Data inversion for cascade impactors: Fitting sums of log-normal distributions

Abstract

The processing of data from cascade impactors is discussed and the effects of non-ideal response characteristics on impactor measurements are shown. Since atmospheric particle size distributions may contain two or three modes, the sparse data from an impactor cannot give detailed information about the shape of individual peaks. Thus a shape must be assumed, and it is suggested that a log-normal form is appropriate. A procedure is described which fits sums of log-normal distributions to impactor data, taking account of the detailed response functions. This procedure has been tested with simulated data including random errors, where it is found to reproduce the input distributions satisfactorily, and with actual atmospheric data.