Inverse estimation of size-distribution parameters of emitted aerosols following the Fukushima accident using FLEXPART simulations and measurements

 The size distribution (SZ) of radioactive aerosols emitted after nuclear accident at nuclear power plants plays a crucial role in assessment of the subsequent atmospheric transport and deposition. However, in reality this distribution in the source is usually unknown. The SZ of particles in the plume also changes with travel time of the plume, because the coarser particles fall out more rapidly than the finer particles. Hence when the measurements of SZ are undertaken at certain distances from the source the SZ could be already altered by plume travel time while it is SZ in the source which is required by atmospheric transport models (ATMs) for simulation of radionuclides atmospheric dispersion and deposition. Also, SZ measurements are usually not available in real time during the accident. More readily available measurements are airborne concentrations. Hence when concentration measurements are available, the SZ parameters of ATMs could be fitted to achieve better agreement between model and measurements.

 In this work, the inverse problem is stated to identify the optimal set of size distribution parameters of the Fukushima source term – activity-averaged mean aerodynamic diameter (d) and geometric standard deviation (σ) which best fit results of FLEXPART ATM to both, local and global measurements datasets. The problem is formulated as multi-objective optimization in which two objective functions. The first objective function J₁ corresponds to model deviations from measurements in the territory of Japan, while the second objective function J₂ corresponds to model deviations from the global observations of CTBTO measurement stations. The combined cost function J=J₁J₂ , characterizing model deviation against measurements in both datasets was also considered. In this way, the estimate of the unknown SZ parameters, which fits both local and global concentration observations is to be found. The method of finding Pareto solution of such multi-objective optimization problem was developed and preliminary results of comparisons of the estimated SZ parameters with SZ measurements, performed following Fukushima accident were obtained.

 The solution of the stated problem leads to reasonable results. The simulations with small values of 1≤σ≤2 led to excellent agreement of estimated mean aerodynamic diameter d of emitted particles between 2 and 3 μm with available measurements of SZ. At the same time if large values of σ were allowed the resulting estimated mean aerodynamic diameter could significantly deviate from the observed values. The use of the small values of mean aerodynamic diameter (d <1μm) in turn did not allow for the minimization of the combined cost function J.