10th International Aerosol Conference September 2 - September 7, 2018 America's Center Convention Complex St. Louis, Missouri, USA
Abstract View
Estimation of Loss Rates from Chamber Experiment Data Using a Statistical Inverse Approach
MATTHEW OZON, Aku Seppänen, Ari Leskinen, Jari Kaipio, Kari Lehtinen, University of Eastern Finland
Abstract Number: 758 Working Group: Aerosol Modeling
Abstract Motivations Deposition is one of the key aerosol microphysical processes, important in large scale when considering both climate and health effects and in a smaller scale, e.g. when analyzing chamber experiments. Deposition mechanisms and rates are strong functions of size: smaller particles deposit mainly by diffusion in contrast to larger ones, which deposit by settling or due to their inertia. In this study, our focus is on estimating deposition rates in experiments, e.g. in a chamber, in which multiple processes are acting simultaneously. One might for example encounter a situation in which the concentration is high enough that coagulation is significant in addition to deposition. Then it is not straightforward to extract size dependent deposition rates just by simply examining the decrease of the number distribution.
Methods A rigorous method to estimate size dependent deposition rates in a chamber has been to perform several separate concentration decay experiments using a monodisperse aerosol. This is, however, not necessary even for a seemingly complex case with other simultaneous disturbing processes if advanced data-analysis methods are used. Here we propose an automated method where the only assumption is that the parameters vary “slowly” over time. Our approach relies on the widely used statistical, dynamical inversion method that accounts for measurement uncertainty in a rigorous way, the Kalman Filter (KF). The KF is stable, minimizes the variance of the estimates and is easy to implement. Our implementation of the KF not only filters — improve the data quality — the size distribution, but it also estimates the global loss rates. Most importantly, this method only relies on a physical evolution model — the aerosol General Dynamic Equation — a measurement model —SMPS— and rough uncertainty estimations that are combined in the KF framework.
Results As a proof of concept of our method we analyze synthetic data generated by solving the GDE with an accurate sectional method. The data are simulated using a time invariant, arbitrary loss rate and a size spectrum ranging from an arbitrary critical cluster size of 2nm to 2μm. In this simulation, there is no vapor present, and thus the particles do not grow by condensation and there is also no nucleation. Additionally, the data are corrupted by additive and multiplicative noise in order to resemble actual measured data. The results show that the global loss rates are well estimated. Finally, the method is applied to SMPS data from a real wall loss experiment. The estimated parameters are of the same order of magnitude as those previously manually computed, and the size dependence is similar to what is expected by the theory.