Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Evaluating processes controlling subtropical
humidity using water vapor isotope measurements
Camille Risi
CIRES, Boulder
Thanks to: David Noone, Sandrine Bony,
TES data: John Worden, Jeonghoon Lee, Derek Brown,
SCIAMACHY data: Christian Frankenberg,
ACE-FTS data: Kaley Walker, Peter Bernath,
ground-based FTIR: Debra Wunsh, Matthias Schneider
TES meeting, 17 June 2010
Uncertainties in humidity changes
I subtropical relative humidity strongly impacts
I water vapor feedback (Soden et al 2008)
I clouds feedbacks (Sherwood et al 2010)
I but high dispersion in climate models (Sherwood et al 2010)
I for present day relative humidityI for projectionsI no link between mean state and projections (Santer et al 2009)
I How to evaluate the credibility of these projections?
2/13
Uncertainties in humidity changes
I subtropical relative humidity strongly impacts
I water vapor feedback (Soden et al 2008)I clouds feedbacks (Sherwood et al 2010)
I but high dispersion in climate models (Sherwood et al 2010)
I for present day relative humidityI for projectionsI no link between mean state and projections (Santer et al 2009)
I How to evaluate the credibility of these projections?
2/13
Uncertainties in humidity changes
I subtropical relative humidity strongly impacts
I water vapor feedback (Soden et al 2008)I clouds feedbacks (Sherwood et al 2010)
I but high dispersion in climate models (Sherwood et al 2010)
I for present day relative humidityI for projections
I no link between mean state and projections (Santer et al 2009)
I How to evaluate the credibility of these projections?
2/13
Uncertainties in humidity changes
I subtropical relative humidity strongly impacts
I water vapor feedback (Soden et al 2008)I clouds feedbacks (Sherwood et al 2010)
I but high dispersion in climate models (Sherwood et al 2010)
I for present day relative humidityI for projectionsI no link between mean state and projections (Santer et al 2009)
I How to evaluate the credibility of these projections?
2/13
Uncertainties in humidity changes
I subtropical relative humidity strongly impacts
I water vapor feedback (Soden et al 2008)I clouds feedbacks (Sherwood et al 2010)
I but high dispersion in climate models (Sherwood et al 2010)
I for present day relative humidityI for projectionsI no link between mean state and projections (Santer et al 2009)
I How to evaluate the credibility of these projections?
2/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
large-s ale subsiden eSherwood et al 1996
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
large-s ale subsiden edry air intrusionsSherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
large-s ale subsiden edry air intrusionslateral adve tionSherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005Pierrehumbert 1998
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
large-s ale subsiden edry air intrusionslateral adve tionsurfa e evaporationboundary layerSherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005Pierrehumbert 1998Couhert et al 2010
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
large-s ale subsiden edry air intrusionslateral adve tionsurfa e evaporationboundary layerrain evaporation
Sherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005Pierrehumbert 1998Couhert et al 2010Folkins and Martin 2005
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
large-s ale subsiden edry air intrusionslateral adve tionsurfa e evaporationboundary layerrain evaporation onve tive subsiden e
Sherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005Pierrehumbert 1998Couhert et al 2010Bretherton et al 2005Folkins and Martin 2005
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
condensate detrainementPierrehumbert and Emanuel 1994 large-s ale subsiden edry air intrusionslateral adve tionsurfa e evaporationboundary layerrain evaporation onve tive subsiden e
Wright et al 2009 Sherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005Pierrehumbert 1998Couhert et al 2010Bretherton et al 2005Folkins and Martin 2005
⇒need complementary evaluation tools
3/13
Processes controlling relative humidity
I dispersion re�ects complexity of humidity controls
800
600
200
100
P (hPa)
0°
400
100030°N
condensate detrainementPierrehumbert and Emanuel 1994 large-s ale subsiden edry air intrusionslateral adve tionsurfa e evaporationboundary layerrain evaporation onve tive subsiden e
Wright et al 2009 Sherwood et al 1996Pierrehumbert and Ro a 98Galewsky et al 2005Pierrehumbert 1998Couhert et al 2010Bretherton et al 2005Folkins and Martin 2005
⇒need complementary evaluation tools3/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation
⇒ record phase changes along
air mass history
HDO
H162 O
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
800
600
200
100
P (hPa)
0°
400
100030°N
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
800
600
200
100
P (hPa)
0°
400
100030°N
onve tive subsiden eRisi et al 2008
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
800
600
200
100
P (hPa)
0°
400
100030°N
rain evaporation onve tive subsiden eWorden et al 2007Risi et al 2008
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
800
600
200
100
P (hPa)
0°
400
100030°N
ondensate detrainementrain evaporation onve tive subsiden eWorden et al 2007Risi et al 2008
Webster and Heims�led 2003
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
800
600
200
100
P (hPa)
0°
400
100030°N
ondensate detrainementlarge-s ale subsiden erain evaporation onve tive subsiden eWorden et al 2007Risi et al 2008Frankenberg et al 1999Webster and Heims�led 2003
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models
4/13
Water stable isotopes
I water isotopes: H162 O, H18
2 O,HDO
I fractionation ⇒ record phase changes along
air mass historyHDO
H162 O
800
600
200
100
P (hPa)
0°
400
100030°N
ondensate detrainementlarge-s ale subsiden erain evaporation onve tive subsiden eWorden et al 2007Risi et al 2008Frankenberg et al 1999Webster and Heims�led 2003
⇒ Goal: use water isotopes to evaluate processes controlling
humidity in climate models 4/13
Isotope dispersion in climate modelsI isotope-enabled GCMs intercomparison (SWING2)
5/13
Isotope dispersion in climate modelsI isotope-enabled GCMs intercomparison (SWING2)
P (
hPa)
nudged
free
−600 −400 −300 −200 −100−500
ECHAM
MIROC
CAM2
LMDZ AR4
LMDZ SWING
HadAM
GSM
1000
900
800
700
600
500
400
300
200
100Subtropi al mean 10◦N-30◦N
δD (h) annual5/13
Isotope dispersion in climate modelsI isotope-enabled GCMs intercomparison (SWING2)
20 25 30 35 40 45 50
−40
−20
0
20
40
60
P (
hPa)
nudged
free
−600 −400 −300 −200 −100−500
ECHAM
MIROC
CAM2
LMDZ AR4
LMDZ SWING
HadAM
GSM
1000
900
800
700
600
500
400
300
200
100Subtropi al mean 10◦N-30◦N
JJA-DJFδD
at450hPa(h)
δD (h) annual
NCEPrelative humidity at 450hPa (%)
5/13
Isotope dispersion in climate modelsI isotope-enabled GCMs intercomparison (SWING2)
20 25 30 35 40 45 50
−40
−20
0
20
40
60
P (
hPa)
LMDZ SWING
LMDZ AR4
nudged
free
−600 −400 −300 −200 −100−500
ECHAM
MIROC
CAM2
LMDZ AR4
LMDZ SWING
HadAM
GSM
1000
900
800
700
600
500
400
300
200
100Subtropi al mean 10◦N-30◦N
JJA-DJFδD
at450hPa(h)
δD (h) annual
NCEPrelative humidity at 450hPa (%)LMDZ-iso (Risi et al in press)
5/13
Isotope dispersion in climate modelsI isotope-enabled GCMs intercomparison (SWING2)
20 25 30 35 40 45 50
−40
−20
0
20
40
60
P (
hPa)
LMDZ SWING
LMDZ AR4
nudged
free
−600 −400 −300 −200 −100−500
ECHAM
MIROC
CAM2
LMDZ AR4
LMDZ SWING
HadAM
GSM
1000
900
800
700
600
500
400
300
200
100Subtropi al mean 10◦N-30◦N
JJA-DJFδD
at450hPa(h)
δD (h) annualwithsaturation limiterNCEP
relative humidity at 450hPa (%)(Codron and Sadourny 2002)adve t min(q,qs)LMDZ-iso (Risi et al in press)5/13
Isotope dispersion in climate modelsI isotope-enabled GCMs intercomparison (SWING2)
20 25 30 35 40 45 50
−40
−20
0
20
40
60
P (
hPa)
LMDZ SWING
LMDZ AR4
nudged
free
−600 −400 −300 −200 −100−500
ECHAM
MIROC
CAM2
LMDZ AR4
LMDZ SWING
HadAM
GSM
1000
900
800
700
600
500
400
300
200
100Subtropi al mean 10◦N-30◦N
JJA-DJFδD
at450hPa(h)
δD (h) annualwithsaturation limiterNCEP
relative humidity at 450hPa (%)withoutsaturationlimiteradve t q(Codron and Sadourny 2002)adve t min(q,qs)LMDZ-iso (Risi et al in press)5/13
Data for 3D isotope evaluation
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
TES(Worden et al 2007)
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES(Worden et al 2007)
(Nassar et al 2007)
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES(Worden et al 2007)
(Nassar et al 2007)
SCIAMACHY(Frankenberg et al 2009)
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES(Worden et al 2007)
(Nassar et al 2007)
SCIAMACHY(Frankenberg et al 2009)
(Wisconsin, Oklahoma)+ground−based FTIR
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES(Worden et al 2007)
(Nassar et al 2007)
in situ(GNIP, Angert et al 2007, Uemura et al 2007)
SCIAMACHY(Frankenberg et al 2009)
(Wisconsin, Oklahoma)+ground−based FTIR
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES
ground−based FTIR profile
(Worden et al 2007)
(Nassar et al 2007)
in situ(GNIP, Angert et al 2007, Uemura et al 2007)
(Izana:Schneider et al 2009)
SCIAMACHY(Frankenberg et al 2009)
(Wisconsin, Oklahoma)+ground−based FTIR
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES
ground−based FTIR profile
(Worden et al 2007)
(Nassar et al 2007)
in situ(GNIP, Angert et al 2007, Uemura et al 2007)
(Izana:Schneider et al 2009)
SCIAMACHY(Frankenberg et al 2009)
(Wisconsin, Oklahoma)+ground−based FTIR
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screening
I GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels
6/13
Data for 3D isotope evaluation
ACE−FTS
TES
ground−based FTIR profile
(Worden et al 2007)
(Nassar et al 2007)
in situ(GNIP, Angert et al 2007, Uemura et al 2007)
(Izana:Schneider et al 2009)
SCIAMACHY(Frankenberg et al 2009)
(Wisconsin, Oklahoma)+ground−based FTIR
800
600
200
100
P (hPa)
0° 30°N
400
1000
I quality screeningI GCM-data comparison: collocated with simulation nudged by
ECMWF; averaging kernels 6/13
Multidataset evaluation: annual mean
Eq 30N60S 60N30S
Eq 30N60S 60N30S
Eq 30N60S 60N30S−100
−200
−300
−300
−200
−100
−200
−100
0
−400
−500
−600
−300
Eq 30N60S 60N30S
Eq 30N60S 60N30S−700
−400
−500
−600
LMDZ−AR4data
δD
(h)δD
(h)
δD
(h)Surfa e
Total olumn600 hPa
Zonal mean, annual average
δD
(h)
δD
(h)500-300hPa
300-200hPa7/13
Multidataset evaluation: annual mean
Eq 30N60S 60N30S
Eq 30N60S 60N30S
Eq 30N60S 60N30S−100
−200
−300
−300
−200
−100
−400
−500
−600
−300
Eq 30N60S 60N30S
Eq 30N60S 60N30S−700
−400
−500
−600
LMDZ−"moist bias"LMDZ−AR4data
−200
−100
0
Surfa eTotal olumn600 hPa
Zonal mean, annual averageδD
(h)δD
(h)
δD
(h)δD
(h)
δD
(h)500-300hPa
300-200hPa7/13
Multidataset evaluation: seasonal
Eq 30N60S 60N30S
Eq 30N60S 60N30S
−100
0
100
−100
0
100
−100
0
100
−100
0
100
−100
100
0
Eq 30N60S 60N30S Eq 30N60S 60N30S
Eq 30N60S 60N30S∆
δD
(h)∆
δD
(h)∆
δD
(h)
∆δD
(h)
∆δD
(h)
LMDZ-AR4data600 hPa
Total olumn and 800hPaSurfa e
300-200hPa500-300hPa
Zonal mean, JJA-DJF
8/13
Multidataset evaluation: seasonal
Eq 30N60S 60N30S
Eq 30N60S 60N30S
−100
0
100
−100
0
100
−100
0
100
−100
0
100
−100
100
0
Eq 30N60S 60N30S Eq 30N60S 60N30S
Eq 30N60S 60N30S∆
δD
(h)∆
δD
(h)∆
δD
(h)
∆δD
(h)
∆δD
(h)
LMDZ-AR4dataLMDZ-"moist bias"600 hPa
Total olumn and 800hPaSurfa e
300-200hPa500-300hPa
Zonal mean, JJA-DJF
8/13
Annual mean in TES
30S
Eq
60N
30N
60S
120W180W 060W 60E 120E 180E
120W180W 060W 60E 120E 180E
−230 −220 −210 −200 −190 −180 −170 −160 −150
30S
Eq
60N
30N
60S
TES data
LMDZ AR4 -31h
δD (h) 600hPa9/13
Annual mean in TES
120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S
30S
Eq
60N
30N
60S
120W180W 060W 60E 120E 180E
120W180W 060W 60E 120E 180E
−230 −220 −210 −200 −190 −180 −170 −160 −150
30S
Eq
60N
30N
60S
same in CAM2(Lee et al 2009)
TES data
LMDZ AR4 -31h LMDZ "moist bias" -31h
δD (h) 600hPa9/13
Annual mean in TES
120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S
30S
Eq
60N
30N
60S
120W180W 060W 60E 120E 180E
120W180W 060W 60E 120E 180E
−230 −220 −210 −200 −190 −180 −170 −160 −150
30S
Eq
60N
30N
60S
same in CAM2(Lee et al 2009)
TES data
LMDZ AR4 -31h LMDZ "moist bias" -31h
δD (h) 600hPa9/13
Seasonal variations in TES
−80 −50 −30 −20 −10 10 20 30 50 80
120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S
120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S
∆δD (h) JJA-DJF
LMDZ AR4 LMDZ "moist bias"
TES data
10/13
Cloud e�ect in TES
−50 −20 −10 −5 −2 5 10 20 502
120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S
120W180W 060W 60E 120E 180E
30S
Eq
60N
30N
60S
∆δD (h) total- lear sky
TES data
LMDZ AR4 LMDZ "moist bias"
11/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
LMDZAR4
bias""moistLMDZ
− weak precip compositeStrong precip composite
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
LMDZAR4
bias""moistLMDZ
− weak precip compositeStrong precip composite
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropics
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropics
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropicsmore enriched in summer
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropicsmore enriched in summer
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropics
moist, enriched subtropics
moist tropics
more enriched in summer
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropics
moist, enriched subtropics
moist tropics
more enriched in summer
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
dry tropics
dry,depleted subtropics
moist, enriched subtropics
moist tropics
more depleted in summer
more enriched in summer
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
Annual mean 15°N−30°N
moist bias 25%
at 500hPa:
dry tropics
dry,depleted subtropics
moist, enriched subtropics
moist tropics
moist bias 15%
more depleted in summer
more enriched in summer
∆δD (h)
∆δD (h)
12/13
Isotopes reveal dehydration pathways
1000
900
700
800
600
500
400
200
100
300
1000
900
700
800
600
500
400
200
100
300
−10 0 10 20 30−20 40
−10 0 10 20 30 40−20
− weak precip compositeStrong precip composite
LMDZAR4
bias""moistLMDZ
saturationlimiter
Annual mean 15°N−30°N
moist bias 25%
at 500hPa:
dry tropics
dry,depleted subtropics
moist, enriched subtropics
moist tropics
moist bias 15%
more depleted in summer
more enriched in summer
∆δD (h)
∆δD (h)
+0.5% hange in 4xCO2
-4% hange in 4xCO2
12/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs
⇒develop GCM simulators of satellite dataI Modelling: water isotopes in climate change inter-comparisons
like CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?
⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13
Conclusion/perspectives
I Water isotopes can help detect and understand biases in water
budgets in climate models
⇒ help evaluate the credibility of projected future relative
humidity changes
I Perspectives:
I Observations: maximum variations and biases in themid/upper troposphere
I Model-data comparison methodology: de�ne consistentlycloudy situations in data and GCMs⇒develop GCM simulators of satellite data
I Modelling: water isotopes in climate change inter-comparisonslike CMIP?⇒ use intra-seasonal, seasonal, inter-annual isotope data as anobservational test for model behavior in climate change?
13/13