Abstract for the 2017 HEI Annual Conference
The Hong Kong D3D Study: A Dynamic Three-Dimensional Exposure Model for Hong Kong
Benjamin Barratt1, Poh-Chin Lai2, Linwei Tian2, Thuan-Quoc Thach2, Robert Tang2, Martha Lee4, Paulina Wong2, Wei Cheng2, Yang Yang2, Anthony Tsui2, Ryan Allen3 and Michael Brauer4
1King’s College London, UK; 2The University of Hong Kong, Hong Kong SAR, China; 3Simon Fraser University, Vancouver, Canada; 4University of British Columbia, Vancouver, Canada
Background High-density high-rise cities have become more prominent globally. There is a need to better understand the extent to which vertical variation in air pollution and population mobility in such cities affect exposure and exposure-response relationships in epidemiologic studies.
Objectives 1) To investigate the behavior and distribution of vehicle emissions in a 3D urban landscape, 2) to develop, evaluate and demonstrate a dynamic 3D air pollution exposure model for Hong Kong (HK) and 3) to create an incremental exposure assessment methodology that can be applied in megacities across Asia.
Experimental Design Two (warm and cool season) street level spatial monitoring campaigns were undertaken to facilitate the creation of two-dimensional (2D) land use regression (LUR) models for NO, NO2, PM2.5 and BC. Continuous vertical air pollution monitoring was carried out at strategic residential locations for two weeks in the warm and cool seasons at four heights above street level. Paired indoor monitoring was included to calculate infiltration efficiencies. A population-representative travel behavior survey (n = 89,385) was used to produce the dynamic component of the model. Mortality risk estimates for an existing elderly cohort of 66,000 HK residents were estimated using increasing exposure model complexity.
Results The 2D LUR modelling predicted spatial patterns of air quality in HK that were consistent with the literature. Model R2 values ranged from 0.46 (NO2) to 0.59 (PM2.5). Vertical pollutant profiles supported the use of a single decay factor (k) for each pollutant across the whole region for derivation of the 3D exposure predictions (k = 0.004 and 0.012 for PM2.5 and BC respectively).
Median particle infiltration efficiencies (Finf
) were higher during the cool (91%) vs warm (81%) season, with a significant negative correlation with air conditioning use. Mean predicted population exposures for the dynamic model were 20% lower than the (non-dynamic) 2D model. Dynamic exposures to PM2.5
, BC, and NO2
were respectively 13%, 39% and 14% higher for those age <18, compared to those aged >65. Application of exposure estimates that incorporated infiltration, vertical and dynamic components produced effect estimates with greater magnitude for all-natural, cardiovascular and respiratory mortality outcomes compared to standard 2D LUR exposure estimates. For example, 2D vs. 3D model hazard ratios per inter quartile range increase of NO2
were 1.00 (0.97, 1.03) vs. 1.06 (95% CI:1.03-1.08) and 1.00 (0.95, 1.05) vs. 1.09 (1.04-1.14) for all-natural cause and cardiovascular mortality respectively.
Conclusion The results from this study provide direct evidence of the benefit to epidemiological studies of considering air pollution exposure in a dynamic 3D landscape. Associations were found between mortality and pollutant exposures that would not have been observed had standard 2D LUR or satellite exposure models been used.