Multifractal analysis of airborne particle count data: The influence of data preprocessing

We analyze particle count data samples (taken from the recordings of aerosol particle sensors in a cleanroom environment measuring particles of size larger than 0.01 mu m) by evaluating their multifractal f(alpha)-spectra. This is done by use of the canonical method. The data samples are preprocessed in various ways: The absolute values of the differences of successive original data values (the ''derivative'' of the original data) are taken as a measure for the multifractal analysis. Furthermore, the original data values are shifted by an additive constant. These shifts generate interesting effects in the resulting f(alpha)-spectra if the size of the shifts is comparable to the mean value of the original data set. Additionally, we vary the range of the multifractal parameter q, - q(max) less than or equal to q less than or equal to q(max). Depending on the structure of a data set, q(max) must be extended to higher values in order to allow a consistent and meaningful interpretation of the resulting f(alpha)-spectra and a useful comparison between the f(alpha)-spectra of different data samples. (C) 1997 American Association for Aerosol Research.