Increase in NOx Emissions from Indian Thermal Power Plants during1996−2010: Unit-Based Inventories and Multisatellite ObservationsZifeng Lu* and David G. Streets

Decision and Information Sciences Division, Argonne National Laboratory, Argonne, Illinois, United States

*S Supporting Information

ABSTRACT: Driven by rapid economic development and growingelectricity demand, NOx emissions (E) from the power sector in Indiahave increased dramatically since the mid-1990s. In this study, wepresent the NOx emissions from Indian public thermal power plantsfor the period 1996−2010 using a unit-based methodology andcompare the emission estimates with the satellite observations of NO2tropospheric vertical column densities (TVCDs) from four space-borne instruments: GOME, SCIAMACHY, OMI, and GOME-2.Results show that NOx emissions from Indian power plants increasedby at least 70% during 1996−2010. Coal-fired power plants, NOxemissions from which are not regulated in India, contribute ∼96% tothe total power sector emissions, followed by gas-fired (∼4%) and oil-fired (<1%) ones. A number of isolated NO2 hot spots are observedover the power plant areas, and good agreement between NO2 TVCDs and NOx emissions is found for areas dominated bypower plant emissions. Average NO2 TVCDs over power plant areas were continuously increasing during the study period. Wefind that the ratio of ΔE/E to ΔTVCD/TVCD changed from greater than one to less than one around 2005−2008, implyingthat a transition of the overall NOx chemistry occurred over the power plant areas, which may cause significant impact on theatmospheric environment.

■ INTRODUCTIONNitrogen dioxide (NO2) and nitric oxide (NO), togetherknown as nitrogen oxides (NOx), are two of the mostimportant gaseous species in the atmosphere, and they areemitted from both anthropogenic sources (e.g., fossil-fuelcombustion and human-induced biomass burning) and naturalsources (e.g., wildfires, soil release, lightning, ammoniaoxidation). NOx plays a crucial role in the formation ofozone and secondary aerosol and is involved in the chemicaltransformation of other atmospheric species (e.g., CO, CH4,VOCs) through feedback on HOx (OH + HO2). Consequently,NOx can have broader effects on human health, atmosphericcomposition, acid deposition, air/water quality, visibility,radiative forcing, etc. As the second largest NOx emittingcountry in Asia (∼16% of total Asian emissions1−5), India is ofincreasing concern. Thermal power plants are the mostimportant point sources in India, accounting for more than30% of the national NOx emissions (transportation, industry,and other sectors contribute ∼35%, ∼20%, and ∼15%,respectively).6 Due to the rapid economic development, thegrowing electricity demand, and the absence (or weakenforcement) of regulations, NOx emissions in the Indianpower sector have been reported to increase at a remarkablerate since the mid-1990s.1−3,6 Although some previous studieshave reported NOx emissions from Indian power plants,1−7

they have various shortcomings. First, none of them presentyear-by-year trends with up-to-date activity rates, especially for

the period after 2000. Second, NOx emission factors arestrongly dependent on combustion conditions, fuel quality,boiler size, and NOx control technology; however, previousstudies simply treated all power plants as one single sector (orsubsectors by fuel type) and applied default NOx emissionfactors prescribed by the IPCC.8 Third, few of them usedactivity rates at the plant or unit level but instead at the countryor state level, which cannot yield spatially accurate emissions.To overcome these disadvantages, we use a unit-basedmethodology to develop new NOx emission inventories forIndian power plants.Due to the short lifetime of NOx in the atmosphere, satellite

NO2 observations are closely correlated to the land surfaceNOx emissions.9−11 In the past few years, tropospheric NO2

columns retrieved from satellites have been successfully appliedto monitor and quantify the spatial distribution (local toglobal),9,10,12,13 temporal variation (diurnal to seasonal),14−18

and interannual trends11,19,20 of NOx emissions from bothanthropogenic and natural sources. Especially, because of theirhigh spatial resolution, the new generation of instruments (e.g.,SCIAMACHY and OMI) have proved to be able to identifyand constrain NOx emissions from large point sources (LPS)

Received: March 1, 2012Revised: June 18, 2012Accepted: June 25, 2012Published: June 25, 2012

Article

pubs.acs.org/est

© 2012 American Chemical Society 7463 dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−7470

such as thermal power plants.9,12−14 However, there have beenvery few applications of satellite retrievals to India. Ghude etal.21,22 and Prasad et al.23 studied the interannual trends andseasonal variations of major NO2 hotspots over India usingNO2 retrievals from GOME, SCIAMACHY, and OMI.However, they only gave a qualitative or semiquantitativeanalysis of the relationship between NOx emissions and satelliteretrievals over NO2 hotspots, partially due to the lack of reliableNOx emission inventories for LPS. In this study, we examinethe India NOx emission trend of thermal power plants duringthe Indian financial years (FY, from 1 April of one year to 31March of the next year) 1996−2010 from the viewpoints ofboth unit-based inventories and multisatellite observations. Wecompare the satellite NO2 data with bottom-up NOx emissionestimates over power plant areas in India. The results areintended to provide a better understanding of the large NOx

releases from the Indian power sector for recent years. In amore general sense, the results contribute to our ongoingattempts to assess the usefulness of satellite retrievals forquantifying LPS emissions.

■ METHODS AND DATA SETS

Unit-Based Methodology for NOx Emission Estima-tion. A bottom-up, unit-based methodology is used to developthe NOx emission inventory for Indian public thermal powerplants during FY 1996−2010. Captive (privately owned) powerplants and public coal-fired units with capacity smaller than 20MW are not included due to the lack of detailed data at unit orplant level. In total, information on more than 800 units wascompiled, including geographical location, boiler size (ca-pacity), fuel type, electricity generation, specific fuel con-sumption, the exact time when the unit came into operationand/or retired, etc. Most of the information was derived from aseries of the Performance Review of Thermal Power Stations24

and various reports published by the Central ElectricityAuthority (CEA), Ministry of Power of India. Where unit-level information was not available, corresponding plant-levelinformation was used. The exact longitude and latitude of eachpower plant was obtained from the Global Energy Observatory(http://globalenergyobservatory.org/index.php), and verifiedthrough Google Earth. As shown in Figure S1 of theSupporting Information (SI), the electricity generation ofthermal power plants in India increased dramatically from 321.9to 684.5 TWh during FY 1996−2010, with an annual averagegrowth rate (AAGR) of 5.5%. On average, about 87% ofthermal electricity was generated from coal-fired power units,followed by gas-fired (∼12%) and oil-fired (∼1%) ones. Coal-fired units with capacity larger than 200 MW are the mainboiler type in India, contributing ∼72% to the total thermalelectricity generation. Figure S1 also shows the officialelectricity generation statistics from the Government ofIndia,24 the United Nations Statistics Division (http://unstats.un.org/unsd/default.htm), and the International EnergyAgency (IEA).25 They match reasonably well with the dataderived from our unit-based database (differences < ±3%),indicating that the current database covers nearly all publicgenerating units in India.Total NOx emissions (E, Mg yr−1) from thermal power

plants for year i are estimated by the following equation:

∑ ∑ ∑ ∑= × × ×

× −

E G( SFC NCV EF

10 )

ij k l m

i j k i j k j j l m, , , , , ,

9(1)

where j, k, l, and m stand for fuel type, generation unit, boilersize, and emission control technology, respectively. Grepresents the electricity generation (kWh yr−1); SFCrepresents the specific fuel consumption per unit electricitygeneration (kg kWh−1 for coal- and oil-fired units, and m3

kWh−1 for gas-fired units); and NCV represents the netcalorific value (MJ kg−1 for coal- and oil-fired units, and MJ m−3

for gas-fired units). EF is the NOx emission factor (g GJ−1).NOx emission factors of power generating units vary with

fuel type, fuel quality, boiler size, and NOx controltechnology.15 However, due to the absence of India-specificmeasurement data, nearly all previous studies2−4,6 used defaultEF values prescribed by the IPCC8 to estimate NOx emissionsfrom the Indian power sector. In India, more than 95% of thecoal-fired boilers are pulverized coal (PC) burning type.2,26 Inthis study, boiler-size-specific and emission-control-specific EFswith PC combustion technology were applied to all coal-firedunits (see Table 1), and they were derived from not only the

IPCC Emission Factor Database (EFDB),27 but also 209 fieldmeasurements of Chinese bituminous and lignite coal reportedby Zhao et al.28,29 (Indian coal is of mostly sub-bituminousrank, followed by bituminous and lignite.) We convert the unitof Zhao et al.’s EFs from g kg−1 to g GJ−1 on the basis of NCVsupplied in the IEA Energy Statistics25 to take into account thedifference in coal heating values between the two countries.Currently, NOx emissions are not regulated in India for coal-

fired power plants. Although some new plants were reported tobe equipped with low-NOx burners (LNB), the actualinstallation rate, operation, and performance of LNB devicesin India are unknown.26 To reflect possible alternative NOxemission situations in the Indian power sector during FY1996−2010, we generate five emission scenarios (Table 1). Asused in previous studies,2−4,6 we first generate the baselineemission scenario (S1), in which the default IPCC EF of 300 gGJ−1 is applied to all coal-fired boilers. In scenarios S2−S4, weassume that no boiler (S2), all boilers with capacity ≥300 MW

Table 1. Summary of NOx Emission Factors and Scenariosfor Coal-Fired Power Units

emission scenarios

boiler size LNBEF (gGJ−1) S1 S2 S3 S4

S5<1996

fS5

>1996g

not classified 300a √<100 MW w/o 308b √ √ √ √

w/ 177c √100−300MW

w/o 330b √ √ √

w/ 188d √ √≥300 MW w/o 410e √ √

w/ 236b,d √ √ √aFrom IPCC.8 bFrom Zhao et al.,28 assuming the NCV of bituminouscoal in China is 21.2 MJ kg−1.25 cEstimated, assuming the averageremoval efficiency of the LNB devices is 43%. dFrom Zhao et al.,29

assuming the NCV of bituminous coal in China is 21.2 MJ kg−1.25eFrom IPCC EFDB,27 average EF of sub-bituminous coal for boilerssize ≥300 MW. fFor units built before 1996. gFor units built after1996.

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707464

(S3), and all boilers with capacity ≥100 MW (S4) are equippedwith LNB, respectively. In scenario S5, we utilize theassumption made in the GAINS inventory2 that no boilerbuilt before 1996 was equipped with LNB, but LNB deviceswere installed in all new boilers built after 1996. Thesescenarios are also used to analyze the sensitivities anduncertainties introduced by the LNB assumptions. For gas-fired and oil-fired power plants, India has NOx emissionstandards varying with the age and size of the unit.30 Takinginto account both the standards and the EFs prescribed by theIPCC,8 EFs of gas-fired units are set to be 150 g GJ−1 for unitsbuilt before June 1, 1999, and 100 g GJ−1, 75 g GJ−1, and 50 gGJ−1 for newly installed units with capacity <100 MW, 100−400 MW, and ≥400 MW, respectively. For naphtha-fired units,EFs are 150 g GJ−1 and 100 g GJ−1 for units commissionedbefore and after June 1, 1999. We assume an EF of 200 g GJ−1

for all other oil-fired power units.Satellite Retrievals of Tropospheric NO2. To compare

with our emission estimates, tropospheric vertical columndensities (TVCDs) of NO2 from GOME, SCIAMACHY, OMI,and GOME-2 are used. The combined operation time of theseinstruments covers the entire period of Indian FY 1996−2010.A comparison of the characteristics of each instrument is shownin Table S1 of the SI. In this work, we used the monthly level-3products of GOME (TM4NO2A v2.0), SCIAMACHY(TM4NO2A v2.0), OMI (DOMINO v2.0), and GOME-2(TM4NO2A v2.1) developed at the Royal NetherlandsMeteorological Institute (KNMI),31,32 which are availablefrom the Tropospheric Emission Monitoring Internet Service(TEMIS) at www.temis.nl. The retrieval processes are similarfor all four instruments. First, NO2 slant columns wereretrieved using the differential optical absorption spectroscopy(DOAS) technique; second, TVCDs were derived with theKNMI combined retrieval-assimilation-modeling approach.31,32

■ RESULTS AND DISCUSSIONNOx Emissions of Thermal Power Plants during FY

1996−2010. Annual trends of NOx emissions from Indianthermal power plants during FY 1996−2010 are shown inFigure 1, and detailed results by fuel type, unit size, and majorpower plants are provided in Tables S2−S3 of the SI. Although

the estimates are different under different scenarios, they allshow similar increasing trends. In the baseline scenario S1, NOxemissions increase rapidly by 97% from 1213 Gg in FY 1996 to2391 Gg in FY 2010, with an AAGR of 5.0%. In other scenarios,the AAGRs are in the range of 3.9% to 5.1%, in agreement withvalues reported in other inventories (2.8%−6.0%).1−3,6 Thedramatic change of NOx emissions can be viewed in the contextof a 113% increase of electricity generation, an 84% increase ofgenerating capacity, and a 95% increase of fuel consumption inthe thermal power sector during FY 1996−2010, reflectingrapid economic and social development driven by growingfossil-fuel use and relatively lax emission control.33 Coal-firedpower plants dominate the total NOx emissions from the powersector (93%−98%), followed by gas-fired (2%−6%) and oil-fired (0−1%) plants. In terms of boiler size, coal-fired unitswith capacity larger than 200 MW are the main contributors,accounting for ∼85% of the NOx emission growth.A Monte Carlo approach is used to determine the

uncertainties of the emission estimates in different scenarios.Unless specified otherwise, the term “uncertainty” in this articlerefers to a 95% confidence interval (CI) around the centralestimate. Input parameters and their corresponding probabilitydistributions were incorporated into a Monte Carlo frameworkwith the Crystal Ball software and 10 000 simulations wereperformed. For fuel consumption data, we applied normaldistributions and assigned uncertainties of 5% and 10% to coaland other fuels. Uncertainties of coal EFs were set to be 15%with normal distribution, on the basis of field measurements ofbituminous coal and lignite by Zhao et al.29 Normaldistributions with 30% uncertainties were assumed for EFs ofother fuels based on the authors’ judgment. LNB-relatedparameters (e.g., application rate, operation rate, and perform-ance) were not included in the Monte Carlo simulation, since itis not possible to obtain the necessary information. Instead, weexamined the sensitivity of the LNB assumptions throughdifferent scenarios to analyze uncertainties introduced by them.Results show that due to uncertainties in activity rates and EFsalone, the uncertainties of estimated NOx emissions arebetween ±11% and 15% under all scenarios (uncertainties ofscenarios S2 and S4 are shown in Figure 1). The uncertaintiesare lower than those reported in the TRACE-P inventory(±21%),4 and mainly attributed to the detailed unit-basedmethodology and reliable unit-level information gathered inthis study. Compared to activity rates and EFs, LNB-relatedparameters seem to be a more crucial factor that influences theaccuracy of NOx emission estimates in India. As shown inFigure 1, the difference between emissions estimated inscenario S2 (no boilers with LNB) and S4 (boiler capacity≥100 MW with LNB) is ∼63%, much higher than theuncertainties introduced by fuel consumption data and EFs.Figure 1 also compares the emissions estimated in this study

to those of other inventories, including GAINS,2 REAS,3

TRACE-P,4 INTEX-B,5 HTAP-EDGAR,7 EDGAR4.2,1 andGarg et al.6 Taking into account the uncertainties in all inputparameters, scenarios S2 and S4 can be treated as the upper andlower bounds of NOx emissions from thermal power plants inIndia. Clearly, most previous results are within this “broad”uncertainty range, except for emission estimates in theEDGAR4.2 inventory.

NOx Emissions of Thermal Power Plants Observedfrom Space. Because of the short lifetime of NOx in the lowermixed layer, surface NOx emissions and NO2 TVCDs areclosely correlated, and changes in NO2 columns are expected to

Figure 1. Comparison of NOx emission estimates of Indian thermalpower plants.

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707465

be approximately proportional to changes in NOx emis-sions.9−11 This makes it possible to identify the NOx sourcesfrom NO2 column maps. Figure 2a and Figures S2a−S4a showthe spatial distribution of average NOx emissions from India,highlighting thermal power plants and all-month averages ofNO2 TVCDs over India from four instruments. A number ofsatellite NO2 hot spots (especially for OMI, SCIAMACHY, andGOME-2, which have finer footprints) are observed over India,and they match the locations and the amounts of NOxemissions of thermal power plants reasonably well. This isconsistent with the findings of Ghude et al.21,22 and Prasad etal.23 because thermal power plants are the major local NOxemitters and power plants in India were often built inpopulated areas where other anthropogenic NOx emissions(e.g., industrial and vehicular) are also high. Influenced by theSouth Asia monsoon meteorology, India is largely subject tofour seasons: winter (December−February), summer/premon-soon (March−June), monsoon (late June−September), andpostmonsoon (October−November).34 The spatial distribu-tions of average NO2 TVCDs by season for each instrument areshown in Figure 2c−f and Figures S2−S4. NO2 columns arehigh in winter and low in the monsoon season due to theseasonal variation of NOx lifetime and the effective wetscavenging during the monsoon season.15,21

Three major factors may impact the comparison betweensatellite NO2 retrievals and emission estimates: uncertainty inthe satellite retrievals, seasonal variation of NOx lifetime, andincomplete conversion of NO to NO2.

14 These factors are thereasons summertime (June−August) NO2 retrievals were

always used in previous studies to quantitatively analyze NOxemissions in the U.S.9,13 and China.12,35 The short NOxlifetime, low snow cover, small solar zenith angle, and strongactinic flux in summer make the retrievals more reliable and thelink between emissions and the NO2 columns more direct thanin the other seasons in these two countries.15 However, forIndia, although the NOx lifetime is short during the monsoonperiod, late June to September is the worst period to observeNO2 from satellites (Figure 2c−f and Figures S2−S4). In India,more than 70% of the annual precipitation occurs in themonsoon season.36 The frequent cloud cover and heavy rainfallresults in insufficient satellite observation samples andextremely low NO2 surface concentrations, and makes thesatellite retrievals highly uncertain. However, the fact that Indiais closer to the equator (the latitude is ∼20° lower) than theU.S. and China brings tremendous advantages to India for theuse of other months’ NO2 retrievals in the quantitative analysis.First, due to the low latitude, temperature and actinic flux arehigh in India, resulting in a relatively short lifetime of NOx allyear round. Second, low snow cover and favorable satellitemeasurement geometry at small solar zenith angle make theNO2 retrievals very reliable. Third, compared to the U.S. andChina, the seasonality of temperature, actinic flux, NOx lifetime,and NO2 retrievals in India is weaker, which means that all-yeardata can be used in the analysis and the NO2 retrievals’uncertainties are further reduced. Taking into account all theabove factors, we select all nonmonsoon months as our studyperiod for India, in contrast to the summer-only NO2 retrievalsused in previous studies of the U.S. and China. An additional

Figure 2. Spatial distribution of (a, c−f) average NOx emissions from thermal power plants (203 plants, based on baseline scenario S1), (b) definedpower plant areas, and (a, b) annual and (c−f) seasonal average of OMI NO2 TVCDs over India during FY 2005−2010.

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707466

favorable condition for quantitative analysis in India is that areaNOx emissions are smaller and individual NOx sources andNO2 hotspots are more isolated and discernible all over thecountry (except for the region around Delhi) in contrast to thedense regional NO2 plumes over the northeastern U.S. andeastern central China.To compare the NOx emissions from thermal power plants

with NO2 TVCDs of different instruments, 81 power plantareas were defined in this study (Figure 2b).Typically, the sizeof an area with a single plant is 0.625° × 0.625°. We combinedthe adjacent plants and adjusted the boundaries of the areasbased on the shapes of NO2 plumes with the intent of capturingthe major satellite signals over all thermal power plants andexcluding signals from nearby sources. The smallest study areais close to or larger than one pixel of SCIAMACHY, GOME-2,and OMI, but smaller than one pixel of GOME in the across-track direction (∼320 km). This implies that GOME may beinsufficient to detect the power-plant signals, and GOMEresults may contain larger uncertainties than the otherinstruments. However, since GOME is the only instrumentcovering the period 1996−2002, we still include it in theanalysis.For power plant area n, we define the NO2 TVCD attributed

to emissions from thermal power plants (TVCDpower) as

= ×

= ×

f

E

E

TVCD TVCD

TVCD

n n n

n

nn

power, power, total,

power,

total,total,

(2)

where TVCDtotal,n is the TVCD averaged over area n, andf power,n is the proportion of power plants’ emissions (Epower,n) tototal NOx emissions (Etotal,n) in area n. In this study, griddedemissions of other anthropogenic sources were taken fromEDGAR4.21 for the year 2005 and scaled to 1996−2010 basedon the GAINS inventory.2 Emissions from natural sources werealso derived from EDGAR4.2 for 2005, but kept constantthroughout the years 1996−2010. Lightning emissions are notincluded in total emissions, because they are negligiblecompared to anthropogenic emissions and they are intensivein the monsoon season, which is excluded from this analysis.Figure 3 presents the scatter plots of annual Epower (based on

baseline scenario S1) versus annual (excluding monsoonperiod) averaged TVCDtotal and TVCDpower over power plantareas for the different NO2 instruments. The results based onthe other scenarios are similar (Figures S5−S8). We excludedata with an average TVCDtotal smaller than 0.5 × 1015

molecules cm−2 since they are lower than the errors of NO2retrievals.31,32 For all data points, Epower correlates better withTVCDpower (R

2 = 0.52−0.71) than TVCDtotal (R2 = 0.34−0.50),

Figure 3. Scatter plots of annual NOx emissions from thermal power plants (based on baseline scenario S1) against annual (corresponding to IndianFY, excluding monsoon period) average TVCDtotal (top) and TVCDpower (middle) over power plant areas for four NO2 instruments. R

2 is the squareof linear correlation coefficient between NOx emissions and TVCDs. (bottom) Relationship between R2 and the proportion of power plants’emissions to total NOx emissions ( f power). Error bars express the ranges of R2 in each 0.1 interval of f power. Data points are arbitrarily fitted byquadratic polynomials, and the goodness of fit is shown as r2.

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707467

implying that the treatment of TVCD using eq 2 is reasonable.Among the four instruments, GOME has the worst perform-ance because of the large pixel size and the relatively low NOxemissions before 2003. Positive intercepts are observed in allscatter plots, and they are probably caused by a combined effectof the exclusion or underestimation of NOx emissions fromsources other than power plants, the import of NO2 fromnearby areas, and the uncertainties in both satellite retrievalsand the emission inventory. The correlations are even better forareas dominated by power plant emissions. As shown in Figure3e−h, R2 values are significantly improved to 0.69−0.91 forareas with f power > 0.85 (∼ 12 power plant areas). To furtherexplore the relationship between NO2 columns and NOxemissions over power plant areas, we show the variation ofR2 with f power in Figure 3i−l. The higher the proportion ofpower plant emissions to total NOx emissions, the better theagreement between NO2 columns and NOx emissions. This isconsistent with previous modeling results of NO2 columns overpower plant areas in the western U.S. and China,13,37 indicatingthat area emissions (normally allocated using various spatialproxies such as population density and road networks) are notas well understood as power plant emissions.Trends of Tropospheric NO2 over Thermal Power

Plant Areas in India during FY 1996−2010. We calculatethe total NO2 burden (Btotal) and NO2 burden attributed topower plant emissions (Bpower) over all Indian power plant areasas

∑= ×B A(TVCD )n

n ntotal(or power) total(or power),(3)

where An is the surface area of power plant area n. Whendivided by ∑An, Btotal (or power) can be further converted to anaverage NO2 TVCD over all power plant areas. Figure 4 showsthe monthly variations and annual evolution of Btotal and Bpower(as well as the corresponding average NO2 TVCD) for the fourNO2 instruments during FY 1996−2010. Theoretically, thetemporal overlap of operation periods provides the possibilityof cross-validation between different instruments; however, dueto the differences in the satellite characteristics, measurementtime, and retrieval algorithms, a comprehensive cross-validationis problematic and beyond the scope of this study. Here, weonly focus on the trends of NO2 retrievals. As shown in Figure4, in line with the dramatic increase of NOx emissions in theIndian power sector, tropospheric NO2 increased during FY1996−2010 with AAGRs of 1.5−3.2% for Btotal and 3.3−5.4%for Bpower, depending on the instrument. Although power plantscontributed only ∼50% to the NO2 burden, they accounted formore than 90% of the burden increment over power plant areasand dominated the NO2 burden trends. Figure 4 also lists thecorrelation coefficients between NO2 burden and unit-basedNOx emission estimates for different instruments. The trends ofNO2 burden are in good agreement with the trends of NOxemissions from the power sector, especially for Bpower (R

2 =0.74−0.98), further confirming the close relationship betweenNO2 columns and NOx emissions over power plant areas.The lifetime of NOx depends on its own concentration,

because reacting with OH is the main sink of NOx, while OHconcentration in the atmosphere depends strongly on NOxconcentration. Taking into account the nonlinear feedback ofNOx emissions on NOx chemistry, Lamsal et al.38 used adimensionless factor β to establish the relationship betweenchanges in surface NOx emissions and changes in NO2 TVCD:

β =Δ

ΔE E/

TVCD /TVCDtotal total

total total (4)

β tends to be greater than one in clean (low NO2) regions,since an increase in NOx emissions increases the OHconcentration through chemical feedback and thus decreasesthe NOx lifetime. In polluted (high NO2) regions, β tends to beless than one, since increased NOx emissions consume OH andincrease the NOx lifetime. In this study, we calculated the βvalues in different periods for four instruments, and the resultsare listed in Table 2. Although there are discrepancies in βbetween different instruments for a certain period, β valuesdecreased from ∼2 to ∼0.7 during FY 1996−2010. This impliesthat the overall NOx chemistry over Indian power plant areashas changed considerably in the past few years due to thedramatic increase in NOx emissions from power plants, as wellas an increase from transportation since many of the powerplants in India are located in the populated places wherevehicular emissions are high.21,33 It should be noted that thechange of VOCs emissions might also affect the NOx−VOCs−O3 chemistry, and further affect β. However, increasing theVOCs emissions by 15% only increases β by 2.8%,38 implyingthat β values are not sensitive to VOCs emissions. Also, basedon the GAINS model,2 VOCs emissions in India increased byonly 6.5% during 2000−2010. Therefore, the overall NOxchemistry change over Indian power plant areas is mainlyattributed to the dramatic increase of NOx emissions. Table 2shows that the transition took place approximately between

Figure 4. Monthly variations and annual evolutions of NO2 burden(Btotal and Bpower, left axis) and corresponding average NO2 TVCD(right axis) over all Indian power plant areas for four NO2 instrumentsduring FY 1996−2010. Annual values are averaged according to IndianFY, and months in monsoon period are excluded. R2 values shown arethe square of correlation coefficients of NO2 burden with unit-basedNOx emission estimates. The dependence of OH concentration andNO2 lifetime (top axis) on NO2 TVCD (right axis) is derived from aHOx−NOx steady-state model, and is replotted according to Figure 1of Valin et al.39.

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707468

2005 and 2008, and this can be confirmed by the dependencecurve of OH concentration and NO2 lifetime on NO2 TVCDshown in Figure 4, which is derived from a HOx−NOx steady-state model.39 Increasing NO2 at low NO2 condition (β > 1)results in an increase in OH and a decrease in NO2 lifetime,while the situation is opposite at high NO2 condition (β < 1).The transition occurred at a NO2 TVCD of ∼2.7 × 1015

molecules cm−2, exactly falling in the period 2005−2008. Thisdiscussion is focused on the overall atmospheric characteristicsover all power plant areas. The NOx chemistry of individualareas with high/low NOx emissions could have changedbefore/after 2005−2008.Our results clearly demonstrate a very high growth of surface

NO2 caused by a continuous increase in NOx emissions frompower plants in India during 1996−2010. More importantly,the overall NOx chemistry over Indian power plants changedaround 2005−2008. Increasing the same proportion of NOx

emissions leads to a greater surface NO2 increase now than inthe past, implying that NOx pollution in India becomes moreand more serious. Undoubtedly, it has led to a change in thetropospheric chemistry relating ozone and aerosol formationand consequently has affected human health, air quality,atmospheric physics, climate forcing, etc. This should befurther investigated with atmospheric chemical transportmodels at local and regional scales. A further implication ofboth the growth in emissions and the change in NOx chemistryis that regulations on NOx emissions from coal-fired powerplants are urgently needed in India, and high-efficiency NOx

control technologies (e.g., selective catalytic reduction) arestrongly recommended to policy makers.

■ ASSOCIATED CONTENT

*S Supporting Information(1) Electricity generation of thermal power plants (Figure S1);(2) Spatial distribution of power plants NOx emissions andNO2 TVCDs (Figures S2−S4); (3) Relationship betweenpower plants NOx emissions and NO2 TVCDs (Figures S5−S8); (4) Characteristics of four NO2 instruments (Table S1);(5) Detailed results of NOx emissions (Tables S2−S3). Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*Phone: (630) 252-9853; fax (630) 252-9868; e-mail: zlu@anl.gov.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

This work was sponsored by the National Aeronautics andSpace Administration (NASA) as part of the Air QualityApplied Sciences Team (AQAST) program. The India emissioninventory was partially funded in support of the Ganges ValleyAerosol Experiment (GVAX) by the Office of Biological andEnvironmental Research in the U.S. Department of Energy,Office of Science. We acknowledge the free use of troposphericNO2 column data from www.temis.nl. Argonne NationalLaboratory is operated by UChicago Argonne, LLC, underContract DE-AC02-06CH11357 with the U.S. Department ofEnergy.

■ REFERENCES(1) Emission Database for Global Atmospheric Research (EDGAR)release version 4.2; Joint Research Centre (JRC)/NetherlandsEnvironmental Assessment Agency (PBL); http://edgar.jrc.ec.europa.eu.(2) Klimont, Z.; Cofala, J.; Xing, J.; Wei, W.; Zhang, C.; Wang, S.;Kejun, J.; Bhandari, P.; Mathur, R.; Purohit, P.; Rafaj, P.; Chambers,A.; Amann, M. Projections of SO2, NOx and carbonaceous aerosolsemissions in Asia. Tellus Ser. B 2009, 61, 602−617.(3) Ohara, T.; Akimoto, H.; Kurokawa, J.; Horii, N.; Yamaji, K.; Yan,X.; Hayasaka, T. An Asian emission inventory of anthropogenicemission sources for the period 1980−2020. Atmos. Chem. Phys. 2007,7, 4419−4444.(4) Streets, D. G.; Bond, T. C.; Carmichael, G. R.; Fernandes, S. D.;Fu, Q.; He, D.; Klimont, Z.; Nelson, S. M.; Tsai, N. Y.; Wang, M. Q.;Woo, J. H.; Yarber, K. F. An inventory of gaseous and primary aerosolemissions in Asia in the year 2000. J. Geophys. Res. 2003, 108, 8809DOI: 10.1029/2002JD003093.(5) Zhang, Q.; Streets, D. G.; Carmichael, G. R.; He, K. B.; Huo, H.;Kannari, A.; Klimont, Z.; Park, I. S.; Reddy, S.; Fu, J. S.; Chen, D.;Duan, L.; Lei, Y.; Wang, L. T.; Yao, Z. L. Asian emissions in 2006 forthe NASA INTEX-B mission. Atmos. Chem. Phys. 2009, 9, 5131−5153.(6) Garg, A.; Shukla, P. R.; Kapshe, M. The sectoral trends ofmultigas emissions inventory of India. Atmos. Environ. 2006, 40, 4608−4620.(7) Hemispheric Transport of Air Pollution-Emission Database forGlobal Atmospheric Research (HTAP-EDGAR); http://edgar.jrc.it/eolo/.(8) Intergovernmental Panel on Climate Change (IPCC). IPCCGood Practice Guidance and Uncertainty Management in NationalGreenhouse Gas Inventories; Cambridge University Press: New York,1996.(9) Kim, S. W.; Heckel, A.; McKeen, S. A.; Frost, G. J.; Hsie, E. Y.;Trainer, M. K.; Richter, A.; Burrows, J. P.; Peckham, S. E.; Grell, G. A.Satellite-observed US power plant NOx emission reductions and theirimpact on air quality. Geophys. Res. Lett. 2006, 33, L22812DOI: 10.1029/2006GL027749.(10) Martin, R. V.; Jacob, D. J.; Chance, K.; Kurosu, T. P.; Palmer, P.I.; Evans, M. J. Global inventory of nitrogen oxide emissionsconstrained by space-based observations of NO2 columns. J. Geophys.Res. 2003, 108, 4537 DOI: 10.1029/2003JD003453.

Table 2. Estimated β Values for Four NO2 Instruments in Different Emission Scenarios

β

period instrument S1 S2 S3 S4 S5

FY 1996−2002 GOME 1.62 1.65 1.65 1.49 1.60FY 2003−2005 SCIAMACHY 2.29 2.41 2.20 2.22 2.24FY 2005−2007 SCIAMACHY 1.04 1.09 1.02 0.97 1.00

OMI 2.38 2.48 2.36 2.21 2.40FY 2007−2010 SCIAMACHY 0.69 0.71 0.68 0.65 0.56

OMI 0.84 0.86 0.84 0.81 0.71GOME-2 0.69 0.71 0.68 0.65 0.57

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707469

(11) Richter, A.; Burrows, J. P.; Nuss, H.; Granier, C.; Niemeier, U.Increase in tropospheric nitrogen dioxide over China observed fromspace. Nature 2005, 437, 129−132.(12) Zhang, Q.; Streets, D. G.; He, K. B. Satellite observations ofrecent power plant construction in Inner Mongolia, China. Geophys.Res. Lett. 2009, 36, L15809 DOI: 10.1029/2009GL038984.(13) Kim, S. W.; Heckel, A.; Frost, G. J.; Richter, A.; Gleason, J.;Burrows, J. P.; McKeen, S.; Hsie, E. Y.; Granier, C.; Trainer, M. NO2

columns in the western United States observed from space andsimulated by a regional chemistry model and their implications forNOx emissions. J. Geophys. Res. 2009, 114, D11301 DOI: 10.1029/2008JD011343.(14) Kaynak, B.; Hu, Y.; Martin, R. V.; Sioris, C. E.; Russell, A. G.Comparison of weekly cycle of NO2 satellite retrievals and NOx

emission inventories for the continental United States. J. Geophys. Res.2009, 114, D05302 DOI: 10.1029/2008JD010714.(15) Zhang, Q.; Streets, D. G.; He, K.; Wang, Y.; Richter, A.;Burrows, J. P.; Uno, I.; Jang, C. J.; Chen, D.; Yao, Z.; Lei, Y. NOx

emission trends for China, 1995−2004: The view from the ground andthe view from space. J. Geophys. Res. 2007, 112, D22306DOI: 10.1029/2007JD008684.(16) Boersma, K. F.; Jacob, D. J.; Eskes, H. J.; Pinder, R. W.; Wang,J.; van der A, R. J. Intercomparison of SCIAMACHY and OMItropospheric NO2 columns: Observing the diurnal evolution ofchemistry and emissions from space. J. Geophys. Res. 2008, 113,D16S26 DOI: 10.1029/2007JD008816.(17) Novotny, E. V.; Bechle, M. J.; Millet, D. B.; Marshall, J. D.National satellite-based land-use regression: NO2 in the United States.Environ. Sci. Technol. 2011, 45, 4407−4414.(18) Yang, Q.; Wang, Y. H.; Zhao, C.; Liu, Z.; Gustafson, W. I.; Shao,M. NOx emission reduction and its effects on ozone during the 2008Olympic Games. Environ. Sci. Technol. 2011, 45, 6404−6410.(19) van der A, R. J.; Peters, D. H. M. U.; Eskes, H.; Boersma, K. F.;Van Roozendael, M.; De Smedt, I.; Kelder, H. M. Detection of thetrend and seasonal variation in tropospheric NO2 over China. J.Geophys. Res. 2006, 111, D12317 DOI: 10.1029/2005JD006594.(20) Russell, A. R.; Valin, L. C.; Bucsela, E. J.; Wenig, M. O.; Cohen,R. C. Space-based constraints on spatial and temporal patterns of NOx

emissions in California, 2005−2008. Environ. Sci. Technol. 2010, 44,3608−3615.(21) Ghude, S. D.; Fadnavis, S.; Beig, G.; Polade, S. D.; van der A, R.J. Detection of surface emission hot spots, trends, and seasonal cyclefrom satellite-retrieved NO2 over India. J. Geophys. Res. 2008, 113,D20305 DOI: 10.1029/2007JD009615.(22) Ghude, S. D.; Kulkarni, P. S.; Kulkarni, S. H.; Fadnavis, S.; vander A, R. J. Temporal variation of urban NOx concentration in Indiaduring the past decade as observed from space. Int. J. Remote Sens.2011, 32, 849−861.(23) Prasad, A. K.; Singh, R. P.; Kafatos, M. Influence of coal-basedthermal power plants on the spatial-temporal variability of tropo-spheric NO2 column over India. Environ. Monit. Assess. 2012, 184,1891−1907.(24) Central Electricity Authority (CEA). Performance Review ofThermal Power Stations 1997−2011; Central Electricity Authority,Ministry of Power, Government of India: New Delhi, 1996−2010.(25) International Energy Agency (IEA). Energy Statistics of Non-OECD Countries; Paris, 2011.(26) Chikkatur, A. P.; Sagar, A. D. Cleaner Power in India: Towards aClean-Coal-Technology Roadmap. Discussion Paper 2007−06; BelferCenter for Science and International Affairs: Cambridge, MA, 2007.(27) Intergovernmental Panel on Climate Change (IPCC). EmissionFactor Database (EFDB); http://www.ipcc-nggip.iges.or.jp/EFDB/main.php.(28) Zhao, Y.; Wang, S. X.; Duan, L.; Lei, Y.; Cao, P. F.; Hao, J. M.Primary air pollutant emissions of coal-fired power plants in China:Current status and future prediction. Atmos. Environ. 2008, 42, 8442−8452.(29) Zhao, Y.; Wang, S. X.; Nielsen, C. P.; Li, X. H.; Hao, J. M.Establishment of a database of emission factors for atmospheric

pollutants from Chinese coal-fired power plants. Atmos. Environ. 2010,44, 1515−1523.(30) Central Pollution Control Board (CPCB), Government ofIndia; http://www.cpcb.nic.in/Industry_Specific_Standards.php.(31) Boersma, K. F.; Eskes, H. J.; Brinksma, E. J. Error analysis fortropospheric NO2 retrieval from space. J. Geophys. Res. 2004, 109,D04311 DOI: 10.1029/2003JD003962.(32) Boersma, K. F.; Eskes, H. J.; Veefkind, J. P.; Brinksma, E. J.; vander A, R. J.; Sneep, M.; van den Oord, G. H. J.; Levelt, P. F.; Stammes,P.; Gleason, J. F.; Bucsela, E. J. Near-real time retrieval of troposphericNO2 from OMI. Atmos. Chem. Phys. 2007, 7, 2103−2118.(33) Lu, Z.; Zhang, Q.; Streets, D. G. Sulfur dioxide and primarycarbonaceous aerosol emissions in China and India, 1996−2010.Atmos. Chem. Phys. 2011, 11, 9839−9864.(34) Lu, Z. F.; Streets, D. G.; Zhang, Q.; Wang, S. W. A novel back-trajectory analysis of the origin of black carbon transported to theHimalayas and Tibetan Plateau during 1996−2010. Geophys. Res. Lett.2012, 39, L01809 DOI: 10.1029/2011GL049903.(35) Wang, S. W.; Streets, D. G.; Zhang, Q. A.; He, K. B.; Chen, D.;Kang, S. C.; Lu, Z. F.; Wang, Y. X. Satellite detection and modelverification of NOx emissions from power plants in Northern China.Environ. Res. Lett. 2010, 5, 044007 DOI: 10.1088/1748-9326/5/4/044007.(36) Krishnamurthy, V.; Shukla, J. Intraseasonal and interannualvariability of rainfall over India. J. Clim. 2000, 13, 4366−4377.(37) Wang, S. W.; Zhang, Q.; Streets, D. G.; He, K. B.; Martin, R. V.;Lamsal, L. N.; Chen, D.; Lei, Y.; Lu, Z. Growth in NOx emissions frompower plants in China: Bottom-up estimates and satellite observations.Atmos. Chem. Phys. 2012, 12, 4429−4447.(38) Lamsal, L. N.; Martin, R. V.; Padmanabhan, A.; van Donkelaar,A.; Zhang, Q.; Sioris, C. E.; Chance, K.; Kurosu, T. P.; Newchurch, M.J. Application of satellite observations for timely updates to globalanthropogenic NOx emission inventories. Geophys. Res. Lett. 2011, 38,L05810 DOI: 10.1029/2010GL046476.(39) Valin, L. C.; Russell, A. R.; Hudman, R. C.; Cohen, R. C. Effectsof model resolution on the interpretation of satellite NO2observations. Atmos. Chem. Phys. 2011, 11, 11647−11655.

Environmental Science & Technology Article

dx.doi.org/10.1021/es300831w | Environ. Sci. Technol. 2012, 46, 7463−74707470