How Accurate A Long Range Weather Forecasts?

Weather Forecast Chart
Weather Forecast Chart
Weather Forecast Chart

Weather forecasting technology has improved significantly over the last few decades. The demand for accurate weather forecasting has resulted in invested in further improvements in research, data centres and supercomputers. For example, the Met Office plan to build a £97m supercomputer accelerate advances in their weather forecasting and climate modeling.

The new facility will operate 13 times quicker than its predecessor and pave the way for a UK forecast model that runs every hour (instead of every 3) and focuses on a resolution of 1.5 km. The new building will become operational in September 2016 and is being build in Exeter, United Kingdom.

Met Office Super Computer
Met Office Super Computer

New Age of Forecast Accuracy

The Met Office CEO, Rob Varley, spoke of a “step change” in forecast accuracy adding that:

“It will allow us to add more precision, more detail, more accuracy to our forecasts on all time scales for tomorrow, for the next day, next week, next month and even the next century,” said Met Office chief executive Rob Varley.

Will The Accuracy of Long Range Weather Forecasts Improve?

The average five-day forecast is as reliable as a two-day forecast 20 years ago. However, long range weather forecasts (made two weeks or more in the future) are considered impossible to predict (Edward Lorenz, 1963) because of the “chaotic nature of the fluid dynamics equations involved. In the models used for weather forecasting, extremely small errors approximately double every 5 days for temperature and wind velocity.

The weather forecasting at the UK Meteorological Office uses multiple computer simulations with slightly different parameters. In the event that the different simulations agree, there is a statically significant chance that the forecast is correct.

However, this approach is only effective when the simulation is over a short period of time. As the simulations are extended there is more divergence in the results. As well as it being significantly less likely to find agreement among the simulations, where there is agreement it has found to be unreliable.