Upload
dinhthien
View
215
Download
0
Embed Size (px)
Citation preview
1
Air Quality Interventions and Spatial Dynamics of Air Pollution in Delhi and its 1 Neighboring Areas 2
3 Naresh Kumar⊗ and Andrew D. Fosterα 4
5 Abstract: In recent years, soaring air pollution levels and their adverse health effects 6 have begun to trigger air pollution regulations in many rapidly expanding cities in 7 developing countries. Enforcements of compressed natural gas (CNG) in Delhi (India) 8 and Lahore (Pakistan) are two examples of such regulations. Using the air pollution data 9 monitored at 113 sites between July and December 2003 this article examines the spatial 10 patterns of air pollution in Delhi and its surroundings with references to the 11 recommended air quality standards, and evaluates the impact of air quality regulations on 12 the spatial patterns of air pollution in Delhi vis-à-vis its surroundings. From the analysis 13 of data three important findings emerge. First, even with the regulations in place, air 14 pollution levels in Delhi and in its surroundings were significantly higher than the 15 standards set by the World Health Organization (WHO), Central Pollution Control Board 16 (CPCB) and Environmental Protection Agency (EPA); the peripheral areas of the city 17 (inside and outside its border) have witnessed two to three times higher levels of air 18 pollution than the EPA standards. Second, air quality regulations in Delhi have resulted 19 in air pollution redistribution and adversely affected the air quality of areas surrounding 20 Delhi, because these areas were not subject to these regulations. Third, industries and 21 trucks continue to the most important sources of fine and coarse particles in the study 22 area. 23 24 Keywords: Air pollution, environmental interventions, air quality, Delhi. 25 26
⊗ Assistant Professor, Department of Geography, University of Iowa, Iowa City, IA 52242, USA & Adjunct Assistant Professor, PSTC, Brown University, Providence, RI 20912. Email: [email protected] or [email protected] α Professor and Head, Department of Economics, Brown University, Providence, RI 20912. Email: [email protected]
2
1. INTRODUCTION: The soaring levels of air pollution that high-income countries 1
witnessed in the 1950s and 1960s have begun to threaten public health in the rapidly 2
growing middle-income countries, including India and China, in recent years. As result, 3
these countries have begun to enforce environmental regulations, such as enforcement of 4
the compressed natural gas (CNG) in Delhi (India) and Lahore (Pakistan) by their 5
respective Supreme Courts are two examples of these regulations, which have drawn the 6
attention of media and researchers worldwide, and there is an increasing interest in the 7
understanding of the impact of these regulations on the time-space dynamics of air 8
pollution and its effect, in turn, on human health. Ideally, time-series data are required to 9
assess the effects of these interventions, but the limited spatial-temporal coverage of air 10
pollution restricts our ability to study the effect of these regulations. Nevertheless, a 11
cross-sectional approach can be adopted in which air quality of an area with the 12
regulations is contrasted with an adjacent area (as a control) without such regulations. 13
Building on this approach, the spatial distribution of air pollution in Delhi (after the 14
regulations) and its surroundings can help us answer three important questions – (a) is the 15
post-regulation state of air quality adequate to protect human health, (b) does air pollution 16
vary geographically within Delhi and its surroundings, and (c) has the air quality 17
regulations in Delhi adversely affected the air quality of areas in its surroundings in the 18
absence of such regulations. Although there are various measures of air pollution, we rely 19
on coarse ≤10μm (PM10) and fine ≤2.5μm (PM2.5) particles, because these are accepted as 20
standard measures of air quality worldwide (WHO, 2000). The use of the term air 21
quality/air pollution in the remaining parts of this paper will be referring to the mass of 22
PM2.5 and PM10 in the ambient environment. 23
24 The remainder of this article is organized into four sections. The first section presents a 25
background of the proposed work, followed by a description on data and methodology in 26
the second section. The fourth section examines the spatial distribution of air pollution in 27
Delhi and its surrounding, and examine the effect of air quality regulation using three 28
different approaches – (a) comparison of air pollution in Delhi with the other mega cities 29
from 2001-2005, (b) air pollution levels in- and out-side Delhi border during the year 30
2003 and (c) air pollution with reference to different sources (of air pollution). The final 31
3
section presents a discussion, concludes the main findings and draws our attention to 1
future research direction in this field. 2
3 2. Background: Rising smoke from chimneys, once used to be a symbol of prosperity, 4
began to pose serious health concerns towards the middle of 20th century. Alarming 5
health effects of air pollution episodes in 1950s, including that of smog in Donora, 6
Pennsylvania in 1948, in Poza Rica, Mexico in 1950 and famous London smog in 1952, 7
not only drew public attention worldwide but also triggered the enforcement of air quality 8
regulations in western countries, such as the UK Clean Air Act in 1956 and the Air 9
Pollution Control Act in US in 1955, which was replaced by the Clean Air Act in 1963. 10
11 Throughout the second half of the 20th century, air pollution and its health effects have 12
been subject to intensive research investigation in western countries. A substantial body 13
of literature has documented the adverse health effects of different types of air pollutants. 14
Over a decade ago, the major focus of research in this field was on the association 15
between aggregate estimates of respirable ambient particulate ≤10µm (PM10) and 16
mortality (Pope and Dockery, 2006, Pope et al., 1995, Schwartz et al., 1995). After 17
controlling for seasonality and other confounders, recent literature reiterates this 18
association between mortality/morbidity and air pollution with the special emphasis on 19
the relevance of fine particulates (PM2.5) in the research on health effects of exposure to 20
ambient air pollution (Dominici et al., 2003, Samet et al., 2000, 21
National Research Council, 2001). A recent review of literature on this topic is available 22
in Davidson et al. (2005). While the health effects of air pollution are examined at length 23
in high-income countries, limited air pollution and health data constrain researchers’ 24
ability to pursue research in the field in low-income countries. 25
26 The main focus of this article is study the spatial patterns of air pollution and the effect of 27
air quality regulations in Delhi on air pollution redistribution in Delhi and its 28
surroundings. The sequential details on these regulations and associated environmental 29
laws are discussed in detail in (Bell et al., 2004). Review of literature suggests that there 30
is limited research available on air pollution distribution in Delhi. Research reported by 31
Kathuria (2002) and Khillare et al. (2004) are particularly relevant to this study. Using 32
4
the data collected by high volume samplers between July 1997 and June 1998 at four 1
locations, (2004) examined spatial and temporal variations in the concentration of 2
suspended particulates matter (SPM) and heavy metals, namely Pb, Cd, Cr, Ni, and Fe, in 3
the atmospheric aerosols in Delhi. Their main findings revealed that the SPM 4
concentration in Delhi was three times higher than the national standards (140µg/m3). 5
Their analysis further showed that (a) the concentration of heavy metals and SPM were 6
land-use specific, for example they found elevated levels of SMP and heavy metals in 7
industrial area, and (b) the main sources of SPM and heavy metals (in ambient aerosol) 8
were emission from automobiles and industries. While the research reported by Khillare 9
et al. (2004) provides valuable insight into air pollution estimates in pre air quality 10
regulation periods, it is difficult to generalize spatial patterns of air pollution for the 11
entire city using the data at monitored at four locations only. 12
13 Kathuria (2002) examined the effects of recent regulations related to vehicular pollution 14
on air quality in Delhi. Based on air pollution data from the central pollution control 15
board (CPCB) before and after the interventions, he concluded that recent interventions 16
have led to little improvement in air quality in the City. This research conveys important 17
messages regarding soaring levels of air pollution in Delhi as compared to WHO 18
standards (WHO, 2000), and required measures to check air pollution from vehicle 19
emission at three different stages – (a) pre-combustion stage, i.e. improvement in the 20
quality of fuel, (b) combustion stage, which refers to engine efficiency, and (c) post-21
combustion stage, i.e. exhaust treatment, e.g. use of catalytic converters. However, the 22
findings regarding the effect of CNG interventions on air quality can be questioned for 23
two important reasons. First, the conversion of commercial vehicles to CNG was 24
completed by the end of year 2002, but the data used in the analysis was until 2001, 25
which fails to capture the full effect of CNG regulations. Second, the analysis is based on 26
air pollution data collected by the CPCB. These data, however, do not capture air 27
pollution scenario for the entire city, because there are only 9 monitoring sites, of which 28
only one near income tax department (ITO) is operational continuously. 29
30 A review of literature suggests that few studies are available on the spatial analysis of air 31
pollution in Delhi. The proposed research will augment a number of literatures. First, it is 32
5
the first systematic attempt to analyze spatially detailed patterns of air quality, measured 1
by fine and coarser particulate matters (PM2.5 and PM10, respectively) in Delhi and its 2
surroundings. Second, this research examines the impact of emission from different 3
sources on air pollution at a given location. Third, this article sheds light on how air 4
quality regulations can guide the course of air pollution redistribution in a city (with the 5
air quality regulations in place) and in its surroundings, largely unaffected by such 6
regulations. Finally, this research identifies the significance of Environmental Kuznets 7
Curve (EKC) in the context of a developing country. The proposed research will answer 8
the following research questions: 9
10 a. What is the spatial distribution of air quality in Delhi and its surroundings, and is 11
the current state of air quality with reference to the recommended air quality 12
standards adequate to protect human health? 13
b. What is the impact of air quality regulations on the spatial distribution of air 14
pollution in the study area, and have the CNG regulations done their job? 15
c. What are the main sources of fine and coarse particles in the study area? 16
17 3. STUDY AREA, DATA AND METHODS: 18 19 3.1 Delhi and its surroundings: This study focuses on air pollution distribution Delhi 20
and its surroundings. Delhi and its 10 neighboring districts of three other states form the 21
National Capital Region (NCR), which was enacted to divert the burden of Delhi’s 22
growth in the surrounding areas (Table 1, Figure 1). Delhi is the second largest 23
metropolitan in India, and its population has increased from 9.4 million in 1991 to 13.2 24
million in 2001 (at the rate of 3.34% annual exponential growth rate). The number of 25
industrial units in Delhi increased from 8,000 in 1951 to 125,000 in 1991, and number of 26
automobiles increased from 235,000 in 1975 to 4,236,675 in 2003-04 27
(Government of India, 2006). Consequently, Delhi has been experiencing a phenomenal 28
increase in air-pollution levels. According to WHO estimates, Delhi was reported to be 29
the tenth most polluted city in the world in terms of suspended particulate matter (NCT 30
Fact Sheet Delhi 1999) that posed a serious threat to human health. In part as a result of 31
6
this concern over air quality in Delhi, there is a need to collect good air pollution 1
longitudinal data for the management and surveillance of air quality in the city. 2
3 Delhi is an instructive location for examining the relationship between air pollution and 4
health in context to recently implemented air quality regulations as a consequence of a 5
series of rulings by the Indian Supreme Court. Of particular importance was the 6
September 2000 ruling directed at so-called “non-conforming areas”. These are areas in 7
which most industrial activity is supposed to be excluded, but in which industrial growth 8
has continued unabated. The potential impact of these rulings is very large. According to 9
a survey of the Delhi government 98,000 of the total of 125,000 industries were in non-10
conforming areas. Subsequent to the decision 38,000 establishments in sectors with 11
particularly high levels of pollution were subject to this ruling. A large number of 12
industrial plots on the periphery of the city were developed to house roughly 24,000 13
establishments. While this ruling has met with substantial resistance and change has 14
proceeded slowly, the ruling remains in force and concerted attempts are being made to 15
relocate industries (TERI, 2001b). It is also anticipated that new investments will have 16
been influenced by these rulings regardless of the eventual extent of their enforcement. 17
An important feature of this ruling is that it means that similarly polluting industries in 18
different parts of the city are differentially affected by the law. 19
20 The second Supreme Court ruling mandated conversion of buses and three-wheel 21
scooters from gasoline/diesel to compressed natural gas (CNG) in 2000 and 2001. It took 22
the state about two years to implement this ruling, and the final bus registered in Delhi 23
was converted to CNG in December 2002. The conversion of a gasoline engine to CNG 24
costs roughly five times the cost of a new gasoline operated vehicle; thus after CNG 25
regulation, most gasoline operated commercial vehicles were migrated to the neighboring 26
districts (in NCR) that were unaffected by CNG regulation. Moreover, it is anticipated 27
that different parts of the region will have been differentially affected by these initiatives 28
given differences in level of traffic through these regions. There is some question about 29
the extent to which these changes have improved air quality as yet, given the phase in 30
time for some components of the new regulations, but a carefully constructed model 31
suggested that total declines in vehicular emissions are likely to become evident starting 32
7
in 2003 (TERI, 2001a). Though the CNG and industrial zoning regulations are expected 1
to reduce the pollution level in Delhi, we expect that the air pollution has been 2
importantly redistributed since most polluting vehicles and industries that were subject to 3
these regulations moved to neighboring states. Spatially detailed air pollution data and a 4
cross-sectional comparison of air pollution in Delhi vis-à-vis outside Delhi border can 5
provide insight into the effect of these regulations on air pollution redistribution. 6
7 3.2 Data: The air pollution data for this research come from two different sources – (a) 8
Central Pollution Control Board (CPCB) and (b) field campaign from July-December 9
2003. 10
11 Air pollution data from CPCB: Air pollution data monitored at a central location for 12
seven mega cities were acquired from CPCB to contrast the trend of air pollution in Delhi 13
with other megacities from 2000 to 2005. Data on three pollutants – SO2, NO2 and 14
particles ≤ 10μm in aerodynamic diameter (PM10) – were collected. PM10 data are 15
collected in three different shifts 6:00 to 14:00, 14:00 to 22:00 and 22:00 to 6:00 (of the 16
next day) using high volume samplers. CPCB facilities, however, are not equipped to 17
monitor PM2.5. There is only one monitoring station in each city where data on these 18
three pollutants are monitored continuously. Therefore, readers should be cautioned about 19
these data, because these may not truly represent the state of air quality of the entire city, 20
though these data may provide insight into the change in air pollution with respect to 21
policy interventions. 22
23 Air sampling field campaign: Given the limited spatial coverage and non-availability of 24
data on fine particulate matter (PM2.5), the existing air pollution data collected by CPCB 25
were inadequate for evaluating the spatial patterns of air pollution. Therefore, a field 26
campaign was conducted from July 23 to December 3, 2003 to monitor ambient particles 27
at 113 sites in Delhi and its surroundings. Since one of the major goals of the campaign 28
was to evaluate spatial variability in the mass of airborne particulate matter (PM), a 29
spatially dispersed sampling design was adopted, in which sample sites were identified 30
using a two-step process. In the first, a rectangular grid was overlaid onto the entire study 31
area, to ensure full coverage of the area. In the second step, a random location was 32
8
simulated within each cell (of size 1x1.5km), and then the simulated locations were 1
transferred to a Garmin Global Positioning System (GPS) in order to navigate them and 2
examine their suitability. Some sites, which were inaccessible, were discarded and re-3
simulated, resulting in a final sample of 113 suitable sites (Figure 2a and 2b). At each site 4
air was sampled at two different times every third day. Each sample involved four 5
readings – two each in mass and count modes. In the mass mode, each reading sampled 6
air for two minutes and in the count mode for one minute. Therefore, each sample 7
involved six minutes of sampling. Although air was sampled at different times between 8
7:30AM and 10:00PM, the PM data for the present analysis was restricted ±150 minutes 9
of the time interval of AOD data, generally 10:30AM, in order to minimize the temporal 10
noise in PM Data. 11
12 The Aerocet 531, a photometric sampler, from Met One Instruments, Inc., was used to 13
collect air pollution data (Met One Inc, 2003). It is an automatic instrument that can 14
monitor particulate mass (PM) in a range of ≤1, ≤2, ≤5, ≤7 and ≤10µm in aerodynamic 15
diameters in mass mode, and PM ≤ 0.5 and PM ≤ 10µm in count mode. The instrument 16
uses laser technology and uses a right angle scattering method at 0.78μm, which is 17
different from gravimetric measurements. The source light travels at a right angle to the 18
collection system and detector, and the instrument uses the information from the scattered 19
particles to calculate a mass per unit volume. A mean particle diameter is calculated for 20
each of the 5 different sizes. This mean particle diameter is used to calculate a volume 21
(cubic meters), which is then multiplied by the number of particles and then a generic 22
density (µgm-3) that is a conglomeration of typical aerosols. The resulting mass is divided 23
by the volume of air sampled for a mass per unit volume measurement (µg/m3). 24
25 This instrument also recorded relative humidity and temperature with every sample. The 26
main flaw of the instrument is that the mass values can be easily inflated with the 27
increase in relative humidity (Thomas and Gebhart, 1994). A standard relationship 28
between photometric and gravimetric measurements as discussed by Ramachandran et al. 29
(2003) was used to calibrate the data for relative humidity: 30
31 D/D0 = 1 + (0.25 (RH2/(1-RH)) (1) 32
9
1 Where D and D0 are wet (photometric estimates) and dry particles (gravimetric 2
measurement), respectively; RH is relative humidity (proportion). Sioutas and others 3
(2000) suggest correction for relative humidity using different particle characteristics 4
including molecular weight of dry particles. These data were not available for the study 5
area. Thus, we relied on the above notation to estimate dry mass of PM2.5, PM10 and TSP, 6
though the analysis was performed on both original and calibrated dataset. 7
8 Data on the source of air pollution: This article focuses on two main sources of air 9
pollutants, namely traffic and industries. Data for traffic were collected as a part of air 10
pollution data. Number of vehicles were counted during each sample and then classified 11
by type – buses, trucks, cars, two wheeler and others. Data on industrial clusters were 12
generated from the Eicher Map of Delhi (EICHER, 2001), which is largely based on 13
Survey of India’s large scale topographic maps. 14
15 3.3 Methods: Out analysis is based on three different methods – (a) proximity analysis 16
for data integration, (b) spatial interpolation of air pollution surfaces and (c) regression 17
modeling. These methods are discussed below: 18
19 3.3.1 Data Processing: As mentioned earlier data were collected at each site at two 20
different times of every third day. On an average we have more than 65 samples at each 21
site, which represent sampling at different times of a day and different days of a week. 22
Each sample included two readings (four minutes of sampling). The data used in the 23
analysis are the averages over July 23 through December 3, 2003. The frequency of 24
vehicles is the average number of vehicles (by their types) every one minute for the same 25
period. Proximity to road and industrial cluster were computed using spatial join in 26
ArcGIS Ver 9.x (ESRI, 2005), which computed straight line distance of all sample sites 27
to the closest sources (roads and industrial clusters in our case). 28
29 3.3.2 Spatial Interpolation: Various methods of interpolation are available to interpolate 30
continuous surface. We employed Kriging, which estimates air pollution at given pixel as 31
an inverse function of distance weighted by spatial autocorrelation among the sample 32
sites (Cressie, 1990, Isaaks and Srivastava, 1989), to interpolate surfaces of PM2.5, PM10 33
10
and TSP. Kriging uses a semivariogram, developed from the spatial structure in the data, 1
to determine the weight. Other advantage of Kriging is that it yields a set of spatial 2
predictions at sampled locations and also provides an associated variance that measures 3
the uncertainty in the predictions. The optimal parameters, such as distance range, 4
distance exponent, were computed by minimizing variance between actual and estimated 5
values at the sample sites (Table 2): 6
7
∑ −=
=n
izz iin 1
22 )ˆ(1||min σ (1) 8
9 where n = number of sample sites 10
z i = estimated value at ith site 11
zi = observed value at ith site 12
13 3.3.3 Air pollution and its sources – regression model: Two different sets of regression 14
models were employed to examine the contribution of different sources (of air pollution) 15
on the levels of air pollution observed at the sample sites. To begin with air pollution at 16
sample site τi was modeled as a function of the selected covariates Xi, which included 17
proximity to the major roads, industrial clusters, frequency of buses and trucks per 18
minute, using an ordinary least square regression (OLS) model: 19
20 τi = α + Xi' β + εi (2) 21
22 Where β is a vector of regression coefficients, and 23
εi = unobservable 24
25 One of the main assumptions of OLS model is that εi ~ N(0,σ2). Since the error term 26
observed spatial autocorrelation, we used spatial autoregressive model. Three different 27
models, namely conditional autoregressive (CAR), simultaneous autoregressive (SAR) 28
and spatial autoregressive models (SAM), are suggested to account for spatial structure in 29
the data. In the first, the spatial dependence in the residuals is expected to have a 30
conditional distribution and joint distribution in the second. In the third one (SAM) we 31
could represent the residuals that are not explained in the OLS with a variable (e.g., iR ), 32
we could treat this new variable as a predictor variable in the appropriate model. This 33
11
way, we would no longer need to assume dependency for the outcome variable. One of 1
the advantages of this approach is that it is easily understandable. Another advantage is 2
that we could avoid the problem of counterintuitive results in SAR and CAR as 3
demonstrated by (Wall, 2004). 4
5
More formally, assuming that the response variable is normally distributed, we could 6
define the model as proposed above as following: 7
8
τ(si) = α + X(si)' β + θRi + εi 9
10
Where iR = information not explained in the OLS 11
θ = parameter coefficients 12
ε'i ~ N(0,σ2) iid 13
14 There are various ways to estimate Ri. Institutively, Ri can be estimated as an inverse 15
distance weighted average of residuals at the neighboring sites, as 16
∑∑ =
−
=
−
=k
jijjk
jij
i drd
R1
1
1 ω
ω 17
Where rj = residual of jth neighboring site 18
k = number of neighboring sites 19
dij = distance between ith site and jth neighboring site, and dij < h 20
h = distance range 21
ω = distance exponent 22
23 The distance range (h) and distance exponent (ω) can be estimated with the aid of 24
empirical semivariogram. The distance range refers to the distance threshold where the 25
semivariogram levels off to nearly a constant value, called as the sill, and then shape of 26
semivariogram can help us determine the distance exponent. For PM2.5 and PM10, 27
distance ranges were 4.5 and 3.0km, respectively, and distance exponent for both was -2. 28
29 4. RESULTS 30
12
1 4.1 Data validation – photometric and gravimetric measurements: Spatially detailed 2
PM data were not available for the study area. Therefore, we conducted a field campaign 3
to collect these data using photometric samplers. Conventionally, however, high volume 4
samplers are employed to monitor ambient PM, which requires a minimum of eight hours 5
of sampling. Filters are weighed before and after the sampling, and based on the mass 6
gained during the sampling period and the amount of air sampled, the mass of PM is 7
computed. Given the cost of gravimetric samplers, it was not plausible to deploy these 8
samplers at a large number of locations. Therefore, we monitored air pollution with the 9
aid of photometric samplers at relatively large number of locations. To assess the 10
accuracy of photometric samplers, first, we compare data from these samplers with the 11
CPCB data. 12
13 During August through November 2003, the daily averages of PM10 from CPCB site and 14
our instrument were 203±26.8μgm-3 and 153±33.5μgm-3, respectively. The photometric 15
estimates were significantly lower than the gravimetric estimates. Given the differences 16
in the method of operation and duration of sampling by gravimetric (24 hour average) 17
and real time photometric measurements (six minutes for each sample), a difference of 18
49.5±31μgm-3 (95% CI) seem reasonable. In addition, the regression analysis suggests a 19
statistically significant positive association in the temporal variability in PM10 measured 20
by both methods (Figure 3). It was not possible to validate PM2.5 because of the non-21
availability of PM2.5 data. Based on other research, the difference between photometric 22
and gravimetric estimates is likely to be smaller for PM2.5, and photometric estimates can 23
be calibrated to gravimetric standards by adjusted for relative humidity as suggested by 24
Ramachandran and others (2003). 25
26 4.2 Descriptive analysis: The concentration of both fine and coarse particles in the study 27
area is much higher than the standards set by the Central Pollution Control Board 28
(CPCB), the Environmental Protection Agency (EPA) of the United States and the World 29
Health Organization (WHO). According to EPA, three year average of PM2.5 must be less 30
than 15µg/m3, while the average mass of PM2.5, PM10 and TSP (five months averages) 31
was recorded as 28.2±1.8µgm-3 (95% CI), 157±18.1µgm-3 and 189±21.8µgm-3, 32
13
respectively (Table 3a). These values are significantly higher than the air quality 1
standards recommended by the World Health Organization, the Environmental Protection 2
Agency and the Central Pollution Control Board. The summary statistics of different 3
sources of air pollutants is presented in table 3b. Among automobiles, cars and trucks 4
recorded the highest (18.9±3.5) and lowest frequency (1.8±0.2), respectively; the average 5
distance to the industrial cluster was 2.4±0.37km. 6
7 As mentioned above, 16 of the 113 sites were located outside Delhi but adjacent to its 8
border. It is interesting to note that the concentration of ambient particles in the area 9
outside Delhi (but closer to Delhi border) is significantly higher than that in Delhi; the 10
average mass of PM2.5, PM10 and TSP outside Delhi were 33.4±6.9µgm-3, 11
213.1±83.6µgm-3 and 250.3±98.9µgm-3 as against 27.3±1.7µgm-3, 147±14.9µgm-3 and 12
178.5±18.3µgm-3, respectively, in Delhi. This clearly indicates that the state of air quality 13
in the areas outside Delhi border is relative poor as compared to that in Delhi. This is 14
further discussed in the section 4.4.2. 15
16 4.3 Spatial pattern of air pollution: Figure 4, 5 and 6 show the spatial distribution of 17
PM2.5, PM10 and TSP in the study area. The trend of spatial distribution is similar in all 18
three maps. In addition, two striking observations emerge from these maps. First, the 19
central part of Delhi experience relatively lower concentration of air pollution as 20
compared to that observed in the peripheral areas, albeit the minimum concentration of 21
ambient particles in- and out-side Delhi is much higher than WHO, CPCB and EPA 22
standards. The absence of air quality regulations and migration of pollution industries and 23
vehicles that were banned in Delhi can be responsible for the elevated concentration of 24
air pollution in areas outside Delhi border. Second, among industrial areas, Ashok Vihar, 25
Sahibabad, Okhla and Industrial areas surrounding Delhi-Gurgaon border (around the 26
intersection of National Highway 8 and Delhi border) witness high concentration of air 27
pollution, which indicates that industries continue to be an important source of air 28
pollution in the study area. 29
30 4.4 Air Quality Regulations and Air pollution: In the absence of spatially detailed time 31
series air pollution data, we rely on three alternate approaches to assess the effect of air 32
14
quality regulation on time-space dynamics of air pollution in Delhi and its neighboring 1
areas. First, we compare air pollution levels in Delhi from 2001-2002 to 2004-2005 with 2
reference to other mega cities. Second, compare air pollution inside and outside Delhi 3
border, especially in the eastern and southern parts where urban growth has continued 4
irrespective of the city boundary. Third, examine air pollution with reference to pollution 5
sources, which have been subject to these regulations. 6
7 4.4.1 Air pollution in Delhi and other mega cities: Air pollution data on three air 8
pollutants, namely SO2, NO2 and PM10, from 2001 to 2005 for seven megacities were 9
acquired from the Central Pollution Control Board. These data come from one 10
monitoring station (located centrally) in each city. The average estimates of air pollutants 11
were computed for the pre-regulation (year 2001-02) and post regulation (2004-05), 12
periods; the summary of these calculations is available in the table 4. Among all cities 13
Delhi recorded the highest levels SO2, NO2 and PM10 during the pre-regulation periods, 14
which undoubtedly indicated very poor state of air quality in Delhi prior to the regulation 15
periods. Although the levels of air pollution during the post regulation period continue to 16
be high, there was not a significant decline in the concentration of PM10 in Delhi. Air 17
pollution data for Mumbai were not available for the year 2001-02, but during the year 18
2004-05 Mumbai, which is the most populous city in India, witnessed the worst type of 19
air quality in terms of the selected three air pollutants – PM10, SO2, and NO2. 20
21 A comparison of the trend of air pollution in Delhi with that in Kanpur can shed some 22
light on the effect of air quality regulation in Delhi, because Kanpur and Delhi are 23
situated in identical topographic and climatic conditions, and Kanpur is the largest city 24
closer to Delhi. The average PM10 in Delhi declined slightly from 240.2±22.7μgm-3 in 25
2001-02 to 239.8±10.9μgm-3 2004-05. In Kanpur, however, the PM10 increased from 26
178.5±12.8μgm-3 in 2001-02 to 198.3±15.3μgm-3 in 2004-05, a net increase of about 27
20.42μgm-3 over a span of just four years. This clearly indicates that while the 28
enforcement of environmental regulations in Delhi did not results in a significant 29
improvement in air quality, these regulations have been protecting air quality from 30
further deterioration. In Kanpur, however, the air quality has deteriorated significantly 31
15
over a span of four years and this deterioration can be attributed to the absence of air 1
quality regulations. 2
3 Interpretation of these results should be used with caution, because PM10 is not a robust 4
indicator of air quality, especially in semi-dry climate where dust is a major contributor 5
of PM10 mass. Data on PM2.5 mass, which results from combustion, are not available for 6
any of the cities. Therefore, using a cross-sectional approach a comparison of PM2.5 mass 7
(and other pollutant) inside and outside Delhi border can shed light on the effect of these 8
regulations on air pollution redistribution. 9
10 4.4.2 Air Pollution inside and outside Delhi border: As evident from the section 4.2, 11
areas in- and out-side Delhi recorded relatively high concentration of ambient particles. 12
In this section, air pollutant is examined in- and out-side Delhi border at different 13
distance intervals. This can allow us to evaluate the effect of recent environmental 14
regulations on air pollution redistribution. The main assumption behind this analysis, 15
however, is that the air quality of areas bordering Delhi (both inside and outside) was 16
same prior to the regulations, because urban growth has spread across Delhi’s border in 17
eastern, south eastern and southern parts (Figure 7), and functional characteristics of 18
these areas were expected to the same. Therefore, prior to the regulations the sources and 19
levels of air pollution should have been the same in these areas. Based on this 20
assumption, we hypothesize that the differences in air pollution levels inside and outside 21
short distances (<2km) of Delhi border are the result of differential impact of air quality 22
regulations. 23
24 The analysis of our data suggests that the levels of air pollution within 2km outside 25
Delhi’s border were significantly higher than the areas inside 2km of Delhi’s border, and 26
the ratio of outside to inside air pollution declines with distance from the border (Table 27
5). This proves our hypothesis and indicates the differential impact of air quality 28
regulations. While the regulations, namely conversion of buses to CNG and closer of 29
polluting industries, are likely to improve the air quality in Delhi, these seem to have 30
adversely affected the air quality of areas outside Delhi border, because these areas seem 31
to have attracted a large number of polluting industries and vehicles that were banned in 32
16
Delhi. As a result the air quality of these areas is likely to deteriorate further in the 1
absence of air quality regulations in place. 2
3 4.4.3 Sources of air pollution: The main assumption behind this analysis is that if the 4
sources of air pollution have been subject to environmental regulations, proximity to 5
these sources should not show any significant association with the levels of air pollution 6
(monitored at spatially dispersed sites). In the study area, there are four main sources of 7
air pollution, namely industries (including three thermal plants), automobiles, cooking 8
and air conditioning buildings. Among these, only diesel buses and a limited number of 9
polluting industries were banned. In order to examine the impact of ban on these 10
activities, we examine air pollution with respect to (indirect) sources of air pollution, 11
namely frequency of different types of vehicles (particularly buses and trucks) and 12
proximity to industrial clusters. Two different sets models were examined – (a) in the 13
first set, the impact of air pollution sources was examined on PM2.5 and PM10 in- and out-14
side Delhi (Table 6a and 6b) and (b) in the second set, PM2.5 and P10 were regressed on 15
the important sources of air pollution using ordinary least square regression model and 16
spatial autoregressive model (Table 7). 17
18 As hypothesized, the frequency of buses, which were converted to the natural compressed 19
gas (during 2001 to 2002) observed an insignificant association with PM2.5 and PM10, 20
particularly inside Delhi. In contrast, the frequency of trucks, which were not subject to 21
CNG regulations, emerged the most important predictor of both PM2.5 and PM10 (Figure 22
8a and 8b); it explains 21 and 12 % of the total variability in PM2.5 and PM10, 23
respectively (Table 7). Among other sources, proximity to industrial clusters emerge the 24
second most important predictor of PM2.5 and PM10 (Figure 8c and 8d) and proximity to 25
road shows a significant positive association with PM10, but its association with PM2.5 26
was insignificant, which indicates that dust is one of the major contributors of PM10. 27
Combustion from trucks and industries are two major contributors of fine particles 28
(PM2.5). These two sources together explain about one third of the total variability in 29
PM2.5 (Table 7). 30
31
17
It is interesting to note that the regression coefficient of the frequency of trucks is 1
stronger outside Delhi as compared to inside Delhi (Table 6a and 6b). The impact of 2
proximity to industrial cluster on PM2.5 and PM10, however, was stronger inside Delhi. 3
The frequency of buses did not register a significant association with fine particles, but 4
showed a negative association with PM10 inside Delhi and a positive association with 5
PM10 outside Delhi, because buses not registered in Delhi were not subject to these 6
regulations, and hence these buses in the areas outside Delhi continue to use diesel and 7
hence likely to be an important source of fine particles (PM2.5). 8
9 Our analysis clearly supports the hypothesis that the sources that were subject to 10
regulation have done their job, for example conversion of buses to compressed natural 11
gas has helped improve air quality in Delhi, but trucks and industries continue to be a 12
major source of fine particles in the study area. PM10 shows a statistically significant 13
spatial autocorrelation, because coarse particles settle by gravity as distance increases and 14
fine particles stay aloft longer distances and for longer duration. 15
16 5. DISCUSSION AND CONCLUSION 17 18
According to the Articles 39 (e), 47 and 48A of the Indian Constitution, it is the 19
responsibility of the State to secure and improve people’s health and protect the 20
environment. The air quality in Delhi continued to deteriorate throughout the 1980s and 21
1990s notwithstanding a number environmental laws were enacted during this period. 22
Because of government’s failure to discharge its constitutional responsibility and growing 23
public discontent over the unabated increase in air pollution, the Indian Supreme Court 24
stepped in and in its series of rulings directed Delhi Administration to improve air quality 25
by adopting such measures as the closure of polluting industries in non-conforming areas 26
and switching commercial vehicles, particularly buses and auto-rickshaws, from 27
conventional fuel (diesel/gasoline) to compressed natural gas (CNG). There is a decade 28
and half long history behind these regulations. A detailed discussion on the background 29
of these regulations and associated environmental laws is available in Bell et al. 30
31
18
From the air quality interventions in Delhi we can learn a number of lessons. First, these 1
regulations clearly indicate how an independent judicial branch in a democratic society 2
can enact environmental laws, generally reserved to legislators and specialized regulator 3
bodies of executive branch, and direct the executive branch to enforce these laws when 4
political will fails to do so (Bell et al., 2004). Second, until recently it was believed that 5
Environmental Kuznets Curve (EKC), which maps the course of air pollution as an 6
inverted U-shape function of economic growth, was applicable to western countries 7
(Krupitsky et al., 2005, Stern, 2004). Developing countries, however, have begun to 8
address environment pollution in recent years (Dasgupta et al., 2002). Delhi, which has 9
enforced two major environmental regulations in recent years, is an example of these 10
attempts. The similar CNG regulations are being enforced in Lahore, Pakistan. These 11
interventions are expected to reverse trend the increasing trend of air pollution in the 12
City, and its entry in the second phase of EKC, in which air pollution declines, to a 13
certain level, with the positive economic growth. Although the validity of EKC in terms 14
of econometric precision in air pollution-economic growth association is questioned 15
(Stern, 2004), the general idea behind it is still relevant for both developed and 16
developing countries. Now the question before us, particularly in the context of 17
developing countries, is to investigate whether the EKC will follow the same course of 18
air pollution (in relation to economic growth) as it did in developed countries. Although 19
this topic is beyond the scope of this research, it will useful to pursue the application of 20
EKC in Delhi for future research. 21
22 Although recent air pollution regulations in Delhi have drawn public attention worldwide, 23
most literature and media coverage has been restricted to such topics as the role of 24
politics and the Indian Supreme Court behind these interventions and the role of 25
government agencies in implementing these regulations, particularly CNG regulation and 26
closure of pollution industries in non-conforming area. Little attention was paid to 27
quantify the state of air quality before and after the regulations and spatial variability in 28
air pollution in- and outside the City, because air pollution data are not available at high 29
spatial-temporal resolutions. Therefore, we attempted to collect spatially detailed air 30
pollution data at 113 sites from July-December 2003, and examined these data to assess 31
19
the spatial patterns of air pollution with reference of the recommended air pollution 1
standards and the effect of these regulations on air pollution redistribution in- and out-2
side Delhi. 3
4 Although these regulations were expected to improve air quality in Delhi, the analysis of 5
our data does not support a significant improvement in air quality within the City. 6
Moreover, substantial variations were observed in the spatial distribution of air pollution 7
in- and out-side Delhi border. These regulations have two major impacts on air pollution 8
redistribution in- and out-side Delhi. First, closure of H-Class - hazardous, noxious, 9
heavy and large polluting – industries in non-conforming areas resulted in either their 10
relocation to newly developed industrial estates, namely Bawana Industrial Estate in 11
North and Kanjawala in North-West, regularization of industries in five non-conforming 12
areas (namely Lawrence Road, Wazirpur, Naraina Phase, Naraina Phase-2 and G.T. 13
Karnal Road) or their migration to the neighboring states, which were not subject to these 14
regulations. As result, air quality at the destination of these polluting industries (whether 15
within the regularized industrial estates in Delhi or outside Delhi) has been affected 16
adversely. Second, most non-CNG buses that were banned in Delhi migrated to 17
neighboring states of Delhi. Therefore, in the absence of stringent air quality regulations 18
the areas surrounding Delhi have served as magnet to attract polluting industries and 19
vehicles, and could explain significantly high concentration of PM2.5, PM10 and TSP. 20
Polluting industries, particularly, prefer proximity to the city center and less stringent 21
environmental regulation. A cross-sectional analysis of data clearly indicates that the 22
elevated concentration of air pollution in the areas outside Delhi could be the result of 23
differential impact of these regulations. Although the air pollution levels in most parts of 24
Delhi and its surroundings were significantly higher than the WHO, EPA and CPCB 25
standards, areas nearing Delhi border (±2km) witnessed three to four times higher than 26
the recommended air pollution levels. 27
28 From the analysis of air pollution and its association with sources of air pollution two 29
important findings emerge. First, the conversion of buses to CNG seems to have been 30
working, because the frequency of buses does not show a significant association with the 31
mass of PM2.5 or PM10, particularly in Delhi. Second, trucks and industrials are the most 32
20
important sources of air pollution, particularly that of fine particles (PM2.5), which are by 1
product of combustion. The proximity to industrial locations and frequency of trucks 2
account for one third of the variability in PM2.5. It seems that land-zoning regulations 3
were not enforced as vigorously as the CNG regulations, and trucks, plying in the city, 4
might not be registered in the Delhi and hence has not been subject to CNG regulations. 5
In addition, a large number of personal vehicles (including diesel cars) are added every 6
year (Waldman, 2005). Therefore, emission reduced by the CNG regulations could have 7
been neutralized by the addition of new diesel based cars and unchecked emission from 8
industries and trucks. 9
10 While more stringent regulations are required to check air pollution from industries, truck 11
and personal vehicles within the city, areas surrounding Delhi also need to implement 12
similar air quality regulations otherwise increasing concentration of air pollution in these 13
areas is like to have severe health effects. In addition, there is also a need for spatially 14
detailed longitudinal data for effective air quality monitoring and management, because 15
the limited sites in Delhi cannot be used to estimate spatially detailed air pollution 16
surfaces. The use of real-time photometric samplers and satellite remote sensing are two 17
substitutes for constructing spatially detailed time-series data (Kumar et al., 2007). These 18
data are critical for computing exposure in micro-environments to study the health effects 19
of air pollution. 20
21
References: 1 2 BELL, R. G., MATHUR, K., NARAIN, U. & SIMPSON, D. (2004) Clearing the Air: 3
How Delhi Broke the Logjam on Air Quality Reforms. Environment Magazine, 4 46, 22-39. 5
CRESSIE, N. (1990) The Origins of Kriging. Mathematical Geology, 2, 239-52. 6 DASGUPTA, S., LAPLANTE, B., WANG, H. & WHEELER, D. (2002) Confronting the 7
Environmental Kuznets Curve. Journal of Economic Perspectives, 16, 147-168. 8 DAVIDSON, C. I., PHALEN, R. & SOLOMON, P. (2005) Airborne Particulate Matter 9
and Human Health: A Review. Aerosol Science and Technology, 39, 737–749. 10 DOMINICI, F., MCDERMOTT, A., ZEGER, S. L. & SAMET, J. M. (2003) Airborne 11
particulate matter and mortality: Timescale effects in four US cities. American 12 Journal of Epidemiology, 157, 1055-1065. 13
EICHER (2001) Delhi: City Map, New Delhi, Eicher Goodearth Ltd. 14 ESRI (2005) ArcGIS, Version 9.1, Redlands, CA, Environmental Systems Research 15
Institute. 16 GOVERNMENT OF INDIA (2006) White paper on pollution in Delhi with an action 17
plan, New Delhi: Ministry of Environment and Forests. 18 ISAAKS, E. H. & SRIVASTAVA, R. M. (1989) An Introduction to Applied 19
Geostatistics, New York, Oxford University Press. 20 KATHURIA, V. (2002) Vehicular Pollution Control in Delhi - Need for Integrated 21
Approach. Economic and Political Weekly, 1147-55. 22 KHILLARE, P. S., BALACHANDRAN, S. & R., M. B. (2004) Spatial and Temporal 23
Variation of Heavy Metals in Atmospheric Aerosol of Delhi. Environmental 24 Monitoring and Assessment, 90, 1-21. 25
KRUPITSKY, E. M., HORTON, N. J., WILLIAMS, E. C., LIOZNOV, D., 26 KUZNETSOVA, M., ZVARTAU, E. & SAMET, J. H. (2005) Alcohol use and 27 HIV risk behaviors among HIV infected hospitalized patients in St. Petersburg, 28 Russia. Drug and Alcohol Dependence, 79, 251-256. 29
KUMAR, N., CHU, A. & FOSTER, A. (2007) An Empirical Relationship between PM2.5 30 and Aerosol Optical Depth in Delhi Metropolitan. Atmospheric Environment 31 (forthcoming). 32
MET ONE INC (2003) AEROCET 531: Operation Manual. Grants Pass: OR. 33 NATIONAL RESEARCH COUNCIL (2001) Research Priorities for Airborne Particulate 34
Matter. Part III. Early Research Progress. Washington, DC. 35 POPE, C. A., 3RD & DOCKERY, D. W. (2006) Health effects of fine particulate air 36
pollution: lines that connect. J Air Waste Manag Assoc, 56, 709-42. 37 POPE, C. A., THUN, M. J., NAMBOODIRI, M. M., DOCKERY, D. W., EVANS, J. S., 38
SPEIZER, F. E. & HEATH, C. W. (1995) Particulate Air-Pollution as a Predictor 39 of Mortality in a Prospective-Study of Us Adults. American Journal of 40 Respiratory and Critical Care Medicine, 151, 669-674. 41
RAMACHANDRAN, G., ADGATE, J. L., PRATT, G. C. & SEXTON, K. (2003) 42 Characterizing Indoor and Outdoor 15 Minute Average PM2.5 Concentrations in 43 Urban Neighborhoods. Aerosol Science and Technology, 37, 33-45. 44
22
SAMET, J. M., DOMINICI, F., CURRIERO, F. C., COURSAC, I. & ZEGER, S. L. 1 (2000) Fine particulate air pollution and mortality in 20 US Cities, 1987-1994. 2 New England Journal of Medicine, 343, 1742-1749. 3
SCHWARTZ, J., DOCKERY, D. & LIPFERT, F. (1995) Particulate Air-Pollution and 4 Daily Mortality in Steubenville, Ohio (Vol 135, Pg 12, 1992). American Journal 5 of Epidemiology, 141, 87-87. 6
SIOUTAS, C., KIM, S., CHANG, M. C., TERRELL, L. L. & GONG, H. (2000) Field 7 evaluation of a modified dataram mie scattering monitor for real-time PM2.5 mass 8 concentration measurements. Atmospheric Environment, 34, 4829–4838. 9
STERN, D. I. (2004) The Rise and Fall of the Environmental Kuznets Curve. World 10 Development, 32, 1419-39. 11
TERI (2001a) Air quality in Delhi. New Delhi, Tata Energy Research Institute. 12 TERI (2001b) Shut shop or shift base. New Delhi, Tata Energy Research Institute. 13 THOMAS, A. & GEBHART, J. (1994) Correlations between gravimetry and light-14
scattering photometry for atmospheric aerosols. Atmospheric Environment, 28, 15 935–938. 16
WALDMAN, A. (2005) All roads lead to cities, transforming India. The New York Times. 17 New York City. 18
WALL, M. M. (2004) A close look at the spatial structure implied by the CAR and SAR 19 models. Journal of Statistical Planning and Inference, 121, 311-324. 20
WHO (2000) Guidelines for air quality. Geneva, World Health Organization. 21 22 23
23
Table-1: Population Growth in National Capital Region, 1981-1991
1991
1981
Annual Exponential
Growth Rate (%)
District/State Name
Total Rural Urban Total Rural Urban Total
Rural Urban
Panipat 677157 467561 209596 492279 340841 151433 3.24 3.21 3.30
Sonipat 754866 576841 178025 593681 467821 125860 2.43 2.12 3.53
Rohtak 1867193 1423133 444060 1516142 1223830 292312 2.10 1.52 4.27
Faridabad 1477240 759727 717513 986076 577446 408630 4.12 2.78 5.79
Gurgaon 1146090 913386 232704 863865 694634 169231 2.87 2.78 3.24
Rewari 623301 528101 95200 496813 437494 59319 2.29 1.90 4.84
Haryana 6545847 4668749 1877098 4948856 3742065 1206786 2.84 2.24 4.52
Alwar 1395513 1115704 279809 1048886 861095 187791 2.90 2.62 4.07
Rajasthan 1395513 1115704 279809 1048886 861095 187791 2.90 2.62 4.07
Meerut 3447912 2171355 1276557 2767185 1903270 863915 2.22 1.33 3.98
Ghaziabad 2703933 1455673 1248260 1843297 1214180 629117 3.91 1.83 7.09
Bulandshahr 2849859 2257064 592795 2358569 1902422 456147 1.91 1.72 2.65
24
Uttar Pradesh 9001704 5884092 3117612 6969051 5019871 19491
80 2.59 1.60 4.81
Delhi 9420644 949019 8471625 6220300 452216 5768084 4.24 7.69 3.92
Delhi UT 9420644 949019 8471625 6220300 452216 5768084 4.24 7.69 3.92
NCR 26363708 12617564 1374614
41918709
31007524
791118
41 3.23 2.28 4.20
Source: Census of India 1981 and 1991
25
Table 2 Parameters used for Kriging to create air pollution surface. Parameter Value
Semivariogram model Exponential
Anisotropic direction 306°
Major range (DD) 0.39
Minor range (DD) 0.18
Lag size (degree decimals) 0.033
No. of points included 12
Neighbors to include 15 or at least 10 for each
angular sector
Angular sector 4
Note: the sill and nugget were computed using the automatic function within ArcMap to obtain the best fit for the semivariogram.
26
Inside Delhi Outside Delhi Total
Variables Inter-Quartile
Range
Mean (± 95% CI)
Inter-Quartile
Range
Mean (± 95% CI)
Inter-Quartile
Range
Mean (± 95% CI)
PM2.5 (μgm-3) Aerosol 11.9 39.1(±2.3) 21.2 46.7(±7.1) 14.0 40.2(±2.3)
PM2.5 (μgm-3) Gravimetric 8.5 27.3(±1.7) 14.9 33.4(±6.9) 8.8 28.2(±1.8)
PM7 (μgm-3) Aerosol 91.7 172.9(±18.0) 210.7 241.1(±66.7) 94.7 183.1(±18.7)
PM7 (μgm-3) Gravimetric 67.7 120.4(±11.8) 82.8 173.1(±63.3) 67.9 128.3(±14.1)
PM10 (μgm-3) Aerosol 113.8 207.1(±22.2) 254.7 291.9(±86.3) 125.8 219.9(±23.4)
PM10 (μgm-3) Gravimetric 91.3 147.1(±14.9) 108.7 213.1(±83.6) 90.2 157.0(±18.1)
Total suspended aerosol (μgm-3) 144.6 250.6(±27.1) 287.9 343.3(±103.9) 149.7 264.6(±28.3)
Total suspended aerosol (μgm-3) Gravimetric
107.4 178.5(±18.3) 129.0 250.3(±98.9) 113.0 189.3(±21.8)
Relative Humidity (%) 5.5 46.8(±0.7) 3.2 45.7(±1.4) 5.5 46.6(±0.6)
Temperature (ºC) 0.8 32.4(±0.1) 0.6 31.9(±0.3) 0.9 32.3(±0.1)
Table 3a: Ambient air pollutants in Delhi and its surroundings July 23 to December 3, 2003 - summary Statistics
27
Inside Delhi Outside Delhi Total
variable Inter-Quartile
Range
Mean (± 95% CI)
Inter-Quartile
Range
Mean (± 95% CI)
Inter-Quartile Range
Mean (± 95% CI)
Two wheeler/minute 15.0 16.8(±2.6) 17.6 13.7(±5.2) 14.9 16.3(±2.3)
Cars/minute 21.8 18.9(±3.5) 18.6 15.3(±6.3) 20.8 18.4(±3.1)
Buses/minute 4.4 2.8(±0.6) 1.8 1.4(±0.7) 4.1 2.6(±0.5)
Trucks/minute 1.8 1.2(±0.2) 2.0 1.5(±0.8) 1.8 1.3(±0.2)
Distance to the closest road (m) 0.98 0.87(±.25) 1.4 1.37(±0.84) 1.11 0.95(±0.25)
Distance to the closest industrial clusters (m)
2.5 2.47(±0.38) 1.14 0.7(±0.39) 2.6 2.2.0(±0.35)
Distance to Delhi border (km) 5.5 6.8(±0.7) 1.4 1.6(±0.6) 6.5 6.1(±0.7)
Distance to the city center (m) 7.5 10.7(±1.1) 6.1 16.3(±2.0) 7.6 11.6(±1.0)
Table 3b: Sources of air pollution in Delhi and Surroundings – summary statistics.
28
City Name 2001 2002 2001-02 2004 2005 2004-05 2001-02 - 2004-05
SO2 13.6(±0.70) 10.5(±0.52) 11.8(±0.45) 8.8(±0.51) 9.0(±0.40) 8.9(±0.32) -2.90(±0.56) NO2 70.6(±2.38) 76.2(±2.30) 73.8(±1.68) 86.8(±3.70) 84.6(±2.36) 85.4(±2.02) 11.55(±2.62) Delhi (ITO B.S.Z.
Marg) PM10 178.7(±21.9) 284.8(±34.7) 240.2(±22.7) 210.1(±13.7) 257.3(±15.0) 239.8(±10.9) -0.83(±25.9)
SO2 4.4(±0.23) 4.0(±0.06) 4.2(±0.11) 4.3(±0.19) 4.1(±0.04) 4.1(±0.07) -0.07(±0.15) NO2 30.4(±2.59) 19.4(±1.24) 24.4(±1.46) 18.6(±1.68) 20.2(±0.97) 19.6(±0.87) -4.80(±1.93) Kanpur (Vikas
Nagar) PM10 200.8(±22.4) 160.3(±14.1) 178.5(±12.8) 185.8(±27.6) 205.2(±18.2) 198.3(±15.3) 20.42(±20.1)
SO2 4.0(±0.07) 4.0(±0.04) 4.0(±0.03) 5.5(±0.57) 5.5(±0.27) 5.5(±0.29) 1.44(±0.36) NO2 15.6(±2.71) 10.0(±0.67) 11.2(±0.85) 12.8(±1.00) 21.3(±1.27) 17.6(±1.02) 6.44(±1.43) Chennai (Adyar)
PM10 60.1(±8.7) 37.3(±3.8) 41.7(±3.8) 58.1(±6.0) 56.5(±4.1) 57.2(±3.5) 15.45(±5.3) SO2 5.2(±0.34) 5.1(±0.34) 5.2(±0.24) 6.7(±1.09) 7.4(±0.63) 7.1(±0.59) 1.89(±0.57) NO2 42.9(±3.13) 42.8(±2.65) 42.8(±2.03) 38.1(±2.96) 48.1(±2.78) 43.9(±2.12) 1.01(±3.00) Kolkata
PM10 74.9(±8.8) 95.9(±8.3) 86.1(±6.1) 157.0(±17.2) 153.1(±16.1) 154.8(±11.8) 68.37(±12.3) SO2 6.6(±0.77) 5.2(±0.47) 5.8(±0.44) 9.2(±2.06) 6.0(±0.85) 7.2(±0.96) 1.43(±0.93) NO2 24.1(±2.98) 25.9(±2.22) 25.1(±1.81) 22.2(±3.15) 20.0(±2.20) 20.8(±1.82) -4.35(±2.73) Vadodara
PM10 72.1(±8.9) 76.9(±6.6) 74.8(±5.4) 66.4(±10.8) 81.8(±7.8) 76.1(±6.4) 1.28(±8.5) SO2 NA NA NA 16.6(±2.03) 24.9(±1.11) 22.0(±1.11) NA NO2 NA NA NA 92.3(±8.81) 91.8(±7.47) 91.9(±5.75) NA Mumbai
PM10 NA NA NA 526.1(±71.0) 213.3(±6.1) 318.8(±30.1) NA Table 4: Ambient air pollutants in Delhi and other mega cities 2001-02 and 2004-05 (μgm-3 (95% CI))
(Source: Central Pollution Control Board)
29
Distance to Delhi Border (km) Delhi Outside Delhi Difference
PM2.5(μgm-3)
<=1 36.9(±10.6; 6) 59.2(±17.4; 5) -22.3 (0.053)
<=2 41.6(±8.8; 11) 50.6(±9.4; 11) -9.0 (0.092)
<=3 40.7(±5.4; 18) 47.5(±7.4; 16) -6.8 (0.075)
All Sites 39.0(±2.4; 95) 47.5(±7.4; 16) -8.4 (0.012)
PM10(μgm-3)
<=1 183.8(±67.7; 6) 423.7(±226.0; 5) -239.8 (0.029)
<=2 244.7(±82.7; 11) 330.0(±119.6; 11) -85.3 (0.131)
<=3 236.8(±54.2; 18) 299.5(±90.5; 16) -62.7 (0.120)
All Sites 207.2(±22.4; 95) 299.5(±90.5; 16) -92.3 (0.007)
TSP(μgm-3)
<=1 226.7(±82.7; 6) 505.8(±271.0; 5) -279.1 (0.032)
<=2 290.8(±93.9; 11) 387.4(±145.2; 11) -96.6 (0.143)
<=3 282.3(±61.5; 18) 352.2(±109.0; 16) -69.9 (0.134)
All Sites 250.6(±27.4; 95) 352.2(±109.0; 16) -101.6 (0.014)Table 5: Distribution of ambient particles with reference to distance from Delhi border, July-December 2003.
30
Covariates (Trucks/Minute)^0.5 (Buses/Minute)^0.5 (Cars/Minute)^0.5
Distance to the closest industrial
cluster (km)
Distance to the closest road
(km)
Model1 Model2 Model1 Model2 Model1 Model2 Model1 Model2 Model1
Model2
0.166 0.162 0.036 0.048 0.014 0.016 -0.004 -0.003 0.002 0.002 Pollution Source (4.47)** (4.48)** -1.400 -1.900 -1.100 -1.310 (3.65)** (2.80)** -1.620 -1.420
0.166 0.197 0.183 0.1 0.168Outside Delhi = 1 (2.73)** (3.00)** (2.79)** -1.44 (2.57)*
3.49 3.469 3.599 3.553 3.593 3.557 3.802 3.761 3.598 3.579Constant
(84.3)** (84.7)** (86.8)** (83.0)** (66.9)** (66.2)** (78.2)** (66.7)** (93.9)**
(94.0)**
Obs 112 112 112 112 112 112 112 112 112 112
R-squared 0.15 0.21 0.02 0.09 0.01 0.08 0.11 0.12 0.02 0.08
Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1% Table 6a: Log(PM2.5) and pollution sources
31
Covariates (Trucks/Minute)^0.5 (Buses/Minute)^0.5 (Cars/Minute)^0.5
Distance to the closest industrial
cluster (km)
Distance to the closest road
(km)
Model1 Model2 Model1 Model2 Model1 Model2 Model1 Model2 Model1
Model2
0.201 0.194 -0.015 0.004 -0.017 -0.013 -0.007 -0.006 0.005 0.004 Pollution Source (2.90)** (2.86)** (0.31) (0.08) (0.75) (0.60) (3.86)** (3.07)** (2.17)* (1.98)
0.29 0.30 0.30 0.15 0.28Outside Delhi = 1 (2.52)* (2.53)* (2.51)* (1.22) (2.40)*
5.07 5.03 5.28 5.21 5.32 5.26 5.55 5.49 5.14 5.11Constant
(65.4)** (65.3)** (70.6)** (66.8)** (55.3)** (54.3)** (64.3)** (54.6)** (75.8)**
(75.5)**
Obs 112.00 112.00 112.00 112.00 112.00 112.00 112.00 112.00 112.00 112.00
R-squared 0.07 0.12 0.00 0.06 0.01 0.06 0.12 0.13 0.04 0.09
Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1% Table 6b: Log(PM10) and pollution sources
32
Log(PM2.5) Log(PM10) OLS SAM OLS SAM
0.258 0.261 0.477 0.46(Trucks/minute)^0.5 (3.66)** (3.71)** (3.82)** (3.76)**
-0.002 -0.002 -0.003 -0.004SQRT(Distance to Industrial Cluster (m)) (1.53) (1.62) (1.46) (1.77)0.002 0.002 0.003 0.003SQRT(Distance to the major road) (1.52) (1.46) (1.51) (1.37)-0.067 -0.068 -0.215 -0.203SQRT(Buses/minute) (1.46) (1.48 (2.63)** (2.55)*0.079 0.086 0.09 0.102Outside Delhi Border = 1 (1.14) (1.25) (0.73) (0.85)
0.24 0.385Spatial Autocorrelation (1.31) (2.46)*
3.52 3.523 5.14 5.172Constant (40.25)** (40.41)** (33.13)** (34.01)**
Observations 111 111 111 111R-squared 0.3 0.31 0.28 0.32
Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1%
Table 7: Ambient particles and sources of air pollution – OLS and Spatial Autoregressive Models.
33
Figure 1: National Capital Regions: population density and CNG regulations.
34
Figure 2a: Delhi and other megacities in Northern India.
35
Figure 2b: Air pollution monitoring sites and sources of air pollution in Delhi and its surroundings.
36
010
020
030
040
0
100 200 300PM10 (High Volume Sampler)
PM
10 (P
hoto
met
ric S
ampl
er)
Figure 3: PM10 from photometric and gravimetric samplers at ITO, Delhi, July 23-December 2003.
37
Figure 4: PM2.5 (μgm-3) in Delhi and its surroundings, July-December 2003.
38
Figure 5: PM10 (μgm-3) in Delhi and its surroundings, July-December 2003.
39
Figure 6: TSP (μgm-3) in Delhi and its surroundings, July-December 2003.
40
Figure 7: Urban sprawl in the southern west part of Delhi’s border.
41
3
3.5
44.
5
0 .5 1 1.5 2 2.5SQRT(Trucks/minute)
log(
PM
2.5
(mic
rogr
am/c
ubic
m))
4.5
55.
56
6.5
0 .5 1 1.5 2 2.5SQRT(Trucks/minute)
log(
PM10
(mic
rogr
am/c
ubic
m))
Figure 8a: PM2.5 and frequency of trucks. Figure 8c: PM2.5 and proximity to industries.
4.5
55.
56
6.5
0 .5 1 1.5 2 2.5SQRT(Trucks/minute)
log(
PM10
(mic
rogr
am/c
ubic
m))
4.5
55.
56
6.5
0 20 40 60 80SQRT(Distance to Industrial Cluster (m))
log(
PM
10 (m
icro
gram
/cub
ic m
))
Figure 8b: PM10 and frequency of trucks. Figure 8d: PM10 and proximity to industries.