Effect of Soil Data Spatial Distribution for Multivariate Spatial Prediction under Uncertainty

Abstract

This paper is intended to explain the multivariate spatial prediction in geostatistical
analysis. The purpose of this paper to develop an empirical methodology for spatial insurance
by studying the fundamental causes for differences in between distributed locations. The
objective of this work is to access an approach of optimal model of prediction under
uncertainty. Cokriging technique used in this paper through empirical estimation of
variograms and cross-variograms in all directions of compass. Multivariate technique applied
to obtain fitting theoretical models of covariance function, with their properties. The data
adopted from Mosul quadrangle/Iraq, include two sets of groundwater viability of soil data.,
each set contains (100) sample. First set primary variable, magnesium (Mg), and second set is
secondary variable (co-variable) is chlorine (Cl). The results of this work shown to suggest fit
models with best prediction, through the small variations and constraints of weights clear
support of accuracy of cokriging prediction. The outcome showed the models of data adopted
nearest of origin models of covariance functions. In conclusion, the prediction of cokriging
found that the origin data are the nearest the prediction of cross variogram. The computations
are carried out by Matlab language