radiometer

1
Calibration of Multi-Satellite
Observations for Climate Studies Microwave
Sounding Unit Dr. Norman Grody
(NOAA/NESDIS) norman.grody_at_noaa.gov
Dr. Konstantin Vinnokov (University of
Maryland, College Park) kostya_at_atmos.umd.edu
Co-investigators Mitch Goldberg , Jerry
Sullivan, Dan Tarpley
2

• Overview of Presentation
• Inter-satellite calibration of MSU N. Grody
• A new a posteriori calibration adjustment
  procedure was developed that accounts for
  instrumental errors. The calibration adjustments
  are obtained by minimizing the differences
  between overlapping satellite observations,
  grouped into two latitude bands.
• Climate analysis of MSU data K. Vinnikov
• (a) For the first time, the latitudinal
  distribution of the MSU-channel 2
• trend is obtained for the period
  1978-2004.
• (b) For the first time, comparisons are made
  with the observed surface
• trend, and GFDL climate model
  simulations.

3
Evolution of Passive Microwave Sensors in the
United States
NPOESS
CMIS
TMI
Nimbus-5
Nimbus-6
Nimbus-7
AQUA
NASA
SCAMS
NEMS
SMMR
TRMM
AMSR
AMTS
ESMR-1
ESMR-2
SSM/T
DMSP F-4, 6, 7, 8, 10, 11, 12, 13, 14, 15
SSM/I
DMSP F-8, 10, 11, 13, 14, 15
DoD
SSM/T2
DMSP F-11, 12, 14, 15
SSMI/S
MSU
TIROS-N , NOAA-6, 7, 8, 9, 10, 11, 12, 14
NOAA
AMSU
NOAA-15, 16, 17
2005
1970
1975
1980
1985
1990
1995
2000
Contains Temperature Sounding Channels
4
The Microwave Sounding Unit (MSU)
MSU was the first operational microwave sensor
developed by NOAA. It has four channels within
the 50-60 GHz oxygen region to profile the
atmospheric temperature from the surface to the
lower stratosphere.
Important characteristics of the MSU instruments
Besides the cloud penetrating property of
microwaves, which lead to its development by
NOAA, of critical importance for climate
applications is the long-term stability and
precise spectral resolution of microwave
radiometers which has been duplicated by each MSU
flown in space.


5
Orbit
Track
6
Tu
tnSecqesTs
tnSecq(1- es )Td
Td
q
Earth Surface Emissivity es
7
MSU channels, frequencies and nadir -
transmittance
Channel Freq. t
(GHz)
4 57.95 0.000 3 54.95
0.001 2 53.74 0.100 1
50.30 0.700
(Ch-4)
(Ch-3)
Tropospheric Temperature
(Ch-2)
Nadir-viewing channel-2 measurements are used to
eliminate the need for any angular corrections.

(Ch-1)
W(pn,q)
8
Block diagram showing the antenna and receiver
systems. Errors in the earth radiation
measurement, Tb, results from errors in the cold
space, Tc , and warm target, Tw , measurements
and uncertainties in the nonlinear parameter, ? .
9
Uncorrected NOAA-11 NOAA-12 MSU Ch-2 globally
averaged measurements
Tb
t
Difference between MSU measurements is 0.0 - 0.4
K mainly due to calibration errors. These
errors must be corrected in order to measure the
global climatic trend, which is about 0.2
K/Decade.
10
Uncorrected NOAA-11, -12 and -14 MSU
instruments.
Calibration corrected NOAA-11, -12 and -14 MSU
instruments.
11
MSU - Calibration adjustments for errors in Tc ,
Tw and m
Warm Target Error,
Cold-Space Error,
Nonlinearity Error.
Bias Correction
Measured
Adjusted
Calibration Correction Equation
Cold-space Warm target Errors
Nonlinear Target Errors
12
NOAA-14 MSU channel 2 calibration factors, K and
Z plotted against one another using global data.
Each data point is the globally averaged
measurements over a 10o latitude band. Also
shown is the straight line fit to the data given
by K a bZ.
13
This figure shows all 12 overlaps for the 9- MSU
satellites. Note that NOAA-9 is a critical
satellite since it interconnects the later MSUs
through NOAA-10.
14
Orbital drift is most pronounced for afternoon
satellites NOAA - 7, 9, 11, 14.


Diurnal temperature variations of overlapping
satellite measurements is reduced by averaging
ascending descending measurements. This removes
all even harmonics.
15
Error-free
Measurements
Measurements
Averaged ascending and descending measurements
results in
so that
zonally pentad averaged measurements for the
overlap between satellites 1 2. (L 30o N
- 30o S , H 30o 85o N S )
where
Consider 1 of the 2 satellites perfectly
calibrated,
so that dT1 dU1 0 and
For this simple case we obtain 2 equations for
the 2 unknowns dT2 , dU2
However, we dont know which, if any, of the 9 -
MSUs is perfectly calibrated. Therefore all dTs
, dUs are obtained using the 12 overlapping
satellite measurements. The over-determined
system of 12x2 equations is solved for the 17
parameters.
16
Satellite, S (K) RMSE (K) (x 10-4 , K-1) RMSE(x 10-4 , K-1)
TIROS-N 0.14 0.06 -0.35 0.09
NOAA-6 0.09 0.05 -0.07 0.06
NOAA-7 0.09 0.05 -0.45 0.05
NOAA-8 -0.07 0.05 -0.40 0.07
NOAA-9 -0.40 0.04 -1.21 0.06
NOAA-10 0 0 -0.53 0.07
NOAA-11 -0.46 0.03 -0.94 0.07
NOAA-12 0.30 0.04 -0.18 0.08
NOAA-14 0.06 0.04 -0.77 0.09
Calibration parameters, dTS , dUS obtained
using 12 overlapping satellites. All offsets,
dTS , are referenced to the NOAA-10 instrument.
17
T-N
N-6
N-7
N-8
N-9
N-10
N-11
N-12
N-14
18
Effects of calibration and
diurnal adjustments on global trend Correction
Equation Calibration Correction
Diurnal Adjustment Global Trend

?T 1st
2nd Harmonics 0.26 K/Decade

?T 1st Harmonic
0.22 K/Decade
?T and ?U
1st Harmonic 0.17 K/Decade




Grody, N., Vinnikov, K., Goldberg, M.,
Sullivan, J. and Tarpley, D., J. Geophys. Res.,
in press (2004).
Vinnikov, K and N. Grody (2003), Science,
269-272.
Effects of reference instrument and
Number of overlapping satellites on global trend
Number of Overlaps Reference MSU
Global Trend .
12 (8) Tiros-N
0.17 (0.16) K/Decade
12 (8) NOAA-6
0.17 (0.14) K/Decade 12 (8)
NOAA-7 0.19 (0.17) K/Decade
12 (8) NOAA-8
0.18 (0.17) K/Decade 12
(8) NOAA-9
0.25 (0.20) K/Decade 12 (8)
NOAA-10 0.20 (0.17) K/Decade
12 (8) NOAA-11 0.24
(0.19) K/Decade 12 (8)
NOAA-12 0.18 (0.15)
K/Decade 12 (8)
NOAA-14 0.22 (0.17) K/Decade

19
Climate analysis of MSU data
1. Determine the latitudinal distribution of the
MSU-channel 2 trend for the period
1978-2004. 2. Compare the observed surface
trend, and GFDL climate model simulations.
3. Significance of the new findings on the
latitudinal distribution of climate trend.

20
Effects of Intersatellite Calibration and Diurnal
Adjustment on Global Trend. MSU Channel 2
Brightness Temperature
 
 
21
Observed and Modeled Time Series of Globally
Averaged MSU Ch. 2 Brightness Temperatures
MODELED
OBSERVED
Trend0.17 K/10 yr
22
(No Transcript)
23
(No Transcript)
24
MSU Ch.2 OBSERVED AVERAGES AND TRENDS
MSU Ch.2 MODELED AVERAGES AND TRENDS
25
AVERAGES AND TRENDS in 3 RUNS of GFDL R-30 MODEL
MSU Ch.2 Brightness Temperature
26
Observed (Surface MSU Ch. 2) and Modeled
(GFDL) Air Temperature Trends
27
Observed and modeled differences in the 1978-2004
zonal air temperature trends at the surface and
in the troposphere (MSU Ch.2).GFDL R30 climate
model ensemble run is forced by CO2Aerosol.
Error bars for difference of modeled trends are
estimated from 900 year control run of the model.
Trend(TSurface) - Trend(TMSU2)
28
Empirical bias adjustment detrends the time
series Tadj(t) Tobs (t) a b
Tw(t) where a and b are estimated by
regression analysis, i.e., Tobs (t) a b
Tw(t) so that adjusted detrended
29
Conclusions

• To measure climatic change, the a posteriori
  calibration of multi-satellite observations must
  be done using physically-based corrections.
• There is no contradiction between the satellite
  and surface observed temperature trends.
• There is no contradiction between the observed
  and model simulated temperature trends.
• Observed decreasing atmospheric stability in the
  tropics and increasing stability at higher
  latitudes was anticipated long ago due to global
  warming

30
SUPPORTING MATERIAL
31
Orbital drift produces a trend-like variation of
TWarm Target
NOAA-14 MSU channel 2 brightness temperature, Tb,
Dicke reference temperature, TDicke and warm
target temperature, TWarm Target during the
lifetime of the instrument. Data are averaged
globally (85oN? 85oS) for low latitudes (30oN?
30oS) and high latitudes (remaining part of the
globe).
32
Instrument Body Effect
Christy found that MSU measurements from
different satellites varied slightly with target
temperature and adjusted the measurements Tb
empirically.
a, b constant, Tw warm load temperature,
To fixed reference
Warm Target, Cold-Space, Nonlineartity
Corrections (DTW , DTC , m)
Grody found that MSU measurements from different
satellites also varied slightly with latitude and
adjusted the measurements Tb using a physical
correction.
where
Measurement
Bias
33

Errors Due to Empirical Equation
Variations in Tw can artificially suppress the
climatic temperature trend data due to errors
in the empirical model as well as the derived ?
parameters. RESULTS Globally averaged
trend 0.26 K/10 yr without any target
temperature corrections. Trend decreases to 0.13
K/10 yr (Mears et al. 2003) and 0.05 K/10 yr
(Christy et al. 2003) after applying the target
temperature adjustment. The smaller trend by
Christy et al. (2003) is due to the larger ?
parameter derived for the NOAA-9 satellite.
CONCLUSIONS 1. The accuracy of the
climatic trend depends on the adjustment
parameters. 2. We believe that the trend also
depends on the statistical method used to derive
the adjustment parameters as well as the
calibration adjustment model itself.
34
From JGR paper by Tsan Mo et al., 2001.
(Note Q - m Z)
35

Averaged brightness temperatures for NOAA-12
(1991-1999) and NOAA-14 (1995-2003) instruments
for two latitude regions (Top). Also shown are
the corresponding Z-factors (Bottom).
36
MSU Satellites
This figure only shows 8 of the possible overlaps
for the 9 MSU satellites.

Note that NOAA-9 has the fewest number of
overlapping pentads (18), which interconnects the
later series of MSUs through NOAA-10 satellite.

This figure only shows 8 of the possible overlaps
for the 9 MSU satellites.

Note that NOAA-9 has the fewest number of
overlapping pentads (18), which interconnects the
later series of MSUs through NOAA-10 satellite.

37
Calibration parameters dT and dU are
determined using overlapping satellite measurement
s averaged lt gt over two latitude bands, L, and
extended time periods comprising many pentads, tS

Averaged measurement
Averaged error-free measurement
Averaged Z-factor
Consider two overlapping satellites (S1, S2) and
neglect any differences in the error-free
measurements when both are averaged over the same
time period, tS1-S2

where
and
38
Calibration Adjustment Algorithm
Consider three overlapping satellite measurements
(S1, S2, S3) and two latitude bands (L l, h),
and express the 6 equations in matrix form, viz.,



where
, etc.,
, etc.,
and
39
However, ?TS1?S2 ?TS2?S3 ?TS1?S3 so that
the three elements are dependent. The element
?TS1?S3 is therefore removed from the equation
so that
The equation is re-written using matrix notation
The solution then becomes
where
40
Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration Un-Adjusted Calibration Adjusted Calibration
Satellite - s - Satellite - k - Overlap Pentads High Latitude (K) Low Latitude (K) High - Low Differences (K) High Latitude (K) Low Latitude (K) High - Low Differences (K)

N-6 T-N 41 -0.24 -0.17 -0.07 0.00 0.00 0.00
N-7 N-6 99 0.26 0.15 0.11 0.00 0.00 0.01
N-8 N-7 79 -0.17 -0.17 -0.00 0.00 -0.01 0.01
N-9 N-6 73 0.29 0.01 0.28 0.00 0.01 -0.01
N-9 N-7 6 -0.05 -0.24 0.20 0.00 0.01 -0.01
N-9 N-8 13 0.12 -0.06 0.19 0.00 -0.01 0.01
N-10 N-9 18 -0.03 0.16 -0.19 0.00 0.00 0.00
N-11 N-10 213 -0.23 -0.34 0.11 0.01 0.01 0.01
N-12 N-10 19 0.01 0.11 -0.09 0.02 -0.00 -0.02
N-12 N-11 316 0.18 0.39 -0.21 0.02 0.03 -0.02
N-14 N-11 55 0.30 0.33 -0.03 0.01 -0.02 0.01
N-14 N-12 280 0.19 0.05 0.14 0.01 0.02 -0.01

minus
minus
Difference between satellites measurements (s,
k) for the high,
, and low,
latitude belts before and
-
as well as
after applying the calibration adjustments. Also
shown are the corresponding differences
the number of overlapping pentads between
different satellites, s and k.
41
Air Temperature Trends Estimates for Three GFDL
R-30 Climate Model Runs Forced by CO2 Aerosol
42
Observed Inter-satellite Differences in
Brightness Temperature
k
?Ts-k,hTs,h-Tk,h High Latitudes 30º-85ºN,S
s
?Ts-k,lTs,l-Tk,l Low Latitudes 30ºS-30ºN
?Ts-k,h-l ?Ts-k,h- ?ts-k,l High Low Latitudes
43
Inter-satellite Differences in Brightness
Temperature after Nonlinear Adjustment
k
?Ts-k,hTs,h-Tk,h High Latitudes 30º-85ºN,S
s
?Ts-k,lTs,l-Tk,l Low Latitudes 30ºS-30ºN
?Ts-k,h-l ?Ts-k,h- ?ts-k,l High Low Latitudes
44
Observed Trends, Surface Temperature
Modeled Averages and Trends, Surface Temperature
45
(Surface Air Temperature )-(MSU Ch.2 Brightness
Temperature)
Difference of Averages, K
Difference of Trends, K/10 yr
46
AVERAGES AND TRENDS in 3 RUNS of GFDL R-30 MODEL
Surface Air Temperature
47
(No Transcript)
48
(No Transcript)