Hierarchical Agglomerative Cluster Analysis Applied to WIBS 5-Dimensional Bioaerosol Data Sets

Introduction Primary biological aerosol particle (PBAP) classification requires discrimination of particles various diverse sources which may have wide reaching effects in the atmosphere. In order to predict these effects under future emissions scenarios it is useful to be able to identify ambient PBAP concentration. To date, this has largely been achieved by the use of off-line techniques, which, whilst allowing accurate identification of different aerosols, are labour intensive and have poor time resolution. To improve on this we have investigated the use of hierarchical agglomerative (HA) cluster analysis applied to single-particle multi-spatial (5-D) datasets comprising optical diameter, particle asymmetry and three induced fluorescence waveband measurements, from two commonly used dual Waveband Integrated Bioaerosol Spectrometers (WIBS), (Kaye et al., 2005). We show that HA cluster analysis, without the need for any a-priori assumptions concerning the expected aerosol types, can reduce the level of subjectivity compared to the more standard analysis approaches for multiparameter aerosol measurements. Methods We use two WIBS– a model 3 and a model 4 (Gabey et al., 2011). In both models the single particle elastic scattering intensity (at 633 nm) is used to infer particle optical-equivalent diameter, DO. A quadrant PMT measures the variation in azimuthal scattering and hence provides a particle asymmetry factor, AF. This measurement triggers pulses from filtered xenon flash-lamps at 280 nm and 370 nm, designed to excite tryptophan and NAD(P)H molecules within the particle. Fluorescence is measured in two wavelength regimes, FL1 & FL2 providing three fluorescence channels; FL1 & FL2 following the 280 nm excitation and FL2 following the 370 nm excitation. The FL1 and FL2 fluorescence detection regimes overlap spectrally in the WIBS3, but have been separated in the WIBS4. A software tool (WIBS Analysis Program, WASP) was developed that applies the averagelinkage HA-cluster analysis algorithm (Everitt, 1993, Robinson et al. 2012) to WIBS data. Average-linkage defines the two most similar clusters as those with the smallest distance across an !-dimensional space, where ! is the number of particle diagnostics. The distance between two clusters is defined as the average squared Euclidian distance between all possible pairs of particles, or