Recalibration and Reprocessing of MSU/AMSU
Observations for Climate Studies
Cheng-Zhi Zou
NOAA/NESDIS/Office of Research and Applications
15th International TOVS Study Conference, Maratea, Italy, October 4-10, 2006
Acknowledgment
Mitch Goldberg
Zhaohui Cheng
Jerry Sullivan
Changyong Cao
Mei Gao
Dan Tarpley
NESDIS/ORA
QSS Group, Inc.
NESDIS/ORA
NESDIS/ORA
CAMS/CMA
NESDIS/ORA
NOAA MSU/AMSU Satellites
ƒ Each satellite has
a life cycle of a few years
ƒ Each satellite overlaps
With other satellites
ƒ LECT gradually changes
With time–
orbital drift phenomenon
ƒ Merging these satellites
is a non-trivial task
Satellite Local Equator Crossing Time (LECT)
vs time
Past MSU CH 2 Trend Studies
ƒ
Spencer and Christy (1992): 0.02 K Dec-1, 1979-1988
0.25
ƒ
Christy et al. (2003):
0.02 K Dec-1, 1979-2002
0.2
ƒ
Prabhakara et al. (2000):
0.13 K Dec-1, 1980-1999
ƒ
Mears et al. (2003, 2005):
0.10 K Dec-1, 1979-2001
ƒ
Vinnikov and Grody (2003): 0.22-0.26 K Dec-1, 1979-2002
ƒ
Grody et al. (2004)
ƒ
Vinnikov et al. (2006):
0.17 K Dec-1, 1979-2002
0.20 K Dec-1, 1978-2004
0.15
0.1
0.05
0
SC
Mears
VG
sfc
Fake Trend Problems
ƒ Intersatellite Biases:
-- 0.1K biases enough to lead
253
252
251
different conclusion
250
ƒ Warm target temperature
Contamination
Trend:
N10 = - 0.40 K Dec-1, N11 = 0.80 K Dec-1,
N12 = 0.36 K Dec-1, N14 = 0.43 K Dec-1
249
1987
1989
1991
1993
1995
1997
1999
2001
Time series of warm target for NOAA10, 11,12, and 14 (ocean averages)
290
--Require brightness temperature
not a function of warm target
temperature
285
280
275
270
265
Trend:
N10 = 1.01 K Dec-1,
N12 = -7.14 K Dec-1,
ƒ Diurnal cycle effect
-- Convert observations of
different local time to a common
time to remove diurnal cycle effect
260
1987
1989
1991
1993
N11 = 9.49 K Dec-1
N14 = 6.99 K Dec-1
1995
1997
1999
2001
2003
2003
SNO Definition
Method to find SNO matchups:
• Use Cao’s (2004) method
to find the orbits that
have intersections
• Use time and location
information in the 1B
file to determine
simultaneity between
two pixels
Schematic viewing the overpasses
between two NOAA satellites
SNO Locations
SNO Temperature range for CH. 2: 200-250 K
Global temperature range for CH. 2: 200-260 K
SNO Bias Characteristics
NOAA 9 and NOAA 10
•Tb difference gets larger
when the SNO pixel distance
gets larger
•SNO numbers increase with
the distance
•Different satellite pairs have
different SNO numbers
because of different overlap
period
NOAA 10 and NOAA 11
0.5
N7-N6
Mean Bias (K)
0.4
N11-N10
0.3
N12-N10
0.2
N12-N11
0.1
N14-N11
0
N14-N12
-0.1
Zero Line
-0.2
-0.3
-0.4
-0.5
0
20
40
60
80
100
120
140
160
180
200
220
Distance (km)
1.4
1.2
STD(K)
1
Range where sensor
noise dominates
N10 and N11
0.8
N10 and N12
0.6
N11 and N12
N11 and N14
0.4
N12 and N14
0.2
"Average"
0
20
40
60
80
100
120
140
160
180
200
Pixel Distance (km)
STD and biases of channel 2 brightness temperature differences between satellite pairs versus center
distance of the nadir overpass pixels. (a) Biases (b) STD . Linear calibration algorithm at level 0 is
used.
220
MSU In-Orbit Calibration Process
Cold Space
T=2.73K
MSU Sensor
Warm Target
Temperature is
measured by PRT
Earth
Conceptual diagram of MSU observational procedure
Level 0 Calibration Equation
Linear Calibration
RL = Rc + S (Ce − Cc )
Slope
Nonlinear Calibration (Mo 1995)
R = RL − δR + μ Z
Z = S (Ce − Cc )(Ce − Cw )
2
Radiance (R)
S
(Cw, Rw)
(Ce, Re)
μZ
{
(Ce, RL)
(Cc , 2.73K)
Digital Counts (C)
SNO Radiance Error Model
Rk = RL ,k − δRk + μ k Z k
R j = RL , j − δR j + μ j Z j
k
Radiance Error Model for SNO Matchup
K and J :
ΔR = ΔRL − ΔδR + μ k Z k − μ j Z j
Z j = β Zk + α
j
SNO Nonlinear Calibration—
Bias Removal
Linear Calibration
Nonlinear calibration using SNO
Scatter plots showing effects of the nonlinear calibration on the error
statistics and distribution of the brightness temperature differences
between NOAA 10 and NOAA 11.
SNO Sequential Calibration
Procedure
• Assuming NOAA 10 as the reference satellite and using
its pre-launch coefficient for reference
• Obtain calibrated radiance (1b) for NOAA 10
• Compute NOAA 11 coefficients from regressions of
N11-N10 SNO
• Obtain calibrated radiance (1b) for NOAA 11
• Repeat above procedure for NOAA 12 with calibrated NOAA 11
as references
……….
Effect on time series and trend
• Intersatellite biases largely reduced
• Trend values are more reliable
253
254
0
Combined
252
-1
Trend = 0.22 K Decade
253
NESDIS
Operational
Calibration
251
250
-1
-1
-1
-1
252
251
Trend: N10 = - 0.40 K Dec , N11 = 0.80 K Dec ,
249
1987
0.22 K Dec-1
N12 = 0.36 K Dec , N14 = 0.43 K Dec
1989
1991
1993
1995
1997
1999
2001
2003
250
1987
1989
1991
1993
1995
1997
1999
2001
2003
(Vinnikov and Grody, 2003)
253
254
-1
SNO
calibration
252
251
252
250
251
-1
0.17 K Dec-1
-1
Trend: N10 = -0.39 K Dec , N11 = 0.58 K Dec
-1
N12 = 0.43 K Dec ,
249
1987
Trend = 0.17 K Dec
253
1989
1991
1993
-1
N14 = 0.31 K Dec
1995
1997
1999
2001
2003
250
1987
1989
1991
1993
1995
1997
1999
2001
2003
Spatial Distribution of Biases
• Ocean OK
• Land needs diurnal
cycle corrections
Warm Target Contamination
Brown Line: Tb differences between NOAA 10 and 11
Blue Line: Tw of NOAA 10
Pink Line: Tw of NOAA 11
Determine Absolute Calibration by
Removing Tw Contamination
• µN10 small or large, large
warm target temperature
contamination
• When µN10 is 25% larger
than its pre-launch value,
averaged warm target
temperature contamination
reaches a minimum (4%)
• Corresponding trend is
0.198 ± 0.02 K Dec-1
• Degree of freedom about
30, correlation is
significant at 95% when
r2>13%
Summary on SNO Calibration
Post-launch coefficients larger than Pre-launch values
10
9
Non-linear coefficient
•
N11
8
7
6
N14
5
4
N12
N10
3
2
1
0
1987
1989
1991
1993
1995
1997
1999
Year
Dots: Pre-launch values
Line: SNO calibration values.
2001
2003
Spatial Distribution of Trend
ARCTIC SEA ICE EXTENT - SEPTEMBER TREND,
1978-2005
(http://nsidc.org/news/press/20050928_trends_fig1.
html)
Global Ocean Zonal Mean Trend (1987-2003)
Trend (K/Decade)
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-90
-60
-30
0
30
Latitude (Degree)
60
90
On-going Work and Future Plans
• Diurnal cycle corrections
--land problems
• Recalibrate and Reprocess all MSU satellites
--TIROS-N to NOAA 14, for all channels 2,3,4
• Connect MSU with AMSU
•
-- Consider different pixel resolutions and frequencies
Impact on Reanalyses
-- Assimilate SNO calibrated MSU/AMSU 1b to next generation reanalyses
-- affect trend of all reanalysis variables
-- diurnal cycle problem resolved
• Compare with other data
-- Radiosonde (Tony Reale)
-- GPS Radio Occultation (Ben Ho, Bill Kuo)
• Other instruments
-- SSM/I and SSMIS (Weng)
Reference and Website
• Zou, C., M. Goldberg, Z. Cheng, N. Grody, J. Sullivan, C. Cao,
and D. Tarpley (2006): Recalibration of Microwave Sounding
Unit for climate studies using simultaneous nadir overpasses, J.
Geophys. Res, 111, doi: 10.1029/2005JD006798
• http://www.orbit.nesdis.noaa.gov/smcd/emb/mscat/mscatmain.h
tm