Determination of the background signal in the UV

As a first step in the determination of the pure emission signal, it has to be sep-arated from the background signal. The background signal is mainly formed by Rayleigh and Raman scattering of solar electromagnetic radiation by air molecules.

Above 70 km the air is thin enough and scattering path lengths are long enough that only single scattering has to be considered. Below 300 nm, ozone in the strato-sphere absorbs the main part of the incoming solar electromagnetic radiation, so that no multiple scattering or surface reflection contributions have to be considered.

Figure 6.1 shows recent measurement results of ozone absorption cross section for typical atmospheric temperatures by Gorshelev et al. (2013) and Serdyuchenko et al.

(2013). The ozone absorption cross section between 200 nm and 800 nm wavelength

200 300 400 500 600 700 800

10⁻²⁶ 10⁻²⁴ 10⁻²² 10⁻²⁰ 10⁻¹⁸ 10⁻¹⁶

Ozone absorption cross section in cm2

Wavelength in nm

293 K 193 K

Fig. 6.1: O3 cross section for the 200 nm to 800 nm range from Gorshelev et al.

(2013); Serdyuchenko et al. (2013). For a typical ozone profile at mid latitudes and a solar zenith angle of SZA = 0^◦, 96 % of the light, incoming from the top of the atmosphere, at 600 nm and at 330 nm is left at 10 km altitude, while at 300 nm only 5 % is left. For the calculated remaining light at 285 nm and 280 nm values of the order of 10⁻⁹ and 10⁻¹⁴, respectively, are obtained. For this study, only the ozone absorption in the stratosphere is relevant: for a typical SCIAMACHY MLT limb state even at 280 nm the remaining signal is still more than 99 % of its original value after passing the whole line of sight for tangent altitudes above 68 km. Therefore, ozone absorption in the mesosphere can be and is neglected in the radiative transfer model.

has three distinguished strong absorption bands. The Hartley band with a maxi-mum absorption cross section at 255 nm shows the strongest absorption. The other bands are the Huggins band between 320 and 360 nm, which overlaps the edge of the Hartley band and the Chappuis band with a maximum absorption cross section at 600 nm. In the mesosphere the ozone density is small enough, so that less than 1%

of the incoming solar electromagnetic radiation at 280 nm is attenuated along a line of sight with a tangent altitude of 70 km. However, for wavelengths closer to 255 nm the ozone absorption is important. Below 70 km altitude the Ring effect, described in the next section, is a larger error source than ozone absorption.

In the stratosphere, the ozone layer is thick enough that far less than 1% of the UV radiation below 295 nm reaches the ground. Therefore, multiple scattering

275 280 285 290 295 300 305 0.2

0.4 0.6 0.8 1 1.2 1.4 1.6

Wavelength in nm

|single−multi|/min(single,multi) in %

Fig. 6.2: Relative difference ^|multiple scattering−single scattering|

min(multiple scat., single scat.) of simulated spectra for multiple scattering for the wavelength region of the Mg/Mg⁺ lines. The simula-tions were performed with SCIATRAN for an equatorial scenario. This and other scenarios with, e.g., a total ozone column density of 150 Dobson units (1 Dobson unit= 2.687× 10¹⁶cm⁻²) show only minor differences of less than 1%. Multiple scattering thus can be neglected for the retrieval of Mg/Mg⁺ densities in the MLT.

from lower altitudes is negligible at 280 and 285 nm. Figure 6.2 shows the relative difference between singly scattered and multiply scattered electromagnetic radiation for a typical limb scan. This was simulated by Vladimir Rozanov with the more sophisticated radiative transfer model SCIATRAN (see, e.g., Rozanov et al., 2005) and shows less than 1% differences between single and multiple scattering at these wavelengths.

The cross section for single Rayleigh scattering for a refractive index of n= 1 is given by the following equation:

σ(λ) = 128π⁵α²

3 · 1

λ⁴ (6.2)

with the polarizability volume α and the wavelength λ. The single Rayleigh scat-tered spectrum, therefore, is an attenuated copy of the solar spectrum with an additional wavelength modulation of λ⁻⁴. For wavelength windows with a much smaller width than the mean wavelength, e.g., a 2 nm window at 280 nm, the ratio of Rayleigh scattered electromagnetic radiation and incoming solar radiation can be well approximated by a linear function.

Figure 6.3 shows the measured ratio of limb radiance and solar irradiance between 275 and 290 nm. The strong peaks in this ratio are clearly identified as

emission signals or inelastic Raman scattering.

275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 2900 0.5

1 1.5 2 2.5 3 3.5x 10

−4

Wavelength in nm

Limb radiance/solar irradiance in sr−1

Fig. 6.3: Limb radiance divided by solar irradiance (average 2008–2012, equator, tangent altitude 89.3 km). Division of both spectra leads to a smooth signal, beside emission lines that are clearly identified. The background signal is separated from the emission lines through a linear fit. Note, that beside the strong Mg⁺at 279.6 and 280.4 nm and the strong Mg line at 285.2 nm very weak NO lines can be observed at the longer wavelength edge of the Mg⁺ lines at 281 nm. However, this influence is negligibly small.

To exclude influence of the inelastic rotational Raman scattering a Ring effect correction, which is described in the next section, is performed before forming the ratio of limb radiance and solar irradiance. The Rayleigh background can be fitted as a straight line of this ratio for a small window around the emission lines, but excluding the emission lines. After subtraction of the Rayleigh background, the ratio is multiplied with the solar irradiance again to obtain the pure emission spectrum.

The result is fitted with the slit function of the instrument to obtain the slant column emission (SCE) signal. These steps are illustrated in Fig. 6.4. For a good fit of the slant column emission signal it is advantageous to have a good signal to noise ratio, low influence of Ring effect and strong emission signals compared to other background signals and no overlap with other nearby emission lines. Figure 6.3 shows a small emission line at 281 nm from the overlapping NOγ(1,6) band. However, this emission is negligibly small.

2820 283 284 285 286 287 288 0.5

1 1.5

2x 10⁹

Radiance I in ph cm⁻²s⁻¹nm⁻¹sr⁻¹ Ring corrected Radiance

2822 283 284 285 286 287 288 4

6 8 10 12 14x 10⁻⁵

I/F in sr⁻¹ Left Right Linear fit

282 283 284 285 286 287 288

−2 0 2 4 6 8 10x 10⁻⁵

I/F − linear fit in sr⁻¹

282 283 284 285 286 287 288

−2 0 2 4 6 8 10x 10⁸

(I/F−linear fit)*F in ph cm⁻²s⁻¹nm⁻¹sr⁻¹ Fit with shape function

Wavelength in nm

Fig. 6.4: Slant column emission (SCE) determination. First, the Ring effect correc-tion is applied to the limb radiance I (top left panel). The limb radiance is divided by the solar irradiance F and the wavelength region left and right to the line are used for a linear fit of the background radiation (top right panel). The linear fit is subtracted from the ratio I/F (bottom left panel) and the spectrum is again mul-tiplied with the solar irradiance F (bottom right panel). Finally, the area of the emission line is fitted with the slit function of the instrument.