Title: Microwave AMSU Calibration Update Bjorn Lambrigtsen Frank Sun Thomas Hearty
1Microwave (AMSU) Calibration UpdateBjorn
LambrigtsenFrank SunThomas Hearty
2Topics
- Radiometric calibration upgrade
- Moon in cold-cal FOV
- Pointing validation
- Radiometric validation
3Radiometric Calibration Upgrade
- The issue
- Cold-cal Tbc is assumed invariant
- Sum of cosmic background and Earth radiation into
sidelobes - Baseline approach (from NOAA) use climatology to
pre-compute Tbc(Earth) - In reality Tbc (Earth) varies
- Strong function of latitude moderate func. of
lon weak func. of season - The effect The computed cold-cal Tbc is
erroneous - Prominent orbital cycle gt Varying calibration
error - The solution
- Take into account variability of Tbc(Earth)
- Time/location-dependent climatology from prior
observations - Tables are in place and can be populated for V4.0
- Will improve absolute radiometric accuracy by up
to 1 K
4Cold-cal observations Ch. 1-4
45 ltgt 1.35 K
Ch. 1 1.35 K
Ch. 3 0.3 K
Ch. 4 0.6 K
Ch. 2 0.5 K
5Cold-cal observations Ch. 5-8
Ch. 5 0.8 K
Ch. 7 1.9 K
Ch. 8 0.3 K
Ch. 6 0.7 K
6Cold-cal observations Ch. 9-12
Ch. 9 0.4 K
Ch. 11 1.0 K
Ch. 12 0.8 K
Ch. 10 0.3 K
7Cold-cal observations Ch. 13-15
Ch. 13 0.9 K
Ch. 15 0.4 K
- Will derive lat/lon climatology
- Based on one 16-day cycle
- Populate existing tables
- Expect residual errors to be small
- lt 0.2 K
- Should improve retrievals
- May help ease sidelobe analysis
Ch. 14 0.8 K
8Moon in Cold-Cal FOV
- The issue
- Moon gets into cold-cal FOV several times a year
- Can cause 3-4 K effect in AMSU (up to 20 K in
HSB) - Baseline approach
- Reject cold-cal if moon-in-FOV is predicted
(computed moon angle lt threshold) - Then use last previously computed cal.
coefficients - But not across granule boundaries
- The result
- If moon is in FOV, it may happen for several
minutes - Because AMSU cold-cal view is snapshot in one
single direction - Large gaps in calibration therefore occur
periodically - This can span across granules - in that case,
large data gaps can occur - The solution
- Account for moons radiometric effect
- Estimate ?Tb from predicted moon angle - add to
Tbc - Continue updating cal. coefficients gt no
calibration gaps gt no data gaps - Can be implemented for V4.0
9Moon observations Ch. 1-4
Ch. 1 3.5 K
Ch. 3 2.6 K
Ch. 4 2.5 K
Ch. 2 2.9 K
10Moon observations Ch. 5-8
Ch. 5 2.6 K
Ch. 7 4.4 K
Ch. 8 2.6 K
Ch. 6 3.0 K
11Moon observations Ch. 9-12
Ch. 9 3.1 K
Ch. 11 3.5 K
Ch. 12 3.5 K
Ch. 10 2.9 K
12Moon observations Ch. 13-15
Ch. 13 3.6 K
Ch. 15 2.9 K
- Results are close to expectations
- Tb 200 25cos(?-40) _at_ 30 GHz
- ?(Aqua) 270 (1/2-waxing) 8 (inclin)
- Tb(moon/Aqua) 185 K _at_ 30 GHz
- ?(moon) 0.52 5
- Use simple functional approximation
- Error lt 1K
- Use flag to alert data users
- Can be improved with better pointing knowledge
Ch. 14 3.7 K
13AMSU Pointing Analysis Method
- Method as described in 2002
- Instrument in nadir stare mode
- Analyze coastal crossings
- Perpendicular gt Pitch error
- Oblique gt Roll error
- Determine obs-calc time lag
- Obs from Tb
- Calc from landfrac
- ?t t(calc)-t(obs)
- ?pitch ?t(6.65 km/sec)/(705 km)
- Works well for window channels
- Results shown here for pitch analysis
- Several dozen crossings
- 5 quasi window channels
- Moon analysis will also be used
- Results to be presented later
Example HSB perpendicular crossing (sampled every
0.02 sec for 1.7 sec, followed by 1-sec gap)
14AMSU Pointing Analysis Results
Summary for 11 selected perpendicular crossings
Example AMSU perpendicular crossing (sampled
every 0.2 sec for 6 sec, followed by 2-sec gap)
15AMSU Pointing Analysis Summary
- Pitch errors are within requirements (10 of
FOV) - Can be largely corrected for in L1A geolocation
processing - Relative tightly clustered
- Small sample, large scatter analysis statistics
can be improved - Roll analysis to be reported on later
- Yaw analysis must use different method using
full-scan data
16AMSU Pointing Analysis Using Moon
- Branches should be equal
- Approach Recede
- At least near closest approach
- Difference due to pointing error
- MoonAng from assumed boresight centroid
- Actual boresight differs
- Complex geometry
- Lunar path not symmetric
- Varying mix of pitch/roll/yaw
- Precise analysis needs work
- Example shown
- Offset 0.35
- Implies pointing error of 0.175
- Combined pitch/roll/yaw
Single lunar encounter - Ch. 1
17Lunar Pointing Analysis Example
Ch. 4
Ch. 15
Small pointing error indicated for ch. 4
Large pointing error indicated for ch. 15
18AMSU Moon Analysis Extra
- Curious behavior of Ch. 14
- Should be identical
- Shared front end
- Ch. 9-13 look very similar
Ch. 10
Ch. 14
19AMSU Radiometric Validation
- Objectives
- Absolute radiometric validation
- Definitive sidelobe analysis
- So far
- Assess effect of clouds
- Assess effect of wind
- Comparison with ECMWF
- Analysis is on-going
- Comparison with radiosondes
20AMSU Obs-Calc vs. Wind speed
Ch. 1
Ch. 2
21AMSU Obs-Calc vs. Wind speed
Ch. 3
Ch. 15