A record audience of 145 was ample evidence of how much everybody is fascinated by the weather - both on a daily basis and also in the long-term, with the prospect of climate change. And the fascinating talk by Peter Gibbs, one of the BBC's leading forecasters, on January 11th offered a wonderful insight into all the complexities of the difficult science of predicting the weather. Peter, who is also co-chairman of Gardeners' Question Time on Radio 4, pointed out that the weather happens on many different scales, both in space and time, ranging from a shower of a short duration to depressions covering hundreds of miles. And answering the simple question - when is it going to rain? - can involve computers making millions of calculations a second as the atmosphere swirls, whilst the Earth continues its rotation.

Just before presenting the weather at 8am on the Today programme on Radio 4, Peter recounted how John Humphrys might suddenly ask him on air about the prospects for a "barbeque summer" or the reality of global warming - all extremely hard to sum up in just a few seconds, and also without "crashing the pips" on the hour, a cardinal sin at the BBC.

One of Peter's heroes is Lewis Fry Richardson, one of the unsung scientists behind early weather forecasting, who published a theory in 1922 in which he applied simple equations of physics and laid down the foundations for the science of modern meteorology. But Richardson was also a pacifist and later abandoned his work when he discovered it was being used to design chemical weapons.

These days, even with the help of powerful computers, it is still far from straightforward to keep ahead of the weather, although new and complex models are constantly being developed to improve forecasting - not only for the next few days, but even for the rest of the century. Chaos theory is a major factor in the process but 10-day forecasts are now routine and this is being pushed out to 15 and there are experiments with monthly predictions. Seasonal forecasts are more difficult for the UK because - unlike with the tropics with the well-defined seasons - factors here are much more variable. But Peter - who is employed by the Met Office - pointed out that existing models make forecasting for the rest of the century better than imagined and it is now possible to estimate the average January temperature in 50 years time. However, long-term forecasting requires the inclusion of mechanisms, like deep ocean currents which transport heat, as well as parameters such as the loss of the Amazon rain forest.

But the $64 million-dollar question - is the climate changing? Peter suggested models tell us the world will get hotter - and hotter air can hold more moisture, implying heavier rainfall. However, these models are not of a fine enough resolution to give regional details and, although it is hard to link a single event to climate change, severe weather is probably more likely.

"Meteorology is such a complex science, with so many variables as the Earth is in a constant state of flux, and yet Peter informed us of the latest advances and predictions with a simple clarity which is his hallmark when broadcasting. Forecasting the weather these days is clearly far more reliable and, with experts like Peter, it is clear that we are in good hands. Another fascinating evening - Vincent Kelly, Oxford.