Minimize SCIAMACHY Product Handbook

Spatial Stray Light

Table of Contents

 


In order to avoid signal-to-noise problems during the spectral stray light measurements, ten wavelength bands were defined separately for s- and p-polarised light leading to a total of 20 bands. For both polarisation directions 9 bands were located in channel 1 to characterise intra-channel stray light and one band covered the signal from channels 2-5 to characterise inter-channel stray light. The channel 1 detector material is not sensitive for light with wavelengths longer than 1000 nm so the SWIR channels did not need to be considered. For each band the stray light contribution to all detector pixels was calculated leading to a 10 x 1024 matrix for both, s- and p-polarised light. The stray light fraction in channel 1 ranges from less than 1% to as much as 10%. The correction has an accuracy of around 25% and reduces the stray light by an order of magnitude leaving at most 1% stray light in the spectrum after correction.

Spatial Stray Light

Shortly after ENVISAT emerges from eclipse and passes the North Pole, the sun shines directly into the limb port. In this orbit region spatial stray light cannot be avoided, i.e. the particular effect was foreseen and the data are flagged accordingly. In order to minimise spatial stray light, the ASM is rotated such that the edge of the mirror/diffuser plate points into flight direction, with the diffuser looking to the instrument side during all measurements using the ESM only.
Both on-ground performance measurements and in-flight limb measurements indicated that there was a small fraction of spatial stray light present. Dedicated in-flight measurements confirmed the performance measurements and indicate periodic structures in the optics before the slit, resulting in a small fraction of the light being dispersed as by a grating. This has no significant impact for nadir measurements, but limb measurements with a dynamic range of several orders of magnitude over a few degrees suffer from the spatial stray light. At the moment there are no corrections for spatial stray light.

4.5 Polarisation

SCIAMACHY is – as all grating spectrometers without a polarisation scrambler – sensitive to the polarisation of the incoming light, i.e. the response will not only depend on the intensity but also on the polarisation of the light. Thus polarisation correction is required. It uses the Mueller matrix approach (see e.g. Azzam and Bashara 1977, Coulson 1988). Measurements of polarised light can be expressed by a so-called Mueller matrix M and the Stokes vector S. Since the detectors only yield a single measurement value per pixel, they can be regarded as polarisation insensitive detectors, and any polarisation sensitivity can be included in the Mueller matrix of the optical components between the incoming light and the detector. Thus, only the top row of the end-to-end Mueller matrix of the instrument is relevant, and can be regarded as a polarisation sensitivity vector (equ. 4-4):


image

SCIAMACHY has multiple viewing geometries, selected by appropriate configuration of the scanner unit containing mirrors and diffusers. After the scanner, the OBM is identical for Limb and Nadir measurements. The end-to-end Mueller matrix of the instrument can be split up in a Mueller matrix for the scanner, which is configuration dependent, and a fixed polarisation sensitivity vector of the OBM. This yields (equ. 4-5):


image


where on the left hand side of the equation a scalar describes the light as detected by the instrument, and on the right hand side we have the polarisation sensitivity vector imageof the OBM, and the Mueller matrix imagedefining the response of the scanner to the incoming light represented by another Stokes vector. The first element of this Stokes vector, I, denotes the total intensity of the light. Q is a measure for the polarisation along the x- or y-axis of a chosen reference frame and can be described as image. U is a measure for the polarisation along the 45¿ direction and is defined as image, and V is the circular polarisation component of the incoming light, which is negligibly small for atmospheric light. Note that the total intensity can be written as imageor image. Often Q and U are normalised to the total intensity I. We will denote normalised fractions with q and u.The polarisation reference frame used in the calibration is defined w.r.t. the direction of the slit, at the plane of the slit: looking in the direction of the light entering the instrument after the scan mirrors, the +45¿ polarisation (u=1) direction is obtained by a 45¿ clockwise rotation from the long side of the entrance slit of SCIAMACHY (q=1). All calibration data use this reference frame.

All imageMueller matrix elements are dependent on wavelength and on the incidence angle of the light on the scan mirror(s) or diffuser(s). In the calibration, ambient measurements on component level and instrument TV measurements have to be combined meaning that the actual instrument matrix has to be calculated by a multiplication of the matrix for the scanner (combination) and the OBM. Note that though the end-to-end circular polarisation sensitivity of the instrument is irrelevant when V is zero, the scanner may through its Mueller matrix introduce a circular polarisation component to the light, which requires the circular polarisation sensitivity of the OBM to be known. Though originally not foreseen for on-ground calibration, an update of the calibration concept introduced the circular polarisation sensitivity into the calibration equations. Defining the relative polarisation sensitivity vector of the OBM as:

image

the response of the instrument to polarised light becomes:

image

Combination of the polarisation response of the scanner and the OBM using

image

and making sure to normalise imageso that the first element image remains unity, gives the equation describing the instrument response as expressed in total intensity and fractional linear polarisation (equ. 4-6)

image

where image is the radiometric sensitivity of the instrument (with implicit dependence on wavelength for the main science channels, and determined as well for all 7 PMDs). The term in brackets is the inverse of the polarisation correction factor, c pol. It depends on the polarisation sensitivity of the instrument and the polarisation of the incoming light q and u.The instrument polarisation correction factor is thus with the definitions given above (equ. 4-7)

 

 


image

The problem of correcting the response of the instrument for polarisation can thus be divided into two parts: (1) determining the polarisation sensitivity of the instrument and (2) determining the polarisation of the incoming light during the science measurements in-flight.

The instrument response was measured on instrument level under TV conditions while the mirror and the mirror/diffuser combination were measured under ambient conditions. The Fresnel equations for reflection off a dielectric medium are used to describe the scanner withthe index of refraction determined from the ambient measurements, and hence used to calculate the Mueller matrix elements of the scanner imagefor arbitrary angles used during TV measurements or in flight. While the OBM polarisation sensitivities image (fig. 4-4) were derived during the on-ground calibration under thermal vacuum, from measurements with fully linearly polarised light at a range of polarisation angles.

 

click to enlarge

fig. 4-4:

image
q (blue) and u (red) sensitivity from equ. 4-6 for nadir (elevation angle of 61° top) and for limb (elevation angle of 11.4° and azimuth angle of 39°, bottom) for channels 1-5. Note that these sen-sitivities are multiplied with the polarisation fractions to get cpol and the correction will thus be smaller than displayed for lower polarisation. (Graphics: SRON)
 


The second step, the determination of the polarisation of the incoming light is done by determining the ratio of the signal in the PMD channels – which is fully polarised due to the Brewster reflection at the pre-disperser prism (see chapter 3.2) – and the corresponding signal in the science channel for each individual measurement. During calibration this ratio was determined for different combinations of u- and q-polarised light. The comparison of the in-flight ratio with the calibration data gives 7 polarisation values for the whole spectrum, one for each PMD channel.

 


Polarisation values q are calculated from PMD A-F needing the corresponding value of u. The ratio u/q, which depends only on the polarisation angle, is assumed to be constant, such that (equ. 4-8)

 


image

 


In the original calibration concept, for UV-VIS wavelengths below 600 nm the polarisation angle from single scattering theory was planned to be used (see below), whereas for higher wavelengths the ratio u/q from PMD D and PMD 45°, both centred around 850 nm, had to be taken. The values of q and u here are derived by iteration, until the u needed to calculate q from PMD D and the q needed to calculate u from PMD 45°, match. In-flight it was noted that PMD 45° delivers systematically signals which are 10-15% higher than expected, even for unpolarised sources such as the sun. As there are indications that this PMD suffers from stray light, it remains currently unused. Instead, u/q is taken from single scattering theory for the complete wavelength range. From POLDER satellite measurements of u/q this appears to be a sufficiently accurate assumption (Schutgens et al. 2004). Note, that for small values of q, this ratio becomes very large, thereby amplifying small measurement errors on qmeasured into large measurement errors on umeasured. However, since the instrument is much more sensitive to q than to u, this has little impact on the radiometric calibration, even though u is notably unreliable. In limb, the calculated polarisation displayed an unexpected drift with increasing tangent height, which is probably due to increasing significance of the spatial stray light contribution as the limb intensity decreases. Therefore the PMD measurements are only used up to 30 km. Radiative transfer calculations show that above this height the depolarisation remains constant (Mc Linden et al. 2002), such that for higher limb tangent heights we are able to scale the measured polarisation at 30 km with a value obtained from single scattering theory.

The PMD channels cover only the instrument channels 2-8. For channel 1 the backscattered radiation is dominated by single scattering as can be inferred from radiative transfer calculations. Similarly to the GOME instrument, a theoretical value based on single scattering geometry is used here (Slijkhuis 2000b, Tanzi 1999, Tilstra et al. 2003). The transition region from single scattering to multiple scattering and/or ground reflection in the region between approximately 300-325 nm requires special attention. For GOME, a parameterisation of the degree of polarisation as a function of wavelength was derived by R. Spurr (Balzer et al. 1996), known as the ‘general distribution function’ (GDF) for polarisation. The GDF is characterised by the single scattering value plus three parameters. These parameters are currently obtained using a simplified version of the algorithm from Schutgens and Stammes (2002) where the dependence on scene albedo and ozone content is neglected. More polarisation information may be derived from the channel overlaps of channels 1-6 (five polarisation points) where the different polarisation sensitivities of each channel leads to two independent measurements for the two variables q and u. However, due to calibration inconsistencies, these polarisation points are currently not reliable.

The polarisation values q and u on the level 1b products are specified in an ‘atmospheric’ coordinate frame which is different from the coordinate frame used for the on-ground calibration and Key Data specification. The ‘atmospheric’ coordinate frame is related to the geometry of the scattering of light in the atmosphere. The choice has been to define q as parallel to the local meridian plane – the plane through satellite, zenith, and centre-of-FoV (where its Z axis points in the travel direction of light, i.e. towards the instrument). This plane is depicted in Figure 5-5 for nadir viewing geometry. For limb viewing geometry, the polarisation plane in the figure is rotated 90° as the line-of-sight is approximately in the flight direction. (fig. 4-5)

 

Table of Contents

 

0 Attachments
2653 Views
Average (0 Votes)
The average rating is 0.0 stars out of 5.
Comments