Upload
destiny-hamilton
View
213
Download
0
Embed Size (px)
Citation preview
© Crown copyright 2007 Page 1
CFMIP2: Options for SST-forced and slab experiments
Mark Ringer, Brian Soden
Hadley Centre,UK & RSMA/MPO, US
CFMIP/ENSEMBLES Workshop: Paris, 11-13th April 2007
© Crown copyright 2007 Page 2
Introduction – consider options for SST-forced and slab model experiments
Some issues to consider prior to this afternoon’s break-out group session on experimental design.
CMIP 1% per year composite ΔSST pattern Climatological or AMIP control? Constructing the composite ΔSST pattern
Other idealized SST perturbation experiments and approaches
Slab model experiments
These are just some ideas to get the discussion going…
© Crown copyright 2007 Page 3
CMIP 1% per year increase in CO2: ΔTs at time of doubling (e.g. Wyant et al. 2006)
Arguably more realistic and relevant to climate change over the 21st century.
However, the perturbation is relatively small (<ΔTs> ~ 1.4K) and there is little spatial structure in the tropics – resembles a uniform ΔTs in the zonal mean.
© Crown copyright 2007 Page 4
Choosing the control experiment – seasonal climatology or varying (AMIP) SSTs?
Climatology (e.g. 10 years) – reaches an equilibrium, which should hopefully reduce the noise in the response.
AMIP – control is potentially more useful for evaluation studies, but the response will be “conditioned” by ENSO and other variability during the period.
© Crown copyright 2007 Page 5
Constructing the composite ΔTs pattern
One alternative to simple averaging might be to “average the patterns”
i.e. ΔTs/<ΔTs>
and then scale by ensemble mean <ΔTs>.
This would reduce biases due to very high or low sensitivity models, for example.
ΔTs ΔTs /<ΔTs>
© Crown copyright 2007 Page 6
In addition, we could also consider…
A simple globally uniform SST perturbation (e.g. the ensemble mean ΔTs) to determine the impact of the SST pattern versus the overall warming – a seasonally varying version of the Cess et al. (1989) method
A zonally uniform SST perturbation to isolate the impact of changes to the meridional gradient in SST (analogous to the aquaplanet experiments in APE)
Reducing the SSTs – this might also provide insights into feedbacks. E.g. Wyant et al. show that +2K/-2K responses are not necessarily symmetric in a given model
Including a composite sea-ice response as well as the SST perturbation – but how?
© Crown copyright 2007 Page 7
Other approaches – e.g. Barsugli et al. (2006)[Climate Dynamics, 27(5), Oct. 2006]
Specified 2K warming/cooling at different locations in the tropics in order to investigate sensitivity to the pattern of ΔSST. Allowed them to separate the “local” and “global” sensitivities to the warming. Example shown is for TOA net radiation – local sensitivity is +ive everywhere but global sensitivity can be either +ive or -ive.
Other techniques include introducing (or suppressing) warming at specific locations believed to be important. E.g. Schneider et al. [JAS, 54, May 1997] held SSTs in the Pacific cold tongue at control values while allowing temperatures elsewhere to vary in response to doubling CO2.
© Crown copyright 2007 Page 8
Slab models – equilibrium 2×CO2 experiments, as in CFMIP1
Although not identical, the control states are relatively similar.
Allows us to consider equilibrium climate sensitivity – if that is still thought to be important!
Also provides continuity with previous studies.
• Will slab models be included in IPCC AR5?
• If not, what will they be “replaced” by (if anything)?
• Should CFMIP2 do them in any case?
© Crown copyright 2007 Page 9
Slab models – continued
Produce larger response – magnitude and spatial pattern – than the 1% per year simulations.
Could produce a composite pattern from these for the SST-forced experiments?
In essence this would mean CFMIP2 looking like CFMIP1 but with patterned SST experiments replacing the Cess et al. (1989) runs plus sensitivity experiments and any other idealized experiments we decide to do…
© Crown copyright 2007 Page 10
Summary of points discussed
Composite pattern derived from…
1%/year runs at time of CO2 doubling
Slab 2×CO2 equilibrium runs
Construction of composite pattern
Simple averaging Average of “patterns” scaled by mean <ΔSST>
Control simulation Seasonally varying climatology
AMIP SSTs, i.e. include interannual variability
Composite sea-ice pattern
Include? Don’t include?
Other idealized experiments
“Classic”, e.g. uniform ΔSST, reduced SST, etc
More sophisticated, e.g. Barsugli et al. method
Slab model experiments
Exclude if not part of IPCC AR5
Perform anyway as part of CFMIP2