Modified ensemble Kalman filter for nuclear accident atmospheric dispersion: Prediction improved and source estimated

• Published in: Journal of hazardous materials Vol. 280; pp. 143 - 155
• Main Authors: Zhang, X.L., Su, G.F., Yuan, H.Y., Chen, J.G., Huang, Q.Y.
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier B.V 15.09.2014; Elsevier
• Subjects: Air Pollution; Applied sciences; Atmospheric dispersion; Computer Simulation; Ensemble Kalman filter; Exact sciences and technology; Models, Theoretical; Nuclear power plant accident; Pollution; Radioactive Hazard Release; Nuclear power plant accident; Ensemble Kalman filter; Atmospheric dispersion; Atmospheric condition; Uncertainty; Meteorological observation; Kalman filter; Prediction; Plume; Modeling; Nuclear accident; Time lag; Surveillance; Nuclear power plant; Environment impact; Planning; Emergency; Data assimilation; Concentration distribution
• ISSN: 0304-3894; 1873-3336
• DOI: 10.1016/j.jhazmat.2014.07.064

•A modified ensemble Kalmen filter data assimilation method is proposed.•The method can consider four main uncertain parameters in the puff model.•The prediction of radioactive material atmospheric dispersion is improved.•The source release rate and plume rise height are successfully reconstructed.•It can shorten the time lag in the response of ensemble Kalmen filter. Atmospheric dispersion models play an important role in nuclear power plant accident management. A reliable estimation of radioactive material distribution in short range (about 50km) is in urgent need for population sheltering and evacuation planning. However, the meteorological data and the source term which greatly influence the accuracy of the atmospheric dispersion models are usually poorly known at the early phase of the emergency. In this study, a modified ensemble Kalman filter data assimilation method in conjunction with a Lagrangian puff-model is proposed to simultaneously improve the model prediction and reconstruct the source terms for short range atmospheric dispersion using the off-site environmental monitoring data. Four main uncertainty parameters are considered: source release rate, plume rise height, wind speed and wind direction. Twin experiments show that the method effectively improves the predicted concentration distribution, and the temporal profiles of source release rate and plume rise height are also successfully reconstructed. Moreover, the time lag in the response of ensemble Kalman filter is shortened. The method proposed here can be a useful tool not only in the nuclear power plant accident emergency management but also in other similar situation where hazardous material is released into the atmosphere.