Application of data assimilation to improve the forecasting capability of an atmospheric dispersion model for a radioactive plume

Modeling an atmospheric dispersion of a radioactive plume plays an influential role in assessing the environmental impacts caused by nuclear accidents. The performance of data assimilation techniques combined with Gaussian model outputs and measurements to improve forecasting abilities are investigated in this study. Tracer dispersion experiments are performed to produce field data by assuming a radiological emergency. Adaptive neuro-fuzzy inference system (ANFIS) and linear regression filter are considered to assimilate the Gaussian model outputs with measurements. ANFIS is trained so that the model outputs are likely to be more accurate for the experimental data. Linear regression filter is designed to assimilate measurements similar to the ANFIS. It is confirmed that ANFIS could be an appropriate method for an improvement of the forecasting capability of an atmospheric dispersion model in the case of a radiological emergency, judging from the higher correlation coefficients between the measured and the assimilated ones rather than a linear regression filter. This kind of data assimilation method could support a decision-making system when deciding on the best available countermeasures for public health from among emergency preparedness alternatives.