Australian Severe Weather Forum

This was the head lines on page 14 & 15 of the QLD Sunday mail (29/03/09). Quoting Directly from the paper:

(Hannah Martin

March 29, 2009 12:00am

SEVERE weather warnings will be issued up to three days in advance thanks to the weather bureau's new super computers.

They will make forecast calculations 11 times faster than existing technology.

The two computers, together worth $30 million, are expected to begin coming on line in late September, just in time for the next storm season, and be fully operational within two years.

They will create highly accurate models and tracking paths of severe weather systems like the destructive storm that flattened parts of The Gap in Brisbane in October, and Cyclone Hamish which swept down the Queensland coast this month.

Cyclones and thunderstorms are among the state's many extreme weather patterns that are analysed and explained in The Sunday Mail-Channel 9 Book of Weather, which will be free inside the paper next Sunday.

The weather bureau 's chief information officer, Philip Tannenbaum, said the faster processing capability of the new computers would allow forecasters to produce an increased number of advanced weather models, based on regional, national and global data.

He said accurate shorter forecasts will be made three days ahead, instead of the current two using existing bureau computers.

"We're one of the top-tier weather organisations in the world and this technology will keep us on par with our peers," he said.

New technology is helping meteorologists track all aspects of the weather better.)





I give the Bureau full credit in trying, but as I read the article I wondered to what extent it will be successful. There is some degree in accuracy, but I am very skeptical of these computer models, as in an earlier thread today, I have posted 2 seven day forecast less than 12 hours apart and have two different predictions. I sarcastically said it changes more than a babies nappy, due to the frustration of the inaccuracies. A lot on this forum gather their own data and predict storm events pretty accurately,then go and gather fantastic photos and video footage, and other evidence, that is what strom chasing is about.

 I know a lot of the younger ones, and maybe older, who may have total trust in computers, will jump down my throat for this, but for the next two months, keep a chart of the forecast and see the percentage of accuracy. Too much reliance on computer generated modelling and putting faith into is not the full answer. Every single storm , every single weather pattern, is different to some degree, and the reality is, it is only been for a brief period in history that we have records for Australia and other countries to their weather systems and patterns. We don't know, the full cycles of storms. We talk about 50 year storms, 100 year storms, from ice records we talk about maybe 1000 + years, but what really happened in those storms. Do our computer models take those factors into there calculations. Last year we had one spokesman for BOM state that their models did not predict what the severe weather actually did. It took a completely different path and dispersed.

It will be intersting to watch and see if these new super computers worth $30 million dollars, do help and how accurate. Another fear factor is, over the past several months there has been a lot of talk of cutting back staff numbers in the stations. There has been talk of one man stations and this caused out rage from the airport, RAAF and other authorities. Maybe this new computer system is part of the staff cut backs.

Coltan,

Too much reliance on computers is like watching Twister - with the scences involving Jonas - you can have all the technology in the world, but it never beats pure instinct and know-how.

Big Pete

               The most important we should always ready and we need to know the updates.

_________________
Manitowoc ice machine

Interesting comment Coltan. I can agree with you in some aspects there, computer reliance can be a bit worrisome, however:
-This supposed additional speed will be used to improve forecast accuracy, as accuracy and model resolution are dictated by the speed by which a model run can be completed. Let just remember this faith process you are talking about fails just as often as a forecast model. Anyone who knows anything about weather forecasts can tell you not to consider forecasts past 4 days...its only on the medias request that its 7 days. Accuracy (skill) for current models at days out is something like:
96% 1 day, 87% 2 days, 75% 3 days, 62% 4 days, 49% 5 days etc etc (approximate values from memory). This is why the priority bureau forecast is 4 day, as this is the main region of accuracy thereafter is no better than a heads or tails guess. If you ever wonder why its so difficult to forecast extremely accurately consider this:
Generally the Bureau uses a MesoLAPS equivalent forecast: thats 12.5km gridded resolution over the surface air of the continent, plus and expanded region of LAPS (50km) out to the edge, nested within a GASP outer global system. Just for the forecast over Australia, without taking into account nested capital models (3km res) we have 12.5kmx12.5km=156.25km^2 grid squares: so over a surface area of 7741220 square kilometers we have 49544 data columns roughly speaking. Now, thats just on one level, MesoLAPS is a 39 level model, so 1932209 data points roughly. For each of those data points (which approximate areas of air which actually compose millions of molecules), we have 12 variables or thereabouts, geopotential, total precip moisture, u and v winds, temperature, dew-point, equiv potential temperature, vertical velocity etc. etc. From this we produce temperature and weather forecasts, and integrals of this gridded data is used to formulate severe weather forecast probabilities (im working on that stuff atm). This is approximating a massive amount of very changeable atmosphere, and running a calculation like this takes time. We also need to remember that numerical approximation used in solving equations of flow at every data point induces errors, depending on how many terms you discard to make the algorithm faster. These add up rapidly when you consider 7 day forecasts and result in the rapid fall of accuracy. Those differences over 12 hours you note are not surprising if you think about that physically being a mass of air thats likely 8-10 thousand kms away over the ocean south of africa....very hard to predict things that far away...its a limitation of meteorlogy.

- I disagree with every storm, every system being different, there is a high degree of similarity in the formative characteristics and enviromental parameters (particularly for thunderstorms and ECLs), and in the end this is what the forecasters rely on. I think better timed sounding would be an improvement though to forecasting for convection. Sometimes real time measurements are very important. However if you ever wondered why convection isnt perfectly resolved, remember any thunderstorm is a billion tonnes or more of moving air, which is still smaller in resolution than the available or models curently in use...hence it becomes very difficult to simulate them exactly...this is the context the bureau person you were talking about. Another statistic I like to talk to lay-people about the difficulties of modelling, remember that the average thunderstorm contains a billion tonnes of air and moisture, at any one time there are roughly 17000 thundestorms over the globe, and at any time thunderstorms probaby cover a maximum of 17% of te globe. Thats alot of moving air particles, and even small parcels, no wonder its so hard to model, even for one time step better lone 6 hour time steps.

-A bit of inside info: The super computers are naught to do with staff cuts, anyone who wants to put the blamely squarely where its due for staff cts should look at the Federal government elected who has taken a razor to the BOM budget (rather than all the other useless gov departments, lot of resentment at CAWCR over this). The bureau has been saving in the budget for quite a time for the new computer system, to update the seriously overburdened computer system. Automated systems seem to be what the government and admin expect for meteorology, regardless of the result.

Happy to talk more about this stuff if you like.

John, that is very good and appreciated information, but as you say as a "lay" persons point of view, I look at the statistics you have produced, and on the very first day you quote it as having 96% accuracy then a diminishing ratio thereafter, however during the course of the day, as BOM releases it forecast, 4 times in that period, "day one and two," time lapsed, will  very often change 4 times.  (I don't know as regards other states, but referring to Brisbane forecast).  This is the area that as a " lay person",  you  start to lack confidence in the systems when you are supposed to have a 96% and 87% accuracy.  As for the rest of the forecast,( especially the 7 day,) over many years of observing, you take it with a pinch of salt. Sometimes you wonder, and as you hinted on, the media wanting a 7 day outlook, that this is the turning point when creditability slowly went out the door. The joke up here in Qld, in talking to any one on the street is " the weather is supposed to be_____ , that is if they ( weather forecasters) got it right." They totally lack confidence in the system.

I don't know, for the older ones, 35+years, that as a kid, the forecast seem to pretty smack on. The seven day forecast was unheard of on the news, I think  the long range was 3 day Max. Then again we could have been in a very predictable weather pattern through out those years.

It seems the weather will be frontier that will never be a 100% understood.  

You stated "Automated systems seem to be what the government and admin expect for meteorology, regardless of the result." That is a very big concern. Are they trying to hide or cover  something up?  

My grudge was, that they will spend $ 30 million on the computers, but cut back staff or have unmanned stations in maybe "crucial places". As you wrote, and know I appreciate, the need for the super computer is to calculate the enormous volume of data that is processed. The articles I had read seem to point to making jobs redundant. Therefore you just wonder, as the climate is shifting/ changing or cycling through uncharted waters for our generation, were the priory lies for some authorities. Should more money be filtered into education such as what you are doing? Do you have grants available to study the weather as deeply as you are? Do you have a lot of fellow students, or just a couple? So many questions to ask.  

I hope you do well in your studies and honestly look forward to your postings, I read them with great interest and appreciation for the effort you have exerted. It is a science on its own, and by reading your post it is a very demanding role. I am really looking forward to the outcome of your findings.

cheers
Col

An article dealing with the super computer and the speed and amount of data it will calculate, it is mind blowing. A much better article than the one I originally posted. John Allen has got me more curious with the calculations etc.

Weather supercomputer announced for BOM and ANU

By Chloe Lake, Technology Editor  NEWS.com.au

March 19, 2009 10:45am

  


    * Supercomputer to help firefighters, scientists
    * One hour's work worth "10 years with PCs"
    * Technology: More computer news, reviews

A $30 million, four-year project to create Australia's biggest weather computer is underway.

The new supercomputing system, being built for the Bureau of Meteorology (BOM) and Australian National University (ANU), will make weather predictions more accurate.

The BOM machine will have the capacity to make about 1.5 trillion complex weather calculations a second as it crunches through weather data from around the country.

It is expected to provide vital information for future firefighting efforts, and will also help predict climate change in the region.

"I think the biggest impact that people will see in the shorter term is that the accuracy of the shorter-term forecasts - days one to four, or five – will get better," Chris Ryan, head of the Bureau's National Meteorological and Oceanographic Centre, told news.com.au.

"That means that – as we saw in the bushfires on the 7th of February – forecast of the wind speed and the changes will be accurate enough to enable planning by people like the firefighters to take more trust in our forecasts and deploy their resources more efficiently."


The BOM's supercomputer will start operating at a massive 50 teraflops per second, according to the Bureau's chief information officer Phil Tannenbaum. "The average computer at home can do maybe... a few billion at most," he said.

"… What you can do on a supercomputer in an hour is what you can do in a decade with a roomful of desktop machines."

The machine, built by Sun Microsystems, will work in tandem with a more powerful twin based in the National Computational Infrastructure (NCI), a Government-funded research section of the Australian National University (ANU).

The ANU machine, which will run at 140 teraflops per second, will be among the world's top 30 high performance computing systems.

As well as calculating climate change with international organisations, the supercomputer will be used in research by Australia’s science community.

Director of NCI Professor Lindsay Botten said it would be used for nanotechnology, photonics, astronomy, and chemistry, including "understanding how Alzheimer's disease progresses and maybe how to stop the onset of that."

The news comes as IBM is reportedly in takeover talks with the machine's maker Sun Microsystems, according to the Wall Street Journal.

The supercomputing system will be ready by the end of the year.


http://www.news.com.au/technology/story/0,28348,25209694-5014239,00.html