Japan attempts 30 years weather prediction, builds expensive paper weight

Does anyone actually believe this is going to work?

Japan is planning ultra long-range 30-year weather forecasts that will predict typhoons, storms, blizzards, droughts and other inclement weather, an official said Tuesday.

The project, to start next year, will harness the powers of one of the world's fastest supercomputers and is an offshoot of ongoing research by the country's science ministry to map global warming trends for the next 300 years.

Using the Earth Simulator supercomputer, housed in a hangar-sized building in Yokohama, just south of Tokyo, Japan's science ministry hopes to calculate long-term patterns in the interaction of atmospheric pressure, air temperatures, ocean currents and sea temperatures, said Tomonori Otake, an official with the ministry's earth environment bureau.

The results will help establish predictable routes for typhoons and identify areas that are recurring targets for heavy rains, abundant snow, high waves, heavy winds, scorching heat or crop-threatening droughts.


The machine tracks global sea temperatures, rainfall and crustal movement to predict natural disasters over the next centuries. As part of the project, Japan eyes forecasts for the entire planet for areas as small as 1.9 square miles.

But don't plan on locking in sunny weather for that planned family picnic in July 2036. These forecasts are only general trends.

"Just like the daily forecast, we can't give a percentage for how accurate they are," Otake said. (Emphasis mine.)

I just can't conceive of how on such a long time scale such a system would be predictable. I don't care how good their computers are. The issue is not one of computation but rather an ever-widdening probability space. Furthermore, if they can't quantitate the size of that probability space -- a percentage of how accurate it is likely to be -- then what good is the prediction?

Come on guys. Haven't you seen The Butterfly Effect? Ashton Kutcher is going to show up and throw all your calculations out of whack.

Hat-tip: Slashdot.

More like this

Apart from Tropical Storm Alberto, the 2006 North Atlantic hurricane season has been pretty slow so far. That's not unexpected; June is never exactly a rockin' month for hurricanes, although last year, July was a rockin' month. So far it's not clear what to expect; I find the best source of day-to-…
This is a huge hurricane/typhoon heading quickly, and imminently, towards taiwan. The storm itself is roughly as wide as the island nation is long, so very little will be left unaffected. The storm is at the very high end of the range of storms in size, strength, etc. It is currently equivalent…
Update: The new forecast track of Neoguri is shown above as well as the location of two nuclear power plants. The forecast track has moved south, and is now in a very good (and here good means bad) position to strike the Sendai nuclear power plant very directly. Keep in mind that this forecast may…
Timed for the Katrina anniversary, the Competitive Enterprise Institute has put out a short report on the hurricane-climate issue (PDF). Without taking a premature stand on who is or isn't winning the scientific debate at the present moment, I'd like to point out how CEI misrepresents the state of…

It's hard to figure out exactly what the real plan is based on the news article. From what it says, I think I could do the same thing for, say, $1 million. All I need are the weather records for the last few decades and then I can tell them where and how much it rains, where typhoons hit and things like that. I think earthquake records can fill in some more of the natural disaster probabilities. If I go back far enough, any normal cyclical climate effects should show up in the records.

Unlike Otake, I think I can give at least an estimate of the accuracy relative to the daily forecast (hint: it will be worse).

By Mark Paris (not verified) on 20 Jul 2006 #permalink

I don't think the idea is to make weather forecasts 30 years in advance, the nutterfly effect would indeed prevent that. Trying to understand how climate change will affect regional distribution of hurricanes and rains aren't an impossible task, however. You may not get an exact date for when it will rain, but you may get an estimate for how high you need to build that dam to prevent flooding.

How well they will succeed is another matter. Japan often starts these megaprojects that fail spectacularly while at the same time they get a lot of less spectacular spin-off results.

By Thomas Palm (not verified) on 20 Jul 2006 #permalink

Global climate models operate on such a large scale that resolving events like typhoons is impossible. It looks like this particular project is confining the spatial and temporal extent so that they may be able to resolve the smaller phenomena, such as typhoons. Not their exact occurence, but their statistics. This is a logical progression in climate modelling.

I still don't see the point. Even today, while hurricane forecasters may estimate the number of storms and major storms for an upcoming season, they don't estimate how many and where hurricanes will reach lan. Hurricane landfall statistics will suffice for that. Trying to predict precipitation in a particular river drainage system is about the same; just look at the records. If you think overall precipitation in a large area will increase, adjust for that. We can't even forecast that well a month in advance now. Of course, what I'm saying is based entirely on an AP story that might not have sufficient details to know what they really plan to do.

By Mark Paris (not verified) on 20 Jul 2006 #permalink

Thomas Palm said: "I don't think the idea is to make weather forecasts 30 years in advance, the nutterfly effect would indeed prevent that."

I like "nutterfly effect." A lot. I think that's my new favorite pseudo-pseudo-scientific-cliche!

Mark, for hurricane landfall locales and regional rainfall, among other phenomena, you can't use historical statistics to predict future events - because the climate's changing. We can't so easily "adjust". However, the information really would still is useful for regional planning.

As for predictions not being done "well", well it is a scientific exercise. We don't know how well we can make predictions until we try. And in doing so, our failures highlight what we don't understand about the processes.

Daniel, I understand, but I don't agree that they can do what the article says they plan to do. I still say that historical data will do the job at least at the level of accuracy that climate models will. In fact, I suspect weather records would do a better job. They also talk about seismic effects, which, I believe, are still not included in climate models. (Read a little good-natured sarcasm here directed at the Japanese researchers.) Also note that although I understand that AGW is changing climate, I stand by my objections, at least to the extent that the AP article is accurate.

By Mark Paris (not verified) on 21 Jul 2006 #permalink

I read the AP article to say that the group "hopes" to develop the ability to get a handle on location-based probabilities of climatic phenomena. They don't yet know if they can do it. Current climate models can't. But it's scientific projects like this that change that.