A New Dynamical Core for the Unified Model: Conservation Properties, Model Grids and Time-Stepping

A test-bed for investigating three essential aspects of a new dynamical core for the Unified Model will be established, building on the current NCAS Microscale Model. The features to be investigated are the conservation properties of the equations and their discretisation, the computational grid and the method for advancing the model in time without exciting acoustic modes. The new approaches will be evaluated and compared with existing approaches, paying particular attention to accuracy and speed in massively parallel implementations. A new implementation of the Microscale Model will be based on a re-casting of the model equations in conservative form; this is likely to be of particular importance for climate modelling. Massively parallel implementations of the Microscale Model will be be created with both rectangular and hexagonal horizontal grid structures and the efficiency of these will be evaluated. In addition to the current split-explicit time-stepping version of the model, a second version will be established using a Helmholtz solver for determination of the pressure, allowing much larger time-steps but simultaneously creating challenges for efficient parallel implementation. The outcomes of these evaluations will feed into the design of the next generation Unified Model.