A Spatial Model of Urban Winter Woodsmoke Concentrations

Abstract

In many urban areas, residential wood burning is a significant wintertime source of PM2.5. In this study, we used a combination of fixed and mobile monitoring along with a novel spatial buffering procedure to estimate the spatial patterns of woodsmoke. Two-week average PM2.5 and levoglucosan (a marker for wood smoke) concentrations were concurrently measured at upto seven sites in the study region. In addition, pre-selected routes spanning the major population areas in and around Vancouver, B.C. were traversed during 19 cold, clear winter evenings from November, 2004 to March, 2005 by a vehicle equipped with GPS receiver and a nephelometer. Fifteen-second-average values of light scattering coefficient (bsp) were adjusted for variations between evenings and then combined into a single, highly resolved map of nighttime winter bsp levels. A relatively simple but robust (R(2) = 0.64) land use regression model was developed using selected spatial covariates to predict these temporally adjusted bsp values. The bsp values predicted by this model were also correlated with the measured average levoglucosan concentrations at our fixed site locations (R(2) = 0.66). This model, the first application of land use regression for woodsmoke, enabled the identification and prediction of previously unrecognized high woodsmoke regions within an urban airshed.

A spatiotemporal land-use regression model of winter fine particulate levels in residential neighbourhoods

“…Our results are different than those reported by Larson et al 2 They developed a LUR model for British Columbia to predict nighttime winter black smoke levels, an indicator of small particles, using selected spatial covariates. As with us, woodstove uses poorly predicted fine particulate levels in the study of Larson et al 2 .…”

“…As with us, woodstove uses poorly predicted fine particulate levels in the study of Larson et al 2 . The best models developed by this group included estimates of socioeconomic status; their R 2 ranged from 0.58 to 0.84.…”

“…In some of the full models, the signs of the coefficients for elevation, wind and temperatures were opposite to those expected, because of the correlation between these predictors of PM1 levels. The use of the compound topographic index, as described by Larson et al, 2 provided similar results as the elevation variable used (see Supplementary Appendix).…”

“…It was assumed that polluted air would be more likely trapped in areas where hills, and thus higher elevation, were present. In secondary analyses, we also used the compound topographic index for grids of 900 and 2400 m, as described in Larson et al 2 Kilometres of major roads in buffers of 1 km around each sampling points were estimated using the categories ''highways, freeways and major roads'' from the DMTI CanMap street files.…”

“…In these models, adjustments were made for variations between evenings (see, e.g., Larson et al 2 ). Here, we developed a model to estimate weekend evening minute (min) levels of particles with median diameter of o1 mg/m 3 (PM1) measured (mobile sampling) in different locations in the province of Quebec, Canada, with consideration given to wood burning emissions.…”