Science communication is a hot topic in the media. The accuracy of weather forecasts, in particular long-range projections, is often found to come under fire. In forecasting there is no single future eventuality we can be certain will occur. There are always uncertainties associated with the observations and models used to make predictions. Nowadays ensemble based methods are widely used to help explore the effects of these uncertainties on the forecast. An ensemble of many forecasts is run which all start from slightly different initial conditions. The starting points are carefully chosen so that the evolutions of the forecasts represent the range of possible outcomes. All the forecasts are then used to calculate probabilities of certain weather events occurring.
Verifying a probabilistic forecast is no easy task. Each forecasted event has a yes or no outcome, so can we say an 85% chance was a skilful prediction if the event occurs? To come up with a meaningful measure we look at the performance of the model over many forecasts. If the event occurs on average 85% of the time that the forecasted chance was 85% we can deem the forecast to be reliable. This will of course mean there are at times ‘false alarms’, when the forecast predicted a high chance of occurrence but the event does not happen in reality.
A well known example of communicating uncertainty in scientific understanding is the IPCC’s ‘likelihood’ scale, with each likelihood category based on a percentage threshold. Ideas of this kind go back many years, a Monthly Weather Review article by W. E. Cooke in 1906 proposed a 5 point scale for describing forecast uncertainty. Of course a system of this kind is far from perfect and individual interpretations of a given phrase will be wide ranging. An interesting review on the history of probability forecasts can be found here.
The World Meteorological Organisation (WMO) says that ‘communicating the uncertainty of the forecast is vital to users. It allows them to make better decisions that are attuned to the reliability of the forecast. It also helps to manage the expectations of users for accurate forecasts.’
Aside from communication aspects, the WMO also make clear that verification is an integral part of the forecast process. If we have any hope of providing clear and useful information to the public we need to understand the models we use and how reliable they are at making the predictions we seek. Increasing media coverage and public interest has highlighted the need for us all to give lucid explanations, whether it be describing the coming weekend’s weather to your parents or discussing future climate change with policy makers. Why don’t you take Andrew’s survey on the communication of seasonal forecasts and have your say?
‘If a man will begin with certainties he shall end in doubts; but if he will be content to begin with doubts he shall end in certainties.’ Francis Bacon