Rapid Prediction of Radioactive Material Deposition Patterns Using Machine Learning
Abstract
In the event of a sudden radioactive material release into the atmosphere, researchers often run physics-based atmospheric and dispersion models to predict the extent and variation of the deposited hazardous material. These physics-based models are accurate and capable, but often run too slowly to be used on low-powered hardware in the field or as forward models in inverse modeling problems. In this work we present a machine learning model that can be used to quickly produce predictions of two-dimensional deposition patterns from an atmospheric release of Cs-137. The model consists of a grid of individual independent linear and logistic regression models. The inputs to our model are the wind direction and release location, and the output is a map of the predicted deposition pattern. We find that we can make predictions that are close in value to the output of physics-based models but with a speedup on the order of 5x106. This work opens the door to using more advanced machine learning models in the future or to running deposition models on low-powered hardware. This work is released under number LLNL-ABS-825166 and was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under contract DE-AC52-07NA27344. Lawrence Livermore National Security, LLC.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.A15E1686G