Spatial variations and development of land use regression models of levoglucosan in four European study areas

Abstract. Relatively little is known about long term effects of wood smoke on population health. A wood burning marker – levoglucosan – was measured using a highly standardized sampling and measurement method in four study areas across Europe (Oslo, the Netherlands, Munich/Augsburg, Catalonia) to assess within and between study area spatial variation. Levoglucosan was analyzed in addition to other components: PM_2.5, PM_2.5 absorbance, PM₁₀, polycyclic aromatic hydrocarbons (PAH), nitrogen oxides (NO_x), elemental and organic carbon (EC / OC), hopanes, steranes and elemental composition. Measurements were conducted at street, urban and regional background sites. Three two-week samples were taken per site and the annual average concentrations of pollutants were calculated using continuous measurements at one background site as a eference. Land use regression (LUR) models were developed to explain the spatial variation of levoglucosan using standardized procedures.

Much larger within than between study area contrast in levoglucosan concentration was found. Spatial variation patterns differed substantially from other measured pollutants including PM_2.5, NO_x and EC. Levoglucosan had the highest spatial correlation with ΣPAH (r = 0.65) and the lowest with traffic markers – NO_x, Σhopanes/steranes (r = −0.22). The correlation of levoglucosan with potassium (K), which is also used as a wood burning marker, was moderate to low (median r = 0.33). Levoglucosan concentrations in the cold (heating) period were between 3 and 20 times higher compared to the warm period. The contribution of wood-smoke calculated based on levoglucosan measurements and previous European emission data to OC and PM_2.5 mass were 13 to 28% and 3 to 9% respectively in the full year. Larger contributions were calculated for the cold period.

The median model R² of the LUR models was 60%. In Catalonia the model R² was the highest (71%). The LUR models included population and natural land related variables but no traffic associated variables.

In conclusion, substantial spatial variability was found in levoglucosan concentrations particularly within study areas. Wood smoke contributed substantially to especially wintertime PM_2.5 OC and mass. The low to moderate correlation with PM_2.5 mass and traffic markers offers the potential to assess health effects of wood smoke separate from traffic-related air pollution¹.


¹ Abbreviations: ESCAPE, European Study of Cohort for Air Pollution Effects; TRANSPHORM, Transport related Air Pollution and Health impacts – Integrated Methodologies for Assessing Particulate Matter; EC/OC, elemental/organic carbon; PAH, polycyclic aromatic hydrocarbons; B[a]P, benzo[a]pyrene, GIS, Geographic Information Systems; LUR, Land Use Regression; NO_x, nitrogen oxides; NO₂, nitrogen dioxide; PM_2.5, mass concentration of particles less than 2.5 μm in size; PM_2.5 absorbance, measurement of the blackness of PM_2.5 filters, this is a proxy for elemental carbon, which is the dominant light absorbing substance; PM₁₀, mass concentration of particles less than 10 μm in size; RB, regional background; S, Street; EPA, United States Environmental Protection Agency; LUR, Land Use Regression; RMSE, Root Mean Squared Error.