The old version of the reflectance factor calibration. The baseline here is curved.

The new version of the reflectance factor calibration using all solar calibration observations (as of January 2023), with all diffraction orders converted and improved calibration at the edges of the detector.

The spectra are calibrated in reflectance factor, where reflectance factor = pi * normalised nadir counts / normalised solar counts / solar-to-nadir scaling factor / cosine(Geometry/MeanIncidenceAngle).

The baseline here is curved.

The "Y" dataset from previous files. This is the radiance calculated from laboratory blackbody measurements, which uses the AOTF and blaze functions to calculate the full radiance hitting each pixel to derive the counts-to-radiance conversion.

This should be used with caution, because 1) it relies on having correct AOTF and blaze functions; and 2) because the blackbody data used for the calibration is not reliable at high diffraction orders (as the planck function decreases and so the signal becomes very small).

The "YAtWavenumber" dataset from previous files. This is the radiance calibrated from laboratory blackbody measurements without using the AOTF or blaze functions.

The counts to radiance conversion is derived from a simple planck function at the temperature of the blackbody and wavenumber of the pixel. This calibration can also be unreliable for high diffraction orders.

Transmittance is computed using a linear extrapolation of the "Sun region" (i.e. where the tangent point is above the atmosphere) for each pixel individually. This method attempts to correct for limb darkening: during a solar occultation, the pointing is not fixed on the same region of the solar disk, and so the signal from the Sun can increase or decrease as the field of view moves towards the centre or edge of the solar disk. This results in a sloped Sun region which is then fit with a linear line of best fit and extrapolated into the "Penumbral region" (i.e. where the tangent point is passing through the atmosphere). The extrapolation is then used to calculate a solar reference spectrum to calibrate each spectrum measured when the line of sight passes through the atmosphere. See Trompet et al. for a more detailed description.

Advantages: the limb darkening effect is reduced by the linear extrapolation

Disadvantages: the extrapolation is done separately for each pixel, therefore noisy pixels could be fit badly, resulting in a poor extrapolation.

The solar reference spectrum is computed per pixel by taking the mean of all the spectra in the defined "Sun region" (i.e. where the tangent point is above the atmosphere).

Advantages: this method averages together multiple spectra, and so the solar reference spectrum is less noisy and less prone to extrapolation errors due to noisy pixels.

Disadvantages: if the "Sun region" is not flat, e.g. due to limb darkening, the solar reference spectrum used to convert all the atmospheric spectra to transmittance will be too high or low. The transmittance should be 1.00 above the atmosphere, but if limb darkening is not accounted for, the calculated transmittances will be higher or lower than this value.

=> this is the agreed dataset used by the team

Science/YErrorMeanRandom

Random errors as propagated through the various stages of the data pipeline

Transmittance is computed using a linear extrapolation of the "Sun region" (i.e. where the tangent point is above the atmosphere) for each pixel individually. This method attempts to correct for limb darkening: during a solar occultation, the pointing is not fixed on the same region of the solar disk, and so the signal from the Sun can increase or decrease as the field of view moves towards the centre or edge of the solar disk. This results in a sloped Sun region which is then fit with a linear line of best fit and extrapolated into the "Penumbral region" (i.e. where the tangent point is passing through the atmosphere). The extrapolation is then used to calculate a solar reference spectrum to calibrate each spectrum measured when the line of sight passes through the atmosphere. See Trompet et al. for a more detailed description.

Advantages: the limb darkening effect is reduced by the linear extrapolation

Disadvantages: the extrapolation is done separately for each pixel, therefore noisy pixels could be fit badly, resulting in a poor extrapolation.

=> this is the agreed dataset used by the team

The solar reference spectrum is computed per pixel by taking the mean of all the spectra in the defined "Sun region" (i.e. where the tangent point is above the atmosphere).

Advantages: this method averages together multiple spectra, and so the solar reference spectrum is less noisy and less prone to extrapolation errors due to noisy pixels.

Disadvantages: if the "Sun region" is not flat, e.g. due to limb darkening, the solar reference spectrum used to convert all the atmospheric spectra to transmittance will be too high or low. The transmittance should be 1.00 above the atmosphere, but if limb darkening is not accounted for, the calculated transmittances will be higher or lower than this value.

Uncertainties on Y, except that the solar reference spectra are normalised before the standard deviation is made. 

=> this is the agreed dataset used by the team

Uncertainties on YMean

 4 parameters from Villanueva et al. 2022:

•  Sinc width (cm-1)
•  Sidelobe factor
•  Asymmetry
•  Gaussian offset intensity

 5 parameters from Villanueva et al. 2022:

•  Intensity of 2nd Gaussian (constant)
•  Resolving power of both Gaussians (constant)
•  Displacement of 2nd Gaussian (cm-1) w.r.t. 3700cm-1 vs pixel number (quadratic coefficients)

 6 coefficients:

• Quadratic for FWHM in cm-1 (diffraction order is input variable) - values are unchanged [0, 0, 22.473422]
• Quadratic for peak position in cm-1 (diffraction order input) - new values [0.0, <FSR>, 0.0] where <FSR> is the temperature + wavenumber corrected value from Villanueva et al. 2022