Data assimilation for short range atmospheric dispersion of radionuclides: a case study of second-order sensitivity

This article presents the application of variational data assimilation to a simple Gaussian plume model for radionuclides. Adjoint modeling is applied to the model in order to minimize discrepancies between contamination observations and model outputs. The interest of such an approach is to get a better estimation of some parameters such as emissions or dispersion parameters. A second-order analysis is also performed to assess the sensitivity of the optimized parameters to some poorly known parameters. Sensitivity with respect to network design is also done.