atlantic multidecadal variability and the role of natural forcing in bcm
DESCRIPTION
Atlantic Multidecadal Variability and the role of natural forcing in BCM. Odd Helge Otterå, Mats Bentsen, Lingling Suo and Helene Langehaug (Nansen/Bjerknes). Bergen Climate Model (version 2). ARPEGE. ARPEGE Resolution: T42, ~2.8x2.8, 31 layers Volcanic aerosols implemented MICOM - PowerPoint PPT PresentationTRANSCRIPT
Helge DrangeGeofysisk institutt
Universitetet i Bergen
Atlantic Multidecadal Variability and the role of natural forcing in BCM
Odd Helge Otterå, Mats Bentsen, Lingling Suo and Helene Langehaug (Nansen/Bjerknes)
Bergen Climate Model (version 2)
• ARPEGE – Resolution: T42, ~2.8x2.8, 31
layers– Volcanic aerosols implemented
• MICOM – Resolution: ~2.4x2.4, 35 isopycnic
layers– New pressure gradient formulation – Reference pressure at 2000 m– Incremental remapping for tracer
advection (better conservation)• Thermodynamic and dynamic sea-ice
modules – GELATO: multi-category ice – NERSC: one ice layer only
ARPEGE
MICOM
Performed simulations with BCM
CONTROL600: All forcings kept constant at pre-industrial (1850) levelNATURAL600: Same as CONTROL600, but with historic total solar irradiance (TSI) and volcanic aerosol variations for the last 600 yearsAll150: Same as NATURAL600, but with variations in well-mixed greenhouse gases and tropospheric sulfate aerosols. Total of 5 ensemble members performed.
Atlantic Merdional Overturning Circulation
CONTROL600
Otterå et al 2009, GMD, in press
16.6 Sv
Comparison to Levitus for control
Southern Ocean problem!
Otterå et al 2009, GMD, in press
Sea ice and NA surface ocean circulation
Otterå et al 2009, GMD, in press
Ventilation sites in BCM (control run)
Late winter Mixed Layer Depth (MLD) averaged over 700 years.
MLD > 1100 m in 10 winters:
3 convection regions
1. Greenland Sea
2. Labrador Sea
3. Irminger Sea
Courtesy of H. Langehaug
Regression of MLD & AMOC
Max MLD in GS ~17yrs after max AMOC
MLDLS is leading AMOC
Max MLD in LS ~8yrs before max AMOC
Regression between the Mixed Layer Depth averaged over the convection regions and the AMOC.
Courtesy of H. Langehaug
Lag=30yrs
Another way to investigate the propagation of intermediate and deep water masses…
Anomalies in the thickness of the intermediate layer (interface σθ=27.75) is regressed with AMOC
Max MLD in LS ~8yrs before max AMOC
Lag=-20yrs Lag=-10yrs Lag=0yrs
Lag=10yrs Lag=20yrs
Max MLD in GS ~17yrs after max AMOC
Courtesy of H. Langehaug
Natural run: Applied forcing(Crowley et al. 2003)
Otterå et al 2009
Effective solar constant
Spörer Minimum
MaunderMinimum
DaltonMinimum
Kuwae
1452 Tambora
1815
Krakatoa
1883
Helge DrangeGeofysisk institutt
Universitetet i Bergen
Reconstructed and observed N Hemisphere temperature
Year
Tem
per
atu
re a
no
mal
y (º
C)
Mann et al. 2008
Settlement on Iceland & Greenland
Little Ice Age
Today
Otterå et al 2009
Simulated NH response
Kuwae
1452 Tambora
1815Krakatoa
1883
Simulated time-latitude variability of SAT (ALL forcing run, relative 1961-1990)
1816 – The year without a summerFollowing the 1815 Tambora eruption
(relative 1500-1899)
Mary Shelley
The winter warming phenomenon
Composite of 10 largest tropical eruptions
Simulated time-latitude variability of SAT (ALL forcing run, relative 1961-1990)
Simulated Early Warming in the Arctic2 m temperature, 60-90oN
Suo et al, in prog
Atlantic Multidecadal Oscillation (AMO)
Sutton & Hodson, 2005, Science
Average SST 75W-7.5W; 0-60N
Observed AMO Simulated AMO
°C per SD-AMO
Observed AMO Simulated AMO
°C per SD-AMO
Similarities between observed and simulated
✓ NH Temperature (1400-2000)✓AMO (1860-2000) and ✓Early Warming (1930-50)
for NATURAL and all members of ALLforcing, but not for CTRL
Natural forcing as a pacemaker for Atlantic multidecadal variability?
Otterå et al 2009
Power spectrum for AMO and AMOC (shading: 60-100 yr)
Otterå et al 2009
Control600 Natural600
More power on 60-100 yr time scales in NATURAL
Variability in the simulated strength of AMOC is – mainly – governed by Labrador Sea mixing with a lag of about 8 years.
Holds for both CTRL and NATURAL.
~ 8 yr lag
Lag-correlations (30 yr filter): AMO vs LabSea/AMOC/RadTOA
CONTROL600 NATURAL600
LS density and AMOC lead by 15 and 8 years
No lag with Rad TOA; LS density and AMOC lag by 5 and 15 years
Otterå et al 2009
Lag-correlations (unfiltered time series): Forcing vs LabSea/AMOC/RadTOA
Volcanoes plays a key role!
NATURAL600
Surface T and Atlantic streamfunction regressed onto AMO
About 90 yr
cycle
AMOC linked to the derivative of the AMO (AMO ROC): Atmosphere link?
SLP regressed onto the AMO ROC index
AMOC
AMO vs other climate parameters
NAO-index: reconstructed vs
model pc1
EOF1
10 yr running mean
Reconstructions from Gardar Drift
Courtesy of Tor Mjell and U. Ninnemann
weak
strong
cold
warm
Overflow
Winter temp
Sortable siltG. inflata
Upper ocean (300 m) temperature regressed on AMO-index (lag 0)
The Gardar Drift region anti-correlates with the AMO-index in the simulation
Simulated winter temperature Gardar drift vs AMO-index
Preliminary summary1. Main features of the observed multidecadal variability in the
Atlantic region are simulated by the model
2. The simulated multidecadal variability is strongly linked to changes in the combined effect of solar irradiance and aerosol variations, rather than to internal variability from the ocean component
3. (2) needs to be supported by other models/studies
4. The simulated AMOC in BCM is out of phase with AMO strong AMOC in cold times and vice versa
5. If these findings are robust, decadal-scale predictability experiments need to take into account future changes in solar irradiance and aerosol variations (volcanoes included)