Statistical separation of natural and anthropogenic signals in observed surface air temperature time...
-
Upload
peregrine-cunningham -
Category
Documents
-
view
217 -
download
0
Transcript of Statistical separation of natural and anthropogenic signals in observed surface air temperature time...
Statistical separation of natural and anthropogenic signals
in observed surface air temperature time series
T. Staeger, J. Grieser and C.-D. Schönwiese
Meteorological Environmental Research / Climatology
Institute for Meteorology and Geophysics J.W. Goethe-University, Frankfurt /M., Germany
1860 1880 1900 1920 1940 1960 1980 2000
tem
pera
ture
ano
mal
ies
in [K
]
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6Global mean temperature 1856 – 2003 after P.D. Jones et al.
Which parts of the variations in observed temperature are assignable to natural and anthropogenic forcings?
Are anthropogenic signals distuingishable from noise?
Approach:
Causes for the structures in the time series under consideration are being postulated.
A pool of potential regressor time series is collected out of the forcings / processes considered.
A selection routine is applied to obtain a multiple linear regression model.
Stepwise Regression
The effects are seen to be linear and additive.
Forcings / processes considered:
- Greenhouse gases (GHG)
- El Niño - Southern Oscillation (SOI)
- Explosive volcanism (VUL)
- Solar forcings (SOL)
- North atlantic oscillation (NAO)
- Tropospheric sulphate aerosol (SUL)
1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
T-A
no
ma
lien
in K
-0,7
-0,6
-0,5
-0,4
-0,3
-0,2
-0,1
0,0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
-0,7
-0,6
-0,5
-0,4
-0,3
-0,2
-0,1
0,0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
GHG + SOL + SOI + VUL
explained variance: 78.9%
global mean temperature 1878 – 2000, annual mean after P.D. Jones
Ges. GHG SUL SOL SOI VUL NAO
proz
ent
uale
erk
lärt
e V
aria
nz
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
explained variance of the complete model and and for single forcings on the global mean temperatur 1878 - 2000
What is noise?
Case 1: noise represents chance:
To obtain the component representing chance, the residual is separated into a structured and unstructered component.
txnoisetxRtxRtxR polytrend ,,,,
The question to be answered here:
Is the greenhouse signal distuingishable from chance?
What is noise?
Case 2: noise comprises of natural variability and unexplained variance
The question to be ansewered here:
Is the greenhouse signal distuingishable from variability of non-anthropogenic origin?
txStxRtxnoise nat ,,,
1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
T-A
no
mal
ie in
K
-0,5
-0,4
-0,3
-0,2
-0,1
0,0
0,1
0,2
0,3
0,4
0,5
0,6GHGSOLVULSOI
99.9%
99.9%
99%
99%
95%
95%
srsch
srsch
Case 1: noise represents chance
1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
T-A
no
mal
ie in
K
-0,5
-0,4
-0,3
-0,2
-0,1
0,0
0,1
0,2
0,3
0,4
0,5
0,6GHGSOLVULSOI
99.9%
99%
99%
95%
95%
srsch
srsch
Case 2: noise = natural variability + unexplained
data field EOF-Transformation
PC
Stepwise Regression
backtransformation
signal fields,
residual field
Treatment of data fields:
GHG signal field for the year 2000 relative to 1901 in [K]:
GHG signal field, seasonal means for 2000 relative to 1901 in [K]:
NH winter NH spring
NH summer NH autum
Ges. GHG SOL SOI VUL NAO
proz
ent
uale
erk
lärt
e V
aria
nz
0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
50
60
70
80
90
100
Explained variance of the full model and of single forcings for the global temperature data field 1878 - 2000
Significance of the GHG signal for 2000 relative to 1901 in percentages:
Case 1: noise represents chance
Case 2: noise = natural variability + unexplained
GHG signal field Europe for 2000 relative to 1878 in [K]:
Significance of the european GHG signal for 2000 relative to 1878 in percentages:
Case 1: noise represents chance
Case 2: noise = natural variability + unexplained
1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
T-A
no
mal
ie in
K
-1,0
-0,8
-0,6
-0,4
-0,2
0,0
0,2
0,4
0,6
0,8
1,0
GHGSOLNAO
90%
90%
srsch
srsch
Signficance of the GHG signal in the german mean temperature 1878 - 2000:
Case 1: noise represents chance
1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
T-A
no
mal
ie in
K
-1,0
-0,8
-0,6
-0,4
-0,2
0,0
0,2
0,4
0,6
0,8
1,0
GHGSOLNAO
90%
90%
srsch
srsch
Signficance of the GHG signal in the german mean temperature 1878 - 2000:
Case 1: noise = natural variability + unexplained
Time moving analysis:
Global mean temperature 1856 - 2003, window width: 100 yr
0
10
20
30
40
50
60
70
80
1856
-195
5
1861
-196
0
1866
-196
5
1871
-197
0
1876
-197
5
1881
-198
0
1886
-198
5
1891
-199
0
1896
-199
5
1901
-200
0
data window
ex
pla
ine
d v
ari
an
ce
[%
]
GES
GHG
NAT
SOL
SOI
Conclusions:
Explained variance is highest in global and hemispheric mean temperatures (ca. 70% - 80%) and is reduced in data sets with high spacial resolution.
On the global scale, GHG forcing is most important and significant.
On the european scale NAO is dominant – GHG forcing is not significant.
Time moving analysis shows a growing meaning of GHG forcing compared to natural forcings, especially since around 1985 on the global scale.