A new software suite for NO2 vertical profile retrieval from ground-based zenith-sky spectrometers

A new software suite for NO2 vertical profile retrieval from ground-based zenith-sky spectrometers

ARTICLE IN PRESS Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333 www.elsevier.com/locate/jqsrt A new software suite for ...

269KB Sizes 1 Downloads 59 Views

ARTICLE IN PRESS

Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333 www.elsevier.com/locate/jqsrt

A new software suite for NO2 vertical profile retrieval from ground-based zenith-sky spectrometers L. Denisa, H.K. Roscoea,, M.P. Chipperfieldb, M. Van Roozendaelc, F. Goutaild a

British Antarctic Survey/NERC, Madingley Road, Cambridge CB3 0ET, UK b Environment Centre, University of Leeds, Leeds LS2 9JT, UK c Belgian Institute for Space Aeronomy (BIRA/IASB), 1180 Brussels, Belgium d Service d’Ae´ronomie du CNRS, BP3, 91271 Verrie`res le Buisson, France Received 17 July 2003; accepted 23 July 2004

Abstract Here we present an operational method to improve accuracy and information content of ground-based measurements of stratospheric NO2. The motive is to improve the investigation of trends in NO2, and is important because the current trend in NO2 appears to contradict the trend in its source, suggesting that the stratospheric circulation has changed. To do so, a new software package for retrieving NO2 vertical profiles from slant columns measured by zenith-sky spectrometers has been created. It uses a Rodgers optimal linear inverse method coupled with a radiative transfer model for calculations of transfer functions between profiles and columns, and a chemical box model for taking into account the NO2 variations during twilight and during the day. Each model has parameters that vary according to season and location. Forerunners of each model have been previously validated. The scheme maps random errors in the measurements and systematic errors in the models and their parameters on to the retrieved profiles. Initialisation for models is derived from well-established climatologies. The software has been tested by comparing retrieved profiles to simultaneous balloon-borne profiles at mid-latitudes in spring. r 2004 Elsevier Ltd. All rights reserved. Keywords: Stratosphere; Inversion; Nitrogen oxides

Corresponding author. Tel.:+44-1223-221400; fax:+44-1223-362616.

E-mail address: [email protected] (H.K. Roscoe). 0022-4073/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.jqsrt.2004.07.030

ARTICLE IN PRESS 322

L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

1. Introduction Ground-based measurements of NO2 are currently used to derive vertical columns. Routine retrieval of vertical profiles of NO2 from the ground-based measurements will greatly increase their information content, which is important because NO2 is such an important sink of ozone in the stratosphere. Another motive for routine retrieval of vertical profiles is to enable a better analysis of trends in NO2. This is important because of the recent discovery of possible discrepancies in its trend relative to that of its source N2O, which might indicate long-term dynamical changes in the global stratosphere. N2O is emitted near the surface, and its trend in the troposphere is about 3% per decade [1]. The NO2 vertical column has increased by 3.4–6.8% per decade since 1981 [2] at the one site with such a long record (Lauder, New Zealand). The change in NO2 is less than linear in N2O [3], and changes in O3, H2O, halogens and temperature cannot explain the difference [3,4]. The trends can be reconciled by a small but significant decrease in background stratospheric aerosol, which is plausible. The trends can also be reconciled by a change in the vertical profile of NO2, hence the operational technique for profile retrieval described here. More than 20 ground-based zenith-sky spectrometers are routinely measuring stratospheric NO2. They cover all latitudes and include the primary stations of the International Network for Detection of Stratospheric Change (NDSC). So far, total columns are deduced from a standard vertical profile. Determining vertical profiles at all stations would help us to deduce more accurate and representative trends. In particular, measurements from polar stations in summer can discriminate against the dependence of NO2 on changes in stratospheric aerosol, as lack of darkness means there is no N2O5 to hydrolyse on aerosol, meaning little change in NO2. Here we present a new tool for operational retrieval of NO2. We use an optimal estimation scheme to change each series of columns measured during a twilight into a profile, coupled with a Radiative Transfer Model (RTM) and the output of a chemical box model to calculate transfer functions between columns and profiles, taking into account the chemical changes of NO2 during twilight. We present an overview of the models, tools and input data used in the retrieval package. The impact of input, resolution and chemistry are quantified, and recommendations for using the software are given. Finally, we have compared retrievals with high-resolution profiles measured simultaneously from a balloon.

2. Retrieval software presentation The software adopts the classical method of detailed and specified forward models coupled to a generalised retrieval module. The RTM is the forward model between profiles and columns. It depends on the box model for the chemistry of NO2 at twilight, and on climatologies for the calculation of atmospheric transmissions. The components are presented here in order of their use. 2.1. The climatologies Climatologies are needed as a constraint to the RTM, to initialise the box model, and to provide a starting profile (a priori) for the retrieval module. The box model is initialised with the

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

323

output of the Cambridge 2D model [5], extended with full heterogeneous chemistry and run in the conditions of 1995. The 2D model output contains mixing ratios of 25 constituents plus temperature and approximate altitude, at 17 pressures from 1000 hPa in steps of 0.5 scale heights (approximately, from the ground to 58 km in 3.5 km steps). They are tabulated in monthly averages from pole to pole at latitude intervals of 101. An initialisation tool reads, interpolates in latitude and date, and saves to the box model initialisation files, the mixing ratios of selected constituents plus temperature. Values are saved at the same pressures as those of the 2D model. The constituents saved are O3, NOy compounds (HNO3, NO, NO2, N2O5, ClONO2, BrONO2), ClO, HCl, HOCl, BrO, BrCl, N2O, CH4, CO, H2O and H2O2. An NO2 climatology based on HALOE data, and developed by Lambert [6] is used as a comparison with the 2D model output. Results showed good agreement in the stratosphere at mid-latitudes. Temperature and pressure from the CIRA 86 model [7] are used for equivalence between number densities and mixing ratios. Finally, a single profile of aerosol extinction is used to calculate scattering in addition to Rayleigh scattering in the RTM [8]. 2.2. The chemical box model The box model is necessary to account for the chemical variation of NO2 during twilight. The model is a non-family version derived from the SLIMCAT 3D Chemistry-Transport Model [9] and is identical to the 3D model in its chemistry. It includes 52 constituents and heterogeneous chemistry on liquid and solid aerosols. Aerosol amounts are based on the H2SO4 distribution and its temperature-dependent equilibrium with HNO3 and H2O. The time-step is chosen to be 1 min. The model is run for 30 days to assure convergence of the NOy compounds within 0.1%, using the solar parameters from a single day throughout the run. It is run independently at each of the 17 pressures of the 2D-model climatology. It is initialised by the files from the initialisation tool described in Section 2.1, plus other constituents initialised to constant values. The output takes the form of a look-up table of NO2 as a function of solar zenith angle (SZA) and altitude, and is used for creating the input profile in the RTM. 2.3. The RTM The RTM used here was developed at Service d’Aeronomie and British Antarctic Survey in the early 1990s for air mass factor (AMF) characterisation [10]. The model has been extensively validated, showing a good agreement with other RTMs in an intercomparison in 1995 [8]. It is made for zenith-sky viewing, and calculates the transmission function by forward ray tracing in a fully spherical atmosphere in shells of 1 km thickness from the ground to 90 km. The model runs in single or double scattering, and uses density, O3, NO2 and aerosol profiles as input. A polar stratospheric cloud (PSC) or a volcanic aerosol layer can be included as an option. The NO2 transmission in each layer above 18 km is chemically weighted using the look-up table calculated by the box model. Chemistry is ignored below 18 km due to the proximity of the tropopause. Sensitivity studies have shown that the change from first to second order of scattering was four times smaller than that induced by a change of the NO2 input profile, and three times smaller than that induced by its chemical variation, and can therefore be ignored. The transmission function is

ARTICLE IN PRESS 324

L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

A = 0.01 A = 0.1 A=1 A=4 A = 10

NO2 A.2.5x1024e-((z-28)/(2x5))

60

80

Altitude (km)

Altitude (km)

80

40 20

60 40 60˚ 80˚ 85˚

88˚

90˚

92˚

94˚

20

0

0 0

5

10

15 24

20 -3

Number density (x10 km )

25

0

5

10

15

20

(SCDpert - SCDinit)/dx

25

30

35

40

(cm-2/cm-3)

Fig. 1. Left, Gaussian profiles of simulated NO2 with various amplitudes, used in the Radiative Transfer Model for weighting function calculations. Right, weighting functions of the Gaussian profile chosen as standard (A=1) at various values of solar zenith angle between 601 and 941.

used to produce simulated slant columns and chemically modified transfer functions from slant columns to vertical profiles (commonly known as weighting functions—see Fig. 1). 2.4. The retrieval module Vertical profiles are deduced from slant columns using an optimal linear inverse method [11,12] calculating the maximum a posteriori solution. It is a weighted mean between an a priori vertical profile (a function of altitude), and the measurement series (a function of SZA) transferred into the vertical space by the matrix of weighting functions. The weights are the errors of the a priori profile, and the error of each measurement is similarly transformed. In our implementation, the vertical profile is determined at vertical intervals equal to the approximate vertical resolution, given by the mean of the full width at half maximum of the averaging kernels (see below). Averaging kernels are calculated on a 1 km vertical grid in the stratosphere. The information content of the retrieved profile is also estimated. The a priori profile can come from a measurement, or from either of the climatologies in Section 2.1. 2.5. Automation and numerical specifications Each module of the retrieval suite is optimised for processing files from a large database in a single step. Parameters are read from separate files that include all information needed for input and output, as many times as there are files to be treated. Certain limitations apply. The atmosphere is limited to 90 km for the RTM and for temperature profiles from CIRA. It is limited to 58 km for the chemical changes from the box model, above 58 km chemical compounds are set to zero. No transport is taken into account. Finally, below 15 km and above 45 km, the information content becomes inherently small and the profile tends towards the a priori profile.

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

325

3. Errors and quality 3.1. Errors and limitations in the RTM Sarkissian et al. [13] assessed the dependence of the RTM results for NO2 on input profiles and on chemical variations during twilight. The main effect was due to changes in the NO2 profile, 1.5 times larger than that due to the chemical variations during twilight. This is why we chose welldefined climatologies of NO2 profile for the RTM input, rather than a fixed profile independent of latitude and season. We chose to ignore second scattering in the RTM as it had a small effect, although the option can still be included at the expense of computer time. Preston et al. [14] analysed information content of the retrieved profile and showed that useful information was limited to altitudes between 10 and 35 km. They showed that an SZA range from 851 to 951 gave better results than 851 to 921. They also showed that the profile was very sensitive to aerosol in the chemical model and to changes in ozone comparable in amplitude to an ozone hole (reinforcing our choice of equilibrium aerosol in the chemical model, and climatological rather than fixed profiles of temperature) but the errors in profile due to typical errors in slant columns were small. We, therefore, chose to process data using monthly input parameters interpolated to the exact latitude of the measurements, to use slant columns between 851 and 951, and to output profiles only from 10 to 35 km. 3.2. Linearity of the retrieval For constituents like ozone, weighting functions are affected by a small variation of the constituent amount, because typical atmospheric amounts of ozone render it optically thick. Hence a first-guess retrieval using a climatological ozone profile must be performed, followed by recalculation of the weighting functions using the retrieved ozone profile. We have investigated the possible non-linearity of weighting functions of NO2 in order to determine if a second iteration is similarly necessary for NO2 retrieval. We calculated weighting functions using a series of NO2 profiles from 100 times smaller to 4 times larger than a profile similar to mid-latitude spring (a Gaussian of amplitude 2.5  109 molecules/cm3 centred at 28 km with half-width 5 km). No chemistry was taken into account for this non-linearity study. Fig. 2 shows the results. For NO2 less than the standard, there is negligible difference at SZA o901, and differences are o2% elsewhere. For NO2 of 4 times the standard, an amount that is so large it is barely conceivable, differences are a maximum of 7% at 941. We, therefore, conclude that non-linearity of weighting functions is not an issue, and that iteration of their calculation during the retrieval process is unnecessary. 3.3. Effect of NO2 chemistry at twilight on the retrieval Diurnal changes of NO2 change the vertical distribution in the profile, and therefore, have an effect on the weighting functions as well as on the outcome of the retrieval. Fig. 3 shows ratios of weighting functions calculated taking into account the chemistry, to weighting functions without taking it into account. Not taking into account the chemistry systematically increases the weighting functions by 10 to 40% between 20 and 40 km. Above 40 km, the increase is much more

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

Altitude (km)

80

x 0.01 x 0.1 x1 x4

88˚

60

80 Altitude (km)

326

40

60 40 20

20

0

0

80

0.95 1.00 1.05 Ratio (WFQvar/WFQstd)

1.10

80

x 0.01 x 0.1 x1 x4

92˚

60

0.97

Altitude (km)

0.90

Altitude (km)

x 0.01 x 0.1 x1 x4

90˚

40

0.98

0.99 1.00 1.01 1.02 Ratio (WFQvar/WFQstd)

1.03

x 0.01 x 0.1 x1 x4

94˚

60 40 20

20

0

0 0.6

0.8 1.0 1.2 Ratio (WFQvar/WFQstd)

0.6

1.4

0.8 1.0 1.2 Ratio (WFQvar/WFQstd)

1.4

Fig. 2. Weighting function ratios using the NO2 standard profile and others defined in Fig. 1. Differences from unity increase with increasing solar zenith angle. For amounts of NO2 four times the standard, the differences are 77% at 941, 75% at 921 and 72% at 901.

SRS

40

SST

50

70˚ 80˚ 88˚ 90˚ 92˚ 94˚

Altitude (km)

Altitude (km)

50

30 20

80˚ 80˚ 88˚ 90˚ 92˚

40 30 20

0.5

1.0 1.5 2.0 Ratio WFno_chem/WFchem_srs

2.5

0.5

1.0 1.5 2.0 Ratio WFno_chem/WFchem_sst

2.5

Fig. 3. Ratios of weighting functions with and without chemical modification for selected SZA between 701 and 941 at sunrise (left) and sunset (right). The large ratios above 40 km are due to the large changes in NO2 there during twilight, but they are not important in our retrieval because of the limited information content and the small amount of NO2 above 40 km.

rapid, up to 2 to 3 times at 55 km, the upper level of the box model, but this is not important because the amount of NO2 is negligible at these high altitudes. At any altitude, the chemical changes are much larger than those caused by varying the NO2 input profiles in the RTM. More

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

327

importantly, the chemical changes are also very variable in altitude and SZA. Therefore, we conclude that it is necessary to include the chemistry in the retrieval.

4. Example and validation at mid-latitudes We applied our retrieval procedure to a series of ground-based measurements between 11 and 21 June 1996 from Observatoire de Haute Provence (OHP, 44.01N, 5.71E). The instrument was a zenith-sky UV-visible spectrometer from the Institut d’Ae´ronomie Spatiale de Belgique (IASB), deployed at OHP with many similar spectrometers as part of an intercomparison campaign [15]. The time of year was chosen partly because of reliable periods with clear skies, at least in the mornings, and also because the stratospheric winds are weak, so that stratospheric composition is slowly varying. Lack of cloud is important for a northern hemisphere site, otherwise one must take care to avoid even minor pollution episodes when measuring total NO2. The IASB instrument was chosen because it has excellent signal-to-noise ratio, and because it performed well in the intercomparison. From 19 to 20 June 1996, a UV–visible spectrometer of the SAOZ design (Syste`me d’Analyse par Observation Ze´nithales), modified for direct-sun observations from a balloon [16] was flown from Gap–Tallard (44.51N, 6.01E). The flight ended near Toulouse (43.61N, 1.91E). Spectra of the sun during occultation were recorded, and they included the sunset of 19 June (observed near Gap) and sunrise of 20 June (observed near Toulouse). Profiles from the balloon flight were retrieved by an onion-peeling method. Because of the short time between spectra, altitude differences between tangent heights were small. Such oversampling leads to the vertical resolution approaching the limit for occultation measurements of 1–2 km in the middle stratosphere near balloon altitude [17]. In the lower stratosphere well below balloon altitude, the smearing of tangent heights caused by the finite size of the solar disc gives a vertical resolution of about 5 km [17]. Throughout the stratosphere, this is finer resolution than that of ground-based profiles. 4.1. Slant column densities Fig. 4 shows all the slant columns measured by the IASB spectrometer. The SZA range is from 221 to 941 and the mean interval between measurements is 11. Sunsets on 11, 14, 15, 16 and 21 June show enhancements of NO2, mostly below 851, indicative of dense cloud plus pollution episodes, common during some afternoons of the period. These twilights were ignored for our validation purposes below. Sunrises show no such enhancements, as expected from the generally clear mornings at the site and from the diurnal cycle of NO2 in polluted areas. From Fig. 4, sunrises also have systematically less NO2 than sunsets, as expected from the conversion of NO2 during the night to NO3 and N2O5, which is then photolysed the following day. 4.2. Retrieval results Figs. 5–7, together with Table 1, show samples and results from various steps in the retrieval process during this validation exercise. Standard RTM and box model initialisation profiles and

ARTICLE IN PRESS 328

L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

NO2 SCD (x1027 molec/km2)

2.5 15sst

2.0

1.5

1.0 21sst

16sst

14sst

11sst

0.5

sunrise sunset

0.0 60

65

70

75

80

85

90

95

SZA (˚)

Fig. 4. Series of NO2 slant column densities (SCD) measured by the IASB zenith-sky ground-based spectrometer at OHP (44.01N, 5.71E) between 11 and 21 June. Dashed lines are at sunrise. Sunsets of 11, 14, 15, 16 and 21 June show perturbation by tropospheric clouds plus pollution and were not used to retrieve vertical profiles.

80

O3 Altitude (km)

Altitude (km)

80 60 40 20

NO2

60 40 20

0

0

0

1

2

3

4

5

Number density (x1012 mol/cm3)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

Number density (x109 mol/cm3)

Fig. 5. O3 and NO2 profiles used for initialisation of the models in the retrieval process, as extracted from the Cambridge 2D model climatology at 441N in June.

values were used as described earlier. In the retrieval, the a priori NO2 profile was taken from the 2D-model climatology. The vertical resolution of the result is about 8 km, as expected. As can be seen in Fig. 7, the NO2 profile shows a day-to-day variation of 25% at sunrise and 15% at sunset. The mean ratio of values at sunset to those at sunrise between 12 and 20 June is 270.2 at 25 km. 4.3. Validation Fig. 8 shows the comparison between retrieved and measured profiles. The standard deviation of the differences is 13%, and agreement is much better than this above 15 km, where groundbased information content is greatest. Given the day-to-day variation in the NO2 profile in Fig. 7,

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

80

44˚N, 20 June Sunrise

Altitude (km)

Altitude (km)

80 60 40 70˚80˚ 85˚ 88˚ 90˚

92˚

94˚

20

329

44˚N, 20 June Sunset

60 40 70˚80˚ 85˚ 88˚ 90˚

92˚

94˚

20

0

0 0

10

20

30

40

50

0

(Ylospert - Ylosinit)/dx ( molec.cm-2/molec.cm-3)

10

20

30

40

50

(Ylospert - Ylosinit)/dx ( molec.cm-2/molec.cm-3)

Fig. 6. Weighting functions used for the retrieval of the profiles at 441N in June. 50

50 Sunrise

Sunset 40

12/06 13/06 14/06 15/06 16/06 17/06 18/06 19/06 20/06 21/06

30 20

Altitude (km)

Altitude (km)

40

10

12/06 13/06 17/06 18/06 19/06 20/06

30 20 10

0

1

2

3

4

5

6

NO2 Number density (x109 molec/cm-3)

0

1

2

3

4

5

6

NO2 Number density (x109 molec/cm-3)

Fig. 7. Retrieved profiles at 441N in June. The larger values at sunset on 12 June correspond to the larger slant columns in Fig. 4 at solar zenith angles above 871. Table 1 Ratios of NO2 at SZA to NO2 at solar zenith angle of 901, at selected solar zenith angles and altitudes, at sunset Altitude (km)SZA(1)

20

25

30

35

40

85 92 94 95

0.82 1.13 1.17 1.16

0.90 1.13 1.28 1.28

0.91 1.15 1.44 1.49

0.92 1.12 1.58 1.95

0.91 1.11 1.60 2.19

these small differences could be due to the difference in location between OHP and Gap and Toulouse, although the vertical resolution of the retrieved profile may also be partly responsible. 4.4. Error analysis Here, we show the error analysis of the profile retrieved for 19 June at sunset. The total error is the sum of three components, calculated according to Rodgers [12]. The first term is the

ARTICLE IN PRESS 330

L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333 50 Retrieved, 19 sunset Retrieved, 20 sunrise Balloon, 19 sunset Balloon, 20 sunrise First guess

Altitude (km)

40

30

20

10 0

1

2

3

4

5

6

NO2 Number density (x109 molec/cm-3)

Fig. 8. Comparison between the profiles retrieved from the ground-based spectrometer in this work and from the balloon-borne spectrometer on 19 June at sunset, and on 20 June at sunrise.

smoothing error Ss, due to the application of averaging kernels during the retrieval process. This error is systematic and also uses the variance assigned to the a priori profile. The second term is the random component Sm due to noise on the measurement error. The third term is the forward model error Sf, due to systematic errors in parameters used in the forward model. These terms are given by: Ss ¼ ðA  IÞSa ðA  IÞT ;

(1)

Sm ¼ Gy Se GTy

(2)

Sf ¼ Gy Kb Sb KTb GTy

(3)

where A is the matrix of averaging kernels, Sa is the error covariance matrix of the a priori profile, Gy is the matrix of contribution functions, Se is the error covariance matrix on the measurements, Kb is derived from the sensitivities of the slant columns to different parameters of the forward models (see [14] for a fuller derivation), and Sb is the error covariance matrix of the forward model parameters. For our calculations, we used the parameter errors adopted by Preston et al [14]. Fig. 9 shows the error profiles. The retrieved noise has the smallest contribution, of the same order than the difference between retrieved profile and SAOZ profile in Fig. 8. Above 30 km, the smoothing error dominates the total error.

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

331

50 Profile Meas. err Smooth. err Model err. Total err.

Altitude (km)

40

30

20

10 0

1

2

3

4

5

6

Number density (x109 mol/cm3)

Fig. 9. Error profiles estimated for the retrieved profile of 19 June at sunset.

4.5. Extension to polar latitudes Section 4.2 has validated our climatology, initialisation and models, for use at mid-latitudes. The question arises: how well do we expect these aspects of the suite to perform in the difficult conditions of polar winter or spring? (a) Use of the CIRA climatology for densities for the RTM has shown that it works well in the Arctic but the AMFs produced have a seasonally dependent error in the Antarctic which affects ozone AMFs by about 5% [18]. (b) The box model is run for enough number of days to be independent of errors in the chemistry initialisation. (c) The chemical reaction suite within the box model itself has been widely used as part of SLIMCAT in studies of ozone-hole chemistry with excellent results, so we are confident that it can be used in the presence of polar stratospheric clouds with good accuracy, provided the temperature is correct so that the clouds are correctly diagnosed by the model. In general, future use of the suite for data in polar winter and spring will have to address some of these issues.

5. Conclusions We have developed a new operational software suite for the retrieval of NO2 profiles from ground-based measurements, using a classical optimal-estimation scheme. The NO2 slant columns measured by zenith-sky UV-visible spectrometers are inverted to vertical profiles at altitudes

ARTICLE IN PRESS 332

L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

adapted to the true vertical resolution of the method. As usual, the method includes a rigorous mapping of random and systematic errors from the measurements and from the input parameters onto the retrieved profile. The suite uses some inputs from climatologies (e.g. vertical profile of density, for the calculation of zenith-sky scattering) and some initialisation parameters from climatologies (e.g. a priori profile of NO2, to start the final inversion). Each climatology is from a well-documented and validated source. The suite is modular, with files that define parameters of each module, and can be edited, for input and output files to pass between modules. In this way flexibility of operation is guaranteed, as well as the ability to diagnose the causes of any unusual features in the retrieved profiles. If the chemical modifications of NO2 near twilight are precomputed to a monthly look-up table, the suite can invert one year’s slant column measurements in a few minutes on a PC. If chemical modifications are computed afresh for each measurement day, the process takes about a day. The software suite is now being evaluated by several European groups who operate ground-based zenith-sky UV-visible spectrometers for routine measurements of NO2. The suite will be used at BAS to help determine trends of global NOy, by inverting profiles of NO2 in Antarctic summer, when the lack of N2O5 minimises the effect of stratospheric aerosol on the relationship between NO2 and NOy.

Acknowledgements This work is supported by the European project EVK2-2000-00545, ‘‘Quantification and Interpretation of Long-Term UV-Visible Observations of the Stratosphere’’ (QUILT) and by the UK Natural Environment Research Council’s Thematic Programme ‘‘UTLS-Ozone’’ via grant NER/T/S/1999/00120.

References [1] Prinn RG, Zander R. Long-lived ozone related compounds, in scientific assessment of ozone depletion: 1998 World Meteorological Organization, global ozone Research Monitoring Project, Report 44. National Oceanic and Atmos. Admin., Silver Spring, MD: 1999, pp. 1.1–1.54. [2] Liley JB, Johnston PV, McKenzis RL, Thomas AJ, Boyd IS. Stratospheric NO2 variations from a long time series at Lauder, New Zealand. J Geophys Res 2000;105:11633–40. [3] McLinden CA, Olsen SC, Prather MJ. Understanding trends in stratospheric NOy and NO2. J Geophys Res 2001;106:27787–93. [4] Fish DJ, Roscoe HK, Johnston PV. Possible causes of stratospheric NO2 trends observed at Lauder, New Zealand. Geophys Res Lett 2000;27:3313–6. [5] Harwood RS, Pyle JA. A two-dimensional mean circulation model for the atmosphere below 80km. Quart J R Met Soc 1975;101:723–47. [6] Lambert JC, Granville J, Van Roozendael M, Sarkissian A, Goutail F, Mu¨ller J-F, Pommereau J-P, Russell III JM. A climatology of NO2 profile for improved air mass factors for ground-based vertical column measurements, Proceedings of the 5th European Symposium Strat. Ozone, Saint Jean de Luz, Sept 1999, Air pollution research report 73. In: Harris NRP, Guirlet M, Amanatidis GT, editors. European Commission, 2000, p. 703. ISBN 92-827-5672-6.

ARTICLE IN PRESS L. Denis et al. / Journal of Quantitative Spectroscopy & Radiative Transfer 92 (2005) 321–333

333

[7] Labitzke K, Barnett JJ, Edwards B, editors. Middle atmosphere program, MAP handbook, vol. 16. Urbana: University of Illinois, 1985. [8] Sarkissian A, Roscoe HK, Fish DJ, Van Roozendael M, Gil M, Chen HB, Wang P, Pommereau J- P, Lenoble J. Ozone and NO2 air-mass factors for zenith-sky spectrometers: intercomparison of calculations with different radiative transfer model. Geophys Res Lett 1995;22:1113–6. [9] Chipperfield MP. Multiannual simulations with a three dimensional chemical transport model. J Geophys Res 1999;104:1781–805. [10] Sarkissian A, Roscoe HK, Fish DJ. Ozone measurements by zenith-sky spectrometers: an evaluation of errors in air-mass factors calculated by radiative transfer model, JQSRT 1995;54:471–80. [11] Rodgers CD. Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation. Rev Geophys Sp Phys 1976;14:609–24. [12] Rodgers CD. Inverse methods for atmospheric sounding. In: Theory and practice, series on atmospheric, oceanic and planetary physics, vol. 2, Taylor FW, editor. University of Oxford, World Scientific Publishing Co. Pte. Ltd., 2000. [13] Sarkissian A, et al. Improved air-mass factors for ground-based total NO2 measurements: a sensitivity study. In: Harris NRP, Guirlet M, Amanatidis GT, editors. Proceedings of the fifth Europ. Symp. on Strat. Ozone, Air pollution research report 73, European commission, 2000, p. 730–3. ISBN 92-827-5672-6. [14] Preston KE, Jones RL, Roscoe HK. Retrieval of NO2 vertical profiles from ground-based UV-Visible measurements: method and validation. J Geophys Res 1997;102:19089–97. [15] Roscoe HK, et al. Slant column measurements of O3 and NO2 during the ndsc intercomparison of zenith-sky uv–visible spectrometers in june 1996. J Atmos Chem 1999;32:281–314. [16] Pommereau J- P, Piquard J. Ozone and nitrogen dioxide vertical distributions by uv-visible solar occultation from balloons. Geophys Res Lett 1994;21:1227–30. [17] Roscoe HK, Hill JGT. Vertical resolution of oversampled limb-sounding measurements from satellites and aircraft. JQSRT 2002;72:237–48. [18] Roscoe HK, Hill JGT, Jones AE, Sarkissian A. Improvements to the accuracy of zenith-sky measurements of total ozone by visible spectrometers II: use of daily air-mass factors. JQSRT 2001;68:327–36.