A Professor I know wants to harness volunteer Internet computers and spare night-time unused classroom computers to run climate and weather programs to better predict the weather. It is kind of like Seti only with an immmediate purpose in mind that may help mankind short and long term.
There are two areas that are separate, but are mutually influential, climate and weather. Predicting climate is much different than and separate to weather, but eventually the two studies may intermingle.
With this harnessing of large-scale distributed, "individual" programs running in background, it may be possible to get accurate, really long term weather predictions for locales. (up to 30 days.) The algorithms are already known to do these kind of predictions but are not fully tested yet at scale. One of this professor's areas of study involved working with flows to analyse causes of turbulence in pipelines carrying fluids. This study determined the cause and cure of turbulence and bubbles in artificial hearts.
The formulae or types of algorithms which were used to predict these flow changes in arteries, can also in part be used to predict the flow, expansion, heating, humidification, dessication, and travel of discrete air masses which travel over ground and interact with other air masses -- which is the phenomenon we percieve as weather. This computational technique is a kind of "finite element analysis of a changing body with flow."
The idea algorithmically is to select several contiguous, say, one kilometer square air masses, and in a mathematical model, subject them to the changes and inputs of the environment that are observed i.e.heat, humidity, ground conditions and windspeed to start, thence predict the air mass reactions according to physical formula and the predictable path over affecting ground conditions (mountains, heated fields, lakes), the movement and subsequent reactions to these inputs, -- and also the interactions with adjacent air masses -- and therefore the subsequent changes in movement and character of the air flow over time. i.e. to see the changing air flow and its "weather effects".
Some of the possible benefits would be to warn people of impending hurricanes, the path of them, possible tornadoes, unusual conditions at sea or in lakes, heat and smog waves, and other weather conditions that may affect safety, comfort or travel plans.
It is thought that accurate ten day predictions for certain locales, say as large as Toronto, may be possible with a good distributed program and perhaps as few as few ten-thousands of computers. Really accurate nation-wide weather prediction could be achieved with about one million harnessed computers. A Popular Seti-like program could achieve this. Think of it ... who would not want to know exactly what weather they would have next weekend at the cottage? Right now, nobody can tell you. But the means to do so is there.
As far back as 1950 it was known that with sufficient computing power, the movement and behaviour of air masses could be predicted. They behaved in certain known ways in interaction with other masses and the ground they travelled over. But the computational power to put in all those geographic conditions, the individual starting air mass conditions, and minutely catalogue and track their interactive recursive changes according to proper formula needed more computer than had been built in the world, running faster and longer than was practicable.
But today it can be done. All it needs is popularization of the use and viability of the concept.
We could call it Weti. Weather Extrapolation and Terrestrial Information ....
What is needed ...
1. A VB, C++ program and Linux GCC program that works on data in background on windows computers. The program mails or ftp's up results and gets more data when it is done. Also adapts to new algorithms by loading them down.
2. Volunteers to interact with and maintain the master program that collates data and examines it, and refeeds data and results to the appropriate distributed computers.. dummy data and overlap would be needed. This takes overall program design for redundancy and data-results handling.
3. Team to telecommute and decide algorithm trials, and overall strategy. First pass may be testing algoriths and data sets. Then test distribution to improve scale, focus and "grain size" of trial methods.
4. Group to work with standing clusters at high schools and universities that are wired and dedicated. This could be used to test methods.
5. Website and chat group to popularize progam and its distribution.
6. Wet mad, rained out, dehydrated, smogged in, or tornado homeless people to put energy into the program's beneficial nature and get people to use it and donate resources.
Anybody who has any ideas may mail me about this at
mailto:echarters@sympatico.ca
The program needs an organizational group meeting, and about ten to 20 volunteers. Sweatshirts, ballcaps, mugs and wall plaques may be provided. Government grants and honorariums may be available as this sort of project can see real use. It has the backing of professors of some academic reputation.
(It strikes me that algorithms that did pattern prediction, regression or cycles analysis could be overlapped here to see if they correlate. Also, factor analysis could be used to see what inputs were more important.. lots to consider in progamming.. Maybe a spin off would be a program where everyone could know what the market would do tomorrow! :) Of course if they knew, it wouldn't do that. )
EC<:-} |
