1: \begin{abstract}

2: The Mat{\'e}rn family  of covariance functions has played a central role in spatial statistics for decades, being  a flexible parametric class with one parameter determining the smoothness of the paths of the underlying spatial field.

3: This paper proposes a  family of spatial covariance functions, which stems from a reparameterization of the

4:  generalized Wendland family. As for the Mat{\'e}rn case, the proposed family allows for a continuous parameterization of the smoothness of the underlying Gaussian random field, being additionally compactly supported.

5: 

6: More importantly, we show that the  proposed  covariance  family generalizes the Mat{\'e}rn  model which is attained as  a special limit case.

7: This implies that    the (reparametrized) Generalized Wendland  model is more flexible than the Mat{\'e}rn model

8: with an extra-parameter that allows for switching from compactly to globally  supported covariance functions.

9: 

10: Our

11: numerical experiments elucidate the

12: speed of convergence of the proposed model to the  Mat{\'e}rn model. We also inspect the asymptotic  distribution of the  maximum

13: likelihood method when estimating the parameters of the proposed covariance models  under  both increasing and fixed domain asymptotics.

14: The effectiveness of our proposal   is illustrated by analyzing a georeferenced dataset of mean temperatures over a region of French, and performing a re-analysis of a large spatial point referenced dataset of yearly total precipitation anomalies.

15: 

16: %and performing a

17:  %re-analysis of  a large spatial point referenced  dataset

18:  %of  yearly total precipitation anomalies.\\

19: 

20:  %The proposed model outperform the Mat{\'e}rn model   in terms of statistical modeling and prediction  and at the same time leads to considerable

21:  %computational savings.

22: \end{abstract}

23: