GOME2 Polarisation Study First results

1
GOME-2 Polarisation StudyFirst results

• L.G. Tilstra (1,2), I. Aben (1), P. Stammes (2)
• (1)SRON (2)KNMI

EUMETSAT, Darmstadt, 29-06-2007
2
Overview

• Q/I versus (Q/I)ss
• verification forward scan versus backward scan
  pixels
• verification PMD readouts versus 187.5 ms
  subpixels
• special points where Q/I Pcos(2?) 0 owing to
• - cos(2?ss) 0
• - backscatter geometry

3
1 Q/I versus (Q/I)ss

• GOME-2 polarisation measurements 256 PMD
  readouts per scan (in 15 spectral bands). A scan
  lasts exactly 6 seconds, of which 4.5 seconds in
  forward scan, and the remaining 1.5 seconds in
  backward scan. Data one level-1B orbit (actually
  consisting of 3-min. chuncks of data) of
  13-APR-2007.
• Total number of NADIR scans 593 scans 151808
  polarisation values per orbit.
• IDEA 0 Q/I (Q/I)ss
• Geometry in the data product is given only 32
  times per scan (for the 187.5 ms subpixels).
• Verification checked (Q/I)ss on the 32-per-scan
  grid in the GOME-2 product with our own
  calculations
• Interpolated viewing and solar angles to the
  256-readouts-per-scan grid
• Calculated (Q/I)ss on the 256-readouts-per-scan
  grid

sheet 1
4
BAND 1 311 nm

• distribution largely outside physical
  regime(where 0Q/I(Q/I)ss)
• there is an offset problem
• 70 of the data points outside interval -1,1
• these data points ALWAYS have the same value of
  2147.4836
• reported wavelengths same value when wrong
• reported errors ALWAYS equal to 0.065535

unphysical regime
unphysical regime
data from 13-APR-2007
5
BAND 2 314 nm

• offset problem smaller

unphysical regime
unphysical regime
data from 13-APR-2007
6
BAND 3 319 nm
unphysical regime
unphysical regime
data from 13-APR-2007
7
BAND 4 325 nm
unphysical regime
unphysical regime
data from 13-APR-2007
8
BAND 5 332 nm
unphysical regime
unphysical regime
data from 13-APR-2007
9
BAND 6 354 nm
unphysical regime
unphysical regime
data from 13-APR-2007
10
BAND 7 381 nm

• distribution again outside physical regime,
  even for (Q/I)ss far from 0

unphysical regime
unphysical regime
data from 13-APR-2007
11
BAND 8 413 nm

• similar

unphysical regime
unphysical regime
data from 13-APR-2007
12
BAND 9 482 nm

• looks pretty ok

unphysical regime
unphysical regime
data from 13-APR-2007
13
BAND 10 558 nm

• Looks ok not a lot of measurements in the
  unphysical regime, and spread around Q0 is
  smaller than before

unphysical regime
unphysical regime
data from 13-APR-2007
14
BAND 11 621 nm
unphysical regime
unphysical regime
data from 13-APR-2007
15
BAND 12 749 nm

• again somewhat out of physical regime
• outliers rainbow?

unphysical regime
unphysical regime
data from 13-APR-2007
16
BAND 13 761 nm

• offset
• outliers

unphysical regime
unphysical regime
data from 13-APR-2007
17
BAND 14 795 nm

• offset problem
• outliers

unphysical regime
unphysical regime
data from 13-APR-2007
18
BAND 15 842 nm

• offset problems are quite severe
• again a lot of outliers
• BAND 15 was not working at all for older versions
  of the data (06-DEC-2006)

unphysical regime
unphysical regime
data from 13-APR-2007
19
Conclusion Q/I versus (Q/I)ss

• At first sight, the data generally look ok
• Data outside interval -1,1 algorithm gives up
  too soon?
• BAND 01 (311 nm) 69.8 out
• BAND 02 (314 nm) 38.5 out
• BAND 03 (319 nm) 27.7 out
• BAND 04 (325 nm) 21.0 out
• BAND 05 (332 nm) 18.5 out
• BAND 06 (354 nm) 17.2 out
• BAND 07 (381 nm) 16.8 out
• BAND 08 (413 nm) 14.0 out
• BAND 09 (482 nm) 12.6 out
• BAND 10 (558 nm) 12.8 out
• BAND 11 (621 nm) 13.1 out
• BAND 12 (749 nm) 12.8 out
• BAND 13 (761 nm) 13.0 out
• BAND 14 (795 nm) 13.4 out
• BAND 15 (842 nm) 13.6 out
• (and having a value of 2147.4836)

20
2 Verification forward scan versus backward scan
21
Sometimes the forward-scan measurement fails
while the backward-scan measurement is ok
sometimes it is the backward-scan that is at
fault. Why are there so many errors and where do
they come from? (TBD)
22
Binning
non-homogeneous but otherwise normal scene
rather inhomogeneous scene

• Good correlation for mildly inhomogeneous scene
  spread probably also determined by improper
  binning of ratios instead of intensities.
  Inhomogeneous scene the higher the wavelength,
  the worse the correlation.
• In conclusion, no problems related to scan
  direction.

23
3 Verification PMD readouts versus 187.5 ms
subpixels
Binning
Conclusion As far as we can tell, there dont
appear to be any obvious problems in the mapping
of the measured polarisation values of the PMD
readouts (256 scan-1, 23.4 ms) to 187.5 ms
subpixels (32 scan-1).
24
4 Special geometries where Q/I 0

• Situations where cos(2?ss) 0 or ?ss45 or
  135
• many situations are found along virtually the
  entire orbit (because of the large range of
  viewing angles and the small pixel sizes in scan
  direction)
• no physical link with data, selection of points
  determined by choice of reference plane (which is
  the local meridian plane)
• these are special situations where (U/Q)ss is
  undetermined, and the data processor treats these
  situations in a special way (!!)
• Backscatter situations (T 180)
• does not depend on the definition of a
  reference plane
• rainbow and sunglint situations are
  automatically filtered out
• situations are only found around the equator
  (ff0180)
• Dependencies pixel number, VZA, SZA, VAZI, SAZI,
  RAZI, SCAT, CHI_SS
• To be done dependency on PMD-p and PMD-s
  intensity

25
4.1 cos(2?ss) 0 pixel number (index) sheet
2
half-way orbit
symmetrical... geometrical effect?
26
4.1 cos(2?ss) 0 VZA sheet 3
?
Viewing angle dependence? Or just indirect?
noisy stable branch
27
4.1 cos(2?ss) 0 SCATTERING ANGLE sheet 5
stable branch turns into noisy branch
Indirect dependence on VZA, SZA, ?
28
4.2 BACKSCATTERING (T 180) pixel number
(index) sheet 6
Mind the small range in pixel number (compare
with sheet 1) Complete agreement with
cos(2?ss) 0 method over the entire mutual pixel
number range!
29
4.2 T 180 DIRECTION OF POLARISATION
CHI_SS sheet 8

• We find no strange behaviour at the points where
  ?ss45 or ?ss135.
• data having this geometry, where cos(2?ss) 0
  and (U/Q)ss does not exist, behave similar to
  other data, despite alternative treatment of data
  processor.
• therefore, the cos(2?ss) 0 method appears to
  be a reliable tool.

?ss45
?ss135
Clearly, more analyses are needed to sort out the
problems