Transcript PPT File - Climate Decision Making Center

Estimated PDFs of climate system properties
including natural and anthropogenic forcings
and implications for 21st century climate change
predictions.
Dr. Chris E. Forest
MIT Joint Program on the
Science and Policy of Global Change
http://web.mit.edu/globalchange/www/
Presentation to:
Climate Decision Making Center Seminar Series
Carnegie Mellon University
http://cdmc.epp.cmu.edu/SEMINARS.htm
April 24, 2006
Calibrated Climate Model Results
• Observed climate changes provide constraints
on climate response to forcings.
• Forest et al. (2006) use observed climate
changes to place probabilistic bounds on
parameters in the IGSM climate component
• These constraints provide bounds for climate
system response to any scenario of future
climate forcings and help provide information for
decision making process.
Major Climate Projection Uncertainties
• Future forcings
– Pathways of climate relevant emissions and
concentrations (GHGs, aerosols, ... ) (IPCC: SRES?)
– How much can pollutants reflect sunlight?
Climate
(Net Aerosol forcing, Faer(IPCC: ??)
System
•
Climate System Response Uncertainty
– Equilibrium temperature change
Properties
• How much will global-mean temperature change after
oceans, ice, or ecosystems adjust?
(Climate Sensitivity to 2x[CO2], S) (IPCC: 1.5-4.5 K)
– Transient climate change
• How fast can oceans (and ice) take up excess heat?
(Rate of heat uptake by the deep ocean, Kv) (IPCC: ??)
Estimating Uncertainty in Climate System Properties:
p(S,Kv,FaerTobs)
1.
Simulate 20th century climate using anthropogenic and
natural forcings while systematically varying the
choices of climate system properties: S, Kv, and Faer
2.
Compare each model response against observed T
as in optimal fingerprint detection algorithm
3.
Compare goodness-of-fit statistics to estimate
p(S,Kv,FaerTobs) for individual T diagnostics
4.
Estimate p(S,Kv,Faer Tobs) for multiple diagnostics and
combine results using Bayes’ Theorem
From: Forest et al. (2002), Science
Climate-change diagnostics (Ti)
1)Upper-air temperature changes, latitudeheight pattern, [1986-1995] - [1961-1980]
(Parker et al. 1997) (M=36x8)
2)Deep-ocean temperature trend, global, 0-3km
(1952-1995) (Levitus et al. 2000, 2005) (M=1)
3)Surface temperature change, latitude-time
pattern, (1946-1995 decadal means, 19061995 climatology, 4 zonal bands) (updated
from Jones, 1994) (M=4 x 5)
Calculations with GSO forcings
Summary of Changes from
GSO  GSOLSV
• Updated Forcings for 1860-2001
–
–
–
–
–
Updated Greenhouse Gas concentrations
Updated Sulfur emissions from 1990-2001
Updated Ozone concentrations
Added Land-use Vegetation Changes
Added Volcanic and Solar forcings
• Updated climate model to 4o resolution and
included new sea-ice model
• T Diagnostics identical to GSO
Constraints on climate parameters set from past
observations: the effects of including volcanic forcing
Greater Sensitivity?
Are
Aerosols &
Volcanoes
Masking the
real effect
of greenhouse
gases?
Less ocean mixing?
Less non-volcanic
aerosol cooling?
from Forest et al. (2002, Science) and
Forest et al. (2006, Geophys. Res. Letters)
Probability Distribution for Climate Sensitivity and
Rate of Deep-ocean heat uptake
from Forest et al. (2006, GRL)
Implication:
99%
Models overestimate rate of ocean heat
uptake for transient response leading
to faster adjustment to climate forcings.
90%
Accepted
Rejected
Slow
Cluster of AOGCMs
(Sokolov et al., 2003)
Sea level rise
Fast
Conclusions from updated PDFs
• Major changes in GSO  GSOLSV
– Higher lower bound on Clim. Sensitivity (~2K)
– Weaker deep-ocean heat uptake indicates a
bias in AOGCM results
– Reduced Net Aerosol forcing strength
– Highlights need for multiple lines of evidence
• Note: Expert priors are justified by including LGM
paleoclimate changes (e.g., Annan and Hargreaves,
2006, GRL)
GSO = Forest et al. (2002), Science
GSOLSV = Forest et al. (2006), GRL
Deep-Ocean Temperature Data
• Higher coverage in NH than SH
• Still poor coverage in SH for surface
• Two alternatives to using Global estimate
– Delete SH data and use trend in NH alone
– Treat hemispheres separately as independent
diagnostics
Zdepth=1km, 1990-1994
Ocean Temperature
Observations at 1km
depth for two 5-yr
periods:
1990-1994 (top)
1955-1959 (bottom)
(from Levitus et al. (2005)
Auxiliary Material)
Zdepth=1km, 1990-1994
From Gregory et al., GRL, VOL. 31, L15312, doi:10.1029/2004GL020258, 2004
Effects of missing data for
ocean heat content anomaly
estimates (HC)
Test two assumptions for
values at missing data points:
1. HC = 0
2. HC = representative
average
If existing data are a good
representation of missing
data, the change in heat
content would have been
larger.
Observed Tocean issues
• Without a detailed analysis, there is no clear
guide for deleting estimates from Southern
Hemisphere although data are sparse. A
judgment call is required.
• Appears to be equal justification for using global
or NH ocean temperatures.
– Natural variability estimated by AOGCMs in NH is
much larger leading to weaker constraints.
• No change in mode indicates AOGCMs’
distribution is still biased.
Implications for future
• Two sensitivity tests
– Effect of reduced oceanic heat uptake (OHU)
• Three runs with different Kv
Mode
2x OHU
~4x OHU = old mode
 Kv = 0.64 cm2/s
 Kv = 2.56 cm2/s
 Kv = 9.2 cm2/s
– Additional volcanic forcing for 21st century
• Repeated past 50 yrs twice
• Added past 100 yrs
Simulations by MIT
IGSM2 (Sokolov et al.,
2005) with reference
emissions scenario from
MIT EPPA4 (Paltzev et
al., 2005). [ S= 2.9 K,
Faer= -0.5 W/m2 ]
Implications for future climate change
• These features indicate that the reduction in the
ocean heat content has a much larger effect on
temperature changes than the volcanic forcing
scenario.
• In terms of temperature changes from the
present, the inclusion of the volcanic forcing
appears to reduce temperature increase only by
at most ~0.5 oC while reducing Kv leads to an
increase of ~1.8 oC by 2100 with S = 2.9 oC and
Faer = -0.5 W/m2 for this reference emissions
scenario.