You are here:
CHARACTERIZATION OF FINE PARTICULATE MATTER
Impact/Purpose:
The goal of this project within the PM Center is to measure the physical and hygroscopic properties of the aerosol in the urban areas where the health studies are being done during the first 4 years and then to incorporate these data into respiratory deposition models and mesoscale transport models. The physical properties to be measured are the number-size distribution from 10 nm to 10,000 nm diameter. The hygroscopic properties are the diameter growth factor for the submicrometric size range for humidities up to 90 percent RH with extrapolation to 100 percent via Koehler theory.
Description:
Size distribution data processing and fitting
Ultrafine, very fine and fine PM were collected nearly continuously from December 2000 through March 2003 at a Washington State Department of Ecology site on Beacon Hill in Seattle. Particle size distributions from 20nm to 800nm were measured using a differential mobility particle sizer (DMPS) and an aerodynamic particle sizer (APS). The DMPS consisted of an electrostatic classifier and condensation particle counter (models 3081 and 3010, TSI, St. Paul MN). The DMPS stepped through the mobility size range over a period of ten minutes. At the end of a stepped scan sequence, the mobility data were inverted with the charge probability matrix to get the Stokes diameter size distribution, (dN(DpStk)/dlogDp). The size distribution from 700 to 5000 nm was measured with an aerosol particle sampler APS (model 3310, TSI, St. Paul, MN) on the same ten minute basis to yield the aerodynamic diameter size distribution (dN(DpAero)/dlogDp). The DMPS measurements were converted to an aerodynamic diameter base by assuming a particle density that varied from 1.0 g/cm3 at 100 nm to 1.8 g/cm3 at 600nm based on measurements at other locations, chemical analysis and several direct measurements of the output of the DMA by the APS over their overlapping measurement range. These inverted DMPS data were merged with the APS aerodynamic size distributions. The ten-minute distributions were edited to eliminate periods of instrument malfunction and high variability in particle concentration due to on-site activity.
The ten-minute distributions were grouped into overlapping, two hour windows. The two- hour window was selected to minimize the stochastic variability in the measurement and to maximize the measure of atmospheric variability. The number-size distributions were then fit by a multi-modal, multi-moment lognormal fitting algorithm The algorithm software was implemented in the R statistical programming system (Ihaka Gentleman, 1996), which minimizes the residual between the measured concentration as a function of size and the fit values simultaneously for three moments (number, surface, and volume) of the distribution rather than for each moment individually. Four lognormal modes were allowed reducing the sixty-seven measured differential concentrations to twelve parameters: concentration, geometric mean , diameter, and standard deviation for each the four modes. The algorithm found a combination of the four lognormal functions that simultaneously minimized the residual of the number, surface area, and volume moments of the distribution. Constraints to the ranges mean diameters and standard deviation for the modes were imposed to prevent unphysical solutions. Volume mean diameters and concentrations were derived from the number-size, multimodal model.
The nanometer mode was included in the algorithm for completeness even though measurement did not include the range; it was not subsequently used due to insignificant frequency of occurrence and volume concentrations even when the algorithm indicated the presence of a nano-mode.
Size distribution and chemical species measurements were combined into a multivariate receptor model of PM2.5 (Larson et al, 2006). The combined model extends the traditional chemical mass balance approach by including a simultaneous set of conservation equations for both particle mass and volume, linked by a unique value of apparent particle density for each source. The model distinguished three mobile source feature, two consistent with previous identifications of “gasoline” and “diesel” sources , and an additional minor feature enriched in EC, Fe, MN and ultrafine particle mass that would have been difficult to interpret in absence of particle size information. This study has also demonstrated the feasibility of defining missing mass as an additional variable, and thereby providing additional useful model constraints and eliminating the posthoc regression step that is traditionally used to rescale the results. Secondly, the very fine particle data were used in a health outcome study evaluating associations between emergency department visits for asthma (Mar et al, 2007, submitted -- see project R827355C002).
URLs/Downloads:
2001 Progress Report1999 Progress Report
covert.html
2000 Progress Report
2003 Progress Report
Final Progress Report