Niceweatherforit.com

Links to weather sites

An examination of climate and weather prediction computer models.

Links to meterological office web sites.

http://www.nws.noaa.gov/ National weather service for USA.

http://www.met-office.gov.uk/ The UK Met office web site

http://www.rmets.org/education/wweb.php The Royal meterological society

http://www.wmo.ch/index-en.html The World meterologcial society

http://www.ecmwf.int/ European Centre for Medium-Range Weather Forecasts

http://www.cmos.ca/ Canadiaan Meterologcial society

Links to weather sites

http://www.weather.com/ A website on weather in the USA where you can tyope in your zip code.

http://www.cnn.com/WEATHER/ A website on CNN weather

http://www.bbc.co.uk/weather/ BBC weather website

An examination of climate and numerical weather prediction computer models

Climate and Numerical weather prediction computer models are computer models used for forecasting weather or climate using physics equations.

Computers are important in prediction due to the scale of physcis involved.

Climate describes long term meterologcial conditions. Weather describes day to day meterologcial conditions. Models can be used to preidct both, but normally are only used for one.

The barriers to climate modelling development have been poor observation data and the ammount of work needed to calculate the weather or climate mathematically. In the early twentith century observation data was inaequate. Improvments continued throughout the century. The satellie is important in weather predition both of obseravtion and for data transfer.

The amount of physcis equations invloved in weather prediction is one of the major motiviations for the development of computer technology. One of the first non computer weather models developed was by Arrhenius in 1896. Bjerknes in 1904 was importan in realsing that forecasting was a problem of non linear partial differential equations involved in physics that did not produce closed answers. Yet these models took months to do by hand. It has been calculated it would take 64,000 people to do even these primitive calculations at the pace at which weather changes. This shows why computers were needed. In 1950 Von Neuman one of the major inovators of computing technology, helped develop the first forecast using a computer model. The model was found to be as accurate as those using empircal techniques. By the mid fifties the computer techniques were found to be superior to the empirical techniques.

Since the development of Von Neumann's computer computers have grown in power, and prcoessing spped. The weather and climate organization allways try to find the most powerful supercomputers round.

The two methods of representing weather are finite difference and finite element modelling.

Finite difference method can be defined as the grid method. Finite element modelling is more difficult to visualize. The surface generally remains a grid, but the atmosphere is represented in waves and then converted back to grid for calculations.

Supercomputers the affect of the models.

From 1876 vector supercomputers were the most powerful suopercomputers. In the 90s parrallel processinf became the dominant supercomputer architecture, At present the most powerful are massively parallel processing supercomputers.

Supercomoputers do not provide more accurate arithemtic, they just produce the calcluations quicker. That is why the BBC feel no shame in asking ordinary people to use their own computing technology to do calculations.

Supercomputer development can affect which meterological organization has the best model. Yet their are different types of supercomputer. The main battle is between scalar and vector processors. Scalar calculates sequentially, while vector, calculates all the equations as a single entity. In the eighties it was realised parrallel processing was the way forward where the processing work would be shared between numerous different processors.

As stated before difference met offices use either finite difference or finite element modelling. Finite differnce works better on scalr prcoessors than finte element. While finite element works better on vector processors. So the process of which model (or wich company) is most successful on weather prediction depends largely on what type of models run best on what is presenlty the most powerful fashion of supercomuter.

In the late nineties massively parrallel processing computers were the most type of scalar processor supercomputer. While vector parallel proccessing supercomputers were the most powerful vector processors.

In the nineties the finite element modelling techniques were more popular, because finite element ran more successfully on the dominant vector processors, than the finite difference model. So MeteoFrance ansd ECMWF had some of the best weather predictions around. Then MPPs became more dominant so that helped the Met Office take over. In terms of physcis it might be preferable for scalar to be dominant as some suggest that some physics problems cannot be vectorised.

Yet as Massively parallel processing supercomputers start to dominate the market, the finite difference modelling techniuqes becomes preferable, as the finite difference modelling techniques run better on MPPs. I use the term better in terms of speed of processing not accuracy.

Quantum computers will be an option for the future, these have a different processing structure to scalar and vecotr processinfg.

When powerful supercomputers

Climate is a physical process whose actions can be exopressed by linear quations. The equations are non linear partial differential eqautions that represent a non linear system.So calculations can only represent approximations. So there are differences expected in the results from the modelling be run, even on same computer.

The 6 basic equations for GCMs are conservation of momentum, moisture, the first law of thermodynamics, the contuniuity equation, the hydrostatic equation and the equation of state.

Observations for climate and NWP models. Initial state is the state the model starts at. Real state is the state the atmosphere is really in. In climate modelling the difference between initial state and real state is not key as forcing affects cancdel out the affect of the initial state. In weather prediction the intial state is obvisouly key. The process of adjusting the model to the intial state is data assimilation.

Alternatives to Computer modelling

Empirical and analog.

Empircial uses evidence from the past to preidct how weather will react in the future, this is till useful in small scale weather prediction. As valleys are often to small to be resprented on a computer model. As computing tecnolgoy increases empirical prediction will not be so useful.

In climate modeeling the alternative is paleaoclimatic evidence using past climate to forecast the future.

Lonympics

http://www.lonympics.co.uk/Cheapchicago.htm

http://www.lonympics.co.uk/CheapSydney.htm

http://www.lonympics.co.uk/CheapNewYork.htm

http://www.lonympics.co.uk/CheapLondon.htm

http://www.lonympics.co.uk/Jokesfreeforusebyrsmith.htm

http://www.lonympics.co.uk/new/Labour-iraq-war-the-movie.htm

http://www.lonympics.co.uk/CelebrityBlog.htm