University of California Transportation Center
UCTC- FR- 2010- 24
Fine Particulate Concentrations Near Arterial Streets: The Influence
of Building Placement and Wind Flow
Marlon Boarnet, Gavin Ferguson, and
Rufus Edwards
University of California, Irvine
Marko Princevac, Christian Bartolome, and
Hansheng Pan
University of California, Riverside
August 2010
Fine Particulate Concentrations Near Arterial Streets: The Influence of Building
Placement and Wind Flow
Submission Date: 7/ 31/ 09
Word Count: 4,103
Number of tables and figures: 12
Marlon Boarnet ( Corresponding Author)
Professor
Department of Planning, Policy, & Design
University of California, Irvine
Irvine, CA 92697
Telephone: ( 949) 824- 7695
Fax: ( 949) 824- 8566
E- mail: mgboarne@ uci. edu
Gavin Ferguson
Graduate Student
Department of Planning, Policy, & Design
University of California, Irvine
Irvine, CA 92697
E- mail: fergusog@ uci. edu
Rufus Edwards
Assistant Professor
Department of Epidemiology
University of California, Irvine
Irvine, CA 92697
Telephone: ( 949) 824- 4731
Fax: ( 949) 824- 0529
E- mail: edwardsr@ uci. edu
Marko Princevac
Assistant Professor
Department of Mechanical Engineering
University of California, Riverside
Riverside, CA 92521
Telephone: ( 951) 827- 2445
Fax: ( 951) 827- 2899
E- mail: marko. princevac@ ucr. edu
Christian Bartolome
Graduate Student
Department of Mechanical Engineering
University of California, Riverside
Riverside, CA 92521
E- mail: christian. bartolome@ email. ucr. edu
Hansheng Pan
Graduate Student
Department of Mechanical Engineering
University of California, Riverside
Riverside, CA 92521
E- mail: hansheng. pan@ email. ucr. edu
ABSTRACT
This paper provides preliminary evidence that the placement of buildings influences the
concentration of fine particulates by altering wind flow. The authors collected measurements of
fine particulate concentration, wind speed, wind direction, and traffic levels around five Southern
California arterials selected to represent a range of building densities. In some cases the
difference in average concentrations between opposite sides of the street was on the order of 10
μg/ m3. In most cases the concentration was higher on the upwind side of the street, where the
wind wakes of buildings limit the dispersion of particulates. Although this work is exploratory in
nature, it reveals that fine particulate concentrations can vary even within a single city block, a
scale finer than those used in current policy models. Given the trend towards infill development
and densification in many places, this is an important topic that warrants further research to more
fully understand the influence of the built environment on air quality.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 1
1 I. INTRODUCTION
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
In this study, we give results that illuminate two gaps in the literature on fine particulate
concentrations in urban areas. First, little is known about fine particulate concentrations along
major arterial streets, as opposed to our relatively greater knowledge of particulate
concentrations near freeways. We offer some of the first evidence on that topic. Second, patterns
of wind transport and hence particulate concentration and dispersion may be affected by building
placement. Computer simulations have shown that fine particulates may be trapped in wakes in
the wind shadows of buildings ( 1), creating fine- grained differences in concentrations that may
be important in higher density, infill developments of the sort that have become popular in the
past decade. The goals of this research are twofold: ( 1) To measure particulate concentrations
near major arterials in five neighborhoods chosen to represent built environments that range from
low density auto- oriented development to dense urban settings, and ( 2) To get exploratory
evidence on factors associated with differences in fine particulate concentrations at a fine level of
geographic detail.
Fine particulate concentrations ( in this paper, particulate matter with diameters 2.5
micrometers or less, PM2.5) have been associated with several thousand premature deaths in
California alone each year ( 2). Fine particulates concentrate near the source; past research near
urban freeways has shown the highest concentrations within 100 meters of the freeway, dropping
to background levels at distances of approximately 300 meters from the freeway ( 3). Most
studies of fine particulates have focused on emissions from and concentrations near freeways ( 3,
4) or along routes that are heavily trafficked by heavy duty diesel trucks ( 5). In the Los Angeles
region ( the context for this study), average annual daily traffic ( AADT) on major arterial streets
can be one- third the level of AADT on the busiest freeways. ( Consider this example from 2005
Caltrans data: AADT on Beach Boulevard in Orange County near Warner is 81,000, compared to
AADT between 270,000 and 311,000 for Interstate 405 in West Los Angeles, between Interstate
105 and Mulholland Boulevard.) As infill and compact development become more popular,
building patterns will place humans in closer proximity to heavily trafficked arterial streets,
increasing the need to understand fine particulate concentrations and how the built environment
can influence particle concentration and dispersion in urban environments.
We hypothesize that the factors associated with fine particulate concentrations include:
( 1) meteorological conditions, especially wind direction and wind speed, ( 2) the built
environment, ( 3) location within the region ( location in the Los Angeles metropolitan area is
strongly associated with wind conditions, due to prevailing onshore breezes, and with the built
environment), and ( 4) traffic. Our results, while preliminary, show variation with the built
environment, with the highest fine particulate concentrations in the study area that is best
characterized as compact development. Beyond variation in fine particulate concentrations
associated with the study areas, we find that wind direction interacts with building placement to
create patterns of higher particulate concentration in building wakes. To date, our preliminary
results do not give evidence of associations with particulate concentrations and traffic flow,
beyond what would be explained by variations across the study areas and times of day, but we
caution that our initial focus has been more descriptive and exploratory.
II. STUDY DESIGN
We chose major arterial streets ( AADT > 40,000) in five study areas, with each study area
representing different development patterns: ( 1) low density ( structures are 1- 2 stories), ( 2) low-rise
( 3- 4 stories), ( 3) mid- rise ( 10- 20 stories), ( 4) high- rise (> 20 stories), and ( 5) strip mall ( large
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 2
surface parking separating buildings and the arterial.) The study neighborhoods, by settlement
type, are in the cities of Anaheim ( low rise, Harbor Boulevard is the arterial street), Pasadena
( low- rise, Colorado Boulevard), downtown Long Beach ( mid- rise, Ocean Boulevard), downtown
Los Angeles ( high- rise, Grand Avenue), and Huntington Beach ( strip mall, Beach Boulevard.)
Each study area is approximately 0.25 square kilometers ( 500 meters by 500 meters) surrounding
the selected arterial. In most areas, there are more than one major arterials of focus within the
study areas; study areas were chosen for settlement patterns and we focus on the small 0.25
square kilometer areas, here denoted by the name of the city. Representative photos of building
patterns in each study area are shown in Table 1. For reference, the Pasadena site is most typical
of compact development.
1
2
3
4
5
6
7
8
9
10
11
12
TABLE 1 Representative Building Types in Each Study Area
Low density
settlement
Low- rise
settlement
Mid- rise
settlement
High- rise
settlement A strip mall
Stories 1 to 2 stories 3 to 5 stories 10 to 20 stories > 20 stories 1 to 2 stories
City Anaheim Pasadena Long Beach Los Angeles Huntington
Beach
Arterials Harbor Blvd. East Colorado
Blvd.
East Ocean
Blvd.
6th Ave. Beach Blvd.
Typical
buildings
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
III. DATA COLLECTION
In each study area, field measurements were conducted on three consecutive days between June
19 and August 1, 2008. Fine particulate ( PM2.5) concentrations were measured using six
stationary DustTrak ( TSI, Inc.) monitors during three time periods on each observation day – the
morning rush hour ( 7 to 9 a. m.), mid- day ( 11 a. m. to 1 p. m.), and the evening rush hour ( 4 to 6
p. m.) Traffic was videotaped using three video cameras ( JVC) and later was counted and
classified by watching the videotapes. Wind speed and direction were measured using a sonic
anemometer located within each study area; whenever possible, the sonic anemometer was
located on the roof of a parking structure or in an elevated or open area to obtain background
wind speed and direction relatively independent of building structures.
In addition to the stationary measurements, at each study site the research team walked
designated routes carrying DustTraks and video cameras to compare particulate concentrations
along street segments. Typically two to three routes were walked per site, approximately three
times per route, and each walkthrough lasted from three to 30 minutes. These walkthrough data
are of particulate interest as, in some study areas, parallel DustTraks were carried simultaneously
on opposite sides of the same street, allowing fine- grained comparisons of PM2.5 concentrations
in small geographic areas, in ways that can give insight into the relationship between fine
particulate concentration, wind direction and speed, and the built environment.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 3
For the walkthrough data, research team members carried a DustTrak in a backpack with
a 32” aluminum tube attached to the DustTrak inlet protruding. The DustTrak data provide
second- by- second measurements of particulate matter concentration. Members of the research
team also recorded traffic conditions with a handheld video camera while walking the routes.
Traffic counts concurrent with the DustTrak readings were obtained from the resulting video
files.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
For traffic counts, vehicles were classified as cars, trucks, or buses to broadly account for
the different emission factors of different classes of vehicles. The car category includes light
trucks and smaller vehicles. The truck category includes delivery trucks and larger vehicles. We
organized the counts into street segments and intersections. The endpoints of a street segment are
the cross- streets at either end, and the traffic count for that segment is the number of vehicles that
passed along the adjacent street in either direction while the person walked from one endpoint to
the other. ( To control for differing lengths of street segments and hence differing time spent
walking the route, traffic counts are converted into traffic per minute in all cases in the analysis
presented here.) The traffic count for an intersection is the number of vehicles that passed
through the intersection in any direction from the time the person reached the intersection to the
time he or she reached the opposite side of the street. Thus, the intersection counts include traffic
passing during the time spent standing at the intersection, time spent waiting to cross the
intersection, and the time spent walking across the street. We obtained the time to the nearest
second at each cutoff point from the video files.
For walkthrough data analysis, background wind speed and direction were obtained from
the sonic anemometer, which gives readings at 0.1 second frequencies. Wind speed was averaged
for the walkthrough time period, and for wind direction we focus on the median direction during
the walkthrough time period.
We augmented the field data with building density measures obtained from parcel records
for the three study areas in Los Angeles County. The parcel records contain the square feet of
floor space in each parcel. We calculated the area of each parcel from the corresponding
shapefile using ArcMap 9.2.
IV. RESULTS
Descriptive statistics for particulate concentrations are shown in Table 2 ( walkthrough data only)
and in Table 3 ( stationary DustTrak data), organized by study area. In general, the Pasadena and
Long Beach study areas have higher PM2.5 concentrations.
TABLE 2 Walkthrough DustTrak PM2.5 Concentration Data ( In Micro- grams per Cubic
Meter, μg/ m3)
City n Mean Mina
1st
Pctl.
25th
Pctl.
50th
Pctl.
75th
Pctl.
99th
Pctl. Max
Anaheim 1,987 29.8 - 7.3 14.7 26.1 28.6 30.8 46.3 1,874.1
Huntington Beach 13,262 31.4 12.0 17.0 28.0 31.8 35.0 46.8 554.9
Long Beach 8,072 36.6 12.5 15.3 19.7 45.2 49.7 60.4 304.3
Los Angeles 21,726 27.8 - 8.2 18.4 24.4 27.0 29.6 50.8 1,418.3
Pasadena 13,097 53.0 26.8 44.1 49.8 52.5 55.2 68.4 1,315.8
aNegative DustTrak readings are rare aberrations. Of the 58,144 walkthrough DustTrak readings, five
were negative values. Discarding those does not change the analysis.
37
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 4
TABLE 3 Stationary DustTrak PM2.5 Concentration Data ( In Micro- grams per Cubic
Meter, μg/ m3)
1
2
City n Mean Mina
1st
Pctl.
25th
Pctl.
50th
Pctl.
75th
Pctl.
99th
Pctl. Max
Anaheim 376,813 39.4 - 90.6 7.7 28.6 35.2 47.3 77.6 2,032.1
Huntington Beach 385,060 41.9 - 208.0 15.5 30.7 38.6 49.2 92.0 3,331.4
Long Beach 322,880 54.1 - 517.9 17.5 41.5 55.0 64.4 95.9 457.5
Los Angeles 324,874 41.7 - 674.6 15.3 28.0 39.8 54.2 77.1 9,585.2
Pasadena 385,280 60.4 - 414.0 21.9 48.0 57.0 74.0 114.8 30,929.0
aNegative DustTrak readings are rare aberrations. Of the 1,794,907 stationary DustTrak readings, 656
were negative values. Discarding those does not change the analysis.
3
4
5
6
7
8
9
10
11
12
13
14
15
A. Differences by Side of Street
In walkthrough routes in two of the study areas, two people walked the same route in tandem,
one on each side of the street. Those parallel routes are shown in Figures 1 and 2 ( respectively
the Huntington Beach and Pasadena study areas), with the median wind direction during each
day’s walkthrough shown on the figures. ( There were typically one to three consecutive passes
through the walkthrough route on each observation day.) Figures 1 and 2 show all walkthrough
routes, which includes both parallel routes ( shown as straight lines) and loop routes. Here we
only focus on the straight- line routes, as those were the only routes where observers with
portable DustTrak devices walked in tandem on opposite sides of the street. Those straight line
routes are Huntington Beach Routes A ( Garfield Avenue) and B ( Beach Boulevard) and
Pasadena Route B ( Colorado Boulevard.).
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 5
1
2
3
FIGURE 1 Huntington Beach study area and walkthrough routes.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 6
1
2
3
4
5
6
7
8
9
10
11
FIGURE 2 Pasadena study area and walkthrough routes.
Table 4 displays the results of two- sample t- tests comparing average PM2.5 concentrations
on opposite sides of the street for each case. The fifth column presents the sample difference in
mean concentration and associated t- statistics for each route on each day. The sixth column
displays the values of the same statistics obtained when we exclude the top one percent of
measurements to eliminate outliers. The last two columns indicate the median wind direction and
mean horizontal wind speed during each of the four cases.
TABLE 4 Two Sample T- Tests, Walkthrough Data for PM2.5 Concentration
City Route Date n
Difference
in mean
conc.
( t- statistic)
Difference
in mean
conc.,
highest 1%
excluded
( t- statistic)
High
conc.
side of
street
Median
wind
direction
Mean
horizontal
wind speed
( m/ s)
Huntington
Beach
A 7/ 16/ 08 2370 3.40
( 11.77)
2.78
( 24.46)
South 216°
( SW)
1.41
A 7/ 18/ 08 1748 13.52
( 20.96)
14.35
( 76.41)
South 228°
( SW)
1.18
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 7
A 7/ 21/ 08 1624 8.19
( 52.74)
8.01
( 56.57)
South 259°
( W)
1.60
B 7/ 16/ 08 2360 4.82
( 5.25)
1.59
( 12.73)
East 221°
( SW)
1.32
B 7/ 18/ 08 2704 6.82
( 41.16)
6.45
( 64.71)
East 228°
( SW)
1.12
B 7/ 21/ 08 2456 10.46
( 98.09)
10.41
( 117.27)
West 259°
( W)
1.58
Pasadena B 7/ 23/ 08 2532 2.51
( 2.44)
1.05
( 7.62)
North 212°
( SW)
1.21
B 7/ 25/ 08 1792 0.39
( 2.65)
0.10
( 0.81)
South 142°
( SE)
0.80
B 7/ 29/ 08 2522 2.40
( 13.75)
2.10
( 13.41)
South 220°
( SW)
1.66
1
2
3
4
5
6
7
8
9
The raw walkthrough PM2.5 concentration data generally display a relatively uniform
pattern throughout the walkthrough routes. As an example, Figure 3 shows walkthrough data, on
both sides of Garfield Avenue, on July 18. The graph plots PM2.5 concentration versus time
separately for each side of the street for comparison. We zeroed the time scale and eliminated
gaps between repeated passes of the same route for the sake of presentation.
FIGURE 3 PM2.5 Concentrations, Huntington Beach Route A, July 18.
E on Garfield
E at Beach & Garfield
E on Garfield
W on Garfield
W at Beach & Garfield
W on Garfield
E on Garfield
E at Beach & Garfield
E on Garfield
0
10
20
30
40
50
60
70
80
90
100
μg/ m ³
00: 00
02: 06
03: 25
04: 53
06: 25
07: 23
09: 30
11: 34
13: 21
14: 34
Zeroed time scale ( mm: ss)
North sidewalk South sidewalk
July 18, 2008
Huntington Beach, Route A
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 8
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
In Huntington Beach for Route A ( Garfield Avenue), the results show uniformly higher
PM2.5 concentrations on the south side of the street, which on all observation days was in the
wind shadow of buildings ( the wind was generally from a southwesterly or, on July 21, a west-southwesterly
direction.) This is consistent with computer modeling results ( 1) that demonstrate
the possibility that fine particulates can concentrate in eddies on the lee side ( in the wind
shadow) of buildings. The results from Route A in Huntington Beach provide empirical evidence
and support our hypothesis that the built environment and meteorological conditions can interact
in ways that trap or disperse particulates in urban settings. For Route B in Huntington Beach
( Beach Boulevard), readings only differed by side of the street on two of the three days, with
higher concentrations once on the east side and once on the west side. The east side of Beach
Boulevard is not in the wind shadow based on wind direction on observation days, hence the July
21 Route B result for Huntington Beach is not consistent with the general pattern of results or
with our hypothesis that particles are trapped in the wind shadow of buildings.
B. Comparison of Paired Stationary DustTrak Measurements
In addition to collecting data while walking, we also collected data for three two- hour periods
each day at fixed locations. In some cases, the stationary DustTrak monitors were on opposite
sides of the street or in a “ windward / leeward” pattern that created the opportunity for DustTraks
to be paired with one another, with one DustTrak in the pair in a building’s wind shadow while
the other DustTrak in the pair is not. The locations of these paired DustTraks are shown in
Figures 4- 6.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 9
1
2
3
FIGURE 4 Paired DustTrak locations in Los Angeles study area.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 10
1
2
3
FIGURE 5 Paired DustTrak locations in Pasadena study area.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 11
1
2
3
4
5
6
7
8
9
10
11
12
13
FIGURE 6 Paired DustTrak locations in Long Beach study area.
Table 5 displays results of two- sample t- tests comparing average particulate matter
concentrations from sites on opposite sides of the street for three sites. In all but one case ( Long
Beach, July 9), statistically significant differences show higher concentrations on the “ wind
shadow” side of the street. Note that for Long Beach, the paired DustTraks are located, relatively
to wind direction, in a manner that is a less good fit with an “ upwind/ downwind” or “ wind
shadow” configuration, and hence less weight might be attached to the Long Beach paired
DustTrak results. For the other two study areas, the paired Stationary DustTraks give results that
are consistently similar to the dominant pattern from the walkthrough data; locations in the wind
shadow of buildings have higher PM2.5 concentrations.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 12
1 TABLE 5 Paired Stationary DustTrak PM2.5 Concentration Comparisons
City Date Street n
Difference
in mean
conc.
( t-statistic)
Difference
in mean
conc.,
highest
1%
excluded
( t- statistic)
High
conc.
side
of
street
Median
wind
direction
Mean
horizontal
wind
speed
( m/ s)
Long
Beach
7/ 2/ 08 Ocean
Blvd.
42,668 14.64
( 86.95)
14.79
( 94.85)
S 262°
( W)
1.10
7/ 7/ 08 Ocean
Blvd.
42,792 7.20
( 50.65)
6.79
( 50.54)
S 256°
( W)
0.86
7/ 9/ 08 Ocean
Blvd.
42,612 11.09
( 91.55)
11.30
(> 100)
N 244°
( SW)
1.05
Los
Angeles
6/ 19/ 08 6th St. 26,340 6.15
( 15.65)
4.25
( 63.94)
SW 198°
( S)
0.69
6/ 23/ 08 6th St. 41,911 1.94
( 14.37)
1.31
( 15.68)
SW 216°
( SW)
0.79
6/ 30/ 08 6th St. 42,017 3.79
( 6.90)
1.55
( 8.69)
SW 207°
( SW)
0.77
Pasadena 7/ 23/ 08 Colorado
Blvd.
42,659 3.10
( 0.25)
2.35
( 9.35)
S 146°
( SE)
1.04
7/ 25/ 08 Colorado
Blvd.
42,200 2.52
( 10.55)
1.64
( 27.64)
S 143°
( SE)
0.74
7/ 29/ 08 Colorado
Blvd.
42780 3.29
( 27.87)
2.93
( 26.01)
S 171°
( S)
0.92
2
3
4
5
6
7
8
9
10
11
12
14
16
17
19
21
22
23
C. Regression Analysis
We next ran several regressions using the data from three cities for which parcel data is
available: Long Beach, Los Angeles, and Pasadena. The dependent variable is the average one-second
particulate matter concentration for each traversal of a street segment or intersection for
all walkthrough data. There are multiple observations for each street segment and intersection
because team members traversed each route multiple times and some routes share common
segments. To provide a buffer between street segments and adjacent intersections, we calculated
the average for street segments using only readings from the middle 50% of the segment.
The independent variables are:
• Cars per minute: Count of cars and light trucks divided by time span of the
13 traversal in minutes
• Heavy trucks per minute: Count of delivery trucks or larger vehicles ( except
15 buses) divided by time span of the traversal in minutes
• Buses per minute: Count of buses divided by time span of the traversal in minutes
• Floor- area- ratio: Sum of built square feet divided by sum of parcel square feet for
18 parcels within 70ft for street segments or 100ft for intersections
• Indicator variables for city, date, and whether the observation corresponds to an
20 intersection
Table 6 displays the results of the regressions. Huber- White standard errors are in
parentheses. Specification ( a) includes only the traffic flow variables, the floor- area ratio, and the
intersection indicator. All of the variables except for car flow have the opposite of the expected
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 13
sign, perhaps due to confounding variables that differ among cities. To control for this, we
include city indicator variables in specification ( b). As expected, the city indicator variables are
significant, but the explanatory variables of interest are no longer significant. In specification ( c)
we attempt to control for unobserved factors more thoroughly by including indicator variables
for each day in addition to each city. The coefficient on floor- area- ratio alone becomes
significant, and it is positive.
1
2
3
4
5
6
7
8
TABLE 6 Regression Results
Variable ( a) ( b) ( c)
Cars per minute 0.19***
( 0.04)
0.03
( 0.03)
0.02
( 0.02)
Heavy trucks per minute - 0.09
( 0.62)
- 0.23
( 0.36)
0.12
( 0.15)
Buses per minute - 1.89***
( 0.32)
- 0.18
( 0.13)
- 0.06
( 0.10)
Floor- area- ratio - 1.45***
( 0.11)
0.07
( 0.06)
0.12*
( 0.06)
Intersection - 1.30
( 1.07)
- 0.45
( 0.70)
- 0.48
( 0.37)
Long Beach 7.31***
( 1.54)
- 8.38***
( 0.86)
Pasadena 24.97***
( 0.59)
22.57***
( 0.56)
Date variables 3 of 4 significant at
0.001 level
Constant 41.20***
( 1.03)
27.59***
( 0.61)
28.73***
( 0.53)
n 654 654 654
*: p < 0.05; **: p < 0.01; ***: p< 0.001
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
The floor- area- ratio ( FAR) variable may, to some extent, be correlated with the study
area dummy variables, but note that the FAR variable as constructed is a much more micro-measure
of the built environment, smaller than the overall study areas.
Overall, we prefer the model that controls for both study area and day of observation. The
clear differences in PM2.5 concentration across study areas suggests the need to control for the
study area. The association with weather patterns ( particularly wind direction) suggests that it
would be wise to control for at least macro meteorological effects through the ( admittedly rough)
method of including day dummy variables. The resulting model, in column ( c), shows a
significantly positive relationship between development intensity ( FAR) and measured PM2.5
concentrations, again providing evidence of an association between fine particulate
concentrations and the built environment. The lack of an association with traffic intensity should
not be regarded as definitive. Due to patterns of atmospheric transport and dispersion, temporal
leads and lags for traffic may be important predictors of fine particulate emissions, and possibly
traffic intensity may have to be interacted with wind speed and direction. We have not
experimented with such considerations yet in a regression model.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 14
1 V. CONCLUSION
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
At this point, the evidence suggests associations between fine particulate concentrations, wind
direction, and building placement. We note that the built environment features that are
apparently associated with PM2.5 concentrations are as fine- grained as building- to- building
variations, at geographic scales smaller than a city block. The differences in concentrations on
opposite sides of the street averaged, in some cases, on the order of 10 μg/ m3. These differences
are outside of current policy models of particulate concentration ( e. g. Caline4), and hence
outside of current policy discussions. Our results should inform both modeling and policy.
We do not believe that PM2.5 concentrations should be used as a reason not to pursue
compact development. The evidence here would not support such a strong policy conclusion, and
any assessment of the appropriateness of compact development should include a comprehensive
assessment of benefits and costs. Even a focus only on air quality would have to weigh the likely
countervailing impacts of compact development on vehicle miles of travel and the effect of
concentrating traffic and development in places where humans would be in closer proximity to
PM2.5 concentrations. Such analyses, and links from concentration to human exposure and to
human health, would require considerably more evidence than is available here.
Our intention is to highlight the interaction of PM2.5 concentrations and the built
environment. The evidence here supports the hypothesis that fine particulate concentrations are
influenced by the built environment in ways that are currently outside of modeling traditions and
policy discussions. Furthermore, our results suggest that the association, while not uniform, may
have enough regularity to lead to policy rules of thumb and modeling methods after further
research.
ACKNOWLEDGEMENTS
This research was funded by the University of California Transportation Center.
Boarnet, Ferguson, Edwards, Princevac, Bartolome, Pan 15
1
2
3
4
5
6
7
REFERENCES
1. Pan, H., M. Princevac, R. Edwards, A. Sfazl, M. Boarnet, J. Wu, and R. Lejano. Field,
Laboratory, and Numerical Study on Flow and Dispersion of PM2.5 in Southern California Cities.
Presented at 89th American Meteorological Society Meeting, Phoenix, AZ, Jan. 11- 15, 2009.
2. California Air Resources Board. Methodology for Estimating Premature Deaths Associated
with Long- term Exposure to Fine Airborne Particulate Matter in California. Staff Report.
October 24, 2008. www. arb. ca. gov/ research/ health/ pm- mort/ pm- mort_ final. pdf. Accessed July
30, 2008.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
3. Zhu, Y., W. C. Hinds, S. Kim, S. Shen, and C. Sioutas. Study of Ultrafine Particles near a
Major Highway with Heavy- Duty Diesel Traffic. Atmospheric Environment, Vol. 36, No. 27,
2002, pp. 4323– 4335.
4. Houston, D., P. Ong, J. Wu, and A. Winer. Proximity of Licensed Child Care Facilities to
Near- Roadway Vehicle Pollution. American Journal of Public Health Vol. 96, No. 9, 2006, pp.
1611– 1617.
5. Houston, D., M. Krudysz, and A. Winer. Diesel Truck Traffic in Low- Income and Minority
Communities Adjacent to Ports: Environmental Justice Implications of Near- Roadway Land Use
Conflicts. Transportation Research Record: Journal of the Transportation Research Board, No. 2067,
Transportation Research Board of the National Academies, Washington, D. C., 2008, pp. 38– 46.