Solar Reference Spectra

An empirical solar linelist has been generated by simultaneous fitting of ATMOS, MkIV, Kitt Peak, Denver U, and TCCON spectra. The telluric absorptions were fitted using the HITRAN linelist and any remaining, airmass-independent absorptions were attributed to the sun. Together with a simple lineshape function, this linelist allows the computation of a solar pseudo-transmittance spectra, for disk-center, disk-integrated, or intermediate cases. These spectra have been validated using MkIV, Kitt Peak, ACE, and GOSAT spectra.

In deriving the Solar Pseudo-Transmittance Spectrum (SPTS), we assumed that each measured solar spectrum, Yi(v), can be expressed as

               Yi(v) = ILSi(v-v') * [ISRi(v') . SPF(v') . SPTS(v') . TTSi(v')]

where ILSi(v-v') is the Instrument Line-Shape for the i'th measured spectrum. ISRi(v') is the Instrument Spectral Response. SPF(v') is the Solar Planck Function. SPTS(v') is the Solar Pseudo-Transmittance Spectrum. TTSi(v') is the Telluric Transmittance Spectrum. The convolution operation is denoted by *. Note that SPTS(v') and SPF(v') are independent of i, whereas all other terms depend on i.

ILSi can be computed from the Max OPD and the internal FOV of the instrument that measured each spectrum, and verified by fitting narrow telluric absorption lines of gases whose vertical profiles are well known. The instrument spectral response, ISRi(v), is not known but varies slowly with wavenumber, as does the Solar Planck Function, SPF(v'). So the product of ISR and SPF were fitted as a low order polynomial. The Solar Pseudo Transmittance spectrum, SPTS(v'), contains the narrow solar Fraunhofer lines and is independent of the measured spectrum (i). The Telluric Transmittance spectrum, TTSi(v'), contains narrow absorption features from gases in the Earth's atmosphere and can be computed using the HITRAN linelist together with knowledge of the vertical distribution of gases in the Earth's atmosphere.

The solar "pseudo-transmittances" have numerical values varying between 0 and 1.0, the latter values representing wavenumber regions in which there are negligible discrete absorption (or emission) lines. We assume that the high-frequency spectral structure is solar (or telluric), whereas the low frequency structure is instrumental (ISR) and due to the solar planck function (SPF). This allows the slowly-varying spectral structure to be fitted as a continuum polynomial, while the solar Fraunhofer absorption is represented by over 40,000 discrete narrow lines.

This assumption (that the solar Fraunhofer lines are narrower than any feature in the ISR) in not valid for solar H-atom absorption lines, which can be as wide as 5 cm-1. Fortunately, the H-atom spectral lines are well characterized in terms of frequency and intensity, allowing them to be distinguished from instrumental artifacts. The low-resolution component of the solar spectrum (i.e. the solar planck function) is not considered here. Nor is the absolute radiometry of the sun (i.e. 1367 W/m2).

At high wavenumbers (> 23,000 cm-1) there are so many strong overlapping Fraunhofer lines that it becomes difficult to tell at which signal level the "continuum" lies, i.e. SPF(v). Unlike the IR, there are few window regions where the Fraunhofer absorption is negligible, and so the decomposition of the solar spectrum into a smooth SPF(v) and a bumpy SPTS(v) becomes more arbitrary and more dependent on the assumed far-wing lineshape.

To convert a solar pseudo-transmittance spectrum into an absolute irradiance spectrum, it is necessary to multiply it by the Solar Planck Function, SPF(v). Unfortunately this is not as simple as it sounds because the effective solar temperature is wavelength dependent, peaking at about 6000 cm-1 where the solar H- opacity is a minimum. We have therefore not attempted to do this. In any case, for most atmospheric remote sensing applications, it is not necessary to know the absolute solar irradiances because the atmospheric gas amounts depend on the fractional depths of the absorption lines, not their absolute depths.


The derived Solar Pseudo-Transmittance Spectrum, SPTS(v), recent versions of which can be downloaded from the table below, was computed using various recent versions of the solar linelist. They cover 600 to 15500 cm-1 (2013 & 2015) and 600 to 26316 cm-1 (2016) on a 0.01 cm-1 grid. [26316 cm-1 corresponds to 380 nm, which is the limit of the Delbouille et al. [1972, 1981] solar atlas [available from:]. In each file the first column is the wavenumber (cm-1) and the second column is SPTS(v).

When fitting/representing spectra of sunlight reflected from the ground, the moon, or diffuser-plates (e.g. SCIAMACHY, GOSAT, OCO-2), we recommend using the disk-integrated solar spectrum. When representing/fitting spectra obtained using direct sunlight from the center of the solar disk (e.g. NDACC-IRWG, TCCON, ATMOS, MkIV, ACE) we reccommend the disk-center solar spectrum.

Important: These are solar pseudo-transmittance spectra, having values between 0 and 1, not a list of solar line positions and intensities.

Disk-Center SpectrumDisk-Integrated Spectrum
2013 version solar_merged_20130402_0600_15500_000.out.gz solar_merged_20130402_0600_15500_100.out.gz
2015 version solar_merged_20150308_0600_15500_000.out.gz solar_merged_20150308_0600_15500_100.out.gz
2016 version solar_merged_20160127_600_26316_000.out.gz solar_merged_20160127_600_26316_100.out.gz
2020 version solar_merged_20200720_600_33300_000.out.gz solar_merged_20200720_600_33300_100.out.gz

Plots illustrating the 2013 and 2015 disk-integrated (DI) solar spectra and their differences (28 pages) are available from: merged_solar_comp_2015.pdf The blue lines show 2013, the red lines show 2015, the green is the difference (x10). All plotted versus wavenumber (cm-1).

Plots illustrating the 2015 and 2016 disk-integrated (DI) solar spectra and their differences plotted versus wavenumber (258 pages) are available from: solar_comp_merged_2016.pdf The blue trace shows the 2015 spectrum, the red trace shows the 2016, the green trace is difference (x10). The main improvements in the 2016 spectrum are in the 15500 to 26316 cm-1 and in regions with strong H2O absorption. In the regions covered by GOSAT and OCO2 there is virtually no change in the 2016 spectrum as compared with 2015.

Plots illustrating the 2016 and 2020 disk-integrated (DI) solar spectra and their differences plotted versus wavenumber (324 pages) are available from: solar_comp_merged_2020.pdf The blue trace shows the 2016 spectrum, the red trace shows the 2020 spectrum, green trace is difference (x10). The improvements are most extensive in the 7600 to 9300 cm-1 region. There are also minor imporovements in the 6153 to 6235 cm-1 region. The 2020 solar linelisa,t used to compute the solar spectra illustrated here, corresponds to the TCCON GGG2020 release.


Toon, G. C., Solar line list for GGG2014, TCCON data archive, hosted by the Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge, Tennessee, U.S.A., doi:10. 14291/, 2014.

Go back to the previous page