1: \begin{abstract}

2: This paper proposes a new preconditioning scheme for a linear

3: system with a saddle-point structure arising from a hybrid approximation 

4: scheme on the sphere, an approximation scheme that combines (local) spherical

5: radial basis functions and (global) spherical polynomials. Making use

6: of a recently derived inf-sup condition \cite{SW08} and the Brezzi

7: stability and convergence theorem for this approximation scheme,

8: we show that the linear system can be optimally preconditioned with a

9: suitable block-diagonal preconditioner. Numerical experiments with a 

10: non-uniform distribution of data points support the theoretical conclusions.

11: 

12: %The combination of (local) radial basis functions and (global) polynomials

13: %for approximation on the sphere is considered. Such a hybrid approximation scheme

14: %is naturally described in terms of a saddle point formulation. The inf-sup

15: %stability of this system is considered and optimal approximation is established

16: %so long as the mesh norm multiplied by the polynomial degree is sufficiently small.

17: %It is shown how this underlying stability leads to optimal preconditioning for the resulting

18: %saddle-point linear systems.

19: 

20: 

21: %The combination of (local) radial basis functions and (global) polynomials

22: %for approximation on the sphere leads to a saddle point linear system

23: %which needs to be solved for the relevant expansion coefficients. In this paper we 

24: %describe preconditioners for use

25: %with iterative methods for the solution of these systems.

26: %Based on the

27: %underlying stability and convergence of the approximation method, we

28: %are able to prove optimality of the resulting preconditioned iterative

29: %methods, provided the mesh norm multiplied by the polynomial degree is sufficiently small.

30: \end{abstract}

31: