Tropospheric nitrogen dioxide inversions based on spectral measurements of scattered sunlight

• Main Author: Vlemmix, T.
• Format: Doctoral or Postdoctoral Thesis
• Language: English
• Published: Technische Universiteit Eindhoven 2011
• Subjects: /dk/atira/pure/sustainabledevelopmentgoals/sustainable_cities_and_communities; SDG 11 - Sustainable Cities and Communities; /dk/atira/pure/sustainabledevelopmentgoals/climate_action; SDG 13 - Climate Action; Aerosol Robotic Network
• Online Access: https://research.tue.nl/en/publications/3061e40c-7e9e-44ec-8888-8de85099514a; https://doi.org/10.6100/IR719874; https://pure.tue.nl/ws/files/3656165/719874.pdf

This thesis describes the development of inversion methods for tropospheric nitrogen dioxide (NO2), based on ground based observations of scattered sunlight with themulti-axis differential optical absorption spectroscopy (MAX-DOAS) technique. NO2 is an atmospheric trace gas which, when present near the surface, is an important component of air pollution. On a global scale, fossil fuel combustion processes (power plants, automobiles) are the main source of NOx (=NO+NO2). NOx has a major impact on air quality, by its essential role in atmospheric photochemistry mechanisms: it influences tropospheric ozone formation as well as the levels of other oxidants (such as the hydroxyl and peroxy radicals), which catalyze the removal of species like carbonmonoxide, methane and other hydrocarbons from the atmosphere. In addition, NOx affects the formation of aerosols. Through the combination of these effects, tropospheric NOx reduces the radiative forcing, and therefore has a cooling effect on climate. A MAX-DOAS instrument measures wavelength spectra in the UV/Vis, with a resolution of less than one nanometer, in multiple viewing directions relative to the horizon. These measurements are analyzed with the differential optical absorption spectroscopy (DOAS) method, which can distinguish between the traces gases absorbing in a certain spectral window by making use of the fact that each gas has a unique spectral fingerprint. MAX-DOAS type of observations are complementary in two ways to other measuring techniques for NO2: Firstly in a temporal sense, relative to space-borne observations from the current generation of polar orbiting satellites, which frequently have no more than one observation per day. Secondly the MAX-DOAS observations are complementary in a spatial sense to in-situ monitors, which can only measure NO2 at the surface. MAX-DOAS instruments are especially suitable to measure tropospheric NO2 columns. This quantity is, more than NO2 concentrations measured at the surface, relevant for studies of transport and of trends in total amounts of tropospheric NO2. In this work it is investigated which information about the total amount and vertical distribution of tropospheric NO2 is contained in the MAX-DOAS measurements, and how this information can be extracted through inversemodeling (retrieval). What are the main error sources and which assumptions have to be made? Special attention is paid to aerosols, which have a large impact on the MAX-DOAS measurements. In addition, the developed retrieval methods are applied to a 14 month data set of MAXDOAS measurements performed in De Bilt. This data set was obtained as part of this research, and is unique for the Netherlands. The results are compared to satellite observations and to an air quality model. The first research described in this thesis (Chapter 3), focuses on the retrieval of tropospheric NO2 columns under clear sky conditions. A method was developed to derive differential air mass factors, by taking into account the effect of aerosols on the NO2 measurements. This was done in a new way: it was demonstrated that the aerosol optical thickness could be derived, for each elevation separately, by solely making use of MAX-DOAS measurements of relative intensity. It was assumed that all NO2 and aerosols were contained in a homogeneously mixed boundary layer of 1 km height. With this method, aerosol corrected air mass factors were derived for the elevations 4 degrees, 8 degrees and 16 degrees. Vertical columns could be derived for each elevation separately. Within this set of elevations, the 4 degrees elevation has the advantage of being most sensitive to trace gases in the boundary layer, but this viewing direction is also most sensitive to errors in the assumed aerosol and NO2 profile shape. With increasing elevation, the sensitivity to traces gases decreases, as well as errors due to wrong assumptions about the profile shapes. The retrieval method was applied to clear sky periods within the 14 month data set of MAX-DOAS observations performed in De Bilt. A comparison with AErosol RObotic NETwork (AERONET) observations of aerosol optical thickness showed good results: correlation of 0.85 was found, and a slope of the linear fit close to one. Comparison with space-borne retrievals of tropospheric NO2 columns retrieved by the OzoneMonitoring Instrument (OMI) shows on average reasonable results (correlations between 0.64 and 0.88 for different subsets), but individual comparisons can differ by more than a factor of two. This was attributed for the largest part to differences in spatial representativity, mostly in the horizontal direction, but also in the vertical. The second study (Chapter 4) focused on the question which information about the vertical distribution of aerosols and NO2 is contained in the MAX-DOAS observations. Vertical profile information derived from MAX-DOAS observations is not only relevant to improve the accuracy of the column retrieval, but it is needed as well in comparisons with other measurement techniques, such as satellite and lidar, and to derive NO2 surface concentrations, which are more directly related to air quality than vertical columns. Profile retrieval is challenging, since the profile information contained in MAX-DOAS measurements is known to be quite low. The topic is currently an active area of research in the MAXDOAS community: a range of approaches is being investigated, and no approach has yet emerged that can be considered a proven concept in all respects, partly because validation is a challenge as well. One of the main research questions addressed in this study, was the question if MAX-DOAS measurements can be used to distinguish between NO2 in the boundary layer and in the free troposphere. The basic retrieval model for tropospheric columns of the first study was expanded with two parameters for NO2, and one for aerosols: the height of the aerosol and NO2 layer above the surface was not longer assumed �fixed, and a second elevated NO2 layer was introduced at a fixed altitude in the free troposphere. In addition O4 measurements were used instead of relative intensity measurements, in order to better characterize the aerosol extinction profile in the boundary layer. Sensitivity studies were performed to investigate the retrieval accuracy for different noise levels, and for aerosol and NO2 profile shapes that were different from those assumed in the retrieval model. This led to the following conclusions: Firstly, if (in reality) NO2 is present above the boundary layer, and if the retrieval model allows an elevated NO2 layer at the same altitude as the real layer, then the amount of elevated NO2 can in principle be retrieved with reasonable accuracy. Secondly, for retrieval models which allow an elevated NO2 layer, tropospheric NO2 may also be retrieved when it is not present in reality. This may be the case for example when the aerosol andNO2 profiles shapes in the boundary layer are not well described by the retrieval model, or when the signal to noise level is low. Finally, despite the fact that MAX-DOAS measurements frequently do not contain more than two or three pieces of information to describe the NO2 profile, a retrieval model with only three free parameters will frequently be too rigid to perform accurate retrievals. In Chapter 6 of this thesis a more flexible approach is proposed. The profile retrieval approach was applied to MAX-DOAS measurements taken at the Cabauw Intercomparison campaign for NItrogen Dioxide measuring Instruments (CINDI). Comparison with independent observations of tropospheric NO2 columns from a lidar and NO2 surface concentrations from an in-situ monitor, showed on average good agreement (an average difference below 5 percent for both), but significant differences for individual cases. The third part of the research (Chapter 5) consisted of a comparison of tropospheric NO2 columns derived from the 14 month data set (that was also used in the first study) with tropospheric columns from the regional air quality model Lotos-Euros, which was run on a resolution of approximately 7x7 km. Whereas the comparison with satellite observations (Chapter 3) could only be performed under cloud free conditions, and at most two times per day, the comparison with the air quality model could be performed by making use of all MAX-DOAS observations. The total data volume therefore increased by a factor of 30, which improved the statistical significance, and allowed more detailed case studies. In order to analyze MAX-DOAS measurements under cloudy conditions, it was required to develop a third retrieval approach. This was based on MAX-DOAS observations at 30 degrees elevation (and the zenith reference), in order to be least sensitive to errors in the assumed NO2 profile shape, and to aerosols (the retrieval of which is difficult under cloudy conditions). Air mass factors were derived using information about the boundary layer height from a meteorological model. In addition, lidar observations of cloud bottom height were used. The comparison between Lotos-Euros and MAX-DOAS showed on average a good agreement (an average difference below 1 percent, and for daily averages over cloud free days a correlation of 0.8). The agreement found was surprising, especially when considering the fact that a bottom-up approach (the model) is compared to a top-down approach (the measurements). Furthermore, a remarkably good agreement was found for the tropospheric NO2 column averaged per sector of the wind direction. This indicates that the average tropospheric NO2 column that is measured in De Bilt, is not dominated by local sources, such as nearby highways with frequent traffic jams (such emissions are difficult to capture in the model), but rather by emissions in densely populated and industrial areas further away, e.g. the cities of Rotterdam, Amsterdam, Antwerp, Brussels and the German Ruhr area. It appears that tropospheric NO2 columns measured with MAX-DOAS can be used for validation of (high resolution) chemistry transport models in urban regions, and the same may be expected for satellite observations with a sufficiently small resolution. This is especially relevant because it is known that comparison between satellite and in-situ is problematic in urban regions (due to the large difference in spatial representativity). No observations were performed in summer months. It may however be expected that because of the shorter lifetime of NO2 in summer, nearby sources would have a larger relative impact on the MAX-DOAS observations, and therefore lead to less agreement with the model than as found in this study. For individual comparisons on an hourly and day-to-day basis, observed differences could be substantial. This is mainly attributed to the fact that within the model actual emissions cannot be described on a high enough spatial and temporal resolution. In addition, differences could be large for example when the observed wind direction or wind speed was different from that in the model, or in the weekend: Observations showed on average a clear decrease in the weekend, compared to the rest of the week, whereas the model showed a less pronounced weekly cycle. The three studies described in this work lead to the following general conclusions about MAX-DOAS observations of NO2: Firstly, it has been demonstrated that long-term MAX-DOAS observations of tropospheric columns are particularly suitable for validation of space-borne observations and air quality models. When averaged over long-enough periods, patterns show up (e.g. as a function of time, wind direction, or another quantity) that would not be seen for individual comparisons, or not even for a few months of observations, due to differences in representativity and limited accuracy. Thorough satellite and model validation therefore requires a large network of MAX-DOAS sites, on locations with a variety of conditions with respect to NO2 and aerosols. There is currently no other ground based method that can provide automated tropospheric NO2 column observations for such a low cost per observation. Secondly, it is concluded that with a simple algorithm, based on a high viewing elevation (e.g. 30 degrees), tropospheric NO2 columns can be retrieved with reasonable accuracy, for a large number of different aerosol and NO2 scenarios, and without strong dependence on a-priori assumptions. With respect to NO2 profile retrieval, the situation is different: due to the many aspects that influence the measurements, the required accuracy, and the relatively low information content, the accuracy of individual retrievals is generally not very high (especially for the free troposphere), and depends strongly on a-priori assumptions, as well as on the atmospheric conditions at the time of measurement. To make optimal use of the information content, it is important that the profile shape parametrization is highly flexible for the lowest 1-2 kilometers of the atmosphere. Profile retrieval accuracy is highest and depends least on a-priori assumptions for cloud free situations, when the aerosol optical thickness is low, when the NO2 is located in a boundary layer of which the top lies between approximately 400 m and 1.5 km altitude, and when in addition the MAX-DOAS instrument is aimed away from the sun. Finally, it is concluded that more validation of MAX-DOAS retrieval methods is needed. This requires long-term observations in the presence of other instruments that can be used for comparisons, such as those present at the CINDI campaign. Such comparisons should be performed in various seasons and under various conditions with respect to the abundance of aerosols and NO2.