ZONATION OF HANFORD FORMATION SEDIMENT AT THE HANFORD SITE (PREPRINT IBRAE-2003-07)

Preprint IBRAE-2003-07

The current work is devoted to application of different approaches to classification (zonation) of the Hanford formation – the uppermost geologic layer in the Hanford site (Washington, USA). Approaches based on machine learning (artificial neural networks and statistical learning theory) were used together with geostatistical methods (indicator approach) for the current task. The problem was significantly complicated by low number of samples, especially for some classes. Additional available information was also used. Advantages and drawbacks of used methods are discussed in the paper. Obtained results were compared with a map built by experts-geologists using additional geologic knowledge on the region under study. The comparison was performed with the help of the simplest methods developed for image analysis.

Bibliographical reference 

Savelieva E., Bolshov L., Pozdnukhov A., Timonin V., Kanevski M., Chernov S., Murray Ch., Thorne P., Xie Y. ZONATION OF HANFORD FORMATION AT THE HANFORD SITE. Preprint IBRAE-2003-07. Moscow: Nuclear Safety Institute RAS, March 2003. 55 p.