Uncertainty in weather and climate prediction: A Royal Society meeting discusses what it means and what to do about it
This week, Carbon Brief was at the Royal Society’s international scientific meeting, ‘Handling uncertainty in weather and climate prediction’. Here’s our rundown of the best bits from the two-day conference.
We’re all familiar with weather predictions that don’t materialise quite as we expect – who could forget the Met Office’s barbecue summer that wasn’t of 2009? But in general, weather and climate forecasters do a pretty good job, considering how complicated the climate system is.
In the past, weather forecasts were based on combining limited observations with scientists’ knowledge of how the climate system behaved. These ‘deterministic’ predictions vastly oversimplified things and weren’t very reliable. But climate research moves on quickly and forecasters now have complex computer models that mimic the earth’s climate system far better. But there are still parts of the system that we don’t yet understand or can’t fully capture which means that the models can never be 100 per cent accurate.
The further in advance the forecast is, the more difficult it is to make an accurate prediction. Two to three day forecasts tend to be more reliable than seasonal ones and predicting changes over climatic timescales is even more problematic.
A group of around 70 climate scientists, meteorologists, weather forecasters, business people, economists and mathematicians gathered at the Kavli Royal Society International Centre in Buckinghamshire on Thursday to discuss the finer points of uncertainty in forecasting. In particular, how we can represent uncertainty in weather forecasts and longer term climate forecasts in a way that is still useful for people who rely on them.
Building resilience
Of course, weather prediction has more important applications than knowing if we’ll have a good summer in the UK or not.
The afternoon session on Thursday highlighted the impact of extreme weather events, such as severe flooding and drought, on developing countries. While no single event can be easily attributed to climate change, research suggests that the frequency of some events – notably heat waves and precipitation events – is likely to increase with further global warming.
So it is important for vulnerable countries to be able to anticipate and adapt to such events, which means factoring in the uncertainty that goes with the forecasts. Professor Tim Palmer, Fellow of the Royal Society and organiser of the conference, explained in an interview with Carbon Brief today:
“Even though the probability of [an event] may be only 10 or 20 per cent, it has such an enormous impact on the [country’s] economy that it’s important to take that into account”.
While scientific research into the impacts of extreme events is rapidly advancing, there is still a disconnection between basic research and potential solutions in places that need them, explained Professor Peter Webster from the Georgia Institute of Technology in yesterday’s session:
“A society that learns to deal with present era hazards will be better equipped to deal with those that occur through future climate change”.
Far-reaching impacts
Several of yesterday and today’s speakers showed how weather and climate predictions affect decision makers in the energy, economics, agriculture, trade, health and hydrology sectors. Take energy. As Dr Renate Hagedorn from the German Meteorological Service said in this morning’s session:
“Due to their strong dependence on weather, [solar and wind energy] are fluctuating in time and it is of great importance to anticipate any possible extreme event or rapid change. The biggest challenge…is to ensure the stability of the grid by continuously balancing production and demand of electricity”.
There’s also health. Weather and climate variability have an important influence on the rise and spread of infectious diseases, as Dr Andy Morse from the University of Liverpool explained in his talk this morning:
“Climate is important for many infectious diseases, especially those that are vector-borne, which means transmitted via water or insects like mosquitoes. But not all diseases are affected in the same way”.
The future
Climate research is evolving all the time. So the climate community is working hard to improve forecast models through more observations, better knowledge of the climate system and greater computer power. But until then it’s important to quantify uncertainty as best as possible. Ken Mylne of the Met Office summarised this in yesterday morning’s session by saying:
“We can’t get away from uncertainty because of the chaotic system, so we have to find ways of communicating it”.
Professor Palmer added in his interview with Carbon Brief that now we have the technology to quantify uncertainty, expressing that uncertainty is the key to having confidence in forecasts:
“If governments are making decisions about putting into place emission policies or wind farm managers are deciding how much electricity to contract or something as trivial as an ice cream vendor wants to decide how much ice cream he’s going to sell, taking account of the estimated uncertainty in the prediction is crucial”.
He continued:
“An unreliable prediction is one without adequate quantification of uncertainty. You can make better decisions in a range of applications if you know what the estimates of uncertainty are and this extends all the way through to climate change”.
The main message of the meeting is that we are already doing well in terms of understanding and dealing with uncertainty in weather and climate models and the future looks bright for improving upon those capabilities. Professor Palmer concluded:
“It’s not as if the science of uncertainty estimation is completely done and dusted, there are still many science questions that need profound exploration…but [sharing] the experiences of the user community is actually quite valuable in helping focus [on] what are the most important science questions to improve our ability to estimate uncertainty in the future”.