The present paper deals with novel developments and application of
Support Vector Machines (Support Vector Classifier SVC and Support
Vector Regression SVR) for the analysis and modeling of spatially
distributed environmental and pollution information (categorical
and/or continuous data). SVC/SVR models are based on the Statistical
Learning Theory or Vapnik-Chervonenkis (VC)-theory. The SVC provide
non-linear classification by mapping the input space into high dimensional
feature spaces where a special type of hyper-planes with maximal
margins (giving rise to good generalizations) are constructed. SVR
provide robust non-linear regression of spatially distributed data.
Real case studies of the present paper deal with binary classification
problem of indicator variables, multi-class classification of soil
types, and prediction mapping of radioactively contaminated territories.
Geostatistical tools (variography) is used to control the performance
of the machines and for better understanding of the results. The
SVC/SVR are well adapted to fuzzy environmental and pollution data.