Key Words particulate matter, pollution sources, air concentrations, human exposure, health effects, uncertainty estimates

• Abstract Indian megacities are among the most polluted in the world. Air con- centrations of a number of air pollutants are much higher than levels recommended by the World Health Organization. In this paper, we focus on Mumbai and Delhi to characterize salient issues in health risks from particulate air (PM₁₀) pollution in Indian cities. We perform a synthesis of the literature for all elements of the causal chain of health risks—sources, exposure, and health effects—and provide estimates of source strengths, exposure levels, and health risks from air pollution in Indian cities. We also analyze the factors that lead to uncertainty in these quantities and provide an overall assessment of the state of scientific knowledge on air pollution in urban India.

INTRODUCTION

In recent decades, urban centers in less-industrialized countries have experienced unprecedented growth, and megacities with populations of 10 million or more people have emerged in many countries. In India alone there are four such cities, with three others expected to join the ranks in the next 20 years. Globally, many rapidly growing cities are being overwhelmed by environmental problems, espe- cially those related to air pollution. Deterioration of air quality is a problem that is directly experienced by a majority of the 300 million urban Indians, who constitute 30% of India’s population.

Megacities of India are no exception to the global pattern of deteriorating urban air quality. Indian cities are among the most polluted in the world, with concentra- tions of a number of air pollutants being well above recommended World Health Organization (WHO) levels (WHO/United Nations Environment Program 1992, Mage et al 1996). Despite the magnitude and urgency of air pollution as a public health issue, scientific understanding of health risks from air pollution in Indian cities is poor, and there is a paucity of scientific studies on the health effects. The few that have been done show much cause for alarm, and it is apparent to scientists and lay people alike that the residents of India’s megacities face significant risks to their health from exposure to air pollutants.

The dearth of data exists across the entire causal chain of risk assessment, from sources of pollution to atmospheric concentrations to human exposures and their health effects. Certainly, more data exist in some of these categories than in others. There is very little known about some sources that contribute to air pollution, while others are better characterized; ambient concentrations of various pollutants are being monitored more systematically than in the past (especially in urban centers), but very few studies have looked at personal exposures to these pollutants. There have also been few epidemiological studies to evaluate the health effects of air pollution in Indian cities, and few studies have attempted to synthesize knowledge regarding human health risks from air pollution in an integrated manner; in other words, there is very little systematic understanding of sources of air pollution, and exposure and effects data have rarely been measured simultaneously and in a single consistent experimental design.

In this paper, we review the literature and characterize salient issues in the calculation of health risks from particulate air pollution in Indian cities. We focus on particulate air pollution because it has become increasingly clear that thoracic particulate matter (PM) is the major cause of human mortality and morbidity from air pollution; studies in the United States (US) have indicated that there are 20,000–100,000 deaths due to particulate pollution per year (US Environmental Protection Agency [EPA] 1996). Particulate pollution in Indian cities is far worse,

Health Effects (Mortality & Morbidity)

. Extrapolations from dose-

.

.

.

.

.

response relationships obtained in US populations Extrapolations from dose- response relationships obtained in China

Time-series epidemiological studies in India

Cross-sectional and longitudinal studies Common confounders in epidemiological studies in India

Bottom-up calculations of mortality

Concentrations and Exposures:

. Exposure effectiveness

.

.

.

.

of sources Ambient

concentrations of PM, metals, PAHs etc.

Time activity patterns of people

Exposures in various micro-environments Personal exposures to PM10 by income class

Sources:

.

. Industry

Vehicles

.

Power plants

.

Domestic fuel combustion Miscellaneous

Figure 1 The causal chain for human health risks. PAHs = polyaromatic hydrocarbons.

so it is likely that per-capita mortality from urban air pollution in India is at least as high as that in the US. In addition to particulate air pollution, other primary pollutants are also a cause for concern in India. Thus, we also provide source inventories for these primary pollutants (CO, nitrogen oxides, and hydrocarbons) when possible.

We have performed a synthesis of the literature for different elements of the health risk in the Indian context and provide an overall assessment of the state of the science. We believe that such a synthesis is best done by performing our own calculations—of source inventories for particulate matter, of human exposure to particulates, and of the health effects of this exposure—in a comprehensive manner. A key challenge for performing such a synthesis for India is in assessing the level of uncertainty and, if possible, suggesting ways for reducing it.

To analyze the causes and consequences of air pollution in India, it is instructive to view the “causal chain” for risk of human mortality. The causal chain describes the stages at which pollutant emissions can be translated into human exposure and ultimately to health effects. We analyze each aspect of the causal chain shown in Figure 1 for its Indian context and summarize important findings in terms of what is known, what is uncertain, and what is missing. Quantitative assessment of the key elements of the risk causal chain allows us to focus on data availability and uncertainty and consequently to characterize future research needs.

We characterize data availability and uncertainty in each element of the causal chain, from sources through health effects. To that end we explicitly characterize uncertainties and use Monte Carlo simulation in all our calculations; the range for each variable presented in this paper is 1σ , which captures about 70% of the distribution. We present source inventories that have been calculated by others,

as well as an inventory that we developed which includes a systematic character- ization of uncertainties. We also examine the importance of the different source categories as they relate to human exposure, using a simplified modeling approach. We follow this with an examination of the link between exposure levels and health effects in Indian cities, and we perform health risk calculations based on our esti- mates of levels of human exposure to PM. We do this for different income classes, since we expect, on the basis of existing studies, that household income is a major determinant of exposure to PM. There have been a few recent studies that have characterized human health effects of air pollution for Indian cities. We review the findings of these studies and compare the results to the results of health risk calculations that we have performed. One consequence of a lack of indigenous data and analysis is that the research on the proximate causes and health conse- quences of air pollution in India uses data from industrialized-country settings. When possible, we analyze the hazards inherent in using data and inferences from different locations, such as those in the US, and transposing them to the Indian context.

Throughout this paper, we focus our attention on Mumbai and Delhi. In choos- ing Delhi and Mumbai, we are not suggesting that urban air pollution problems in India are localized to megacities. In fact, measurements have shown that the air quality in many small towns is just as poor [Tata Energy Research Institute (TERI) 1997]. We chose to focus on Mumbai and Delhi simply because the problem of air pollution has been studied more in these urban centers than it has in other places in India. We assume that these two cities are somewhat representative of the larger Indian cities and that the conclusions we draw, in a broad sense, hold for other cities as well.

The rest of the paper is structured as follows. In the second section we provide an inventory of air pollution sources in Indian cities. In the third section we assess the state of the science as it relates to air concentration and exposure data in major Indian cities. Here we compare exposure to PM among the different income groups and compare various sources of air pollution in terms of resulting exposure levels in these groups. In the fourth section we examine the epidemiological evidence for human health risks in Indian cities. In the fifth section we conclude our review with lessons for scientific analyses and research on air pollution health risks in India.

SOURCES OF URBAN AIR POLLUTION IN INDIA

The main categories of urban air pollution sources in India are vehicular emissions, industrial emissions, fuel use for domestic purposes such as cooking, and a poten- tially large miscellaneous category, which includes burning of household wastes and emissions from small businesses and cremation grounds. Natural sources of PM are also significant, depending on location and season. While particu- lates from natural sources are not conventional pollutants, their contributions are typically taken into account in inventories of total suspended particulates (TSP),

since natural sources can be both a major contributor to polution and a source of uncertainty. However, natural dust particles are coarse and do not contribute sig- nificantly to PM fractions that actually get deposited in human lungs. In Delhi, for example, dust-laden winds from the western desert during the dry season increase the TSP levels, although they have a much smaller impact on particles <10 µm in diameter. In this paper we do not provide inventories for natural source particles. We focus on inventories for particles <10 µm in diameter (PM₁₀). This fraction very closely approximates the particles that can penetrate to the thoracic region of the human lung and thus is a health-relevant measure of exposure.

In the present section, we synthesize findings from the literature and present a tentative inventory for Mumbai and New Delhi, based on a simple accounting model. The inventory includes uncertainties in total emissions and highlights un- certainties in the parameters within and across source categories needed to build an emissions inventory.

Vehicular Emissions

Rapid urbanization in India has led to an increase in transportation demand that public transport systems have been unable to adequately meet. Consequently, the use of personal vehicles has increased dramatically, as seen in Figure 2. Between 1986 and 1991, the total number of vehicles in India increased roughly threefold, from about 9 million to 25 million, and it was estimated that the number of vehicles would reach well above 40 million by the year 2000 (Government of India [GOI] 1993). Roughly half of these vehicles are in three major metropolitan cities: Delhi (∼30%), Mumbai (∼12%), and Calcutta (∼8%). Nationwide, about 70% of the vehicles are gasoline-fueled personal vehicles, two- or three-wheeled vehicles that have two-stroke engines. Other gasoline-fueled vehicles, mostly cars and motorcycles with four-stroke engines, make up 14% of the fleet, and diesel-fueled trucks and buses make up ∼8% of the total (see Table 1).

The number of two- and three-wheeled vehicles, which also represents the largest fraction of all vehicles, has been growing at the rate of ∼20% annually and, between 1987 and 1993, increased threefold, from 7 million to ∼20 million. The number of two-wheelers is expected to keep rising, with a projected 36 million by 2000. Passenger cars and diesel-fueled vehicles, although fewer in numbers, will double in the same time period (see Figure 2). In the mid-1980s, the introduction of cars by Maruti, a public-sector company jointly launched by the Indian govern- ment and Suzuki of Japan, gave impetus to car purchasing by members of India’s upper classes. Because the government’s liberalization program, launched in the early 1990s, has encouraged car production by multinationals in India, there has been an even more dramatic rise in the number of passenger cars in the country. Recent figures showed that, between the late 1980s and 1997, the annual sales of automobiles increased >10-fold, from 40,000 to 400,000 (Chakravarti 1998).

In keeping with the increase in numbers of vehicles, the vehicular use of gasoline and diesel fuel more than doubled over the time period of 1981–1994, increasing

30000

Two- and three-wheelers Cars

Buses Trucks Others

All vehicles

25000

20000

Number in 1000s

15000

10000

5000

0

1965 1970

1975 1980

Year

1985 1990

1995

Figure 2 Growth in the number of vehicles over the past three decades in India. Close to 50% of these increases have occurred in three cities—Delhi, Mumbai, and Calcutta, with Delhi alone accounting for ∼30% of the increase.

from 1.5 and 7.2 million tons, respectively, in 1981, to 3.5 and 14.8 million tons in 1994 (Agrawal et al 1997). Although >80% of the vehicular fleet consists of vehicles that use gasoline, the total amount of diesel fuel consumed in India exceeds the usage of gasoline by close to a factor of five. Diesel fuel is the primary fuel for buses, trucks, and other commercial vehicles, which consume larger quantities of fuel per road mile and also constitute a larger share of road miles traveled. In addition, a significant amount of diesel fuel is consumed in the generation of power by captive power plants (which supply roughly 10% of the energy consumed in India). The distinction between gasoline and diesel fuel is important because the contributions of both fuel types to multiple pollutants and air pollution health risks

TABLE 1 Percentages of different types of vehicles in major metropolitan Indian citiesa

2–3 Wheelers (%) Cars (%) Buses, trucks (%) Others (%)

aSource, Biswas & Dutta 1994. Note that, as cities with functioning train and subway systems, Mumbai and Calcutta have lower fractions of two and three wheelers than the others, perhaps reflecting the role that lack of public transport plays in emissions increases.

are significant, but the technical and policy solutions for reducing these emissions may be quite different.

The principal pollutants emitted by vehicles are carbon monoxide (CO), NOx, particulate matter (PM₁₀), volatile organic compounds, and semivolatile polyaro- matic hydrocarbons (PAHs). Sulfur oxides are emitted in various quantities de- pending on the sulfur content of the fuel. Exhaust gases from gasoline-fueled vehicles also contain lead (Pb) additives, which continue to be used although there has been a recent move toward unleaded gasoline in the major cities.

In this paper, we do not include ozone in any of our discussions. The conditions for significant ozone pollution—the presence of volatile organic compounds, NOx sources, abundant sunlight, and meteorological conditions such as winter inversion layers—exist in many Indian cities, particularly in Delhi and Calcutta; yet the magnitude of the ozone problem is almost entirely unknown. To our knowledge there have been no measurements of ozone in any of the major Indian cities.

From Vehicular Growth to Growth in Pollution Translating numbers of vehi- cles to gross emissions of pollutants requires the development of representative values, called emission factors, that for each vehicle category relate the quantity of a pollutant released into the atmosphere to the level of activity associated with each vehicle type.

An emission factor, F, can be defined as

F = E/A, 1.

where E is the amount of pollutant released (in grams) and A is the level of activ- ity, typically measured in kilometers driven. The emission factor, F, depends on a number of variables, such as the intrinsic nature of the technology; the power of the vehicle; and vehicle operating conditions, including engine temperature, vehicle speed, ambient temperature, deterioration from age and usage, quality of maintenance, and the quality of the fuel. For all these reasons, emission factors within vehicle categories are subject to a good deal of variability. Emission factors

are typically developed for each vehicle type, and an assessment of the variability must be performed to derive a range of values for emissions under different operat- ing conditions. Typically, emission factors are reported for a set of ideal conditions (e.g. for a particular speed, operating temperature, and vehicular age) and with cor- rection factors that can be used to extrapolate to other operating conditions (US EPA AP-42 manual 1985).

There are several reasons to expect larger emission factors for vehicles in India than their equivalents in industrialized countries. A majority of vehicles in India are not equipped with pollution control equipment, and only recently has the government mandated installation of catalytic converters on cars sold in the major metropolitan areas, including Mumbai and Delhi. In addition, maintenance of vehicles is poor, and there is very little monitoring and enforcement of emission standards.

Average traffic speeds, particularly in Delhi, have dropped dramatically over the past decade as vehicle density (the number of registered vehicles per square kilo- meter) has risen rapidly (GOI 1993). The average speed of ∼20 km/h (TERI 1993) in Delhi is far slower than the US average of >50 km/h. Slower speeds (<30 km/h) with increased phases of acceleration and deceleration result in significantly larger emission factors than do higher cruising speeds.

Another key issue in the derivation of emission factors is the quality of fuel used. In the Indian context, the quality of fuel, especially adulteration of gasoline by kerosene, is a particularly important, if understudied, issue. This problem is almost universal among motorized three-wheeled vehicles (auto-rickshaws), which for the most part are not owned by their operators. Kerosene is heavily subsidized because it is seen as an important fuel for household cooking for lower-income groups. Since the price of gasoline per liter exceeds that of kerosene by a factor of five, auto-rickshaw operators typically adulterate their gasoline fuel with as much as 30% kerosene. In addition, operators mix excessive amounts—as much as 10% (as opposed to the typical operating value of 1.5%–2%)—of lubricating oil to compensate for poor fuel quality resulting from addition of kerosene. While this can shorten engine life, the vehicles are typically not owned by the operator but rented from the owner, so there is little incentive on the part of the operators to not adulterate the fuel. Lubricating oil, which is sold in unpackaged form mainly for use in motorized two- and three-wheeled vehicles, is also adulterated. Furthermore, gasoline is not the only petroleum product that is adulterated. Diesel fuel is also adulterated with kerosene, since the price of the former is almost twice that of the latter. Kerosene and lubricating oil increase emission factors, although there are few measurements of the extent to which they do so.

In the United States, the EPA has developed the infrastructure to create a regu- larly updated inventory of emission factor data for a wide range of vehicular and other pollutant sources (US EPA 1985). However, as noted earlier, the vehicular fleet status and operating conditions in India are quite different from those in the US, and US data for the most part are not very helpful in evaluating emission factors under Indian conditions. In India, emission factor data are not as extensive and

TABLE 2 A compilation from the literature of emission factors for different criteria pollutants and for different vehicle categories

Pollutant emission (g/km)a

Vehicle category

1996, Shah & Nagpal 1997, IIP 1994

a—, Negligible or unknown level of emission.

are less readily available, although some data are available from studies conducted by the Indian Institute of Petroleum and the Automotive Research Association of India (Table 2). In Table 2, we provide a compilation of emission factors from various sources and compare them with emission factors evaluated by the US EPA.

While the quantities in Table 2 are uncertain, some patterns can still be observed. It is apparent from Table 2 that emission factors representing Indian conditions are generally higher than those derived for US conditions. The differences are particularly acute for emissions of PM, many of which are an order of magnitude higher for India than they are for the US for some categories. For gasoline cars as a category, emissions of most pollutant types except for PM₁₀ are almost an order of magnitude higher than in the US. This is not surprising since the majority of Indian cars do not have catalytic control equipment; the age and maintenance levels of the Indian fleet also lead to greater emissions. Furthermore, emission factors representing Indian conditions may be underestimates because vehicles operating in field conditions may show greater deterioration than those tested in laboratories.

In Table 3 we provide emission factor probability distributions for different vehi- cle types for major pollutants. The distributions are subjective and based on ranges and median values presented in the literature. The distributions include variability in emission factors in vehicle populations as well as measurement uncertainties. The distributions are conservative; the lower bounds of the distributions are the emission factors that are typical for US vehicles, the modes represent values typi- cal for Indian vehicles, and the upper bounds are subjective estimates based on the field expertise of the authors. Note that the emission factors in Table 3 are for ex- haust emissions alone. In addition to exhaust, automobiles can release pollutants from crankcase emissions.

Inventory of Emissions from Vehicular Sources Emissions of major pollutants depend on emission factors and levels of usage of different vehicles. Annual emis- sions from a category of vehicle can be calculated, as shown in Equation 1, by multiplying the emission factor provided in Table 3 (in grams per kilometer) by the

TABLE 3 Probability distributions for emission factors for different vehicle types in India

Emission factors (g/km)a

Vehicle

category PM10 CO Hydrocarbons NOx

Gasoline cars T(0.3, 0.4, 0.5) T(5, 25, 40) T(0.3, 5, 10) T(0.5, 3, 6)

Diesel cars T(0.5, 1.5, 2.0) T(0.5, 2, 4) T(0.25, 1, 2) T(0.8, 1.1, 1.4)

Heavy-duty, T(0.5, 1.5, 3.0) T(1, 3, 5) T(0.5, 2, 4) T(5, 10, 15)

diesel vehicles (trucks and buses)

Two-stroke engines T(0.2, 0.4, 0.6) T(5, 15, 20) T(5, 10, 25) T(0.2, 0.3, 0.5)

(scooters, motorcycles, and three-wheelers)

aT(a,b,c) refers to a triangular distribution with mode b, and a and c as lower and upper bounds, respectively. The distributions are subjective and based on ranges and median values presented in the literature (see Table 2).

TABLE 4 Probability distributions for usage of different vehicle types in India

Vehicle (N ) in 1995 in city Vehicle category Usage per vehicle (in km)a Delhi Mumbai

aUsage range provides the two end points of a uniform distribution. Usage levels per vehicle are assumed to be similar for Delhi and Mumbai. U (l, u) is a uniform distribution in the range with l and u as upper and lower bounds respectively.

annual usage (in kilometers) for that category. In Table 4, we provide a subjective estimated range for daily usage for each of the vehicle categories.

The entries in Tables 3 and 4, along with Equation 1, enable us to calculate a zero-order inventory for exhaust emissions from automobiles in Delhi and Mumbai. Uncertainties in emission factors and levels of usage can also be included in such an inventory through Monte Carlo simulation. Results of this analysis are shown in Table 5, which provides an inventory of annual particulate emissions from vehicles in Delhi and Mumbai. Table 5 also provides the results for other key pollutants from vehicular usage for these two cities.

The most striking pattern that can be observed from the data is the rapidity of the increase in the numbers of vehicles (Figure 2). Vehicle numbers in major urban centers have increased by more than fivefold over the past two decades, a rate that is faster than that of population growth in these cities. It is not surprising, therefore, that vehicular pollution has grown accordingly.

No single category of vehicles dominates the gross amounts of emissions of all primary pollutants (Table 5). This implies that air pollution control strategies will have to include consideration of simultaneous control of the different pollutants across all categories of vehicles. For instance, since 1996, catalytic converters have been required for all cars registered in the major cities where unleaded gaso- line is now available. Yet, this will not solve the problem of pollution from carbon monoxide and hydrocarbons, even if all cars are equipped with the new technol- ogy, since the most significant contributions for both of these pollutants are other sources, such as two-stroke engines. A similar issue may also arise for ozone control because different vehicle types contribute differently to the precursors of ozone—two-stroke vehicles are the major contributors of hydrocarbons, while cars and diesel-fueled trucks and buses are responsible for much of the NOx. When the analysis is repeated on a per-passenger basis, vehicles with two-stroke engines and cars are, by far, worse than buses, highlighting the virtues of mass transportation (Table 6). This raises issues of environmental and energy equity, because the pro- ducers of pollution are middle- and higher-income groups that use two-wheelers

TABLE 5 Emissions of key pollutants for vehicular categories for Delhi and Mumbaia

Particulates Hydrocarbons Carbon monoxide Nitrogen Oxides Vehicle category Delhi Mumbai Delhi Mumbai Delhi Mumbai Delhi Mumbai

aData are thousands for metric tons per year. Estimates were obtained by using emission factor ranges in Table 3 and usage ranges in Table 4.

TABLE 6 Expected values of emissions of key pollutants per passenger for different vehicular categories for Delhi and Mumbai

Emissions per passenger (g/passenger km)a

aIt is assumed that the average number of passengers on a typical trip is 1.5 for a two-stroke vehicle,

2.5 for cars, and 40 for buses.

and cars while the health consequences are shared by lower-income groups that cannot afford these amenities.

Cities are impacted to different degrees by the size and composition of their ve- hicular stock. The total number of vehicles in Delhi well exceeds that in Mumbai. In fact, there are more total vehicles in Delhi than in the other three major metropoli- tan areas—Mumbai, Calcutta, and Chennai—combined (TERI 1997). Thus, one would expect the magnitude of gross pollutant emissions from vehicular traffic in Delhi to far exceed those in Mumbai and other cities. This is confirmed in Table 5, which shows estimates of levels of primary pollutants from vehicular sources in Delhi and Mumbai. Cars, which have higher levels of emissions per kilometer for the principal pollutants, make up a much larger fraction of vehicles in Mumbai than in Delhi, where vehicles with two-stroke engines have become predominant. Consequently, vehicles with two-stroke engines dominate emissions in Delhi.

In Table 7, we show a comparison of net vehicular emission data from this study [which uses a range of emission factors, daily usage (i.e. kilometers tra- veled per day), and total vehicular population in each category) with total emissions

TABLE 7 Comparison of net emissions from the transport sector from three different studies

Net emissions (thousands of metric tons/year) This study

calculated for two other studies [Central Pollution Control Board (CPCB) 1994, Shah & Nagpal 1997]. Our estimates for emissions of PM₁₀ and NOx are con- siderably higher than those of either of these studies. The difference lies in our assumptions regarding emission factor data which we have compiled from the lit- erature. We believe that the emission factors used in the individual studies may be systematically lower than the synthesized estimates used here.

The above review is focused on synthesizing what is known of source inventories for vehicles in Indian cities. A detailed analysis of the policy issues at the heart of this subject is outside the scope of this paper. The interested reader is referred to an article by Agrawal et al (1996), which provides a good overview of the policy issues related to vehicular air pollution in India.

Power Plant and Industrial Emissions

After gaining independence in 1947, India embarked on a path of rapid industri- alization in all the major manufacturing sectors—iron and steel, heavy manufac- turing, industrial and petrochemicals, and agricultural and paper products. Today, despite its label as a “less-industrialized country,” India is heavily industrialized, with a thriving manufacturing sector that until recently was largely indigenous. The CPCB has catalogued over 1500 large-scale industrial units in 17 industrial categories (CPCB 1995), accounting for about 60% of India’s industrial output. Small-scale industries are an important part of the Indian economy and account for the remaining 40% of the industrial output. At present, India has over three million small factories (Confederation of Indian Industry 1996).

Mumbai and Delhi are both major industrial centers with many large- and small-scale industries. In addition to being India’s financial and commercial cap- ital, Mumbai is also India’s most industrialized city. The industrial belt in and around Mumbai is responsible for more than 10% of India’s industrial produc- tivity, with a substantially greater fraction of the country’s chemical, petrochem- ical, and drug manufacturing. Not surprisingly, residents of Mumbai, particularly those in the eastern suburbs where the larger industries are concentrated, face a disproportionate burden of industrial emissions. Small-scale manufacturing is also spread over the entire greater Mumbai region. Although exact numbers are hard to establish, one estimate suggests that there are 40,000 small-scale plants and big industries in the Mumbai area, of which 32 are classified as hazardous (Shah & Nagpal 1997). Industries contributing to air pollution include textile mills, chemical and pharmaceutical engineering units, and foundries. While reg-

ulations have limited the growth of large industrial plants in the capital terri- tory of Delhi, this has not affected the growth of number of small-scale units.1

1According to the CPCB (1995), the state of Maharashtra had 321 major industrial units (in all industrial categories) in 1995, of which almost 40% (i.e. close to 130) were located in Mumbai. The same source indicated that Delhi has only five such units. Admittedly the data on Delhi refer only to the Union Territory of Delhi and not the adjoining industrial belt consisting of Uttar Pradesh and Haryana.

Small-scale manufacturing units in Delhi include everything from metalworking to food processing. One estimate places the number of small-scale units at 93,000 (CPCB 1993).

Industries emit a wide variety of process-specific pollutants—gaseous organic and inorganic compounds, complex vapors that undergo phase transformation after emission into the atmosphere, and PM with process-specific composition (e.g. heavy metals and PAHs). The presence of a large number of small-scale industrial plants makes pollution control more difficult because small-scale oper- ations are more financially strapped and less technologically capable than large- scale ones, and their numbers make the already lax monitoring and enforcement of pollution control laws even more difficult.

Criterion pollutants (SOx, NOx, CO, HC, and PM) that are released as a part of industrial combustion may be quantified to the first order on the basis of overall estimates of fuel used and average emission factors for various industrial activ- ities. Determination of the levels of specific toxic substances released into the environment, on the other hand, requires plant-specific data for each toxic com- pound. Construction of inventories for specific chemicals is a resource-intensive and difficult activity owing to its process-specific nature. Detailed inventories for each industry and each plant can be calculated only through regular monitor- ing of emissions. Curiously, while a catastrophic industrial disaster in India—the Bhopal tragedy—was the catalyst for enactment of the Emergency Planning and Community-Right-to-Know Act (EPCRA) in the US, leading to the creation of a Toxics Release Inventory (TRI), such databases are not available for India. The result is that data for estimating industrial emissions are sparse.

While it is a safe assumption that Mumbai and Delhi have substantial emis- sions of toxic substances and heavy metals, many small towns also may have local manufacturing units with potentially high-level toxic emissions. For exam- ple, Moradabad, a small town in the state of Uttar Pradesh, is a center for brass production (smelting, electroplating, cutting, scraping, and machining), and am- bient measurements there have revealed very high levels of heavy metals such as Pb, Cd, Cu, and Zn (Tripathi et al 1989). It is likely that many other places face similar localized issues regarding the release of toxic substances.

One consequence of the lack of detailed inventories and the heterogeneity of emissions is that simple scaling of information from one situation cannot be used to determine industrial emissions in another situation. In the rest of our discussion of industrial emissions we will construct an inventory of emissions from two categories—power plants and other industries—for Mumbai and Delhi.

Power Plants Coal-fired power plants generate two-thirds of India’s electric power (GOI 1996). Its coal-fired power capacity is expected to grow from 55 GW in 1996 to ∼80 GW by 2002. Indian steam coal is high in ash content (30%–50%) but low in sulfur (<0.5%). More than 99% of the coal used in the generation of elec- tric power in India is domestic steam coal. Additionally, the ash is very high in silica and aluminum (>90%). This results in very high resistivity for the fly ash (10¹³ to 10¹⁵ ▲-cm), which makes it difficult for conventional electrostatic precipitators

(ESPs) to collect fly ash efficiently (Lookman & Rubin 1998). Conventional ESPs are the only devices used in Indian power plants for control of PM; thus, more efficient pre- and postcombustion methods such as coal washing and the use of flue gas conditioning are not being applied. Existing efficiencies of ESPs of 85%–95% result in emissions of >45 million tons of fly ash from Indian power plants each year (Confederation of Indian Industry 1996). The main method of disposal of fly ash from power stations is mixing it with water; the resultant slurry is pumped through pipes to ash disposal ponds. Coal combustion in thermal power plants also emits a variety of toxic heavy metals, such as Pb, Zn, Ni, Co, Cd, Cr, and Cu.

Delhi has three power plants, all coal fired, located within its city limits: the 235-MW Indraprastha Power Station (IPP station); the 135-MW Rajghat Power House, owned by Delhi Electric Supply Undertaking; and the 720-MW Badarpur Thermal Power Station (BTPS). The total quantity of fly ash from the three power plants is about 6000 tons per day (Indraprastha 1200–1500, Rajghat 600–800, and Badarpur 3500–4000 tons per day [GOI 1997]). In these power plants, ash is collected by ESPs which have collection efficiencies that are higher than average for India—99.3% (IPP station), 99.7% (Rajghat Power House), and 98% (BTPS) (Mehra et al 1998). Nonetheless, there are episodes of major particulate pollution around the power stations from fly ash dispersal. The larger IPP plant also has shorter stacks (60 m) than the Rajghat Power House plant (160 m) and thus is more likely to cause human PM exposures. We calculated PM₁₀ emissions from power plants in Delhi by using the above estimates for fly ash production and a range for ESP collection efficiency of 97.5%–99.5%. The resulting estimate for PM₁₀ emissions is between 45,000 and 125,000 tons/year.

In contrast, Mumbai has one major thermal power plant, the Tata thermal plant, located in the industrial eastern section of the city. In addition, a 360-MW nuclear power plant is also located near Mumbai, in Tarapur. The Tata thermal station has a total generation capacity of 1 GW (2 units of 500 MW) and can use multiple fuel types—coal, natural gas, and oil—with typical usage of these fuels in the ratio of 1:2:3. Ash and PM are collected by ESPs, and at 278 m the plant’s stacks are tall. The URBAIR study (Shah & Nagpal 1997) estimated that the total PM₁₀ emitted from power plants in Mumbai is roughly 1500 tons/year. This is considerably less than the estimate for Delhi, and partly because the Delhi power plants are completely coal based whereas the TATA thermal plant uses coal for only one- fifth of its fuel needs. The rest of the fuel requirement is met by using distillate oil and gas. Furthermore, the URBAIR study uses US emission factors, which may underestimate emissions in India by an order of magnitude. The particulate emission limit for Indian power plants is set at 150 mg/Nm³, although the general level of compliance is acknowledged to be poor. By contrast, in the United States, the New Source Performance Standard is 30 mg/Nm³ while the current best- practice level is roughly 5 mg/Nm³ (Lookman & Rubin 1998). We assume that the per-megawatt PM₁₀ emissions from coal-based power in Mumbai are the same as in Delhi, and PM₁₀ emissions from Mumbai are estimated to be in the range of

7000–20,000 tons per year.

TABLE 8 Industrial and power plant emissions of PM₁₀ in Delhi and Mumbai

Emissions from sourcea (thousands of metric tons/year)

Study Power plants Other industries

Delhi

CPCB 1994 18.25 21.9

Saxena & Dayal 1997 6.0–24.3 3.5–10.2

This study 45–125 45–125

Mumbai

URBAIR (Shah & Nagpal 1997) 1.5 2.4

This study 7–20 ∼45–125

aRanges capture one standard deviation around the mean of the distribution.

Emissions from Other Industries Particulate emissions resulting from indus- trial combustion of fossil fuels have been characterized for Mumbai and Delhi in some studies. Table 8 shows levels of PM₁₀ emissions by power plants and other industries in these two cities, estimated by three different studies besides this one. For Delhi, while the three other studies have different estimates of total emissions, they show roughly equal contributions by power plants and other industries to PM₁₀ emissions. A GOI white paper (GOI 1997) on air pollution in Delhi also estimated that PM₁₀ emissions from other industries were roughly equal to power plant emissions. There is insufficient information on the assumptions that went into the inventory calculations of the two other studies (CPCB 1994, Saxena & Dayal 1997).

In the case of Mumbai, the World Bank URBAIR study (Shah & Nagpal 1997) developed a detailed emission inventory for industrial sources for the metropolitan Mumbai region. Based on amounts of fuel used and representative emission factors for different fuel types, the URBAIR study calculated levels of PM₁₀ emissions from industries and emissions from marine shipping. Since no specific data on industrial processes were available, process and fuel combustion emissions were not separated. The PM₁₀ emissions for Mumbai were estimated to be 2377 tons per year. We find it surprising that the estimates for Mumbai derived in the URBAIR study are lower by at least an order of magnitude than the available estimates for Delhi even though Mumbai is a more industrialized city. This discrepancy may be partly due to the emission factors used in the URBAIR study, which are based on US EPA AP-42 data (US EPA 1985).

In the absence of reliable information on industrial emissions of PM₁₀ in Mumbai, we made the assumption that industrial emissions in Mumbai are roughly equal to those in Delhi, although the range of uncertainty in both cities is rather high (40,000—125,000 tons/year). Mumbai is a more industrialized city than

Delhi, so we expect emission levels in Mumbai to be higher than those for Delhi. The high range for industrial emissions in both of the cities admits this possibility, although, given the available information, we cannot estimate emissions from the two cities with greater accuracy.

The industrial-pollution estimates must be seen only as order-of-magnitude assessments. Admittedly, they represent a crude approximation of a vastly more complicated picture, illustrating the difficulties in estimating reliable inventories in the absence of systematic monitoring of emissions from various industrial sectors on a plant-by-plant basis. Given the substantial uncertainties for the two cities that have been the focus of most studies, we expect uncertainties for other in- dustrial cities and towns, many of which have not been studied at all, to be even greater.

Domestic Fuel Combustion

Fuel combustion from domestic sources is a major cause of pollution in India. Although health risks from these sources appear to be greatest in rural areas, there are significant emissions in cities as well. Smoke emissions from burning wood, coal, cattle dung, and other biomass fuels are a significant source of indoor PM in many cities, although these sources can also contribute to outdoor PM. The combustion of biomass and coal is usually incomplete and often occurs in simple stoves, which are either small pits or open clay boxes. The resulting emissions contain large quantities of PM, carbon monoxide, and unburned hydrocarbons. Emissions from biomass combustion also contain a large number of polyaromatic hydrocarbons, such as benzo(a)pyrene, that are mutagenic and carcinogenic. In addition to these organic substances, coal and kerosene smoke also contains SOx and trace metals.

Besides biomass, coal, kerosene, and liquefied petroleum gas (LPG) are the main fuels for domestic uses such as cooking and heating. From a health viewpoint, however, it is the use of solid-biomass-based fuels that are burned inefficiently and vented in close proximity to people that is the cause for greater concern. In this section, we examine emissions from the domestic sector in urban India and estimate the range of total emissions of PM. The general question of the effect of indoor air pollution on the entire Indian population, particularly in rural areas, has been studied by Smith and coworkers (Smith 1987, 1994, 1999; Parikh et al 1999).

Household fuel usage is the key determinant of domestic air pollution, with fuels such as dung and wood that are lower on the “energy ladder” being signif- icantly more polluting than modern clean-burning fuels such as LPG and elec- tricity. Intermediate fuels such as kerosene are less polluting and more efficient than biomass-based fuels. PM₁₀ and other emissions from biomass fuels can be two or more orders of magnitude higher than those for modern fuels for the same level of end-use energy provided. The use of more efficient smokeless stoves and better ventilation mechanisms can reduce the levels of emissions and

exposure. However, although improved stoves help mitigate pollution exposures, they do not eliminate them. In the long term, only clean fuels can help eliminate exposure.

Determining the amounts of different fuels used domestically in Indian cities and the extent of biomass usage is critical to the understanding of health implica- tions of air pollution. We proceed to do this in the following section.

Quantitative Estimates of Domestic Fuel Usage Roughly three-quarters of all Indian households use unprocessed biomass as their primary fuel, mostly for cook- ing but with small amounts used for space heating (International Institute for Popu- lation Sciences 1995). Of these households, >90% use wood or animal dung as their primary fuel. Although the majority of rural households use unprocessed biomass fuels, there is significant household usage of biomass in cities, too, mostly restricted to the large numbers of urban poor.

There are a number of studies that examine the form of energy consumption in cities as a function of economic class. De Koning et al (1985) estimated that 100% of the urban poor in India use wood fuels, cattle dung, and crop residues for domestic cooking, although later studies provided lower estimates. A study by Raiyani et al (1993a) of a lower-income area of Ahmedabad found that 60% of the study population used biomass fuels, 20% used coal, and 15% used kerosene. The URBAIR study in Mumbai (Shah & Nagpal 1997) assumed that 20% of the urban poor used firewood and 70% used kerosene while the remaining 10% used LPG. While the above-described studies show large variations, other studies that sys- tematically characterize the use of different types of domestic fuels as a function of economic class show a more nuanced variation (Reddy 1997, Kulkarni et al 1994). From a health perspective, differentiating the use of fuels for different economic groups is critical, since indoor domestic pollution leads to much greater exposure per unit emission than other sources (Smith 1993). In this paper we have divided the population into three economic groups (low income, middle income, and high income). We have used evidence from the literature to arrive at total usage for

each group and for each fuel type.

We assume that 40%–50% of the population falls in the low-income cate- gory, 30%–40% falls in the middle-income category, and the remaining 10%–30% falls in the high-income category. The numbers for the low-income category are consistent with numbers for “slum” populations of between 40%–45% for many Indian cities. The numbers for high- and middle-income groups are consistent with projections of 150 million–200 million for the middle-class population in India. Assuming that three-quarters of the members of this middle class live in cities, roughly half of the urban population of 300 million could be classified as belonging to the middle class and above.

Table 9 shows estimates by Reddy (1997) and Kulkarni et al (1994) which suggest a systematic pattern for fuel use based on economic class. Higher-income classes tend to move up the energy ladder and prefer cleaner, more efficient fuels

TABLE 9 Summary of usage by fuel type for different income classes in Indian cities

% Households using fuel type (income classa)

aL, low income; M, medium income; H, high income.

bIncluding coal, dung, and waste.

such as LPG and electricity. Those in the lower economic groups primarily use wood for cooking and kerosene. Middle-income groups tend to use a combination of kerosene and LPG and small amounts of wood fuel (Banerjee et al 1999). Fur- thermore, compared with cooking, a greater fraction of the fuel used for heating water is biomass-based fuel. The use of coal for domestic purposes is generally small, with ranges of 5%–10% in a number of studies. Fuel usage within different economic groups differs in two ways. First, as is evident in Table 9, the fractions of households using different fuel types in the three economic classes differ. Second, the amounts of each fuel type used in the three economic classes also differ. For example, wealthier households tend to use larger quantities of “clean” fuels such as kerosene than those poor households that can afford to use some amount of that fuel.

We calculate usage of fuel by income class using per-capita emissions (averaged over all income classes) and the relative levels of usage among different income classes gleaned from the literature (see Table 10). Per-capita consumption of fire- wood has been estimated in some studies. Shah & Nagpal (1997) estimated usage of 180–200 kg/person/year for low-income groups, which is also the assumption

TABLE 10 Annual per-capita fuel usage for income classes in urban areas

Income class Biomass (kg) Coal (kg) Kerosene (liters) LPG (kg)

High income —a — 15–20 35–45

a—, Negligible amount.

used in this study.2 We also assume that the use of firewood for middle-income groups is half that of low-income groups (60–70 kg/person/year) and that firewood usage by high-income groups is negligible. This is consistent with observations in a number of cities, including Bangalore (Reddy 1997) and Hyderabad (Alam et al 1994). Coal use data for Mumbai and Delhi were extrapolated using data from other cities. Data from Ahmedabad (Raiyani et al 1993a) indicate that an average household of six consumes roughly 2 kg of coal/day. We assume that a small proportion (10–20%) of the low-income population in Delhi and Mumbai uses coal, resulting in an annual per-capita usage of 15–25 kg/person/year for that group; furthermore, we assume that 5–10% of the middle-income group uses coal, resulting in a usage of 7–12 kg/person/year. It is assumed that high-income groups do not use any coal.

The total consumption of kerosene in Mumbai was estimated in 1992 at 550,000 kl/year or 44 liters/person/year assuming uniform usage of kerosene in the popula- tion (Shah & Nagpal 1997). This is roughly in accord with the 50-liters/person/year normative assumption but is significantly higher than the 12-liters/person/year usage rate for all of India (Parikh et al 1999). The usage of kerosene follows an inverted-U pattern. While low-income groups tend to use small amounts of kerosene and much larger amounts of biomass, middle-income groups substitute kerosene and some LPG for the same and use smaller amounts of firewood. High- income groups mostly use LPG and electricity, with small amounts of kerosene. In this work we assume average usages of 20–25 liters/person/year (low income), 30– 35 liters/person/year (middle income), and 15–20 liters/person/year (high income). LPG is the preferred fuel for those who can afford it. High-income groups use LPG and, to some degree, electricity, which is used almost exclusively for cooking and water heating. Low-income groups, on the other hand, use very small quantities of LPG. The average per-capita usage of LPG in Mumbai was

assumed to be between 20 and 25 kg/year.3 Based on this assumption, the average

2Ravindranath & Ramakrishna (1997) estimate a range of 200–300 kg/person/year for cooking with fuel wood (two meals per day), the variation being accounted for by differences in stove efficiencies. Alam et al (1994) estimate average fuel wood use in Hyderabad to be 14 kg/person/year in 1994, down from 67 kg/person/year in 1984.

3The total domestic consumption of LPG in Mumbai in 1993 was about 300,000 metric tons/year or about 23 kg/year/person.

usages for the three income classes were determined to be 5–10 kg/person/year (low income), 20–25 kg/person/year (middle income), and 35–45 kg/person/year (high income).

Quantitative Estimates of Domestic Particulate Emissions The devices used for burning different fuel types have very different characteristics. Biomass-burning stoves are thermally inefficient and emit large quantities of various pollutants, and studies have quantified the emissions from these stoves, albeit under controlled laboratory conditions. As can be expected, various studies differ in their conclu- sions and there is a range of predicted emission factors depending on the type of fuel and stove. Ahuja et al (1987) determined emission factors of unvented metal stoves burning firewood. These values ranged from 1.1–3.9 kg/metric ton of wood, which are similar to results obtained by Joshi et al (1989). We assume a conservative range of 4–6 kg/ton for biomass. For coal, kerosene, and LPG, we used emission factors of 1–2, 0.5–0.8, and 0.05–0.1 kg/ton, respectively.

Annual emissions in Mumbai and Delhi for each fuel category and for each income class were calculated using the total annual consumption of fuel in each income class and the relevant emission factor. The results are shown in Table 11. Not surprisingly, emissions from biomass usage dominate the total emissions from the domestic sector. There are also sharp differences in the levels of pollutants emitted for different economic classes—for instance, high-income groups, who can afford to use clean-burning fuels, use them. Their usage leads to lower net domestic emissions of PM. Those who cannot (i.e. low-income groups) continue to use biomass-based fuels that result in larger emissions. Middle-income groups tend to use some amounts of traditional biomass-based fuels and thus face some potentially large exposures.

These results, however, must be treated with caution. The domestic-fuel sector in cities is undergoing rapid changes, and several of the studies we used to assess fuel usage were carried out over the past decade; hence, the numbers for biomass usage are likely to be overestimates. As Alam et al (1994) point out, there has been a rapid downward trend in biomass use in Indian cities as households have shifted to using less biomass and larger amounts of cleaner-burning fuels. At the same time, the demand from the nonhousehold sector—bakeries, crematoria, and wedding halls, as well as small-scale industries—has compensated for this decrease. Often these businesses are in densely populated neighborhoods and probably contribute significantly to poor air quality. However, emissions from these sources and re- sulting human exposures have, for the most part, not been studied.

A potentially large and virtually undocumented source of PM₁₀ is the burn- ing of household waste, leaves, and garbage in city streets, as well as burning in municipal waste dumps. The emission factors for such sources are likely to be extremely high, with estimates of 8–37 kg/ton (WHO 1993, Semb 1986). Burn- ing of waste in neighborhoods also creates potentially high-level local exposures. Furthermore, burning of garbage in large government waste dumps can create high particulate levels over large areas. Measurements of roughly 2000 µg/m³

AIR POLLUTION IN URBAN INDIA 651

TABLE 11 PM₁₀ emissions from domestic fuel usage by income class in Mumbai and Delhi

Domestic fuel usage (tons/year) in city by income classa

aRanges capture one standard deviation around the mean of the distribution. Total emissions for Mumbai are 5500–8000 tons/year and for Delhi 3900–6000 tons/year.

b—, Negligible usage.

TABLE 12 PM10 emissions from different source categories for Delhi and Mumbaia

Source Delhi Mumbai

aData are thousands of metric tons per year (mean ± SD).

have been made around a 100-hectare garbage dump in eastern Mumbai. The URBAIR (Shah & Nagpal 1997) study estimated that waste disposal results in

∼4000 tons of emissions/year in Mumbai, which is roughly the same order of magnitude as indoor biomass emissions. Given the proximity of these sources to people, their impact on human health could be very large. Shah & Nagpal (1997) also estimated firewood consumption in Mumbai by bakeries and crematoria to be ∼160,600 tons/year and 32,120 tons/year, resulting in PM emissions of 400 and 800 tons/year, respectively. This level of consumption is of the same order of magnitude as domestic-fuel consumption.

Table 12 summarizes the contributions by various sources to particulate emis- sions in tons per year. It is clear that substantial uncertainties exist in such char- acterizations, and this is reflected in the large error bars for the estimates. For both Delhi and Mumbai, the power sector and industrial sources are the biggest contributors while biomass combustion is the smallest contributor. However, as we shall see in the next few sections, the total contribution by a given source is a misleading indicator of its impact on human exposures.

HUMAN EXPOSURES TO PM₁₀

In the previous sections, we described the emissions of PM₁₀ from different source categories into the atmosphere. However, the relevant quantity from a human health standpoint is the amount of this airborne material that is actually breathed in by human beings—in other words, the personal exposure. We begin our analysis of exposure issues by ranking the sources of pollution in Mumbai and Delhi in terms of human exposure that results from these emissions. We use the “exposure effectiveness” concept developed by Rowe (1985), Ramousset & Smith (1990), and Smith (1993) that helps rank the different sources of air pollution for different income classes. This is followed by a review of air quality as it relates to PM and its constituents in Delhi and Mumbai. Air quality measurements performed outdoors are only part of what determines personal exposures to PM. Evaluation of personal exposure requires an empirical assessment of domestic, outdoor, and occupational

exposures. To this end, we perform a quantitative analysis of exposure to PM among different income classes.

Ranking the Overall Societal Impact of Sources

Exposure effectiveness of a source provides a way to calculate the overall societal impact of the source’s PM emissions. It comprises three components: the concen- trations of PM, in the immediate vicinity of people, that are a direct result of unit emissions from that source; the total number of people that are affected; and the length of time that people are exposed to pollutant concentrations. Mathematically, exposure effectiveness can be expressed as

EE(Kilo − Exposure Units/ton) =

PM(in µg/m³) × number exposed (in person) × Length of exposure (in years) Source emissions (in tons)

2.

In general, calculations of exposure effectiveness require detailed modeling of the specific physical pathways by which particulate emissions find their way to the vicinity of those who are potentially at risk of exposure. For example, deter- mination of the exposure effectiveness of a particular power plant in New Delhi would require a detailed model of emission characteristics (e.g. stack height and emissions profile), meteorology (e.g. wind speed and direction), and local condi- tions and demographics (e.g. population density). However, such detailed analyses can be time-consuming and data intensive. An alternative is to perform order- of-magnitude assessments based on simplifying assumptions that are nonethe- less appropriate for the task of ranking and comparing exposures from different sources. For example, it may be possible to adapt exposure effectiveness mea- sures developed for one context for use in another by using appropriate scaling procedures.

In Table 13, we provide estimates of exposure effectiveness of source categories as derived in the literature. We also provide a range for the numbers we have used in this study. We derived the numbers for power plant and industry emissions by using calculations for China performed by Wang & Smith (1999), but we scaled their numbers for the population densities of Mumbai and Delhi (∼50,000 and 25,000 people/km², respectively) and obtained ranges of 8000—12,000 for Mumbai and 4000–6000 for Delhi. In addition, we assumed that the exposure effectiveness of vehicular traffic is twice that of power plants and industry while that of biomass combustion is two orders of magnitude greater.

The overall impact of a source can be calculated by multiplying the exposure effectiveness of a source and the total emissions of that source. The product, which we call exposure impact, is a measure of the concentration of PM that could lead to exposure averaged over the number of individuals exposed and the time and levels of exposure. Thus, the exposure impact measurements for different sources can be compared and the relative magnitudes of their effects can be obtained. While this

TABLE 13 Exposure effectiveness (EE) of various sources

Sourcea Reference EE ( µg/m³ × persons × year//tonnes) Comment

US power plant Rowe 1985 100 Based on a study of several hundred US power plants

LDC power plant Smith 1993 1000 Based on scaling US data to LDC population density

Delhi power plant Saxena & Dayal 1997 23,800b Based on scaling US data to Delhi population density

Biomass cook stoves Ramousset & Smith 1990 150,000 Based on model calculations

China power plant Wang & Smith 1999 260 Based on China-specific model

China indoor coal cook stoves Wang & Smith 1999 12,000 Based on China-specific model LDC vehicles Smith 1993 ∼2000

Power plant and industry This study 8000–12,000 (Mumbai) Chinese data (Wang & Smith 1999) (Delhi and Mumbai) 4000–6000 (Delhi) scaled by population density

Vehicles (Delhi and Mumbai) This study 16,000–24,000 (Mumbai) Assumed to be the twice that of

8000–12,000 (Delhi) industrial sources

Indoor combustion This study 100,000–200,000 Based on Roumasset & Smith 1990, (Urban India) with added subjective range

aLDC, less developed country.

bNote that this estimate uses average US population density and the population density of Delhi for scaling. This results in a gross overestimate of exposure effectiveness since the appropriate scaling requires a comparison of the population density of Delhi with those regions in the US that are in the vicinity of power plants.

measure is by itself not a very intuitive one, it provides a useful means for com- paring the relative societal impacts of different sources. In Table 14, we provide a comparison of the relative exposure impacts of different source categories— biomass, industry, and vehicles—on different income groups. The results suggest that there are striking differences in exposure to PM between income groups, and they raise significant environmental and energy equity issues. Emissions from indoor combustion dominate the exposure for low-income groups, while expo- sure for high-income groups to the same is lower by two orders of magnitude. High-income groups face little risk from indoor combustion of cooking fuels, and the majority of their exposure is a result of industrial and vehicular emissions, which are responsible for roughly equal levels of exposure in this income cate- gory. Middle-income classes are exposed to indoor and outdoor emissions, all of which are responsible for similar order-of-magnitude effects. Overall, low-income groups face exposure levels that are a factor of two or more higher than those faced by middle- and high-income groups, due mostly to their use of cheaper biomass fu- els. Conversely, middle-income and high-income groups get significant fractions of exposure from industrial and vehicular sources.

It is instructive to compare the results in Table 14 with those in Table 12. Even though biomass combustion emissions are very small in terms of sheer mass, their contribution to human exposure is the greatest in terms of exposure effectiveness.

Ambient Concentrations of Particulate Matter

From a health effect perspective there are three characteristics of PM that are crit- ical. These are concentration, size distribution, and composition. Human beings can breathe in particles that range in diameter from a few nanometers to ≥100 µm. In the early years of air pollution monitoring in the US and the United Kingdom,

TABLE 14 Relative levels of exposure impact of PM₁₀ from source categories for populations in different income classes

Mean exposure (kiloexposure unitsa) from source by city Biomass Vehicular Industry

aObtained by multiplying net emissions from biomass, industry, and vehicular sources as calculated in this study (shown in Table 12) with exposure effectiveness for these sources given in Table 13. Exposure from biomass of low-income groups in each city scaled, without loss of generality, to 1000; that is, low-income groups in Delhi and Mumbai are assumed to have relative exposure impact values of 1000.

agencies typically measured total suspended particles (TSP). The original US National Ambient Air Quality Standard for TSP, promulgated in 1971, was 260 µg/m³ for a 24-h period, not to be exceeded more than once per year, and a 75-µg/m³ annual geometric mean. This metric came to be defined by instrument characteristics, with a concomitant lack of standardization.

Over the next 30 years, scientists began to understand that most of the larger particles are trapped in the upper airways of the respiratory tract (the extrathoracic region) and are then removed through various mechanisms. Thus, larger parti- cles are now considered to be of limited relevance to most respiratory diseases involving the tracheobronchial and alveolar regions (which, together, comprise the thoracic region) of the respiratory tract, and the focus has shifted to smaller particles. The thoracic fraction is now approximated by the PM₁₀ criterion of the US EPA (particles with a median aerodynamic diameter of <10 µm) and is closely related to thoracic criteria adopted by agencies providing international standards (International Standards Organization 1992, Comite´ Europe´en de Nor- malisation 1992). In the US, the US National Ambient Air Quality Standard was revised in 1987 with a new PM₁₀ standard of 150 µg/m³ for a 24-h period (not to be exceeded more than once per year) and an annual average of 50 µg/m³. Recent epidemiological studies in the US have indicated that an even finer partic-

ulate fraction, the PM2.5 (particles with a median aerodynamic diameter of <2.5 µm) may be more relevant to health. On the basis of the totality of evidence, the US EPA has proposed a new 65-µg/m³ 24-h ambient standard for PM2.5.

WHO4 has lagged somewhat and continues to provide guidelines for TSP lev-

els, with a 24-h-average limit of 150–230 µg/m³ and an annual-average limit of 60–90 µg/m³.

Typically, in the US, the 90th percentile of 24-h PM₁₀ concentrations ranges between 65 and 80 µg/m³, with annual averages ranging between 4 and 11 µg/m³. Studies in the US show that the PM₁₀/TSP ratio is typically equal to 0.55, yielding 24-h TSP concentrations of 120–145 µg/m³ and annual averages of 8–20 µg/m³. Big cities like Los Angeles have 24-h TSP levels of ∼180 µg/m³. By compari- son, megacities in developing countries, such as Beijing (∼250–450 µg/m³) and Mexico City (∼200–550 µg/m³), have much higher annual average TSP concen- trations (Mage et al 1996).

In India, the CPCB maintains a network of monitoring stations and collects air quality measurements in about 100 cities and towns. Table 15 is a summary of TSP concentrations in five major cities for 1993–1994. It is clear that all of these cities routinely exceed WHO guidelines for TSP levels. Figure 3 shows the monthly variation of TSP concentrations in Delhi and Mumbai for 1994.

4These guidelines were issued in 1979 (WHO 1979). More recent assessments do not include air quality guidelines due to the “fact that a threshold for the onset of health effects could not be detected.” Instead linear relationships between the PM10 or PM2.5 concentrations and various types of health effects were provided. For details on WHO guidelines for particulate matter see http://www.who.org/peh/air/airguides2.htm.

TABLE 15 Total suspended particle (TSP) concentrations in selected Indian cities during 1993– 1994a (Central Pollution Control Board, 1995)

TSP 24-h average concentrations ( µg/m³)

Industrial areas Residential areas

aSource, Central Pollution Control Board 1995.

Monthly concentrations are systematically higher in Delhi than in Mumbai. Consequently, the average annual TSP concentration is almost twice as high in Delhi (∼400 µg/m³) as in Mumbai (200 µg/m³).

Unfortunately, the utility of data gathered by the CPCB is somewhat limited because it measures TSP, which includes particles of all sizes. Although few PM₁₀ measurements are available for Indian cities, including Mumbai and Delhi, it is very likely that PM₁₀ concentrations in Indian cities are very high. For example,

800

Total Suspended Particles ( μg/m³

)

700

600

500

Delhi

Mumbai

400

300

200

100

0

January

February

March

April

May

June

July

August

September

October

November

December

Month (1994)

Figure 3 Monthly variation in total-suspended-particulate concentrations in Delhi and Mumbai for 1994. The error bars show the monthly maxima and minima.

using a PM10/TSP ratio of 0.55, which is typical for US cities, results in PM10 concentrations that are greater than the WHO limits by a factor of two for Mumbai and three for Delhi. However, this assumption, often made in studies relating to PM in India, is seldom validated. One of the few studies that actually measured this ratio in India shows a PM₁₀/TSP ratio of 0.87 to 0.89 (Sharma & Patil 1991), which would increase PM₁₀ levels even higher. While the ratio may be lower for Delhi owing to seasonal dust storms, there are, in fact, no actual measurements to document this.

The chemical composition of PM can also have a major impact on the eventual human health risks of exposure. The two major categories we consider below are metal-containing particulates and PAHs, both of which are known to have health consequences.

Heavy Metals There have been some studies measuring heavy-metal concen- trations in the atmosphere in India, although few have focused specifically on Mumbai and Delhi. Tripathi et al (1989) measured concentrations of Pb, Cd, Cu, and Zn in Mumbai and Moradabad in the atmosphere as well as in the blood and teeth of residents. Bandhu et al (1996) measured metal concentrations in Chandigarh, and Tripathi (1994) measured airborne lead levels in Varanasi. These values are compared with airborne levels of metal concentrations in selected US cities in Table 16. While the US studies date from the early 1980s, with the excep- tion of that of Riverside, CA, which was conducted in the early 1990s, the Indian studies are more recent, although fewer in number and more scattered. In general, concentrations of most metals are higher in Indian cities than in the US, sometimes by several orders of magnitude, as for Cd, Zn, and Pb.

Many airborne metals are known to be carcinogenic, and WHO has developed lung cancer risk values for some of these metals. For example, the lifetime risk from breathing 1 µg of arsenic/m³ is 1.5 × 10⁻³, while for nickel it is 3.8 × 10⁻⁴. While the acute and long-term health effects of exposures to airborne-metal aerosols are well known from occupational-exposure studies, recent studies in the US have led to the hypothesis that first-row transition metals (As, Fe, Ni, V, Cu, and Zn) may be causal agents for increased risks in ambient settings as well. These metals, which exist both as soluble salts and complexed to insoluble components of PM₁₀, produce free radicals in the lungs. Production of free radicals results in lung injury, inflammation, and alterations in pulmonary host defense. If this hypothesis is true, then it is important to control metal concentrations on a regular basis in urban settings.

Polycyclic Aromatic Hydrocarbons PAHs are formed by the incomplete com- bustion of fossil fuels. The primary sources for urban PAHs are vehicular emis- sions, industrial oil combustion, and cooking fuel (kerosene, coal, wood, and other biofuels such as dried cow manure) combustion. Although they are present both in particle and in gas phases, the particle-phase PAHs are more toxic to human health. Over 100 PAHs have been identified in urban air in India, although much

AIR POLLUTION IN URBAN INDIA 659

TABLE 16 Comparison of selected metal geometric mean concentrations (24-h averages) in Indian and US cities. Data sources: (a) Tripathi et al (1989); (b) Bandhu et al (1996); (c) Tripathi (1994); (d) Krishnamurti and Viswanathan, 1991; (e) Saltzman et al (1985); (f) Clayton et al (1993)

attention is focused on a few selected ones, such as benzo(a)pyrene, fluoranthene, pyrene, benzo(a)anthracene, and chrysene (Venkataraman 1998). Source appor- tionment studies have shown that the relative contributions from each of these sources vary geographically as well as seasonally. Automobiles are the biggest contributor of PAHs in New Delhi, while automobiles and industry appear to be significant contributors in Mumbai.

Mohan Rao et al (1982) and Aggarwal et al (1982) measured PAH levels to evaluate their carcinogenic risk. Studies in Mumbai, New Delhi, Ahmedabad, and Nagpur have shown that PAH levels in Indian cities are 10- to 50-fold higher

than those reported internationally (Venkataraman 1998).5 Raiyani et al (1993b)

reported that total PAH concentrations in Ahmedabad ranged between 90 and 195 ng/m³ for industrial areas and between 20 and 70 ng/m³ for residential areas. Annual average total PAH concentrations in four locations in New Delhi dominated by vehicular traffic range from 150 to 1800 ng/m³ (Ravi Shankar 1990 as reported by Venkataraman 1998). Reported concentrations in Mumbai range from 20 to 95 ng/m³ (Pandit et al 1996), while those for Nagpur range from 125 to 190 ng/m³, with over 70% of the mass in particles <10 µm in diameter.

WHO provides estimates of cancer risks from exposure to PAHs. WHO es- timates that the lifetime risk of lung cancer from breathing 1 ng of the PAH benzo(a)pyrene/m³ is 8.7 × 10⁻⁵. Smith & Liu (1994) have calculated lung can- cer mortality in rural areas of developing countries, using this WHO estimate. To our knowledge, there have been no studies that specifically examine PAH expo- sures in the Indian context.

Personal Exposure to Particulate Matter

Personal exposures to PM are calculated using the following expression for time- weighted averaging:

Σm C_i t_i

i =1

E = Σ_m t , 3.

i =1 i

where C_i is the PM10 concentration in the ith microenvironment and t_i is the time spent in that microenvironment. The relevant microenvironments are indoors dur- ing cooking, indoors while not cooking, outdoors, and occupational environments. Evaluating the time-averaged exposure, E, requires a knowledge of the time spent by individuals in the relevant microenvironments and the concentrations of the particles in these microenvironments. We did this calculation for representative individuals in each economic group since different economic groups can have con- siderably different exposures. The exposure calculated using Equation 3 is only a model representation of the true level of personal exposure. In reality, personal

5Kamens et al (1990) have found that the half-lives of particle-bound PAHs increase dramat- ically from about1h to several days in the presence of sunlight. This may have significant implications for Indian conditions with abundant sunshine most of the year.

exposures are best evaluated by measuring PM concentrations over 24-h periods with sampling devices that are attached to the individual. Very few studies have actually performed such measurements in India.

Time Activity Patterns Time activity patterns help in determining the fraction of time people spend in different environments. In the US, people typically spend 90% of their time indoors (Sexton & Ryan 1988). There are no similar studies reporting time fractions spent indoors in India. However, there is some evidence that people in India also spend a majority of their time indoors. Cooking, in particular cooking that uses biomass fuels, alone can take from 3–6 h per day (Smith et al 1983). Women, who do the majority of the cooking, are more exposed to particulates from indoor combustion (Parikh et al 1999). Infants, young children, and older family members are also more likely to be exposed to higher indoor levels of PM. Smith (1993) proposed using the fraction of the urban population living in slum housing as an indicator of time spent outdoors in urban areas, with the caveat that the immediate outdoor surrounding is likely to be more polluted than areawide averages might indicate. His data suggest that people in urban areas in developing countries like India spend 80% of their time indoors and only 20% of their time outdoors. Kulkarni & Patil (1999) estimate that time spent outdoors for nonoccupational purposes is 1–2 h/day (i.e. ∼10%).

In Table 17 we provide estimates of the amount of time spent by the rep- resentative individuals in three different income classes in indoor, outdoor, and occupational environments. In each case we assume that the individual is exposed to indoor emissions from cooking, as well as outdoor emissions and emissions in occupational settings. Thus, the individual best classified as the most exposed individual is the working woman, who has housework duties at home but also a full-time job outside the house.

PM₁₀ Concentrations in Indoor, Outdoor, and Occupational Environments Evaluation of personal exposure requires the estimation of PM concentrations and time spent in these microenvironments for low-, medium-, and high-income groups. PM₁₀ concentrations in some microenvironments, such as indoor environ- ments for low-income groups, have been studied in depth (Smith 1993, Kulkarni & Patil 1999). Furthermore, outdoor environments, for which similar exposure levels occur in all income groups, have also been well studied, since most air pollution monitoring is focused on outdoor measurements and epidemiological studies focus on relating these outdoor levels to mortality and morbidity effects. However, indoor residential environments (during noncooking periods) and occu- pational environments (which may be similar to indoor residences for middle- and high-income groups but very different for low-income groups) have not been stud- ied. In this section, we review the literature on concentrations of PM in various microenvironments in India and use the review as a basis for suggesting ranges for PM concentrations for the three income classes. These ranges are shown in Table 17.

TABLE 17 Exposure levels of 24–h average PM₁₀ and time spent in different microenvironment settings as a function of income class in urban Indiaf

Indoor ( µg/m3) Outdoor ( µg/m3)

Occupational

Income class Cooking Noncooking Delhi Mumbai ( µg/m³)

Low inc

aRaiyani et al (1993b). bRaiyani et al (1993b). cKulkarni & Patil (1999).

dCPCB, Pollution Statistics (1994).

eOccupational concentration set equal to noncooking indoor concentrations.

fAll exposure ranges, except outdoor exposures, were obtained from the referenced papers as PM10. Outdoor exposure numbers obtained as TSP concentrations from CPCB were converted to PM₁₀ equivalents using a PM₁₀/TSP range of 0.55–0.9.

Urban indoor PM concentrations are significantly affected by indoor as well as outdoor sources. Air exchange rates in Indian homes6 are relatively high, ranging from 4 to 10 h⁻¹ (calculated from data in Smith et al 1983 and Ahuja et al 1987). However, even these high air exchange rates are not sufficient to keep PM concentrations low in small houses. Furthermore, cooking areas themselves are usually poorly ventilated, and about half the households in cities do not have a separate kitchen. Thus, typical concentrations on the order of 5000–7000 µg/m³ are seen during cooking with biomass fuels. This is not surprising since wood and coal combustion in residences produces 300- and 100-fold more TSP, respectively, than natural-gas combustion for the same amount of energy (De Koning et al 1985). In a study of personal exposures to PM in western India, Smith et al (1983) found PM levels reaching >55,000 µg/m³ and averaging ∼7000 µg/m³. The same study reported BaP levels of 400 ng/m³. Aggarwal et al (1982) found similar indoor levels in Ahmedabad. In a more recent review, Smith (1993) reported that short-term (15-min) indoor PM levels due to biomass combustion range from 4000 to 21,000 µg/m³. In urban households in China—which has economic conditions not too dissimilar from those of India—a number of indoor coal smoke studies have shown 12- and 24-h PM concentrations ranging from 270 to 2800 µg/m³

(Smith 1993).

In addition to high indoor particulate levels, residential biomass fuel combustion leads to high ambient concentrations. Using fuel use rates, population densities,

6Air exchange rates in residences in the US range from 0.2 h−1 for tightly sealed homes to greater than 2.0 h−1 for homes with open windows (Wilson & Suh 1997). Two studies give a median of 0.5 h−1 for the US (Koontz & Rector 1995, Murray & Burmaster 1995).

and the existence of limited mixing heights in a simple equilibrium model, Smith et al (1981) have calculated ambient concentrations. For example, one semiur- ban area had an ambient concentration of 600 µg/m³ during an inversion period in the evening (Smith et al 1983). Although these calculations were done for a nonurban setting, they provide a measure of the potential contributions of biomass combustion to urban air pollution.

Nearly all the particles from indoor combustion are <10 µm in aerodynamic diameter and are thus able to penetrate to the lung. Size distribution measurements of particulates from biomass combustion show that a majority of these particles (96%) are <9 µm in aerodynamic diameter and are therefore almost identical to PM10 (Raiyani et al 1993a), implying that the PM10/TSP ratio is close to unity for biomass combustion.

The above-described studies suggest a systematic pattern of high indoor PM₁₀ concentrations during cooking in low-income households, which tend to use biomass fuels. Table 17 provides a summary of concentration and time activ- ity assumptions for different microenvironments and income classes. High- and middle-income homes tend to have far lower concentrations. Based on the lit- erature, we have assessed concentrations of 1000–4000, 350–600, and 200–300 µg/m³ for low-income, medium-income, and high-income households, respec- tively, during cooking. During other periods of indoor exposure (noncooking hours) we assume ranges of 650–800, 200–300, and 200–300 µg/m³ for low- income, medium-income, and high-income households, respectively. The range for outdoor concentrations is easier to justify because concentration data are more widely available. For all income classes, we suggest an outdoor concentration range of 200–700 µg/m³. This is consistent with the range for monthly annual

concentrations of PM measurements shown in Figure 3.7 Limited studies have

indicated that nonoccupational and occupational environments contribute equally to the daily personal exposures of people in a range of occupations, such as small- scale industrial workers working in the open, shopkeepers, traffic constables, and telephone booth operators (Kulkarni & Patil 1999). We used this finding in our estimates of occupational exposures.

Personal Exposure Calculations The concentration and time activity data shown in Table 17 were used in conjunction with Equation 3 to calculate personal exposure levels for the three income groups. The calculated daily personal exposure levels of PM10 for Delhi are 750–1050 µg/m³ for low-income groups, 260–310 µg/m³ for middle-income groups, and 250–290 µg/m³ for high-income groups. The cal- culated daily personal exposure levels in Mumbai are marginally lower because outdoor concentrations in Mumbai are lower than those in Delhi. These levels of ex- posure in both cities are significantly higher than exposure levels in industrialized

7We are interested in a range of PM concentrations when exposure is the highest, i.e. during daylight hours when individuals tend to be outdoors. It is also during this time that emissions from vehicular traffic are the highest.

countries such as the US. For example, in a southern California study, Clayton et al (1993) reported outdoor concentrations of 91 ± 7 µg/m³, residential concen- trations of 95 ± 6 µg/m³, and daily personal exposure levels of 150 ± 9 µg/m³. There are only a few studies against which our estimates can be compared.

Kulkarni & Patil (1999) calculated personal exposures of low-income groups in Mumbai to be 870 ± 455 µg/m³. These exposure levels are roughly in the same range as our estimates. Saxena & Dayal (1997) used a model to estimate PM exposures in urban nonslum populations (roughly corresponding to middle- and high-income groups) to be 150 µg/m³ and exposures in slum populations to be 375 µg/m³, values which are considerably smaller than our estimates.

In a recent study of the middle-class population in Delhi, Shukla (1999) found high levels of daily personal exposure to pollutants. The study focused on an East Delhi suburb, in the vicinity of industrial sources and highways, and measured personal exposure levels for different members of the middle-class population. The daily average TSP exposure of housewives was found to be 590 µg/m³, compared with 440 µg/m³ for infants. The study also determined that average daily TSP exposure levels for school and college students and office employees were 700, 860, and 660 µg/m³, respectively. While the study provides an indication that exposure levels are high even in middle-class settings, it measured particles of all sizes and not PM₁₀ fractions. Unfortunately, there have been no studies in Delhi which have measured the relationship between TSP and respirable fractions such as PM10. Assuming a TSP/PM10 ratio of 0.55 results in exposure levels of 300–400

µg/m³ for different categories of individuals in the population, which are slightly

higher values than the range we calculated.

Overall, the few available field study data and our own calculations suggest that personal exposure levels in the Indian population are high. The exposure levels increase with decreasing income, with the poor facing the greatest levels of exposure. Personal exposure levels for low-income groups are extremely high— over an order of magnitude greater than WHO standards. Personal exposure levels in middle- and high-income groups are higher than these same standards by factors of three or more. From a public health perspective, such high levels of personal exposure can have significant negative consequences.

EPIDEMIOLOGY AND EVALUATION OF HEALTH RISKS FROM EXPOSURE TO PARTICULATE MATTER

A number of diseases have been associated with inhalation exposure to airborne PM: respiratory disorders whose effects range from minor symptoms such as coughs and dyspnea to severe ones such as acute respiratory infections (ARI), asthma, and pneumonia, chronic obstructive lung diseases such as bronchitis, car- diovascular disease, tuberculosis, lung cancer, and blindness. In addition, perinatal effects such as stillbirths and low birth weights are also associated with air pollu- tion. However, the health end point that is most clearly defined is death, and many

epidemiological studies in developed countries focus on obtaining relationships between mortality rates and ambient levels of pollution.

High levels of chronic morbidity exact their own toll and pose severe strains on the health care infrastructure. For instance, one study estimated that the incidence of respiratory diseases in Delhi is 12-fold higher than that for the rest of the coun- try. A preliminary study of a middle-class population in East Delhi also found that 23% of the population suffered from severe respiratory disorders and 54% of the population suffered from some form of respiratory disease (Pandey 1998). The study also found that the sale of drugs that help combat respiratory disease, like citrizine (an antihistamine), salbutamol (a bronchial dilator), bromzine (mucolac- tice; liquifies sputum), amoxicillin, and erythromycin (antibiotics for respiratory tract infections), is increasing at the rate of 20% per year, which is much faster than the current rate of population growth.

As for other parts of the causal paradigm (Figure 1), very few epidemiological studies have been conducted on the health risks of exposure to PM in Indian cities. Much of the science of exposure and effects assessment has been developed in industrialized countries, particularly the US. This understanding is routinely relied upon for making extrapolations to other contexts, such as India, where the science is less well studied. For example, several assessments of mortality from air pollution in India use simple reduced-form expressions, derived from US studies, linking mortality to PM concentration. There are good reasons to believe that the science developed in one context may be applicable in others. Certainly the prevailing scientific view is that epidemiological studies performed in regions with similar levels and types of exposure, demographics, and statuses of public health result in findings of similar effects and human mortality in these regions. However, the differences between urban India and industrialized settings like the US are large enough that extrapolations of US findings to India are likely to produce misleading analyses. This could result in errors in aggregate calculations of human mortality, in the magnitude of health risks faced by specific vulnerable groups, and ultimately in analyzing the ways in which exposure reduction can be achieved.

In this section, we provide a synthesis of the science as it relates to air pollution studies in India. We critically evaluate past studies and provide a summary of estimates of health risks (mortality and selected indices of morbidity) from past studies as well as our own estimates.

Confounders

Tobacco smoking, occupational exposures to air pollution, health status, and copol- lutants can be significant confounders in any epidemiological study involving lung disease. We thus begin with a discussion of some common confounders in epi- demiological studies of air pollution as they relate to India.

Tobacco Smoking WHO estimates that in 1997 India produced more than 750,000 million manufactured cigarettes, accounting for 13.5% of the world total

(WHO 1997). In the same year, 7.0% of our world’s total tobacco was produced in India, making it the world’s third largest tobacco-growing country. The most common manner of tobacco consumption in India is smoking bidis. Bidis are unfiltered smokes that do not use paper but are rolled into a leaf from the native tendu vine. Bidis account for ∼40% of tobacco consumption in India, while 20% of the total tobacco consumed in that country is in the form of cigarettes and the rest is divided among other tobacco mixtures.

Consumption patterns of tobacco show major differences across regions. In general, bidis are more harmful to human health than cigarettes; they have much higher tar levels (45–50 mg) than cigarettes (18–28 mg), and their use has risen substantially during the last three decades. Cigarette smoking increased until the 1970s and remained stationary or declined somewhat during the 1980s. Other forms of tobacco use have declined considerably over the years. It is estimated that 65% of all men use some form of tobacco (∼35% smoke regular tobacco, 22% use smokeless tobacco, and 8% use both). However, the overall prevalence of bidi and cigarette smoking among women is ∼3%. Smoking rates also tend to be higher in rural areas than in urban areas.

Clinical observations in some areas have revealed that >60% of heart disease patients <40 years of age are tobacco users; over half of the patients aged 41–60 are also smokers (WHO 1997). Tobacco is responsible for a significant amount of morbidity and mortality among middle-aged adults in India. Tobacco-related cancers account for about one-half of all cancers among men and one-fourth among women. India has one of the highest rates of oral cancer in the world, and the rate is still increasing. Oral cancer accounts for one-third of the total cancer cases, with 90% of the patients being tobacco chewers. As in other parts of the world, the use of tobacco is also a significant confounder in assessing the impacts of air pollution on health. For India, however, the percentage of women who smoke is low, making it easier to assess the impact on that population.

Occupational Exposures to Particulate Matter A significant fraction of those in lower-income groups work in manufacturing and production industries (such as cement, coal, iron and steel, asbestos, and grain processing) as well as small-scale operations (welding, metal plating, etc) where they are exposed to high levels of PM. In urban India, typically the workforce in such sectors lives close to the work place and is exposed to ambient emissions from these industries as well. Occupational exposures to a wide variety of airborne dusts, gases, and fumes (such as grain dust, wood dust, and various metal fumes [e.g. nickel, chromium, and cobalt]) can cause asthma—a common health outcome studied in air pollution epidemiology. Other respiratory diseases arising from occupational exposures that could be misattributed to ambient exposures are pneumoconiosis (from coal dust), pulmonary fibrosis (from silica or asbestos exposures), and lung cancer (e.g. from a variety of heavy metals).

In the well-known Bombay Air Pollution Health Study (Kamat 1984), four com- munities in and around Bombay were chosen as “urban high”, “urban medium,”

“urban low”, and “rural” based on pollution levels. Despite the “urban medium” area having lower SO₂ and PM levels, it had higher morbidity for common colds, intermittent coughs, and dyspnea. This may have been attributable to a higher fraction of this population being occupied in dusty trades (in addition to hav- ing a higher fraction of children under age 5 years, a segment of the population considered more susceptible). In this work, we attempt to incorporate a measure of occupational exposure into calculations of total exposure (Table 17), although more fine-grained analyses of occupational exposure are needed to isolate its role.

Health and Socioeconomic Status While socioeconomic factors are also impor- tant confounders, very few studies have looked at them systematically. Again, the study by Kamat (1984) provides some evidence of the importance of these factors. Of the four communities studied, the rural area had the lowest level of outdoor air pollution, but it showed an intermediate degree of morbidity. The villages studied typically had no sanitation, no protected water supply, poor housing, poor nutrition, widespread intestinal parasitism, and poor quality of medical care. These factors, over a long period of time, may have accounted for the poorer health status and lung function of the residents. Mishra et al (1997) showed that persons living in households with a separate kitchen had lower risks of tuberculosis than persons liv- ing in houses without a separate kitchen. Educational levels were also found to be strongly linked to tuberculosis prevalence. These findings are unlike observations from studies conducted in the West (Mostardi & Leonard 1974, Ferris et al 1979); those studies show no effect of socioeconomic status on lung function or other measures of health status.

A number of authors, however, argue that exposure to PM is an important determinant of mortality, even when socioeconomic status is taken into account. Globally, ARI is a major cause of infant mortality, killing 4.3 million children per year (WHO 1992). ARI is the single largest disease category in India, accounting for one-eighth of the national disease burden (Smith 1999). In a Brazilian study, Penna & Duchiade (1991) observed statistically significant associations between average annual levels of PM and infant mortality from pneumonia after controlling for socioeconomic factors such as family income level. Researchers in Kerala have also found an association between pneumonia, a number of socioeconomic variables, and air pollution (Shah et al 1994). Smith (1993) summarized studies conducted in five different lesser-developed countries which indicated that the relative risk for severe ARI from smoke exposures might be in the range of two to six. Clearly, socioeconomic status and exposure to air pollution are heavily linked in India. In this work, we analyze this linkage by studying the effects of income on exposure and, consequently, on health.

Copollutants A number of studies in the US have found that the observed health effects of PM are confounded by other pollutants commonly occurring in community air, such as SO₂, NO₂, O₃, and CO (Samet et al 1995, Moolgavkar et al 1995, Moolgavkar & Luebeck 1996). The contention is that these pollutants,

acting in concert, cause the reported health effects, rather than PM acting alone. If there are high correlations and synergisms between concentrations of differ- ent pollutants, it becomes difficult to separate the effects of any one pollutant from those of any other. While the amount of confounding is still unclear, there are no systematic measurements of these pollutants (especially O₃) by the CPCB or National Environmental Engineering Research Institute in Indian cities. Addi- tionally, measurements of SO₂ and TSP in some cities show opposing trends: in Mumbai, SO₂ levels have been decreasing since 1978 from levels of ∼100 µg/m³ to ∼30 µg/m³ in 1996. Corresponding TSP levels have either remained constant or increased in suburban Mumbai over the same period.

Premature-Mortality Estimates

A survey of the literature shows that very few systematic studies have been con- ducted to evaluate the health risks of exposures to PM and, more generally, air pollution in urban India. Two broad approaches to studying the health effects of PM in developing countries can be identified in the literature. These methods rely on both (a) the findings from and (b) the methodologies used for studies in devel- oped countries. However, exposure patterns and the composition of air pollution in India are quite different from those in Western industrialized countries, and hence studies conducted in the West may be of limited value for assessing health effects in India. There is a need for methods tailored for Indian patterns of pollution and population, similar to those devised for China (Florig 1997) or other developing countries. In this section, we present estimates of premature mortality, obtained using various assumptions (see also Table 18), and critically assess the various approaches used to obtain these estimates.

Studies That Use Dose-Response Relationships from Developed Countries This approach evaluates health impacts using dose-response functions developed from epidemiological studies done in a few developed countries, chiefly the US (Pope et al 1992, Schwartz et al 1996, Dockery et al 1993). According to these stud- ies, there is a 1% increase in total mortality for every 10-µg/m³ increase in PM₁₀ concentration. The first such exercise, conducted by Ostro (1994) for Jakarta, has been the model for similar studies of India (Brandon & Homman 1995, Shah & Nagpal 1997). The appendix lists the dose-response equations developed by Ostro (1994). Brandon & Homman (1995) used these dose-response relationships to estimate reductions in mortality/morbidity in 36 Indian cities when PM (and SO₂, Pb, and NOx) levels were reduced to WHO-recommended annual average standards. Scaling these numbers to the entire urban population of India results in 150,000 excess deaths per year (see Table 18).

The assumption implicit in this approach is that the population and exposure characteristics in India and in developed countries—the baseline health status of the population, size distribution and chemical composition of PM, and exposure attributes such as the relationship between indoor and outdoor concentrations of

TABLE 18 Estimates of annual excess deaths using various assumptions

Study assumptions Exposure data Population Excess deathsa

Assumption 1:1.12% increase in mortality for every 10 µg/m³ increase in PM₁₀

Assumption 2:0.23% increase in mortality for every 10 µg/m³ increase in TSP

Cropper et al (1997) for TSP = 375 µg/m³ 9 million 5070 Delhi

aRanges capture one standard deviation around the mean of the distribution. We used time-weighted (daily average) personal exposures calculated for various subpopulations for mortality calculations. Personal exposures for subpopulations were assumed to be identical for Mumbai and Delhi. The daily average personal exposures levels of PM10 in Delhi are 750– 1050 µg/m³ for low-income groups, 260–310 µg/m³ for middle-income groups, and 250–290 µg/m³ for high-income groups. The numbers are only marginally different for Mumbai.

PM—are identical. Additionally, the spatial relationships that exist between PM monitors and the population in the original studies are similar in Indian and US cities. The only parameter that is assumed to be different is the PM concentration.

Even the very limited information that is available on these factors shows that a number of the above assumptions are unlikely to hold for the urban Indian situation. The population distributions for India and the US are very different (see Figure 4). The median and mode of the Indian population (∼24 years and

∼10 years, respectively) are lower than those for the US (∼34 years and ∼39, respectively). Additionally, the life expectancy in the US is higher than that in India, as evidenced by the longer tail to the population distribution in the US.

1.00

China India US

0.90

Cumulative age distribution

0.80

0.70

0.60

0.50

0.40

0.30

0.20

0.10

0.00

1 4 9 14 19 24 29 34 39 44 49 54 59 64 69 74 79 84 89

Age

Figure 4 Cumulative age distributions for the populations of India, China, and the US.

These differences may lead to different susceptibilities. In the US, the health effects of PM are more pronounced in the elderly population, whereas in India the same situation may be hard to detect because the comparable population of the elderly is more limited. At the other end of the population distribution, infant mortality is very high in India compared to the US, especially that attributable to respiratory illnesses. The age-dependent susceptibilities to air pollution are different for the two populations because the baseline health statuses are different.

While little is known about the exact role of differences in chemical composition in determining health effects, it is quite clear from the literature that particles with diameters of >10 µm do not significantly contribute to mortality. The issue for India is that ambient concentrations for particles of all sizes, and not for PM₁₀ alone, have been typically measured. Thus, using the Ostro relationship for India requires a translation from TSP to PM₁₀. Brandon & Homman (1995) and the URBAIR study (Shah & Nagpal 1997) assumed the size distribution, as quantified by the PM₁₀/TSP ratio, of 0.55 developed by the US EPA (1986) for US cities. In fact, there is very little in the peer-reviewed literature on the PM₁₀/TSP ratio in India. Only one peer-reviewed article describes the size distribution of the ambient aerosol in Mumbai, and it shows a PM₁₀/TSP ratio of 0.87 to 0.89 (Sharma & Patil 1991). While the ratio may be lower for Delhi owing to seasonal dust storms, there are, in fact, no actual measurements to support this contention. If the ratio is much higher than 0.55, then the PM₁₀ levels (which are used in the dose-response calculations)

are much higher than has been assumed, and hence the health effects calculated by all such ecological studies may be underestimates.

Finally, the Ostro equation is based on regressing ambient concentrations of PM₁₀ and mortality in the US and may not necessarily be a causal relationship. A more direct causal relationship would relate personal exposure to mortality. Studies in India that use ambient concentration data implicitly assume that the relationship between ambient concentrations and personal exposure in that country is similar to their relationship in the US. Studies in the US have examined the ratios of personal exposures to outdoor PM₁₀ levels (e.g. Clayton et al 1993), and the observed ratios vary between 1.0 and 4.0, with a midrange of 1.5–2.0. In India, these ratios may hold for high- and middle-income classes but may be vastly different for low- income groups, whose members face very high indoor exposure. Table 18 shows our own estimates that we have developed on the basis of Ostro’s work. While we simply cannot overcome the many caveats discussed above, our estimates incorporate the role of personal exposure in different income classes. For each income class we developed an “effective” ambient concentration by converting personal exposures for each income class into outdoor concentrations, using a range for the personal exposure-to-ambient concentration ratio of 1.5 to 3.0. This allowed us to incorporate the role of indoor air pollution, particularly that from biomass-based sources. Not surprisingly, our estimates for mortality (500,000– 800,000 annual deaths) are far higher than those of Shah & Nagpal (1997) and Brandon & Homman (1995), which simply ignore the effects of indoor exposures. In addition to the mortality for all urban areas, we also have calculated the mortality for Delhi (16,000–27,000) and Mumbai (21,000–35,000).

The population of Mumbai is about 30% larger than that of Delhi, and hence the analysis shows greater mortality in Mumbai, although daily average TSP con- centrations are higher in Delhi by a factor of 1.5–2.0. This is explained by the fact that higher outdoor concentrations have a small effect on personal exposure be- cause adults spend only 1–2 h a day outdoors. However, this calculation does not account for the effect of ambient concentrations on indoor levels, and it assumes similar indoor exposures for populations in similar income classes in both cities. If outdoor concentrations have a significant influence on indoor concentrations, then it is possible that our mortality figure for Delhi has been underestimated and that for Mumbai has been overestimated. Detailed personal-exposure analyses that explore the relationships between indoor and outdoor concentrations are needed to clarify this issue.

It seems more appropriate to extrapolate to the Indian situation dose-response relationships from epidemiological studies in other developing countries, such as China. China and India have comparable populations in many respects—they have similar health and economic statuses, population age distributions (see Figure 4), and average outdoor and indoor TSP concentrations. A number of stud- ies conducted in the 1980s (summarized by Hong [1996]) showed average levels of TSP ranging between 200 µg/m³ and >1100 µg/m³. A study of air pollution due to TSP and mortality was conducted by Xu et al (1994) in Shenyang. The data

showed a 35% increase in mortality for each 100-µg/m³ increase in TSP, which is considerably higher than the estimates based on studies done in countries with developed economies. However, the use of data from this Chinese study results in unrealistically high mortality estimates; the number of deaths from air pollution estimated using this approach exceeds the total number of annual deaths from all causes in urban India.

Studies That Use Epidemiological Methods Appropriate for Industrialized Countries While most studies estimating mortality have used dose-response esti- mates, one recent study, by Cropper et al (1997), using field data, focused on the ef- fect of TSP levels on daily nontrauma deaths in Delhi between 1991 and 1994. They found a 2.3% increase in mortality for every 100-µg/m³ increase in TSP. Based on an average TSP level of 375 µg/m³, they calculated 5070 excess deaths per year in Delhi alone. The Cropper et al (1997) study was patterned on similar time series ecological studies done in the US, in which a time series of particle concentrations is obtained from measurements made daily and compared with another time series consisting of an adverse health outcome (e.g. death). Such studies are ecological if data on the exposures or health outcomes are available only at the level of the en- tire population and not from individuals. For example, in the Cropper et al (1997) study, the entire population was assigned an aggregate TSP concentration value for each day of the study.

Ecological studies can lead to significant measurement error—not just analytical measurement error but also the error introduced by using a central monitor to estimate population-level exposures. These errors can cause substantial biases in the associations between exposure and health effect. Unfortunately, Cropper

et al (1997) presented no estimates of these errors.8 Gamble & Lewis (1996)

provide a detailed criticism of such studies and make the case that such studies provide an unreliable estimate of risk due to the “ecologic fallacy,” in other words, a fallacy inherent in estimating individual risk on the basis of group data.

In their work, Cropper et al (1997) found relatively low-level effects of air pollution on mortality. They found a 0.23% increase in total mortality for every 10-µg/m³ increase in PM₁₀ concentration, as opposed to the 1% increase found in US studies. Cropper et al explained the low-level effects by suggesting that

8While all the coefficients in the model used by Cropper et al (1997) are statistically significant at the 95% level, this significance can be misleading. The significance level associated with a coefficient refers to the term being statistically significantly different from zero, and it is related to the number of observations. Their analysis included ∼1400 data points, and the statistical significance of their coefficients might just be an indication of the statistical power of their study. However, they do not report their R² value, which is a measure of how much of the variability in their data is explained by their model. In similar studies in the US (including Steubenville, which Cropper et al cite in comparison), in spite of having statistically significant P values for model coefficients, the models had R² values of .01–0.1; that is, the models explained 1%–10% of the variability (Gamble & Lewis 1996).

deaths in Delhi occur, on an average, earlier than they do in industrialized countries and that older, more vulnerable populations constitute a smaller proportion in Delhi than in cities in North America. However, this argument may capture only part of the picture. Mortality due to ARI and other respiratory diseases in children is very high in India, and the same is probably true of Delhi, which is also reflected in the Cropper et al data but was left unanalyzed by them.

The use of cause-specific mortality rates may have shown a different pattern. Studies in the US that use cause-specific mortality data (e.g. Schwartz & Dockery 1992) show that 40% of all deaths are in the 0–4-year age group, with a large proportion of deaths being from respiratory causes. Furthermore, the rather small increase in mortality could also be due to saturation in the slope of the dose- response curve at the high TSP concentrations found in Delhi. Despite the above criticisms, the Cropper et al study makes a start in the right direction by attempting to reproduce known results in developing-country settings. More in situ studies are needed if a better understanding of air pollution health effects is to be gained. We calculated estimates for overall mortality from air pollution in India by using the Cropper et al finding of a 2.3% increase in mortality for every 100- µg/m³ increase in TSP, assuming a range of 0.5–0.9 for the PM₁₀/TSP ratio. These estimates are in the range of 110,000–190,000 deaths per year on an all- India basis. In addition to the mortality for all urban areas, we also calculated the mortality for Delhi (3800–6200) and Mumbai (5000–8000). These numbers are of the same order of magnitude as those of other studies with larger increases (1%) in mortality for every 100-µg/m³ increase in PM₁₀ (Brandon & Homman 1995, Shah & Nagpal 1997). However, our estimates are based on far greater levels of exposure, implying that dose-response studies can arrive at similar mortality estimates for India starting from vastly differing assumptions regarding levels of exposure and the slope of the dose-response curve. These uncertainties make a strong case for new studies that integrate personal exposure and mortality in

arriving at dose-response relationships for the Indian population.

Studies That Use Data on Prevalence of Air-Pollution-Related Diseases in India Smith (1999) proposed a “bottom-up” approach in which the actual disease pattern in India is used to estimate the proportion caused by indoor exposures to PM. He used data from the Global Burden of Disease study (Murray & Lopez 1996) and se- lected diseases that contribute ≥1% to the national burden of disease and in whose etiology air pollution might play a role. These diseases included ARI, chronic obstructive lung diseases, tuberculosis, blindness, perinatal effects, and cardiovas- cular diseases. Odds ratios (for exposed vs unexposed populations) for mortality from each of these diseases have been calculated for developing countries by using meta-analyses of previous studies. Exposed populations included children under the age of 5 years and women suffering from the specific diseases listed above. Smith used the mortality data and the odds ratios to estimate the occurrence of between 410,000 and 790,000 excess deaths (with a median value of 500,000) due to particulate air pollution in India (rural and urban), with contributions

from ARI (71%), cardiovascular disease (12%), tuberculosis (12%), and chronic obstructive lung diseases (5%). Assuming that 30% of the total population is ur- ban and that the levels of exposure in the urban poor (women as well as children under age 5 years) are similar to those faced by the poor in rural areas, we estimate 60,000–120,000 excess urban deaths per year based on Smith’s all-India estimate. The bottom-up approach provides a way to evaluate the reasonableness of mortality estimates derived using dose-response relationships. At the same time, estimates based on these approaches have their own accounting problems. The approach primarily involves isolating exposed populations and attributing the mor- tality from certain diseases to air pollution-related causes. While air pollution may play an important role in many such cases, there may be other correlated reasons, particularly related to socioeconomic status, why diseases such as ARI are so prevalent in India. This explains the rather high uncertainty ranges on estimates

by Smith (1999).

The bottom-up approach has been carried out for the population of women and children in India, who are easy to isolate as a group that is predominantly exposed to very high levels of indoor air pollution. The method works for this group because adult men do not face the same levels of high exposure; mortality from respiratory diseases in that group may be attributable to smoking and exposures other than indoor air pollution. While this method has not been applied to different populations in urban India, such cause- and exposure-specific analyses are a powerful way to provide bounds for mortality from air pollution in different income, gender, and age groups.

Morbidity Impacts Due to Particulate Matter Pollution Estimating nonmortal- ity health end points is very problematic owing to underreporting, misclassifica- tion, and misdiagnosis. These problems are aggravated by resource constraints in most developing countries. In addition, attributing health effects to air pollution requires knowledge of baseline incidence rates of various health end points, such as chronic bronchitis, asthma attacks, and other respiratory symptoms, that is not usually available. Again, Ostro (1994) developed regression fits to data from air pollution morbidity studies in the US and Canada which have been used by others (e.g Brandon & Homman 1995) in the Indian context (see appendix).

A number of studies have examined the link between air pollution and mor- bidity (e.g. Awasthi et al 1996, Behera & Jindal 1991, Sharma et al 1998) in the Indian context. These studies have all found systematic linkages between a number of respiratory system-related health outcomes and air pollution. One study (Kamat 1984), however, deserves particular mention. Kamat and coworkers conducted 3-year prospective studies in Bombay that looked at the relationships between air pollution, smoking, diet, and various morbidity indices. The stud- ies by Kamat et al characterized health and nutritional status, smoking patterns, occupational status, and seasonal factors, as well as TSP, SO₂, and NO₂ levels and indices of morbidity ranging from lung function, abnormal chest symptoms such as coughs, colds, and dyspnea, ischemic heart disease, and carboxyhemoglobin

levels in blood, in four communities. Mortality data were not studied, but morbid- ity data in four populations classified by exposure to different levels of SO₂ and TSP—“urban high,” “urban medium,” “urban low,” and “rural”—were compared. In the urban areas, higher pollution levels were associated with higher prevalence of dyspnea, chronic cough, and obstruction of lung function. There was a strong correlation between TSP, SO₂, NO₂, and abnormal chest symptoms. All urban subjects showed expiratory obstruction, especially in the high-exposure areas. Another significant finding was that in urban medium and rural areas, children

<10 years of age suffered from a higher frequency of colds and cough and chest symptoms, and females had a higher frequency of the same than males. This might be due to higher indoor exposures to biomass combustion aerosols. The study con- cludes strongly that air pollution is a major cause of chest morbidity in Bombay.

The Kamat (1984) study also examined the relationship between fluctuations of daily air pollution levels and respiratory symptoms in the same four communities. While the statistical treatment of the data was limited and respiratory symptoms were determined subjectively by health diaries rather than clinical assessment, they found that fluctuations in cold and cough frequency corresponded broadly to TSP in some years and to SO₂ in other years. Despite the lack of sophisticated statistical analyses of the collected data to help tease out the effects of air pollution from those of other factors, this series of studies should serve as a model for future epidemiological research.

URBAN AIR POLLUTION IN INDIA: The Known and

The Known Unknown

Indian cities are undergoing a rapid risk transition. Residents of urban India face multiple risks during the current period of accelerating growth, when both tradi- tional and modern risks coexist. Such risk overlaps are most frequent among the urban poor, who often bear the brunt of traditional air pollution risks resulting from conventional sources of fuel such as biomass. Modern forms of air pollution, resulting from vehicular traffic and industries, tend to affect both the poor and the rich, although little is known about actual exposures to pollution from these sources in different segments of society. This paper has focused on the sources and human health consequences of PM pollution, which is seen to be largely respon- sible for mortality from air pollution in industrialized countries. Measurement of exposure to the thoracic fraction of PM (PM₁₀) is probably a good overall means of examining health risks in the Indian context as well, although the influence of gaseous copollutants and the toxicity of the particles themselves are not very well known. In what follows, we summarize the major findings presented in this paper and discuss implications for data collection and analysis.

A striking feature of urban air pollution in India is the imbalance between the state of knowledge of the issue and its importance from a public health perspective. Overall knowledge and scientific interest regarding urban air pollution is scant, yet

the health of millions is at stake. A similar observation with regard to rural indoor air pollution in India has been eloquently made by Kirk Smith. Smith (1996) points out that “.. .there is yet no research program directly addressing this problem in any national or international organization concerned with health, development, housing, or environment [This] may one day be seen as a shocking public

health neglect of a serious problem for a large population with relatively little political and economic clout.” The same general conclusion also holds true for urban air pollution, although urban residents have significantly more economic and political clout than those living in rural areas.

The sources of air pollution in urban India are poorly understood. Inventories developed either by the government and international agencies or by academic institutions are not publicly available for most cities, Delhi and Mumbai being among the exceptions. Even for these megacities, source inventories are sketchy because elements of data needed to perform inventory calculations are either miss- ing or simply extrapolated from studies carried out in the West. Thus, as they stand today, sources inventories are more akin to educated guesses than to rigorously collected and validated data.

While the actual numbers for PM₁₀ emissions may be uncertain, there is little doubt that vehicular pollution is a major and rapidly growing problem. Although some local data on emission factors have been recently made available, most data on emission factors for vehicles are based on measurements outside India, and there does not seem to be a systematic program for determining emissions reflecting Indian conditions. This is a serious problem because emission factors tend to show wide variability depending on the technology used, the age and maintenance of the vehicle, and fuel quality. The situation for industrial emissions is even worse. Automobiles can be categorized into different types of vehicles, and the usage and emission factors for the categories, however imperfect, are available; industries show a much greater variation in scale and type. This makes it difficult to assess the total numbers of industrial units, let alone calculate their emissions. Levels of emissions from power plants and industry are extremely uncertain but potentially very high.

Emissions from indoor sources are easier to characterize because there is a longer history of measurements from indoor sources and also because emissions from cooking stoves have been extensively studied. However, domestic urban fuel usage is rapidly changing as biomass fuels are being replaced by cleaner fuels. On the other hand, the demand for biomass fuels has not fallen because they continue to be used in small-scale commercial settings. The net effect of these dynamics on biomass emissions must be taken into account because they have a profound effect on human exposures to PM. At a minimum, inventories of indoor air pollution need to characterize fuel usage (and corresponding device efficiencies) in the population and across different income groups. Furthermore, emissions from biomass combustion related to waste disposal might be responsible for very high local-exposure levels for large fractions of the population, yet they remain virtually unquantified.

First-order inventories, of the sort developed in this paper, are useful in illu- minating the key sources of pollution. They also provide a way to assess the relative impact of these sources on humans by using simple models such as those described above. It is evident to us, however, that the utility of simplified invento- ries is limited and that developing and maintaining systematic and comprehensive inventories of pollutants is a matter of some urgency.

Studies of exposures to pollutants are fairly limited. However, there is sub- stantial literature on indoor personal exposure from biomass combustion during cooking in rural India. The situation for low-income groups in cities is probably no different; the houses are small and the fuels and stoves similar, leading to very high exposure levels, on the order of thousands of micrograms per cubic meter. On the other hand, data on personal exposure from other outdoor source categories are very limited for low-income as well as other income categories. While middle- and high-income groups are not subjected to the large doses of exposure from indoor biomass sources, a recent study done in a highly polluted part of Delhi measured a high daily average personal exposure (>500 µg/m³) in these groups as well.

The limited information on exposure at hand suggests two broad patterns. First, like the poor in rural India, urban poor face very high levels of exposure to biomass sources. Biomass sources are most likely the largest single cause of exposure for roughly half the urban population, although this number does not take into account changing patterns of fuel use in cities. Second, levels of exposure to vehicular and industrial sources are high and appear to cut across income classes, although this contention needs to be empirically tested. Furthermore, there are virtually no measured data on exposures for the large sections of society whose constituents either work in outdoor settings in busy locations and are likely to face high-level vehicular exposure or live and work close to industrial sources. In summary, levels of exposure to pollutants are high in virtually all subpopulations, several times higher than the prescribed international limits. However, the patterns of exposure across differences in income class, age, gender, and occupation are still sketchy. Numerous basic personal exposure data must be collected and synthesized in a manner that is useful for epidemiological studies as well as studies of exposure mitigation.

Given the high levels of exposure, there is little doubt that there are major health effects attributable to air pollution in Indian cities. Most in situ epidemiological studies in India have focused on morbidity, and there have been few studies based on mortality data in urban India. The morbidity studies have found a systematic linkage between air pollution and a variety of respiratory, cardiovascular, and eye- related health outcomes. However, a majority of the studies have neither involved a comprehensive research design to account for various alternative causes and confounds nor used statistical analyses to isolate the effects of air pollution in terms of the excess risks it poses.

Studies of mortality tend to use dose-response relationships derived from other settings, particularly those in industrialized countries. There are a number of haz- ards in making such extrapolations because exposure levels and public health

indicators in India are radically different from those in industrialized countries. Thus, mortality estimates are likely to have large uncertainties, although it is dif- ficult, in practice, to assess the true extent of these uncertainties. More recently, some studies have begun to use mortality data from within India to estimate dose- response relationships between ambient air quality and mortality (Cropper et al 1997) and air pollution-related mortality estimates (Smith 1999). However, much needs to be done within the Indian context to assess the relationship between sources, exposures, and mortality in Indian cities.

In this review we have not paid attention to questions related to mitigation of air pollution in urban India. Clearly, there is also a real need for studies that assess mitigation options in terms of technology choices, mitigation costs, and barriers to policy implementation. Recent studies of indoor air pollution (Parikh et al 1999) and urban air pollution (Shah & Nagpal 1997, Bose 1998) have begun to link the analyses of air pollution to broader domains of energy and transportation policy. While a more detailed analysis of science policy issues is outside the scope of this paper, we end with a few brief comments on the role of scientific institutions and analysis in shaping policy decisions.

In industrialized countries, scientific institutions have been instrumental in rais- ing societal concerns about environmental problems. In the US, scientists have contributed in a number of ways to increasing scientific input in public policy debates—as employees of governmental agencies, as members of organized sci- entific bodies such as the National Academy of Sciences, and as researchers in universities and environmental nongovernmental organizations or, conversely, in industries. While there is some debate about whether too much or too little science is reflected in actual policy making, few will deny that significant human and insti- tutional resources are expended in an effort to make scientific analyses responsive to policy needs.

In India the situation is quite different. Despite the presence of a large and tal- ented scientific community, problems related to air pollution do not seem to have received the levels of scientific attention and scrutiny they deserve. While the underlying reasons for this may be quite complex, the lack of scientific attention has resulted in vast gaps in knowledge regarding urban air pollution issues. Fur- thermore, scientific institutions and organized scientific bodies have rarely played a role in synthesizing and summarizing knowledge regarding air pollution with the intent of educating politicians and sensitizing the public. Governmental agencies responsible for air pollution monitoring and control also rarely perform compre- hensive scientific assessments. The role of synthesizing knowledge and informing the public is being performed by a few committed nongovernmental organizations, most notably the Center for Science and the Environment in Delhi. However, the task is too large and scientifically demanding to be carried out by nongovernmental organizations alone.

There is no guarantee that the acquisition of more complete scientific know- ledge on the issue will result in cleaner air in Indian cities. However, it is unlikely that India’s air pollution problems will be resolved easily without the engagement

of the scientific community both in scientific research and in the policy processes aimed at reducing air pollution.

ACKNOWLEDGMENTS

We acknowledge discussions with Kirk Smith on human health and exposure issues. We thank Madhav Badami for extensive discussions and information on vehicular pollution in India. This work was funded in part by the National Science Foundation (grant number SBR 95-21914 to Milind Kandlikar).

APPENDIX

The following equations, developed by Ostro(1994), are used to calculate changes in health effects due to incremental changes in ambient PM₁₀ concentrations. Thus, personal PM10 exposure values must be converted to ambient PM10 concentrations before being used in these equations.

Mortality

Excess deaths = 0.0012 × ∆[PM₁₀] × P × c, where P is the number of people exposed and c is the crude mortality rate (0.0076 for Mumbai).

Morbidity

Change in yearly cases of chronic bronchitis per 100,000 persons = 6.12 ×

∆[PM10].

Change in restricted-activity days/person/year/µg/m³ of PM10 = 0.0575 × ∆[PM10]. Change in respiratory-related hospital admissions per 100,000 persons = 1.2 ×

∆[PM10].

Change in annual risk of bronchitis in children <18 years old = 0.00169 ×

∆[PM10].

Change in daily asthma attacks per asthmatic person = 0.0326 × ∆[PM₁₀]. Respiratory-symptom days per person per year = 0.183 × ∆[PM₁₀].

For example, if we consider only the urban population (300 million people) in India, a 100-µg/m³ decrease in PM₁₀ concentration would lead to the following decreases:

Total cases of chronic bronchitis per year = 1,836,000 [(6.12 × 100/100,000) ×

300 × 10⁶)];

Total restricted-activity days per year = 1,725,000,000 (0.0575 × 100 × 300 ×

10⁶);

Total respiratory-related hospital admissions per year = 360,000;

Annual risk of bronchitis in children <18 years old = 16.9% decrease from the baseline risk of getting bronchitis. Thus, if the baseline risk is 6.47% (Ostro 1994), then the new risk is 5.37%.

Change in daily asthma attacks per asthmatic person = 3; Respiratory-symptom days per person per year = 18.

Visit the Annual Reviews home page at www.AnnualReviews.org

LITERATURE CITED

Aggarwal AL, Raiyani CV, Patel PD, Shah PG, Chatterjee SK. 1982. Assessment of exposure to benzo(a)pyrene in the air for various population groups in Ahmedabad. Atmos. Environ. 16(4):867–70

Agrawal A, Sharma A, Roychowdhury A. 1996. Slow Murder: The Deadly Story of Vehicu- lar Pollution in India. New Delhi: Cent. Sci. Environ.

Ahuja DR, Joshi V, Smith KR, Venkataraman

C. 1987. Thermal performance and emission characteristics of unvented biomass-burning cookstoves: a proposed standard method for evaluation. Biomass 12:247–70

Alam M, Sathaye J, Barnes D. 1994. Urban household energy use in India: efficiency and policy implications. Energy Policy 26(11): 885–91

Automot. Res. Assoc. India. 1997.

Awasthi S, Glick HA, Fletcher RH. 1996. Ef- fect of cooking fuels on respiratory diseases in preschool children in Lucknow, India. Am.

J. Trop. Med. Hyg. 55(1):48–51

Bandhu HK, Puri S, Shahi JS, Mehta D, Garg ML, et al. 1996. An evaluation of the sources of air pollution in the city of Chandigarh, India: a study using EDXRF technique. Nucl. Instr. Methods Phys. Res. B 114:341–44

Banerjee R, Inamdar AB, Phulluke S, Pateriya

B. 1999. Decision support system for energy planning in a district: residential module. Econ. Polit. Week. In press

Behera D, Jindal SK. 1991. Respiratory symp- toms in Indian women using domestic cook- ing fuels. Chest 100(2):385–88

Biswas D, Dutta SA. 1994. Combating the smog and noise in cities. In The Hindu Survey of the Environment, pp. 41–45. Madras: Hindu Press

Bose RK. 1996. Energy demand and environ- mental implications in urban transport—case of Delhi. Atmos. Environ. 30(3):403–12

Bose RK. 1998. Automotive energy use and

emissions control: a simulation model to ana- lyse transport strategies for Indian metro- polises. Energy Policy 26(13):1001–16

Brandon C, Homman K. 1995. The cost of inaction: valuing the economy-wide cost of environmental degradation in India. Proc. Symp. Model. Glob. Sustain. Tokyo: United Nations Univ.

Cent. Pollut. Control Board. 1993. Pollution Statistics. New Delhi: Cent. Pollut. Control Board

Cent. Pollut. Control Board. 1994. Pollution Statistics. New Delhi: Cent. Pollut. Control Board

Cent. Pollut. Control Board. 1995. Pollution Statistics. New Delhi: Cent. Pollut. Control Board.

Chakravarti S. 1998. Deals on wheels. India Today, Jan. 19

Clayton CA, Perritt RL, Pellizzari ED, Thomas KW, Whitmore RW. 1993. Particle total ex- posure assessment methodology (PTEAM) study: distributions of aerosol and elemen- tal concentrations in personal, indoor, and outdoor air samples in a southern Cali- fornia community. J. Expo. Anal. Environ. Epidemiol. 3:227–50

Com. Eur. Norm. 1992. Workplace Atmo- spheres: Size Fraction Definitions for Mea- surement of Airborne Particles in the Work- place. CEN Stand. EN 481. CEN

Confed. Indian Ind. 1996. Indian Environmen- tal Legislation: Guide for Industry and Busi- ness. New Delhi: Confed. Indian Ind.

Cropper ML, Simon NB, Alberini A, Sharma PK. 1997. The Health Effects of Air Pollu- tion in Delhi, India. Policy Res. Work. Pap. No. 1860. World Bank, Washington, DC

De Koning HW, Smith KR, Last JM. 1985. Biomass fuel combustion and health. Bull. WHO 63(1):11–26

Dockery DW, Pope CA III, Xu X, Spengler JD, Ware JH, et al. 1993. An association between

air pollution and mortality in six U.S. cities.

N. Engl. J. Med. 329:1753–59

Faiz A, Sinha K, Walsh M, Warma A. 1990. Au- tomotive Air Pollution: Issues and Options for Developing Countries. Washington, DC: World Bank

Ferris BG Jr, Speizer FE, Spengler JD. 1979. Effects of sulfur oxides and respirable parti- cles on human health: methodology and de- mography of populations in study. Am. Rev. Respir. Dis. 120:767–79

Florig KF. 1997. China’s air pollution risks.

Environ. Sci. Technol. 31:274–79

Gamble JF, Lewis RJ. 1996. Health and res- pirable particulate (PM10) air pollution: a causal or statistical association? Environ.

Health Perspect. 104:838–50

Gov. India. 1993. Pocket Book on Transport Statistics in India. New Delhi: Minist. Surf. Transp., Gov. India

Gov. India. 1996. Annual Report, 1995-1996. New Delhi: Minist. Power, Gov. India

Gov. India. 1997. White Paper on Pollution in Delhi with an Action Plan. New Delhi: Min- ist. Environ. For., Gov. India

Hong CJ. 1996. A review of air pollution and its health effects in China. In Air Pollution and Health Effects in China: A Monograph, pp. 1–46. Geneva: WHO

Indian Inst. Pet. 1985. State of the Art Report on Vehicle Emissions, Dehradun: Indian Inst. Pet.

Indian Inst. Pet. 1994. Vehicle Emissions and Control Perspectives in India—A State of the Art Report. Dehradun: Indian Inst. Pet.

Int. Inst. Popul. Sci. 1995. National Family Health Survey (MCH and Family Planning): India 1992–1993. Bombay: Int. Inst. Popul. Sci.

Int. Stand. Organ. 1992. Air quality—particle size fraction definitions for health-related sampling. Techn. Rep. ISO/TR/7708-1983

(E). Geneva: ISO

Joshi V, Venkataraman C, Ahuja DR. 1989. Emissions from burning biofuels in metal cook-stoves. Environ. Manage. 13(6):763– 72

Kamat SR. 1984. Bombay Air Pollution He- alth Study. Bombay: Munic. Corp. Greater Bombay

Kamat SR, Mahashur AA. 1997. Air pollu- tion: slow poisoning. In The Hindu Survey of the Environment. pp. 57–63. Madras: Hindu Press

Kamens RM, Guo Z, Fulcher JN, Bell DA. 1990. Influence of humidity, sunlight and temperature on the daytime decay of polyaro- matic hydrocarbons on atmospheric soot par- ticles. Environ. Sci. Technol. 22(1):103–8

Koontz MD, Rector HE. 1995. Estimation of Distribution of Residential Air Exchange Rates. Final Rep., Off. Pollut. Prev. Toxics, US EPA, Washington, DC

Krishnamurti CR, Viswanathan P, eds. 1991. Toxic Metals in the Indian Environment. New Delhi: Tata McGraw-Hill

Kulkarni A, Sant G, Krishnayya JG. 1994. Ur- banisation in search of energy in three Indian cities. Energy 19(5):549–60

Kulkarni MM, Patil RS. 1999. Monitoring of daily integrated exposure of outdoor work- ers to respirable particulate matter in an ur- ban region of India. Environ. Monit. Assess. 56:129–46

Lookman AA, Rubin ES. 1998. Barriers to adopting least-cost particulate control strate- gies for Indian power plants. Energy Policy 26(14):1053–63

Mage D, Ozolins G, Peterson P, Webster A, Orthofer R, et al. 1996. Urban air pollution in mega-cities of the world. Atmos. Environ. 30(5):681–86

Mehra A, Farago ME, Banerjee DK. 1998. Im- pact of fly ash from coal-fired power sta- tions in Delhi, with particular reference to metal contamination. Environ. Monit. Assess. 50(1):15–35

Mishra V, Retherford RD, Smith KR. 1997. Ef- fects of Cooking Smoke on Prevalence of Tu- berculosis in India. East-West Center Work. Pap., Popul. Ser., No. 92. East-West Center, Honolulu, Hawaii

Moolgavkar SH, Luebeck EG. 1996. A criti- cal review of the evidence on particulate air

pollution and mortality. Epidemiology 7(4): 420–28

Moolgavkar SH, Luebeck EG, Hall TA, Ander- sen EL. 1995. Particulate air pollution, sulfur dioxide and daily mortality: a reanalysis of the Steubenville data. Inhal. Toxicol. 7:35–44 Mostardi RA, Leonard D. 1974. Air pollution and cardiopulmonary function. Arch. Envi-

ron. Health 29:325–28

Murray C, Lopez A. 1996. Global Burden of Disease. Cambridge, MA: Harvard Univ. Press

Murray DM, Burmaster DE. 1995. Residential air exchange rates in the United States: em- pirical and estimated parametric distributions by season and climatic region. Risk Anal. 15:459–65

Ostro B. 1994. Estimating the Health Effects of Air Pollutants: A Method with Application of Jakarta. Policy Res. Work. Pap. 1301. The World Bank, Washington, DC

Pandey M. 1998. Mother nature under attack.

Asian Aff., Dec.

Pandit GG, Sharma S, Rao AMM, Krish- namoorthy TM. 1996. Chromatographic methods for the estimation of polycyclic aro- matic hydrocarbons in atmospheric partic- ulates. Proc. 5th Natl. Symp. Environ., pp. 133–36

Parikh JK, Smith KR, Laxmi V. 1999. Indoor air pollution: a reflection on gender bias. Econ. Polit. Week., Feb. 27

Penna MLF, Duchiade MP. 1991. Air pollution and infant mortality from pneumonia in the Rio de Janeiro metropolitan area. Bull. Pan Am. Health Organ. 25:47–54

Pope CA III, Schwartz J, Ransom MR. 1992. Daily mortality and PM10 pollution in Utah valley. Arch. Environ. Health 47:211–17

Raiyani CV, Jani JP, Desai NM, Shah JA, Kashyap SK. 1993a. Levels of polycyclic aromatic hydrocarbons in ambient environ- ment of Ahmedabad. Indian J. Environ. Prot. 13:206

Raiyani CV, Shah SH, Desai NM, Venkaiah K, Patel JS, et al. 1993b. Characterization and problems of indoor pollution due to cooking

stove smoke. Atmos. Environ. A 27:1643–55 Ramousset JA, Smith KR. 1990. Exposure trad- ing: an approach to more efficient pollution control. J. Environ. Econ. Manage. 18:276–

91

Rao AMM, Rajgopalan R, Doshi MH, Vohra KG. 1982. Measurement of benzo(a)pyrene in the city of Bombay for carcinogenic risk. Sci. Total Environ. 22:105–13

Ravindranath NH, Ramakrishna J. 1997. En- ergy options for cooking in India. Energy Policy 25(1):63–75

Ravi Shankar V. 1990. Characterization of at- mospheric polynuclear aromatic hydrocar- bons with special reference to automobile ex- haust. PhD thesis. Jawaharlal Nehru Univ., New Delhi

Reddy BS. 1997. Flow of energy in an urban society. Energy Sources 19(3):271–94

Rowe MD. 1985. NO₂ exposure from vehicles and stoves. J. Transport. Eng. 111(6):679–91 Saltzman BE, Cholak J, Schafer LJ, Yeager DW, Meiners BG, Svetlik J. 1985. Concentrations of six metals in the air of eight cities. Environ.

Sci. Technol. 19:328–33

Samet JM, Zeger SL, Berhane K. 1995. The association of mortality and particulate air pollution. In Particulate Air Pollution and Daily Mortality: Replication and Validation of Selected Studies. Cambridge, MA: Health Eff. Inst.

Saxena S, Dayal V. 1997. Total exposure as a ba- sis for the economic valuation of air pollution in India. Energy Environ. Monit. 13(2):93– 102

Schwartz J. 1993. Air pollution and daily mor- tality in Birmingham, Alabama. Am. J. Epi- demiol. 137:1136–47

Schwartz J, Dockery DW. 1992. Increased mor- tality in Philadelphia associated with daily air pollution concentrations. Am. Rev. Respir. Dis. 145:600–4

Schwartz J, Dockery DW, Neas LM. 1996. Is daily mortality associated specifically with fine particles? J. Air Waste Manage. Assoc. 46:927–39

Semb A. 1986. Measurements of Emissions

from Open Burning. NILU Rep. AS/SBH/ 08622/10. Lillestrøm, Norway

Sexton K, Ryan PB. 1988. Assessment of hu- man exposure to air pollution: methods, mea- surements and models. In Air Pollution, the Automobile, and Public Health. Washington, DC: Natl. Acad. Press

Shah J, Nagpal T, eds. 1997. Urban Air Qual- ity Management Strategy in Asia (URBAIR). Greater Mumbai Rep. World Bank Tech. Pap. No. 381, World Bank, Washington, DC

Shah N, Ramankutty V, Premila PG, Sathy

N. 1994. Risk factors for severe pneumo- nia in children in south Kerala: a hospital- based case-control study. J. Trop. Pediatr. 40(4):201–6

Sharma S, Sethi GR, Rohtagi A, Chaudhary A, Shankar R, et al. 1998. Indoor air quality and acute lower respiratory infection in In- dian urban slums. Environ. Health Perspect. 106(5):291–97

Sharma VK, Patil RS. 1991. In situ measure- ments of atmospheric aerosols in an in- dustrial region of Bombay. J. Aerosol Sci. 22:501–7

Shukla N. 1999. Exposure Assessment to Par- ticulate Matter in the Middle Class Sub- population of East Delhi. Res. Rep., Environ. Monit. Train. Res. Cent., New Delhi

Singh MP, Goyal P, Panwar TS, Agarwal P, Nigam S, Bagchi N. 1990. Predicted and ob- served concentrations of SO2, SPM, and NOx

over Delhi. Atmos. Environ. A 24:783–88

Smith KR. 1987. Biofuels, Air Pollution and Health: A Global Review. New York: Plenum Smith KR. 1993. Fuel combustion, air pollution exposure, and health: the situation in devel- oping countries. Annu. Rev. Energy Environ.

18:529–66

Smith KR. 1994. Health risks from air pol- lution in India: implications of new stud- ies. Proc. Workshop Energy-Environ. Nexus: Indian Issues and Global Impacts, Univ. Penn. Philadelphia: Cent. Adv. Study India

Smith KR. 1996. Indoor air pollution in India.

Natl. Med. J. of India 9(3):103–4

Smith KR. 1999. The national burden of disease

from indoor air pollution in India. Presented at Indoor Air ’99, Edinburgh, UK

Smith KR, Aggarwal AL, Dave RM. 1983. Air pollution and rural biomass fuels in develop- ing countries: a pilot village study in Indian and implications for research and policy. Atmos. Environ. 17:2343–62

Smith KR, Liu Y. 1994. Indoor air pollution in developing countries.. In The Epidemiology of Lung Cancer, ed. J Samet , pp. 151–84. New York: Marcel Dekker

Smith KR, Ramakrishna J, Menon P. 1981. Air Pollution from the Combustion of Tra- ditional Fuels: A Brief Survey. Work. Pap. WP-81-5. East-West Resour. Syst. Inst., Honolulu, HI

Tata Energy Research Institute. 1993. Im- pact of Road Transportation on Energy and Environment—An Analysis of Metropolitan Cities of India. New Delhi: TERI

Tata Energy Research Institute. 1997. India 2047: Looking Back to Think Ahead. New Delhi: TERI

Tripathi A. 1994. Airborne lead pollution in the city of Varanasi, India. Atmos. Environ. 28:2317–23

Tripathi RM, Khandekar RN, Raghunath R, Mishra UC. 1989. Short communication: as- sessment of atmospheric pollution from toxic heavy metals in two cities in India. Atmos. Environ. 23:879–83

US Environ. Prot. Agency. 1973. Mobile Source Emission Standards Summary. Washington, DC: US EPA. 2nd ed.

US Environ. Prot. Agency. 1982. Review of the National Ambient Air Quality Standard for Particulate Matter: Assessment of Scientific and Technical Information. Research Trian- gle Park, NC: US EPA

US Environ. Prot. Agency. 1985. Compilation of Air Pollutant Emission Factors. EPA Man- ual AP-42, Suppl. A. Research Triangle Park, NC: US EPA. 4th ed.

US Environ. Prot. Agency. 1986. Fuel oil com- bustion. In Compilation of AirPpollutant Emission Factors, pp. 1.3-1–1.3-11, Suppl. Research Triangle Park, NC: US EPA. 4th ed.

US Environ. Prot. Agency. 1996. Review of the National Ambient Air Quality Standards for Particulate Matter. Policy Assessment of Sci- entific and Technical Information. OAQPS Staff Pap. EPA-452/R-96-013. Washington, DC: US EPA

Venkataraman C. 1998. Atmospheric behaviour of polycyclic aromatic hydrocarbons in urban aerosols. In Aerosols: Generation and Role in Medicine, Industry, and Environment, ed. KS Nambi, BK Sapra. New Delhi: Allied Publ.

Wang XD, Smith KR. 1999. Secondary benefits of greenhouse gas control: health impacts in China. Environ. Sci. Technol. 33(18):3056– 61

Wilson WE, Suh HH. 1997. Fine particles and coarse particle concentration relation- ships relevant to epidemiologic studies. J. Air Waste Manage. Assoc. 47:1238–49

WHO. 1992. Programme for Control of Acute

Respiratory Infections. WHO/ARI/92.22. Geneva: WHO

WHO. 1993. Assessment of Sources of Air, Wa- ter, and Land Pollution. A Guide to Rapid Source Inventory Techniques and Their Use in Formulating Environmental Control Stra- tegies. Part 1. Rapid Inventory Techniques in Environmental Pollution. WHO/PEP/ GETNET/93.1-A. Geneva: WHO

WHO. 1997. Tobacco or Health: A Global Sta- tus Report. Geneva: WHO

WHO. 1997. Health and Environment in Sus- tainable Development. Geneva: WHO

WHO/UN Environ. Programme. 1992. Urban Air Pollution in Mega-cities of the World. Oxford: Blackwell. 39 pp.

Xu Z, Chen BH, Kjellstrøm T, Xu X, Lin Y, Daqian Y. 1994. Study of severe air pollu- tion and mortality in Shenyang, China. In Air Pollution and Its Health Effects in China: A Monograph, pp. 47–88. Geneva: WHO

Annual Review of Energy and the Environment Volume 25, 2000

CONTENTS

CONVERGING PATHS LEADING TO THE ROLE OF THE OCEANS

IN CLIMATE CHANGE, Wallace S. Broecker 1

ENERGY IN THE TWENTIETH CENTURY: Resources, Conversions,

Costs, Uses, and Consequences, Vaclav Smil 21

PHOSPHORUS IN THE ENVIRONMENT: Natural Flows and Human

Interferences, Vaclav Smil 53

TECHNOLOGIES SUPPORTIVE OF SUSTAINABLE

TRANSPORTATION, A. Dearing 89

OPPORTUNITIES FOR POLLUTION PREVENTION AND ENERGY EFFICIENCY ENABLED BY THE CARBON DIOXIDE

TECHNOLOGY PLATFORM, Darlene K. Taylor, Ruben Carbonell,

Joseph M. DeSimone 115

WINDPOWER: A Turn of the Century Review, Jon G. McGowan,

Stephen R. Connors 147

THE POTENTIAL OF BIOMASS FUELS IN THE CONTEXT OF

GLOBAL CLIMATE CHANGE: Focus on Transportation Fuels, Haroon

S. Kheshgi, Roger C. Prince, Gregg Marland 199

GEOENGINEERING THE CLIMATE: History and Prospect, David W.

Keith 245

THE ENGLAND AND WALES NON-FOSSIL FUEL OBLIGATION:

History and Lessons, Catherine Mitchell 285

INDUSTRIAL SYMBIOSIS: Literature and Taxonomy, Marian R.

Chertow 313

INTEGRATED ANALYSIS FOR ACID RAIN IN ASIA: Policy

Implications and Results of RAINS-ASIA Model, Jitendra Shah, Tanvi Nagpal, Todd Johnson, Markus Amann, Gregory Carmichael, Wesley Foell, Collin Green, Jean-Paul Hettelingh, Leen Hordijk, Jia Li, Chao

Peng, Yifen Pu, Ramesh Ramankutty, David Streets 339

CAPACITY DEVELOPMENT FOR THE ENVIRONMENT: A View for

the South, A View for the North, Ambuj D. Sagar 377

WATER VAPOR FEEDBACK AND GLOBAL WARMING, Isaac M.

Held, Brian J. Soden 441

ENGINEERING-ECONOMIC ANALYSES OF AUTOMOTIVE FUEL ECONOMY POTENTIAL IN THE UNITED STATES, David L.

Greene, John DeCicco 477

HEALTH AND PRODUCTIVITY GAINS FROM BETTER INDOOR ENVIRONMENTS AND THEIR RELATIONSHIP WITH BUILDING

ENERGY EFFICIENCY, William J. Fisk 537

INDOOR AIR QUALITY FACTORS IN DESIGNING A HEALTHY

BUILDING, John D. Spengler, Qingyan Chen 567

PUBLIC HEALTH IMPACT OF AIR POLLUTION AND IMPLICATIONS FOR THE ENERGY SYSTEM, Ari Rabl, Joseph V.

Spadaro 601

THE CAUSES AND CONSEQUENCES OF PARTICULATE AIR

POLLUTION IN URBAN INDIA: A Synthesis of the Science, Milind

Kandlikar, Gurumurthy Ramachandran 629

ENERGY AND MATERIAL FLOW THROUGH THE URBAN

ECOSYSTEM, Ethan H. Decker, Scott Elliott, Felisa A. Smith, Donald

R. Blake, F. Sherwood Rowland 685

GREENHOUSE IMPLICATIONS OF HOUSEHOLD STOVES: An

Analysis for India, Kirk R. Smith, R. Uma, V.V.N. Kishore, Junfeng

Zhang, V. Joshi, M.A.K. Khalil 741

METHYL tert-BUTYL ETHER AS A GASOLINE OXYGENATE:

Lessons for Environmental Public Policy, Serap Erdal, Bernard D.

Goldstein 765