A new regionalization method for spatially dependent functional data based on local variogram models: an application on environmental data

Regionalization is to divide a set of spatial objects into a number of spatially contiguous regions while optimizing an objective function, which is usually a homogeneity (or etereogenity) measure of the derived regions. In this paper we propose a new regionalization method for spatial functional data by means of a Dynamic Clustering Algorithm. The method optimizes a criterion of spatial association among functional data, it is such that the prototype is a trace-semivariogram model chosen from a set of mathematical functions that describes spatial relationships. The appropriate model is chosen by matching the shape of the curve of the experimental variogram to the shape of the curve of the mathematical function. Performances of these methods are checked through an application on real data.