No abstract

Partial least squares regression (PLS) is proposed for solving air pollution source apportionment problems as an alternative method to the frequently used chemical mass balance technique. A discriminant PLS is used to calculate linear mixing proportions for a synthetic ambient aerosol data set where the truth is known. Without sacrificing orthogonality of the source profiles, PLS can resolve the emission sources and accurately predict the emission source contributions. Further extensions of the PLS approach to environmental receptor modelling are discussed. KEY WORDS Partial least squares Receptor modelling Colinearities SNTNF Senior Visiting Scientist. Current address: EPA Environmental Research Lab., 200 S.W. 35th Street, Corvallis, OR 97333, U.S.A. 0886-9383/88/04028 1-16$08 .OO

A B S T R A C T An experiment was conducted to measure aerosol turbulent fluxes to a grass field. A new high-flow-rate aerosol sensor was deployed from a tower to make eddy correlation (EC) measurements of aerosol turbulent flux and deposition velocity. The EC data were screened and analysed for uncertainties associated with advection, boundary layer growth, instrument separation and counting particles. An apparent bias in the aerosol flux due to particle hygroscopic growth was evaluated from chemical and microphysical measurements and removed from the results based on derived corrections. The resulting aerosol deposition velocity for 0.52 µm diameter particles depended on atmospheric stability with values of 0.3 cm s −1 during near-neutral stability, 0.44 cm s −1 during unstable periods and 0.16 cm s −1 during stable periods with an estimated uncertainty of ±0.07 cm s −1 due to chemical composition and particle counting.

An eddy correlation technique was used to measure the turbulent flux of cloud droplets to complex, forested terrain near the coast of Washington State during the spring of 1993. Excellent agreement was achieved for cloud liquid water content measured by two instruments. Substantial downward liquid water fluxes of-1 mm per 24 h were measured at night during "steady and continuous" cloud events, about twice the magnitude of those measured by Beswick et al. in Scotland. Cloud water chemical fluxes were estimated to represent up to 50 % of the chemical deposition associated with precipitation at the site. An observed size-dependence in the turbulent liquid water fluxes suggested that both droplet impaction, which leads to downward fluxes, and phase change processes, which can lead to upward fluxes, consistently are important contributors to the eddy correlation results. The diameter below which phase change processes were important to observed fluxes was shown to depend upon 'hi L, the relative standard deviation of the liquid water content (LWC) within a 30-min averaging period. The crossover from upward to downward L W flux occurs at 8 µm for steady and continuous cloud events but at-13 µm for events with a larger degree of L WC variability. This comparison of the two types of cloud events suggested that evaporation was the most likely cause of upward droplet fluxes for the smaller droplets (dia < 13 µm) during cloud with variable LWC (a LIL> 0.3).

Occult deposition to vegetation, via mechanical interception of wind-blown cloud water, can be a significant fraction of total ionic chemical deposition for some forests. Applying micrometeorological methods to the estimation of cloud water deposition requires particular consideration since cloud droplets are not conservative but are subject to material change (phase change); sedimentation also affects fluxes of droplets. The budget equation for liquid water (LW) in orographic cloud predicts that LW fluxes will diverge due to condensation during mean ascent. For hilltop measurements such as these, other factors can contribute to changes in the vertical flux with height above the surface.Fluxes measured concurrently at two heights exhibit a persistent and significant divergence, while laterally separated measurements are found to agree. A LW budget equation is presented and simplified by scale analysis. Surface uptake is estimated by extrapolation of the measured fluxes. Estimated surface deposition is found to be substantially different from the flux measured at a reference height (10 m), often by a factor of two and occasionally with a different sign. This difference in estimated surface uptake extends to estimates of chemical as well as water deposition. The turbulent flux of LW is shown to be dependent on two criteria for describing 'steadycloud' conditions, thus presumably minimizing the effects of entrainment. An often used model relating droplet deposition to the 'deposition velocity' for momentum is found to be inappropriate for application in complex terrain. Deposition of LW is estimated to range from 5 to 50 mg m-' s-I during the third field campaign of the Cloud and Aerosol CHemistry Experiment (CACHE-3), with an average of 19 mg m-'s-' (nearly 2 mm per in-cloud day) in late summer at this Pacific coastal site in North America. Numerical recipes in FORTRAN: the art of scientiJc computing.Cambridge University Press, Cambridge, UK Drag and drag partition on rough surfaces. Boundary-Layer Meteoml., 60,375-395 Simplified expressions for vegetation roughness length and zeroplane displacement as functions of canopy height and area index. Boumkry-Layer Meteoml., 71,211-216 Temperature and humidity fields and fluxes over low hills. Q. J. R. Meteoml. SOC., 118, 191-225 A neglected water resource: the Camanchaca of South America. Bull. Am. Meteoml. SOC., 69, 138-147 Footprint prediction of scalar fluxes from analytical solutions of the diffision equation. Boundary-Layer Meteoml., 50,355-373 A one-dimensional theoretical description of the vegetationatmosphere interaction. Boundary-Layer Meteoml., 10,273-302