Our Forecasting System's Slower Than A PC, Says Thai Meteorological Department

[webfact]

Our forecasting system's slower than a PC, says Met Dept

The Nation

File photo. Source: wikimedia

The accuracy of weather forecasts by the Meteorological Department is only around 80 per cent because of its obsolete computer systems, deputy director-general Somchai Baimuang said yesterday.

The department's system - now working slower than a modern personal computer - would need a Bt2 billion budget to be replaced.

"Our systems might be little better than today’s PCs, in terms of systems’ stability. A forecast could be accurate for a seven-day period, but in a period longer than that, we can deliver only speculation," he added.

The longer the period of forecasting the less accurate the results, he said. "The daily forecast’s accuracy was around 75-79 per cent, while the accuracy for a seven day period was down to 60 per cent. The most accurate forecasting is for a three hours’ or six hours’ period, at around 90 per cent," he added.

In terms of range, the department’s computers could forecast weather at only provincial level, but could not do so at district or village level. "The department has been capable of forecasting in a seven day period and at the provincial level. This is our limitation which we live with, as we lack newer computers," he added.

"The accuracy of any forecasts done in advance for a period longer than 10 days is very low. A new system would help with forecasting accuracy over a period longer than 10 days, or even up to a year," Somchai said.

In a new system, five components would be replaced or put in place - beginning with a measuring mechanism comprising subordinate tools in a national network. "Our current system cannot yet fully link the entire network," he said.

The second would be a better processing mechanism to analyse data from the networks outside Bangkok, or from abroad. "We need a supercomputer to do this job, which would make forecasts error-free, or at least better than our existing mechanism," he said.

"An improved processing mechanism would enable forecasts down to district, tambon and village level, in a period longer than those we are capable of, with greater accuracy. This is what we need," he added. Personnel development is also needed to improve staff performances to cope with advanced technology.

If entirely improved, the five new components could ensure hourly forecasts and all types of weather reports at all channels available, including a proposed television channel run by the Meterological Department, instead of only on its website, which is not accessible to the entire audience or population, he said.

A budget request made by the department has regularly been ignored or under-served, he said. "A budget next year for Bt900 million has been earmarked for salaries and routine spending, leaving us a Bt200 million to spend or invest on new projects or machines."

The acting director of Geo-Informatics and Space Technology Development Agency, Anond Sanitwong na Ayutthaya, said a database should be processed and made available under a uniform format, or changed into an entirely new format to standardise information for communal use.

-- The Nation 2012-01-24

[pedro01]

Why don't they just look on CNN web site for the forecast...

[Soutpeel]

"The longer the period of forecasting the less accurate the results, he said. "The daily forecast’s accuracy was around 75-79 per cent, while the accuracy for a seven day period was down to 60 per cent. The most accurate forecasting is for a three hours’ or six hours’ period, at around 90 per cent," he added"

"The accuracy of any forecasts done in advance for a period longer than 10 days is very low. A new system would help with forecasting accuracy over a period longer than 10 days, or even up to a year,

Which is pretty much on par with the rest of the world as regards accuracy, Khun Somchai, what do you really want the cash for ?

[Buchholz]

Why don't they just look on CNN web site for the forecast...

There's no way to pocket millions and millions of baht that way.

need a Bt2 billion budget to be replaced.

Edited January 24, 2012 by Buchholz

[thaicbr]

He's just telling the truth of it.. the system is way outdated.

also he's covering his back for when we get flood 2012.. he can just say ..sorry. but i told them we could not foresee this because the equipment is old.

[1weatherman]

Since this is my area of expertise (over 30 years in weather), I shall reply to some sections of the Director Generals statements. First off, the "supercomputer" the Met Dept. has to crunch all the numbers is 15 years old. 15 years old is like dead in human years. And I'd say a lot of higher end personal computers now are more stable. I have a couple of weather programs on my laptop which might not run on their computer. I've seen it, it needs to be replaced. It's archaic. Thus the extra money they need to ask the govt. for, in addition to the regular budget, will be for the purchase of a new supercomputer.

As for this statement: "The longer the period of forecasting the less accurate the results, he said. "The daily forecast’s accuracy was around 75-79 per cent, while the accuracy for a seven day period was down to 60 per cent. The most accurate forecasting is for a three hours’ or six hours’ period, at around 90 per cent," he added. All true but even in the USA the % is probably only 5% better at these time periods. And: "The accuracy of any forecasts done in advance for a period longer than 10 days is very low. A new system would help with forecasting accuracy over a period longer than 10 days, or even up to a year," Somchai said. Yes you'd get a little better out to 7 to 10 days, but up to one year? Somchai is being the politician on this one. Also: "We need a supercomputer to do this job, which would make forecasts error-free, or at least better than our existing mechanism," he said. Umm. When it's going to be clear blue skies out and no wind, otherwise he's dreaming. I really don't know what he means for error-free. It would be better than the existing mechanism for sure. As thus we come to: If entirely improved, the five new components could ensure hourly forecasts and all types of weather reports at all channels available, including a proposed television channel run by the Meterological Department, instead of only on its website, which is not accessible to the entire audience or population, he said. Thailands version of the Weather Channel. For those of you concerned that the on-air weather personalities giving said weather reports will be old men, there are a few lovely females that work there.

With all this being said, weather forecasting in the tropics isn't easy. It's easier to forecast the weather for the states from over here than it is to forecast Thailand's weather from here. I'm sure a new supercomputer will help the Thai forecasters. It certainly wouldn't hurt. And they certainly need intregration of the various components as stated in the article into one main system. As for the Met Dept. getting the money in this fiscal years budget for a new computer, sorry that money will go to the purchase of the tablet PC's for the children of Thailand. Better luck next year.

[IsaanUSA]

80% sounds amazing!

[hkt83100]

My usual forecast for tomorrow is "same like today" and I am right 90%. Where are my billions?

[sparebox2]

No need to spend billions.

Just go to Pantip to pick up a high end PC.

100,000 Baht should be enough, and that includes a 3 in one (color inkjet printer, scanner + fax) thrown in.

[Old Man River]

It was only about 5 years ago that they reported in one of the newspapers that their accuracy on daily forecasts was about 50%. Hence, it looks like they are doing much better than before. Still, given the importance of what they do on so many levels, if they are using archaic equipment then it needs to be updated quickly and it will be costly.

[Pib]

I sure wish they could improve their forecasts for rain in specific areas. Ex: Forecast may be 80% chance of rain in Bangkok today. Well, I've seen such forecasts many, many times and it never rained in my part of Bangkok...of course the weather service response would be, Well, it did indeed rain in at least one part of Bangkok so our forecast was accurate. Then there are many cases were the forecast is 20% chance of rain (which is low) but it rains big time for hours. I guess when it comes to issuing rain forecasts during rainy season trying to figure out if that thunderstorm is going to pass over your area is just a crap shoot.

[harrry]

Maybe the forceasts are now good for a very big area....it says a province. A new computer may enable them to forecast for smaller areas making them much more useful.

[belg]

2 billion, hahah, 15.000 baht for the one computer and the rest in the large pockets of the many many receipients...

555

[bigbamboo]

80% sounds amazing!

Agree............ in most countries people can only dream of that level of accuracy! And after upgrading the system by buying a decent PC you'd get plenty of change from two billion baht that could be far better spent elsewhere.

Edited January 24, 2012 by bigbamboo

[R10wota]

If the Computer is slow.. Replace it.

If you need more sophisticated equipments .. Get it.

You can get a lot in Bt 2 Billion...

Except.. Common Sense and Logic.

The rainy season is around the corner, The Biggest Dam is 90% full. The second Dam is 85% full. Now you guys need a "Computer"... well a "FAST COMPUTER" to tell that you need to drain some water out?

Billions won't be enough to have that logic.. and common sense.

[R10wota]

2 billion, hahah, 15.000 baht for the one computer and the rest in the large pockets of the many many receipients...

555

Well even if you get a Supercomputer (not that it is required but still) you only spend about 5% of the Quoted Budget... hahaha

[anterian]

Short range forecasts up to 24 hours, I watch the birds, ants, winged insects and the clouds, I rarely get it wrong. For longer range forecasts, several days, I look at the wind strength and direction and what is happening in adjacent areas. For longer ranges I look at satellite weather maps and previous patterns.

[bigbamboo]

Short range forecasts up to 24 hours, I watch the birds, ants, winged insects and the clouds, I rarely get it wrong. For longer range forecasts, several days, I look at the wind strength and direction and what is happening in adjacent areas. For longer ranges I look at satellite weather maps and previous patterns.

Or for an even better indication see what the girls are wearing. If it's a thin,skimpy top, it's going to rain soon!

[americaninbangkok]

Maybe they can just beat up the 6-year olds that are the priority target audience for the first round of tablet computers to be distributed by the government?

[bkk_mike]

error-free weather forecasts...???

Weather is the classis chaotic system where an immeasurable difference can have a huge impact on the forecast, especially as you try to forecast further out...

[bytebuster]

It was only about 5 years ago that they reported in one of the newspapers that their accuracy on daily forecasts was about 50%.

50% answering the question "rainy or sunny"? Well, a 1-baht coin would deliver exactly the same accuracy.

[Old Man River]

It was only about 5 years ago that they reported in one of the newspapers that their accuracy on daily forecasts was about 50%.

50% answering the question "rainy or sunny"? Well, a 1-baht coin would deliver exactly the same accuracy.

Actually, what I told my friends at that time was that I could open the window and stick my arm out and be 100% correct. I found it interesting that the next day in the other English language paper (I forget which) it was roughly the same article, but had been changed to 60% accuracy.

Thinking back, it was pre tsunami, but definitely in the past 10 years.

[bytebuster]

Thinking back, it was pre tsunami, but definitely in the past 10 years.

That might be the case.

However... Wait a minute... How then the people forecasted weather forty years ago? "Viking" and "Pioneer" spacecrafts had weaker processors than my coffee machine has (I've opened it). And still, people wrote programs delivering accuracy much better than 50 or 60%.

[stoneyboy]

I appreciate that tropical climates may be unpredictable,but is really that difficult,obviously they don't have as many options as say the UK,possible snow,sleet,hail,fog,mist,ice of various types etc etc.

I mean if I was to say the weather for the remainder of the week is likely to be hot and humid,with possible showers I doubt I'd be far wrong would I,with an average high of say 32 and depending on location a low of 21.

Can't see what all the fuss is about.

And today it's going to be hot,and tomorrow it's likely to be hot,and Thursday is going to be hot with the chance of a shower in the afternoon.

[animatic]

There are fully funded and top-level weather monitoring and projection services that cover ALL of Thailand and the region, that are as simple as $99 a year to be subscribed to and would give accurate as possible up to date and long term forecasting. 2-3 of these services would be more than enough with the existing systems to give accurate reports.

Of course that leaves nothing concrete to be the boss of,

or more importantly not enough purchasing budget to skim enough to become a real player.

So WE NEED TO BUY NOW!!!!

Where's our slice of the P.T.Pie !!!

You also can EASILY imagine that the guys at Pantip could build

an over-clocked, distributed processing, super computer network,

for under US$25,000 Let alone hand the brief to a Uni to build one.

Again not enough room for a good skim-off.

Edited January 24, 2012 by animatic

[Soutpeel]

There are fully funded and top-level weather monitoring and projection services that cover ALL of Thailand and the region, that are as simple as $99 a year to be subscribed to and would give accurate as possible up to date and long term forecasting. 2-3 of these services would be more than enough with the existing systems to give accurate reports.

Of course that leaves nothing concrete to be the boss of,

or more importantly not enough purchasing budget to skim enough to become a real player.

So WE NEED TO BUY NOW!!!!

Where's our slice of the P.T.Pie !!!

You also can EASILY imagine that the guys at Pantip could build

an over-clocked, distributed processing, super computer network,

for under US$25,000 Let alone hand the brief to a Uni to build one.

Again not enough room for a good skim-off.

Yeap plenty of these around....strangely enough the offshore boys use weather services out of Australia for their forecasts in the gulf of Thailand... ....pretty sure if the goverment offered the company concerned 1/10 of the budget proposed sure they would provide forecasts to the Thai goverment...LOL

[bkk75]

Yeap plenty of these around....strangely enough the offshore boys use weather services out of Australia for their forecasts in the gulf of Thailand... ....pretty sure if the goverment offered the company concerned 1/10 of the budget proposed sure they would provide forecasts to the Thai goverment...LOL

Then they would lose face if they outsourced. Kon Thai tham dai na

[ResX]

He's just telling the truth of it.. the system is way outdated.

also he's covering his back for when we get flood 2012.. he can just say ..sorry. but i told them we could not foresee this because the equipment is old.

The truth is dealing with seasonal floods using majors flood control dams you don't need any forecasting system. To date there is nothing better than using the dedicated rule curve for each dam. In my country we have done this over the last 25 years. We have improved the rule curves tremendously over the last 8 years.

Thailand also has used the rule curves approach over the last 20 years that I can recall. But I think there are corrections need to be done here and there. We have gone through the same learning curve over the last 25 years.

No, he can't cover his back using that argument