NO 2 lidar profile measurements for satellite interpretation and validation

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114,, doi: /2009jd012441, 2009 NO 2 lidar profile measurements for satellite interpretation and validation H. Volten, 1 E. J. Brinksma, 2,3 A. J. C. Berkhout, 1 J. Hains, 2 J. B. Bergwerff, 1 G. R. Van der Hoff, 1 A. Apituley, 1 R. J. Dirksen, 2 S. Calabretta-Jongen, 1,2 and D. P. J. Swart 1 Received 7 May 2009; revised 11 August 2009; accepted 24 August 2009; published 19 December [1] Satellite instruments are efficient detectors of air pollutants such as NO 2. However, the interpretation of satellite retrievals is not a trivial matter. We describe a novel instrument, the RIVM NO 2 mobile lidar, to measure tropospheric NO 2 profiles for the interpretation and validation of satellite data. During the DANDELIONS campaign in 2006 we obtained an extensive collection of lidar NO 2 profiles, coinciding with OMI and SCIAMACHY overpasses. On clear days and early mornings a comparison between lidar and in situ measurements showed excellent agreement. At other times the in situ monitors with molybdenum converters suffered from NO y interference. The lidar NO 2 profiles indicated a well-mixed boundary layer, with high NO 2 concentrations in the boundary layer and concentrations above not differing significantly from zero. The boundary layer concentrations spanned a wide range, which likely depends on the wind directions and on the intensity of local (rush hour) traffic which varies with the day of the week. Large diurnal differences were mainly driven by the height of the boundary layer, although direct photolysis or photochemical processes also contribute. Small-scale temporal and spatial variations in the NO 2 concentrations of the order of 20 50% were measured, probably indicative of small-scale eddies. A preliminary comparison between satellite and lidar data shows that the satellite data tend to overestimate the amount of NO 2 in the troposphere compared to the lidar data. Citation: Volten, H., E. J. Brinksma, A. J. C. Berkhout, J. Hains, J. B. Bergwerff, G. R. Van der Hoff, A. Apituley, R. J. Dirksen, S. Calabretta-Jongen, and D. P. J. Swart (2009), NO 2 lidar profile measurements for satellite interpretation and validation, J. Geophys. Res., 114,, doi: /2009jd Introduction [2] Satellite instruments have great potential for detecting tropospheric pollutants such as NO 2 with global coverage within one or a few days. The validation and interpretation of satellite retrievals under polluted conditions is not a trivial matter (see, e.g., Celarier et al. [2008] for an extensive overview of validation measurements and methods). [3] One of the major sources of uncertainty in satellite NO 2 tropospheric column retrieval is the assumed vertical distribution of NO 2, since the shape of the vertical profile of NO 2 influences the (physical) air mass factors [Boersma et al., 2004, 2002; Bucsela et al., 2006]. For example, the Dutch OMI NO 2 algorithm [Boersma et al., 2007] uses online simulations of NO 2 vertical distributions from the TM4 chemistry transport model; these simulated profiles 1 Centre for Environmental Monitoring, Netherlands National Institute for Public Health and the Environment, Bilthoven, Netherlands. 2 Climate Research and Seismology Department, Royal Netherlands Meteorological Institute, De Bilt, Netherlands. 3 Now at TEC-EE and EOP-SM, ESTEC, ESA, Noordwijk, Netherlands. Copyright 2009 by the American Geophysical Union /09/2009JD affect the retrieved total and tropospheric columnar NO 2 amounts. [4] Measured NO 2 profiles are scarce. So far, most efforts to measure profiles pertain to aircraft-based in situ measurements [Heland et al., 2002; Martin et al., 2006; Bucsela et al., 2008; Boersma et al., 2008], which are unable to obtain NO 2 concentrations in the lower boundary layer where most of the NO 2 resides. Schaub et al. [2006] reported groundbased NO 2 profiles constructed from surface measurements at different altitudes in the Alps. These measurements covered a large spatial domain of 400 km by 250 km, so horizontal gradients may be present in the obtained profile. [5] In this paper, our first aim is to provide a description of a novel instrument, the NO 2 mobile lidar developed at Netherlands National Institute for Public Health and the Environment (RIVM). This instrument is unique because it was explicitly designed to measure complete lower tropospheric NO 2 profiles for use in satellite validation efforts; most NO 2 mobile lidars are intended to detect local sources of NO 2 [Fredriksson and Hertz, 1984; McDermid et al., 1990]. [6] Our second purpose is to show NO 2 lidar results obtained with the new instrument during the DANDELIONS campaign in September 2006 in Cabauw [Brinksma et al., 1of18

2 Figure 1. Simplified scheme of the lidar system. The emitter unit is formed by a pulsed Nd:YAG pump laser and a dye laser. The receiver unit consists of a telescope, an interference filter, and a photomultiplier tube (PMT). The spectrophotometer unit consists of a beam sampler (M1) with a reflectance of 1% and a beam splitter (M2) with a reflectance of 50%. EM1 and EM2 are energy meters, and GC indicates a gas cell filled with NO ]. The lidar measurements of NO 2 concentration profiles ranged from a few meters above the ground up to about 2.5 km, well above the top of the boundary layer and they coincided with satellite overpasses of either OMI or SCIAMACHY. The measurements gave insight to the actual shape of the NO 2 profiles, and how the profiles may vary from day to day, depending, for example, on the intensity of local (rush hour) traffic and on the wind direction. In addition, the NO 2 lidar data contained unique information on how the profiles may vary during the day, i.e., for different overpass times in the morning (SCIAMACHY) and afternoon (OMI), but also on the time scales of minutes. We studied the spatial inhomogeneity and the vertical and temporal variability of NO 2 in detail by measuring lidar time series and lidar measurements alternating between two azimuth directions at one height. These inhomogeneities of the atmosphere in time and space are highly relevant for the interpretation and validation of satellite measurements [Celarier et al., 2008]. [7] Our lidar measurements were directly compared with collocated ground-based measurements, in particular with in situ values of two chemiluminescence NO 2 monitors with molybdenum converters used for monitoring activities in the framework of the RIVM Dutch National Air Quality Monitoring Network (LML). These monitors were located on the tower at Cabauw, at ground level and at a height of 200 m. [8] We compared integrated boundary layer NO 2 concentrations obtained with the lidar (using boundary layer height observations) with satellite tropospheric columns. The impact of the assumed profile shape on the satellite retrievals is described quantitatively elsewhere [Hains et al., 2009]. [9] In section 2 we describe the RIVM mobile lidar system. We proceed with a brief description of in situ chemiluminescence NO 2 monitors with molybdenum converters, the boundary layer lidar, and the satellite instruments OMI and SCIAMACHY. In section 3 we present the measured vertical NO 2 lidar profiles and show examples of the observed differences from day to day. Further, we demonstrate the diurnal variability measured by the NO 2 lidar. In addition we performed measurements dedicated to the study of small-scale temporal and spatial heterogeneity in the NO 2 concentrations. A comparison between lidar NO 2 values and in situ NO 2 monitors is given in section 4. In section 5 we explain how to obtain a provisional comparison between NO 2 lidar and satellite data. We end with conclusions and recommendations in section Experimental Methods 2.1. RIVM Mobile Lidar System [10] We measured NO 2 profiles with a lidar system developed at RIVM [Swart and Bergwerff, 1989; Berkhout et al., 2006]. The entire system is housed in a truck, constituting a fully self-supporting mobile laboratory. Below we describe the design of the instrument, and we explain how the instrument is operated to obtain NO 2 vertical profiles Design [11] The RIVM mobile lidar (light detection and ranging) system uses the DIAL (DIfferential Absorption Lidar) technique to measure trace gas concentrations in the atmosphere [Measures, 1984; McDermid et al., 1990]. Most systems of this type [e.g., Fredriksson and Hertz, 1984; Edner et al., 1987] are built for the remote monitoring of industrial emissions within the framework of the enforcement of environmental regulations. In contrast, the lidar system described in this paper has a more flexible design to make it suitable for air-quality or climate-related process studies and the validation of satellite instruments as well. 2of18

3 Figure 2. Spectrum of NO 2 as measured with the spectrophotometric system. On and off wavelengths are indicated. For the latter application the lidar is designed to be able to measure vertical NO 2 concentration profiles. [12] The lidar system design is outlined in Figure 1. The lidar system consists of an emitter unit, a receiver unit, a spectrophotometer unit, and a processing and control unit (not shown). [13] The emitter unit consists of a pulsed pump laser dye laser combination (see Figure 1). The pump laser is a Spectra-Physics Quanta-Ray Pro 250 Nd:YAG laser. It runs at 30 Hz and produces pulses of 8 10 ns, with an energy per pulse of about 420 mj at 355 nm. This light is used to pump a Sirah PrecisionScan PRSC D 30 dye laser. The resonator of the dye laser uses a single 1800 lines/mm grating, which results in a line width of 2.4 pm. For measuring NO 2 we use the dye Coumarin 2 (also known as Coumarin 450). With this dye we produce pulses of about 40 mj at 448 nm, with a pulse duration of about 10 ns. The divergence of the laser beam is about 0.5 mrad. We tune the dye laser to the proper wavelength using a stepper motor. In addition, we use a piezoelectric actuator to rapidly detune and retune the resonator by 0.8 nm. This detuning takes place between two laser pulses and enables the system to alternate continuously between the two wavelengths required by the DIAL technique, in our case nm and nm, thus generating on-off pairs at 15 Hz. The latter wavelength is absorbed more strongly by NO 2 than the former. The on and off signals are thus measured virtually simultaneously. Since the on and off wavelengths are so close together, the wavelength dependence of aerosol backscatter and extinction can be safely neglected over this wavelength range. [14] We direct the laser beam to the atmosphere through a set of mirrors and prisms. The final two mirrors are mounted on a movable platform, together with the receiver optics. When measuring, this platform is lifted out of a hatch in the laboratory roof, allowing access to the atmosphere. When traveling, it is lowered and the hatch is closed, protecting the inside from the elements. The beam can be pointed at any direction: in the vertical plane from the horizon to the zenith, in the horizontal plane almost 360 around. [15] The receiver unit consists of a telescope, an interference filter, and a photomultiplier tube (Figure 1). The telescope is a Celestron C11 Schmidt-Cassegrain telescope. It has a clear aperture of 280 mm and a field of view of 1.1 mrad. The interference filter blocks out the ambient light, its passband is centered at nm and has a width of 3.4 nm (FWHM). The filter is mounted in a temperature controlled housing because the passband wavelength depends on the temperature. Most of the daylight is blocked by the interference filter. The photomultiplier tube is a Thorn-EMI 9818 QA tube. [16] The light emitted by the laser has a line width narrower than small features in the NO 2 spectrum. This makes the instrument sensitive to small variations in wavelength calibration, such as might occur if the temperature varies in the truck where the laser is located. Also the line width is different from that of instruments used to determine literature values for the NO 2 cross sections. To address these issues, a spectrophotometer unit (Figure 1) is used to determine and monitor the cross section of the trace gas under investigation, in our case NO 2. [17] In this spectrophotometer unit, a small portion of the outgoing laser beam is diverted and split in two. One part is directed onto an energy meter, capable of measuring the energy of single laser pulses (EM1 in Figure 1). The other part is passed through a gas cell containing the trace gas (GC in Figure 1) before being measured by a second energy meter (EM2 in Figure 1). By dividing the measurements of EM1 and EM2, an absorbance or transmittance value is obtained. Repeating this procedure at a range of wavelengths enables the measurement of a spectrum (Figure 2). The spectral resolution is determined mainly by the line width of the laser. Because the NO 2 concentration in the gas cell is well known, we can derive from these spectra an instrument-specific differential cross section, in this case of 3142 cm 2 /g ± 10%. This system is running during a DIAL 3of18

4 Figure 3. Combining measurements at different elevation angles (dir1 and dir2) into a vertical profile; r1 is the closest measurement range (typically 300 m), and r2 is the longest measurement range (typically 2.5 km). Lidar measurements obtained at a single elevation angle may yield more than one NO 2 concentration data point (indicated in red). measurement for online monitoring of the cross section; if it changes, the wavelength is retuned. The cross sections of the NO 2 molecules in the atmosphere depend on temperature and pressure. Radio sonde measurements (see AVDC) showed that the temperature and pressure range over the altitude spanned by the lidar measurements is relatively small since we probe only the lowest part of the atmosphere. Differences in temperature or pressure may cause a deviation in the cross sections relative to the spectrophotometer derived values of at maximum 3 4% [Harwood and Jones, 1994; Vandaele et al., 1998]. [18] The lidar instrument is fully computer controlled. All functions instrument control, data acquisition, data processing are handled by an integrated dedicated program, written on the LabVIEW development platform. Data acquisition is done using a 40 MHz 12 bit transient recorder (Licel TR ). This samples the signals with a range resolution of 3.75 m. [19] The entire system is housed in a custom built mobile laboratory, 8 m long, 2.5 m wide and 2.3 m high, mounted on a vehicle. It is fully self-contained: it has a generator, climate control and a system to provide the laser with cooling water. The complete emitter and receiver units, including the movable telescope, are mounted on a single aluminum frame. When traveling, this frame is supported by actively damping bellows, filled with pressurized air, to reduce the risk of damage and loss of alignment due to shock or vibration. When measuring, the air bellows are emptied of air and the frame rests firmly on the laboratory floor. This floor is stabilized by underpinning the entire vehicle with hydraulically retractable supports Measurement Method of NO 2 Vertical Profiles [20] The lidar instrument measures backscattered signals, I on and I off, at the on and the off wavelengths. From these signals a DIAL curve is derived as follows [Measures, 1984]: dial ¼ ln I on I off : The DIAL curve has a range resolution of 3.75 m. The slope of the DIAL curve is a measure for the NO 2 concentrations in the atmosphere as follows: ½NO 2 Š¼ ð1þ slope ; ð2þ c where c is the differential cross section in cm 2 /g. The slope of the DIAL curve changes as a function of range, depending on the way the NO 2 gas is distributed over the atmosphere. To obtain data points, we select range intervals of the DIAL curve that have a constant slope, i.e., a constant NO 2 concentration. Generally, one or two data points are obtained per DIAL curve. The noise on the DIAL curve and the length of the selected range intervals determine the precision of the concentration data points. [21] In principle, a vertical concentration profile can be made by pointing at zenith. However, the lidar only yields results when the laser beam is fully in the field of view of the telescope, which occurs in our system beyond a distance of 300 m. Thus, the lidar yields no results for the first 300 m. When measurements near the surface are required, the emitter section and receiving telescope are tilted under various angles and we build profiles from a number of data points, obtained from DIAL curves each measured at a different elevation angle. This also enhances the vertical resolution and increases the precision. The principle is 4of18

5 Figure 4. A NO 2 profile is obtained from a sequence of measurements at six elevations, in the following order: 0.75, 3, 12, 24, 6, and 1.5. This sequence is repeated typically 50 times. An additional measurement directed at 90 is performed after the sequences are finished. The elevation angle of 12 probes the atmosphere roughly over an altitude range from 70 m to 500 m and is the default for measurements of spatial and temporal variability. shown schematically in Figure 3 for two elevation angles; operationally NO 2 profiles are obtained from measurements at seven different elevation angles. [22] First, a DIAL measurement is done in the direction dir1. This takes 6.7 s and includes 200 laser shots in 100 onoff pairs. The process is repeated for the direction of dir2 at a different elevation angle. The sequence dir1 dir2 is then repeated, typically 50 times for a profile measurement, and the results averaged to reduce the noise. The NO 2 concentration determined from the slope of the DIAL curve over the range interval from r1 to r2 is then attributed to the vertical interval from h1 to h2, and the concentration measurement over the range interval from r2 to r3 is attributed to the vertical interval from h2 to h3. [23] A measurement sequence during the DANDELIONS campaign consisted of measurements at six elevations, in the following order 0.75, 3, 12, 24, 6, and 1.5 (Figure 4). Data points obtained from DIAL curve measurements at these elevation angles may overlap in altitude range. The sequence of elevation angles was chosen to minimize the time needed for changes in elevation angle during the measurements. Elevations close to the horizontal yield NO 2 concentrations at low altitudes but pertaining to a certain horizontal extent away from the instrument (for the slant measurements, typically between about 300 m and 2.5 km away from the instruments), whereas a zenith observation yields NO 2 concentrations exactly above the instrument. [24] A sequence takes 50 s, and is repeated 50 times, taking 42 min in total. Then an additional measurement directed at the zenith (90 ) is performed, and repeated 50 times, taking about 6 minutes with an added 2 min for maneuvering the platform with the receiver optics. The latter elevation angle is not included in the sequence of the other 6 directions since it would entail repeatedly moving the live laser beam over a large angle, which is deemed irresponsible for safety reasons. A full profile measurement is thus completed in 50 min. [25] The method described above to construct a vertical profile of NO 2 from lidar scans at multiple elevations angles only truly works when the atmosphere is horizontally homogeneous and the scans are interleaved. However, because the vertical measurement is 300 m to 2.5 km from the region probed by the low-angle scans, and because the 6 min vertical measurement is done after the 42 min lowangle scan sequence, changes in NO 2 concentration, for example, due to a rapidly rising boundary layer and subsequent dilution of NO 2, may cause differences in the measured NO 2 concentration at 90 compared to the earlier shallow-angle measurements. Therefore, we only use the vertical measurement to obtain NO 2 concentrations well above the boundary layer top; NO 2 values below the boundary layer top are rejected. [26] We obtained profiles consisting of about six values inside the planetary boundary layer, and one or two above this layer. A vertical profile provides data for altitudes 5of18

6 simultaneously measures NO 2 concentrations of two volumes of air at the same altitude but approximately 2 km apart. [28] Temporal inhomogeneities are investigated by measuring in one azimuth direction for 20 to 40 min and determining averages over an interval of 2.5 or 3.75 min. The area of the atmosphere that is sampled during such a measurement depends on the range of the lidar beam r2 r1 (see Figure 3), the duration of the measurement T, the direction of the wind z, relative to the lidar direction q, and the wind speed v (see equation (3)). A ¼ðr2 r1þv T jsinðq zþj: ð3þ Figure 5. Vertical resolutions of the NO 2 lidar instrument as functions of altitude during the DANDELIONS campaign in Apart from the altitude the vertical resolution strongly depends on how the NO 2 is distributed in the atmosphere, resulting in a broad area of possible vertical resolution values, in particular at high altitudes. As long as the NO 2 concentrations are constant over a certain altitude range, the choice for a longer altitude interval, i.e., a worse vertical resolution, will result in a higher precision. Therefore the choice for a certain vertical resolution is a trade-off with precision in NO 2 concentration. ranging from a few meters, for small elevation angles, up to approximately 2500 m for an elevation angle of 90 (see Figure 4) [see also Brinksma et al., 2008], with a precision of mg NO 2 m 3 (1 mg NO 2 m 3 = molecules/m 3 = 0.49 ppb [ K, hpa]). The vertical resolution of a profile varies, it depends on the projection of the elevation angle, but also on how the NO 2 is distributed in the atmosphere. This results in a broad area of possible vertical resolution values, in particular at high altitudes. In Figure 5 we plotted the actual vertical resolutions during the DANDELIONS campaign as a function of altitude. As long as the NO 2 concentrations are constant over a certain altitude range, the choice for a longer altitude interval, i.e., a worse vertical resolution, will result in a higher precision. Therefore, the choice for a certain vertical resolution is a trade-off with precision in NO 2 concentration. This choice is made during the analysis. From the same data, higher resolution profiles can be obtained at the cost of precision Measurements of Spatial and Temporal Variability [27] In addition to profile measurements, we used the lidar to investigate spatial and temporal changes in the horizontal NO 2 field. In this alternative operational mode, the lidar instrument measures at a fixed elevation angle of 12 (see Figure 4). At this elevation angle the atmosphere is probed roughly over an altitude range from 70 m to 500 m; the exact range covered may differ from measurement to measurement, since it depends on the atmospheric conditions, for example, the boundary layer height and the way NO 2 is distributed in the atmosphere. Spatial inhomogeneities in the NO 2 concentrations are studied by every 3.75 s alternating between two azimuth directions, 75 apart, for approximately half an hour. In this way, the lidar nearly The largest area is covered when the wind direction is perpendicular to the lidar direction; when the wind blows in the same direction as the lidar beam the area covered is at its smallest (see Figure 6). If, for example, the lidar measured over a range of 2.2 km, the wind speed was 5 m/s, and the wind direction was perpendicular to the lidar direction, in 3.75 min, we sampled an area of 2.2 km by 1.1 km. By choosing time intervals of a few minutes, for example, 2.5 or 3.75 min, we study inhomogeneities of roughly 1 km 2. [29] Wind speeds and wind directions are listed in Table 1 for around 1200 UTC on each measurement day In Situ NO 2 Monitors [30] During the DANDELIONS campaign in 2006, two in situ NO 2 monitors were used. The first one operated at ground level in Cabauw as part of monitoring activities in the framework of the RIVM Dutch National Air Quality Monitoring Network (LML, see which consists of 45 sites where nitrogen oxides are monitored on a continuous basis with 1 min intervals, using automatic analyzers based on chemiluminescence, and using molybdenum converters (Thermo Electron 42W). The second one is an identical system, which for the purposes of the campaign was placed on top of the Cabauw tower, at 200 m altitude. The instruments measure NO x and NO concentrations and derive the NO 2 concentration by subtraction: [NO x ] [NO] = [NO 2 ]. The measuring range of Figure 6. The area of the atmosphere that is sampled during a measurement depends on the useful range section of the lidar beam (red line), the direction of the wind relative to the lidar direction, and the wind speed. The largest area is covered when the wind direction (Wind 1) is perpendicular to the lidar direction (indicated in blue). If the wind direction is 30 (Wind 2) relative to the lidar direction, the area is halved (indicated in black), and when the wind blows along the lidar beam (Wind 3), the area covered is at its smallest (indicated in green). 6of18

7 Table 1. Dates and Times When NO 2 Lidar Measurements Were Performed, Average Wind Speeds and Wind Directions Around 1200 UTC, Type of Lidar Measurement, and, for Lidar Profile Measurements, the Corresponding Satellite Overpass Time at Cabauw a Date v (m/s) z ( E of N) Measurement Time (UTC) Satellite Overpass Time (UTC) 6 Sep profile OMI Sep azimuth scan time series profile OMI 1135 profile OMI Sep profile OMI 1218 time series profile OMI Sep profile OMI 1123 time series profile OMI Sep azimuth scan time series profile OMI 1205 azimuth scan profile OMI 1344 time series Sep profile SCIA 0959 azimuth scan time series profile OMI 1248 profile OMI Sep azimuth scan time series profile OMI 1153 time series azimuth scan profile OMI Sep profile SCIA 1047 profile OMI 1200 profile OMI Sep profile SCIA 1016 azimuth scan time series profile OMI 1242 time series Sep profile SCIA 0944 time series profile OMI 1148 profile OMI 1325 a Average wind speed, v; average wind direction, z. The local civil time in Cabauw is UTC + 2 h. The local solar time is UTC + 20 min + (equation of time), where the equation of time is on the order of +1 to +7 min between 6 and 22 September. the monitor is mgno 2 m 3 with a detection limit of 1 mg NO 2 m 3 [Van Elzakker, 2001]. Similar monitors are used in monitoring networks all over the world. Due to technical problems during the 2006 campaign, no data were recorded at surface level before 5 September and from 7 to 10 September. [31] Previous studies have shown that NO 2 concentration values from chemiluminescence instruments with molybdenum converters may have a positive bias. This bias results from a sensitivity to NO y components; that is, they experience interference from other oxidized nitrogen compounds such as peroxyacetyl nitrate and nitric acid (HNO 3 )[Ordóñez et al., 2006; Dunlea et al., 2007; Steinbacher et al., 2007] Boundary Layer Lidar [32] Information about the top of the boundary layer is important for the interpretation of the measured lidar NO 2 profiles. This information was obtained using an aerosol backscatter lidar that provided continuous information on boundary layer height and vertical aerosol distribution with high temporal resolution (5 min) and with a typical precision of 50 m [Apituley et al., 2000; Schaap et al., 2009]. An overview of all instrumentation, their data products and the use of the suite of measurements at CESAR is described by Russchenberg et al. [2005] and at Satellite Instruments [33] In this paper we use data obtained with two different satellite instruments, OMI and SCIAMACHY. Both retrieve NO 2 from the VIS part of the spectrum using DOAS-type methods. SCIAMACHY passes over Cabauw about every three days, between 0900 and 1100 UTC. OMI has one or two overpasses over Cabauw daily, between 1100 and 1400 UTC OMI [34] The Ozone Monitoring Instrument (OMI) is a spaceborne nadir viewing imaging spectrometer operating in the UV-VIS wavelength range ( nm). Among the primary objectives of the OMI instrument is to measure the concentrations of ozone (O 3 ) and nitrogen dioxide (NO 2 ) for the ongoing monitoring of the Earth s strato- 7of18

8 spheric ozone layer and for the monitoring of tropospheric air quality. The 2600 km wide swath of the OMI instrument, which enables daily global coverage, is divided into 60 ground pixels that measure km along track and km across track, depending on the viewing angle. Information regarding the OMI mission objectives is given by Levelt et al. [2006a] and an elaborate instrument description is given by Levelt et al. [2006b] and Dobber et al. [2006]. [35] In this study we used the Dutch OMI NO 2 (DOMINO) data product, version 1.0.2, retrieved with collection 3 Level 1B data [Dobber et al., 2008], available from temis.nl [Boersma et al., 2007, also Dutch OMI NO2 (DOMINO) data product HE5 data le user manual, 2008, TEMIS Web site, which has been validated in other studies [Boersma et al., 2009] SCIAMACHY [36] SCIAMACHY (SCanning Imaging Absorption SpectroMeter for Atmospheric CartograpHY) is an imaging spectrometer designed to measure sunlight, transmitted, reflected and scattered by the Earth s atmosphere or surface in the ultraviolet, visible, and near infrared wavelength regions ( nm) at moderate spectral resolution ( nm) [Bovensmann et al., 1999]. [37] SCIAMACHY measurements yield the amounts and distribution of several trace gas species, such as O 3 and NO 2. Atmospheric columns of NO 2 are retrieved from SCIAMACHY nadir measurements using the DOAS method [Boersma et al., 2004; Blond et al., 2007]. SCIAMACHY achieves global coverage within three or six days, depending on the scanning mode. More on the instrument and the mission concept can be found, on the IUP Bremen Web site ( and on the TEMIS web site ( The tropospheric NO 2 column data used in this paper are created by TM4- NO2A, version 1.04, and have been obtained from the TEMIS project Web site ( 3. Results [38] The NO 2 lidar results were obtained during the dandelions campaign in DANDELIONS (Dutch Aerosol and Nitrogen Dioxide Experiments for validation of OMI and SCIAMACHY) is a project that encompasses validation of NO 2 measurements by OMI and SCIAMACHY, and of aerosol measurements by OMI and the Advanced Along-Track Scanning Radiometer (AATSR) [Smith et al., 2001], using an extensive set of ground-based and balloon measurements over the Netherlands. [39] The 2006 campaign took place from 1 to 30 September, at the Cabauw Experimental Site for Atmospheric Research (CESAR, N, 4.93 E, 0.70 m below mean sea level) [Russchenberg et al., 2005]. The lidar was located about 200 m from the meteorological measurement tower of Cabauw. An overview of the extensive data sets obtained during the 2005 and 2006 DANDELIONS campaigns is presented by Brinksma et al. [2008]. The ground data, including the NO 2 lidar data, are publicly available through the Aura Validation Data Center (AVDC; nasa.gov/) in numerical and graphical form. The NO 2 profiles were measured coinciding with satellite overpasses of either OMI or SCIAMACHY. [40] All NO 2 lidar measurements presented here were obtained on days with less than 20% cloud coverage during the DANDELIONS campaign in September 2006 as estimated by trained observers from the meteorological institute, and verified with boundary layer lidar measurements. We obtained 22 profile measurements in total; 4 profile measurement coincided with SCIAMACHY overpasses, and 18 profile measurements coincided with OMI overpasses. Profile measurements typically took about 50 min. In addition, we obtained 7 azimuth scans and 11 time series measurements. These measurements lasted about 20 to 40 min. In Table 1 we list the dates and times when different types of lidar measurements were performed. For the profile measurements we also list the corresponding satellite overpass time in Cabauw. [41] We first focus on data for two days, 9 September, which was a clean day, and 12 September, which was a polluted day. The data for these two days illustrate most of the general behavior we find on the other days. In section 3.2 we show the diurnal variability of the NO 2 profiles. In sections 3.3 and 3.4 the small-scale spatial and temporal variabilities occurring in the boundary layer are studied NO 2 Profiles on a Clear and a Polluted Day [42] We show examples of profile measurements for a relatively clean day, i.e., 9 September 2006, for two different OMI overpasses (Figure 7, top) and for a strongly polluted day, 12 September 2006, for a SCIAMACHY overpass and an OMI overpass (Figure 7, bottom). On both days the measurements illustrate the general observation that the NO 2 concentration is rather constant within the boundary layer and drops to zero within the precision of the measurement above the boundary layer. The development of the boundary layer on these two days was measured by the boundary layer lidar at Cabauw [van Pul et al., 1994], as shown in Figures 8 and 9. The yellow/green colors indicate the presence of aerosols in the boundary layer. The top of the boundary layers are marked by red dashed lines, as guides to the eye. On 9 September 2006, the boundary layer rose in the morning until about 1000 UTC and remained quite stable during the rest of the morning and the afternoon. On 12 September 2006, the boundary layer height again rose quickly during the morning, and kept rising until well into the afternoon, at least until about 1600 UTC. We derived hour values for the boundary layer heights from these boundary layer lidar measurements (see AVDC for the data in tabular form). They are indicated in Figure 7 by dashed lines. On many days the NO 2 concentrations dropped as the boundary layer rose; this was also the case on 12 September, the boundary layer concentrations dropped from about 50 mg NO 2 m 3 during the SCIAMACHY overpass in the morning when the boundary layer top was at 200 m, to about 23 mgno 2 m 3 during the OMI overpass in the afternoon when the boundary layer height had risen to about 500 m (Figure 7) and NO 2 -free air is mixed in from the free troposphere. In addition, direct photolysis of NO 2 may play a role in diminishing the NO 2 concentrations in the course of the day. [43] The NO 2 concentrations in the atmosphere depend on local sources, such as near-by traffic on local highways 8of18

9 Figure 7. Lidar NO 2 profiles on log scales (black circles and squares) and NO 2 monitor values (open and solid red circles) measured in Cabauw. The blue squares, arbitrarily placed at a height of 100 m, indicate estimates for NO 2 boundary layer concentrations derived from the tropospheric column observed by SCIAMACHY or OMI by assuming that all the NO 2 in the tropospheric column is evenly distributed throughout the boundary layer. Horizontal bars indicate ±1s values for the concentrations. Where no error bars are visible, they are smaller than the symbols plotted. Vertical bars indicate the height intervals over which concentrations have been determined. Boundary layer heights (BL) are indicated by grey dashed lines. (top left) Lidar profiles measured at 9 September 2006, coinciding with the first OMI overpass. (top right) Lidar profiles measured at 9 September 2006, coinciding with the second OMI overpass. (bottom left) Lidar profiles measured at 12 September 2006 around the time of the SCIAMACHY overpass. Since the boundary layer height passed the 200 m point around 1000 UTC, we determined NO 2 monitor values for a time interval before (open red circles) and after (solid red circles) 1000 UTC. (bottom right) Lidar profiles measured at 12 September 2006, coinciding with the OMI overpass. Due to technical problems the profile was measured in 24 min instead of 50 min. [e.g., Beirle et al., 2003; Bucsela et al., 2007; Boersma et al., 2009], and sources farther away, such as industries close to the harbor of Rotterdam and the industrial Ruhr area in Germany. To establish how NO 2 sources farther away contribute, we calculated 24 hour back trajectories to see where the air parcels over Cabauw originated. The back trajectories were calculated for different altitudes using the NOAA Air Resources Laboratory (ARL) HYbrid Single- Particle Lagrangian Integrated Trajectory (HY-SPLIT) model (Version 4) [Draxler and Rolph, 2003] ( and NCAR/NCEP Reanalysis meteorology. [44] The back trajectory calculated for 9 September 2006 at 1400 UTC showed that on this clean Saturday, the wind was coming from the north curling to the east, bringing in clean air from over the North Sea, resulting in low NO 2 concentrations (see Figure 7, top). By contrast, the back trajectory calculated for 12 September 2006 at 1300 UTC showed that the wind was coming from the southeast, bringing polluted air from the industrialized Ruhr area in Germany, resulting in quite high NO 2 concentrations (see Figure 7, bottom). [45] On the other days, the NO 2 values in the boundary layer were in between those measured on 9 and 12 September 2006 (see section 3.2). Back trajectory calculations for those days indicated that the winds were coming predominantly from southern directions, bringing in air from over France and Belgium, with the exception of 10 September On this Sunday, few local sources were expected but back trajectory calculations showed that the air was coming from 9of18

10 Figure 8. Boundary lidar measurements for 9 September 2006 as a function of time. Yellow and green indicate the presence of aerosols. A red dashed line has been drawn to indicate the top of the boundary layer. Solid red lines indicate the times of the satellite overpasses. The boundary layer remained relatively stable between the overpasses. the same direction as on 12 September 2006, i.e., from the industrialized Ruhr area in Germany. On the 10th the NO 2 concentrations were substantially higher than on the 9th, when clean air was brought in from the North Sea. However, NO 2 concentrations on the 10th were substantially lower than on the polluted 12th, because on a Sunday local sources produce less NO Diurnal Variations in the NO 2 Profiles [46] When we look at all the profiles and differentiate between lidar profiles measured coinciding with SCIAMACHY overpasses, and OMI overpasses before and after 1300 UTC, we see that the four profiles coinciding with the SCIAMACHY overpasses between 0900 and 1100 UTC span a wide range of observed concentrations, roughly between 10 and 50 mg NO 2 m 3. Average values centered around 20 mg NO 2 m 3 in the boundary layer (Figure 10). [47] The nine profiles coinciding with the OMI overpasses that took place before 1300 UTC show less spread (Figure 11). Minimum and maximum NO 2 concentrations are, respectively, about 2 mg NO 2 m 3 and 23 mg NO 2 m 3 at low heights. The average concentration in the boundary layer lay roughly at 10 mg NO 2 m 3. Figure 9. Boundary lidar measurements for 12 September 2006 as a function of time. Yellow and green indicate the presence of aerosols. A red dashed line has been drawn to indicate the top of the boundary layer. Solid red lines indicate the times of the satellite overpasses. The boundary layer height rose quickly in the morning. 10 of 18

11 Figure 10. Lidar NO 2 profiles on a log scale coinciding with SCIAMACHY overpasses, which took place between 0900 and 1100 UTC. Vertical bars indicate the height intervals over which concentrations have been determined. Horizontal bars indicate ±1s values for the concentrations. Where no error bars are visible, they are smaller than the symbols plotted. The boundary layer heights between 0900 and 1000 UTC (for 12, 21, and 22 September) or 1000 and 1100 UTC (for 20 September) are indicated by lines in the corresponding color for each profile. [48] The seven profiles coinciding with the OMI overpasses after 1300 UTC continued the observed trend (Figure 12). The maximum boundary layer NO 2 concentration was at about 17 mg NO 2 m 3 somewhat lower than for the earlier OMI overpass profiles, whereas at about 2 mgno 2 m 3 the minimum was the same. The average concentration in the boundary layer was lower than 10 mg NO 2 m 3. Figure 11. Lidar NO 2 profiles on a log scale coinciding with OMI overpasses that took place before 1300 UTC. Vertical bars indicate the height intervals over which concentrations have been determined. Horizontal bars indicate ±1s values for the concentrations. Where no error bars are visible, they are smaller than the symbols plotted. The boundary layer heights between 1200 and 1300 UTC are indicated by lines in the corresponding color for each profile. Figure 12. Lidar NO 2 profiles on a log scale coinciding with OMI overpasses that took place after 1300 UTC. Vertical bars indicate the height intervals over which concentrations have been determined. Horizontal bars indicate ±1s values for the concentrations. Where no error bars are visible, they are smaller than the symbols plotted. The boundary layer heights between 1300 and 1400 UTC are indicated by lines in the corresponding color for each profile. [49] When looking at the individual profiles, we often observe the NO 2 concentrations in the boundary layer decreasing slightly with height, up to the top of the boundary layer (see, e.g., the profiles for 11 and 12 September in Figure 11). This effect (up to a decrease of 10% in concentration) is probably partly due to decreasing atmospheric pressure, or because of mixing-in of NO 2 -free air from the free troposphere. In addition, the decrease in the NO 2 over NO x ratio with altitude (due to a slower NO x cycling at lower temperatures) may play a role. Also the presence of local sources may occasionally cause the NO 2 concentrations to be higher at low altitudes. Above the top of the boundary layer the NO 2 concentrations rapidly drop to zero (within the precision of the measurements). [50] A diurnal effect we frequently observe is that NO 2 concentrations drop in the course of the day as the boundary layer top rises, because the gas is being mixed upwards, and because after the morning rush hour traffic, an important NO 2 source, has decreased in intensity. [51] To show to what extent the concentration differences during the day may be attributed to the effect of the upward mixing of gas, we plotted in Figure 13 the average concentrations in the boundary layer during the OMI overpass before 1300 UTC, [NO 2 ] OMI<13:00, versus the corresponding average concentrations in the boundary layer during the other overpasses, [NO 2 ] x, normalized by boundary layer height (BLH) as follows: ½NO 2 Š norm ¼½NO 2 Š x BLH OMI<13:00 BLH x : ð4þ A linear fit, y = 0.8(0.2)x + 0.6(3.1) (1-s errors given in parentheses), to the data points is indicated by a solid line. The 1:1 line is indicated by a dotted line. If we assume a static box with a fixed amount of NO 2, vertical mixing and an increasing boundary layer, we would expect all data 11 of 18

12 Figure 13. The average concentrations in the boundary layer during the OMI overpass before 1300 UTC, [NO 2 ] OMI<13:00, versus the corresponding average concentrations in the boundary layer, [NO 2 ] x, during the SCIAMACHY overpasses (open circles) and the OMI overpasses after 1300 UTC (solid squares) normalized by boundary layer height (BLH). We assumed errors of 2 mg NO 2 m 3. The 1:1 line is indicated by a dotted line. A linear fit, y = 0.8(0.2)x + 0.6(3.1) (1-s errors given in parentheses), to the data points is indicated by a solid line. The two outliers (circled in red) pertain to the same day, 20 September. points to lie on the 1:1 line. The linear fit in Figure 13 indicates that a large part of the variations during the day may be attributed to this effect, but other effects are important as well. Changes in the sources of NO 2, or direct photolysis of NO 2 and photochemical loss of NO 2 due to reactions with OH and/or O 3 may contribute to the variations in the NO 2 concentrations observed over the day [Boersma et al., 2008, 2009]. Also, the box is not static: More (or less) polluted air may be advected. For example, the two outliers in Figure 13 (circled in red) pertain to the same day, 20 September, when during the day wind from the direction of Rotterdam may have carried additional NO 2. Excluding the outliers does not significantly change the results of the linear fit, although the 1-s errors are reduced Temporal Variations [52] The profile measurements give an indication of the variability in NO 2 concentrations from day to day, and during a day, as a function of height. Variability also occurs on shorter time scales. To study the temporal variation in the NO 2 concentrations we performed time series measurements by fixing the elevation angle at 12 and the azimuth direction at the default value of 289 E of N. We let the lidar run for min, and averaged the measurement over intervals of 2.5 or 3.75 min. Examples of the results are shown in Figures 14 and 15. [53] On 9 September, a clear day, we find variations in the concentrations between two ten-minute intervals from 2 to 4 mg NO 2 m 3. On 12 September, we were able to obtain data points at two altitudes from one measurement run. At both altitude intervals we see variations in concentrations of up to 10 mg NO 2 m 3 between 10 min intervals. [54] To obtain a more general measure for the temporal variability as observed in the time series measurements, we divided the maximum value by the minimum value for each time series measurement and plotted this ratio as a function of time and date (Figure 16). In this manner, we get an indication of the maximum variability of 2.5 to 3.75 min averages over a time span of 25 to 37.5 min. The average of this ratio over the whole campaign is 1.5 ± 0.2. Thus in the boundary layer the variability in NO 2 concentrations over time scales of a few minutes is about 50%. These smallscale variabilities probably are indicative of small eddies of roughly 1 1 km occurring in the atmosphere, ensuring the well-mixed boundary layer when averaging over an air volume of several km 3, such as observed in the profile results presented in section Spatial Variability [55] To look at the spatial variability in NO 2 concentrations we performed azimuth scans. We fixed the lidar at an elevation of 12, and switched every 6.7 seconds between two azimuth directions. The azimuth direction 289 EofNis the default direction, that is also used for the profile measurements and the time series measurements, and the second azimuth direction was chosen 75 further at 214 E of N (see Figure 17). On average the measured volumes were roughly 2 km apart. The azimuth scans took about 20 min. Depending on the atmospheric conditions we obtained per azimuth direction data points for one or more altitude ranges (see Figure 18). 12 of 18

13 Figure 14. Lidar NO 2 time series measurements taken at an elevation angle of 12 on 9 September 2006 showing short time-scale variations in the NO 2 concentrations. The lidar measurements have been averaged over 2.5 min. During this measurement an area of 0.36 km 2 is sampled in 2.5 min. Vertical bars indicate ±1s values for the concentration. Where no error bars are visible, they are smaller than the symbols plotted. Horizontal bars indicate the time intervals over which concentrations have been determined. The observed small-scale variations are probably indicative of eddies occurring in the atmosphere. [56] An exceptionally large spatial variation of about 20 mgno 2 m 3 is observed on 12 September; on other days the variation is on the order of a few mg NO 2 m 3. [57] In Figure 19 we plotted the NO 2 concentrations measured for azimuth direction 289 E of N divided by Figure 15. Lidar NO 2 time series measurements taken at an elevation angle of 12 on 12 September 2006 showing short time-scale variations in the NO 2 concentrations in two different altitude intervals. The lidar measurements have been averaged over 3.75 min. During this measurement an area of 0.14 km 2 is sampled in 3.75 min at the lower altitude range, and an area of 0.19 km 2 is sampled at the higher altitude range. Vertical bars indicate ±1s values for the concentration. Where no error bars are visible, they are smaller than the symbols plotted. Horizontal bars indicate the time intervals over which concentrations have been determined. The observed small-scale variations are probably indicative of eddies occurring in the atmosphere. Figure 16. Maximum NO 2 concentrations over minimum NO 2 concentrations measured in a time span between 25 and 37.5 min. Vertical bars indicate ±1s values for these ratios. Where no error bars are visible, they are smaller than the symbols plotted. The concentration measurements themselves are averages over either 2.5 or 3.75 min, and are for different altitude ranges, but all inside the boundary layer. These plotted ratios indicate short time-scale variations of up to a factor 2. The average of this ratio over the campaign is 1.5 ± 0.2. those measured for azimuth direction 214 E of N. The concentration measurements are for different altitude ranges in the boundary layer. The differences between the results for these azimuth directions may be several tens of percents. We see no apparent systematic behavior in the sense that concentrations are always higher or lower in one particular direction. Because the measurements are averages over 20 minutes, it is unlikely that they are caused by the same small-scale eddies as mentioned in section 3.3. Possibly, these azimuth variations are caused by differences in local sources for different azimuth directions. [58] During the DANDELIONS campaign the Heidelberg MAX-DOAS telescopes observed the atmosphere under three different azimuth angles. Although these observations are of a different type it is interesting to note that also these observations indicated large horizontal gradients of the NO 2 concentration field [Celarier et al., 2008; Brinksma et al., 2008]. 4. Comparison of NO 2 Lidar Data With in Situ Data [59] To our knowledge there are no well-validated instruments that provide data equivalent to the lidar NO 2 profile data for a direct comparison. Therefore, we compared lidar measurements with point measurements by NO 2 in situ monitors located at two altitudes, at ground level (3.5 m) and on the tower at 200 m. [60] We plotted in Figure 7 in situ NO 2 monitor data (red circles) together with the corresponding lidar data. From 9 September on, the NO 2 monitor located at 200 m was operational. The ground-level NO 2 monitor only became fully operational on 13 September, but some useful data was obtained on the 12th as well. We averaged the NO 2 -monitor minute values over the time interval that the lidar integrated. 13 of 18

14 Figure 17. Map of the measurement site at Cabauw. The different azimuth directions, 289 E ofn (default direction) and 214 E of N, are indicated with red lines. We also indicated the distances of 300 m and 2.5 km away from the measurement site. The tower of Cabauw with the in situ monitors at 3.5 m and 200 m altitude is indicated by a blue dot. The agreement between the NO 2 monitor data points and the lidar profile data points is good early in the morning during SCIAMACHY overpasses. For example, on 12 September, excellent agreement between the lidarderived NO 2 concentrations and the in situ data is observed around the SCIAMACHY overpass at 0959 UTC. Both the NO 2 monitor and the NO 2 lidar measured a high value of around 50 mg NO 2 m 3 at surface level. The value of the NO 2 monitor placed at 200 m rose from around 10 to 50 mg NO 2 m 3 during the 50 min lidar integration time. This is due to the boundary layer rising from below 200 m to higher altitudes during this period (Figure 9). We determined two averages for the NO 2 monitor data at 200 m; the Figure 18. Lidar NO 2 measurements averaged over 20 min taken in different azimuth directions, 289 E of N and 214 E of N, and at one elevation angle, 12. Horizontal bars indicate ±1s values for the concentration. Where no error bars are visible, they are smaller than the symbols plotted. Vertical bars indicate the height intervals over which concentrations have been determined. In particular, at lower altitudes the concentration differences for different azimuth directions is very large, about 20 mg NO 2 m 3. The observed variations are probably caused by differences in local sources for different azimuth directions. Figure 19. NO 2 concentrations measured in an azimuth direction of 289 E of N divided by NO 2 concentrations measured in an azimuth direction of 214 E of N, i.e., differences in NO 2 concentrations in air volumes separated by approximately 2 km. Measurements are averages over 20 min. The concentration measurements are for different altitude ranges in the boundary layer. Vertical bars indicate ±1s values for the concentration ratios. Where no error bars are visible, they are smaller than the symbols plotted. 14 of 18