Study of Spatial Relationships Between 2 Sets of Variables Using Multivariate Geostatistics

This paper describes a method based on multivariate geostatistics and redundancy analysis for studying spatial asymmetric relationships between two sets of variables. The method involves fitting a linear model of coregionalization to all experimental variograms, analyzing coregionalization matrices which describe relationships between variables at different spatial scales, and cokriging linear combinations of the predictor variables that account for most of the variance in the set of dependent variables for a given spatial scale. The procedure is illustrated with an analysis of relationships between soil and vegetation variables. Because it takes into account the regionalized nature of the variables, the geostatistical approach allows to distinguish between local relationships and regional relationships due to the presence of different soil types.