Sensitivity and uncertainty analysis in predicting hygroscopic growth of aerosols for atmospheric and nuclear relevance

Abstract

Sensitivity and uncertainty analysis serves as a tool for modeling physical phenomenon and evaluating the model validity. It finds applications in simulating the dynamical behavior of aerosols in different contexts. Hygroscopic growth of aerosol particles is one such domain where it can optimize the accuracy of model output greatly. Accurate prediction of hygroscopic growth characteristics depends on the intrinsic parameters and the size-dependent evolution of growth factors. In reactor accident scenarios, growth characteristics of fission/activation product aerosols play a significant role in determining the airborne or deposited fraction. Köhler theory is fundamentally used to predict the growth behavior of aerosols with a detailed physical explanation of critical size and supersaturation characteristics. Alternative approaches including modified Köhler theories and semi-empirical models rely on the linkage of physical and chemical composition-based parameters. The variation of involved parameters and the choice of working range influence the outcome of the theory greatly. This study is focused on sensitivity analysis of the modified Köhler theory given by Brechtel and Kreidenweis and the evaluation of sensitivity and associated uncertainty in the parameters. Monte Carlo approach has been employed, and sensitivity indices were obtained. Identified sensitive parameters were dry particle diameter, solute density, molecular weight, and chemical composition dependent factor (b0). These parameters require careful attention while designing experimental measurements and/or preprocessing model inputs. All sensitive parameters were subjected to uncertainty analysis and the overall output of the aerosol growth code thus obtained has been observed to follow log-normal law with median size at 335.2 nm with a geometric standard deviation of 1.54. This finds applications in the case of studies aiming to predict the growth of aerosols in the reactor component systems during accident conditions. In such conditions, there lies no choice of selection of input parameters and one cannot be sure about their variation range. This provides a conservative estimate for such a random scenario as uncertainty analysis performed in this study covers all the random combinations of input parameters and the overall most probable grown droplet size is estimated.