Functional estimation of spatiotemporal heterogeneities

Heterogeneity analysis is one of the most complex concerns in spatiotemporal statistics, due to the different sources of variability present in real phenomena. In this paper, a functional approximation to the multiscale structure of a certain class of non‐homogeneous (non‐stationary) spatiotemporal processes, based on a multiple‐window atomic decomposition, is studied. The multiscaling exponent, defining the local regularity and correlation of the processes, is estimated from such a decomposition. Copyright © 2007 John Wiley & Sons, Ltd.