Ozone columns obtained by ground-based remote sensing in Kiev for Aura Ozone Measuring Instrument validation



[1] Ground-based observations with a Fourier transform spectrometer in the infrared region (FTIR) were performed in Kiev (Ukraine) during the time frames August–October 2005 and June–October 2006 within the Ozone Monitoring Instrument (OMI) validation project 2907 entitled “OMI validation by ground based remote sensing: ozone columns and profiles” in the frame of the international European Space Agency/Netherlands Agency for Aerospace Programmes/Royal Dutch Meteorological Institute OMI Announcement of Opportunity effort. Ozone column data for 2005 were obtained by modeling the ozone spectral band at 9.6 μm with the radiative transfer code MODTRAN3.5. Our total ozone column values were found to be lower than OMI Differential Optical Absorption Spectroscopy (DOAS) total ozone column data by 8–10 Dobson units (DU, 1 DU = 0.001 atm cm) on average, while our observations have a relatively small standard error of about 2 DU. Improved modeling of the ozone spectral band, now based on HITRAN-2004 spectral data as calculated by us, moves our results toward better agreement with the OMI DOAS total ozone column data. The observations made during 2006 with a modernized FTIR spectrometer and higher signal-to-noise ratio were simulated by the MODTRAN4 model computations. For ozone column estimates the Aqua Atmospheric Infrared Sounder satellite water vapor and temperature profiles were combined with the Aura Microwave Limb Sounder stratospheric ozone profiles and Tropospheric Emission Monitoring Internet Service-Koninklijk Nederlands Meteorologisch Instituut climatological profiles to create a priori input files for spectral modeling. The MODTRAN4 estimates of ozone columns from the 2006 observations compare rather well with the OMI total ozone column data: standard errors are of 1.11 DU and 0.68 DU, standard deviation are of 8.77 DU and 5.37 DU for OMI DOAS and OMI Total Ozone Mapping Spectrometer, respectively.

1. Introduction

[2] It is common knowledge that the stratospheric ozone layer is very important for sustaining life on Earth. The ozone layer protects life on Earth from the harmful and damaging ultraviolet solar radiation. Ozone in the lower atmosphere, or troposphere, is also an important greenhouse gas. Ozone is not emitted directly by any natural source. However, tropospheric ozone is formed under high ultraviolet flux conditions from natural and anthropogenic emissions of nitrogen oxides (NOx), volatile organic compounds (VOCs), methane and carbon monoxide. The assessment of the ozone warming potential is a difficult task since both temporal and spatial distributions should be taken into account. Marenco et al. [1994] computed a relative radiative forcing for ozone at least 1200 times higher than for carbon dioxide (CO2), much higher than the other greenhouse gases except for chlorofluorocarbons (CFCs) which range up to 5800 times the radiative forcing of CO2. However, the emissions of CFCs are expected to decrease to near zero in the next decades following international agreements on phase-out measures. In this way, ozone becomes (most probably after 50 years) the most important contributor in the greenhouse effect after carbon dioxide and methane. With its current concentrations, ozone would contribute 22% of the global warming in the Northern Hemisphere and 13% in the Southern Hemisphere.

[3] Remote sensing is used to quantify and understand key processes in the global ozone budgets. At the present-day, satellite remote sensing observations are widely available for total ozone column and ozone atmospheric profiles. Nevertheless, ground-based monitoring is needed to validate and complement space-based measurements and to clarify local/regional specific sources and sinks of this important gas. Space and ground-based observations in synergy enable studies of the dynamical behavior of atmospheric trace gases and air pollution and to check compliance to the pollution transport models. These observations will also serve the development of environmental policies, with greenhouse gas regulation policies in particular, on a local and regional scale. Our first attempts to retrieve total ozone columns from the observations by our Fourier transform spectrometer in the infrared region (FTIR) instrument from the Main Astronomical Observatory of the National Academy of Sciences of Ukraine was successful [Shavrina and Veles, 2004]. This success allowed to us to submit our proposal 2907 entitled “OMI validation by ground based remote sensing ozone columns and profiles” in response to the International Ozone Monitoring Instrument (OMI) Announcement of Opportunity call, where it was accepted.

2. FTIR Spectrometer for Atmospheric Monitoring

[4] Our work is performed with a Fourier transform infrared (FTIR) spectrometer, model “Inralum FT 801,” which was modernized for the task of monitoring the atmosphere by direct sun observations. The “double cat eye” interferometer scheme [Egevskaya et al., 2001] is shown in Figure 1. The main advantage of this device is its small size and small sensitivity of the optical arrangement to vibrations. The working spectral range of the FTIR spectrometer is 2–12 μm (800–5000 cm−1) with the highest possible spectral resolution of about 1.0 cm−1. The accuracy of the wavelength calibration in the FTIR spectra is determined by the stability of the wavelengths of a helium-neon laser used as the reference source of light in the reference channel. The corresponding accuracy is 0.001 cm−1. The exposure time needed to obtain one experimental spectrum is roughly 3 s. When obtaining the 2005 data set, we averaged 6–8 spectra obtained within 1.5–2 min to improve signal-to-noise ratio (S/N) to exceed 100. Following the modernization in 2006 of our Fourier spectrometer and the software for the initial treatment of the registered spectra, the system now allows us to average 2–99 individual spectra during the observation period. We averaged 4 single spectra as was recommended by the developers of the spectrometer device to avoid a degradation of the averaged spectrum due to the recording of atmospheric instabilities at longer exposure times. Our averaged spectra have S/N of 150–200. We registered 3–4 averaged spectra during 2–3 min of recording time. Prior to further treatment of observed spectra we checked the repeatability of these 3–4 spectra and choose the spectrum with the best S/N to be fitted with the model spectra ([Egevskaya et al., 2001]).

Figure 1.

Structural and optical scheme of the FTIR spectrometer: 1, fast lens; 2, 4, ellipsoids for shift of image; 3, interferometer; 5, detector; 6, electronics block.

[5] Direct sun observations of solar radiation transmitted through the Earth's atmosphere were performed with the FTIR spectrometer during the time periods August–October 2005 and June–November 2006 at the Main Astronomical Observatory (MAO), National Academy of Sciences of Ukraine (NASU). The geographical coordinates of the Observatory are 50°21′50.90″N and 30°29′51.34″E, located at 186 m above sea level.

[6] In 2005, the spectral resolution of the setup was 4.0 cm−1. Following a major system update, the FTIR spectrometer was operated with an improved spectral resolution of 2.0 cm−1 and a higher S/N ratio of 130–200 during the observational period of June–November 2006. On sunny days with clear skies, observations were made during daylight over a wide range of solar zenith angles (from 26° up to 84°). Starting from July 2006, the Ozone Analyzer Thermo Environmental, Model 49 UV, for performing “surface” ozone measurements was operating at the same location as the FTIR spectrometer, namely, on the roof of the observatory building. An example of the observed spectra on 27 June 2006, when measurements were performed from 0847 LT up to 1905 LT (z = 30.26–72.80°, resolution of 2.0 cm−1) is shown in Figure 2.

Figure 2.

A selection of FTIR spectra recorded on the 27 June 2006. Observations started at 0847 LT and ended at 1905 LT (z = 30.26–72.80°, resolution R = 2.0 cm−1). Each spectrum takes roughly 3 s to record. The ozone band is located at 9.6 μm or 1042 cm−1.

3. MODTRAN Spectral Modeling

[7] The vertical column amounts of ozone molecules were retrieved with the help of the radiative transfer code MODTRAN3.5 [Abreu and Anderson, 1996] from the observations of 2005. MODTRAN4.3 (MODTRAN4 Version 3 Revision 1 [Berk et al., 1999]) was provided to us in the autumn 2006 and was used for modeling of the observations of 2006. Both codes use a moderate spectral resolution model of optical transmission, described by Abreu and Anderson [1996] and Bernstein et al. [1996]. MODTRAN4.3 has new options for modeling atmospheric radiation transport and other improvements [Berk et al., 1999]. MODTRAN3.5 band model was based on HITRAN-96 molecular data [Rothman et al., 1998]; MODTRAN4.3 uses HITRAN-2001 (http://www.vs.afrl.af.mil/ProductLines/IR-Clutter/modtran4.aspx). The codes are widely applied to the interpretation of ground based, airborne and space-borne remote observations of the Earth's atmospheric spectra. These programs calculate the atmospheric transmission and irradiation at electromagnetic radiation frequencies from 0 up to 50,000 cm−1. The modeling program uses the spherical source function for solar and moon light scattering and the standard and user defined model atmospheres. The users specify the a priori atmospheric profiles of trace gases, aerosols, clouds, fog and even rain. To model the shape of molecular absorption bands, both MODTRAN programs use the so-called “smeared lines” approximation, namely the band model approximation [Abreu and Anderson, 1996; Bernstein et al., 1996], instead of line-by-line calculations. The band model is calculated once as molecular absorption in 1 cm−1 spectral bins for six temperatures and standard pressure. The computed table is further applied to calculate the shape of molecular bands in model spectra using temperature interpolation and pressure correction. The calculation of the band model is based on a large array of previously accumulated data of spectral lines in the HITRAN database for 12 molecules (H2O, CO2, O3, N2O, CO, CH4, O2, NO, SO2, NO2, NH4, and HNO3). In addition, the programs account for optical spectral absorption by heavy molecules: CFC (9 molecules) and ClONO2, HNO4, CCl4, and N2O5). The calculations are performed in local thermodynamic equilibrium (LTE) approximation for the moderate spectral resolution (2.0 cm−1) which fortunately corresponds to the resolution of our observed FTIR spectra.

[8] For the new set of spectral data recorded during 2006, the model spectra were calculated with MODTRAN4 code [Berk et al., 1999] and new band model input file which was prepared by us based on the HITRAN-2004 database of Rothman et al. [2005], the most recent and more complete molecular line database. The band model parameters, namely, average absorption coefficient S/d and line density parameter 1/d were calculated as defined by Bernstein et al. [1996]

equation image

where Sj is the temperature-dependent line strength, Δω is the width of the spectral interval (Δω = 1 cm−1 in this study), and the summation includes only lines within the interval.

[9] In Figure 3 we show the spectra of the ozone band at 9.6 μm calculated with MODTRAN3 and FASCODE (line-by-line modeling) from Bernstein et al. [1996]. (The code FASCODE is not available for us, our own line-by-line code is now under development.) We estimate the value of the sample variance S of these two spectra fit in the range of 981–1078 cm−1 as S = Σ(FobsFcomp)2/N = 6.9 × 10−4, where Fobs, Fcomp are the observed and computed residual intensities, N is the number of considered wavelengths. In the following, we use this value of S as our coincidence criterion to choose the best fits of our model spectra to the observed ozone bands in the minimization procedure.

Figure 3.

Comparison of MODTRAN (band model) and FASCODE (line-by-line) spectral radiance calculation for ozone 9.6 μm band [from Bernstein et al., 1996]. We estimate the value of the sample variance of these two spectra fit in the range of 981–1078 cm−1 as 6.9 × 10−4. Please, consult the text of section 3 for further explanation.

4. Input Profiles for Modeling

[10] Measurements of surface ozone concentrations by the collocated ozonometer together with data from the Atmospheric Infrared Sounder instrument aboard NASA EOS-Aqua (AIRS, http://disc.gsfc.nasa.gov/AIRS/) and the Microwave Limb Sounder (MLS) aboard EOS Aura (http://avdc.gsfc.nasa.gov/Data/Aura/) were used for construction of atmospheric ozone, temperature and water vapor input profiles for the MODTRAN4.3 code. We used the level 3 daily data products of AIRS to obtain water vapor profiles, temperature profiles and geopotential height estimates at 24 standard pressure levels over the location of MAO in Kiev. The level 3 files contain geophysical parameters averaged and binned into 1° × 1° grid cells. Grid maps correspond to −180.0° to +180.0° of longitude starting from zero meridional (Greenwich) to the west and the east correspondingly and are separated into the ascending (A) and descending (D) portion of the satellite orbit, where “ascending or descending” refers to the directions of the subsatellite points on the satellite tracks at the equatorial crossing. The level 3 daily files are provided in the Hierarchical Data Format (HDF)-EOS format. To extract data from the HDF files for the location of MAO in Kiev (latitude 50°21′9″N, longitude 30°30′E), the software package MathLab 6.5 was used.

[11] AIRS vertical profiles of water vapor at 12 atmospheric layers, labeled H2OVapMMRA and H2OVapMMRD (ascending and descending, respectively) are reported as water vapor mass mixing ratio (gm/kg dry air). Temperature profiles are reported in Kelvin at 24 standard pressure levels, labeled temperatureA and temperatureD. Geopotential heights are reported in meters at 24 standard pressure levels, labeled GPHeightA and GPHeightD. These data were used to construct the MODTRAN temperature and water vapor input profiles for each observing day. MLS data over the location of the MAO observatory in Kiev were downloaded as ASCII files from the “station overpass” section from the Aura Validation Data Center (AVDC) (http://avdc.gsfc.nasa.gov/Data/Aura/). Station overpass data files contain satellite data observations spatially collocated with the ground station coordinates ±5° in latitude and ±8° in longitude. The downloaded MLS data product was ozone profiles (MLS-L2GP-03) of version V1.5. The data for Kiev are of 6–8 profiles typically per day and were averaged for each day of the observations. Prior to using MLS data we downloaded the “Version 1.5 level 2 data quality and description document” (v1-5-report-29jul05b.pdf) from the MLS Web site. The MSL ozone profiles were used over the range 215–0.46 hPa were the L2gpPrecision field was taken into account as a weighing factor. We also interpolated these data for the exact coordinates of the point of observation (MAO NASU). These profiles differ from averaged very little. We preferred the averaged ozone profiles due to the rather low precision of MLS v1.5 data. From these data together with the Tropospheric Emission Monitoring Internet Service (TEMIS) climatological ozone profiles (http://www.temis.nl/data/fortuin.html), the ozone input files for the MODTRAN modeling were constructed. AIRS water vapor (H20) data are given for the range 1100–150 hPa, and AIRS temperature (T) data are given for the range 1100–1.5 hPa. These data sets were expanded by MODTRAN standard summer water vapor and temperature profile data for the midlatitude standard model atmosphere.

5. Software Developments

[12] We developed in-house a sophisticated procedure to analyze the observed ozone spectral band at 9.6 μm. To obtain the best fit to the observed spectral band, we adhered to the following procedure:

[13] 1. We use the standard summer or the averaged summer-winter MODTRAN model atmosphere as a reference model to compute a grid of Earth model atmospheres with different atmospheric columns and profiles of ozone. The construction of ozone input profiles, AIRS water vapor and CO2 profiles are iterated in the case of a poor fit of model spectra to the observed spectrum (at the end of the process) by a procedure in which these profiles are determined by two parameters: scaling (or fitting) factor and location (altitude) of the maximum of the molecular densities. Additionally, for ozone we adopt two parameters determining the curvatures of upper and lower parts of the profile corresponding to upper and lower atmosphere layers near the ozone density maximum.

[14] 2. Using the MODTRAN4 code, we compute a grid of the theoretical spectra associated with the grid of the model atmospheres.

[15] 3. To determine the best fit parameters, we compare the observed and computed spectra with a two-step optimization procedure. First, we determined the best fit to observed water vapor and CO2 lines in the region 800–1240 cm−1; that is, we exclude the ozone spectral band from the analysis. Second, we fit the observed spectrum with the grid of calculated ozone spectral band profiles using water vapor and CO2 atmospheric profiles selected on the previous step. We choose the best fit in accordance with a minimal value of the sample variance.

6. Ozone Column Data Analysis

[16] Ozone columns for various dates in 2005 were obtained using MODTRAN3.5 [Abreu and Anderson, 1996] modeling of the ozone spectral band shape at 9.6 μm with the original band model based on the HITRAN-96 database [Rothman et al., 1998]. As mentioned above, MODTRAN4.3 version was available to us in autumn 2006 only, band model based on HITRAN-2004 was prepared by us nearly to the same time. We compared our ozone columns with spatially and temporally collocated Aura OMI total ozone column data downloaded from the Aura Validation Data Center, OMDOAO3 level 2 data and OMTO3 data. The OMDOAO3 data product contains total ozone and ancillary information produced by the OMI Differential Optical Absorption Spectroscopy (DOAS) algorithm. The OMI DOAS algorithm is described in the OMI Algorithm Theoretical Basis Document (http://avdc.gsfc.nasa.gov/Data/Aura/OMI/OMDOAO3/ieee-ozone-doas_20050427.pdf). OMTO3 contains total ozone, aerosol index, and ancillary information produced from the Total Ozone Mapping Spectrometer (TOMS) Version 8 (V8) algorithm applied to OMI global mode measurements. (http://avdc.gsfc.nasa.gov/Data/Aura/OMI/OMTO3/OMTO3ReleaseDetails.html). In Table 1 we present the statistical results of our comparisons. Our total ozone values are found to be somewhat lower than the OMI-DOAS data by 8–10 Dobson units (DU, 1 DU = 0.001 atm cm) on average with a standard deviation of 10.5 DU (about 3.5%) (see Table 1 and Figure 4). Our total ozone values have a relatively small standard error ∼2 DU. Incorporating our new band model based on the HITRAN-2004 database [Rothman et al., 2005] as calculated by us improves the agreement of our ozone column data with OMI DOAS data to near zero difference on average with a similar standard deviation. Our total ozone values are lower than the OMI TOMS data by 3 DU on average with a standard deviation of 13.4 DU (about 4%) and standard error ∼2.5 DU.

Figure 4.

(top) Time series of OMI total ozone (OMI TOMS and OMI DOAS) and ground-based FTIR total ozone data for 2005. (bottom) Time series of difference between satellite and ground-based ozone columns. Average difference of satellite minus ground-based amounts to 8.45 DU and 3.19 DU for OMI DOAS and OMI TOMS respectively, with a 10.50 DU and 13.41 DU standard deviation (1.98 DU and 2.53 DU standard errors).

Table 1. Descriptive Statistics and Regressive Analysis of the Ground-Based FTIR Total Ozone Column Versus the OMI DOAS and OMI TOMS Total Ozone Column Data for 2005 and 2006
 Descriptive Statistics
Standard Error1.98/2.531.11/0.68
Standard Deviation10.50/13.418.77/5.37
Sample Variance110.19/179.8476.96/28.89
 Regressive Analysis
FTIR versus DOAS/FTIR versus TOMSFTIR versus DOAS/FTIR versus TOMS

[17] FTIR observations during the time frame July 2006 to October 2006 with the modernized FTIR spectrometer provided observations with a much better signal-to-noise ratio. The recorded spectra were simulated using MODTRAN4.3 and our new band model based on the HITRAN-2004. The AIRS's water vapor and temperature profiles, Aura MLS stratospheric ozone profiles and TEMIS climatological ozone profiles [Fortuin and Kelder, 1998] were used for the creation of the a priori input files. The MODTRAN4.3 ozone columns in 2006 fit rather well with OMI TOMS and OMI DOAS data, see Table 1 and Figure 5. As coincidence criteria we consider the values of standard errors and standard deviation (Table 1). The standard errors are of 1.11 DU and 0.68 DU with a standard deviation of 8.77 DU (2.7%) and 5.37 DU (1.8%), for OMI DOAS and OMI TOMS, respectively. The number of collocated observations of FTIR and satellite was 28 and 62 in 2005 and 2006, respectively. We only found significant differences between ground-based and satellite observations under conditions of insufficiently clear skies (clouds).

Figure 5.

(top) Time series of the OMI total ozone and the ground based FTIR total ozone data of 2006 for Kiev (MAO). (bottom) Time series of difference of satellite minus ground-based ozone columns. Average difference of satellite minus ground-based amounts to 0.37 DU and −0.25 DU for OMI DOAS and OMI TOMS, respectively, with a 8.77 DU and 5.37 DU standard deviation (1.11 DU and 0.68 DU standard errors).

[18] The total ozone column estimates were in general obtained from the FTIR spectra observed at lower zenith angles of the Sun, 1100 to 1400 LT. Their values are mainly determined by stratospheric ozone profiles, namely by their maximum heights and widths near the concentration maximum. Their dependence on tropospheric ozone variations is rather weak.

7. Conclusion

[19] In this paper we have presented our extensive data record of FTIR direct sun observations and the ozone columns retrieved from these spectra, covering two periods in the time frame 2005–2006. Our estimates of the total ozone columns in the 2006 time frame obtained with the MODTRAN4 modeling agree well with the OMI total ozone column data. We noted some significant differences for the data reordered under the insufficiently clear skies (clouds) which exemplifies the care one must to take to perform accurate FTIR measurements of the total ozone column.

[20] The algorithm for automatic modification of the atmospheric ozone input profiles, particularly tropospheric profiles requires further development. Our calculations with the band model of MODTRAN4.3 must be tested through modeling with the line-by-line radiative transfer code (like FASCODE). We plan to create such code for the line-by-line calculations with the HITRAN-2004 molecular line lists and the available model atmospheres.


[21] This work was performed in the framework of the International ESA/KNMI/NIVR OMI “Announcement of Opportunity for Calibration and Validation of the Ozone Monitoring Instrument,” providing early access to provisional OMI data sets and guidance to public OMI data. The Dutch-Finnish built OMI instrument is part of the NASA EOS Aura satellite payload. The OMI project is managed by NIVR and KNMI in the Netherlands. Provisional OMI data were obtained from the NASA Goddard Aura Validation Data Center (AVDC). Public OMI data were obtained from the NASA Goddard Earth Sciences (GES) Data and Information Services Center (DISC), home of the GES Distributed Active Archive Center (DAAC). We thank the AIRS team, the Aura team, the MLS team, and the TEMIS KNMI team for their data. The work of the authors from MAO NASU was supported by the grant of STCU.