Abstract
Atmospheric dispersion (AD) modeling is the imperative part of accident consequence assessment in nuclear leakage accident. The emergency decision-making depends on the AD model forecast result. The combination of the observation data and forecast result will improve the AD prediction. This paper practices a data assimilation (DA) method in AD based on image registration. The observation data is used to adjust the result of Gaussian plume model, and image affine transformation is used in the DA. The DA method is simple and rapid, and the discrepancy between AD model prediction and true situation is significantly reduced.