Use of spatial factors in the analysis of heavy metals in sediments in a Brazilian coastal region

Environmental phenomena often generate complex data which are not only multivariate, but where spatial structure and dependency may also be important. One potential area of interest is that of identifying combinations of variables which exhibit distinctive spatial structure in a set of multivariate and geographically indexed observations, e.g. long, as opposed to short, range spatial contiguity. Such components may serve not only to characterize the main determinants in the geographical distribution of a set of environmental measurements, but may also provide useful insight into the physical or environmental processes underlying and governing that distribution. This article suggests one form of multivariate technique which may be useful in such a context. We demonstrate the method in relation to data comprising measures of various heavy metals in sediments in a Brazilian coastal region. Copyright © 2005 John Wiley & Sons, Ltd.