A matrix-free fixed-point iteration for inverting cascade impactor measurements with instrument's sensitivity kernels and hardware
Valtonen, Laura; Saari, Sampo; Pursiainen, Sampsa (2021)
Avaa tiedosto
Lataukset:
Valtonen, Laura
Saari, Sampo
Pursiainen, Sampsa
Taylor & Francis
2021
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2022012610340
https://urn.fi/URN:NBN:fi-fe2022012610340
Tiivistelmä
This study focuses on advancing the inversion of aerosol data measured by a cascade
impactor. Our aim is to find and validate a comprehensive and robust mathematical
model for reconstructing a particle mass distribution. In this paper, we propose a
fixed-point iteration as a method for inverting cascade impactor measurements with
a relatively simple measurement hardware, which is not optimized for handling ad-
vanced linear algebraic operations such as large matrices. We validate this iteration
numerically against an iterative L1 norm regularized iterative alternating sequential
inversion algorithm. In the numerical experiments, we investigate and compare a
point-wise (matrix-free) and integrated kernel-based approach in inverting five dif-
ferent aerosol mass concentration distributions based on simulated measurements
and sensitivity kernel functions.
impactor. Our aim is to find and validate a comprehensive and robust mathematical
model for reconstructing a particle mass distribution. In this paper, we propose a
fixed-point iteration as a method for inverting cascade impactor measurements with
a relatively simple measurement hardware, which is not optimized for handling ad-
vanced linear algebraic operations such as large matrices. We validate this iteration
numerically against an iterative L1 norm regularized iterative alternating sequential
inversion algorithm. In the numerical experiments, we investigate and compare a
point-wise (matrix-free) and integrated kernel-based approach in inverting five dif-
ferent aerosol mass concentration distributions based on simulated measurements
and sensitivity kernel functions.