Subscriber access provided by CORNELL UNIVERSITY LIBRARY
Article
Relationships between Changes in Urban Characteristics and Air Quality in East Asia from 2000 to 2010 Andrew Larkin, Aaron van Donkelaar, Jeffrey A. Geddes, Randall V Martin, and Perry Hystad Environ. Sci. Technol., Just Accepted Manuscript • DOI: 10.1021/acs.est.6b02549 • Publication Date (Web): 21 Jul 2016 Downloaded from http://pubs.acs.org on July 21, 2016
Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.
Environmental Science & Technology is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.
Page 1 of 26
Environmental Science & Technology
Relationships between Changes in Urban Characteristics and Air Quality in East Asia from 2000 to 2010
1 2 3
Andrew Larkin1*, Aaron van Donkelaar2, Jeffrey A. Geddes2, Randall V. Martin2,3, Perry Hystad1
4 5 6 7 8
1
College of Public Health and Human Sciences, Oregon State University, Corvallis, OR, USA, 97331
9 10
2
11
3
Department of Physics and Atmospheric Science, Dalhousie University, Halifax, Nova Scotia, Canada Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts, USA
12 13
*Corresponding email:
[email protected] 14 15
*Corresponding telephone: 001-541-737-2743
16 17 18 19 20 21 22 23 24 25 26 27 1 ACS Paragon Plus Environment
Environmental Science & Technology
28
Page 2 of 26
ABSTRACT Characteristics of urban areas, such as density and compactness, are associated with local
29 30
air pollution concentrations. The potential for altering air pollution through changing urban
31
characteristics, however, is less certain, especially for expanding cities within the developing
32
world. We examined changes in urban characteristics from 2000 to 2010 for 830 cities in East
33
Asia to evaluate associations with changes in nitrogen dioxide (NO2) and fine particulate matter
34
(PM2.5) air pollution. Urban areas were stratified by population size into small (100,000-
35
250,000), medium, (250,000-1,000,000) and large (>1,000,000). Multivariate regression models
36
including urban baseline characteristics, meteorological variables, and change in urban
37
characteristics explained 37%, 49%, and 54% of the change in NO2 and 29%, 34%, and 37% of
38
the change in PM2.5 for small, medium and large cities, respectively. Change in lights at night
39
strongly predicted change in NO2 and PM2.5, while urban area expansion was strongly associated
40
with NO2 but not PM2.5. Important differences between changes in urban characteristics and
41
pollutant levels were observed by city size, especially NO2. Overall, changes in urban
42
characteristics had a greater impact on NO2 and PM2.5 change than baseline characteristics,
43
suggesting urban design and land use policies can have substantial impacts on local air pollution
44
levels.
45 46 47
2 ACS Paragon Plus Environment
Page 3 of 26
48 49
Environmental Science & Technology
BACKGROUND For the first time in human history more than 50% of the world’s population resides in
50
urban areas, with an expected increase to 66% by 2050.1 Much of this growth is from, and will
51
continue to occur in, rapidly urbanizing East Asia (China, Japan, Indonesia, Malaysia, Vietnam,
52
Thailand, Korea, Philippines, Taiwan, Cambodia, Loa). In China alone the urban population
53
grew from 191 million in 1980 to 622 million in 2009.2 Only 36% of East Asia's population
54
currently lives in urban areas, suggesting substantial future urban growth in this region.3
55
Important public health gains have resulted from urban growth, such as increased sanitation, safe
56
drinking water, and access to education and healthcare; however, negative influences have also
57
emerged, including increases in outdoor air pollution levels.
58
Global burden of disease estimates for 2013 include 2.9 million premature deaths and 70
59
million years of healthy life lost from exposure to outdoor fine particulate matter (PM2.5) air
60
pollution.4 This places outdoor PM2.5 air pollution ahead of such risk factors as physical
61
inactivity, alcohol use, and high body-mass index in terms of disease burden in East Asia. Air
62
pollution levels in this part of the world have also deteriorated from 2000-2010,5 potentially due
63
to rapid urbanization and accompanying air pollution emissions from energy production,
64
industry, home heating and vehicles.
65
Planning of urban areas and land use policies can have important implications for local
66
air pollution concentrations. A number of studies, primarily in developed countries, have
67
documented that urban form characteristics, such as compactness (or alternatively sprawl),6
68
green space7, population density8 and road density9 can significantly improve or degrade local air
69
pollution levels. For example, in 111 US urban areas, an interquartile range change in population
70
density, centrality and transit supply was associated with 4-14% change in PM2.5 air pollution
3 ACS Paragon Plus Environment
Environmental Science & Technology
Page 4 of 26
71
concentrations; impacts equivalent in size to those from meteorological conditions.10 However,
72
studies examining urban characteristics and air pollution in rapidly developing countries are rare.
73
One study conducted using 83 global cities in the World Bank’s Urban Growth Management
74
Initiative observed that, on average, cities with highly contiguous urban form had lower nitrogen
75
dioxide (NO2) air pollution (a marker for traffic-related air pollution), while urban compactness
76
was not associated with NO2 concentrations.8 Importantly, no analyses have been conducted on
77
the change in air pollution associated with urban expansion and changing urban characteristics –
78
information vital to informing urban planning. Much of the existing literature also focuses on
79
large urban areas, despite the fact that approximately 50% of the world’s urban population reside
80
in relatively small cities of less than 500,000 inhabitants,1 and that small urban areas may be
81
more amenable to policy and design changes that could reduce air pollution.
82
The degree to which urban growth increases air pollution is complex and an important
83
policy question that has not been fully explored, especially in rapidly developing countries. Here
84
we examined changes in NO2 and PM2.5 air pollution from 2000 to 2010 using satellite derived
85
estimates for 830 cities in East Asia with 2010 populations’ ≥100,000, and examine associations
86
with urban expansion levels and changes in urban characteristics over this ten-year period.
87 88
METHODS
89
Study Region and Urban Areas
90
The geographical extent of our study region covers all of Eastern Asia, including Oceania
91
island nations such as Marshall Islands and Micronesia (Figure 1, created using ArcGIS 10.3.1).
92
We used World Bank 2000 and 2010 East Asia urban area classifications3 to define urban areas
4 ACS Paragon Plus Environment
Page 5 of 26
Environmental Science & Technology
93
based on the extent of urban development rather than municipal boundaries. City footprints were
94
esimated using a slight modification of an equation originally derived by Angel et al.11
95
96
1 if classified as urban by the World Bank Urban Expansion dataset Footprint status = 1 if 20% or more of the area within a 1 km radius is classified as urban 0 otherwise
97
Urban area population size was calculated by multiplying footprint size and mean
98
population density from Gridded Population of the World population density estimates.12 Urban
99
areas with 2010 populations ≥100,000 were selected for analysis, which included 830 urban
100
areas (Figure 1). Urban areas were stratified by 2010 population size based on US Census
101
Bureau city size (by population) classification due to preliminary observations of differing
102
typology-air pollution relationships and covariate correlations between urban areas with small
103
(100,000-250,000), medium (250,000-1,000,000) and large (> 1,000,000) populations (Figure
104
S1)
105
Air Pollution Estimates
106
Ground-based air pollution monitoring data are not available for most urban areas in East
107
Asia in 2010 and were extremely sparse in 2000. We therefore used satellite-derived estimates of
108
NO2 and PM2.5 to determine changes from 2000 to 2010. Satellite data offer global coverage and
109
a uniform approach to measuring air pollution concentrations. Estimates of NO2 for 1999-2001
110
and 2009-2011 were derived from the GOME and SCIAMACHY satellite instruments, which
111
provide measurements of NO2 atmospheric column abundance with global coverage every three
112
and six days respectively. Three-year averages were used to reduce the effects of inter-annual
113
variability that may in part arise from random error. A global chemical transport model (GEOS-
114
Chem) was used to relate the column observations to surface concentrations as described in
115
Geddes et al.13 In their approach, GOME observations were downscaled to achieve the finer 5 ACS Paragon Plus Environment
Environmental Science & Technology
Page 6 of 26
116
resolution captured by SCIAMACHY by applying the mean spatial structure observed by
117
SCIAMACHY at 0.1° x 0.1° over 2003-2005. There is no evidence of a systematic discontinuity,
118
and a high correlation (r=0.89) between the two instruments was observed during a period of
119
overlap.13 PM2.5 was estimated from retrievals of aerosol optical depth from three satellite
120
instruments (MODIS, MISR, and SeaWIFS) with coincident aerosol vertical profiles from
121
GEOS-Chem.14 Three year average surfaces (at a 10x10 km resolution) were also created (1999-
122
2001 and 2009-2011). These estimates have shown good agreement with ground-based
123
measurements and have an expected 1-sigma uncertainty of 1 µg/m3 +25%.15 Similar satellite-
124
derived PM2.5 estimates have been used in recent global epidemiological studies,16,17 and were
125
the basis of the 2013 global disease burden estimates.4,5
126
Calculations of Change in Air Pollution and Urban Characteristics
127
Air pollution, urban expansion, and urban typology variables are listed in Table 1, with
128
data sources listed in Table S1. Variables are expressed as absolute rather than relative levels of
129
change from 2000-2010. Air pollution variables include change in mean PM2.5 and NO2 from
130
2000-2010. Urban expansion variables include change in population and urban footprint area
131
from 2000-2010. Urban characteristic variables include changes in compactness index, a
132
measure of urban sprawl (Figure S2) and changes in mean and within-city-variation in
133
population density, green space (calculated from the normalized difference vegetation index
134
(NDVI)), impervious surface area (ISA), and percent lights at night, a satellite-based relative
135
measure of global nighttime light emissions often used as an indirect measure of economic
136
activity.18 Measures of within-city-variation were calculated in addition to mean levels to
137
evaluate the potential influence of spatial heterogeneity in urban typology on air pollution. For
6 ACS Paragon Plus Environment
Page 7 of 26
Environmental Science & Technology
138
example, increased variation in population density is an indirect indicator of urban sprawl (low
139
density suburban areas contrasting with high density urban centers). Change in mean air pollution and urban typology variables were calculated as the
140 141
difference in means for the corresponding 2000 and 2010 urban footprint: & #'()($'()( − !"#'((($'((( ∆!"#$ = !
142 143
Eq. 1
Where ∆!"#$ denotes the change from 2000-2010 in the variable i mean for urban area j
144 145
!"#'((($'((( denotes the year 2000 variable i mean for the area covered by the year 2000
146
footprint of urban area j
147
Similarly, changes in within-city-variation were calculated as the difference in standard deviation
148
for the corresponding 2000 and 2010 urban footprint: ∆+#$ = +#'()($'()( − +#'((($'(((
149 150
Eq. 2
Where
151
∆+#$ denotes the change from 2000-2010 in the standard deviation of variable i
152
for urban area j
153
+#'((($'((( denotes the year 2000 variable i standard deviation for the area covered by the
154
year 2000 footprint of urban area j
155
Cities with urban footprints smaller than 10km2 in area were assigned a within-city-variation
156
value of zero due to the spatial resolution of several urban characteristic datasets.
157
Due to urban spillover there were several regions where a single 2010 urban footprint
158
covered multiple distinct year 2000 urban areas. For these instances, changes in air pollution,
159
urban expansion, and typology variables were calculated using an area-weighted average of
160
corresponding 2000 urban areas. 7 ACS Paragon Plus Environment
Environmental Science & Technology
& #'()($'()( − ∑19: -8./0001/000∗34567/000 ∆!"#$ = ! ∑ ; ∗? 8
161 162
"
19:
1/000
Page 8 of 26
Eq. 3
Where
163
∆!"#$ denotes the area-weighted change from 2000-2010 in the variable i mean for urban
164
area j
165
!"#'(((@'((( denotes the year 2000 variable i mean for the area covered by the year 2000
166
footprint of urban area k
167
Areak2000 denotes total area of year 2000 footprint for urban area k
168
And n is the number of year 2000 urban areas covered by the year 2010 urban area j
169 170
Statistical analysis First, we tested for differences between 2000 and 2010 levels of independent variables
171
using multiple paired t-tests with Bonferroni-adjusted confidence intervals. Second, linear
172
regression models were developed for all univariate combinations of dependent variables
173
(change in NO2 and PM2.5 air pollution) and independent variables (urban expansion and change
174
in urban typology) stratified by small, medium, and large urban areas. Third, multiple,
175
multivariate linear regression models were created with model structures consisting of 1) all of
176
the independent variables, or 2) a subset of variables selected via bidirectional stepwise
177
regression. Adjustment factors in the multivariate models include baseline (year 2000) urban
178
characteristics, annual mean rainfall and temperature, and distance to coast.
179
RESULTS
180
Locations of the 830 urban areas in East Asia examined in our study are illustrated in
181
Figure 1, along with the surface estimates for changes in NO2 and PM2.5 concentrations and an
182
example of the urban expansion footprint for Beijing, China. Table 1 summarizes the air
183
pollution concentrations, urban expansion levels, changes in urban characteristics (both in terms 8 ACS Paragon Plus Environment
Page 9 of 26
Environmental Science & Technology
184
of average measures and within-area variability) and baseline (year 2000) characteristics by
185
urban size category. On average the change in NO2 and PM2.5 levels ranged from 1.21 to 2.25
186
ppb and 10.54 to 11.02 µg/m3 based on urban size. The size of urban expansion, in terms of
187
geographic area and population, was large for all urban size categories. On average, the area of
188
urban expansion for small, medium and large cities was 6.9, 49.3, and 379 km2, while the
189
population size expansion was 23,470, 342,400, and 2,178,000, respectively. Generally, urban
190
expansion was associated with decreasing average compactness index and increasing population
191
density, NDVI and lights at night. In small and medium cities, urban expansion was associated
192
with decreases in the percent impervious surface (due to sprawl) but in large cities were
193
associated with an increase. Change in variable heterogeneity showed similar patterns across
194
urban size categories with the exception of lights at night that showed a decrease in
195
heterogeneity with increasing urban size.
196 197 198
Table 1. Descriptive statistics of change in NO2 and PM2.5, urban expansion and urban characteristic variables from 2000-2010 stratified by urban population. 100,000-250,000 (n=235) Mean Std dev
Change in Air Pollution NO2 (ppb) 1.21 1.75*** PM2.5 (µg/m3) 10.54 10.19*** Urban Expansion Area (sq km) 6.93 7.67*** Population (1,000s) 23.47 32.29 Change in Urban Characteristics Compactness Index (0-1) -1.72x10-2 4.65x10-2*** 2 2 Population Density (pop/km ) 7.72x10 1.12x103*** NDVI (-1-1) 0.35 0.10*** Impervious Surface Area (%) -0.10 10.96*** Lights at Night (%) 12.09 7.83*** Change in Urban Characteristic Heterogeneity Population Density 5.21x102 1.03x103*** -2 NDVI (-1-1) 8.24x10 2.83x102*** Impervious Surface Area (%) -15.96 8.92*** Lights at Night (%) 0.69 2.84** Baseline (2000) Urban Characteristics NO2 (ppb) 1.58 1.27
Urban Size 250,000-1,000,000 (n=340) Mean Std dev
> 1,000,000 (n=255) Mean Std dev
2.07 12.53
2.15*** 10.91***
2.25 11.02
3.03*** 12.42***
49.27 3.42x102
63.86*** 2.25x102***
3.79x102 2.18x103
8.72x102*** 1.96x104***
-2.87x10-2 2.76x103 0.33 -0.13 10.62
8.43x10-2*** 2.94x103*** 0.11*** 11.10*** 6.26***
-2.35x10-2 4.56x104 0.31 10.16 8.79
9.09x10-2*** 2.93x105 0.12*** 11.15*** 7.11***
3.31x103 8.32x102 -16.82 -1.47
2.76x103*** 3.57x10-2*** 7.92*** 3.34
6.33x103 9.05x10-2 -13.43 -3.20
5.13x103*** 3.77x10-2*** 7.13*** 3.81***
1.77
1.28
2.51
1.27
9 ACS Paragon Plus Environment
Environmental Science & Technology
PM2.5 (µg/m3) Area (sq km) Population (1000s) NDVI (-1-1) Impervious Surface Area (%) Lights at Night (%)
28.05 53.54 52.18 1.41x10-2 24.46 35.49
19.15 43.90 73.05 3.06x10-3 9.35 11.00
31.42 1.14x102 1.41x102 1.33x10-2 28.27 37.75
21.07 93.51 2.17x102 2.88x10-3 11.21 11.31
Page 10 of 26
23.86 5.50x102 1.68x103 1.28x10-2 33.14 11.16
17.81 43.90 7.31 3.06x10-3 9.35 11.00
199 200 201 202 203
* p-value < 0.05, **p-value < 0.01, *** p-value < 0.001. Differences between 2000 and 2010 levels were tested for significance using multiple paired t-tests with Bonferroni-adjusted confidence intervals.
204
changes in urban characteristics are shown in Table 2. For all urban areas the geographic size of
205
expansion was associated with increased NO2, but only in small urban areas was expansion size
206
associated with increased PM2.5. Population growth in small and medium urban areas was
207
associated with increases in both NO2 and PM2.5, while no statistically significant association
208
was seen for large cities. A number of significant associations were observed between changes in
209
air pollution concentrations and urban characteristics that correspond to hypothesized effect
210
directions. For example, NO2 increased with increases in population density, % impervious
211
surface, and lights at night, while NO2 decreased with increasing compactness and NDVI.
212 213 214
Table 2. Univariate associations between NO2 and PM2.5 change, urban expansion, and change in urban characteristics.
Univariate associations between changes in NO2 and PM2.5, urban expansion, and
100,000-250,000 Coefficient Std Error NO2 Urban Expansion Area (sq km) 3.82x10-2 9.56x10-3*** -2 Population (1000s) 2.02x10 2.14x10-3* Change in Urban Characteristics Compactness Index (0-1) -2.86 1.60 Population Density 4.08x10-4 6.42x10-5*** NDVI (-1-1) -3.97 0.71*** Impervious Surface Area (%) 2.09x10-2 6.73x10-3** -2 Lights at Night (%) 9.76x10 8.54x10-3*** Change in Urban Characteristic Heterogeneity Population Density 2.64x10-4 7.12x10-5*** -3 NDVI (-1-1) -1.21x10 2.58x10-4*** -2 Impervious Surface Area (%) 3.21x10 8.22x10-3*** -2 Lights at Night (%) -6.25x10 2.61x10-2*
Urban Size 250,000-1,000,000 Coefficient Std Error
> 1,000,000 Coefficient Std Error
9.32x10-3 2.78x10-3
2.37x10-3*** 6.70x10-4*
1.24 x10-3 -2.50x10-7
3.45x10-4*** 1.62x10-7
-4.58 -1.06x10-5 -3.47 2.85x10-2 8.00x10-2
1.84* 5.37x10-5 1.38* 1.41x10-2* 2.45x10-2**
-1.26 -1.91x10-6 -2.74 4.81x10-2 0. 24
3.55 1.08x10-6 2.68 2.85x10-2 3.79x10-2***
2.35x10-4 -1.46x10-3 6.60x10-2 -0.13
5.45x10-5*** 4.29x10-4*** 1.93x10-2*** 4.62x10-2**
2.48x10-4 -2.47x10-3 4.77x10-2 -0.3.1
5.71x10-5*** 8.15x10-4** 4.50x10-2 7.80x10-2***
10 ACS Paragon Plus Environment
Page 11 of 26
Environmental Science & Technology
PM2.5 Urban Expansion Area (sq km) 0.12 5.64x10-2* Population (1000s) 0.10 1.27x10-2*** Change in Urban Characteristics Compactness Index (0-1) -23.35 9.29* Population Density 3.24x10-3 3.63x10-4*** NDVI (-1-1) -1.25x10-3 4.20x10-4** -2 Impervious Surface Area (%) 2.91x10 3.96x10-2 Lights at Night (%) 0.421 5.25x10-2*** Change in Urban Characteristic Heterogeneity Population Density 1.624x10-3 4.15x10-4*** NDVI (-1-1) Impervious Surface Area (%) Lights at Night (%)
-6.10x10-3 4.61x10-2 0.20
1.52x10-3*** 4.86x10-2 0.15
2.28x10-2 1.53x10-2
1.24x10-2 3.38x10-2***
9.85x10-4 -7.00x10-6
1.51x10-3 6.66x10-6
-8.44 3.53x10-4 -1.43x10-3 7.11x10-2 0.56
9.50 2.72x10-4 7.05x10-4* 7.201x10-2 0.12***
9.78 5.90x10-6 -8.01x10-4 0.13 0.83
14.45 4.44x10-6 1.09x10-3 0.12 0.17***
1.53x10-3
2.68x10-4***
9.77x10-4
2.35x10-4***
-6.65x10-3 0.17 0.52
2.20x10-3** 0.10 0.24*
-1.01x10-2 0.20 -0.78
3.32x10-3** 0.18 0.34*
215
* p-value < 0.05, **p-value < 0.01, *** p-value < 0.001.
216
Stepwise multivariate models of change in NO2 and PM2.5 are shown in Tables 3 and 4,
217
respectively. Saturated multivariate models are shown in Tables S3 and S4. The reduced
218
stepwise model variables included urban expansion, change in urban characteristics (mean and
219
variability measures), baseline (2000) urban characteristics, and other adjustment factors
220
(distance to coast and meteorological variables). For change in NO2, the overall model explained
221
37%, 49% and 54% for small, medium and large urban areas. The largest predictors were lights
222
at night (R2=0.20) and annual rainfall (R2=0.16) for small urban areas; lights at night (R2=0.13)
223
and the annual rainfall (R2=0.13) for medium urban areas; and heterogeneity in population
224
density (R2=0.12) and annual rainfall (R2=0.12) for large urban areas.
225
Models of PM2.5 change generally had lower performance than NO2 models. The overall
226
models explained 29%, 34% and 37% of change in PM2.5 for small, medium and large urban
227
areas. The largest predictors were annual rainfall (R2=0.10) and baseline urban area (R2=0.07)
228
for small urban areas; lights at night (R2=0.10) and baseline impervious surface area (R2=0.05)
229
for medium urban areas; and lights at night (R2=0.08) and population density heterogeneity
230
(R2=0.11) for large urban areas. The change in urban characteristics and urban characteristic
11 ACS Paragon Plus Environment
Environmental Science & Technology
231
heterogeneity explained 14.7-28.8% and 4.0-20.5% of the decadal change in NO2 and PM2.5,
232
respectively, compared to 3-13% and 4-14% explained by baseline urban characteristics.
Page 12 of 26
12 ACS Paragon Plus Environment
Page 13 of 26
233
Environmental Science & Technology
Table 3. Multivariate associations between NO2 change and urban expansion and urban characteristics – reduced model. Urban Size 250,000-1,000,000 (n=340)
100,000-250,000 (n=235) Coefficient
Std Error
3.84x10-2
1.23x10-2**
R2
> 1,000,000 (n=255)
Coefficient
Std Error
R2
Coefficient
Std Error
0.02
1.38x10-2
4.72x10-3**