SCIENCE-BASED MODELLING OF CHERNOBYL FALLOUT TREND (PREPRINT IBRAE-2002-08)

Preprint IBRAE-2002-08

Kanevski M., Savelieva E., Demyanov V., Chernov S., Sorokovikova O., Belikov V.

This work is devoted to the science based modeling of ¹³⁷Cs contamination due toChernobyl fallout, this fallout is an example of data with non-linear trend and high local variability and that’s why they are very efficient for testing non-linear estimators. Science based atmospheric dispersion modeling is performed on the base of the Lagrange model atmospheric transport model and with the help of software NOSTRADAMUS, developed at IBRAE. Nevertheless of detailed preliminary preparation of model parameters the final result shows only the main orientation of the trace. It does not reproduce the spotted structure presented in measured data. Such situation is caused by not exact knowledge of parameters important for atmospheric modeling – Chernobyl example seems to be too difficult for pure science-based modeling. The residuals of science-based model were studied using Artificial Neural Networks (ANN). Such hybridization allows to improve the results by introducing some “hot spots” distributed not in the main trace.

Bibliographical reference 

Kanevski M., Savelieva E., Demyanov V., Chernov S., Sorokovikova O., Belikov V. SCIENCE-BASED MODELLING OF CHERNOBYL FALLOUT TREND. Preprint IBRAE-2002-08. Moscow: Nuclear Safety Institute RAS, 2002. 23 p. — Refs.: 15 items.