1: \begin{abstract}

2:   The Met Office's weather and climate simulation code the Unified

3:   Model is used for both operational Numerical Weather Prediction and

4:   Climate modelling. The computational performance of the model

5:   running on parallel supercomputers is a key consideration. A Krylov sub-space solver is employed

6:   to solve the equations of the dynamical core of the model, known as

7:   ENDGame. These describe the

8:   evolution of the Earth's atmosphere. Typically, 64-bit precision is

9:   used throughout weather and climate applications. This work presents

10:   a mixed-precision implementation of the solver, the beneficial effect on

11:   run-time and the impact on solver convergence. The complex interplay of errors arising from

12:   accumulated round-off in floating-point arithmetic and other

13:   numerical effects is discussed. A careful analysis is required, however,

14:   the mixed-precision solver is now employed in the operational forecast to

15:   satisfy run-time constraints without compromising the accuracy of

16:   the solution. 

17: \end{abstract}

18: