Transcript MatCollins_RMetS_palaeo

Palaeo-Constraints
on Future Climate
Change
Mat Collins, School of Engineering, Mathematics and Physical
Sciences, University of Exeter
Tamsin Edwards (Bristol), Tom Russon (Edinburgh), James Pope
(Leeds) + many others
#RMetSMeet
Current Status of Global Projections
Reto Knutti and Jan Sedláček
Nature Climate Change, 2012
Climate Models and Modellers
• A significant fraction of climate science is in modelling,
understanding, prediction and projection of future
climate change
• “What can studying palaoclimates do for us?”
• Palaeoclimate reconstructions have been influential in
showing that the climate could have been much different
• And have motivated modellers to reconfigure their
models and run simulations of palaeoclimate
• But can they help improve models and reduce
uncertainties in projections in a quantitative way?
Example:
Last
Glacial
Maximum
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
CS=3.7K
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
5-95%:
2.3K-5.0K
5-95% range
2.3-4.3K
Sexton et al. 2012 constrained by
present-day means and trends
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
5-95%:
1.5K-5.8K
QUMP/PalaeoQUMP Simulations
Example:
Last
Glacial
Maximum
5-95%: 0.4K-7.7K
See also Schmidt et al. 2013 Clim. Past Discuss.
QUMP/PalaeoQUMP Simulations
Quantitative use of Palaeoclimate
Data in Constraining Projections
• Potentially large ‘signal’, perhaps correlated
with things we might want to project
• Also potentially large ‘noise’ arising from
uncertainties in palaeo-reconstructions and
uncertainties in forcings/boundary conditions
• Three examples
• Natural variations in the El Niño Southern
Oscillation
• Pliocene
• Joint constraints from Mid-Holocene and Last
Glacial Maximum
El Niño Conditions
(SST Anomalies)
upwelling
© Crown copyright Met Office
El Niño Southern Oscillation (ENSO) Variability during
the Last Millennium
Western Cold Tongue
NINO3
Palmyra Atoll
Russon et al.
submitted
El Niño Southern Oscillation (ENSO)
Variability during the Last Millennium
Russon et al. submitted
Mid Pliocene Warm Period (3.33.0 million years BP)
• Characterised by high (natural) concentrations
of CO2 (405 ppmv)
• Continental configuration similar to present day
• Palaeo reconstructions available from ocean
sediments and plant fossils
• Simulations with perturbed parameter versions
of HadCM3 using PRISM2 boundary conditions
(reduced Rockies, Greenland and Antarctica,
vegetation changes)
James Pope
Temperature difference from STD
B
D
F
H
I
J
K
L
M
N
P
Q
James Pope
Data-Model Comparison: Temperature
Ensemble Member B
Standard
Ensemble Member P
James Pope
Data-Model Comparison: Biomes
B
J
P
James Pope
CONSTRAINTS FROM MID-HOLOCENE AND
LGM
STATISTICAL MODEL
Rougier, Goldstein and House (in review.): Second-order exchangeability analysis for
multi-model ensembles. Journal of the American Statistical Association.
We want to make as few assumptions and judgements as
possible
We don’t feel confident weighting simulators by their relative
success
so we use the ensemble mean
don’t include multiple versions of same
simulator
and check for outliers
We work with the large-scale patterns of change
UPDATE CLIMATE SENSITIVITY
still watching this space
Edwards et al. in prep
Summary and Conclusions
• Palaeo-data can be used in a quantitative way in
climate science
• The main issues are those of signal-to-noise and
having the right simulations, observations and
techniques
• Three examples
• Reconstructions of ENSO show that the period of modern-day
observations is quite anomalous in comparison with the
previous 1000 years
• Higher-sensitivity Pliocene simulations are more consistent
with observations than lower sensitivity models
• Statistical frameworks are required to synthesise models and
data (just like they are for modern-day observations and
simulations)
#RMetSMeet