This work presents a brief review of spatial data analysis methods
and their application to radioactive contamination of territories.
Two methods are described in the paper and applied to real data on
soil contamination with strontium 90 (Sr90) and cesium 137 (Cs137)
? cokriging model and general regression neural networks. Cokriging
is a geostatistical predicator based on multivariate linear regression,
which allows inclusion of data on correlated variables in the joint
estimation procedure, in order to improve the prediction quality
and to reduce estimation errors. General regression neural network
is a nonparametric estimator, which is fast and produces high quality
results on extremely variable Chernobyl data. Such research with
adaptation of these methods to the characterization of radionuclides
data has been under progress for 5 years by the senior authors group.