INTEGRATION BETWEEN GEOSTATISTICAL METHODOLOGIES AND GIS ENVIRONMENTAL GEO-DATA: THE FACTORIAL KRIGING

This paper deals with factorial kriging (FK), the methodology proposed by Matheron in 1982. FK is capable of tripping the crude measured data of a land survay of beckground noises or anything that may hide the relationships between the variables. The paper describes an application of FK to soil science, the aim being to study the variability of soil characteristics. Coregionalization analysis, that is the procedure behind the FK, was carried out using over 600 samples distributed over an area of about 1600 sq. km. Several variables were measured on the samples, but only six were considered for the present study. Four variability structures were identified. The variance/covarianza matrix was adjusted for each scale and principal component analysis (PCA) was carried out on the corresponding correlation matrix. The results show that correlation can vary considerable for somo of the variables and may even be entirely the opposite, passing from one scale to the other. This demonstrates the importance of decompose the variables into their spatial components since the crude data often mask certain behaviours.