Estimation of aerosol particle number distributions with Kalman Filtering - Part 1: Theory, general aspects and statistical validity
- ISSN: 16807316
- DOI: 10.5194/acp-12-11767-2012
Aerosol characteristics can be measured with different instruments\nproviding observations that are not trivially inter-comparable. Extended\nKalman Filter (EKF) is introduced here as a method to estimate aerosol\nparticle number size distributions from multiple simultaneous\nobservations. The focus here in Part 1 of the work was on general\naspects of EKF in the context of Differential Mobility Particle Sizer\n(DMPS) measurements. Additional instruments and their implementations\nare discussed in Part 2 of the work. University of Helsinki\nMulti-component Aerosol model (UHMA) is used to propagate the size\ndistribution in time. At each observation time (10 min apart), the time\nevolved state is updated with the raw particle mobility distributions,\nmeasured with two DMPS systems. EKF approach was validated by\ncalculating the bias and the standard deviation for the estimated size\ndistributions with respect to the raw measurements. These were compared\nto corresponding bias and standard deviation values for particle number\nsize distributions obtained from raw measurements by a inversion of the\ninstrument kernel matrix method. Despite the assumptions made in the EKF\nimplementation, EKF was found to be more accurate than the inversion of\nthe instrument kernel matrix in terms of bias, and compatible in terms\nof standard deviation. Potential further improvements of the EKF\nimplementation are discussed.