In this paper we describe a generalization of the Shapiro and Botha <1991> approach for the case of anisotropy in two components with a view to obtaining flexible stationary variogram models suitable for modeling spatiotemporal dependency. Additionally, instead of choosing the empirical semivariogram estimator, the use of a nonparametric kernel estimator of the semivariogram is proposed. The work also includes an application to wind speed data and a simulation study. It can be observed in the simulations that the use of flexible models avoids problems of bad specification that may arise when parametric models are used and, in addition, that the local linear semivariogram estimator produces significantly better results than those obtained with the empirical estimator.