In fact, the Indigenous communities up there do note the stronger swells which the cyclones cause. Obviously, there are not very many other major wind systems in the tropics which can cause swells, especially around this area, so once you see swells you figure there is a cyclone out there somewhere.

When it came through, Maningrida was affected probably by category 2 strength winds, so in the end there would have been probably recorded some sort of category 2 cyclone that might have just hit the coast, if reports came out of Maningrida alone.

There was a large swathe, about 30 kilometres wide, of devastation which would have affected crocodiles, cane toads and whatever else happened to be up there – as well as the trees, which were stripped bare. But population numbers were low. In 1906 that would have gone down as a very, very insignificant cyclone.

Now let's consider intensity. Now we are getting onto the more subtle stuff, with intensity, structure, system definitions, wind definitions, instrumentation and personalities.

(Click on image for a larger version)

The intensity is defined by the maximum mean surface wind. This correlates well with pressures, but it is affected by other influences.

Measuring winds in a cyclone can be somewhat problematic – generally instruments get blown away, which is a bit of a problem. And often the cyclone won't go over an instrument; that is another problem.

(Click on image for a larger version)

Robust anemometers can be used, but they are rarely in the right place for the cyclone. If they are even 50 kilometres down the track they can get nothing. If Cyclone Tracy, for instance, had moved 40 kilometres one way or the other, Darwin would have been unaffected. (Cyclone Tracy is one of the smallest ever recorded cyclones, if not the smallest.)

Pressure measurement: we get that more often, but the wind/pressure relationships do have problems. We have to use estimated relationships and there is still research going on into those.

Damage assessment: even though there is a lot of research going into damage assessment to determine the strength of events, it is not helped by the fact that different buildings have different strengths and so forth.

Wind guesstimates: on the slide I say 'estimates' but it is usually more like guesstimates. Witnesses say, 'Oh, it's blowing like the clappers'; we have different people's opinions. Observers on the ground are not experts.

Aircraft reconnaissance is the best way of doing the measurements, and they have done magnificent work in the North Atlantic – they have lots of money up there!

We have a limited Doppler radar network at this stage. We have five, including one at Berrimah, Darwin, with another two going in. Again the cyclone has to be pretty close to a Doppler radar for us to get any useful information.

And satellite sensors are coming more into play over the last 10 to 15 years. The slide shows Quickscat winds, where we have whole swathes of data over the ocean. There are problems with those; they are not perfect – they can be contaminated by rain and so forth.

(Click on images for larger versions)

Intensity analysis is done by a pattern analysis scheme: you look at the spiral effect, in certain cases looking at the spiral bands to see how far it goes out until the eye forms, and then you look at the temperature of the cloud tops around the eye compared with those in the middle of the eye, and the difference is an estimate of the intensity.

(Click on image for a larger version)

The various basins – the Pacific, the Atlantic and so forth – have different DVORAK T-Number pressure/wind relationships, probably because of various issues. For instance, in the Atlantic it is a bit further north, further away from the Equator than the other parts. We will go into that in a moment.

The significant factors are pressure, latitude (the Coriolis effect varies as you go further from the equator), system size (bigger cyclone, smaller cyclone), system movement and intensity trends.

(Click on image for a larger version)

Firstly, Coriolis effects: the Coriolis 'acceleration' is due to the rotation of the Earth. It is not really acceleration, it is how it rotates, and the effective force is related to the latitude, therefore you cannot have one equation for all latitudes.

(Click on image for a larger version)

Similarly, with different pressure gradient structure at the size, you get a little cyclone like this one with the same central pressure, the pressure gradient is going to be greater and that will generate stronger winds.

The system movement has some more subtle effects not just relating to translational speed effects – that is, on one side there are going to be stronger winds and not weaker winds, because the whole system is moving – but also due to varying surface friction effects, which are second degree effects.

The system intensity trend is again demonstrable, but not easy to derive scientifically.

We know a lot more than we did in 1975, for instance, so at present we are going through a period of valuable research but past analysis will be based on the more limited research, so we actually have to go back and look at that data as well. Often much of the information which we had then is now lost – but not all of it; there are good records kept in some situations.

Data and equipment trends: much more data is becoming available now with, as I mentioned, the satellites and so forth. Also, better equipment is coming on-line. These can have subtle effects on data. I will show two examples.

(Click on image for a larger version)

The first example is satellite resolution. The blue dots indicate satellite resolution of one pixel every one kilometre, and the red ones represent every two kilometres, for instance. What I have done is to use a line of numbers, and the red ones are the average of the various ones through there. And you are looking at the older satellites.

The peak temperature difference would not be the same as the difference we find with the newer ones, where you can actually see deeper into the eye. Therefore, you can now get a much deeper intensity estimate.

(Click on image for a larger version)

For wind gust measurements we have a problem in that we are changing the instrumentation. The old-style dynes monitor uses the Bernoulli effect – pressure difference – compared with the new cup anemometers, which are less expensive. These are two different ways of measuring wind; each one of them has its positives and negatives, but they use a different method.

This trace is from Cyclone Tracy. From the actual observer the gust that we have taken (with the exception of one when it came apart) is at the top of the spikes, whereas the actual three-second gust from a cup anemometer would probably be slightly below that – not to mention the fact that you have got different physics involved in the two situations, different dynamics. The method of gusts measurement has not been thoroughly examined.

(Click on image for a larger version)

We will have a look at the various timescales listed here, using different equipment.

(Click on image for a larger version)

In recent records there is an apparent downward trend. I would note that there is a trend in the severe cyclones: the ones shown in red seem to be greater in the centre of the graph than they are through the right-hand end. That one is probably real, because we are only looking at the more valid data of more recent times.

(Click on image for a larger version)

If we have a look at the Atlantic major hurricanes, however, in the information since 1974, especially in the last few years, we see a nice big increasing trend. But if we look at a bit more data, going back to 1944, we find that maybe there is some sort of cyclic trend.

(Click on image for a larger version)

To dig deeper, we can have a look at 20th century northern Queensland. This is actually looking at isotope ratios from stalagmites and naturally observed cyclones at this place in Chillagoe, which is inland from Cairns. There is a reasonable correlation between cyclones and the actual ratio figure. We say, 'Okay, there is a slight increase at one part of the graph, but it is difficult to say.'

(Click on image for a larger version)

But let's have a look over the last millennium, because we do have stalagmite isotope ratios for that period. You can see that perhaps we are in the middle of a quiet period.

So there are lots of other effects, which won't be global warming, of course, because we didn't have it then. This information tells us that there are many other effects which have to be taken into account.

(Click on image for a larger version)

To look at the coastal impacts from 1909 to 2002: the red dots represent severe cyclones and the black ones are just category 1 or 2 cyclones, at impact. So the mid-north-west coast looks fairly dangerous, which it is. There hasn't been too much across the very top, except for Tracy. That is almost 100 years we are looking at there.

Let's have a look at the severe impacts in 2003–07. The red arrows indicate three along the mid-north-west coastline – one at Faraway Bay, which was Ingrid; one up at Cape Don, which was also Ingrid; and Monica, also in that area looking reasonably safe – one right up at Lockhart River, Ingrid again (the first cyclone to be a category 3 or above in three different states); then more impacts by Monica as it was coming across to the east; and finally Larry. So you can see that the cyclones across the northern coast have actually affected areas which you might otherwise have thought were reasonably safe, on the limited data we had.

What do we expect? Let's have a look at some simplistic and complicating factors and climate models.

On a simplistic outlook: cyclones need warm oceans to form; the warmer the ocean, the greater the potential for the cyclone; the oceans are getting warmer. More cyclones? No.

(Click on image for a larger version)

Complications include, firstly, more El Niño effects. There has been a suggestion that global warming could be producing more El Niño effects. That is not 100 per cent sure either; there are lots of uncertainties involved. But it does actually reduce the cyclone impacts in Australia.

Changes to ocean currents: this is partly El Niño that we are talking about, with such drastic things as perhaps the collapse of the Gulf Stream. If the Gulf Stream did collapse, the whole global ocean current structure would change. We don't know what that would do down here.

Greater warming at the Poles: temperatures are actually rising faster at the Poles than they are at the equator. One of the things that cyclones do, as well as most weather patterns, is to help move the energy from the equator towards the Poles. If the Poles are warming up faster, then there is less necessity for energy transport, because the Poles are warming more because of the reduction of the albedo than the greater energy transport, with the melting of the ice.

Greater warming of the stratosphere: the stability of the atmosphere is going to be improved if the stratosphere warms more than the lower ones, which would tend to suggest that that would stymie cyclone developments.

As to upper wind patterns, we are not quite sure how they are going to evolve. A cyclone doesn't just need energy and warmth; it also needs the right wind patterns to make sure it doesn't get blown apart by upper strong winds.

And we have looked at long-term oscillations.

Climate models are necessarily broad-scale, so the finer nuances of tropical cyclone frequency and intensity are not likely to be accurately depicted. Many climate models are coming up with Doomsday – or, more recently, I have heard a couple saying there are going to be fewer cyclones – but it is difficult to actually assess that in view of the necessity for the broader scale analysis.

The recent conclusions from the World Meteorological Organisation and the Intergovernmental Panel on Climate Change are that there is no obvious trend in tropical cyclone frequency; there is some evidence of greater tropical cyclone intensity. But there are too many data quality and complex climate change issues to allow for a definitive long-term forecast at this time.

(Click on image for a larger version)

This slide shows the locations of those reports.

But not all is wrong with the Katrina-led bandwagon. Other issues in the global warming debate are based on more obvious scientific evidence which is coming to the fore. So we can't complain too much about people using Katrina as leverage to get global warming into the mainstream. It has had a positive side.

(Click on image for a larger version)

The danger is, however, that you give the sceptics something to attack legitimately and the whole argument is potentially compromised.

So you have to be aware when you are doing science that short-term databases do have their limitations.

Varying technologies and knowledge over the years can make data almost useless as is. Errors also exist in databases, and you have to be aware of those when you are using them. There is a need to repair the database, using 21st century knowledge. And users of any database should realise its limitations.

Discussion

Question 1 – I work a lot with the cyclone network and I have noticed that you do record the position codes and intensity code, which gives an estimate of how the cyclone's position is fixed and how the intensity [inaudible], which is quite useful…I am just wondering, however, if it is possible to go back where those items are missing, through some of the old data, and go through any paper records or other data you have that is not in the database, to fill in some of the missing values for those codes.

Alan Sharp – We are actually looking at a project which will do just that. We are going to go right back as far as we can – thoroughly go through the system. It is going to be an expensive project, and we are hoping to get some money to do that.

Blair Trewin has been working on an initial repair of the database which gets rid of the obvious mistakes, but to do the more subtle stuff is going to require going over original records which are kept in the archives, looking at it thoroughly, using 21st century knowledge.