This website uses cookies. Read about how we use cookies.

  1. As the national hunt racing season draws to an end Kevin McConway, Emeritus Professor of Applied Statistics at The Open University, helps to explain the nature of probability and how weather forecasting and horse racing are unlikely partners when it comes to beating the odds.

How accurate are our public forecasts?

Rainbow storm

We have an open and transparent policy on how well our public weather forecasts are performing. See how we are doing against our targets.

Global comparisons

The World Meteorological Organization compares similar statistics among national meteorological services around the world. These show that the Met Office is consistently one of the top two operational services in the world.

Research confirms that accurate forecasts are valued highly and improvements in accuracy enable society to take a wider range of weather related decisions with more confidence. We monitor the performance of our forecasts by looking at trends in accuracy over the longer term.

Verification involves comparing our forecasts against relevant observations to determine how accurate they are. Verification, therefore, underpins the targets set by government and the Public Weather Service Customer Group. These forecast accuracy targets are part of our ongoing long-term commitment to improving the accuracy of our forecasts to the public. 


Forecasts for both maximum and minimum temperature are compared to the actual values observed at an agreed list of 119 sites across the UK. The sites used for verification are those where we have quality-controlled data and where we produce forecasts for.

The early morning forecast on our website is used to produce a percentage number of the times when the forecast is accurate to within +/- 2°C. This is based over a rolling 36-month period to smooth out extremes and give a representative average.

Temperature forecast - performance

This information will be updated every month.

Minimum temperature - first night of forecast

  • 86.5% of minimum temperature forecasts are accurate to within +/- 2°C on the first night of the forecast period (36-month average).


Maximum temperature - second day of forecast

  • 92.4% of maximum temperature forecasts are accurate to within +/- 2°C on the next day (36-month average).

Three-hourly temperature

Three-hourly temperature forecasts are monitored in the same way as maximum and minimum temperatures, but using our three-hourly forecasts for the current day.

  • 95.1% of three hourly temperatures are accurate to within +/- 2°C on the current day (36-month average).


We also monitor how well we forecast weather types, e.g. 'rain' or 'sun'.

The forecast weather type is compared to the actual weather observed at the same agreed list of 45 stations across the UK. The early morning website forecast is used to produce a percentage number of the times when we correctly forecast the weather as 'rain' or 'sun'. This is based on our three-hourly forecasts for the current day and is based over a rolling 12-month period, as it is a relatively new measure.


  • 0.530 Equitable Threat Score (ETS) of three hourly weather is correctly forecast as 'rain' on the current day.
  • It is difficult to forecast rainfall at a pinpoint location, therefore the target for this element is not as high as for others. For example, one location may have rain and a location a mile away may remain dry.


To verify the 'rain' weather type, any hourly observations of precipitation will count towards a 'rain' observation.

The following symbols would be included in the definition of 'rain':

Light rain shower (night) light rain shower (night) Heavy rain shower (night) heavy rain shower (night)
Light rain shower (day) light rain shower (day) Heavy rain shower (day) heavy rain shower (day)
Drizzle drizzle Heavy rain heavy rain
Light rain light rain  


  • 0.537 Equitable Threat Score (ETS) of three-hourly weather is correctly forecast as 'sun' on the current day.


To verify the 'sun' weather type, hourly observations of cloud cover of between 0 and 2 oktas count towards a 'sun' observation.

Comparing forecast accuracy

We continue to increase our forecast accuracy through research, investment in satellite remote sensing and supercomputing technology.

For many years we have verified our forecasts by comparing forecasts of mean sea-level pressure with subsequent model analyses of mean sea-level pressure. These comparisons are made over an area covering the North Atlantic; most of western Europe, and north-eastern parts of North America. From this long-term comparison an average forecast error can be calculated.

The graph shows how many days into a forecast period this average error is reached compared to a baseline in 1980. This graph shows that a four-day forecast today is more accurate than a one-day forecast in 1980.

Last updated: Feb 15, 2017 9:55 AM