All enjoyable things must come to an end and so it is that I find myself thinking what to write about in my last article for theWeather Club. Over the years I’ve covered all things hot to cold, calm to stormy, hurricanes to tornadoes, so I thought it was about time I wrote about communicating the weather, or the weather forecast to be more precise. It’s all very good making an outstanding forecast, but if you cannot get the forecast across to the user of the information then to all intents and purposes it is worthless, at least to that user.
Let’s be fair, communicating weather forecasts is a tough job. I remember once, many years ago, we decided to do a public survey. It was on using probability in forecasts - you know the kind of thing: tomorrow there will be 30% chance of rain. On continental Europe and in the US they have, for many years, used probabilities, but we have done much less in the UK. Many obvious technical things spring to mind. First how do you tell if probability forecasts are accurate? You can, but it’s more complicated than checking if a ‘rain or no rain’ forecasts was correct. On the other hand the gains are much greater, if you think about it. There is no way forecasts will ever be 100% accurate (we wish) so communicating the uncertainty is very helpful. For example, if you say you say 95% chance of rain compared to 50% chance of rain that tells you something about the levels of risk. It raises another question about what probability forecasts are telling you, that is, what does 50% chance of rain actually mean? One thing it doesn’t mean is that the forecaster flipped a coin and is hedging their bets. A better interpretation would be that in these types of weather conditions the forecaster (or the computer model making the forecast) would expect to see rainfall 50% of the time.
Sharing uncertainty helps make decisions on what to do because they allow you to personalise the impact to you and your circumstances. And it allows us to share forecasts that are much further out in to the future and might have inherently more uncertainty. However in some instances we want the decision (or at least aspects of it) made for us, such as in weather warnings. So in certain circumstances and on certain timescales you have to resort to more directed information.
The choice of language is of course key. That’s why it is standardised as much as possible, particularly in the forecasts that are safety critical. The shipping forecast for example has a standardised set of terms and the scripts even set to a limit of 350 words. Whereas in the TV forecasts they can afford to be a bit more relaxed and talk about ‘spits and spots’ of rain. It’s not just the words. Who the presenter is, how they are dressed, if there have changed the hairstyle, all have an impact. I remember a fascinating study on this. Basically when we watch the forecast we spend about the first 30 seconds seeing if we recognise the presenter, their tie or dress, and then spend the next 15 + seconds trying to find our town on the map. More than 50% of the people surveyed could not locate their town on the weather map within 30 km. So there is a lot of work goes in to the format of how forecasts are presented and how that changes from TV to radio to the internet.
Back to the survey I mentioned earlier. We showed 1,000 people 4 versions of the same forecast but presented in slightly different ways. One used symbols that were fainter or bolder depending on whether they were less or more likely. Another used the same idea but in words, where the more likely the bigger and bolder the words. The third used percentages, 30%, 40%, 50% etc; the fourth was much more scientific and used error bars. That is we gave forecast values and then lines above and below the value of varying lengths to show the range of uncertainty in the numbers; something used a lot in the presentation of scientific data rarely used in data shown to the public. Can you guess which the 1,000 people voted for. Would you be surprised if I told you the votes we spread almost equally across the four options. The useful lesson I learned was that people are very different in how they absorb information and no one way of doing this will work for everyone. So it’s useful to have some standards, it’s useful to think of the format and layout, and you have to deliver the same information in a variety of ways if it is going to appeal to a wide audience. And to do all of that in ever decreasing amounts of time or space in programming schedules. I’m biased but I think our forecasts in the UK are some of the best presented and communicated in the world.
Now I can’t close without noting the new weather graphics on the BBC website, which had a good test over the winter snow period. I like them a lot and I think the BBC team have done a great job with these, but as with most forecasts presentations it will be the public who have the final say, and that’s as it should be.
I thought I would just leave you with this quote (c. 1800), the curse of all those brave enough to have taken on the task of communicating weather forecasts:
The Duke of Rutland urged the folk to pray for rain;
the rain came down the next day.
The pious marvelled, the sceptics murmured fluke,
the farmers late with hay shouted, ‘Damn the Duke’.
That’s me signing off (… with a distant sound of a mic drop).