Assessing Dynamic Exposure to Air Pollution - Space-Time Integration in Geography and GIScience - page 280

exposures make up a disproportionately large proportion of individual exposure
to ambient (and especially traffic-related) air pollution. A static analysis of air
pollution exposure may thus lead to serious underestimation of people's actual
exposures.
Additionally, proximity-based assessment is heavily based on traffic emissions
that are generated by a combination of various types of vehicles. But levels of
emissions may vary considerably with different mix of cars and trucks that occupy
the roadways (Kanaroglou et al. 2000 ; Gertler 2003 ). Since most previous studies
did not distinguish vehicle type when making emission estimates, assessments of
people's air pollution exposure may be inaccurate. Further, dispersion characteris-
tics of air pollution may be influenced by topographical and meteorological factors.
The isotropic dispersion assumption that air borne particles experience the same
dispersion pattern in all directions surrounding the study area can be invalid in
most situations (Jerrett et al. 2005 ). It is unrealistic to assume that wind speed,
wind direction, weather patterns, atmospheric stability, and land use patterns have
no effect on air pollution dispersal. Thus, proximity-based assessment tends to
oversimplify the analysis of air pollution exposure. Reality encompasses much more
complex scenarios - for example, human activities take place at many locations
throughout the day and external factors influence the dispersal of particulate matter
concentrations.
Geostatistical interpolation is another approach to pollution assessment. It uses
deterministic or stochastic geostatistical techniques to estimate pollution levels at
unsampled locations across a study area, and uses these values to interpolate a con-
tinuous surface of pollution concentrations for the entire area (e.g., Farhang 1983 ;
Pengelly et al. 1984 ; Finkelstein et al. 2003 ). A critical component of this method is
a dense network of monitoring stations that is distributed throughout the study area.
Several interpolation techniques were used to estimate pollution concentrations,
including inverse distance weighting, splines, Theissen triangulation, and kriging.
Among them kriging is the most commonly used method in air pollution studies
(Jerrett et al. 2001 ). An advantage of kriging interpolation over the other methods is
the establishment of both predicted values at unsampled locations and their standard
errors. Having access to error estimates helps inform the researchers about where
predicted values may be more or less reliable.
Implementing interpolation techniques in air pollution studies is a more sophisti-
cated approach when compared to proximity-based assessments. But this approach
also has its limitations. While interpolation methods utilize real pollution measure-
ments for computing exposure estimates, data availability may be an issue. Jerrett
et al. ( 2005 ), for instance, estimated that a dense network of sampling sites ranging
from 10 to 100 stations is required for meaningful interpolation results depending
on the size of the study area. But it is difficult to obtain an adequate geographic
coverage of pollution concentration readings for certain areas since government
monitoring stations are mostly located in heavily populated areas or near major
roadways or industrialized areas and are sparsely scattered in rural areas (Bell
2006 ). Further, the temporal resolution of monitored data is often low (e.g., on 6-day
intervals or at particular hours of the day). As a result, estimates for days or times