Probability-based forecasts can better inform weather-based decision-making, according to new Met Office research.
For the first time, the Met Office has brought together 25 years of research to explain why probabilistic forecasting is becoming central to UK weather prediction.
The research, funded by Public Weather Service and published in the Royal Meteorological Society’s Weather, is a comprehensive analysis of how probability-based forecasts, which capture the inherent uncertainty of predicted weather patterns, can provide enhanced forecasts that better inform decision-making.
In addition, the peer-reviewed research discusses how public understanding of probability-based forecasts should not be a barrier to uptake as has been previously thought.
Uncertainty in forecasting
Probability forecasts, based on ensemble forecasting as it’s referred to in meteorological science, is a fundamentally different approach to many weather forecasts that appear on TV.
While on traditional broadcasts, presenters stand in front of a map and show a single (deterministic) projection of future weather patterns, ensemble forecasts use slightly different starting conditions to run the forecast forward multiple times. This approach produces an ensemble forecast – typically 20-50 simulations – offering a richer and more nuanced picture of possible outcomes.
The graph gives a representation of how ensemble forecasting works. Lines closer together indicate greater certainty, while large deviations represent uncertainty in possible outcomes.
The Met Office has pioneered use of ensemble predictions, initially for month-ahead predictions, with research as far back as 1986 on running the forecast multiple times to better understand the likely scenarios for the atmosphere. Just small changes in starting conditions can result in big changes to forecasts, usually at longer ranges but sometimes at short ranges, so an ensemble forecast is a method of capturing and communicating that uncertainty.
Author of the new research, Met Office Science Fellow Ken Mylne, said: “Ensemble forecasts have often operated as a supplementary system for meteorologists, running alongside single deterministic model runs to provide a measure of uncertainty.
“However, studies over many years show how ensembles provide better predictive skill than single deterministic runs and could, with greater focus on ensembles, capture the range of uncertainty to provide the public with the information they need to make better decisions.”
Understanding probability in forecasts
The research also addresses whether people can understand uncertainty in forecasts, something that is crucial to realise full benefits from a different approach to producing a weather forecast.
Communicating uncertainty is often done by presenters with their language, though the most common way uncertainty is communicated through many apps is the ‘% chance of rain’ that is displayed. This is most people’s interaction with ensemble forecasts, communicating the percentage chance of rain falling in a particular hour or day, informed by model runs which capture that uncertainty.
Example ideas for how precipitation probabilities could be presented clearly for rapid assimilation and good user comprehension.
Ken explained how previous assumptions around a lack of understanding of probabilities may not be true. He said: “Most previous discussions on expressing probabilities in forecasts started from an assumption that they can be hard for people to understand and that expressing uncertainty could undermine people’s confidence in the forecast and therefore undermine their ability to make decisions.
“However, this research suggests that this assumption is wrong. People can understand probabilistic forecasts and could indeed find it more useful for informing weather-based decisions.”
The future of forecasting
The Met Office Research and Innovation Strategy highlights a move to include forecasts inclusive of uncertainty at the heart of endeavours, with an aim to expand use of ensembles in research, operational forecasts and warnings. This will allow users and customers to make even better-informed risk-based decisions on all timescales.
Met Office Blended Probabilistic Forecast data, which consists of a method of turning ensembles into user-ready forecasts, has already been rolled out across the Met Office website and app.
In addition, Met Office in-depth forecasts available on YouTube, such as Deep Dive and 10 Day Trend, regularly explore ensemble forecasts, explain nuances and get to the core of expressing uncertainty in forecasting to better inform decision making.
Met Office Chief Information and Data Officer Charles Ewen said: “These papers bridge the gap between cutting-edge research and developing an understanding with customers and users about what it means for them.
“It’s crucial to educate and inform people that any prediction has in-built uncertainties, but our use of ensembles is a method to calculate and communicate uncertainties in a useful way. This allows our users; public, industry and government, to make better decisions based on that uncertainty.”
