To model the atmospheric dispersion of radionuclides released from nuclear accident is very important for nuclear emergency. But the uncertainty of model parameters, such as source term and meteorological data, may significantly affect the prediction accuracy. Data assimilation (DA) is usually used to improve the model prediction with the measurements. The paper proposed a parameter bias transformation method combined with Lagrangian puff model to perform DA. The method uses the transformation of coordinates to approximate the effect of parameters bias. The uncertainty of four model parameters is considered in the paper: release rate, wind speed, wind direction and plume height. And particle swarm optimization is used for searching the optimal parameters. Twin experiment and Kincaid experiment are used to evaluate the performance of the proposed method. The results show that the proposed method can effectively increase the reliability of model prediction and estimate the parameters. It has the advantage of clear concept and simple calculation. It will be useful for improving the result of atmospheric dispersion model at the early stage of nuclear emergency.

对核事故释放的放射性核素的大气扩散进行建模对于核应急非常重要。但模型参数的不确定性，如源项和气象数据，可能会显着影响预测精度。数据同化 (DA) 通常用于通过测量改进模型预测。论文提出了一种结合拉格朗日粉扑模型进行DA的参数偏置变换方法。该方法使用坐标变换来近似参数偏差的影响。论文考虑了四个模型参数的不确定性：释放率、风速、风向和羽流高度。粒子群优化用于搜索最优参数。使用孪生实验和 Kincaid 实验来评估所提出方法的性能。结果表明，所提方法能够有效提高模型预测的可靠性和参数估计值。它具有概念清晰、计算简单的优点。对改进核应急初期大气扩散模型的结果具有重要意义。