Evidence that the global climate is changing is now unequivocal, but what these changes mean for regional and local weather events, both now and in the future, remains uncertain. Today climateprediction.net, the world's largest climate forecasting experiment, and the Met Office launch a new international project called weatherathome.net, allowing anyone with a computer and internet access to take part and help understand how climate change may be producing damaging – or beneficial – weather events around the world.

Weather, which can change in a matter of hours, contrasts with climate, which describes the trend in weather patterns over much longer time periods.

Since its launch in 2003, climateprediction.net has harnessed computer time donated by hundreds of thousands of volunteers to run Met Office climate models and explore the uncertainties in global climate predictions in unprecedented depth. This new experiment takes results from these and other modelling experiments and uses them to drive much more detailed models simulating the changing weather in, initially, three target regions: Europe, the western United States and southern Africa.

Understanding the impact of climate change on extreme weather eventsis challenging for two reasons. Firstly, the models used for simulating global climate change still do not resolve many relatively small-scale features, such as weather fronts associated with storms and intense rainfall.

Secondly, since extreme weather events are, by definition, rare, many model-years of simulation are required in order to gather reliable statistics. "It's like trying to work out if a roulette wheel is truly fair and random: you have to watch the wheel spin thousands of times to work out if some numbers are coming up more often than they should," explainsMyles Allen at the University of Oxford Department of physics, principal investigator of climateprediction.net.

The new experiment uses a regional climate model which provides information on weather events in much finer detail than is typically provided by global climate models. Because of the cost of computing time, the regional model covers only a specific area and is supplied with values of weather variables – such as winds, temperature and humidity – around its edges, so that it feels the influence of large-scale weather in other parts of the world.

Participants in this new experiment will, therefore, actually be running two models: a global model to simulate large-scale weather, and a regional model to simulate detailed events in a specific part of the world. "Regional models add the detail of how different types of weather will change to the broad-scale changes provided by global models," explains Richard Jones, head of regional modelling at the Met Office, who led the development of the new experiment. "Watching them run you can see how weather fronts and local features such as mountains interact to produce extreme rainfall."

Both models have been developed by the Met Office, and results from different regions will be used directly by professional scientists specialising in the climates of those regions. The European region will be analysed by the Met Office Hadley Centre and Oxford, Edinburgh and Leeds universities, southern Africa by the University of Cape Town, and the western USA by Oregon State University. Results will also be made available to scientists interested in climate impacts in the various regions.

The experiment, which is supported by Microsoft Research, the UKNatural Environment Research Council and the European Commission, is in five parts. The first is a kind of calibration. A large number of different versions of the global and regional models will be used to simulate the period from 1960-2010 using observed changes in sea-surface temperatures, sea ice, atmospheric greenhouse gases and aerosols. The simulated climates and patterns of change in weather events from these models will then be compared with observations over the same period to select a range of realistic versions and document their behaviour. If, for example, it is observed that a particular version of the model tends to over-estimate the number of storms, it is then possible to take account of this when using this version to forecast future changes in storminess.

The second experiment is to produce a forecast of changes in weather events in the 2020s and 2030s. Using the output from many different models with evolving oceans to provide the forecast sea-surface temperatures, the regional model will tell us in unprecedented detail about the potential changes to patterns of weather events through the next three decades. Particularly interesting features are likely to be changes in the likelihood of droughts, floods, heatwaves and cold snaps.

The third experiment runs beyond the 2030s and provides detailed information about changes in weather features in a world that is 2, 3 and4C warmer, globally, than today. This represents a range of climates that might be encountered towards the end of the 21st century or beyond. This experiment will provide some of the most detailed information to date on regional weather in such possible future worlds, which is essential to assess the range of potential impacts of climate change.

The fourth experiment returns to changes seen since the 1960s, and attempts to quantify to what degree these changes can be attributed to the effects of human interference in the climate system. The driving conditions fed into the models are modified to reflect what they would have been like if we had not produced the greenhouse gas and aerosol emissions that we have over the past century. The difference between these simulations and the initial "baseline" runs will provide the basis for assessing the human contribution to recent weather trends.

Finally, the fifth experiment also looks back into the past – looking at snapshots of the weather at intervals over the past 10,000 years, a period of Earth's history called the "Holocene". This is the first time large numbers of regional models will have been applied to such "paleoclimate" simulation: an unprecedented opportunity to explore the evolution of the weather over recent Earth history.

This project is as much about exploring the limitations of current climate models as it is about exploring their potential, and the project team will release the first of two online information packs on the use and interpretation of regional climate models to ensure results are interpreted realistically.

"Climate models are not crystal balls, but they can be useful tools provided people understand the uncertainties," says Neil Massey, technical coordinator of climateprediction.net. Marion Manton, at Oxford University's department for continuing education and project leader of the climateeducation.net initiative, adds: "Anyone can use these information packs to learn about how climate simulation works, and potential users can explore how – and how not - to use our results in planning how to adapt to climate change."