Abstract
Twomey's nonlinear iteration method was applied to retrieve the size distribution of atmospheric aerosols from data obtained by an Andersen low-pressure impactor. Effects of some factors on the inversion results were investigated using numerically generated data for unimodal and bimodal lognormal functions. The factors were weighting functions, first-guess distributions and treating the ends of the distribution. The pth power of response function was adopted as a weighting function and a value of 0.7 was selected for the inversion of data in most cases. Smoothed initial distribution with the treatment of the ends of the distribution gave good results in case of the distribution showing a peak near the first stage of the impactor, while uniform initial guess resulted in an oscillatory distribution. We proposed a new stopping criterion based on experimental errors which represent random errors in mass measurement and sample fluctu- ation. and the criterion is useful for inverting the data involving random errors. The procedure was also applied to estimate the size distribution for ambient aerosol samples.
