Multi-grid methods for speeding up weather forecasts
The Met Office uses the Unified Model (UM) to give 5 day forecasts of the weather, every 6 hours. Central to the UM is its dynamical core, which is used to solve the equations of large-scale fluid flow in the atmosphere on a computational mesh. A computationally very expensive part of the dynamical core is the pressure correction at each point during a timestep (which is needed to ensure mass conservation).
To do this it is necessary to solve a very large number of coupled linear equations. This computation is a major bottleneck in the solution process, and takes a significant amount of time.
Recently, a collaboration between the Met Office and the IMI secondee Eike Mueller led to the development and implementation of a more efficient solution method for the pressure correction equation in the UM based around the use of multi-grid methods.
The key idea is to use conditional coarsening for the multi-grid iterations in the North-South direction while increasing the mesh spacing uniformly in the East-West direction. This new approach reduces the total runtime of the model significantly and has been operationally implemented, leading to more accurate weather forecasts in a shorter time. The new multigrid solver will allow higher resolution forecasts to be run in the future and it is already allowing better utilisation of Met Office supercomputer resources.
The IMI secondment of Eike Mueller enabled him to spend more time at the Met Office to develop these ideas and, accordingly, improve codes of the project partner.
In August 2020 the Met Office issued a Press Release describing how the use of multi-grid solvers had reduced the time taken to make the weather forecast