Weather forecast models and apps

Rupert Holmes unpicks the tech behind the data that underpins all weather apps and highlights his favourite software

- Rupert Holmes

To understand the forecasts provided by different weather apps and websites we first need to have a basic appreciation of the data they present. Obviously they can only be as good as their data inputs, but where does that come from?

The answer lies in the raw data output of numerical weather prediction (NWP) models. We therefore also need to understand the inherent strengths and weaknesses of these forecasts.

As there aren’t that many different models that provide free or relatively low-cost data, often different apps or websites are simply presenting the same data in different ways.

Understanding numerical weather prediction

The key premise behind weather modelling is that the mathematical equations governing motion, fluid dynamics and thermodynamics can be used to predict the future state of the atmosphere.

Key variables within each model include wind strength and direction in three dimensions, air density, temperature, pressure and humidity.

This makes it a hugely data-intensive process, which necessitates dedicated use of some of the world’s most powerful supercomputers, which are often replaced on two- or three- yearly timescales. Even then, a single model run can take up to six hours to process.

The accuracy of model output depends on many factors including the size of the horizontal grid and the number of vertical atmospheric ‘slices’ considered.

A feature, whether meteorological or geographical, needs to cover at least three grid points in order to register in the model. Therefore, a model with a 22km grid will ignore weather features that are less than 65km across.