Machine learning-derived dose response factors (DRFs) for 137Cs ground deposition in urban environments

Read full text Summarise with VIRAT
Authors: Xibo Ma, Hikaru Kobayashi
Cited by: 0
References: 46

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Monte Carlo photon transport simulations can be employed to predict urban dose distributions resulting from radionuclide ground deposition following nuclear accidents, using measured or estimated radiation source data and urban geometry information. However, large-scale calculations that aim to reproduce realistic environments remain challenging owing to the significant modeling and computational demands. To address this limitation, we developed dose response factors (DRFs) for ground-deposited 137 Cs employing machine-learning models trained on randomly generated urban layouts. These factors quantify the spatial dose response of each small-area source under the consideration of complex shielding inherent to the local urban morphology. Accordingly, the dose distribution within the area of interest could be calculated by integrating the DRFs of the surrounding sources weighted by their respective strengths. Using building cover ratios between the source and response areas as primary features to characterize the urban morphology composed of wooden houses with diverse dimensions and locations, the trained model achieved high accuracy. Using the physical simulation results as the reference values, the predicted results yielded a mean absolute percentage error below 5% across locations up to 90 m from the source on a 3 m planar grid.

Fields of Study & Subjects

Radioactive contamination and transfer Nuclear and radioactivity studies Radiation Shielding Materials Analysis

Publication details

Journal / source
Nuclear Engineering and Technology
Volume
58
Issue
7
Pages
104272–104272
Publication date
30 March 2026
Language
EN