Convergence monitoring of Markov chains generated for inverse tracking of unknown model parameters in atmospheric dispersion
- Title
- Convergence monitoring of Markov chains generated for inverse tracking of unknown model parameters in atmospheric dispersion
- Author
- 이재기
- Keywords
- atmospheric dispersion; source reconstruction; Markov chain; convergence monitoring; autocorrelation; potential scale reduction factor
- Issue Date
- 2011-02
- Publisher
- 日本原子力学会
- Citation
- Progress in NUCLEAR SCIENCE and TECHNOLOGY, Vol. 1, p.464-467 (2011)
- Abstract
- The dependency within the sequential realizations in the generated Markov chains and their reliabilities are monitored by introducing the autocorrelation and the potential scale reduction factor (PSRF) by model parameters in the atmospheric dispersion. These two diagnostics have been applied for the posterior quantities of the release point and the release rate inferred through the inverse tracking of unknown model parameters for the Yonggwang atmospheric tracer experiment in Korea. The autocorrelations of model parameters are decreasing to low values approaching to zero with increase of lag, resulted in decrease of the dependencies within the two sequential realizations. Their PSRFs are reduced to within 1.2 and the adequate simulation number recognized from these results. From these two convergence diagnostics, the validation of Markov chains generated have been ensured and PSRF then is especially suggested as the efficient tool for convergence monitoring for the source reconstruction in atmospheric dispersion.
- URI
- http://www.aesj.net/publish/jnst/volume-index-pnst/pnst1https://repository.hanyang.ac.kr/handle/20.500.11754/72708
- DOI
- 10.15669/pnst.1.464
- Appears in Collections:
- COLLEGE OF ENGINEERING[S](공과대학) > NUCLEAR ENGINEERING(원자력공학과) > Articles
- Files in This Item:
There are no files associated with this item.
- Export
- RIS (EndNote)
- XLS (Excel)
- XML