A Monte-Carlo Approach to Estimating
Uncertainty in MSU/AMSU Climate Data
Carl Mears and Frank J. Wentz, Remote Sensing Systems
Peter Thorne and Dan Bernie, U.K. Met Office
Merged Climate Data Records for the MSU/AMSU series of microwave sounders are used extensively in
climate research.
Data have been used in global, regional, and even local studies. Both short time scale (e.g. ENSO) and
long time scale (e.g. decadal trends) phenomena have been investigated.
WE NEED TO KNOW ABOUT UNCERTAINTY ON DIFFERENT
SPATIAL AND TEMPORAL SCALES!
Error structure is complicated due to spatial and temporal correlation in the error sources -- too hard to
do analytically -
USE MONTE CARLO!
260
255
NOAA-11
Mean of Hourly Data
NOAA-12
Mean, NOAA-11 Sampling 242.33
“Diurnal” adjustments
for measurement time drifts
ERROR INPUTS
242.88
Mean, NOAA-12 Sampling 242.71
250
10:00
Equator Crossing Time
245
240
235
230
0
5
10
15
20
25
30
These are two different simulations
of the sampling error made using
real satellite sampling (from different years) of hourly output of a
climate model (CCM3)
Days since Jan 1
NOAA-11
08:00
NOAA-06
Different satellite sampling, combined with CCM3 output from
different years are combined to
make 400 realizations of the sampling noise
04:00
NOAA-15
TIROS-N NOAA-07
NOAA-09
NOAA-11
NOAA-14
02:00
1985
1990
TLT
1995
Year
2000
2005
2010
We use the output from climate
models to determine a diurnal cycle
climatology that we use to adjust the
satellite data to solar noon.
0.0
TMT
Here we use the results from two
different climate models (CCM3 and
HADGEM1) to obtain two different
diurnal climatologies.
0.4
0.0
1996
NOAA-12
HADGEM
CCM3
1.0
NOAA-08 NOAA-10
06:00
1980
Global Diurnal Adjustment (K)
Simulated Brightness Temperature (K)
Sampling noise
1998
2000
2002
2004
1996
1998
2000
2002
2004
Weighted combinations of these two
climatologies are used to create the
error realizations.
Year
400 realizations
of estimated error
Errors in Merging Parameters
e.g. Nonlinearity Adjustment
0.02
TMT
σ(α)
0.01
0.00
Histograms of Regional Trends with Uncertainty
Global
mean = 0.153
σ = 0.022
20
10
0
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
Channel
TLT (lower
troposphere)
TMT (middle
troposphere)
(red - sampling only, yellow - diurnal adjustment included)
Estimated Error (K)
NOA
A-15
NOA
A-14
NOA
A-12
NOA
A-11
NOA
A-10
NOA
A-09
Estimates of Trend Errors
Error in TLT Trends (1979-2009)
Global Time Series Errors
TMT
Trend Error (K/decade)
-0.2
TLT
0.0
0.0
0.2
Global (80S to 80N)
Tropics (20S to20N)
0.153 +/- 0.044
0.147 +/- 0.034
0.093 +/- 0.042
0.114 +/- 0.038
Fine Print:
The results of this study consist of a set of 400 realizations of the possible error in for each RSS MSU/AMSU dataset.
Each error realization has the same temporal and spatial resolution of our dataset (144 x 72 x 396) (Longitude x Latitude x Number of Months)
Each error realization can be added to our MSU/ASMU dataset to obtain a realization of the measured data with possible error.
These can (and should) then be used to propogate the uncertainty through whatever subsequent analysis is planned.
We have performed this analysis for our 4 MSU/AMSU products: TLT (lower trop.) TMT (mid trop), TTS (tropopause), TLS (lower strat.)
0.0
-0.2
NOA
A-08
The uncertainty in the input data leads to uncertainty in the merging
parameters (intersatellite offsets and non-linearity adjustments),
which in turn lead to even more uncertainty in the final results. This
additional uncertainty is very likely to be correlated in time and
space.
Trend (K/decade)
-0.2
NOA
A-07
Continental USA
mean = 0.193
σ = 0.093
NOA
A-06
Merging Algorithm
0
0.00
S-N
40
0.01
TIRO
Number of Realizations
80
TLT
0.02
0.4
Note that this error is different (and perhaps in addition to) the errors the have been calculated in climate assessments such as IPCC and CCSP.
In these resports, the errors for MSU/AMSU have been determined by a “goodness of fit to a line” criteria.
References:
Mears, CA, FJ Wentz, 2009, Construction of the RSS V3.2 lower tropospheric dataset from the MSU and AMSU microwave sounders, Journal of
Atmospheric and Oceanic Technology, 26, 1493-1509.
-0.2
1980
1990
Year
2000
2010
Mears, CA, FJ Wentz, 2009, Construction of the Remote Sensing Systems V3.2 atmopsheric temperature records from the MSU and AMSU
microwave sounders, Journal of Atmospheric and Oceanic Technology, 26, 1040-1056.