1: \begin{abstract}

2:   Approximation/interpolation from spaces of positive definite or

3:   conditionally positive definite kernels is an increasingly popular

4:   tool for the analysis and synthesis of scattered data, and is

5:   central to many meshless methods. For a set of $N$ scattered sites,

6:   the standard basis for such a space utilizes $N$ \emph{globally}

7:   supported kernels; computing with it is prohibitively expensive for

8:   large $N$. Easily computable, well-localized bases, with

9:   ``small-footprint" basis elements -- i.e., elements using only a

10:   small number of kernels -- have been unavailable. Working on

11:   $\sphere$, with focus on the restricted surface spline kernels

12:   (e.g. the thin-plate splines restricted to the sphere), we construct

13:   easily computable, spatially well-localized, small-footprint, robust

14:   bases for the associated kernel spaces. Our theory predicts that

15:   each element of the local basis is constructed by using a

16:   combination of only $\mathcal{O}((\log N)^2)$ kernels, which makes

17:   the construction computationally cheap.  We prove that the new basis

18:   is $L_p$ stable and satisfies polynomial decay estimates that are

19:   stationary with respect to the density of the data sites, and we

20:   present a quasi-interpolation scheme that provides optimal $L_p$

21:   approximation orders. Although our focus is on $\sph^2$, much of the

22:   theory applies to other manifolds -- $\sph^d$, the rotation group,

23:   and so on. Finally, we construct algorithms to implement these

24:   schemes and use them to conduct numerical experiments, which

25:   validate our theory for interpolation problems on $\sphere$

26:   involving over one hundred fifty thousand data sites.

27: \end{abstract}

28: