1.3 Air Pollution
2.1.1 Paper Analysis - The DOAS Tomography State of the Art
2 . 1 . D OA S T O M O G R A P H Y
Table 2.3: Selection filters in use for this study’s search.
Criterium Definition
Exc. Criteria EC1 Satellite data papers are not accepted Inc. Criteria
IC1 Results must be journal papers
IC2 Results must be about Tomographic DOAS IC3 Results must be written in English
Start
Google Scholar
Web of Science Scopus
In GS? IC and EC Pass?
N
Discard Add to selected
N Discard
End For each paper in library
Figure 2.1: Conduction stage flowchart. Libraries are searched independently through Publish or Perish [44], but results must be checked to ensure they are not counted twice, due to the imbalance of power between GS’s search engine and the others.
Table 2.5 presents all selected papers, as well as their score according to the formula of Equation 2.1. The following subsection includes my observations on the hypothesis explored through the selected papers and how I can use this in my research.
Table 2.4: Search results summary.
Library Results Excluded In GS Final Weight in study
GS 37 25 0 12 92,31%
Scopus 15 3 12 0 0%
WoS 9 1 7 1 7,69%
Total 61 29 19 13 100,00%
CH A P T E R 2 . L I T E R AT U R E R E V I E W
Table 2.5: References and relevant data for the selected papers, including their score, which was calculated according to Equation 2.1.
Number Reference Quartile Citations Year Score
1 [42] 1 35 2006 2,69
2 [43] 1 2 2005 0,14
3 [59] 1 30 2004 2,00
4 [72] 2 1 2005 0,05
5 [63] 3 6 2003 0,19
6 [36] 1 8 2006 0,62
7 [52] 2 26 2009 1,95
8 [67] 4 0 2003 0,00
9 [75] 1 12 2006 0,92
10 [74] 1 42 2005 3,00
11 [66] 4 1 2003 0,02
12 [62] 1 4 2006 0,31
13 [19] 1 2 2017 1,00
kind. As will be described in Section 3.2, active DOAS systems do have better analyt-ical capabilities than their passive counterparts, although that comes at the cost of increased instrument complexity and operational costs.
Another immediate conclusion is that there is a "dominant"study. Almost half of the papers found originated from the BABII campaign, in which a group of researchers set out to quantify pollution through DOAS tomography along a busy German motorway, in the beginning of the 21stcentury [42, 43, 59, 61, 72, 73, 75].
All of the active DOAS systems were purposely built for their corresponding exper-iment (or group of experexper-iments). BABII researchers used two telescopes with around 200mm diameter and 1m focal length to simultaneously illuminate 8 retroreflectors that were assembled onto two towers located on each side of the road. In one of the papers associated with this initiative, the same telescope instrumentation was used to validate the 2D reconstruction technique that was going to be used in the other papers. The campaign’s main assembly is illustrated in Figure 2.2.
Another important initiative with respect to DOAS tomography was the study conducted in 2016 by Stutz et al [84]. The approach in this case was to use a similar telescope to detect the light emitted by a narrow interval UV LED light source (290nm) to create a fence line monitoring system for Benzenes, Toluene and Xylene. The team managed to apply this system in a successful manner in refineries in Los Angeles and Houston. One of the most interesting aspects of this study is that it details a tomographic system that could easily be commercially deployed.
Another type of DOAS tomography system was proposed by researchers in the Cork Institute of Technology [63, 66, 67]. In their three papers, the authors describe 1)a new multipath instrument that significantly increases the amount of projection information in this kind of application; 2) a tomographic reconstruction algorithm
2 . 1 . D OA S T O M O G R A P H Y
Figure 2.2: BABII assembly geometry. In this experiment campaign, the telescopes illuminated retroreflecting targets that were positioned in two steel towers on both sides of a busy motorway in Germany, connecting Heidelberg to Mannheim [73].
based on evolutionary algorithms; and 3)the application of DOAS tomography to a simulated urban canyon scenario. Although all three papers present technological innovation, it would not be fair not to say that from a strictly literary point of view, these were among the weakest retrieved by the search process.
Regarding passive DOAS applications, the two papers we have found come with two completely different paradigms. The first article [52] was written in 2009 and details the application of a tomographic inversion algorithm to a scanning DOAS application, designed to work with trace gas plumes like the ones above volcanoes or power stations.
The team present a system composed of two DOAS devices, with sufficient distance with themselves as to allow tomographic reconstruction, but sufficiently small to allow the light path to be considered a straight line from the point of last scattering to the detector. The authors applied an adapted version of the Lower Third Derivative (LTD) algorithm to the projections obtained by pointing the set of fixed DOAS apparatus towards the plume in different angles. Besides simulations for their proposed method, the authors have also conducted practical experiments, both over a power plant in Spain and a volcano in Italy. Results from these experiments display a good agreement between reality and simulation results, proving the technique’s validity.
The second Passive DOAS application is a paper published by Frins et al. [36]. In this study, the researchers detail a particular application in which they measure light coming from bright and nonreflecting sun-illuminated objects in their field of view.
They use this light to retrieve column density values for a number of trace gases. The proposed method also includes a way with which to remove the stratospheric contri-bution that appears in the measured light besides the target column. The authors discuss how radiative transfer can influence measurements, but they also present a number of approaches to mitigate this problem, ensuring the validity of their ap-proach. Besides presenting the method, the authors also describe an experiment they conducted by assembling and manoeuvring a DOAS system on top of a building in Heidelberg, Germany.
CH A P T E R 2 . L I T E R AT U R E R E V I E W tion of the received light is reflected by the target, or,
alternatively, black targets can be used; then the received light is scattered solely within the air vol-ume between the instrvol-ument and target. In either case, the novel technique combines the advantages of both passive and active systems; no artificial light sources involving relatively complicated optical set-ups are needed. In addition, the interpretation of the observations can be referred to well-defined light paths and even allows tomographic inversions.26
The main limitations of the method are the restric-tions to wavelengths ⬎300 nm and to daytime. In particular, most aromatic compounds and nighttime chemistry cannot be investigated. However, there is a large class of species including NO2, HCHO, SO2, H2O, Glyoxal, BrO, and others that can be detected.
The paper is organized as follows: In Section 2 we introduce the proposed measurement technique, de-scribe the composition and information content of the measured light, and the specific requirements and subsequent steps of the spectral analysis. In Section 3 the various measurements strategies and possible results (including tomographic techniques) are de-scribed. In Section 4 the first measurement examples are presented. Finally, in Section 5 we summarize our findings and give recommendations for basic and advanced measurement strategies and for the specific design of advantageous experimental setups.
2. Multiaxis Differential Optical Absorption Spectroscopy Observations of Illuminated Targets In many aspects tomographic MAX-DOAS observa-tions can be treated very similarly to tomographic DOAS measurements using artificial light sources and well-defined light paths.1–3,27 In particular, we will show in this paper that tomographic MAX-DOAS eventually yields an intermediate product very similar to active long-path DOAS: the path-integrated (or av-erage) trace gas concentration between the instrument and the target. However, there are also two fundamen-tal differences: First, MAX-DOAS observations use the Sun, which is an extraterrestrial light source; second, atmospheric scattering processes between the instru-ment and the target can also contribute to the mea-sured signal. In the following subsections the resulting effects and their correction are described in detail.
A. Tomographic Multiaxis Differential Optical Absorption Spectroscopy Measurements
MAX-DOAS observations directed to targets illumi-nated by sunlight receive photons that have traversed two basic sections in the atmosphere: the distance be-tween the top of the atmosphere and the target, and the distance between the target and the instrument (see Fig. 1).
Depending on the investigated problem, the prop-erties of the measuring site, and the number of instruments and targets, a large variety of mea-surement geometries can be developed. In the sim-plest case, one instrument is directed to one target;
then the average concentration between the instru-ment and the target can be derived. Targets at
dif-ferent distances or altitudes allow the retrieval of horizontal or vertical gradients (see Fig. 1). Higher-dimensional tomographic setups can even yield the spatial trace gas distribution.
Irrespective of the final inversion strategy, the ba-sic quantity that is derived from the spectral DOAS analysis is the (differential) optical depthof a se-lected trace gas, from which the so-called slant col-umn density (SCD) S, the trace gas concentration integrated along the light pathL, can be derived by dividing it by the respective absorption cross section at the same wavelength:
S⫽
冕
0L共l兲dl⫽共共兲兲. (1) It should be noted that for the DOAS analysis the so-called differential optical depth is usually ana-lyzed, which, in the simplest case, is the difference of the optical depth at different wavelengths.14,28In analogy, a differential cross section can be defined, which then has to be applied in Eq. (1). It might, for example, indicate the difference of the absorption cross section (or optical depth) inside and outside an absorption band. For these differential quantities we will hereafter use the terms=and=.The analyzed total SCD can be expressed as the sum of the partial SCDs of both sections introduced above (see Fig. 1):
Stot⫽STarget⫹SAtm. (2)
Fig. 1. (Color online) Experimental setup for tomographic MAX-DOAS observations. In contrast to conventional MAX-MAX-DOAS obser-vations the telescope is directed toward targets that are illuminated by the Sun. The total signal contains the trace gas absorptions of two sections of the total light path: between the targets and the top of the atmosphere (paths A1, A2, A3) and between the instrument and the targets (paths B, C, D). After correction for the absorption from the section between the targets and the top of the atmosphere (see Subsection 2.B) the average trace gas concentration between the instrument and the target can be analyzed. If several targets are used, even multidimensional trace gas distributions can be re-trieved.
6228 APPLIED OPTICS兾Vol. 45, No. 24兾20 August 2006
(a) Schematic representation of Frins’s as-sembly [36].
Here STarget is the trace gas concentration inte-grated between the instrument and the target, and SAtm is the trace gas concentration integrated be-tween the target and top of the atmosphere. For many trace gases such as NO2, BrO, or O3, a substan-tial or major fraction of the total atmospheric column is located in the stratosphere; for such trace gases, SAtmwill be independent from the MAX-DOAS view-ing direction because the absorption paths through the stratosphere are determined by the solar zenith angle. Even for trace gases located in the free or upper troposphere this assumption is valid with only small restrictions.
It should be noted that the sensitivity of the mea-surement for the trace gases between the instrument and the target strongly depends on the brightness of the target and the atmospheric conditions. Only for white targets is the measured SCD almost similar to the true SCDSTarget(see Subsection 2.C.1).
Since we are interested inSTarget, we have to remove SAtmfrom the total signal. This can be easily achieved and will be explained in Subsection 2.B. Another fun-damental question is to determine the sensitivity of the measurement for the trace gas between the in-strument and the target, in particular the depen-dence of the sensitivity as a function of the distance from the instrument. For an ideal measurement the sensitivity would be constant along the light path between the instrument and the target as, for exam-ple, for traditional long-path DOAS observations. We will show below that for many cases the sensitivity can be expressed as a simple functional expression of the distance from the instrument. For bright targets, for example, the sensitivity becomes similar to that for long-path DOAS observations. Nevertheless, in general, this sensitivity decreases with the distance from the detector because additional sunlight is scat-tered into the line of sight of the instrument (see Fig. 1). For the correct interpretation of the measure-ment, this sensitivity has to be characterized (see Subsection 2.C) and corrected. If no simple correction is possible, radiative transfer modeling has to be ap-plied (see Subsection 2.D).
B. Removal of the Signal From the Region Between the Target and Top of the Atmosphere (SAtm)
The correction of the partial SCD between the target and the top of the atmosphere共SAtm兲can be easily (and quasi automatically) achieved by two procedures. The main aspect of both methods is that for the DOAS analysis of scattered sunlight, a so-called Fraunhofer spectrum has to be included for the correction of the strong solar Fraunhofer lines.8,9,11,23 For this pur-pose, usually a measurement taken with the same instrument but under different measurement condi-tions (e.g., at a low solar zenith angle) is used. Since this Fraunhofer spectrum contains not only the solar Fraunhofer lines but also atmospheric absorptions, the result of the spectral DOAS analysis is the dif-ference between the SCDs of the measurement and the Fraunhofer spectrum. The trick for the correction of signal from between the target and the top of the
atmosphere共SAtm兲is to select a Fraunhofer spectrum that contains onlySAtm. This can easily be achieved by two simple approaches. The first approach is to point the telescope of the instrument toward the zenith.
It then receives photons that have traversed almost similar paths as those between the target and the top of the atmosphere. Especially for trace gases that (in addition to the boundary layer) are located only in the stratosphere, a zenith-scattered spectrum is already sufficient for the correction ofSAtm. If trace gases are also located in the free troposphere it might be better, however, to choose a refined approach. For the mea-surement of the Fraunhofer spectrum the telescope should then be directed to a target close to the instru-ment, which has similar properties compared to the real targets. Particularly the orientation (the angle of the reflecting surface with respect to the instrument and the Sun) and the reflectivity (the albedo) of the nearby target should be similar to the target at far distance. This procedure ensures that the photon paths in the lower free troposphere (and thus the SAtm) of both the actual measurement and the Fraun-hofer spectrum are almost identical.
In some cases it might be sufficient to use a mea-surement from one (far) target (e.g., meamea-surement D in Fig. 1) as the Fraunhofer spectrum for the analysis of a measurement from another (far) target that is placed in the same direction but at a different distance (e.g., measurement C in Fig. 1). Then the resultingSTargetis the integrated trace gas concentration along the path between both targets.
C. Sensitivity of the Measurement for Trace Gases Between the Instrument and the Target
For the correct interpretation of the trace gas absorp-tion determined by the DOAS analysis the sensitivity of the MAX-DOAS observation for the trace gas be-tween the instrument and the target (e.g., as a func-tion of the distance from the instrument) has to be known. We show in Subsections 2.C.1–2.C.3 that for many cases simple functional expressions for this de-pendence can be found. The light received by the telescopeI共,L兲contains two parts (see Fig. 2):
I共, L兲⫽IAtmos共,L兲⫹ITarget共, L兲, (3) whereITarget共,L兲is the measured radiance caused by the reflected sunlight by the target, andIAtmos共,L兲is Fig. 2. Composition of the light received by the telescope: sunlight can be reflected by the target or scattered by molecules and aero-sols between the target and the instrument.
20 August 2006兾Vol. 45, No. 24兾APPLIED OPTICS 6229
(b) The physical principle behind Frins’s pa-per [36].
Figure 2.3: Erna Frins’s paper [36] proposes a very relevant passive DOAS application that can conceptually be employed in a DOAS tomography scenario. In this 2006 paper, the authors use multi-axis measurements of sun-illuminated targets to estimate absorption paths without using radiative transfer models.
In summary, the search has found that active tomographic DOAS is far more com-mon than the passive counterpart (11 out of 13 articles discussed this method). This preference can be explained by the fact that the results produced by this kind of sys-tem are generally superior to those obtained by passive methods. However, passive applications are normally much less demanding on a technical level, and are simpler to run and assemble. Much as a result of this, we have also identified that the systems used in the literature were not mobile or had a very low mobility level which in turn caused that all the systems were working with low projection numbers. This should be taken into account in future research on the topic. As a final note, we would also like to point out that there is no commercially available systems for this kind of application, although some of the articles, like the one by Stutz in 2016 [84] detail systems which could easily be adapted to that end.