IAA – Accelerating climate- and weather-forecasts with faster multigrid solvers

The overarching aim of this project was to implement novel multigrid solvers in the Met Office climate- and weather forecast models.

The success of the Met Office models relies on fast numerical methods for the solution of very large systems of equations; this can account for 25%-50% of the model runtime. Forecasts are generated under tight time constraints and efficient algorithms are crucial. For its next generation forecast model (codenamed “LFRic”) the Met Office is exploring a so-called “hybridised” solver algorithm, which has shown great promise but only achieves its full potential when combined with multigrid techniques developed by the PI. A recent publication [Betteridge et al., JPC Vol 426, 109948 (2021)] demonstrates the efficiency of the hybridised multigrid approach for an idealised model system and a related discretisation.

The main purpose of this IAA project was to generate impact by accelerating solvers with hybridised multigrid methods in the next-generation “LFRic” code. In addition, it was also planned to improve the multigrid solvers in the current Met Office model.