Finite Element-Based Extended Kalman Filter and Smoother for Retrieval of Aerosol Size Distributions and Process Rates
TEEMU SALMINEN, Kari E.J. Lehtinen, Matti Niskanen, Pietari Mönkkönen, Jari P. Kaipio, Aku Seppänen, University of Eastern Finland
Abstract Number: 621
Working Group: Aerosol Physics
Abstract
Aerosol particles have a major effect on the climate as they reflect and absorb incoming solar radiation and act as cloud condensing nuclei, indirectly affecting the albedo of the Earth. However, these effects are highly uncertain due to uncertainties in aerosol dynamics modeling in global climate models as both aerosol process rate approximations and applied aerosol dynamics models are typically rather crude.
Temporal evolution of the aerosol number distribution can be described with the General Dynamic Equation of aerosols (GDE) which is an integro-partial differential equation that accounts for processes affecting an aerosol particle population. However, the process rates (growth rate, deposition rate, and nucleation rate) and mechanics behind them are partially unknown. To approximate the evolution of aerosol number distributions accurately with GDE, these process rates should be approximated accurately.
Extended Kalman filter (EKF) and Fixed interval Kalman smoother (FIKS) are powerful statistical tools which can be used to estimate variables and their uncertainties from the sequential measurements in the Bayesian state-space framework. We apply the EKF and FIKS, which uses Finite Element Method (FEM) approximation of GDE in prediction step, to estimate the temporal evolution of aerosol size distribution and process rates and their variances in simulated case and for a chamber measurement. We investigate whether the EKF and FIKS lead to more accurate process rate estimates than customarily applied ones.
Our preliminary results indicate that FEM- based EKF and FIKS provide feasible estimates for the aerosol number distribution and aerosol process rates in both cases. Moreover, we obtain approximations for uncertainties of our estimates. Future goal is to apply the methodology to atmospheric aerosol data to obtain estimates for the process rates with uncertainties which then can be used in the climate model parametrizations after the feasibility of methodology is established.