Why are weather forecasts not always accurate?

Over the past 30 years or so, forecasting accuracy has been improving in leaps and bounds with the Met Office five-day forecast now as accurate as its one-day forecast was 30 years ago.

However, weather is a ‘chaotic’ system, this means small differences in conditions now, such as a shift in temperature, wind or humidity, can have a significant impact on the weather conditions we might see over the coming days. Even seemingly small discrepancies in the current conditions can lead to inaccuracies that grow as the forecast runs further into the future.

To generate forecasts, meteorologists run computerised numerical weather models which analyse billions of observations to work out how the current conditions will develop into the future. Some weather models can be more or less accurate, some are better at forecasting certain weather types, others at forecasting at different lead times.  Therefore, meteorologists never rely on one run of one weather simulation model but use ensemble modelling in their forecasts. This is where many different computer model runs are used, each starting with slightly different weather conditions. Sometimes different models are also combined to take advantage of their different skills. 

Highly skilled and experienced meteorologists look at these range of outcomes and determine which is the most likely, as well as assessing the risks associated with other possible outcomes.

However, many weather apps and website forecasts, including those of the Met Office, are calculated using automated data taken directly from a combination of weather prediction models without any input from meteorologists. 

This automated approach works very well in most situations and allows us to issue localised forecasts for a huge number of specific locations, but some automated location-based forecasts can struggle under some circumstances, particularly under extreme circumstances such as we saw in July 2022 when temperatures topped 40C at a number of locations.

Based on research into the best use of ensemble modelling, the Met Office will be rolling out improvements to its automated forecasting system later this year.  

These changes will be latest innovation from the Met Office. Predicting weather and its impacts has become one of the most important areas of scientific endeavour and there have been significant advances in forecasting skill over recent years. Technology is changing at an ever-increasing pace, providing huge opportunities for innovation in the field of weather and climate science. Taking advantage of these new capabilities requires new ways of working but is resulting in better forecasts, with longer lead times helping us all to make better decisions to stay safe and thrive and to minimise impact damage. 

Met Office Director of Science and Technology, Professor Stephen Belcher, said "There are changes in our world that we must also respond to. From technological innovation to the need for clean growth, to the change in the climate itself. We need to ensure that our exceptional research and innovation continue to drive improvements in our science, technology and operations – and vice versa." 

The Met Office plays a key role in the development of weather and climate science and is at the forefront of many cutting-edge advances across a wide range of different industries using its weather and climate expertise to push forward innovation in a range of different areas. For example, we are involved in the safe development of autonomous vehicles and unmanned drones, we are supporting government and industry to plan for changes to our climate and the potential impacts from extreme weather. 

From supercomputers to satellites, from TV to VR, technology and innovation influences weather forecasting, climate prediction. As faster supercomputers with more processing power are developed, harnessing this power and speed for the benefit of improving weather and climate projections will continue to underpin the Met Office's capability.