Highest resolution weather model

The highest resolution forecasting model is the IBM Global High-Resolution Atmospheric Forecasting System (GRAF), which subdivides its global model into cells that cover an area of as little as 3 km2 (1.16 sq mi). This model was created by IBM (and its subsidiary, The Weather Company) in partnership with the National Center for Atmospheric Research (all USA), and unveiled on 8 January 2019.

As a general rule, doubling the resolution of a forecasting model requires a tenfold increase in computing power, so even small increases in resolution represent major investments in hardware and software optimisation. GRAF achieves its record-setting resolution though a mixture of sheer computing power and resource-efficient design. In particular, it uses cells of varying sizes for different parts of the globe – over the oceans, for example, where detailed information about weather patterns isn't needed, the cells are an average of 12 km2 (4.63 sq mi), only dropping to 3 km in areas over land where high levels of detail are needed.

Getting the resolution down to 3 km2 is notable because this means that the model can resolve the behaviour of individual storm cells, rather than providing an aggregate estimate of the behaviour of multiple potential storms in a given area.

The next highest resolution model is operated by the European Centre for Medium-Range Weather Forecasts at their computing centre in Bologna, Italy. It has a resolution of between 18 and 9 km (11.18 and 5.59 mi). Scientists from the Korea Meteorological Administration have discussed the possibility of using their Guru and Maru supercomputers (the most powerful dedicated weather forecasting computers) to increase the resolution of forecasts to 5–8 km (3.11–4.97 mi) in key areas, but these model upgrades do not appear to have yet gone live.