•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
1/14
Ocean variability:
intrinsic/chaotic vs atmospherically-driven
What can we learn from ocean ensemble-simulations?
Stephanie Leroux
and the OCCIPUT team:
T. Penduff (PI), L. Bessières,
B. Barnier, J-M. Brankart, PV.
Huot, A. Jaymond, J-M. Molines, G. Sérazin, L. Terray, E AL.
5-day mean SSH (1–5 jan. 2015) from the OCCIPUT 50× ensemble simulation (1/4deg)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
2/14
Motivation | Why ocean ensemble simulations?
−→
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
3/14
Motivation | From laminar to eddy-permitting Ocean-GCMs
Atmospheric Forcing
Laminar NEMO ocean
(2 deg.)
▶ The atmosphere modulates a
passive ocean.
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
3/14
Motivation | From laminar to eddy-permitting Ocean-GCMs
Atmospheric Forcing
Laminar NEMO ocean
(2 deg.)
▶ The atmosphere modulates a
passive ocean.
Atmospheric Forcing
Turbulent, eddy-permitting
NEMO ocean
(1/4 deg.)
▶ The atmosphere modulates a
turbulent, chaotic ocean.
▶ How can we disentangle
the intrinsic and the
atmospherically-forced
contributions of the variability?
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
3/14
Motivation | Intrinsic chaotic variability in eddy-permitting NEMO
Penduff et al (2011), Sérazin et al (2015),Gregorio et al (2015) :
Climatological simulations:
Intrinsic interannual AMOC Std deviation:
Climatological forcing
(no interannual variability)
o
latitude ( )
Turbulent, eddy-permitting
NEMO ocean
(1/4 deg.)
▶ “Pure”, isolated intrinsic variability:
– mesoscale turbulence,
– cascading to interannual timescales
and larger spatial scales,
– potential impact on climate variability.
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
4/14
Motivation | Intrinsic chaotic variability in eddy-permitting NEMO
Penduff et al (2011), Gregorio et al (2015), Sérazin et al (2015) :
Climatological simulations:
▶ How does this intrinsic
Climatological forcing
(no interannual variability)
variability look like when
simulated under realistic
Turbulent, eddy-permitting
NEMO ocean
(1/4 deg.)
atmospheric forcing?
▶ How do magnitudes compare
(forced/intrinsic)?
▶ “Pure”, isolated intrinsic variability:
– mesoscale turbulence,
▶ Ensemble simulations:
provide a way to disentangle
– cascading to interannual timescales
the forced and intrinsic
and larger spatial scales,
– potential impact on climate variability.
contributions.
▶ Consistent with idealized studies
E.g. Sevellec & Huc 2015, Dijkstra & Ghil 2005
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
OCCI
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
5/14
| 50× global 1/4o ocean/sea-ice hindcasts 1960-2015
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
OCCI
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
6/14
| 50× global 1/4o ocean/sea-ice hindcasts 1960-2015
1960
1961
2015
50× members (1/4o ) driven by same atmospheric forcing (ERAi/DFS5.2)
one-member spin-up ∼20 yrs
Ensemble mean
(→ forced variability)
Ensemble dispersion
(→ intrinsic variability)
Probability Density Function
evolving with time
Initial perturbation strategy:
50× stochastic equation of state
applied for ONE year (Brankart et al 2013)
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
OCCI
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
6/14
| 50× Eddy-permitting ocean/sea-ice hindcasts 1960-2015
1960
1961
2015
50× members (1/4o ) driven by same atmospheric forcing (ERAi/DFS5.2)
one-member spin-up ∼20 yrs
Ensemble mean
(→ forced variability)
Ensemble dispersion
(→ intrinsic variability)
Probability Density Function
evolving with time
Initial perturbation strategy:
50× stochastic equation of state
applied for ONE year (Brankart et al 2013)
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
OCCI
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
6/14
| 50× Eddy-permitting ocean/sea-ice hindcasts 1960-2015
1960
1961
2015
50× members (1/4o ) driven by same atmospheric forcing (ERAi/DFS5.2)
one-member spin-up ∼20 yrs
Ensemble mean
(→ forced variability)
Ensemble dispersion
(→ intrinsic variability)
Probability Density Function
evolving with time
Initial perturbation strategy:
50× stochastic equation of state
applied for ONE year (Brankart et al 2013)
▶ Ensemble-NEMO: N members run simultaneously in one single executable,
▶ On-line ensemble statistics (could be re-injected directly in the on-going integration),
▶ 50×Synthetic obs datasets (e.g. ENAC /EN EMBLE in-situ data, Jason2)
▶
∼19 million CPU h. Full-time OCCI
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
engineer L. Bessières, +JM. Molines, JM. Brankart.
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
7/14
A probabilistic description
of the simulated ocean
→ Ex: SSH mesoscale variability and AVISO.
5-day mean SSH (1/4deg)
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
OCCI
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
8/14
| A probabilistic description of the simulated ocean
Ex: Daily SSH variability in the Gulf of Mexico:
0
•Summary
Ratio intrinsic/forced:
1
[On-going work – PV Huot’s masters thesis]
3.5
PDFs (all mbs, all timesteps) :
SSH timeseries at grid-point LC (intrinsic dominates):
50×members
2008
2009
2010
E-Mean
2011
AVISO
2012
2013
2012
2013
SSH timeseries at grid-point YU (forced dominates):
2008
2009
2010
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
2011
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
OCCI
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
8/14
| A probabilistic description of the simulated ocean
Ex: Daily SSH variability in the Gulf of Mexico:
0
•Summary
Ratio intrinsic/forced:
1
[On-going work – PV Huot’s masters thesis]
3.5
PDFs (all mbs, all timesteps) :
SSH timeseries at grid-point LC (intrinsic dominates):
50×members
2008
2009
2010
E-Mean
2011
AVISO
2012
2013
2012
2013
SSH timeseries at grid-point YU (forced dominates):
2008
2009
2010
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
2011
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
9/14
How much does
this intrinsic variability matter
on longer timescales
and larger spatial domains ?
→ Focus on the interannual variability of a basin-integrated quantity:
the Atlantic Meridional Overturning Circulation (AMOC).
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
10/14
Interannual timescale | How chaotic is AMOC variability?
AMOC (Sv)
AMOC at 34.5o S (annual mean anomalies)
34.5o S
2.0
1.5
1.0
0.5
0.0
−0.5
−1.0
−1.5
−2.0
1960
Ind. members
1964
1968
1972
1976
1980
1984
1988
1992
1996
SAMOC
From 50×
Tim e-st d of E-m ean
AMOC St d (Sv)
global ens.
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1960
1964
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
1968
E-st d
1972
1976
Em ean and dist rib. of t he individual t im e-st d
1980
1984
[email protected]
1988
1992
1996
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
10/14
Interannual timescale | How chaotic is AMOC variability?
AMOC (Sv)
AMOC at 34.5o S (annual mean anomalies)
34.5o S
2.0
1.5
1.0
0.5
0.0
−0.5
−1.0
−1.5
−2.0
1960
En emble-mean
1964
1968
Quantiles 25-75%
1972
1976
1980
1984
+/- 1 Ensemble-Std
1988
1992
1996
SAMOC
From 50×
global ens.
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
10/14
Interannual timescale | How chaotic is AMOC variability?
AMOC (Sv)
AMOC at 34.5o S (annual mean anomalies)
34.5o S
2.0
1.5
1.0
0.5
0.0
−0.5
−1.0
−1.5
−2.0
1960
En emble-mean
1964
1968
Quantiles 25-75%
1972
1976
1980
1984
+/- 1 Ensemble-Std
1988
1992
1996
SAMOC
global ens.
AMOC Std (Sv)
From 50×
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1960
Time-std of E-mean
E-std
Emean and distrib. of the individual time-std
80%
1964
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
1968
1972
1976
1980
1984
[email protected]
1988
1992
1996
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
11/14
Interannual timescale | How chaotic is AMOC variability?
At all latitudes:
65
55
45
La i ude (o)
latitude (o )
35
25
Time-Std of the ensemble-mean: forced variability.
50%
Ensemble-Std (averaged over 1960-1999): intrinsic variability
15
30%
5
— Distribution of the 50 individual total Time-Std (mean, min,
max, q25, q75): total variability.
−5
−15
−25
−35
80%
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Std (Sv)
Time
-Std 2016:
of E-mean
Journées GMMC, Toulon,
Jun.
The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
12/14
OCCIPUT Ensemble-mean
OCCIPUT intrinsic mb#1
CLIM run
Forced var.
Fully-forced intrinsic var.
Pure intrinsic var.
65
65
55
55
55
45
45
45
65
35
35
25
25
15
5
latitude (o)
35
25
latitude (o)
o
latitude
latitude( (o))
Interannual timescale | AMOC spatio-temporal patterns:
15
5
15
5
−5
−5
−5
−15
−15
−15
−25
−25
−25
−35
1961 1965 1969 1973 1977 1981 1985 1989 1993 1997
time (years)
−35
1961 1965 1969 1973 1977 1981 1985 1989 1993 1997
time (years)
time (years)
−1.8
−1.2
−0.6
0.0
(Sv)
0.6
1.2
1.8
time (years)
−0.9
−0.6
−0.3
0.0
0.3
0.6
0.9
−35
1961 1965 1969 1973 1977 1981 1985 1989 1993 1997
time (years)
time (years)
−0.9
−0.6
−0.3
0.0
0.3
(Sv)
(Sv)
→ “Fully-forced”
and “pure” intrinsic variability:
0.6
0.9
similar space-time organisation.
→ Forced and intrinsic variability: similar space-time
organisation south of ∼25N, different to the north.
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
13/14
Interannual timescale | Ocean Heat Content variability
0-700m OHC anomalies (BP filtered T= 2–22 yrs)
Forced (E-Mean)
Intrinsic (mb#5)
[On-going work – A. Jaymond’s masters thesis]
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
•Interannual-to-multidecadal variability
•Summary
14/14
Summary
▶ At eddy-permitting resolution: ocean variability is NOT fully
−→
determined by the atmospheric forcing.
▶ A substantial intrinsic and chaotic component adds up, cascading
from mesoscale turbulence to interannual timescales.
▶ Ensemble simulations provide:
- a probabilistic description of the simulated ocean under realistic
atmospheric forcing (e.g. →comparison with obs),
- a way to disentangle intrinsic and forced contributions.
▶ SST, SSH, AMOC, OHC show substantial intrinsic variability on
interannual timescales (→potential impact on climate variability).
On-going work...
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
[email protected]
(MEOM/LGGE, Grenoble)
•Motivation
•OCCI
•Probabilistic approach
Journées GMMC, Toulon, Jun. 2016: The OCCIPUT Project
•Interannual-to-multidecadal variability
•Summary
[email protected]
14/14
(MEOM/LGGE, Grenoble)