GEOSTATISTICAL TOOLS FOR CHARACTERIZING THE SPATIAL VARIABILITY OF MICROBIOLOGICAL AND PHYSICOCHEMICAL SOIL PROPERTIES

This paper reviews the main applications of geostatistics to the descr iption and modeling of the spatial variability of microbiological and physico-chemical soil properties. First, basic geostatistical tools su ch as the correlogram and semivariogram are introduced to characterize the spatial variability of each attribute separately as well as their spatial interactions. Then, the key issue of fitting permissible mode ls to experimental semivariograms is addressed for the univariate and multivariate situations. Capitalizing on this model of spatial depende nce, the value of a soil property can be predicted at unsampled locati ons using only observations of this particular property (kriging) or i ncorporating additional information provided by other correlated prope rties (cokriging). Factorial kriging allows one to discriminate the di fferent sources of spatial variation in soil on the basis of the scale at which they operate, and it often enhances relations between soil a ttributes which were blurred in a traditional correlation analysis whe re the different sources of variations are mixed. Geostatistics can al so be used to assess the risk of exceeding critical values (regulatory thresholds, soil quality criterion) at unsampled locations, and to si mulate the spatial distribution of attribute values. All the different tools are illustrated using two transects of 100 pH and electrical co nductivity values measured in pasture and forest.