intercomparison of sciamachy and omi - harvard...

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. ???, XXXX, DOI:10.1029/,

Intercomparison of SCIAMACHY and OMI1

tropospheric NO2 columns: observing the diurnal2

evolution of chemistry and emissions from space3

K. Folkert Boersma,1

Daniel J. Jacob,1

Henk J. Eskes,2

Robert W. Pinder,3

Jun Wang,1

Ronald J. van der A2

K. F. Boersma, Division of Engineering and Applied Sciences, Harvard University, Pierce Hall,

29 Oxford St., Cambridge, MA 02155, USA. ([email protected])

1Division of Engineering and Applied

Sciences, Harvard University, Cambridge,

Massachusetts, USA.

2Royal Netherlands Meteorological

Institute, De Bilt, the Netherlands.

3Atmospheric Sciences Modeling Division,

Air Resources Laboratory, National Oceanic

and Atmospheric Administration, in

partnership with National Exposure

Research Laboratory, U.S. Environmental

Protection Agency, Raleigh, North Carolina,

USA.

D R A F T April 15, 2007, 10:46am D R A F T

X - 2 BOERSMA ET AL.: SCIAMACHY VS. OMI TROPOSPHERIC NO2

Abstract. Concurrent satellite measurements of tropospheric NO2 columns4

from OMI aboard Aura (13:30 local overpass time) and SCIAMACHY aboard5

Envisat (10:00 local overpass time) offer an opportunity to examine the con-6

sistency between the two instruments and the effect of different observing7

times. For scenes with tropospheric NO2 columns <5.0×1015molec.cm−2, SCIA-8

MACHY and OMI agree within 1.0–2.0×1015molec.cm−2, consistent with the9

detection limits of both instruments. We find evidence for a low bias of 0.2×1015molec.cm−210

in OMI observations over remote oceans. Over the fossil fuel source regions11

at northern mid-latitudes, we find that SCIAMACHY observes up to 40%12

higher NO2 at 10:00 local time than OMI at 13:30. Over biomass burning13

regions in the tropics, SCIAMACHY observes up to 40% lower NO2 columns14

than OMI. These differences are present in the spectral fitting of the data15

(slant column), and are augmented in the fossil fuel regions, and dampened16

in the tropical biomass burning regions by the expected increase in air mass17

factor as the mixing depth rises from 10:00 to 13:30. Using a global 3-D chem-18

ical transport model (GEOS-Chem), we show that the 10:00–13:30 decrease19

in tropospheric NO2 column over fossil fuel source regions can be explained20

by photochemical loss, dampened by the diurnal cycle of anthropogenic emis-21

sions that has a broad daytime maximum. The observed NO2 column increase22

from 10:00 to 13:30 over tropical biomass burning regions points to a sharp23

midday peak in emissions, and is consistent with a diurnal cycle of emissions24

derived from geostationary satellite fire counts.25


BOERSMA ET AL.: SCIAMACHY VS. OMI TROPOSPHERIC NO2 X - 3

1. Introduction

Nitrogen oxides (NOx = NO + NO2) play an important role in atmospheric chemistry.26

Near the Earth’s surface, emissions of NOx in the presence of hydrocarbons and sunlight27

lead to the urban and regional-scale formation of ozone. In the presence of ammonia28

(NH3), oxidation of NOx to nitric acid leads to the formation of aerosol nitrate (NH4NO3).29

Both ozone as well as fine particles at ground levels have a detrimental effect on public30

health. Moreover, there is a general interest in NOx as the limiting precursor for ozone in31

the global troposphere, as ozone is an efficient greenhouse gas and a source of hydroxyl32

(OH). To obtain a better understanding of how NOx is influencing air pollution and33

contributing to the greenhouse effect, accurate and precise measurements of NOx over34

large spatial domains with sufficient temporal sampling are indispensable.35

Ground-based and air-borne measurements of NOx are temporally and spatially limited.36

Satellite instruments provide measurements with global coverage and fixed spatial and37

temporal resolution. Data sets of tropospheric NO2 columns retrieved from the Global38

Ozone Monitoring Experiment (GOME), the Scanning Imaging Absorption Cartography39

(SCIAMACHY) and the Ozone Monitoring Experiment (OMI) now span more than ten40

years (1996-2006) and have been used for trend studies [Richter et al., 2005; van der A et41

al., 2006]. GOME and SCIAMACHY tropospheric NO2 data has been used to estimate42

surface emissions of NOx [Leue et al., 2001; Martin et al., 2003; Jaegle et al., 2004, 2005;43

Muller and Stavrakou, 2005; Bertram et al., 2005; Toenges-Schuller et al., 2006; Konovalov44

et al., 2006; Martin et al., 2006; Kim et al., 2006] and lightning NOx production [Boersma45

et al., 2005; Beirle et al., 2006; Martin et al., 2007].46



The latest generation of satellite instruments observes the Earth with horizontal reso-47

lutions that allow a detailed view of the NOx pollution patterns. SCIAMACHY [Bovens-48

mann et al., 1999] on board ESA’s Envisat has a horizontal resolution of 30 × 60 km249

and OMI [Levelt et al., 2006] on board NASA’s EOS-Aura satellite has a nadir pixel50

size of 13 × 24 km2. There is a growing interest from various environmental and public51

health agencies to start using these satellite observations to verify emission inventories,52

investigate trends in emissions, and forecast air quality. It is important to obtain confi-53

dence in the satellite observations by validation and error characterization. The quality54

of SCIAMACHY NO2 data retrieved by KNMI/BIRA has previously been validated by55

Petritoli et al. [2005]; Schaub et al. [2007] and Blond et al. [2007], with SCIAMACHY56

columns generally being within 20% of the validation data. Furthermore, OMI NO2 re-57

trievals from KNMI [Boersma et al., 2006] are shown to be consistent within 5% with58

aircraft vertical profiles during the INTEX-B campaign in situations where assumptions59

on the unobserved part of the troposphere are limited [Boersma et al., 2007].60

We present here an intercomparison of SCIAMACHY and OMI observations using sim-61

ilar KNMI retrievals. Of particular interest is to interpret the comparison in terms of the62

implied diurnal variation of NOx emissions and chemistry. Both SCIAMACHY and OMI63

are in sun-synchronous orbits, with local overpass times of 10:00 and 13:30 local time64

respectively. NOx emissions from transport are expected to peak during the morning65

commute, while power plants and industrial sources have in general little diurnal varia-66

tion. Fire and soil emissions are expected to peak in the middle of the day. Chemical67

loss of NOx also peaks in the middle of the day. Resolving the compounded effects of68

these different diurnal variations is critical for the application of satellite observations69



to estimate NOx emissions, although this has not been properly recognized in previous70

studies. The different overpass times of SCIAMACHY and OMI offer an opportunity to71

test our current understanding.72

2. SCIAMACHY and OMI tropospheric NO2 retrievals

2.1. Common algorithm

SCIAMACHY and OMI are UV/Vis spectrometers that measure direct sunlight and73

backscattered light from the Earth’s atmosphere. Slant columns of NO2 are obtained by74

fitting the absorption cross-section spectra of NO2 to the observed reflectance spectra75

(defined as the ratio between the Earth radiance and solar irradiance spectra) around 44076

nm. The retrieved slant column represents the column of NO2 (in molecules per cm2)77

along the average photon path from the Sun, through the atmosphere, to the satellite78

instrument.79

To convert retrieved slant columns into vertical, tropospheric NO2 columns (Nv), we

follow the approach described in Boersma et al. [2004]:

Nv =Ns −Ns,st

Mtr

, (1)

where Ns,st is the stratospheric component of the NO2 slant column Ns, and Mtr represents80

the tropospheric air mass factor (AMF) that depends on cloud fraction, cloud pressure,81

surface albedo, and the a priori NO2 profile.82

For both SCIAMACHY and OMI retrievals, we estimate the stratospheric slant column83

by data assimilation of NO2 slant columns in the TM4 chemical transport model (CTM)84

[Dentener et al., 2003]. The estimate of stratospheric NO2 thus adjusts to observed NO285

over regions dominated by stratospheric NO2 (e.g. oceans), while stratospheric winds in86



TM4 ensure the propagation of the assimilated information. See Boersma et al. [2006] for87

more details.88

Removal of the stratospheric slant column Ns,st from the total slant column Ns defines89

the tropospheric slant column, which we convert to the tropospheric vertical column by90

applying the tropospheric AMF (Mtr) as in Eq. 1. We compute tropospheric AMFs91

following the formulation of Palmer et al. [2001] and Eskes and Boersma [2003] with the92

DAK radiative transfer model [Stammes, 2001], and take into account the effects of clouds93

(from the SCIAMACHY FRESCO [Koelemeijer et al., 2001] and OMI O2-O2 [Acarreta94

et al., 2004] cloud algorithms), surface albedo (from the Herman and Celarier [1997] and95

Koelemeijer et al. [2002] data sets as described in Boersma et al. [2004]), and the a priori96

vertical profile shape from TM4. No independent aerosol correction is applied as it is97

effectively accounted for in the cloud retrievals [Boersma et al., 2004].98

A crucial aspect in comparing SCIAMACHY and OMI NO2 sets, is that they be re-99

trieved in a consistent manner. Slant columns have been retrieved by non-linear least-100

square fitting of backscattered radiance spectra for both SCIAMACHY (by BIRA) and101

OMI (by KNMI/NASA). For both retrievals, we use the data assimilation of slant columns102

in TM4 to estimate the stratospheric columns. The AMFs have been computed with the103

same radiative transfer model, the same surface albedo dataset, and the same TM4 model104

to obtain vertical profile shapes. Some aspects pertaining to the retrievals are character-105

istic to each instrument, and we give here a short overview of the main differences.106

2.2. Differences between SCIAMACHY and OMI

2.2.1. Instrumental differences107



Table 1 gives an overview of essential instrumental characteristics of the SCIAMACHY108

and OMI tropospheric NO2 retrievals. Important differences are:109

• The horizontal resolution of SCIAMACHY nadir viewing scenes (pixels) when the110

instrument is in nominal operation mode is 30 × 60 km2. The horizontal resolution of111

OMI in nominal operations mode is 13 × 24 km2 for scenes observed with a viewing angle112

of 0◦ (nadir) to approximately 150 × 24 km2 for scenes observed at the edges of the swath113

with a viewing angle of 57◦ (see Figure 5 in Levelt et al. [2006]).114

• SCIAMACHY alternately takes limb and nadir measurements. The observations in115

the limb mode are not sensitive to the troposphere so that tropospheric (nadir) NO2116

retrievals from SCIAMACHY are not continuous. This, combined with SCIAMACHY’s117

field of view that corresponds to 1000 km on the Earth’s surface, leads to a global cov-118

erage that is achieved within approximately 6 days. OMI has a 57◦ field of view, which119

corresponds to an approximately 2600 km wide swath on the Earth’s surface. Because of120

the wide swath of the 14–15 orbits per day, OMI achieves global coverage in one day.121

• SCIAMACHY measures direct and backscattered solar radiation between 220 nm and122

2380 nm with a spectral resolution of 0.44 nm around 440 nm. OMI measures between123

270 and 500 nm, so that cloud retrieval using the O2-A band (760 nm) is not possible124

with OMI, and the O2-O2 absorption feature (470 nm) is used instead.125

2.2.2. Retrieval differences126

Table 2 summarizes the main differences in the fitting procedures followed for SCIA-127

MACHY and OMI. NO2 retrievals from SCIAMACHY are the result of a collaboration128

between KNMI and BIRA (Belgian Institute for Space Aeronomy). Slant columns are129

obtained by BIRA by fitting a modelled spectrum to the measured reflectance in the130



426.3–451.3 nm region. For OMI, a wider fit window (405–465 nm) is used to account for131

lower signal-to-noise ratios (∼1400 for nominal mid latitude conditions) than in SCIA-132

MACHY measurements (signal-to-noise ∼2000). In Boersma et al. [2002], we did not find133

a strong sensitivity of the fitting results for various combinations of fitting windows, fitting134

techniques, or inclusion of species with minor absorption features such as H2O and O2-O2.135

Different NO2 absorption spectra are used in SCIAMACHY [Bogumil et al., 2003] and136

OMI [Vandaele et al., 1998] but the differences are less than 2% at all temperatures. We137

do not expect significant differences between SCIAMACHY and OMI total slant columns138

from the fitting differences in Table 2.139

SCIAMACHY reflectances have been found to be systematically underestimated by140

approximately 13% [Acarreta and Stammes, 2005], largely due to radiometric offsets in141

solar irradiances [Skupin et al., 2005]. In the KNMI/BIRA retrieval, NO2 slant columns142

are therefore retrieved by spectral fitting with an Earth radiance spectrum observed over143

the Indian Ocean as the reference spectrum. This reference spectrum is assumed to144

contain the signature of a 1.5×1015 molec. cm−2 stratospheric NO2 column [van der A et145

al., 2006]. For OMI, a solar irradiance spectrum is used for the spectral fitting procedure,146

and a correction for a NO2 signature in the reference spectrum is not necessary.147

Calibration errors in the OMI reflectance measurements lead to spurious across-track148

variability. This variability is significantly reduced, but not removed altogether, by the149

correction procedure described in Boersma et al. [2006].150

Cloud parameters for SCIAMACHY and OMI are retrieved with different algorithms.151

The FRESCO and O2-O2 cloud algorithms have been compared in Boersma et al. [2006],152

which showed that mean cloud fractions were on average the same within 0.011. Cloud153



pressures from the O2-O2 algorithm (OMI) were found to be higher (+58 hPa) than from154

FRESCO (SCIAMACHY).155

3. Intercomparison of SCIAMACHY and OMI tropospheric NO2 columns

Figure 1 shows the monthly mean tropospheric NO2 fields derived from SCIAMACHY156

and OMI for August 2006, averaged over 0.5◦ × 0.5◦ grid cells. We averaged data for all157

days on which both SCIAMACHY and OMI took a cloud-free observation (cloud radiance158

less than 50%) in the same 0.5◦ × 0.5◦ grid cell.159

The spatial distributions of both have a correlation coefficient r=0.77 (n=1.9×105).160

Both observe the highest values over the urban regions of North America, Europe, and161

China, and also high levels over the biomass burning regions of South America, southern162

Africa, and Indonesia. The OMI mean tropospheric NO2 field is much smoother than for163

SCIAMACHY because it is based on more pixels.164

Figure 2 shows absolute differences between the SCIAMACHY and OMI monthly mean165

tropospheric NO2 columns for August 2006, and Figure 4 shows the corresponding scat-166

terplot. We estimate a 1-sigma uncertainty for individual retrievals as the sum of as a167

base absolute component (from spectral fitting and the stratospheric background) and a168

relative component from the AMF [Boersma et al., 2004, 2006]. For OMI, the base compo-169

nent is 0.5–1.5×1015 molec.cm−2 (depending on the AMF). For SCIAMACHY retrievals,170

which have a better fitting precision, the base component is 0.3–1.0×1015 molec.cm−2.171

SCIAMACHY is in general higher than OMI over fossil fuel source regions, and this is172

particularly pronounced in large urban areas such as megacities (Los Angeles, Mexico City,173

Moscow, Riyadh, Tehran, Hong Kong). This would be expected from the combination of174

morning peak in emissions and midday maximum in NOx chemical loss. In contrast, OMI175



observes higher NO2 columns over biomass burning regions in Brazil, southern Africa, and176

Indonesia, and we attribute this to a midday peak in emissions as discussed below. Higher177

OMI than SCIAMACHY values are also sometimes seen downwind of isolated urban178

areas (this is particularly manifest for Riyadh and Johannesburg), which could reflect179

transport of the urban plume. Another area where OMI is higher than SCIAMACHY is180

the southeastern United States, and possible reasons for this will be discussed below.181

Figure 5(a) shows the probability distribution functions (PDFs) for the monthly mean182

SCIAMACHY and OMI data sets. The median of the OMI PDF is left-shifted by 0.2×183

1015 molec.cm−2 relative to SCIAMACHY. This bias is a background feature, as shown in184

Figure 5(b) by a PDF subset for the clean Pacific Ocean. We find that OMI NO2 within185

this region is on average 0.17 × 1015 molec.cm−2 lower than SCIAMACHY, and the mean186

is negative. We conclude that OMI suffers from a small negative bias over unpolluted187

regions. We find a similar negative bias from a validation study over the remote Gulf of188

Mexico [Boersma et al., 2007, this issue]. This bias is inconsequential for polluted regions.189

Negative tropospheric NO2 columns may occur if the slant stratospheric column ex-190

ceeds the total slant column (see Eq. 1). This would be caused by bias in the assimilation191

procedure used for the stratospheric column estimate. Figure 3 shows that SCIAMACHY192

(normalized) total slant columns are systematically smaller than OMI (normalized) total193

slant columns, i.e. the distribution of differences peaks at -0.59 × 1015 molec.cm−2. This194

is likely due to the differences in reference spectrum for the SCIAMACHY (backscat-195

ter spectrum with residual NO2 signature) and OMI (solar irradiance spectrum) spectral196

fitting procedures (see section 2.2.2). Figure 3 also shows that the differences between197

SCIAMACHY and OMI total slant colums are closely followed by the differences be-198



tween SCIAMACHY and OMI (geometric AMF-normalized) stratospheric slant columns.199

On average, (SCIAMACHY-OMI) stratospheric slant column differences are -0.58×1015200

molec.cm−2, very close to slant column differences. This is expected from the assimila-201

tion procedure: biased slant columns propagate into similarly-biased stratospheric slant202

columns (described in section 2.1). The relative bias between SCIAMACHY and OMI203

cancels (Eq. 1) and only marginally affects the tropospheric slant columns for any of the204

two data sets. This is confirmed by the curve with the solid line in Figure 3 that shows a205

median difference corresponding to 0.0×1015 molec.cm−2.206

4. SCIAMACHY vs. OMI differences in NOx source regions

We now investigate the cause of the large differences between OMI and SCIAMACHY

over polluted and biomass burning regions (Figure 2). A first issue is to determine

whether they are driven by differences in tropospheric slant columns (Ns-Ns,st, herafter

slant columns) or in AMFs. Difference in the slant columns points to information in the

actual spectra, while difference in the AMF points to information in the physics of the

retrieval (such as higher mixing depths at 13:30 than at 10:00). As slant columns and

AMFs are both proportional to the length of the light path, we normalize them by the

local geometrical AMF, (Mg=1

cos(θ)+ 1

cos(θ0)where θ is the satellite viewing angle and θ0

is the solar zenith angle) to remove the simple effect of optical geometry. Tropospheric

vertical columns (Nv) can thus be expressed as:

Nv =Ns −Ns,st

Mg

Mg

Mtr

(2)



Here, Ns−Ns,st

Mgand Mtr

Mgare the normalized slant columns, and normalized AMFs, respec-207

tively. We successively examine these two terms, focusing on the NOx source regions208

delineated by boxes in Figure 2.209

Table 3 shows the ratio of observed SCIAMACHY and OMI tropospheric NO2 columns210

for regions with strong NOx sources where we observe the largest differences between211

SCIAMACHY and OMI tropospheric NO2 columns , and the contributions from the slant212

column and the AMF, for the source regions of Figure 2. For the fossil fuel source regions213

where SCIAMACHY is higher than OMI, slant columns and AMFs both make comparable214

compounding contributions to the difference in tropospheric columns. For the biomass215

burning and southeastern United States regions where OMI is higher than SCIAMACHY,216

the difference is mainly in the slant columns with only a small effect from the AMFs.217

The AMF difference between SCIAMACHY and OMI deserves some discussion. Both218

AMF calculations use the same surface albedos (Herman and Celarier [1997] and Koele-219

meijer et al. [2002]) and the cloud fractions are similar within 1%. Thus the normalized220

AMF differences in Table 3 must reflect either differences in cloud pressures or in the221

NO2 vertical shape profile. The OMI and SCIAMACHY cloud retrieval algorithms show222

a small difference in cloud pressures, with OMI clouds having cloud pressures on aver-223

age higher by 60 hPa [Boersma et al., 2006]. Since temporal variation in global cloud224

fraction and cloud pressure between 10:00 and 13:30 is small [Bergman and Salby, 1996],225

this difference represents an algorithmic bias. Figure 5(b) in Boersma et al. [2004] shows226

increasing AMFs with increasing cloud pressure, especially for situations with low clouds227

and polluted boundary layers. As we exclude scenes with cloud radiance fractions >50%,228

the impact of systematically different cloud pressures on AMF calculations remains lim-229



ited. We computed AMFs with different cloud pressures (OMI +60 hPa) but otherwise230

identical forward model parameters (albedo, cloud fraction, and profile shape) as in the231

actual retrievals for August 2006, and find OMI AMFs to be higher by up to 10% for232

scenes with high NO2.233

Both SCIAMACHY and OMI use the TM4 CTM to estimate vertical NO2 profiles, and234

growth of the atmospheric mixed layer between 10:00 and 13:30 local time would lead to a235

greater vertical extent of NO2 seen by OMI and hence a higher AMF. In TM4, boundary236

layer mixing follows the vertical diffusion parameterisation of Holtslag and Moeng [1991].237

Diffusion coefficients and mixing depths are updated every 3 hours as described in Krol238

et al. [2005]. We compared tropospheric AMFs computed with NO2 profiles sampled at239

10:00 and at 13:30 local time, but otherwise identical forward model parameters (albedo240

and cloud inputs) as in the actual retrievals for August 2006. We find tropospheric AMFs241

to be up to 15% larger at 13:30 than at 10:00 local time.242

In conclusion, it is clear that the differences between SCIAMACHY and OMI in source243

regions cannot be ascribed to a retrieval artifact. Most of the difference is in the spectra.244

Although the retrieval (through the AMF) contributes to yielding lower tropospheric245

columns in OMI for a given slant column, this mostly reflects well understood physics of246

diurnal mixed layer growth. We now turn to explaining these differences on the basis of247

diurnal variation in NOx emissions and chemistry.248

5. Diurnal variations in tropospheric NO2 columns

5.1. Model vs. satellite observations

We use the global GEOS-Chem CTM (see appendix) to examine if the 10:00–13:30 dif-

ferences in tropospheric NO2 columns observed by SCIAMACHY vs. OMI are consistent



with our understanding of diurnal variations in NOx emissions and chemical loss. Con-

ceptually, and in the absence of transport, Eq. 3 shows that, neglecting transport, the

diurnal variation of tropospheric NO2 columns (Nv) depends on the diurnal cycle of NOx

emissions and chemical loss:

dNv

dt= α(t)(E(t)− k(t)Nx(t)), (3)

with α(t) the NO2:NOx column ratio, E(t) the NOx emission strength, and k(t) the249

chemical loss rate constant for NOx columns Nx(t) as a function of time-of-day t. Figure 6250

conceptually illustrates the normalized diurnal variation of Nv(t), α(t), E(t) and k(t)251

for a typical fossil fuel source region, where k(t) mainly represents the chemical loss of252

NOx to HNO3 (through the gas phase NO2+OH reaction and by hydrolysis of N2O5253

in aerosols). Since α(t) shows little difference between 10:00 and 13:30, NO2 columns254

increase in time if emissions add more NOx to the atmosphere than is chemically lost255

through k(t)Nx(t). Chemical loss occurs throughout the diurnal cycle but is strongest256

at midday, when OH concentrations are highest. Over urban regions, emissions show257

a broad daytime maximum (E(t) represents the U.S. mean diurnal variation from the258

Environmental Protection Agency [1989] in 4-hourly steps as applied worldwide in GEOS-259

Chem), whereas the integrated chemical loss between 10:00 and 13:30 is larger than the260

diurnal average, so that we expect NO2 columns to be higher at 10:00 than at 13:30. Over261

regions dominated by biomass burning or soil NOx emissions, emissions E(t) are likely262

higher in the afternoon than in the morning, and 10:00–13:30 column differences are263

expected to be smaller than over urban regions or even negative. The GEOS-Chem CTM264

includes all of the processes described above together with transport for a description of265



how NO2 tropospheric columns would be expected to vary between 10:00 and 13:30 local266

time on the basis of current understanding.267

5.2. Diurnal variation in NO2 columns over fossil fuel source regions

We compare the GEOS-Chem simulated diurnal variations in tropospheric NO2 columns268

to the OMI–SCIAMACHY results for the different NOx source regions in Figure 2). The269

diurnal cycle of emissions is shown in blue in the left panel of Figure 7: strongest emissions270

take place between 06:00 and 18:00 local time, related to highest anthropogenic activity,271

without much variation within that time frame.272

For observations and simulations to be comparable at the GEOS-Chem horizontal res-273

olution of 2◦ latitude × 2.5◦ longitude, we require that 2◦ × 2.5◦ grid cells were sampled274

only on days when more than 50% of the (2◦ × 2.5◦) geographical cell area was covered275

by SCIAMACHY and OMI observations with (observed) cloud radiances <50%. This276

corresponds to more than 52 OMI pixels and more than 14 SCIAMACHY pixels within277

a 2◦ × 2.5◦ grid cell.278

Figure 7 (left panel) shows the average diurnal cycle of tropospheric NO2 columns279

simulated by GEOS-Chem over the northeastern United States for August 2006. NO2280

columns in the simulation with constant emissions show a minimum in the late afternoon,281

reflecting the diurnal cycle of chemical loss. The standard simulation with diurnally282

varying emissions has a weaker minimum a few hours earlier, and a less pronounced283

diurnal cycle (10:00–13:30 ratio 1.31, constant emissions: 1.59). It agrees better with the284

SCIAMACHY vs. OMI observations (ratio 1.18).285

Table 3 summarizes the results for all fossil fuel source regions at northern mid-latitudes286

of Figure 2. GEOS-Chem simulates a 10:00–13:30 NO2 column ratio of 1.3 for all regions,287



reflecting the model uniformity in the diurnal cycle of fossil fuel emissions and the similar-288

ity in the chemistry. It overestimates the observed 10:00–13:30 ratios in the northeastern289

United States and northern Europe, reproduces observations in northeastern China, and290

gets the trend wrong in the southeastern United States where the observed ratio is less291

than 1.292

For comparison, we analyse the diurnal trend of observed surface NO2 concentrations293

observed at EPA Aerometric Information Retrieval System (AIRS) urban sites for the294

summer of 2004. For six stations in the northeastern United States (Chicago, Detroit,295

Duluth, New York, Pittsburgh, and Philadelphia), the average 10:00–13:00 ratio in NO2296

concentrations is 1.43, close to the August 2006 GEOS-Chem ratio. A 10:00–13:30 ra-297

tio >1.0 has also been found in summertime tropospheric NO2 columns obtained with298

DOAS zenith-sky observations in Bologna, Italy [Petritoli et al., 2005], especially during299

weekdays.300

The larger simulated than observed 10:00–13:30 ratio over the northern United States301

can likely be reconciled with an improved diurnal cycle of emissions that has recently be-302

come available. Figure 7 compares summertime normalized hourly emissions from the EPA303

National Emissions Inventory 2001 over the northeastern United States with the 4-hour304

diurnal variation from the Environmental Protection Agency [1989] used in GEOS-Chem.305

In NEI2001, the emissions in the hours before 10:00 are lower than in the hours before306

13:30, whereas they are similar in Environmental Protection Agency [1989], implying a307

smaller 10:00–13:30 ratio with the improved diurnal cycle of emissions (Eq. 3).308

Urban areas in the southeastern United States (Houston, Dallas, and Atlanta, see Fig-309

ure 2) show 10:00–13:30 ratios >1.0, similar to the other fossil fuel regions. But for most of310



the rural Southeast, NO2 is lower at 10:00 than at 13:30. We hypothesize that the opposite311

diurnal variation between the observed (0.70) and modelled (1.31) 10:00–13:30 ratios over312

the southeastern United States likely reflects a different diurnal pattern of emissions in313

the rural Southeast in summer. The most important NOx sources in the rural Southeast314

are power plants, soils, biomass burning, and lightning. According to CEM (Continuous315

Emission Monitoring by the U.S. EPA) data for August 2006, the third warmest August316

since instrumental records began [NOAA, 2006], NOx power plant emissions in the south-317

eastern United States are highest in the afternoon. Soil NOx emissions (dependent on soil318

temperature) are strongest in the afternoon [Ludwig et al., 2001], and there is evidence319

that these emissions are significantly underestimated in the model [Jaegle et al., 2005].320

Furthermore, Park et al. [2007] found strong biomass burning activity throughout the321

Southeast and GEOS-Chem does not take into account the diurnal cycle of fires (see next322

section). In summary, all these NOx sources have a strong afternoon maximum and either323

their source strength (soils, lightning [Hudman et al., 2007]) or their diurnal cycle (power324

plants, biomass burning) is underestimated. The observed 10:00–13:30 ratios <1.0 over325

the southeastern U.S. appear consistent throughout the summer as we also found them for326

July 2006 (not shown), while we did not find them for the non-summer month of October327

2004 (not shown). Obviously, there is a need for more aloft measurements to help explain328

the differences between simulated and observed 10:00–13:30 differences in NO2 over the329

southeastern United States.330

5.3. Diurnal variation in NO2 columns over biomass burning regions

High NO2 columns in southern Africa, Brazil, and Indonesia in Fig. 1 reflect consid-331

erable biomass burning activity during the local dry season. Biomass burning dominates332



over soil NOx emissions or lightning NOx production in the dry season [Jaegle et al.,333

2005]. In these regions, NO2 columns from OMI are higher than from SCIAMACHY,334

but the standard GEOS-Chem simulation with no diurnal variation in biomass burning335

emissions shows the reverse. Recent satellite- and ground-based observations have shown336

that there is a distinct and consistent diurnal cycle in the agricultural practices driving337

tropical biomass burning (Prins et al. [1998]; Kauffman et al. [2003]; Giglio et al. [2003])338

with a minimum at night and a maximum in the early to late afternoon. Therefore we339

implemented a diurnal cycle for biomass burning emissions in GEOS-Chem. This cycle is340

based on hourly fire counts over Central America detected by the Geostationary Opera-341

tional Environmental Satellite (GOES-8) for 2002 [Wang et al., 2006] and it is consistent342

with the diurnal distribution of fires observed by GOES-8 over South America in August343

1995 [Prins et al., 1998]. The four-hour normalized emission factors that we adopted are344

shown in blue in Figure 7 (right panel).345

This figure also shows the diurnal variation of NO2 columns over southern Africa in346

August 2006, modelled by GEOS-Chem vs. observed by SCIAMACHY and OMI. We see347

that the observed 10:00–13:30 increase can be reproduced by GEOS-Chem by accounting348

for the diurnal variation of emissions. Table 3 shows that similar results are obtained349

over Brazil and Indonesia, though the simulated 10:00–13:30 ratio is still not as low350

as observed , possibly because even stronger diurnal variability needs to be included in351

biomass burning emissions [Wang et al., 2006].352

A number of studies (Martin et al. [2003]; Muller and Stavrakou [2005]; Jaegle et al.353

[2005]) have used GOME or SCIAMACHY NO2 observations in the 10:00–10:30 local time354

window as top-down constraint on NOx emissions from biomass burning. These studies355



report emission estimates lower than bottom-up estimates based on carbon burned and356

NOx emission ratios, but this likely reflects in fact the unaccounted diurnal variation of357

fire emissions. A mid-morning local overpass times is sub-optimal for the observation of358

fires, and any top-down analysis needs to account for the large diurnal variation of fire359

emissions. Using our imposed diurnal cycle in the right panel of Fig. 7, we find that360

a top-down constraint on biomass burning emissions based on 10:00 observations would361

have to be corrected upward by 35% to account for this diurnal cycle. This difference is362

of the same order as the decreases reported on by Martin et al. [2003] and Muller and363

Stavrakou [2005], which supports the a priori inventories (6-7 Tg N yr−1 globally) that364

they used, rather than than the a posteriori optimized values that they derived.365

6. Summary and conclusions

We have compared satellite retrievals of tropospheric NO2 columns from two different366

instruments (SCIAMACHY and OMI) using similar retrieval methods for August 2006.367

The purpose of this comparison is to (1) gain confidence in the OMI NO2 data product368

(available from http://www.temis.nl), (2) examine if OMI NO2 data can continue the369

(long-term record of tropospheric NO2 columns initiated by GOME and) SCIAMACHY370

to monitor NOx emission trends [van der A et al., 2006], and (3) examine the information371

offered by satellite NO2 observations at different times of day as additional constraint372

on the type and diurnal variation of NOx emissions. An essential difference between373

SCIAMACHY and OMI observations is that they are taken at different times of day374

(SCIAMACHY - 10:00, OMI - 13:30 local time).375

The comparison of SCIAMACHY and OMI retrievals for locations and days when both376

instruments observed clear-sky scenes (less than 50% cloud radiance fraction) shows that377



the instruments observe similar spatial patterns of NO2 (r=0.77). Differences between378

SCIAMACHY and OMI outside of the major source areas (tropospheric NO2 column379

<5.0× 1015 molec.cm−2) are generally within the measurement uncertainties of both in-380

struments (1.0–2.0× 1015 molec.cm−2). We find evidence for a low bias of approximately381

0.2 × 1015 molec.cm−2 in the OMI retrievals over oceans.382

SCIAMACHY observes higher NO2 than OMI (up to 40%) in most industrial regions383

of northern mid-latitudes, while it observes lower NO2 than OMI (up to 35%) in tropical384

biomass burning regions. We show that these observed differences are not due to retrieval385

errors. Most of the difference is in the slant column, which simply reflects spectral fitting.386

There is also some contribution from the air mass factor (AMF) describing mixed layer387

growth from 10:00 to 13:30, but it is small (typically less than 10%) and in any case it388

has some physical basis. Cloud pressures in OMI are 60 hPa lower than in SCIAMACHY,389

which could lead to some small AMF differences; these are difficult to characterize and390

are in any case less than 10%.391

We used a global 3-D chemical transport model (GEOS-Chem) to interpret the 10:00–392

13:30 diurnal trends in NO2 columns seen by SCIAMACHY and OMI as driven by diurnal393

variations in chemistry and emissions. Chemistry alone would cause higher NO2 columns394

by a factor of 1.5 at 10:00 than at 13:30 over source regions in the tropics and in mid-395

latitudes summer, reflecting the photochemical sink from oxidation by OH. We find from396

GEOS-Chem that this diurnal variation is dampened by typically a third due to diur-397

nal variation in fossil fuel emissions with a broad daytime maximum. This decrease is398

consistent with 20-40% decreases observed by SCIAMACHY and OMI for the same days399

and locations, and furthermore with 10:00–13:00 decreases in NO2 surface concentrations400



observed by AIRS in summertime urban regions over the northeastern United States.401

Over urban regions in the southeastern United States, SCIAMACHY and OMI observe402

a 10:00–13:30 decrease that is similar to the other fossil fuel source regions. But over403

the Southeast, the observed 10:00–13:30 ratio (0.70) is opposite to the modelled (1.31).404

The most important NOx sources in the rural Southeast are power plants, soils, biomass405

burning, and lightning, and all these have strongest emissions in the afternoon. We hy-406

pothesize that these different diurnal patterns of emissions may explain the observed NO2407

column increase from 10:00 to 13:30.408

The 35% increase in tropospheric NO2 columns between 10:00 and 13:30 local time over409

tropical biomass burning regions is opposite of the diurnal cycle from photochemical loss.410

A diurnal cycle for biomass burning emissions that accounts for much stronger afternoon411

fire activity can resolve that discrepancy and is consistent with geostationary satellite412

observations of fire counts. Such a cycle is generally not included in chemical transport413

models but it should be. Top-down estimates on biomass burning NOx emissions based414

on NO2 columns observed at 10:00–10:30 (SCIAMACHY, GOME) increase by 35% if the415

diurnal variation in fire emissions is taken into account. Previous studies that did not416

account for this diurnal variation in their top-down constraints incurred a corresponding417

error, leading them to conclude erroneously that prior bottom-up emission estimates were418

too high.419

Scaling factors between SCIAMACHY and OMI can be derived in order to continue420

trend analyses from GOME and SCIAMACHY into the OMI era. In the future, scaling421

factors (10:00–13:30 ratios) should be evaluated for the complete seasonal cycle. There422

is a need to better understand the 10:00–13:30 ratios in terms of the diurnal variation in423



emissions (and possibly chemistry) in order to warrant continuity between the GOME,424

SCIAMACHY, and OMI tropospheric NO2 data sets.425

Appendix A: The GEOS-Chem CTM

For simulation of the diurnal variation of tropospheric NO2 columns, we use the426

2.5◦ × 2.0◦ version of the 3-D chemical transport model GEOS-Chem (v7-04-06; www-427

as.harvard.edu/chemistry/trop/geos/index.html) [Bey et al., 2001; Park et al., 2004].428

GEOS-Chem has been used previously for inverse modelling of NOx emissions using GOMe429

and SCIAMACHY satellite data by Martin et al. [2003],Jaegle et al. [2005],Martin et al.430

[2006], and Sauvage et al. [2006] and we refer to these studies for more detailed descriptions431

of the model.432

GEOS-Chem is driven by assimilated meteorological fields (GEOS-4) from the NASA433

Global Modeling Assimilation Office (GMAO) that are provided every 6 hours (3 hours434

for surface fields and mixing depths). There are 55 vertical sigma levels, extending up435

from the surface to 0.01 hPa, and including 5 levels in the lowest 2 km of the atmosphere.436

The chemical simulations are conducted for August 2006 after a 7-month initialization.437

GEOS-Chem uses national emission inventories for anthropogenic NOx where avail-438

able, and otherwise default values from the Global Emission Inventory Activity (GEIA)439

[Benkovitz et al., 1996] scaled to 1998 [Bey et al., 2001]. For Europe we use the most recent440

NOx emissions (2004) from the European Monitoring and Evaluation Program (EMEP)441

following Auvray and Bey [2005]. Over the U.S., (NOx) emissions are taken from the442

EPA 1999 National Emissions Inventory (NEI99, U.S. Environmental Protection Agency443

[2001]). Over Mexico, emissions are from BRAVO [2004]. The biomass burning inventory444

is based on satellite observations of fires by van der Werf et al. [2006], and emission factors445



from Andreae and Merlet [2001]. Soil NOx emissions are computed following Yienger and446

Levy [1995] with canopy reduction factors described by Wang et al. [1998]. The produc-447

tion of NOx by lightning is based on the cloud-top scheme of Price and Rind [1992] and448

vertical distribution from Pickering et al. [1998] with a global source of 4.7 Tg N yr−1449

GEOS-Chem includes a detailed simulation of O3-NOx-hydrocarbon-aerosol chemistry450

in detail. The aerosol and gaseous simulations are coupled through the formation of451

sulphate and nitrate, the HNO3/NO−3 partitioning of total inorganic nitrate, and the452

uptake of N2O5 by aerosols in the presence of water vapor (the main night-time sink of453

NOx, modelled as in Evans and Jacob [2005]). The chemical timestep in the model is 1454

hour, short enough to sample the 3.5 hour time difference between the SCIAMACHY and455

OMI overpass times.456

Acknowledgments. This work was funded by the NASA Atmospheric Composition457

Modeling and Analysis Program. The OMI project is managed by NIVR and KNMI in458

The Netherlands. The OMI NRT tropospheric NO2 data are obtained from the KNMI459

TEMIS website (http://www.temis.nl). The NASA GSFC SIPS team is thanked for the460

production of the NRT NO2 slant columns. The research presented here was performed461

under the Memorandum of Understanding between the U.S. Environmental Protection462

Agency (EPA) and the U.S. Department of Commerce’s National Oceanic and Atmo-463

spheric Administration (NOAA) and under agreement DW13921548. This work consti-464

tutes a contribution to the NOAA Air Quality Program. Although it has been reviewed465

by EPA and NOAA and approved for publication, it does not necessarily reflect their466

policies or views.467



References

Acarreta, J. R., J. F. De Haan, and P. Stammes (2004), Cloud pressure retrieval468

using the O2-O2 absorption band at 477 nm, J. Geophys. Res., 109, D05204,469

doi:10.1029/2003JD003915.470

Acarreta, J. R., and P. Stammes (2005), Calibration Comparison Between SCIA-471

MACHY and MERIS Onboard ENVISAT, IEEE Geosc. Rem. Sens. Lett., 2(1), 31–35,472

doi:10.1109/LGRS.2004.838348.473

Andreae, M. O. and P. Merlet (2001), Emission of trace gases and aerosols from biomass474

burning, Global Biogeochem. Cycles, 15, 955–966, 2001.475

Auvray M., and I. Bey (2005), Long-Range Transport to Europe: Seasonal Variations476

and Implications for the European Ozone Budget, J. Geophys. Res., 110, D11303,477

doi:10.1029/2004JD005503.478

Benkovitz, C. M., M. T. Scholtz, J. Pacyna, L. Tarrason, J. Dignon, E. C. Voldner,479

P. A. Spiro, J. A. Logan, and T. E. Graedel (1996), Global gridded inventories for480

anthropogenic emissions of sulphur and nitrogen. J. Geophys. Res., 101, 29,239–29,253.481

Beirle, S., N.Spichtinger, A.Stohl, K.L.Cummins, T.Turner, D.Boccippio, O.R.Cooper,482

M.Wenig, M.Grzegorski, U.Platt, and T.Wagner (2006), Estimating the NOx produced483

by lightning from GOME and NLDN data: a case study in the Gulf of Mexico, Atmos.484

Chem. Phys., 6, 1075-1089. SRef-ID: www.atmos-chem-phys.net/6/1075/2006.485

Bergman, J. W., and M. L. Salby (1996), Diurnal Variations of Cloud Cover and Their486

Relationship to Climatological Conditions, J. Climate, 9, 2802–2820.487

Bertram, T. H., A. Heckel, A. Richter, J. P. Burrows, and R. C. Cohen (2005), Satellite488

observations of daily variations in soil NOx emissions, Geophys. Res. Lett., 32, L24812,489



doi:10.1029/2005GL024640.490

Bey, I., D. J. Jacob, R. M. Yantosca, J. A. Logan, B. D. Field, A. M. Fiore, Q. Li, H. Y.491

Liu, L. J. Mickley, and M. G. Schultz (2001), J. Geophys. Res., 106(D9), 23973–23095.492

Blond, N., K. F. Boersma, H. J. Eskes, R. J. van der A, M. Van Roozendael, I. De Smedt,493

G. Bergamatti, and R. Vautard (2007), Intercomparison of SCIAMACHY nitrogen diox-494

ide observations, in situ measurements and air quality modeling results over Western495

Europe, J. Geophys. Res., accepted.496

Boersma, K. F., E. J. Bucsela, E. J. Brinksma, and J. F. Gleason (2002), NO2, in OMI497

Algorithm Theoretical Basis Document, vol. 4, OMI Trace Gas Algorithms, ATB-OMI-498

04, Version 2.0, edited by K. Chance, pp. 1336, NASA Distrib. Active Archive Cent.,499

Greenbelt, Md., Aug. 2002.500

Boersma, K. F., H. J. Eskes, and E. J. Brinksma (2004), Error analysis for tropospheric501

NO2 retrieval from space, J. Geophys. Res., 109, D04311, doi:10.1029/2003JD003962.502

Boersma, K. F., H. J. Eskes, E. W. Meijer, and H. M. Kelder (2005), Estimates of lightning503

NOx production from GOME satellite observations, Atmos. Phys. Chem., 5, 2311-2331.504

SRef-ID: 1680-7324/acp/2005-5-2311.505

Boersma, K. F., H. J. Eskes, J. P. Veefkind, E. J. Brinksma, R. J. van der A, et al. (2006),506

Near-real time retrieval of tropospheric NO2 from OMI, Atmos. Chem. Phys. Discuss.,507

6, 12301-12345. SRef-ID: 1680-7375/acpd/2006-6-12301.508

Boersma, K. F., D. J. Jacob, E. J. Bucsela, A. Perring, R. Dirksen et al., Validation of509

OMI tropospheric NO2 observations during INTEX-B to constrain NOx emissions over510

the contiguous U.S. and Mexico, submitted to J. Geophys. Res..511



Bogumil, K., J. Orphal, T. Homann, S. Voigt, P. Spietz, O. C. Fleischmann, A. Vogel,512

M. Hartmann, H. Bovensmann, J. Frerick, and J. P. Burrows (2003), Measurements513

of molecular absorption spectra with the SCIAMACHY Pre-Flight Model: instrument514

characterization and reference data for atmospheric remote-sensing in the 230–2380 nm515

region, Journal of Photochemistry and Photobiology A: Chemistry 157, 167–184.516

Big Bend Regional Aerosol and Visibility Observational Study Final Report (2004),517

September 2004.518

Bovensmann, H., J. P. Burrows, M. Buchwitz, J. Frerick, S. Noel, V. V. Rozanov, K. V.519

Chance, and A. P. H. Goede (1999), SCIAMACHY: Mission Objectives and Measure-520

ment Modes, J. Atmos. Sci., 56(2), 127150.521

Bucsela, E. J., E. A. Celarier, M. O. Wenig, J. F. Gleason, J. P. Veefkind, K. F. Boersma,522

and E. J. Brinksma (2006), Algorithm for NO2 vertical column retrieval from the Ozone523

Monitoring Instrument, IEEE Trans. Geo. Rem. Sens., 44(5), 1245–1258, Special Issue524

on the EOS Aura Mission.525

Dentener, F., W. Peters, M. Krol, M. van Weele, P. Bergamaschi, and J. Lelieveld, In-526

terannual variability and trend of CH4 lifetime as a measure of OH changes in the527

1979-2003 period, J. Geophys. Res., 108(D15), 4442, doi:10.1029/2002JD002916, 2003.528

Dirksen, R., M.R. Dobber, R. Voors and P. Levelt (2006), Pre-launch characterization529

of the Ozone Monitoring Instrument transfer function in the spectral domain, Appl.530

Optics, 45(17), 3972–3981.531

Environmental Protection Agency (1989), The 1985 NAPAP emission inventory (version532

2): Development of the annual data and modeler’s tapes, EPA-600/7-89-012a, Research533

Triangle Park, N. C.534



Environmental Protection Agency (2002), Nonroad Model, U.S. Environmental Protec-535

tion Agency, Office of Transportation and Air Quality, Ann Arbor, MI, USA, 2002.536

http://www.epa.gov/oms/nonrdmdl.htmi.537

Environmental Protection Agency (2002), User’s Guide to MOBILE6.1 and MO-538

BILE6.2: Mobile Source Emission Factor Model, Office of Transportation and539

Air Quality, Ann Arbor, MI, USA, October 2002; Report No. EPA420-R-02-028.540

hhttp://www.epa.gov/otaq/m6.htmi.541

Electric Reliability Council of Texas (2006), Record Electricity Demand Expected Today,542

www.ercot.com/news/press releases/2006/nr08-17-06-1.html.543

Eskes, H. J., and K. F. Boersma (2003), Averaging kernels for DOAS total-column satellite544

retrievals, Atmos. Chem. Phys., 3, 1285-1291.545

Evans, M. J., and D. J. Jacob (2005), Impact of new laboratory studies of N2O5 hydrolysis546

on global model budgets of tropospheric nitrogen oxides, ozone, and OH, Geophys. Res.547

Lett., 32, L09813, doi:10.1029/2005GL0022469.548

Giglio, L., J. D. Kendall, and R. Mack (2003), A multi-year active fire dataset for the549

tropics derived from TRMM VIRS, Int. J. Remote Sensing, 24 (22), 4505–4525.550

Hao, W. M., and M. H. Liu (1994), Spatial and temporal distribution of tropical biomass551

burning, Global Biogeochem. Cycles, 8, 495–503.552

Herman, J. R., and E. A. Celarier (1997), Earth surface reflectivity climatology at 340–380553

nm from TOMS data, J. Geophys. Res., 102, 2800328011.554

Heue, K.-P., A. Richter, T. Wagner, M. Bruns, J. P. Burrows, C. v. Friedeburg, W.555

D. Lee, U. Platt, I. Pundt, P. Wang (2005), Validation of SCIAMACHY tropospheric556

NO2–columns with AMAXDOAS measurements, Atmos. Chem. Phys., 5, 1039-1051.557



Holtslag, A. A. M., and C.-H. Moeng (1991), Eddy diffusivity and counter-gradient trans-558

port in the convective atmospheric boundary layer, J. Atmos. Sci., 48, 1690–1698.559

Honore, C., R. Vautard, and M. Beekmann (2000), Photochemical regines in urban at-560

mospheres: The influence of dispersion, Geophys. Res. Lett., 27(13), 1895–1898.561

Hudman, R. C., D. J. Jacob, S. Turquety, E. M. Leibensperger, L. T. Murray et al. (2007),562

Surface and lightning sources of nitrogen oxides in the United States: magnitudes,563

chemical evolution, and outflow, J. Geophys. Res., in press.564

Jaegle, L., Martin, R. V., Chance, K. V., Steinberger, L., Kurosu, T. P. Jacob, D. J.,565

Modi, A. I., Yoboue, V., Sigha-Nkamdjou, L., Galy-Lacaux, C. (2004), Satellite map-566

ping of rain-induced nitric oxide emissions from soils, J. Geophys. Res., 109, D21310,567

10.1029/2004JD004787.568

Jaegle, L., L. Steinberger, R. V. Martin, and K. Chance (2005), Global partitioning of NOx569

sources using satellite observations: Relative roles of fossil fuel combustion, biomass570

burning and soil emissions, Faraday Discuss., 130, 407–423, doi:10.1039/b502128f.571

Kauffman, J.B., M.D. Steele, D.L. Cummings, and V.J. Jaramillo (2003), Biomass dy-572

namics associated with deforestation, fire, and conversion to cattle pasture in a Mexican573

tropical dry forest, Forest Ecology and Management, 176, 1-12, 2003.574

Kim, S.-W., A. Heckel, S. A. McKeen, G. J. Frost, E.-Y. Hsie, M. K. Trainer, A.575

Richter, J. P. Burrows, S. E. Peckham, and G. A. Grell (2006), Satellite-observed U.S.576

power plant NOx emission reductions and their impact on air quality, Geophys. Res.577

Lett.,Vol.33(22),L22812, doi:10.1029/2006GL027749.578

Koelemeijer R. B. A., P. Stammes, J.W. Hovenier, and J.F. de Haan (2001), A fast method579

for retrieval of cloud parameters using oxygen A-band measurements from the Global580



Ozone Monitoring Instrument, J. Geophys. Res., 106, 3475–3490.581

Koelemeijer, R. B. A., P. Stammes, J. W. Hovenier, and J. F. de Haan (2002), Global582

distributions of effective cloud fraction and cloud top derived from oxygen A band583

spectra measured by the Global Ozone Monitoring Experiment: Comparison to ISCCP584

data, J. Geophys. Res., 107(D12), 4151, doi:10.1029/2001JD000840, 2002.585

Konovalov, I. B., M. Beekmann, A. Richter, and J. P. Burrows (2006), Inverse modelling586

of the spatial distribution of NOx emissions on a continental scale using satellite data587

(2006), Atmos. Chem. Phys., 6, 1747–1770.588

Krol, M., S. Houweling, B. Bregman, M. van den Broek, A. Segers, P. van Velthoven, W.589

Peters, F. Dentener, and P. Bergamaschi (2005), The two-way nested global chemistry-590

transport zoom model TM5: algorithm and applications, Atmos. Chem. Phys., 5, 417–591

432.592

Leue, C., M. Wenig, T. Wagner, O. Klimm, U. Platt, and B. Jahne (2001), Quantitative593

analysis of NOx emissions from Global Ozone Monitoring Experiment satellite image594

sequences, J. Geophys. Res., 106, 5493-5505.595

Levelt, P. F., G. H. J. van den Oord, M. R. Dobber, A. Malkki, H. Visser, J. de Vries,596

P. Stammes, J. O. V. Lundell, and H. Saari (2006), The Ozone Monitoring Instrument,597

IEEE Trans. Geosci. Remote Sens., 44(5), 10931101.598

Li, Q., D. J. Jacob, R. Park, Y. Wang, C. L. Heald, R. Hudman, R. V. Martin,599

and M. Evans (2005), North American pollution outflow and the trapping of con-600

vectively lifted pollution by upper-level anticyclone, J. Geophys. Res., 110, D10301,601

doi:10.1029/2004JD005039.602



Lobert, J. M., W. C. Keene, J. A. Logan, and R. Yevich (1999), Global chlorine emis-603

sions from biomass burning: Reactive Chlorine Emission Inventory, J. Geophys. Res.,604

104(D7), 83738390.605

Ludwig, J. F. X. Meixner, B. Vogel, and J. Forster (2001), Soil-air exchange of nitric606

oxide: An overview of processes, environmental factors, and modelling studies, Biogeo-607

chemistry, 52, 225–257.608

Martin, R.V., K. Chance, D.J. Jacob, T.P. Kurosu, R.J.D. Spurr, E. Bucsela, J.F. Gleason,609

P.I. Palmer, I. Bey, A.M. Fiore, Q. Li, R.M. Yantosca, and R.B.A. Koelemeijer (2002),610

An improved retrieval of tropospheric nitrogen dioxide from GOME, J. Geophys. Res.,611

107(D20), 4437, 10.1029/2001JD001027.612

Martin, R. V., D. J. Jacob, K. V. Chance, T. P. Kurosu, P. I. Palmer, and M. J. Evans613

(2003), Global inventory of Nitrogen Dioxide Emissions Constrained by Space-based Ob-614

servations of NO2 Columns, J. Geophys. Res., 108, 4537, doi:10.1029/2003/JD003453.615

Martin, R. V., D. D. Parrish, T. B. Ryerson, D. K. Nicks Jr., K. Chance, T. P. Kurosu,616

D. J. Jacob, E. D. Sturges, A. Fried, and B. P. Wert (2004), Evaluation of GOME617

satellite measurements of tropospheric NO2 and HCHO using regional data from air-618

craft campaigns in the southeastern United States, J. Geophys. Res., 109, D24307,619

doi:10.1029/2004JD004869.620

Martin, R. V., C. E. Sioris, K. Chance, T. B. Ryerson, T. H. Bertram, P. J. Wooldridge,621

R. C. Cohen, J. A. Neuman, A. Swanson, and F. M. Flocke (2006), Evaluation of622

space-based constraints on global nitrogen oxide emissions with regional aircraft mea-623

surements over and downwind of eastern North America, J. Geophys. Res., 111, D15308,624

doi:10.1029/2005JD006680.625



Martin, R.V., B. Sauvage, I. Folkins, C.E. Sioris, C. Boone, P. Bernath, and J.R. Ziemke626

(2007), Space-based constraints on the production of nitric oxide by lightning, J. Geo-627

phys. Res., in press.628

Muller, J.-F., and Stavrakou (2005), Inversion of CO and NOx emissions using the adjoint629

of the IMAGES model, Atmos. Chem. Phys., 5, 1157–1186.630

Murphy, J. G., D. A. Day, P. A. Cleary, P. J. Wooldridge, and R. C. Cohen (2006),631

Observations of the diurnal and seasonal trends in nitrogen oxides in the western Sierra632

Nevada, Atmos. Chem.. Phys., 6, 5321–5338.633

NOAA National Environmental Satellite, Data, and Information Service (2006), Emission634

Inventory Mapviewer, http://map.ngdc.noaa.gov/website/al/emissions/viewer.htm.635

NOAA (2006), www.ncdc.noaa.gov/oa/climate/research/2006/aug/national.html.636

van Noije, T. P. C., H. J. Eskes, F. J. Dentener, D. S. Stevenson, K. Ellingsen, et al.,637

Multi-model ensemble simulations of tropospheric NO2 compared with GOME retrievals638

for the year 2000, Atmos. Chem. Phys., 6, 2943–2979.639

Olivier, J., J. Peters, C. Granier, G. Petron, J.-F. Muller, and S. Wallens (2003), Present640

and future surface emissions of atmospheric compounds, POET Report#2, EU project641

EVK2-1999-00011.642

Palmer, P. I., D. J. Jacob, K. Chance, R. V. Martin, R. J. D. Spurr, T. P. Kurosu, I. Bey,643

R. Yantosca, A. Fiore, and Q. Li (2001), Air-mass factor formulation for spectroscopic644

measurements from satellites: Application to formaldehyde retrievals from the Global645

Ozone Monitoring Experiment, J. Geophys. Res., 106, 14539-14550.646

Park, R. J., D. J. Jacob, B. D. Field, R. M. Yantosca, and M. Chin (2004), Nat-647

ural and transboundary pollution influences on sulphate-nitrate-ammonium aerosols648



in the United States: Implications for policy, J. Geophys. Res., 109, D15204,649

doi:10.1029/2003JD004473.650

Park, R. J., D. J. Jacob, and J. A. Logan (2007), Fire and biofuel contributions to annual651

mean aerosol mass concentrations in the United States, submitted to Atmos. Environ.652

Petritoli, A., E. Palazzi, C. Volta, and G. Giovanelli (2005), Validation of NO2 tropo-653

spheric column from space in the Po valley (Italy), in Tropospheric sounding from654

space ACCENT-TROPOSAT-2 in 2004-5, Burrows, Borrell eds., 308–312.655

Pickering, K. E., Y. Wang, W.-K. Tao, C. Price, J.-F. Muller (1998), Vertical distributions656

of lightning NOx for use in regional and global chemical transport models, J. Geophys.657

Res., 103(D23), 31203-31216, 10.1029/98JD02651.658

Platt, U., Differential optical absorption spectroscopy (DOAS) (1994), in Air monitoring659

by spectroscopic techniques, M.W. Sigrist, Ed., Chemical Analysis Series, 127, John660

Wiley and Sons, Inc661

Price, C., and D. Rind (1992), A simple lightning parameterization for calculating global662

lightning distributions, J. Geophys. Res. 97, 9919–9933.663

Prins, E. M., J. M. Feltz, W. P. Menzel, D. E. Ward (1998), An overview of GOES-8664

diurnal fire and smoke results for SCAR-B and 1995 fire season in South America, J.665

Geophys. Res. Vol. 103(D24), 31821–31835 (98JD01720).666

Quelo, D., V. Mallet, and B. Sportisse (2005), Inverse modelling of NOx emissions at667

regional scale over northern France: Preliminary investigation of the second-order sen-668

sitivity, J. Geophys. Res., 110, D24310, doi:10.1029/2005JD006151.669

Rappengluck, B., P. Oyola, I. Olaeta, P. Fabia (2000), The Evolution of Photochemical670

Smog in the Metropolitan Area of Santiago de Chile, J. Appl. Met., 39, 275–290.671



Richter, A., J. P. Burrows, H. Nuss, C. Granier, and U. Niemeier (2005), Significant672

increases in tropospheric nitrogen dioxide over China observed from space, Nature, 437,673

129–132,doi:10.1038/nature04092.674

Sauvage, B., R.V.Martin, A.vanDonkelaar, X.Liu, K.Chance, L.Jaegle, P.I.Palmer, S.Wu,675

and T.-M.Fu (2006), Remote sensed and in situ constraints on processes affecting trop-676

ical tropospheric ozone, 11465–11520. SRef-ID: 1680-7375/acpd/2006-6-11465.677

Schaub, D., K. F. Boersma, J. Keller, D. Folini, D. Brunner, B. Buchmann, H. Berresheim,678

and J. Staehelin (2007), SCIAMACHY tropospheric NO2 over the Alpine region and679

importance of pixel surface pressure for the column retrieval, Atmos. Chem. Phys.680

Discuss., 7, 429468.681

Skupin, J., S. Noel, M. W. Wuttke, M. Gottwald, H. Bovensmann, M. Weber, and J. P.682

Burrows (2005), SCIAMACHY solar irradiance observation in the spectral range from683

240 to 2380 nm, Adv. Space Res., 35(3), 370–375, doi:10.1016/j.asr.2005.03.036.684

Stammes, P. (2001), Spectral radiance modeling in the UV-Visible range, IRS2000: Cur-685

rent problems in atmospheric radiation, edited by W L. Smith and Y. J. Timofeyev,686

385–388, A. Deepak, Hampton, Va, USA.687

Staudt, A. C., D. J. Jacob, F. Ravetta, J. A. Logan, D. Bachiochi, T. N. Krishna-688

murti, S. Sandholm, B. Ridley, H. B. Singh, and B. Talbot, Sources and chemistry689

of nitrogen oxides over the tropical Pacific (2003), J. Geophys. Res., 108(D2), 8239,690

doi:10.1029/2002JD002139.691

Toenges-Schuller, N., O. Stein, F. Rohrer, A. Wahner, A. Richter, J. P. Burrows, S.692

Beirle, T. Wagner, U. Platt, and C. D. Elvidge (2006), Global distribution pattern of693

anthropogenic nitrogen oxide emissions: Correlation analysis of satellite measurements694



and model calculations, J. Geophys. Res., 111, D05312, doi:10.1029/2005JD006068.695

United States Environmental Protection Agency (U.S. EPA) (2001), National Air Quality696

and Emissions Trends Report, 1999, EPA 454/R-01-004, Off. Of Air Qual. Planning and697

Stand., Research Triangle Park, N.C.698

Vandaele, A. C., C. Hermans, P. C. Simon, M. Carleer, R. Colin, S. Fally, M. F. Merienne,699

A. Jenouvrier, and B. Cocquart (1998), Measurements of the NO2 absorption cross-700

section from 42000cm-1 to 10000 cm-1 (238–1000 nm) at 220 K and 294 K, J. Quant.701

Spectrosc. Radiat. Transfer, 59, 171184.702

van der A, R. J., H. J. Eskes. M. Van Roozendael, I. DeSmedt, N. Blond, K. F. Boersma,703

and A. Weiss (2006), Algorithm Document Tropospheric NO2, TEM/AD1/001, De Bilt,704

the Netherlands, January 2006. http://www.temis.nl/airpollution/no2.html705

Van der Werf, G. R., J. T. Randerson, L. Giglio, G. J. Collatz, P/ S. Kasibhatla, and706

A. F. Arellano (2006), Interannual variability in global biomass burning emissions from707

1997 and 2004, Atmos. Chem. Phys., 6, 3423–3441.708

Wang, Y., D. J. Jacob, and J. A. Logan (1998), Global simulation of tropospheric O3-709

NOxhydrocarbon chemistry: 1. Model formulation, J. Geophys. Res., 103, 10,713–710

10,726.711

Wang, J., S.A. Christopher, U.S. Nair, J.S. Reid, E.M. Prins, J. Szykman, and J.L. Hand712

(2006), Mesoscale modeling of Central American smoke transport to the United States,713

part I: ”top-down” assessment of emission strength and diurnal variation impact, J.714

Geophys. Res., 111, D05S17, doi:10.1029/2005jd006720.715

Yarwood, G., T. E. Stoeckenius, J. G. Heiken, and A. M. Dunker (2003), Modeling Week-716

day/Weekend Ozone Differences in the Los Angeles Region for 1997, J. Air & Waste717



Figure 1. Monthly mean tropospheric NO2 columns in August 2006 from SCIAMACHY and

OMI, in situations when both observe a cloud-free (cloud radiance <50%) scene. Grey grid cells

were not observed or had persistent cloud cover.

Figure 2. Absolute difference between monthly mean SCIAMACHY and OMI tropospheric

NO2 columns for August 2006. Red colours indicate higher retrievals from SCIAMACHY than

from OMI, blue colours vice versa. The black rectangles enclose the regions studied in section 5.

Figure 3. Distribution of SCIAMACHY-OMI differences in total slant columns (dashed line),

in stratospheric slant columns (dashed-dotted line), and in tropospheric slant columns (solid line)

for August 2006. All slant columns have been normalized by their geometrical air mass factors.

Figure 4. Scatterplot of OMI vs. SCIAMACHY tropospheric NO2 columns in August 2006.

Each point represents the monthly mean for a 0.5◦ × 0.5◦ grid cell. The dashed-dotted blue lines

show the error bounds from the combined uncertainties in the SCIAMACHY and OMI retrievals.

The red line shows the OMI running average over 0.5×1015 molec.cm−2 wide SCIAMACHY NO2

bins.

Figure 5. (a) Global probability distribution functions of monthly mean tropospheric NO2

columns for SCIAMACHY and OMI in August 2006. (b) Subset for the Pacific Ocean (180◦W–

90◦W, 45◦S–15◦N) only, and Gaussian curves fitted to the data.

Manage. Assoc., 53, 864–875.718

Yienger, J. J., and H. Levy (1995), Empirical model of global soil-biogenic NOx emissions,719

J. Geophys. Res., 100, 11,447–11,464.720



Figure 6. Conceptual illustration of the diurnal variation of the tropospheric NO2 column

(Nv), NO2:NOx column ratio (α(t)), NOx emissions (E(t)), and chemical loss rate constant (k)

for NOx columns (see Equation 3) in a typical fossil fuel source region (August). Chemical loss

of NOx takes mainly place by the gas phase NO2 + OH reaction and by hydrolysis of N2O5 in

aerosols. The left y-axis holds for Nv and α, and the right y-axis holds for E(t) and k(t).

Figure 7. Average diurnal variation of tropospheric NO2 columns modelled by GEOS-Chem

(black lines), and observed by SCIAMACHY at 10:00 and OMI at 13:30 local time (diamond

symbols) for the northeastern U.S. and African regions of Figure 2 in August 2006. Tropospheric

NO2 columns are normalized relative to the 13:30 value. The solid line indicates the GEOS-Chem

simulation with diurnal variations in NOx emissions. The dashed line indicates the GEOS-Chem

simulation with constant emissions over the diurnal cycle. The solid blue lines shows the diurnal

variation of anthropogenic NOx emissions from the Environmental Protection Agency [1989]

(left), and the diurnal variation of biomass burning emissions based on GOES fire counts [Wang

et al., 2006]. The blue dashed lines show constant emissions.

Table 1. Instrument characteristics relevant for SCIAMACHY and OMI NO2 retrievals

Local equator Nadir Global Spectral SpectralInstrument Satellite crossing time field of view coverage range resolution (at 440 nm)

SCIAMACHY Envisat 10:00 30×60 km2 6 days 220–2380 nm 0.44 nmOMI EOS-Aura 13:30 13×24 km2 1 day 270–500 nm 0.63 nm



Table 2. Spectral fitting characteristics for SCIAMACHY and OMI NO2 retrievals.

Instrument SCIAMACHY OMI

Fitting window 426.3–451.3 nm 405.0–465.0 nmFitted species NO2, O3, Ring, O2-O2, H2O NO2, O3, RingSpectral resolution 0.44 nm 0.65 nmFitting method Non-linear least squares Non-linear least squaresNO2 cross section spectrum SCIAMACHY PFM Vandaele et al. [1998]

[Bogumil et al., 2003] convolved with OMI ITFa

a Dirksen et al. [2006]

Table 3. 10:00–13:30 ratios of tropospheric NO2 columns for selected source regionsa.

Simulation: Simulation: Simulation:Observed Normalized Normalized standard constant diurnal biomass

Regionb column slant columnc AMFc rund emissionse burningf

Northeastern U.S. 1.18 1.06 0.90 1.31 1.59 1.31Northern Europe 1.05 1.03 0.98 1.30 1.50 1.30Northeastern China 1.37 1.25 0.91 1.29 1.46 1.29Southeastern U.S. 0.70 0.72 1.07 1.29 1.44 1.29

Southern Africa 0.69 0.65 0.94 1.50 1.50 0.84Brazil 0.65 0.59 0.90 1.67 1.67 0.89Indonesia 0.63 0.62 1.01 1.76 1.76 1.02

a Monthly mean values for August 2006.

b Regions as indicated in Figure 2

c Tropospheric NO2 slant columns and tropospheric AMFs have been normalized by geo-

metrical AMFs to correct for differences in viewing angles between SCIAMACHY and OMI as

discussed in the text.d Standard GEOS-Chem simulation with diurnally varying anthropogenic emissions and con-

stant biomass burning emissions.e GEOS-Chem simulation with all emissions constant in time.

f GEOS-Chem simulation with diurnally varying anthropogenic [Environmental Protection

Agency, 1989] emissions and biomass burning emissions (Fig. 7).


SCIAMACHY

OMI

-3-2

-10

12

3D

iffer

ence

in s

lant

col

umn

(1015

mol

ec. c

m-2)

0

1•10

4

2•10

4

3•10

4

4•10

4

Number of occurencesS

lant

col

umn

diffe

renc

esS

trat

osph

eric

sla

nt c

olum

n di

ffere

nces

Tro

posp

heric

sla

nt c

olum

n di

ffere

nces

05

1015

2025

30S

CIA

MA

CH

Y tr

opos

pher

ic N

O2

(1015

mol

ec. c

m-2)

051015202530

OMI tropospheric NO2 (1015 molec. cm

-2)

y=x

Err

or b

ound

sR

unni

ng a

vera

ge

05

1015

20T

ropo

sphe

ric N

O2

colu

mn

(1015

mol

ec c

m-2)

110100

1000

1000

0

OccurencesO

MI

SC

IAM

AC

HY

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Tro

posp

heric

NO

2 co

lum

n (1

015 m

olec

cm

-2)

0

500

1000

1500

2000

2500

OccurencesO

MI

SC

IAM

AC

HY

25

811

1417

2023

26Lo

cal t

ime

(hrs

)

0.5

1.0

1.5

0.0

0.0

Normalized Nv and α

0.5

1.0

1.5

2.0

0.0

0.0

Normalized E and k

Nv

α E k

Nor

thea

ster

n U

nite

d S

tate

s

25

811

1417

2023

26Lo

cal t

ime

(hrs

)

0123

Normalized NO2 column

0.5

1.0

1.5

2.0

0.0

0.0

Normalized NOx emission strength

GE

OS

-Che

m E

PA

diu

rnal

em

issi

ons

GE

OS

-Che

m c

onst

ant e

mis

sion

sS

atel

lite

obse

rvat

ions

EP

A d

iurn

al e

mis

sion

sC

onst

ant e

mis

sion

sN

EI2

001

diur

nal e

mis

sion

s

Afr

ica

25

811

1417

2023

26Lo

cal t

ime

(hrs

)

0.0

0.5

1.0

1.5

2.0

2.5

Normalized NO2 column (1014 molec.cm

-2)

024681012

Normalized NOx emission strength

GE

OS

-Che

m d

iurn

al fi

re e

mis

sion

sG

EO

S-C

hem

con

stan

t em

issi

ons

Sat

ellit

e ob

serv

atio

nsD

iurn

al fi

re e

mis

sion

sC

onst

ant e

mis

sion

s

intercomparison of sciamachy and omi - harvard...

Documents