level 1.5 processing updates martin bates, ral gist 23 28 th april, 2005
DESCRIPTION
Gains LFLF V Space view Black body view 1 / G TOT TOTAL Instrument continually calibrated in-flight using views of (warm) internal Black Body and (cold) space Calibration can go wrong if bright objects encroach on space view Stray light and the moon are two sources of contaminationTRANSCRIPT
Level 1.5 Processing Updates
Martin Bates, RALGIST 23
28th April, 2005
Contents
• Gain calculation• Geolocation• ARG Duplication
GainsLF
VSpace view
Black body view
1 / GTOT
TOTAL Instrument continually calibrated in-flight using views of (warm) internal Black Body
and (cold) space
Calibration can go wrong if bright objects encroach
on space view
Stray light and the moon are two sources of
contamination
Gain Calculations
• Currently flag affected data– stray light flagging implemented as part of new
product confidence flags (11-Jan-2005)– but moon not flagged
• Original suggestion was offline gains– gains are stable– but problem of keeping up with changes
• Two other ideas– split gains– running averages
Split Gains• What are split gains ?
– Calculate average response (DS-BB) per pixel in western and eastern halves of image
– if either half exceeds the other by more than a set amount, remove that half from gain calculation
(DS – BB)west > (DS – BB)east
Use only eastern pixels for gain calculation
TOTAL Response (All pixels)
Stray light bar
Stray light in blackbody
Lunar Data
Lunar Data
without sd cut
with sd cut
Typical response ~ -2200 to -2600
0.5 % => need to be sensitive to change of ~ 10 counts
Centring Problem
• Change in MSG rotation rate Jan 2004.– GERB looks further west– Not enough space pixels to east at
equator (for split gains)• Corrected itself following 08 Feb
2005 manoeuvre
26 Oct 2004
E-W Centring Spin Period
Split Gains Summary
• Cannot be used on 2004 data• Monitor centring more tightly to ensure
feasability of use for future data• Need to cut at ~ 10 counts
– level of background this will produce would need to be measured
Running Averages• Maintain a running
average gain for each pixel.
• Update running average with latest measurement
• Weight of any given measurement drops away over 24-48 hours
• No updates ~ midnight (check stray light flags)
• Lunar effects washed out• Similar for BB SW Meas
Radiance.
Gain = ra_factor * Inst_Gain
+ (1- ra_factor) * Ave_Gain
ra_factor = 0.008
TOT Gain 05-06 Feb 2005 Pix 200
TOT Gain 05-06 Feb 2005 Pix 200
TOT Gain 05-06 Feb 2005 Ave Pix
TOT Gain 01 Jan 2005 Pix 24
Summary• Advantages of running gains:
– successfully handles stray light and lunar data– uses more measurements
• takes advantage of stability of gains• gain measurement less noisy (particularly central pixels)• (split gains would actually use fewer measurements)
– still tracks longer term (> few hours) drifts in gains– should work on 2004 data and nrt data
• Disadvantages ?– Time dependence to processing ?
• Stability of gains means order of processing not critical– Start-up conditions ?
• Write gains to file after each scan processed – file can be read if processing stopped and started
• Revised weighting scheme for startup until a stable running average is derived• Should only ever come into play once
• Proposal:– Switch on running averages for current level 1.5 processing.– Switching to split gains remains an option if suitable cuts can be devised.
Stray Light in Black Body
• Occurs ~ midday near equinoxes
• Problem because each EV pixel reading needs corresp. BB reading subtracted: (VEV – VBB)
• Could use a known gain and deep space reading:
• LBB = (VDS – VBB) / Gain
• but would need VBB to be stable over a scan (is it ?)
• Use VIS ?• Otherwise, just flag as now
Sep 27
Jul 05
Geolocation
• Last time:– Comparisons between GGSPS L1.5 GEO and RGP level 2 GEO
(nrt and 90 min)– time dependent, unphysical corrections to optical model, variations
at 0.1 – 0.2 level– noise in level 1.5 column spacings (SOL_SOE_Delay)
• Plan:– devise a “good” optical model correction– implement level 1.5 GEO in level 2 processing for a pre-defined
test period– still to be done – however, new problems have arisen in Jan/Feb 2005.
E/W Shift (Spin rate)
Spin rate change associated with 08 Feb manoeuvre
Effect of spin rate changes
MSG SLOWS DOWN
MSG SPEEDS UP
A known consequence of
de-spin mechanism
But predicted magnitude out by
~1 pixel in 7.
N/S Shift (??)
N/S Shift commencing 15 Jan 2005
N/S Shift Jan 2005
Coincident with increase in spin axis tilt
Also follows SEVIRI
decontamination
Geolocation
• Instantaneously, we have had a 1 pixel shift E/W in level 1.5 geolocation (manoeuvre).
• Over a 2 week period, we have had a 1 pixel shift N/S in level 1.5 geolocation– not associated with any manoeuvre– coincident, if not connected, with increase in spin axis
tilt (and end of SEVIRI decontamination period)– may be connected with changing axis misalignments ?
Effects of Axis Misalignments
Effect of axis misalignments on determination of optical model
Getting knowledge of axis misalignments now a high priority.
No misalignment
0.3 degrees (radial
rotation)
0.3 degrees (transverse rotation)
L15G 90 min fits
20050208 20050227
20050115 20050207
ARG Duplication
Renaming
Duplication
G2_SEV1_L20S_20050101_145013_V002.hdf
G2_SEV1_L20S_20050101_150711_V002.hdf
G2_SEV1_L20S_20050101_144500_V002.hdf
G2_SEV1_L20S_20050101_151500_V002.hdf
G2_SEV1_L20S_20050101_150000_V002.hdf
15
17
SEVIRI
GERB
ARG Duplication
• Having second thoughts about this scheme:– Current status (renaming but no duplication)
• looks as though there is missing data• filename misleading as to data contents
– ARG Duplication• seemingly independent files actually contain the same data• filename still misleading as to data contents
• How feasible/desirable is it to abandon scheme and name ARGs according to data content ?– remember designed for level 1.5 ARG, but main product affected
is level 2 ARG.– (technical): can RGP avoid renaming to a precise 15 minute slot ?