Hannah Christensen receives a £1m Leverhulme Research Leadership Award

The computer models used to make weather forecasts are also needed to make climate predictions. There is an old Danish saying: "it is difficult to make predictions, especially about the future". Nowhere is this more evident than the practice of weather prediction.

Millions of us check the forecast every day and make decisions based on what we see. Should we take a raincoat? Suncream? Plan a BBQ for the weekend? A range of industries also rely on weather forecasts to make decisions, ranging from the energy sector predicting renewable yields to local councils deciding whether to grit the road.

It is this acting on the predictions that makes us very aware when the predictions go wrong. Yet weather forecasts are made using sophisticated computer models run on some of the world's biggest supercomputers. The physical laws which govern the movement of the air around the planet, and the way it interacts with the oceans and the land among other things, are encoded into a computer model – a simulator of the planet – and the simulation run forward in time to predict the future.

So what leads to this all-too-frequent divergence between forecast and reality? This is one of the big questions which will be addressed by Hannah Christensen and her research group in a new million-pound project funded by the Leverhulme Trust. Dr Christensen explains why the forecasts are not exact; and that, despite this, they can be used to improve climate predictions.

There are two main reasons why a forecast might be wrong. The first is that the atmosphere is chaotic - this introduces an element of unpredictability. We should therefore make our predictions probabilistic, indicating the chance of some event happening, to express how certain we are in the forecast.

The second reason is that the computer model is just that – a model. Different approximations have been made when building the model which can lead to the model always making the same mistakes. We must unpick and understand these two issues to improve weather forecasts.

There is another important reason to understand errors in weather forecasts. The computer models used to make weather forecasts are also used to make climate predictions.

When it comes to climate, instead of predicting the particular atmospheric conditions on a given day, the simulations produced by the models are used to understand how the statistics of weather patterns might change in the future.

A novel aspect of the project is using weather forecasts to improve climate predictions. The problem with climate predictions is that it is hard to check their quality - we don't want to wait 50 years to see if we were right or not. But by carefully evaluating weather forecasts we can identify ways to improve the underlying computer models, or failing that, to pinpoint aspects of our changing climate that we are more, or less, certain of.

And that's the key point. We will never be able to predict the future perfectly, be it days or decades into the future. But what we can do is reliably indicate, on a case by case basis, how good we think our predictions are.

And why is this important? Because without knowing the chance of a severe storm next week, or the changing likelihood of such events as we look decades into the future, we have no way of responding to the forecast, be this is issuing weather warnings, or changing building standards. Only the user of a forecast knows how much risk they can tolerate. 10% chance of rain tomorrow? – I'd leave my umbrella at home. But if it were 30%, I'd take it just in case.