Why weather forecasters (like me) often seem to be wrong

The Challenges of Weather Forecasting

Sometimes, while walking through a supermarket, a shopper will approach me in the aisle. “I hosted a barbecue on Saturday and you told me it was going to rain,” they might say. “And it didn’t. Why did you get it wrong?” Or the opposite: they planned for a day of sunshine, only to be disappointed by grey skies. A parent might ask me in March what the weather might be like for their son’s wedding—scheduled for September.

These interactions are always friendly, and they highlight an interesting truth about weather forecasting. Over my career, which has spanned three decades, forecasting has improved dramatically. We can now predict the weather with much higher accuracy and more detailed granularity than when I started in the mid-1990s.

Liz Bentley, a professor of meteorology at Reading University and chief executive of the Royal Meteorological Society, notes that a one-day forecast is correct over 90% of the time. However, despite these advancements, there are still gaps in public trust.

A recent survey by YouGov found that 37% of British adults said they didn’t trust the weather forecast “very much” or “at all.” On the other hand, 61% expressed trust in forecasters like myself. Jokes about weather forecasts are common. For instance, during the 2012 Olympics opening ceremony, a clip from 1987 was shown where Michael Fish, a weather forecaster, assured viewers there wouldn’t be a hurricane, only for a storm to hit hours later. Although Fish was technically correct, the incident became a symbol of forecast errors.

Great Accuracy and Great Expectations

Part of the challenge lies in rising expectations. In our world of constant access to information, we can adjust our fridge temperature or detect car issues via smartphones in seconds. So why can’t we know with 100% certainty whether it will rain on our street at 2pm on Sunday?

Another factor is how the data is communicated. Meteorology generates a vast amount of data, which is difficult to condense into a concise, TV or app-friendly prediction. Even if we’re technically correct, some viewers may still feel confused.

The science itself is complex. It’s a delicate field where even minor inaccuracies in initial data can lead to significant discrepancies. Every day, across the British Isles, forecasters collect data from over 200 weather stations run by the Met Office. This data is then fed into mathematical models using powerful supercomputers.

Earlier this year, the Met Office introduced a new cloud-based supercomputer, marking a shift from physical machines. This advancement is expected to improve forecasts and support climate research globally.

However, as with any science, there are limitations. The atmosphere is a chaotic system, meaning small errors in data can drastically alter predictions. This concept is known as Chaos Theory or the Butterfly Effect, where a butterfly flapping its wings in Brazil could influence the atmosphere in northern Europe days later.

The Challenge of Small-Scale Predictions

Predicting weather over small geographic areas remains difficult. In the 1990s, weather events needed to be larger than 100 miles before they could be observed. Today, the UK-wide model used by the Met Office can map events as small as 2 miles. However, predicting phenomena like heavy fog, which affects only a 1km area, is still challenging.

Even with technological advancements, glitches can occur. Last autumn, a website displayed impossibly high winds and temperatures, which were later corrected. These rare incidents can erode public confidence.

The Art of Communicating Uncertainty

One of the biggest challenges in my job is synthesizing complex data into digestible television segments. Scott Hosking, a director of environmental forecasting at the Alan Turing Institute, notes that weather forecasting is unique in how it's judged by the public. It’s as complex as nuclear fusion physics, yet most people don’t encounter it daily.

Forecasters now use ensemble forecasts, running multiple models to assess uncertainty. If all models align, confidence is high; if they differ, confidence is lower. This is why apps often show a 10% chance of rain.

Rethinking Forecast Communication

There’s a growing need for more creative communication methods. Dr. Hosking suggests moving away from percentages and using a "storyline approach," such as comparing current conditions to past events. The Met Office, for example, names storms to make them more relatable.

Prof. Bentley argues that numbers are powerful and that American consumers are comfortable with percentage-based forecasts. However, AI is set to revolutionize weather prediction. Google DeepMind’s AI models can predict weather 15 days in advance, and Cambridge University’s Aardvark Weather program uses significantly less computing power than traditional models.

Despite these advances, AI has limitations. It relies on historical data, making it difficult to predict unprecedented events like record-breaking temperatures. Prof. Turner acknowledges this challenge and is working on solutions.

The Future of Weather Forecasting

In the future, forecasts will become more detailed, focusing not just on whether it will rain but also on its impact—on travel, gardening, etc. This "so what" factor aims to provide practical advice, such as suggesting a barbecue at lunchtime to avoid afternoon rain.

Audiences are increasingly interested in understanding the science behind weather. Viewers want to know why heatwaves occur or why climate change leads to larger hailstones. As AI improves, forecasts may become longer-term and more accurate, potentially allowing me to give better answers to parents planning weddings months in advance.

While AI offers exciting possibilities, it also presents risks, such as overwhelming users with too much data. Human forecasters will remain essential in communicating weather clearly and effectively.