CAI Zhao-Nan, LIU Yi, and LIU Xiong

1Key Laboratory of Middle Atmosphere and Global Environment Observation, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China

2Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, USA

Sensitivity Studies for Monitoring Tropospheric Ozone from Space Using the Ultraviolet, Visible, and Polarization Bands

CAI Zhao-Nan1, LIU Yi1, and LIU Xiong2

1Key Laboratory of Middle Atmosphere and Global Environment Observation, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China

2Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, USA

The authors analyzed the retrieval sensitivity of tropospheric ozone using simulated the Global Ozone Monitoring Experiment-2 (GOME-2) measurements. The retrieval sensitivity was evaluated by the degree of freedom for signal (DFS). The combination of the ultraviolet (UV), UV polarization (UVPOL), and visible (VIS) bands enhances DFS of tropospheric ozone and improves the vertical resolution of the retrieved ozone profile. UVPOL reduces the dependence on solar zenith angle, mainly increases the sensitivity in upper troposphere. Polarization increased the DFS by 20% on the eastern side of the GOME-2 orbit, with little improvement on the western side because the increase in DFS due to polarization is dependent on the relative azimuth angle. The inclusion of the visible band reduces significantly the dependence on viewing geometry, and mainly increases the DFS in the lower troposphere (0?6 km) by a factor of two. It was possible to retrieve several independent pieces of tropospheric ozone information from GOME-2 UV/UVPOL/VIS measurements, especially in the lower troposphere.

tropospheric ozone, retrieval sensitivity, GOME-2

1 Introduction

Ozone is an important trace gas in the troposphere. Tropospheric ozone plays important roles: as an air pollutant, greenhouse gas, and the primary source of hydroxyl radicals. The ozone profile from the troposphere to stratosphere can be retrieved from backscattered ultraviolet (UV) measurements due to wavelength-dependent ozone absorption.

A space-based spectrometer with an adequate signal-to-noise ratio (SNR) and sufficient spectral resolution offers the possibility to monitor tropospheric ozone using ozone absorption in the Huggins band (Chance et al., 1997). Observations from space can provide ozone information globally. After careful wavelength and radiometric calibration and extensive improvement to the forwardmodel, Liu et al. (2005, 2010) demonstrated that valuable tropospheric ozone information can be retrieved from the global ozone monitoring experiment (GOME) and ozone monitoring instrument (OMI) UV measurements. Whatever retrieval method is used, the retrieval sensitivity to tropospheric ozone is limited for using UV measurements only from GOME, OMI, and GOME-2 (Liu et al., 2005, 2010; Cai et al., 2012).

The solar radiance reflected by the atmosphere is generally polarized. The strong absorption of ozone in the Huggins band reduces the possibility of photons' being multiply scattered in the lower atmosphere. The sensitivity of polarization to tropospheric ozone reaches a maximum value in the troposphere. In addition to UV radiance, the inclusion of polarization measurements (UVPOL) can increase tropospheric ozone sensitivity in the free and upper troposphere (Hasekamp and Landgraf, 2002). Natraj et al. (2011) proposed to improve the vertical sensitivity in the retrieval of tropospheric ozone information using a multi-spectra approach from geostationary orbit and suggested that the combination of visible (VIS) and UV measurements significantly enhances the sensitivity for low-level (0?2 km) ozone. A limited number of atmospheric were investigated, and the dependence of sensitivity on observation geometry was not included. Cuesta et al. (2013) improved the information obtained for lowermost ozone (0?3 km) by combining the thermal infrared (TIR) measurements of the Infrared Atmospheric Sounding Interferometer (IASI) and UV measurements of GOME-2.

Four main GOME-2 channels measure backscattered UV/VIS radiance from 270 to 790 nm with a spectral resolution (full width at half maximum, FWHM) of 0.24?0.53 nm. This covers the Hartley and Huggins bands in the UV and the Chappuis band in the VIS. Moreover, polarization-sensitive instruments such as GOME-2 require a polarization correction of radiance. The polarization measurement devices (PMD) of GOME-2 measure light polarized parallel and perpendicular to the reference plane for four wavelengths (312.7, 318.0, 325.3, and 332.6 nm) in the Huggins ozone absorption band. Thus the GOME-2 instrument can provide UV, VIS, and UVPOL measurements in the same field of view (FOV).

In this study, we focused on the retrieval sensitivity for the combination of UV, VIS, and UVPOL measurementsfrom simulated GOME-2 measurements. The simulated spectra were generated by a forward model based on the GOME-2 instrument configuration and sunsynchronous low earth orbit (LEO) geometries. The sensitivity dependencies on viewing geometry are discussed.

2 Forward model and analysis tools

2.1 Radiative transfer model and inputs

The full Stokes vectorI,Q(0°/90° polarization),U(± 45° polarization),V(circular polarization), and their analytic weighting functions with respect to atmospheric or surface parameters were calculated using the Vector LInearized Discrete Ordinate Radiative Transfer model (VLIDORT) version 2.4, which is a fully linearized multiple scattering multi-layer radiative transfer model (Spurr, 2006).

VLIDORT was used to simulate the upwelling radiance on the top of the atmosphere and provide weighting functions with respect to the ozone profile, surface albedo, total water vapor column, and NO2column. Following the definition of GOME-2, the polarization of light is described by stocks fractionq=Q/Iand the weighting functions ofqare then calculated from the chain rules.

The improved high-resolution (0.04 nm FWHM) solar reference spectrum was taken from (Chance and Kurucz, 2010). A priori ozone profiles and their standard deviations were taken from climatology data (McPeters et al., 2007). For the first guess, we considered a pure Rayleigh scattering atmosphere without scattering of aerosol and cloud.

2.2 Instrument parameters

Table 1 summarizes the spectral configurations used in the sensitivity study. For the UV band, according to the ozone profile retrieval (Cai et al., 2012), two sub-bands (290?307 nm and 324?340 nm) with a high SNR and good quality radiometric calibration in were chosen. The SNR was estimated from GOME-2 measurements with an independent SNR calculator (refer to Cai et al. (2012) for more detail). As an approximation, we assumed Gaussian slit functions as the instrument line shapes (ILS) for UV and VIS bands and assumed tophat Gaussian slit functions for the UVPOL band. The simulated spectra were generated as follows. First, radiance and the associated weighting functions were calculated using a forward model in a finer wavelength grid and were then convolved to the instrument resolution. Finally, the oversampled spectra/weighting functions were interpolated to the instrument coarse wavelength grid. Solar zenith angle (SZA), viewingzenith angle (VZA), and relative azimuth angle (RAA) were taken from real GOME-2 orbit.

Table 1 Spectral configuration for the ultraviolet (UV), UV polarization (UVPOL), and visible (VIS) bands.

2.3 Analysis tools

The inverse model is based on the optimal estimation technique (Rodger, 2000). Inverse problems are usually non-linear and ill posed. An optimal solution is the average of the a priori informationxaand measurementsxweighted by a priori error and measurement noise, respectively. Then, the retrieved state ?xcan be written as:

whereAis the averaging kernel matrix, andIis the identity matrix.

The averaging kernelAis given by

whereSais the a priori covariance matrix,Syis the measurement error covariance matrix,Kis the weighting function matrix of the forward model with respect to the state vector, andGis the contribution function matrix. The retrieval statexacontains ozone columns in 24 layers, first order surface albedo for each fitting window, total water vapor column, and total O4column. The a priori covariance matrix was constructed assuming a correlation length of 6 km.

A row ofAat a given layer indicates the sensitivity of retrieval at that layer to the variations at all layers. The trace ofAis the total degree of freedom for signal (DFS), which describes the number of independent pieces of information from measurements.

For the purpose of sensitivity analysis, the linear estimate is a good and fast approximation of the non-linear retrieval. It allows rapid assessment of the retrieval characteristics.

3 Results

We examined the DFS of the tropospheric ozone column (0?12 km, noted as DFS12) and the ozone column in the lower troposphere (0?6 km, noted as DFS06) for four scenarios: UV only, UV+UVPOL, UV+VIS, and UV+UVPOL+VIS (Fig. 1). For the UV-only retrievals, the tropospheric ozone sensitivity varies greatly with SZA, but showes little dependence on VZA and RAA. The inclusion of polarization information in addition to UV radiance information can increase the tropospheric ozone sensitivity. Figures 1a and 1d show that the inclusion of UVPOL, compared with UV-only retrievals, reduces the dependence on SZA and significantly increases DFS12 for higher SZA, but not too much in the low-level troposphere. It should be noted that for GOME-2 there are only five broad bands of polarization measurement; the DFS can be improved further by including polarization with much higher spectral resolution. The inclusion of UVPOL shows a dependence on RAA, mainly because the degree of polarization is generally dominated by the single scattering of molecules, i.e., the scattering angle. For the example presented in Fig. 1c, the scattering angle was 90° for RAA = 0°, where the degree of polarization was largest. The UVPOL had a small dependence on VZA.

Figure 1 The degree of freedom for signal (DFS) of ((a), (b), (c)) the tropospheric ozone column and ((d), (e), (f)) the lower tropospheric ozone column as a function of solar zenith angle (SZA), viewing zenith angle (VZA) and relative azimuth angle (RAA) for four scenarios: UV, UV+UVPOL, UV+VIS, and UV+UVPOL+VIS.

The inclusion of VIS reduces significantly the dependence on SZA and increases significantly sensitivity in the low-level troposphere. VIS increases the DFS12 by 50% and increases the ozone information in the lower troposphere (DFS06) by a factor of ～ two.

Figure 2 shows the retrieval averaging kernels at a large SZA for each layer. UV-only retrieval loses sensitivity in the low-level troposphere. UV+UVPOL improves the DFS in the free and upper troposphere but had little effect in the lowermost troposphere because the increasing possibility of multiple scattering decreases the degree of polarization (DOP) in that region and thus reduce the information contained. UV+VIS increases significantly the sensitivity of ozone in the boundary layer even for large SZA. Because photons at longer wavelength can penetrate to the near surface; thus, the backscattered VIS measurement containes ozone information in the boundary layer. The inclusion of the UVPOL and VIS bands gives a narrower FWHM of the averaging kernels, i.e., it improves the vertical resolution of the retrieved profile. In the analysis, we assumed a first order wavelength-dependent surface albedo and retrieved the albedo offset and slope to characterize the albedo. These parameters were correlated with tropospheric ozone and reduced the sensitivity in the lower troposphere, especially in the VIS band.

GOME-2 covers the UV, UVPOL, and VIS bands in one instrument. These bands are consistent in their viewing geometry, FOV. It has been shown that the improvement in retrieval sensitivity to tropospheric ozone by using a multi-spectra approach depends on observation geometries (Fig. 1). To further investigate the feasibility of using GOME-2 multi-band measurements in principle, we calculated DFS12 for GOME-2 pixels with SZA, VZA, and RAA taken from one orbit of GOME-2 (orbit number 9265). Figure 3a shows the difference in DFS12 between UV+UVPOL and UV-only retrievals. For SZA < 65°, DFS12 increases by ～ 20% for the pixels of the easternside of the orbit and by 5% in the western pixels, with almost no increase for the nadir pixels. This is mainly due to the change in relative azimuth angle between solar and satellite angles. GOME-2 has a morning orbit, with a local crossing time of 9:30 a.m. The pixels of the eastern side have the largest DOP, whereas on the other side, the DOP is much smaller due to the geometry are close to being fully backscattering. For a SZA > 65°, there is a significant improvement in the DFS12.

Figure 2 Averaging kernel for high SZA = 70° and off-nadir viewing direction VZA = 25° using, and RAA = 90°: (a) UV, (b) UV+UVPOL, and (c) UV+VIS. The averaging kernelAvalues have been normalized by the a priori error. Different colors indicate the altitude of the averaging kernels.

Figure 3 The increase in DFS12 compared with UV-only retrievals by using (a) UV+UVPOL, (b) UV+VIS, and (c) UV+UVPOL+VIS for one the Global Ozone Monitoring Experiment-2 (GOME-2) orbit on 1 August 2008.

Figure 3b shows the difference in DFS12 between UV+VIS and UV-only retrievals. The increase in the DFS12 is generally around 40%, and is larger than 60% for higher SZAs. The inclusion of the VIS band does not show much cross-track dependence.

4 Conclusions

Tools for GOME-2 measurement simulation and linear error analysis for tropospheric ozone retrieval have been developed. The differences in DFS between UV, UV+UVPOL, UV+VIS, and UV+VIS+UVPOL were compared. The DFS analysis revealed that the inclusion of polarization and visible measurements in addition to the UV band improves the retrieval sensitivity to tropospheric ozone and the vertical resolution of the retrieved ozone profile. The inclusion of GOME-2 polarization measurements increases the ozone information in the middle and upper troposphere and decreases the dependence of DFS on SZA. For GOME-2, the increase in DFS12 by using UV and UVPOL is strongly cross-track dependent. The increase is as much as 20% for eastern-side scans and almost zero for nadir scans. The DFS for lower tropospheric ozone increases by a factor of two when the combination of UV and VIS bands is used. The VIS band significantly enhances the sensitivity in the boundary layer. Both UVPOL and VIS improves the DFS at higher SZAs.

Acknowledgements. This work was supported by National Natural Science Foundation of China (Grant Nos. 41205018 and 41375035).

Cai, Z., Y. Liu, X. Liu, et al., 2012: Characterization and correction of Global Ozone Monitoring Experiment 2 ultraviolet measurements and application to ozone profile retrievals,J. Geophys. Res., 117, D07305, doi:10.1029/2011JD017096.

Chance, K., and R. L. Kurucz, 2010: An improved high-resolution solar reference spectrum for earth's atmosphere measurements in the ultraviolet, visible, and near infrared,J. Quant. Spectrosc. Radiat. Transf., 111(9), 1289-1295.

Chance, K. V., J. P. Burrows, D. Perner, et al., 1997: Satellite measurements of atmospheric ozone profiles, including tropospheric ozone, from ultraviolet/visible measurements in the nadir geometry: A potential method to retrieve tropospheric ozone,J. Quant. Spectrosc. Radiat. Transf., 57(4), 467-476.

Cuesta, J., M. Eremenko, X. Liu, et al., 2013: Satellite observation of lowermost tropospheric ozone by multispectral synergism of IASI thermal infrared and GOME-2 ultraviolet measurements over Europe,Atmos. Chem. Phys., 13(19), 9675-9693.

Hasekamp, O. P., and J. Landgraf, 2002: Tropospheric ozone information from satellite-based polarization measurements,J.Geophys. Res., 107(D17), 4326, doi:10.1029/2001JD001346.

Liu, X., P. K. Bhartia, K. Chance, et al., 2010: Ozone profile retrievals from the Ozone Monitoring Instrument,Atmos. Chem. Phys., 10(5), 2521-2537.

Liu, X., K. Chance, C. E. Sioris, et al., 2005: Ozone profile and tropospheric ozone retrievals from the Global Ozone Monitoring Experiment: Algorithm description and validation,J. Geophys. Res., 110(D20), D20307, doi:10.1029/2005JD006240.

McPeters, R. D., G. J. Labow, and J. A. Logan, 2007: Ozone climatological profiles for satellite retrieval algorithms,J. Geophys. Res., 112(D5), D05308, doi:10.1029/2005JD006823.

Natraj, V., X. Liu, S. Kulawik, et al., 2011: Multi-spectral sensitivity studies for the retrieval of tropospheric and lowermost tropospheric ozone from simulated clear-sky GEO-CAPE measurements,Atmos. Environ., 45(39), 7151-7165.

Rodgers, C. D., 2000:Inverse Methods for Atmospheric Sounding: Theory and Practice, World Scientific, Singapore, 238pp.

Spurr, R. J. D., 2006: VLIDORT: A linearized pseudo-spherical vector discrete ordinate radiative transfer code for forward model and retrieval studies in multilayer multiple scattering media,J. Quant. Spectrosc. Radiat. Transf., 102(2), 316-342.

:Cai, Z.-N, Y. Liu, and X. Liu, 2014: Sensitivity studies for monitoring tropospheric ozone from space using the ultraviolet and visible and the polarization bands,Atmos. Oceanic Sci. Lett., 7, 198-202,

10. 3878/j.issn.1674-2834.13.0092.

Received 25 November 2013; revised 3 January 2014; accepted 3 January 2014; published 16 May 2014

CAI Zhao-Nan, caizhaonan@mail.iap.ac.cn