Integrating travel behavior with land use regression to estimate dynamic air pollution exposure in Hong Kong.

BACKGROUND: Epidemiological studies typically use subjects' residential address to estimate individuals' air pollution exposure. However, in reality this exposure is rarely static as people move from home to work/study locations and commute during the day. Integrating mobility and time-activity data may reduce errors and biases, thereby improving estimates of health risks. OBJECTIVES: To incorporate land use regression with movement and building infiltration data to estimate time-weighted air pollution exposures stratified by age, sex, and employment status for population subgroups in Hong Kong. METHODS: A large population-representative survey (N = 89,385) was used to characterize travel behavior, and derive time-activity pattern for each subject. Infiltration factors calculated from indoor/outdoor monitoring campaigns were used to estimate micro-environmental concentrations. We evaluated dynamic and static (residential location-only) exposures in a staged modeling approach to quantify effects of each component. RESULTS: Higher levels of exposures were found for working adults and students due to increased mobility. Compared to subjects aged 65 or older, exposures to PM2.5, BC, and NO2 were 13%, 39% and 14% higher, respectively for subjects aged below 18, and 3%, 18% and 11% higher, respectively for working adults. Exposures of females were approximately 4% lower than those of males. Dynamic exposures were around 20% lower than ambient exposures at residential addresses. CONCLUSIONS: The incorporation of infiltration and mobility increased heterogeneity in population exposure and allowed identification of highly exposed groups. The use of ambient concentrations may lead to exposure misclassification which introduces bias, resulting in lower effect estimates than 'true' exposures.

Land use regression modelling of air pollution in high density high rise cities: A case study in Hong Kong.

Land use regression (LUR) is a common method of predicting spatial variability of air pollution to estimate exposure. Nitrogen dioxide (NO2), nitric oxide (NO), fine particulate matter (PM2.5), and black carbon (BC) concentrations were measured during two sampling campaigns (April-May and November-January) in Hong Kong (a prototypical high-density high-rise city). Along with 365 potential geospatial predictor variables, these concentrations were used to build two-dimensional land use regression (LUR) models for the territory. Summary statistics for combined measurements over both campaigns were: a) NO2 (Mean=106µg/m3, SD=38.5, N=95), b) NO (M=147µg/m3, SD=88.9, N=40), c) PM2.5 (M=35µg/m3, SD=6.3, N=64), and BC (M=10.6µg/m3, SD=5.3, N=76). Final LUR models had the following statistics: a) NO2 (R2=0.46, RMSE=28µg/m3) b) NO (R2=0.50, RMSE=62µg/m3), c) PM2.5 (R2=0.59; RMSE=4µg/m3), and d) BC (R2=0.50, RMSE=4µg/m3). Traditional LUR predictors such as road length, car park density, and land use types were included in most models. The NO2 prediction surface values were highest in Kowloon and the northern region of Hong Kong Island (downtown Hong Kong). NO showed a similar pattern in the built-up region. Both PM2.5 and BC predictions exhibited a northwest-southeast gradient, with higher concentrations in the north (close to mainland China). For BC, the port was also an area of elevated predicted concentrations. The results matched with existing literature on spatial variation in concentrations of air pollutants and in relation to important emission sources in Hong Kong. The success of these models suggests LUR is appropriate in high-density, high-rise cities.