Tropospheric nitrogen dioxide inversions based on spectral
measurements of scattered sunlight
Citation for published version (APA):
Vlemmix, T. (2011). Tropospheric nitrogen dioxide inversions based on spectral measurements of scattered
sunlight. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR719874
DOI:
10.6100/IR719874
Document status and date:
Published: 01/01/2011
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. • Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.
If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy
If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
Printed by Universiteitsdrukkerij TU Eindhoven, Eindhoven, The Netherlands A catalogue record is available from the Eindhoven University of Technology Library Vlemmix, Tim
Tropospheric Nitrogen Dioxide Inversions based on Spectral Measurements of Scattered Sunlight/ by Tim Vlemmix. - Eindhoven: Technische Universiteit Eindhoven, 2011.
-Proefschrift.
ISBN: 978-90-386-3039-7 NUR: 924
Trefwoorden: stikstofdioxide / atmosferische samenstelling / remote sensing / spectroscopie Subject headings: nitrogen dioxide / atmospheric composition / remote sensing / spectroscopy
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magnificus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor Promoties
in het openbaar te verdedigen op donderdag 15 december 2011 om 16.00 uur
door
Tim Vlemmix
prof.dr. P.F. Levelt en
prof.dr. H.M. Kelder
Copromotor: dr. A.J.M. Piters
1 Introduction 1
1.1 Overview of this Chapter 1
1.2 The atmosphere and its interaction with solar radiation 1 1.2.1 General composition of the atmosphere 1
1.2.2 Solar Radiation 4 1.3 The importance of NOx 6
1.3.1 The role of NOxin tropospheric chemistry 7
1.3.2 The role of NOxin relation to climate 9
1.3.3 Effects on human health 10 1.3.4 Sources, sinks and trends 11
1.4 Measurement Techniques for NOxand NO2 12
1.4.1 In-situ techniques 13
1.4.2 Remote sensing from the surface 13 1.4.3 Remote sensing from space 14
1.4.4 Spatial representativity and validation 15 1.5 The retrieval process 16
1.5.1 Forward simulations 17
1.5.2 Inversion and the inverse problem 19 1.6 MAX-DOAS and this thesis 21 1.6.1 MAX-DOAS - a unique instrument 21 1.6.2 Published MAX-DOAS results 22 1.6.3 Research objectives 24
1.6.4 Thesis overview 24
2 MAX-DOAS instrument and DOAS method 27
2.1 Overview of this chapter 27
2.2 Instrument description and characterization 28 2.2.1 Field of view 31
2.3 DOAS method 32 2.3.1 General description 32
2.3.2 Practical implementation of DOAS 35 2.3.3 CINDI results 35
2.4 Interpretation 38
2.4.1 Radiative transfer simulations of slant columns 38 2.4.2 The influence of aerosols on MAX-DOAS observations 41 2.4.3 Inversion methods 43
3 Retrieval of tropospheric NO2columns with a correction for aerosols 49
3.1 Introduction 49
3.1.1 Validation of satellite NO2 49 3.1.2 MAX-DOAS method 50 3.1.3 This paper 52
3.2 Measurements 52
3.2.1 Mini MAX-DOAS instrument 52 3.2.2 Operations 53
3.2.3 DOAS analysis of spectra 54 3.2.4 Relative intensity observations 56 3.3 Retrieval algorithm 56
3.3.1 Radiative transfer modeling 57 3.3.2 Sensitivity study 60
3.3.3 Error sources 63 3.4 Results 65
3.4.1 AOT and tropospheric NO2for selected days 65
3.4.2 Comparison with geometrical air mass factor approximation 67 3.4.3 Verification of AOT with AERONET data 70
3.4.4 Comparison with satellite observations 70 3.5 Conclusions 72
4 Ability of the MAX-DOAS method to derive profile information for NO2 75
4.1 Introduction 76
4.1.1 Background and motivation 76 4.1.2 Profile retrievals with MAX-DOAS 77
4.2 MAX-DOAS measurements and uncertainties 78 4.3 Retrieval algorithm 80
4.3.1 Retrieval model 80 4.3.2 Forward modeling 82 4.3.3 Inversion 85 4.4 Sensitivity study 88
4.4.1 Homogeneous boundary layer 89
4.4.2 Inhomogeneous distribution in boundary layer 90 4.4.3 Elevated layers 92
4.4.4 Limitations of MAX-DOAS profiling potential and dependence on the retrieval approach 97 4.5 Application to measurements performed at the CINDI campaign 98
4.5.1 Retrieval results for selected days 99
4.5.2 Comparison to other measurement techniques for NO2 100
4.6 Summary and Conclusions 105
5 MAX-DOAS tropospheric NO2columns compared with Lotos-Euros 109
5.1 Introduction 109
5.2 MAX-DOAS 112
5.2.1 Measurements 112 5.2.2 Retrieval 112
5.3 Lotos-Euros 117 5.4 Comparison 117
5.4.1 Examples of individual comparisons 118 5.4.2 Quantitative analysis 120
5.4.3 Diurnal, Weekly and Monthly cycles 123 5.4.4 Dependence on meteorological conditions 124 5.5 Conclusions 129
6 Closing Remarks and Outlook 133
6.1 Overview of this Chapter 133 6.2 Retrospective and discussion 133 6.2.1 Columns and profiles 133
6.2.2 Comparisons with other data sets 138
6.2.3 Synthesis of sensors and additional data sources 139 6.2.4 Optimal estimation versus parametrized profiles 140 6.3 Final Conclusions 142 6.4 Outlook 144 6.4.1 Final remark 145 Bibliography 147 Summary 157 Curriculum vitae 161
Introduction
1.1 Overview of this Chapter
Tropospheric nitrogen dioxide (NO2) is a trace gas that plays an important role in atmospheric
chem-istry and is a major component of air pollution. In this thesis a method is studied to measure column amounts and vertical profiles of NO2in the troposphere. The method used is known as the MAX-DOAS
technique, which is a combination of a measurement technique, and a spectral analysis technique: dif-ferential optical absorption spectroscopy (DOAS). The power of the DOAS method lies in its ability to separate the total absorption of light, which is a sum of the absorption by many different species and effects, into fractions that can be ascribed to individual trace gases. DOAS can be applied in many con-texts. Here it is applied to scattered sunlight observations observed from the ground in multiple viewing directions, performed with a MAX-DOAS type of instrument, see Fig. 1.1 and Fig. 3.1. A detailed intro-duction to the MAX-DOAS instruments used and the DOAS technique is given in the following chapter. In this chapter an overview is given of relevant topics which describe the broader context of this work: the research field of atmospheric composition. First some basic information is given about the compo-sition of the atmosphere, and of radiation, i.e. sunlight, which, when measured at the surface, contains information about the atmospheric constituents that absorb in the ultra violet and visible parts of the wavelength spectrum. Subsequently, the role of NO2, and NOx(= NO+NO2) in atmospheric chemistry
is described, and its relevance in the context of air pollution and climate. Different measurement tech-niques for NO2are described and compared in Sect. 1.4. The following section describes the retrieval
process, i.e. the general procedure used to extract the desired information from the measurements. The second to last section focuses on the use of the MAX-DOAS method: what makes it unique, compared to other techniques, and what is the current state of knowledge published in the scientific literature? We end this section with a list of research objectives for this thesis. This chapter is concluded with an overview of the thesis.
1.2 The atmosphere and its interaction with solar radiation
The MAX-DOAS instruments used in this work measure spectra of sunlight scattered by the atmo-sphere. These spectra contain information on the composition of the atmoatmo-sphere. In this section some general background knowledge is provided of the atmosphere and its interaction with solar radiation.
1.2.1 General composition of the atmosphere
The atmosphere is the blanket of air separating the surface of the Earth from the vacuum of outer space. Generally speaking, the atmosphere is in hydrostatic equilibrium: on the one hand its particles do not escape in large numbers, because they are attracted by gravity, on the other hand they do not fall to the
Figure 1.1: MAX-DOAS instruments of three institutes (from left to right: MPI-Mainz, KNMI, BIRA) participating in the Cabauw Intercomparison campaign of Nitrogen Dioxide measuring Instruments (CINDI) 2009. They were mounted at the 20 meter balcony of the more than 200 meters high Cabauw tower, the shadow of which is visible in the field.
surface because they are in the gas phase, repel one another, and take in a volume dictated by tempera-ture and pressure according to the ideal gas law. The atmospheric pressure decreases exponentially with altitude (see Fig. 1.2), whereas the temperature shows a very different altitude dependence. The atmo-sphere is constituted mainly of air, a mixture of nitrogen (78% of its volume for dry air), oxygen (21 %), argon (1%) and carbon dioxide (0.04%). These numbers are very constant over most of the atmosphere. The variability of water vapor is much larger, but on average it is 1% of the total volume. Trace gas con-centrations (see below) are typically several orders of magnitude lower and especially near the surface highly variable.
One can divide the atmosphere in vertical layers (Fig. 1.2): the troposphere, which extends typically from 0 to 9 km at the poles and from 0 to 17 km at the equator, the stratosphere, starting above the troposphere up to about 50 km, and higher layers: mesosphere, thermosphere, and exosphere which are increasingly thin in terms of the gas particle number density.
The troposphere can be separated in a boundary layer and the free troposphere. The boundary layer is the layer closest to the surface. During the day its height is determined for a large part by thermal
convection: the sun heats the surface, and causes air to rise. The rise of air usually stops at a temper-ature inversion (where tempertemper-ature starts to increase with height). This inversion is generally located close to the surface in the morning (about 100 m above the surface) and rises during the day, up to about 500 m in winter, and 2000 m in summer. Air pollution is mostly produced at the surface and because of the temperature inversion, transport to the free troposphere is limited, which explains why free-tropospheric concentrations are usually much lower than boundary layer concentrations. For these reasons, well mixed block-shaped boundary layers are assumed in the simplified atmospheric models in the Chapters 3-5.
The top of the troposphere is called the tropopause height. The separation between troposphere and stratosphere is due to a temperature rise from this altitude onwards which is mainly due to the absorption of sunlight by ozone. Although the atmosphere contains only very little ozone compared to its total volume (if all ozone would be brought to sea level pressure, the ozone layer would be 3mm thick), the presence of ozone is crucial for living species because it absorbs harmful UV radiation. At the same time, when present near the surface, ozone is toxic to humans.
Ozone is one example of an atmospheric trace gas, SO2, CH4, CO, H2O, and NO2are others. They
gen-erally have very low concentrations, but still may have a large impact for example on the radiation (Fig. 1.3), air quality and the chemistry of the atmosphere (Sect. 1.3.1). A typical surface concentration for NO2
is 5 parts per billion. The total amount of a certain gas cannot be expressed in terms of concentration. It is frequently expressed as a vertical column: the total number of molecules above one square centime-ter of surface. Above the Netherlands the average tropospheric NO2column is 13.5⋅ 10
15molec/cm2
. If all those molecules above 1 square centimeter would be brought down to the surface, they would form a layer of 0.005 mm thick. In addition to trace gases, the atmosphere also contains water droplets, ice crystals (clouds), and aerosols (small particles). Trace gases, aerosols and clouds may vary strongly from place to place. The spatial variations in the stratosphere are generally much less pronounced than in the troposphere, which directly receives the emissions from the surface. Sources (e.g. urban regions) and sinks (e.g. removal by rain) of trace gases are not evenly spread across the surface.
A concept that will be frequently referred to in this work is ’the state of the atmosphere’. This will sometimes refer to the real atmosphere, the atmosphere that surrounds us, and on other occasions to the state of the atmosphere as described in (radiative transfer) models (Sect. 1.5.1). It is not straight-forward to give an adequate description of what one imagines as the state of the real atmosphere. The ideal definition would describe for each atom, molecule and photon its quantum mechanical state. It is needless to say that this is a definition without practical use because of the huge number of particles involved. However, we want to use our observations to retrieve at least some aspects of the state of the atmosphere, e.g. the tropospheric NO2column. For this, we use a simplified description of the state
of the atmosphere, which contains sufficient detail to explain those aspects of our measurements that are related to the physical quantity of our interest, e.g. the vertical NO2profile. The complexity of the
three-dimensional atmosphere, which varies constantly with time, thus needs to be reduced. Since the interaction of radiation and matter occurs with the speed of light, we frequently analyze ’snapshots’ of the state of the atmosphere. The spatial simplification is typically done by describing the composition of the atmosphere in one dimension only (the vertical). The bulk of the atmosphere, the air consisting of nitrogen and oxygen molecules, is described in terms of pressure and temperature. Often we use fixed vertical profile shapes that are scaled with the surface temperature and pressure at the time of observa-tion. For trace gases and aerosols, vertical profile descriptions are chosen that are appropriate for the specific problem one wants to solve. The aerosol profile for example is frequently described in this work only by a layer starting at the surface with a certain height and a certain aerosol optical thickness,
as-normalized vertical profile
altitude [km]
Figure 1.2: Normalized vertical profiles (US standard mid latitude summer atmosphere) of temperature (T), pressure (p) and ozone volume mixing ratio (XO3
). Note the logarithmic scale on the vertical axis. The peak values for these profiles are: T=294.2 K, p=1013.0 hPa and 8.9 ppm for XO3
. The pressure de-creases exponentially with altitude, the temperature shows a very different behavior which is partly due to UV absorption in the ozone layer. The transition levels between boundary layer, free troposphere and stratosphere are indicative: the boundary layer height shows a strong diurnal variation, varying from less than a hundred meters after cold nights to approximately 2 km in summer afternoons. The tropopause height, which indicates the bottom of the stratosphere, shows a seasonal and latitudinal variation.
suming homogeneous mixing throughout the layer. This description may be sufficiently accurate for the retrieval of tropospheric NO2columns (Chapters 3 and 5), but it is not accurate enough for the retrieval
of tropospheric NO2profiles (Chapter 4).
1.2.2 Solar Radiation
Solar radiation, observed from the surface of the earth, contains information about the composition of the atmosphere. In this thesis we analyze the information contained in the spectrum of visible (solar) light in order to infer as much information as we can about atmospheric NO2. This type of analysis
is called ’passive remote sensing’, in contrast to ’active remote sensing’ where artificial light sources are used, e.g. a laser. The sun has a surface temperature of 5778K and emits electromagnetic radiation (light) according to Planck’s law, see Fig. 1.3. Part of the light is in the UV (λ< 400 nm) and part in the for the human eye visible part of the electromagnetic spectrum, ’the visible’ (or Vis): 400< λ < 700 nm.
There are generally speaking two types of interaction between radiation and matter: scattering and absorption. In the visible wavelength region (Fig. 1.3) the most dominant effect of the atmosphere on the solar radiation is known as Rayleigh scattering. Molecules in the air (mainly nitrogen and oxygen) cause the incident solar light to scatter in all directions, with a relative amount Φ varying per scattering angle ϕ according to
Φ=3
4⋅ (1 + cos
2ϕ) ,
Figure 1.3: The spectrum of sunlight at the top of the atmosphere and at sea level, computed for a US standard atmosphere with an air mass of 1.5 (solar zenith angle of approx. 48○). The difference between the spectra can mostly be explained by: Rayleigh and Mie scattering [A], Rayleigh, Mie scattering and ozone absorption [B], absorption by water vapor [C], by CO2[D] and by oxygen [E]. Absorption by NO2
(290-600 nm) cannot directly be recognized from this figure. The spectra are modeled with SMARTS2 version 2.9.2 [1, 2], and distributed by the American Society for Testing and Materials (ASTM).
and a total amount quantified by the scattering cross section σRayleigh(λ) [3]. σRayleigh(λ) depends
on the number density of air (Nair=2.4⋅ 1019molec/cm3at 20○C, 1 atm), the wavelength-dependent
refraction index of air n0(=1.000293 at 550 nm) and the wavelength λ according to:
σRayleigh(λ) = 1.061 ⋅ 8π3 3λ4N2 air (n0(λ) 2− 1)2 . (1.2)
If there would be no Rayleigh scattering, i.e. no atmosphere, our sky would be like that of the moon: we would see darkness and only see light from bright white stars (mainly the sun). Rayleigh scattering is stronger for shorter wavelengths, and therefore the specific combination of the Planck spectrum from the sun, the Rayleigh scattering, and the sensitivity of our human eyes, allows us to see the sky in blue. The reddish color of the sun at sunrise and sunset is caused by the same mechanism: from the originally white solar light the blue part is filtered out on its way through the atmosphere to the observer (by Rayleigh scattering) and mostly the long wavelengths survive (the red part).
Another form of scattering is Mie scattering. Rayleigh and Mie scattering both apply to different regimes determined by the ratio of wavelength and particle size. Air molecules and UV/Vis light fall in the Rayleigh regime, whereas UV/Vis light and aerosol or cloud particles fall in the Mie regime. Mie scattering is generally more complex than Rayleigh scattering: the intensity of scattered light varies more strongly with scattering angle (sometimes by several orders of magnitude) and in a way that is more specific to the exact particle size (distribution), geometrical symmetries and chemical composition. The difference between Rayleigh and Mie scattering in the MAX-DOAS (UV/Vis) context is that Rayleigh scattering is always present (it is due to the molecules of the air), well-predictable and usually sufficient to make first order estimates, whereas Mie scattering properties of the atmosphere are much more variable and more difficult to describe quantitatively. The scattering angle dependence of both Rayleigh and Mie
scattering, together with the vertical profiles of scatterers and absorbers, lead to differences with viewing direction of the wavelength spectra detected at the surface. These directional differences are exploited by the MAX-DOAS method (see Chapter 2).
Sofar we have described how the radiation-based information transport mechanism is affected by scattering: scattering changes the direction of propagation of light from the sun to the surface. However, as mentioned above, in the UV/Vis the scattering is mainly caused by molecules in the air, of which NO2
forms only a very tiny fraction. It is therefore not possible to measure the effect of NO2on radiation by
studying its relative contribution on scattering. The ’information’ that is transported is thus not scatter-ing information, but another type, namely absorption (Sect. 1.5.1). The effect of NO2and many other
gases on the observed spectrum in terms of absorption of radiation is far more important than in terms of scattering. The amount of absorption by an individual NO2molecule depends on the wavelength of
incident light, and is described by the absorption cross section (Fig. 2.7). The differential optical absorp-tion spectroscopy (DOAS) method (Chapter 2), used throughout this work, is specifically designed to use the spectral variations in absorption (by NO2and other trace gases) to pull the essential information
out of the measured spectrum.
To summarize: in this study we measure absorption of solar radiation by NO2, relative to a
back-ground of scattered light which is determined mainly by molecular Rayleigh scattering, but also by Mie scattering (clouds and aerosols), which is often equally important, but less predictable.
1.3 The importance of NOx
Atmospheric NOx resides in the stratosphere as well as in the troposphere. The current section will
focus on the latter since the MAX-DOAS instrument, being the main subject of this thesis, is sensitive primarily to this part of the atmosphere (see e.g. Fig. 4.1 in Chapter 4).
With respect to stratospheric NOxit is only relevant to note that it has a different source than
tropo-spheric NOx: it is formed by reaction of stratospheric UV generated oxygen radicals with N2O, which
is a strong greenhouse gas. N2O is an inert gas, which is the reason it can reach the stratosphere
de-spite the fact that it mainly originates from the nitrification and de-nitrification processes in the soil. Stratospheric NOxis thus not directly related to anthropogenic NOx sources discussed in Sect. 1.3.4.
The relevance of stratospheric NOxlies in the fact that it acts as a catalyst for ozone loss, whereas in the
troposphere it acts as a catalyst for ozone formation.
Tropospheric NOxis an important component of air pollution. In addition, since NOxabundances
mostly coincide with a range of other pollutants, NOxor NO2measurements can be interpreted as a
marker for air pollution in general. NOxplays a key role in a variety of tropospheric chemical reactions
or cycles.
In this section the focus will sometimes shift from NOxto NO2and vice versa. Some measurement
techniques are sensitive to NOx, others, mainly optical techniques, are sensitive primarily to NO2.
Dur-ing the daytime NO and NO2are in equilibrium, as will be explained below. If NOxis present, a
signif-icant fraction of it will be in the form of NO2.
The following is for a large part based on the textbooks [3] and [4], both introducing the basics of atmospheric chemistry.
Figure 1.4: Air pollution above Utrecht trapped in a shallow boundary layer after a cold night. NO2is
visible as a brownish gas. Picture: K. Vlemmix.
1.3.1 The role of NOxin tropospheric chemistry
This section gives an overview of tropospheric chemistry, with the focus on the role of NOx. First we
describe some molecular properties of the NO and NO2molecules that are relevant for their chemical
and optical behavior. Then we describe the role of NOxas a catalyst in the formation of tropospheric O3
(an important air quality gas causing health problems, and also a greenhouse gas), and its interaction with OH, a free radical, which is the main oxidizer of tropospheric trace gases. Finally we mention the role of NOxin aerosol formation, through which it indirectly affects air quality and climate.
The NO and NO2molecule
The nitrogen dioxide molecule consists of two oxygen atoms each attached to a nitrogen atom by co-valent bonds, having a bond length of 0.120 nm. The molecule is non-linear with a bond angle of 134○. Since NO2has a critical temperature of 157.8K, it exists in the atmosphere only in the gas phase. At
temperatures below 420 K, NO2exists in equilibrium with N2O4[5]. The NO2molecule essentially has
one unpaired electron and therefore a radical character (high chemical reactivity). NO2absorbs
elec-tromagnetic radiation in the UV and visible, roughly between 250 and 650 nm. This gives the gas NO2a
red-brownish color, which may even be recognized by eye on clear days with high NO2abundances in
a shallow boundary layer, see Fig. 1.4.
The nitrogen oxide molecule has one unpaired electron and is a relatively stable free radical, reacting primarily with molecules having unpaired electrons, typically other free radicals such as OH. In contrast to NO2, it is a colorless gas: it absorbs in the UV, mainly below 210 nm with a peak at 100 nm [see, e.g.,
6, 7]. Since the sun emits little light below 250 nm, and at the surface all UV radiation below 290nm is absorbed by the ozone layer or scattered by Rayleigh scattering (Fig. 1.3), the passive remote sensing technique used in this work is unsuitable to detect NO.
The NOxand HOxcycles
The interaction of several chemical reaction mechanisms, in which NOxplays a key role, leads to the
formation of tropospheric ozone, see Fig. 1.5. Atmospheric ozone resides primarily in the stratosphere (approximately 80% at mid latitudes [8]), where its production is initiated by the photolysis of oxygen (< 240 nm). In the troposphere it is the photolysis of NO2(< 420 nm) which leads to initial formation
NO2+ hν Ð→ NO + O( 3P),
(1.3) which is followed rapidly by the reaction:
O(3P) + O
2+ M Ð→ O3+ M, (1.4)
where M is any other molecule. After its formation, O3is usually oxidized by NO to form NO2again:
O3+ NO Ð→ NO2+ O2 (1.5)
During the day time, NO and NO2are both present in the atmosphere because the formation of NO2is in
equilibrium with its photochemical destruction. Their concentrations are typically within one order of magnitude. During the night, in the absence of sunlight, NOxis present almost solely in the form of NO2.
The dependence of the above relation on the presence of sunlight leads to different photo-stationary regimes, where the O3is related to the ratio of the concentrations of NO and NO2, as expressed by the
Leighton relationship [9]: [NO] [NO2] = j k⋅ [O3] , (1.6)
where j denotes the photolysis frequency of NO2(typical values of j for the Netherlands vary between
0 and 8⋅10−3
s−1, depending on the actinic flux [10], and k denotes the rate constant for the reaction of O3with NO, which is temperature dependent: 7.6⋅10
11⋅ e−4180/RT
cm3molec−1 s−1 [11]. The above reactions do not lead to a net ozone production. However, every other reaction that converts NO to NO2, without oxidizing O3, does lead to a net production of O3(see Fig. 1.5). The interaction of NO
with the hydroperoxyl radical (HO2), provides such a mechanism:
HO2+ NO Ð→ NO2+ OH. (1.7)
This process would not be so relevant if not an additional mechanism would be present to form HO2,
because otherwise the formation of O3would stop after a single reaction between HO2and NO.
How-ever, there is such a mechanism: the reaction of OH with CO or hydrocarbons leads to the production of HO2:
OH+ CO + O2Ð→ CO2+ HO2. (1.8)
The hydroxyl radical OH is a highly reactive radical, with an average lifetime of about 1 second. OH is a strong oxidizer, for example by H-subtraction of H-containing molecules, or by reaction with molecules having oxygen atoms with double bonds. It is initially formed mainly by UV photolysis of O3(λ< 320
nm, and λ< 420 nm), creating electronically excited oxygen atoms from which a certain part will react with water to OH. The HO2formed in reaction 1.8 can, in the presence of NO, be converted back to OH.
NO therefore catalyzes oxidation by OH. In this way it prevents, through OH, CO and hydrocarbon concentrations in the atmosphere to reach toxic levels. Thus whereas the interaction of the NOxand
HOxcycle on the one hand leads to removal of certain toxic species, it also plays an important role in
the formation of another toxic gas, namely tropospheric ozone.
The availability energy suppliers for the NOxand HOxcycles (i.e. sunlight for the first, and CO and
hydrocarbons for the second), can lead to different regimes of ozone and OH levels. Ozone production can be NOxlimited as well as CO or hydrocarbon limited. In city centers the availability of both NOx
CO or hydrocarbons, e.g. CH4 CO2+ H2O OH HO2(RO2) NO2 NO HOX NOX O3+ H2O + sunlight NO2 HNO3 sunlight (l<420 nm) O O3 O2 O2
Figure 1.5: The HOxand NOxcycle, acting together in the formation of tropospheric ozone. CO,
hy-drocarbons and sunlight are the energy providers for the HOx/NOx tropospheric ozone production
mechanism (adapted from Fig. 2.5 in [3]).
[12] that the relative concentrations of NOx, CO and hydrocarbons are more optimized for ozone
pro-duction not in the direct vicinity of emission sources, but rather downwind of such regions. Starting from either a NOxlimited regime, or a hydrocarbon limited regime, the difference in lifetime between
NOxand hydrocarbons may lead to a more optimal combination for ozone production several hours
after emission [13]. Three different NOxregimes can be distinguished with respect to OH levels: for low
NOxconcentrations (<0.1 ppb) no significant mechanism is present to reform OH after it has been
con-verted to HO2, it is lost by radical-radical interaction. OH concentrations therefore stay low, and with
it the oxidizing capacity of the atmosphere. For medium NOxconcentrations (1 ppb), HO2is efficiently
converted to OH, such that it can be used again for oxidation. At high NOxconcentrations (100 ppb)
most of the OH radicals react with NO2to form HNO3which is a permanent sink for both species: it is
removed from the atmosphere either by dry or by wet deposition.
1.3.2 The role of NOxin relation to climate
In order to understand the climate of the Earth, its natural fluctuations and those induced by anthro-pogenic activities, it is important to know the relative contribution of the factors affecting the radiative balance of the Earth-Atmosphere system. Whereas clouds, water vapor, carbon dioxide and methane provide the largest contribution to the radiative balance [14], a study by [15] has demonstrated that the contribution of NOxand some other gases is considerable as well. We focus on the contribution of NOx,
which has both a direct and an indirect radiative forcing effect.
The direct radiative forcing by the broad band absorption by NO2of UV/Visible light has been
in-vestigated by [16]. The global mean contribution by NO2on atmospheric heating is small, 0.05 W/m2
compared to a value of 1.69 W/m2for CO2. However, in heavy polluted places atmospheric heating
Through its effect on tropospheric chemistry mechanisms (Sect. 1.3.1), NOxaffects the atmospheric
composition and therefore indirectly the radiative forcing of the atmosphere in various ways, see [15]. They find that radiative forcing due to NOxemission (through its effect on other compounds such as
sulfate, nitrate, methane and ozone) is -0.29±0.09 W/m2; as opposed to +1.6±0.16 for CO
2. One such
mechanism involves OH: reduced NOxlevels lead to lower OH levels (Sect. 1.3.1), and therefore to higher
levels of methane, which is a strong greenhouse gas. NOxis thus a powerful cooling agent through its
indirect effects. Only SO2could have a stronger cooling effect through the highly uncertain aerosol
indirect effect on clouds. Reducing SO2and NOxemissions may be expected to lead to a higher total
radiative forcing, i.e. a warmer climate.
A second important effect of NOxis its role in the formation of aerosols, directly as well as indirectly.
Aerosols potentially have a large effect on the atmosphere’s radiation budget not only by direct scattering and absorption, but most of all by their effect on the lifetime and scattering properties of clouds, [see e.g. 17, 18]. The magnitude of those effects is however very uncertain [14]. NOxinduced changes in oxidant
levels affect the conversion from SO2to sulfate aerosol (indirect effect of NOx). A more direct effect
of NOxon the formation of aerosols is that it provides a source for HNO3, which in the presence of
ammonia leads to the semi-volatile ammonium nitrate:
NH3+ HNO3←→ NH4NO3, (1.9)
from which aerosols can be formed. Nitrate aerosol is formed through heterogeneous reactions of ni-trogen radicals such as N2O5, NO3, HNO3on aerosol surfaces [19]. Nitrate aerosols are mainly present
in the fine aerosol mode [20], have the strongest scattering efficiency in the UV/Vis, and therefore a sig-nificant climate forcing (cooling). The scattering properties of nitrate and ammonium aerosols also have an indirect effect: they perturb the photochemical oxidant production by altering photolysis frequencies [21].
1.3.3 Effects on human health
There is no doubt that NO and NO2are important components of air pollution [22], but it is nevertheless
an epidemiological challenge to determine which fraction of air pollution induced health effects, such as mortality or reduction of lung and cardiovascular function, can be ascribed to NOxalone. Urban
air pollution is a cocktail of NOx, O3, particulate matter (PM), SO2and volatile organic compounds,
all of which are harmful by themselves. Not only are they often emitted in a certain mixture, but in addition their interactions are complex as described above. It is therefore almost impossible to isolate these species in epidemiological studies. For example, NOxreductions may be advantageous in order
to have low ozone concentrations, but at the same time the decreased NOxmay lead to lower OH levels
and therefore more air pollutants in the form of volatile organic compounds.
Some studies find evidence that NO2enhances the effect of allergens, increases bronchial reactivity
and leads to increased hospital admissions for respiratory disease [23]. A meta-analysis of mortality has shown consistent associations with NO2[24]. High concentrations of NO2may induce long term
dam-age on lungs. Secondary products formed from NO2within the body, also in combination with other air
pollutants, can be carcinogenic. NOxmay lead to the formation of the nitrate radical (NO3), nitroarenes,
and nitrosamines which may cause biological mutations. Prolonged exposure to tropospheric ozone (for which NOxis a catalyst) is found to cause chronic damage to the lungs, and is associated with lung
Table 1.1: Global sources of NOxin the year 2000. Combined from [27] and [26].
Source Strength (TgN yr−1) Fossil fuel combustion 25.6
Lightning 8.9 Biomass Burning 5.8 Soils 8.9 Biofuels 2.2 Aircraft 0.5 Stratosphere 0.1
and vice versa. High concentrations of particulate matter, mostly (but not only) the fine mode, may lead to a marked reduction in life expectancy. The reduction in life expectancy is primarily due to increased cardio-pulmonary and lung cancer mortality [22].
1.3.4 Sources, sinks and trends
NOxemission estimates can be derived in a bottom-up manner, by considering the various processes
in which NOxis formed as well as their relative magnitudes and spatial distribution, and in a top-down
manner for which space borne observations are crucial. Tab. 1.1 gives an overview of NOxemission
es-timates from a study using a combination of both techniques, see e.g. [26] and [27]. The main global source of NOxis combustion by anthropogenic activities. NOxis formed in fossil fuel combustion
pro-cesses: power stations, industry and home heating. Automobiles also form an important source. NOxin
addition is emitted in biomass burning events, with a natural or anthropogenic origin.
Initial formation of NOx takes place after thermolysis of oxygen molecules at temperatures above
2000 K, present e.g. in combustion chambers or with lightning:
O2+ (heat) ←→ O + O (1.10)
O+ N2←→ NO + N (1.11)
N+ O2←→ NO + O (1.12)
These three equilibria shift towards the right with increasing temperatures. It is interesting to note that the air pollutant NOxis not created from oxidation of the fuel itself, but instead from the molecules of
clean air. An exception is formed by the combustion of coal, where also the nitrogen contained in the fuel leads to the formation of NOx.
Important natural NOxsources are lightning and soils. Soil NOxmainly originates from the tropics
(70%), [28]. Large pulses of NO emissions are created by rain falling on dry soils, e.g. savannas. The NOx
induced by lightning is due to thermolysis, similar as for combustion.
Figure 1.6 shows the global distribution of NO2 derived from OMI satellite observations. Densely
populated and industrialized regions clearly show up, but also biomass burning regions, for example in Africa. The NOx emission budgets per continent are summarized in Tab. 1.2. In the year 2000 the
Table 1.2: Surface NOxemissions for different parts of the world in the year 2000. From [26].
Region Emissions (TgN yr−1) United States 7.4
Europe 6.2
East Asia and Japan 7.0 SE Asia and India 4.3
Mideast 1.3
Africa 6.3
Cent. and S. America 4.3
Australia 1.2
Global 40.3
The Netherlands produce about 2.2% of European NOxemissions. Its surrounding countries:
Ger-many 10%, Belgium 1.8% and Great Britain 10%, [29]. Between 2003 and 2007 the European NOx
emis-sions decreased with 5.8%, about four times less than in the Netherlands (22% decrease). It was estimated that the NOxconcentration in the Netherlands is determined for 30% by sources across the border, [30].
The principal sink for NOxis oxidation by OH and O3, forming HNO3. HNO3itself deposits relatively
fast due to its high solubility in water. Reaction of NOxwith carbonyls, that can have an anthropogenic
as well as a biogenic origin, leads to the formation of PAN (peroxyacyl nitrate). PAN is another sink for NOx, but is less soluble in water and therefore has a longer lifetime. It can be thermally dissociated,
leading again to the formation of NOx. Under cold conditions PAN provides a long-range transport
mechanism for NOx.
Lifetime
For summertime daylight conditions, the typical lifetime of individual NO2molecules is approximately
2 minutes [31]. This fact is however not related to the lifetime of NOxabundances in general, since NO2
molecules are not only photo-dissociated but also recreated from NO and O3, see the previous section.
The average lifetime of NOxis determined by permanent removal mechanisms, such as wet and dry
deposition of HNO3after reaction with OH.
An overview of NO2lifetimes found in various studies is given in [32]. These studies are based on very
different approaches: power plant and urban plumes are measured from the surface, chemical transport models are used, as well as observations from space. For summertime conditions, the estimated lifetime of NO2varies between 3 and 6 hours. Winter lifetimes may increase up to 20 hours. Lifetime is also
affected by the presence of clouds [33]: the presence of clouds reduces the amount of sunlight reaching the boundary layer, leading to lower concentrations of OH, and therefore to longer lifetimes for NO2.
1.4 Measurement Techniques for NOxand NO2
There are many different types of measurement techniques for NOxand NO2. These techniques have
different domains of applicability. Some are especially useful for monitoring of street-level concentra-tions, others are better suited for monitoring of the vertical distribution of NO2. We discuss the various
methods in the section below since they are essential for intercomparisons with MAX-DOAS, in or-der to establish its quality. It should be noted however, that none of these measurement techniques can provide a golden standard due to differences in representativity, see Sect. 1.4.4.
1.4.1 In-situ techniques
Various in-situ techniques exist for monitoring of NO2or NOx. A distinction can be made between
optical techniques which leave the sampled air unaffected, such as long-path DOAS and optical cavity based techniques (see below) on the one hand, and on the other hand techniques that are based on a chemical analysis which influences the chemical composition of air: chemiluminescence techniques.
The chemiluminescence NOxmonitoring techniques have the longest tradition and are widely used
in national-scale air quality networks, see e.g. [34]. The essence of the technique is the detection of light produced by luminescence of activated NO2, itself being formed from NO after reaction with artificially
supplied ozone. Chemiluminescence monitors can be equipped with molybdenum converters or with photolytic converters, [see e.g. 35, 36]. The latter are more specific to NO2, see e.g. [37].
In contrast to the chemiluminescence techniques, active optical methods can be used to analyze trace gas abundances in the air without altering the concentrations. The long-path DOAS method [38], which is not really an in-situ method in a strict sense, but included in this section because it is precise and has a relatively local character, uses a light source at one location aimed towards a distant detector (typically at several hundreds of meters). Current implementations of this method often use broad-band light sources, see e.g. [39]. Like other DOAS applications, this technique has the advantage of being specific, i.e. the NO2measurement is completely insensitive to NO abundances.
A different type of active optical technique are cavity-enhanced techniques, which use highly re-flective mirrors to create light paths that are hundreds of meters in length within a relatively small instrument. Examples of cavity-based techniques applied to NO2observations are: cavity ring-down
spectroscopy using a pulsed laser, [see e.g. 40, 41], and LED based cavity-enhanced DOAS [42]. An advantage of the chemiluminescence and cavity enhanced techniques is that they can be used in airplanes to measure vertical profiles of NO2, [see e.g. 43, 44]. Such in-situ measurements of vertical NO2
profiles are highly desired since the vertical distribution of NO2has a large impact on ground and space
based remote sensing observations of tropospheric columns, even though this vertical distribution is poorly constrained in those observations. A-priori profile assumptions have to be made when analyzing the remote sensing observations, and these should preferably be based on a climatology of observed NO2
profiles. A disadvantage of the use of airplanes is the high cost per vertical NO2profile. A much more
low cost solution is provided by the recent development of a light-weight chemiluminescence based measurement technique that can be used in combination with radiosondes [45].
1.4.2 Remote sensing from the surface
Various techniques exist for surface based remote sensing of tropospheric and stratospheric NO2. A
distinction can be made between active and passive methods.
The RIVM NO2lidar [46] is an example of an active system. It sends out laser pulses at two carefully
selected wavelengths which together are sensitive to NO2and not to other absorbers. NO2
strengths of this method are (i) it has a unique sensitivity to NO2, thus does not suffer from
interfer-ences by other species, (ii) the light path is well known, and therefore one can clearly select different vertical regions. A disadvantage of the technique is that the sensitivity decreases with distance. Not only clouds, but also aerosol layers can greatly affect the accuracy in the retrieval through shielding. As a consequence, the system is primarily sensitive to the boundary layer, where it is well validated [47], and much less to the free troposphere (see also Sect. 1.4.4 and Chapter 4).
The passive techniques use sunlight (or sometimes moonlight) as a light source. First DOAS type measurements of scattered sunlight were zenith sky observations of stratospheric gases (Sect. 1.6.2). The variation of the NO2absorption with varying solar zenith angle, which is largest around sunrise and
sunset, can be used to infer stratospheric NO2columns. Based on zenith-sky observations, also
tro-pospheric NO2columns have been retrieved [48]. The DOAS method (Chapter 2) requires a reference
spectrum in addition to the spectrum which is to be analyzed. In the case of zenith sky measurements, the reference is usually the zenith noon spectrum, which has the shortest light path through the strato-sphere. Since this spectrum is taken at another time of the day, temporal variations in tropospheric and stratospheric NO2columns (that are strongly correlated with solar zenith angle) should carefully be
considered.
Direct sun observations are sensitive to the total NO2column, thus to the sum of stratospheric and
tropospheric NO2. Based on these observations alone, one cannot distinguish between the two. A major
advantage of this technique is that (in the absence of clouds) the light path through the atmosphere is well-known. The selection of a reference spectrum is less trivial for direct sun than for other passive DOAS applications [49].
The MAX-DOAS technique, [see e.g. 50, 51], which is explained in detail in Chapter 2, uses scattered sunlight observations obtained for multiple viewing elevations (Fig. 3.1). Since it uses the zenith spec-trum as a reference, it is primarily sensitive to NO2in the troposphere, see Fig. 4.1 and 4.10. MAX-DOAS
instruments can be used to derive tropospheric columns (Chapters 3 and 5), or simplified vertical pro-files (Chapter 4). Especially for profile retrieval the use of low viewing elevations is important. For these elevations it is important to consider the effect of aerosols on the NO2measurements, see Chapters 2-4.
MAX-DOAS instruments can also be used in the zenith-sky mode, in order to retrieve stratospheric NO2amounts (see above). Only some newly developed MAX-DOAS instruments have an option to
perform direct sun observations. For an overview of such instruments, see [52].
1.4.3 Remote sensing from space
NO2remote sensing from space was first done with the SAGE instrument [REF], but the solar
occulta-tion technique used was not suitable for the retrieval of tropospheric NO2. Tropospheric NO2
observa-tions from space started with the GOME-1 instrument (1995-2011), which for the first time demonstrated the potential of DOAS type retrievals from space [53]. After GOME-1, the SCIAMACHY instrument was launched in 2002 [54], OMI in 2004 [55], and GOME-2 in 2006 [56], all of them in low-earth sun-synchronous polar orbits, such that they cross the equator approximately 13-15 times per day (depending on the height of the platform), each time at the same local time. Instruments differ in many aspects: op-tical design, scanning procedure, wavelength range, and pixel size (which vary from 320x40 km for GOME-1, to 24x13 for OMI nadir pixels).
Such space borne observations allowed, for the first time, daily global mapping of tropospheric trace gases measured with a single instrument, which is essential for an independent comparison of NO2
OMI tropospheric NO2column density [1015molecules/cm2] OMI t roposp heri c NO 2 co lum n de ns ity [ 10 15 m ole cu le s/ cm 2]
Figure 1.6: Left panel: A global map of tropospheric NO2retrieved by OMI (Domino-Product v.2, [57])
showing the different source areas. Anthropogenic emissions in densely populated regions (mainly in industrialized countries) clearly show up. Ship emissions of NO2are so large that major routes can be
seen, for example in the Mediterranean. In central Africa and the Amazon NO2is emitted in biomass
burning events. The OMI data is averaged over 48 months, starting in 2005. Right panel: A map of tropo-spheric NO2, retrieved by OMI, for the Netherlands and surroundings. For each of the years 2005-2009
only the four months May-August were used. In this warm season the lifetime of NO2is relatively low,
typically a few hours, and therefore the map shows most NO2close to its source. Courtesy of Boersma
and Vinken.
abundances at different locations worldwide, see also Fig. 1.6. The central idea of observations from space is illustrated in Fig. 2.10: sunlight is reflected by the atmosphere and surface of the earth, and detected by the satellite. Unlike surface observations, satellites can also perform direct observations of the sun, which can be used as a reference spectrum in the DOAS analysis (Chapter 2).
There are several challenges in the retrieval of tropospheric NO2from space, namely to separate the
total NO2column in a stratospheric part and a tropospheric part, to obtain a reliable estimate the NO2
vertical profile shape (this usually comes from a global chemistry and transport model), and to account for clouds, aerosols, the surface albedo and terrain height (orography) [58].
1.4.4 Spatial representativity and validation
In order to assess the quality of the different NO2measurement techniques described above, validation
is crucial. For the in-situ techniques this can for a large part be done in the laboratory, although compar-isons with other techniques under atmospheric conditions will always be needed, for example to exclude interferences by unknown species.
For atmospheric remote sensing techniques, validation is more challenging since they can mostly not be tested in a laboratory under representative conditions. In addition, most available techniques have a different spatial representativity and a different temporal resolution, as shown in Tab. 1.3. The table shows that there is not a single instrument which has a high accuracy (to NO2) at all altitude levels in
the troposphere. Some techniques may be very accurate for some specific conditions, but often they cannot determine by themselves if those conditions occurred at the time of measurement.
true profile (real atmosphere) assumed profile ( ‘retrieval model’ ) retrieved profile or vertical column
State Space Measurement Space
forward modeling
interaction of radiation with real atmosphere
and instrument
simulation of interaction of radiation with modeled atmosphere and instrument
real measurements
>> differential slant columns from DOAS spectral fit >> spectra (uncommon)
simulated measurements
>> differential slant columns from single wavelength simulations >> spectra (uncommon)
inversion
iterative minimization of differences between real and simulated measurements by changing the parameters of
the assumed profile
trace gas amount
trace gas amount trace gas amount
he ig ht he ig ht he ig ht measurement
Figure 1.7: Schematic of a retrieval scheme. In the context of MAX-DOAS observations, it is common to minimize differences between measured and simulated differential slant NO2columns. The measured
differential slant column is derived from a spectral (DOAS) fit that is applied to measured spectra, the simulated slant column is determined with single wavelength radiative transfer simulations, see Chap-ter 2.
Because all the monitoring techniques have their specific strengths and shortcomings, it is essential to compare them in dedicated campaigns, see [see e.g. 52, 59, 60], and to take many additional data sources into account in order to understand the origin of the observed differences.
1.5 The retrieval process
In this section we will describe in general terms how remote sensing measurements can be interpreted. The MAX-DOAS measurements used in this research are indirect measurements: the basic retrieval quantities, ’differential slant NO2columns’, cannot be converted to what we want to know, i.e. a vertical
NO2column or profile, in a simple, unambiguous manner. An inversion method is needed to convert the
measurement to the physical quantity of interest. This procedure of measurement and inversion is what will be called the ’retrieval process’, or simply the retrieval. It is one of the main subjects studied in this thesis to find out which retrieval method can be used to interpret the MAX-DOAS measurements. In this section we will give a general description of the various steps of a retrieval process: taking measurements, simulate measurements using a radiative transfer model and applying an inversion method.
The general procedure of a retrieval is illustrated in Fig. 1.7. A distinction is made between the ’state space’ (see also Sect. 1.2) and the ’measurement space’. In the following we define the state space as a space with parameters that describe the vertical NO2profile, and other important factors which
in-fluence the measurements: the aerosol profile, temperature profile, etc. Since measurements, especially MAX-DOAS measurements, are frequently not sensitive to many aspects of the state of the real
atmo-sphere, a simplified atmospheric model can be defined which contains just enough detail to explain the observations. The complexity of this model depends on the quantity one wants to retrieve, and the mea-surements that one wants to use: a retrieval model used for tropospheric NO2column retrieval (Chapter
3 and 5) can generally be more simple than a model for profile retrieval (Chapter 4 and Fig. 1 in Chapter 6).
Measurements are performed in ’measurement space’. The transformation from state space to mea-surement space is essentially described by the physical interaction between the atmosphere and the radiation passing through it, and by the detection of radiation by the instrument. This process takes place in reality but it can also be simulated, in which case it is called ’forward modeling’ (Sect. 1.5.1). This forward modeling is performed with a radiative transfer model, which simulates radiation that propa-gates through the simplified atmospheric model (the retrieval model). Depending on the application, also the interaction with the instrument is simulated.
Ultimately, we want to know a quantity in state space. This is done in a procedure called ’inversion’ (Sect. 1.5.2), where real and simulated measurements are compared to find the state parameters of the retrieval model which best describe the observations. If we found these best parameters, we call it ’the solution’, or ’the retrieved quantities’.
1.5.1 Forward simulations
Unlike in the laboratory, the conditions in the atmosphere cannot be controlled, and most of the time we do not know its state. As a consequence, one cannot find out empirically which state parameters affect our measurements. This should be known to allow a transition from measurement to state space. The solution is to simulate measurements using a radiative transfer model (RTM), and in this sense RTMs are equally important for the retrieval process as instruments (the hardware). Atmospheric RTMs calcu-late intensity of electromagnetic radiation (e.g. sunlight) that has propagated through the atmosphere. Depending on the model and its settings, this output can be specified per wavelength, viewing direc-tion, solar posidirec-tion, instrument position (surface based or space based). In the model the atmosphere itself can be defined, usually only in the vertical dimension, by specifying pressure, temperature, aerosol properties, trace gas amounts, clouds, etc. for each vertical layer of a certain vertical grid.
RTMs are generally used for two reasons: (i) they can be used to perform sensitivity studies: through forward modeling we learn how the measurements are affected by the state of the atmosphere because we can ’control’ the state of the model atmosphere; (ii) if we know which factors are most important to interpret our observations, we can define a simplified retrieval model which is used to simulate mea-surements, and which parameters we want to retrieve by combining it with real observations.
There are several types of RTMs: e.g. discrete ordinate type models (e.g. [61]) which numerically solve the radiative transfer equation (see below), ’backward’ Monte-Carlo type models (e.g. [62]), which de-scribe radiation using statistical techniques applied to a large ensemble of paths simulated for individual photons, and models based on the doubling adding principle (e.g. [63], and [64]). The studies reported in this thesis were all performed with a single model of the last type: ’doubling adding KNMI’ (DAK).
The central equation in most radiative transfer models is the ’radiative transfer equation’, which de-scribes scattering and absorption of radiation. When applied to UV and visible wavelengths (thus in the absence of thermal emission), it is written as:
dI(λ) ds = − (єa(λ) + єs(λ)) ⋅ I(λ) + єs(λ) ⋅ ∫ π 0 ∫ 2π 0 I(λ, θ, ϕ) ⋅ S(θ, ϕ) 4π dϕ⋅ sinθdθ,
Figure 1.8: Schematic of laboratory experiment to measure absorption by a trace gas in a cell. The cell is filled with air, and a trace gas, which is homogeneously mixed throughout the cell. [A] shows a direct ob-servation, where the far majority of detected photons has the same path towards the detector (a straight line). In this situation the Lambert-Beer law can be applied for a wide range of concentrations. [B] shows an observation of scattered light, where the scattering occurs mainly due to the air molecules. The de-tected photons can have many different light paths. The Lambert-Beer law can only be used to measure deviations with respect to a reference concentration of the trace gas, and applies only (approximately) to a small range of concentrations around the reference.
where λ is the wavelength, dI(λ)/ds represents the change in the spectrum of incoming radiation that moves through an infinitesimally thin layer with thickness ds. єaand єsrespectively denote the
absorp-tion and scattering coefficient and S(θ, ϕ) denotes the dimensionless scattering funcabsorp-tion, see e.g. [3].
Lambert-Beer law
The Lambert-Beer law (or Bouguer-Lambert-Beer law, see [65]) is the central law in studies of absorp-tion by trace gases in the atmosphere. In the MAX-DOAS context it is used to describe the absorpabsorp-tion of scattered sunlight. The Lambert-Beer (LB) law is used in combination with measurements, as well as in combination with radiative transfer simulations, both are described in detail in Chapter 2. Before considering its atmospheric application, we here introduce the LB law first in the simple context of a laboratory set-up with a light source aimed at a detector, shining through a transparent cell which is filled mainly with air, but in addition with an absorbing gas (see Fig. 1.8). The intensity of the light in the absence of the absorber is denoted I0(λ), and called the reference measurement. In the presence of
the absorber, the intensity is reduced according to the LB law, which basically follows from integrating the absorption part of Eq. 1.5.1 along the light path:
I(λ) = I0(λ) e−σ(λ)⋅n⋅L. (1.13)
Here σ(λ) denotes the absorption cross section of the gas in [cm2/molecule], n denotes the gas concen-tration in [molecules/cm3
] and L denotes the path length of the light through the cell, e.g. in [cm]. The product σ(λ) ⋅ n ⋅ L is called the optical thickness (dimensionless), and σ(λ) ⋅ n the extinction [m−1
]. The product n⋅ L is called the column, and in the case of scattered light it is called the slant column. If a trace gas concentration and path length are accurately known, then its absorption cross section can be determined, as done in a laboratory set-up. Absorption cross sections obtained in this way are used throughout this work. Conversely, if the absorption cross section is known, then the same law can be used to determine the concentration of the absorber within the cell. This application of the LB law de-pends on the possibility to remove the gas from the cell in order to take a reference measurement of the
Figure 1.9: Cartoon illustrating the inverse problem. Can we find the dragon from its tracks? In this research the dragon is the vertical NO2profile. The tracks are the MAX-DOAS observations. One could
say that the tracks are blurred by those of another animal: the aerosol profile. Sometimes we do not desire to find exactly which dragon created the tracks, but only want to know its ’weight’: the tropospheric NO2
column. See also Fig. 1.7. Adapted from [66].
light source without the absorber in between. In atmospheric research, this can only be simulated in a model.
In the above example, the LB law is written for a situation where all detected photons have the same path: straight from the light source to the detector. In addition, the light path is the same with and without the absorbing gas. Furthermore, the law is written for a single wavelength. Strictly speaking, the LB law only applies to such conditions. In most practical situations, there are various reasons why the LB law does only apply by approximation, since these conditions are violated. In general this is the case when the optical thickness varies over the various light paths from the source to the detector. In that case, the LB law applies for each light path individually, but not for the sum of light paths. In [65] (p. 55) this is summarized in the phrase: ’The sum of exponentials is not an exponential.’ Since MAX-DOAS measurements are always based on scattered light, one should be careful with the use of the LB law (see also Chapter 2). For weak absorptions (τ<< 1) the LB law can be applied to scattered light observations with no large errors.
1.5.2 Inversion and the inverse problem
One of the main challenges in the field of remote sensing is to tackle the inverse problem, see Fig. 1.9. This problem can be stated as follows: how to give the best estimate of the state of the atmosphere that corresponds to the available measurements, taking their uncertainty into account? When applied to the research described in this thesis, the problem could be stated as: which retrieval algorithm can retrieve with adequate accuracy a tropospheric NO2column, or a (simplified) vertical NO2profile, given a set
of MAX-DOAS differential slant NO2column measurements at various viewing elevation angles.
The profile problem could be solved relatively easy if there were many independent measurements, and if there was a 1-1 correspondence between the set of MAX-DOAS NO2measurements and the NO2
profile, for example if the measurements for each viewing elevation would be sensitive to one specific altitude range. This is however not the case: the MAX-DOAS vertical sensitivity functions are very broad and almost flat at high altitudes. The measurements therefore contain very little altitude information (see
e.g. Fig. 4.1 in Chapter 4). In addition, the measurements are not completely independent of one another, or more frequently very strongly related, given the various sources of uncertainty.
In Fig. 1.10 it is illustrated for a two-dimensional retrieval problem how two measurements can be not independent, given their uncertainty ranges. A two-dimensional retrieval problem can in principle be solved by two suitable observations, provided that they are independent. The two observations used in Fig. 1.10 are independent in the sense that they are based on different physical principles: the relative intensity measurements is primarily influenced by scattering properties of the atmosphere, whereas the O4measurement is primarily an absorption measurement. Both are sensitive to the presence of aerosols,
but in contrast to the relative intensity measurement the O4measurement is also quite sensitive to the
vertical distribution of aerosols, at least, for higher aerosol optical thicknesses. For low AOTs the ob-servations have almost the same information content (just as other viewing elevations not shown in the figure), and contain only one piece of information, especially if their uncertainties are considered.
The complexity of the NO2inversion problem considered in this work is increased by the fact that
MAX-DOAS NO2measurements are sensitive as well to aerosols, clouds and other factors affecting
the light paths of detected photons. This implies that, even under cloud free conditions, not only the NO2-state of the atmosphere needs to be retrieved, but also the aerosol state.
Even if in theory there is only one set of observations corresponding to a certain state of the atmo-sphere, then, in the presence of noise, there may be many possible states leading to the same set of observations, as illustrated in the above example. It is therefore of key importance to develop an es-timate of the uniqueness of the solution, an issue which is closely related to the retrieval uncertainty estimate. A distinction can be made between two types of retrieval uncertainty. Firstly the uncertainty within the retrieval model, which is based on a simplified description of the atmosphere. This uncer-tainty estimate describes the set of possible model states that might have lead to the measurements. The second uncertainty estimate also considers the effect of atmospheric states not included in the retrieval model. In the next Chapter two retrieval methods are introduced: Optimal estimation based retrieval (OE) and retrieval based on a profile parametrization with a low number of free parameters (PP) (see Chapter 2 for references). In the OE approach, the two types of retrieval error are generated as standard output of the formalism. In the PP approach the uncertainty estimate within the retrieval model can be derived for example by using an ensemble approach. The effect of atmospheric states not included in the retrieval model should be investigated in dedicated sensitivity studies, for both see Chapter 4.
Whether the inverse problem can be solved, and with which accuracy, will depend on the desired vertical resolution, be different for different vertical regions, will depend on the noise level (and thus on the amount of NO2in each vertical layer, the availability of sunlight and on the instrument sensitivity),
and the state of the atmosphere, for example the abundance of aerosols. All these factors need to be considered in the design of retrieval algorithms. In the chapters 3-5 each time a different retrieval algo-rithm is used: (i) one to derive tropospheric vertical NO2columns (for satellite validation) under clear
sky conditions, using various viewing elevation angles and correcting for aerosols (Chapter 3), (ii) one to retrieve information on the vertical distribution of NO2to explore the potential of MAX-DOAS in
this respect (Chapter 4), and (iii) another algorithm for tropospheric NO2column retrieval, which can
be applied also under cloudy conditions, based on a relatively high viewing elevation which is almost insensitive to aerosols.
Figure 1.10: An example of a two-dimensional retrieval problem ’solved’ for two different cases (26 De-cember 2008: the two solid lines, and 21 March 2009: the two dashed lines), by combining a MAX-DOAS O4absorption measurements (darkest) with an observation of relative intensity (lightest), both at α= 4○
and for λ= 360 nm. Each measurement consists of one value, that is constant along a certain contour (each line in the plot represents a contour) in this two-dimensional space. The shape of the contour is calculated with a radiative transfer model. In both cases the retrieval problem is solved in a strict sense, i.e. the solution is given by the point where the lines cross. It is determined by a value for the AOT and a value for the boundary layer height. However, the retrieval of the boundary layer height is very inac-curate, when the uncertainty estimates are considered, especially on 26 December. On this last date, the two measurements almost contain the same information: both of them quite accurately determine the AOT, whereas they are both quite independent of the boundary layer height.
1.6 MAX-DOAS and this thesis
This section describes on a high level what can (potentially) be achieved using the MAX-DOAS method, and what it contributes to other measurement techniques (a detailed description of instrument and method is given in the next Chapter). Then we give a short overview of the relative short history of MAX-DOAS research. We end this section with a list of the research objectives of this thesis.
1.6.1 MAX-DOAS - a unique instrument
Retrievals based on measurements performed with the MAX-DOAS instrument can be used to derive: • Trace gases absorbing in the UV-Vis: qualitatively by identifying the differential absorption struc-tures of trace gases in the DOAS fit; tropospheric columns; simplified vertical profiles; strato-spheric columns when using the zenith sky mode.
